Interpreting Key Trials

In treating hypertension, lower systolic pressure is better than higher—with caveats. This is the message of the Systolic Blood Pressure Intervention Trial (SPRINT),¹ a large, federally funded study that was halted early when patients at high cardiovascular risk who were randomized to a goal systolic pressure of less than 120 mm Hg were found to have better outcomes, including lower rates of heart failure, death from cardiovascular causes, and death from any cause, than patients randomized to a goal of less than 140 mm Hg.

See related editorial

The caveats: the benefit came at a price of more adverse events. Also, the trial excluded patients who had diabetes mellitus or previous strokes, so it is uncertain if these subgroups would also benefit from intensive lowering of systolic pressure—and in earlier trials they did not.

This article reviews the trial design and protocol, summarizes the results, and briefly discusses the implications of these results.

BEFORE SPRINT

Hypertension is very common in adults in the United States, and is a risk factor for heart disease, stroke, heart failure, and kidney disease. The estimated prevalence of hypertension in the 2011–2014 National Health and Nutrition Examination Survey (NHANES) was 29%, and the prevalence increases with age (7.3% in those ages 18 to 39, 32.2% in those ages 40 to 59, and 64.9% in those ages 60 and older).² Isolated systolic hypertension (ie, systolic blood pressure > 140 mm Hg with diastolic pressure < 90 mm Hg) is the most common form of hypertension after age 50.³

Clinical trials have provided substantial evidence that treating hypertension reduces the incidence of stroke, myocardial infarction, and heart failure.^4,5 Although observational studies show a progressive and linear rise in cardiovascular risk as systolic blood pressure rises above 115 mm Hg,⁶ clinical trials in the general population have not documented benefits of lowering systolic pressure to this level.^7–11 However, clinical trials that directly evaluated two different blood pressure goals in the general population showed benefit with achieving systolic blood pressure less than 150 mm Hg,^7,9 with limited data on lower blood pressure targets.^10–12

No benefit found in intensive systolic lowering in diabetes or after stroke

The Action to Control Cardiovascular Risk in Diabetes-Blood Pressure (ACCORD BP) trial¹³ in patients with type 2 diabetes found no benefit in lowering systolic pressure to less than 120 mm Hg compared with less than 140 mm Hg in terms of the trial’s primary composite cardiovascular outcome (ie, nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes). However, the intensively treated group in this trial did enjoy a benefit in terms of fewer stroke events.

The Secondary Prevention of Small Subcortical Strokes (SPS3) trial¹⁴ in patients with stroke found no significant benefit in lowering systolic pressure to less than 130 mm Hg compared with less than 150 mm Hg for overall risk of another stroke, but a significant benefit was noted in reduced risk of intracerebral hemorrhage.

Current guidelines, based on available evidence, advocate treatment to a systolic goal of less than 140 mm Hg in most patients, and recommend relaxing this goal to less than 150 mm Hg in the elderly.^15,16

SPRINT was stopped early due to better outcomes in the intensive treatment group

Given the uncertainty surrounding optimal systolic targets, SPRINT was designed to test the hypothesis that a goal of less than 120 mm Hg would reduce the risk of cardiovascular events more than the generally accepted systolic goal of less than 140 mm Hg.17 Patients with diabetes and stroke were excluded because a similar hypothesis was tested in the ACCORD BP and SPS3 trials, which included patients with these conditions.

SPRINT DESIGN

SPRINT was a randomized, controlled, open-label trial sponsored by the National Institutes of Health and conducted at 102 US sites.

Inclusion criteria. Participants had to be at least 50 years old, with systolic pressure of 130 to 180 mm Hg, and had to have at least one cardiovascular risk factor, eg:

• Clinical or subclinical cardiovascular disease (other than stroke)
• Chronic kidney disease, defined as estimated glomerular filtration rate (eGFR), calculated by the Modification of Diet in Renal Disease (MDRD) study equation, of 20 to less than 60 mL/min/1.73 m²
• Framingham risk score of 15% of more
• Age 75 or older.

Major exclusion criteria included:

• Diabetes
• Stroke
• Polycystic kidney disease
• Chronic kidney disease with an eGFR less than 20 mL/min/1.73 m²
• Proteinuria (excretion > 1 g/day).

Intensive vs standard treatment

Participants were randomized to receive intensive treatment (systolic goal < 120 mm Hg) or standard treatment (systolic goal < 140 mm Hg). Baseline antihypertensive medications were adjusted to achieve blood pressure goals based on randomization assignment.

Doses of medications were adjusted on the basis of an average of three seated office blood pressure measurements after a 5-minute period of rest, taken with an automated monitor (Omron Healthcare Model 907); the same monitor was used and the same protocol was followed at all participating sites. Blood pressure was also measured after standing for 1 minute to assess orthostatic change.

Intensive treatment required, on average, one more medication than standard treatment

Lifestyle modifications were encouraged in both groups. There was no restriction on using any antihypertensive medication, and this was at the discretion of individual investigators. Thiazide-type diuretics were encouraged as first-line agents (with chlorthalidone encouraged as the primary thiazide-type diuretic).

Outcomes measured

The primary outcome was a composite of myocardial infarction, acute coronary syndrome not resulting in myocardial infarction, stroke, acute decompensated heart failure, and cardiovascular mortality.

Secondary outcomes included individual components of the primary composite outcome, all-cause mortality, and the composite of primary outcome and all-cause mortality.

Renal outcomes were assessed as:

• Incident albuminuria (doubling of the urinary albumin-to-creatinine ratio from less than 10 mg/g to more than 10 mg/g)
• Composite of a 50% decrease in eGFR or development of end-stage renal disease requiring long-term dialysis or kidney transplantation (in those with baseline chronic kidney disease)
• A 30% decrease in eGFR (in those without chronic kidney disease).^1,17

SPRINT also recruited participants to two nested substudies: SPRINT MIND and SPRINT MIND MRI, to study differences in cognitive outcomes and small-vessel ischemic disease between intensive treatment and standard treatment.

STUDY RESULTS

Older patients at risk, but without diabetes

Of 14,692 participants screened, 9,361 were enrolled in the study between 2010 and 2013. Baseline characteristics were comparable in both groups.

Demographics. The mean age of the participants was 67.9, and about 28% were 75 or older. About 36% were women, 58% white, 30% black, and 11% Hispanic.

Cardiovascular risk. The mean Framingham risk score was 20% (ie, they had a 20% risk of having a cardiovascular event within 10 years), and 61% of the participants had a risk score of at least 15%. Twenty percent already had cardiovascular disease.

Blood pressure. The average baseline blood pressure was 139.7/78.2 mm Hg. One-third of the participants had baseline systolic pressures of 132 mm Hg or less, another third had pressures in the range of 132 to 145, and the rest had 145 mm Hg or higher.

Renal function. The mean serum creatinine level was about 1.1 mg/dL. The mean eGFR was about 71 mL/min/1.73 m² as calculated by the MDRD equation, and about 28% had eGFRs less than 60. The mean ratio of urinary albumin to creatinine was 44.1 mg/g in the intensive treatment group and 41.1 in the standard treatment group.

Other. The mean total cholesterol level was 190 mg/dL, fasting plasma glucose 99 mg/dL, and body mass index nearly 30 kg/m².

Blood pressure during treatment

People in the intensive treatment group were taking a mean of 2.8 antihypertensive medications, and those in the standard treatment group were taking 1.8. Patients in the intensive group required greater use of all classes of medications to achieve goal systolic pressure (Table 1).

Study halted early due to efficacy

Throughout the 3.26 years of follow-up, the average difference in systolic pressure between the two groups was 13.1 mm Hg, with a mean systolic pressure of 121.5 mm Hg in the intensive treatment group and 134.6 mm Hg in the standard treatment group. The mean diastolic blood pressure was 68.7 mm Hg in the intensive treatment group and 76.3 mm Hg in the standard treatment group.

Although the study was planned to run for an average follow-up of 5 years, the National Heart, Lung, and Blood Institute terminated it early at a median of 3.26 years in view of lower rates of the primary outcome and of heart failure and death in the intensive treatment group (Table 2).

The effects on the primary outcome and mortality were consistent across the prespecified subgroups of age (< 75 vs ≥ 75), sex (female vs male), race (black vs nonblack), cardiovascular disease (presence or absence at baseline), prior chronic kidney disease (presence or absence at baseline), and across blood pressure tertiles (≤ 132 mm Hg, > 132 to < 145 mm Hg, ≥ 145 mm Hg).

Follow-up for assessment of cognitive outcomes (SPRINT MIND) and small-vessel ischemic disease (SPRINT MIND MRI) is ongoing.

WHAT DOES THIS MEAN?

SPRINT is the first large, adequately powered, randomized trial to demonstrate cardiovascular and mortality benefit from lowering the systolic blood pressure (goal < 120 mm Hg) in older patients at cardiovascular risk but without a history of diabetes mellitus or stroke.¹

Most SPRINT patients had reasonably controlled blood pressure at baseline (the mean systolic pressure was 139.7 mm Hg, and two-thirds of participants had systolic pressure < 145 mm Hg). Of note, however, this trial excluded patients with systolic pressure higher than 180 mm Hg. There was excellent separation of systolic pressure between the two groups beginning at 1 year, which was consistent through the course of the trial.

The cardiovascular benefit in the intensive treatment group was predominantly driven by lower rates of heart failure (a 38% reduction in the intensive treatment group, P = .0002) and cardiovascular mortality (a 43% reduction in the intensive treatment group, P = .005), while there was no significant difference between the two groups in myocardial infarction or stroke. The beneficial effect on heart failure events is consistent with results from other trials including the Systolic Hypertension in the Elderly Program,⁷ Systolic Hypertension in Europe,⁸ and Hypertension in the Very Elderly Trial,⁹ all of which showed greatest risk reduction for heart failure events with systolic pressure-lowering (although to higher systolic levels than SPRINT).^7–9 It is unclear why there was no beneficial effect on stroke events. The reduction in all-cause mortality in the intensive treatment group in SPRINT was greater than the reduction in cardiovascular deaths, which is also unexplained.

Although the study was terminated early due to efficacy (which introduces the possible bias that the estimated effect size will be too high), the number of primary end points  reached was large (562 in the two groups combined), providing reassurance that the findings are valid. There was no blinding in the study (both participants and study investigators were aware of treatment assignment and study medications), but there was a structured assessment of outcomes and adverse events, with adjudication done by blinded reviewers.

SPRINT used an automated device for blood pressure measurement, which is known to reduce the “white coat” effect and correlates tightly with average daytime blood pressure done by ambulatory blood pressure monitoring.¹⁸ However, in clinical practice automated devices may not be available and a strict protocol for correct measurement may not be followed, with the possible result that blood pressure may be overestimated and overtreated.

What about diastolic pressure?

The trial, by design, focused on lowering systolic pressure (given the greater prevalence of isolated systolic hypertension with age), and the implications of lowering diastolic pressure are unclear. The issue of a J-shaped relationship between diastolic pressure and cardiovascular risk is debated in the literature: patients with a diastolic pressure of 60 to 65 mm Hg, especially those with existing coronary artery disease, may not tolerate aggressive blood pressure-lowering.^19,20 Further analysis of this association (if any) from SPRINT will be helpful.

What about patients with diabetes?

Patients were excluded from SPRINT if they were under age 50, were at low cardiovascular risk, or had diabetes, raising the question of whether the results apply to these groups as well.

The question is particularly relevant in diabetes, as the ACCORD BP study, which used the same blood pressure targets as SPRINT, did not show a significant difference in the primary cardiovascular outcome between the intensive and standard treatments in patients with diabetes (Table 3).¹³ In ACCORD BP, the rate of the primary outcome was 12% lower in the intensive treatment group than in the standard treatment group, but the 95% confidence interval was –27% to +6%, so the finding was not statistically significant. However, the wide confidence interval does not exclude the possibility of a benefit that was comparable to that observed in SPRINT.

It has been speculated that ACCORD BP was underpowered to detect significant differences in the primary outcome.²¹ An analysis combining data from both trials indicated that effects on individual outcomes were generally consistent in both trials (with no significant heterogeneity noted).²² Also, the primary composite outcome in ACCORD did not include heart failure, which is particularly sensitive to blood pressure reduction.

Additionally, ACCORD BP had a 2 × 2 factorial design involving a simultaneous comparison of intensive vs standard glycemic control, which may have influenced the effects due to blood pressure. Indeed, a post hoc analysis showed that there was a significant 26% lower risk of the primary outcome in ACCORD BP patients who received intensive systolic pressure  control plus standard glycemic control than in those receiving standard systolic control plus standard glycemic control.²³

Are more adverse events an acceptable trade-off?

Adverse events, including acute kidney injury, were more frequent in the intensive therapy group in SPRINT.

Acute kidney injury was coded as an adverse event on the basis of this diagnosis being included in the hospital discharge summary (as a primary or main secondary diagnosis) and if considered by the safety officer to be one of the top three reasons for admission or continued hospitalization. Further analysis of renal events should be forthcoming.

People in the intensive treatment group, on average, needed one more medication than those in the standard treatment group. Some of the adverse events may be related to the antihypertensive medications taken (eg, electrolyte abnormalities such as hyponatremia and hypokalemia due to diuretic use), and others may be related to blood pressure-lowering (eg, acute kidney injury due to renal hypoperfusion).

At this point, the long-term effects of these adverse events, especially on kidney function, are not known. Patients enrolled in clinical trials tend to be healthier than patients seen in clinical practice; thus, the rate of adverse events reported in the trial may be lower than one would see in the real world.

Does lower systolic pressure protect or harm the kidneys?

SPRINT included patients with stage 3 and 4 chronic kidney disease (ie, with eGFR 20–50 mL/min/1.73 m²), but it was designed to assess cardiovascular outcomes, not the progression of chronic kidney disease. The trial excluded patients with diabetic nephropathy or high degrees of proteinuria.

Only about half of hypertensive adults have their blood pressure under control, ie, < 140/90

Earlier randomized trials that focused on chronic kidney disease progression, including the MDRD²⁴ and the African American Study of Kidney Disease and Hypertension,²⁵ did not show benefit with more aggressive blood pressure-lowering (except in patients with higher degrees of proteinuria), and these trials were not powered to assess effects on cardiovascular outcomes.^24,25

The Irbesartan Diabetic Nephropathy Trial,^26,27 which was done in patients with overt diabetic nephropathy, showed that a progressively lower achieved systolic pressure down to 120 mm Hg predicted lower rates of heart failure, cardiovascular mortality, and renal events (although the trial target was ≤ 130/85 mm Hg and few participants achieved systolic pressure lower than 120 mm Hg).

IMPLICATIONS FOR MANAGEMENT

The recent estimates of hypertension prevalence and control from NHANES show that only about 53% of hypertensive adults have their blood pressure under control (defined as systolic pressure < 140 mm Hg and diastolic pressure < 90 mm Hg).² Analysis of the NHANES 2007–2012 data showed that 16.7% or 8.2 million US adults with treated hypertension meet the eligibility criteria for SPRINT.²⁸

Although the SPRINT results support the notion that “lower is better,” the risks and benefits of intensive control will need to be balanced in individual patients. Table 4 shows the number needed to treat and number needed to harm in the trial.

More aggressive management of hypertension is challenging. The median systolic pressure achieved in the intensive group in SPRINT was just over 120 mm Hg, which implies that at least half of the participants in the intensive group did not achieve the goal of less than 120 mm Hg. While it may be reasonable to aim for systolic pressure of less than 120 or 125 mm Hg in patients who fit the SPRINT criteria and can tolerate intensive blood pressure lowering, it would be prudent to aim for a more conservative goal in elderly patients who are frail and at risk for falls, considering the higher incidence of specified adverse events in the intensive group.

Results of cognitive outcomes, as well as data related to quality of life, are still awaited. Long-term renal outcomes are also unclear.

As noted above, the question of generalizability of SPRINT results to patients with diabetes is open to debate. In our opinion, with currently available evidence, it is difficult to conclusively answer the question of whether a lower systolic target provides cardiovascular benefit in diabetes. It is also unclear whether similar beneficial results would be seen with intensive treatment in a population at low cardiovascular risk. The American Heart Association and the American College of Cardiology are in the process of formulating new hypertension guidelines, and evidence from  SPRINT will inform any new recommendations.

As more medications will likely be needed for intensive systolic blood pressure control, side effects and tolerability of medications with polypharmacy and potential nonadherence with increasing complexity of medication regimens should be kept in mind. Lifestyle modifications will need to be emphasized and reinforced, with greater use of combination antihypertensive therapy.

The data from SPRINT indicate that lower systolic pressure is better, as long as untoward clinical events can be monitored and avoided or easily managed. Careful monitoring will likely entail more frequent clinic visits and more frequent assessment of renal function and electrolyte levels (participants in the intensive group in the trial were seen every month until goal was achieved). A team approach that includes pharmacists and nurse practitioners, along with optimal use of best practice algorithms and remote monitoring technology, will need to be implemented for efficient and effective care.

In treating hypertension, lower systolic pressure is better than higher—with caveats. This is the message of the Systolic Blood Pressure Intervention Trial (SPRINT),¹ a large, federally funded study that was halted early when patients at high cardiovascular risk who were randomized to a goal systolic pressure of less than 120 mm Hg were found to have better outcomes, including lower rates of heart failure, death from cardiovascular causes, and death from any cause, than patients randomized to a goal of less than 140 mm Hg.

See related editorial

The caveats: the benefit came at a price of more adverse events. Also, the trial excluded patients who had diabetes mellitus or previous strokes, so it is uncertain if these subgroups would also benefit from intensive lowering of systolic pressure—and in earlier trials they did not.

This article reviews the trial design and protocol, summarizes the results, and briefly discusses the implications of these results.

BEFORE SPRINT

Hypertension is very common in adults in the United States, and is a risk factor for heart disease, stroke, heart failure, and kidney disease. The estimated prevalence of hypertension in the 2011–2014 National Health and Nutrition Examination Survey (NHANES) was 29%, and the prevalence increases with age (7.3% in those ages 18 to 39, 32.2% in those ages 40 to 59, and 64.9% in those ages 60 and older).² Isolated systolic hypertension (ie, systolic blood pressure > 140 mm Hg with diastolic pressure < 90 mm Hg) is the most common form of hypertension after age 50.³

Clinical trials have provided substantial evidence that treating hypertension reduces the incidence of stroke, myocardial infarction, and heart failure.^4,5 Although observational studies show a progressive and linear rise in cardiovascular risk as systolic blood pressure rises above 115 mm Hg,⁶ clinical trials in the general population have not documented benefits of lowering systolic pressure to this level.^7–11 However, clinical trials that directly evaluated two different blood pressure goals in the general population showed benefit with achieving systolic blood pressure less than 150 mm Hg,^7,9 with limited data on lower blood pressure targets.^10–12

No benefit found in intensive systolic lowering in diabetes or after stroke

The Action to Control Cardiovascular Risk in Diabetes-Blood Pressure (ACCORD BP) trial¹³ in patients with type 2 diabetes found no benefit in lowering systolic pressure to less than 120 mm Hg compared with less than 140 mm Hg in terms of the trial’s primary composite cardiovascular outcome (ie, nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes). However, the intensively treated group in this trial did enjoy a benefit in terms of fewer stroke events.

The Secondary Prevention of Small Subcortical Strokes (SPS3) trial¹⁴ in patients with stroke found no significant benefit in lowering systolic pressure to less than 130 mm Hg compared with less than 150 mm Hg for overall risk of another stroke, but a significant benefit was noted in reduced risk of intracerebral hemorrhage.

Current guidelines, based on available evidence, advocate treatment to a systolic goal of less than 140 mm Hg in most patients, and recommend relaxing this goal to less than 150 mm Hg in the elderly.^15,16

SPRINT was stopped early due to better outcomes in the intensive treatment group

Given the uncertainty surrounding optimal systolic targets, SPRINT was designed to test the hypothesis that a goal of less than 120 mm Hg would reduce the risk of cardiovascular events more than the generally accepted systolic goal of less than 140 mm Hg.17 Patients with diabetes and stroke were excluded because a similar hypothesis was tested in the ACCORD BP and SPS3 trials, which included patients with these conditions.

SPRINT DESIGN

SPRINT was a randomized, controlled, open-label trial sponsored by the National Institutes of Health and conducted at 102 US sites.

Inclusion criteria. Participants had to be at least 50 years old, with systolic pressure of 130 to 180 mm Hg, and had to have at least one cardiovascular risk factor, eg:

• Clinical or subclinical cardiovascular disease (other than stroke)
• Chronic kidney disease, defined as estimated glomerular filtration rate (eGFR), calculated by the Modification of Diet in Renal Disease (MDRD) study equation, of 20 to less than 60 mL/min/1.73 m²
• Framingham risk score of 15% of more
• Age 75 or older.

Major exclusion criteria included:

• Diabetes
• Stroke
• Polycystic kidney disease
• Chronic kidney disease with an eGFR less than 20 mL/min/1.73 m²
• Proteinuria (excretion > 1 g/day).

Intensive vs standard treatment

Participants were randomized to receive intensive treatment (systolic goal < 120 mm Hg) or standard treatment (systolic goal < 140 mm Hg). Baseline antihypertensive medications were adjusted to achieve blood pressure goals based on randomization assignment.

Doses of medications were adjusted on the basis of an average of three seated office blood pressure measurements after a 5-minute period of rest, taken with an automated monitor (Omron Healthcare Model 907); the same monitor was used and the same protocol was followed at all participating sites. Blood pressure was also measured after standing for 1 minute to assess orthostatic change.

Intensive treatment required, on average, one more medication than standard treatment

Lifestyle modifications were encouraged in both groups. There was no restriction on using any antihypertensive medication, and this was at the discretion of individual investigators. Thiazide-type diuretics were encouraged as first-line agents (with chlorthalidone encouraged as the primary thiazide-type diuretic).

Outcomes measured

The primary outcome was a composite of myocardial infarction, acute coronary syndrome not resulting in myocardial infarction, stroke, acute decompensated heart failure, and cardiovascular mortality.

Secondary outcomes included individual components of the primary composite outcome, all-cause mortality, and the composite of primary outcome and all-cause mortality.

Renal outcomes were assessed as:

• Incident albuminuria (doubling of the urinary albumin-to-creatinine ratio from less than 10 mg/g to more than 10 mg/g)
• Composite of a 50% decrease in eGFR or development of end-stage renal disease requiring long-term dialysis or kidney transplantation (in those with baseline chronic kidney disease)
• A 30% decrease in eGFR (in those without chronic kidney disease).^1,17

SPRINT also recruited participants to two nested substudies: SPRINT MIND and SPRINT MIND MRI, to study differences in cognitive outcomes and small-vessel ischemic disease between intensive treatment and standard treatment.

STUDY RESULTS

Older patients at risk, but without diabetes

Of 14,692 participants screened, 9,361 were enrolled in the study between 2010 and 2013. Baseline characteristics were comparable in both groups.

Demographics. The mean age of the participants was 67.9, and about 28% were 75 or older. About 36% were women, 58% white, 30% black, and 11% Hispanic.

Cardiovascular risk. The mean Framingham risk score was 20% (ie, they had a 20% risk of having a cardiovascular event within 10 years), and 61% of the participants had a risk score of at least 15%. Twenty percent already had cardiovascular disease.

Blood pressure. The average baseline blood pressure was 139.7/78.2 mm Hg. One-third of the participants had baseline systolic pressures of 132 mm Hg or less, another third had pressures in the range of 132 to 145, and the rest had 145 mm Hg or higher.

Renal function. The mean serum creatinine level was about 1.1 mg/dL. The mean eGFR was about 71 mL/min/1.73 m² as calculated by the MDRD equation, and about 28% had eGFRs less than 60. The mean ratio of urinary albumin to creatinine was 44.1 mg/g in the intensive treatment group and 41.1 in the standard treatment group.

Other. The mean total cholesterol level was 190 mg/dL, fasting plasma glucose 99 mg/dL, and body mass index nearly 30 kg/m².

Blood pressure during treatment

People in the intensive treatment group were taking a mean of 2.8 antihypertensive medications, and those in the standard treatment group were taking 1.8. Patients in the intensive group required greater use of all classes of medications to achieve goal systolic pressure (Table 1).

Study halted early due to efficacy

Throughout the 3.26 years of follow-up, the average difference in systolic pressure between the two groups was 13.1 mm Hg, with a mean systolic pressure of 121.5 mm Hg in the intensive treatment group and 134.6 mm Hg in the standard treatment group. The mean diastolic blood pressure was 68.7 mm Hg in the intensive treatment group and 76.3 mm Hg in the standard treatment group.

Although the study was planned to run for an average follow-up of 5 years, the National Heart, Lung, and Blood Institute terminated it early at a median of 3.26 years in view of lower rates of the primary outcome and of heart failure and death in the intensive treatment group (Table 2).

The effects on the primary outcome and mortality were consistent across the prespecified subgroups of age (< 75 vs ≥ 75), sex (female vs male), race (black vs nonblack), cardiovascular disease (presence or absence at baseline), prior chronic kidney disease (presence or absence at baseline), and across blood pressure tertiles (≤ 132 mm Hg, > 132 to < 145 mm Hg, ≥ 145 mm Hg).

Follow-up for assessment of cognitive outcomes (SPRINT MIND) and small-vessel ischemic disease (SPRINT MIND MRI) is ongoing.

WHAT DOES THIS MEAN?

SPRINT is the first large, adequately powered, randomized trial to demonstrate cardiovascular and mortality benefit from lowering the systolic blood pressure (goal < 120 mm Hg) in older patients at cardiovascular risk but without a history of diabetes mellitus or stroke.¹

Most SPRINT patients had reasonably controlled blood pressure at baseline (the mean systolic pressure was 139.7 mm Hg, and two-thirds of participants had systolic pressure < 145 mm Hg). Of note, however, this trial excluded patients with systolic pressure higher than 180 mm Hg. There was excellent separation of systolic pressure between the two groups beginning at 1 year, which was consistent through the course of the trial.

The cardiovascular benefit in the intensive treatment group was predominantly driven by lower rates of heart failure (a 38% reduction in the intensive treatment group, P = .0002) and cardiovascular mortality (a 43% reduction in the intensive treatment group, P = .005), while there was no significant difference between the two groups in myocardial infarction or stroke. The beneficial effect on heart failure events is consistent with results from other trials including the Systolic Hypertension in the Elderly Program,⁷ Systolic Hypertension in Europe,⁸ and Hypertension in the Very Elderly Trial,⁹ all of which showed greatest risk reduction for heart failure events with systolic pressure-lowering (although to higher systolic levels than SPRINT).^7–9 It is unclear why there was no beneficial effect on stroke events. The reduction in all-cause mortality in the intensive treatment group in SPRINT was greater than the reduction in cardiovascular deaths, which is also unexplained.

Although the study was terminated early due to efficacy (which introduces the possible bias that the estimated effect size will be too high), the number of primary end points  reached was large (562 in the two groups combined), providing reassurance that the findings are valid. There was no blinding in the study (both participants and study investigators were aware of treatment assignment and study medications), but there was a structured assessment of outcomes and adverse events, with adjudication done by blinded reviewers.

SPRINT used an automated device for blood pressure measurement, which is known to reduce the “white coat” effect and correlates tightly with average daytime blood pressure done by ambulatory blood pressure monitoring.¹⁸ However, in clinical practice automated devices may not be available and a strict protocol for correct measurement may not be followed, with the possible result that blood pressure may be overestimated and overtreated.

What about diastolic pressure?

The trial, by design, focused on lowering systolic pressure (given the greater prevalence of isolated systolic hypertension with age), and the implications of lowering diastolic pressure are unclear. The issue of a J-shaped relationship between diastolic pressure and cardiovascular risk is debated in the literature: patients with a diastolic pressure of 60 to 65 mm Hg, especially those with existing coronary artery disease, may not tolerate aggressive blood pressure-lowering.^19,20 Further analysis of this association (if any) from SPRINT will be helpful.

What about patients with diabetes?

Patients were excluded from SPRINT if they were under age 50, were at low cardiovascular risk, or had diabetes, raising the question of whether the results apply to these groups as well.

The question is particularly relevant in diabetes, as the ACCORD BP study, which used the same blood pressure targets as SPRINT, did not show a significant difference in the primary cardiovascular outcome between the intensive and standard treatments in patients with diabetes (Table 3).¹³ In ACCORD BP, the rate of the primary outcome was 12% lower in the intensive treatment group than in the standard treatment group, but the 95% confidence interval was –27% to +6%, so the finding was not statistically significant. However, the wide confidence interval does not exclude the possibility of a benefit that was comparable to that observed in SPRINT.

It has been speculated that ACCORD BP was underpowered to detect significant differences in the primary outcome.²¹ An analysis combining data from both trials indicated that effects on individual outcomes were generally consistent in both trials (with no significant heterogeneity noted).²² Also, the primary composite outcome in ACCORD did not include heart failure, which is particularly sensitive to blood pressure reduction.

Additionally, ACCORD BP had a 2 × 2 factorial design involving a simultaneous comparison of intensive vs standard glycemic control, which may have influenced the effects due to blood pressure. Indeed, a post hoc analysis showed that there was a significant 26% lower risk of the primary outcome in ACCORD BP patients who received intensive systolic pressure  control plus standard glycemic control than in those receiving standard systolic control plus standard glycemic control.²³

Are more adverse events an acceptable trade-off?

Adverse events, including acute kidney injury, were more frequent in the intensive therapy group in SPRINT.

Acute kidney injury was coded as an adverse event on the basis of this diagnosis being included in the hospital discharge summary (as a primary or main secondary diagnosis) and if considered by the safety officer to be one of the top three reasons for admission or continued hospitalization. Further analysis of renal events should be forthcoming.

People in the intensive treatment group, on average, needed one more medication than those in the standard treatment group. Some of the adverse events may be related to the antihypertensive medications taken (eg, electrolyte abnormalities such as hyponatremia and hypokalemia due to diuretic use), and others may be related to blood pressure-lowering (eg, acute kidney injury due to renal hypoperfusion).

At this point, the long-term effects of these adverse events, especially on kidney function, are not known. Patients enrolled in clinical trials tend to be healthier than patients seen in clinical practice; thus, the rate of adverse events reported in the trial may be lower than one would see in the real world.

Does lower systolic pressure protect or harm the kidneys?

SPRINT included patients with stage 3 and 4 chronic kidney disease (ie, with eGFR 20–50 mL/min/1.73 m²), but it was designed to assess cardiovascular outcomes, not the progression of chronic kidney disease. The trial excluded patients with diabetic nephropathy or high degrees of proteinuria.

Only about half of hypertensive adults have their blood pressure under control, ie, < 140/90

Earlier randomized trials that focused on chronic kidney disease progression, including the MDRD²⁴ and the African American Study of Kidney Disease and Hypertension,²⁵ did not show benefit with more aggressive blood pressure-lowering (except in patients with higher degrees of proteinuria), and these trials were not powered to assess effects on cardiovascular outcomes.^24,25

The Irbesartan Diabetic Nephropathy Trial,^26,27 which was done in patients with overt diabetic nephropathy, showed that a progressively lower achieved systolic pressure down to 120 mm Hg predicted lower rates of heart failure, cardiovascular mortality, and renal events (although the trial target was ≤ 130/85 mm Hg and few participants achieved systolic pressure lower than 120 mm Hg).

IMPLICATIONS FOR MANAGEMENT

The recent estimates of hypertension prevalence and control from NHANES show that only about 53% of hypertensive adults have their blood pressure under control (defined as systolic pressure < 140 mm Hg and diastolic pressure < 90 mm Hg).² Analysis of the NHANES 2007–2012 data showed that 16.7% or 8.2 million US adults with treated hypertension meet the eligibility criteria for SPRINT.²⁸

Although the SPRINT results support the notion that “lower is better,” the risks and benefits of intensive control will need to be balanced in individual patients. Table 4 shows the number needed to treat and number needed to harm in the trial.

More aggressive management of hypertension is challenging. The median systolic pressure achieved in the intensive group in SPRINT was just over 120 mm Hg, which implies that at least half of the participants in the intensive group did not achieve the goal of less than 120 mm Hg. While it may be reasonable to aim for systolic pressure of less than 120 or 125 mm Hg in patients who fit the SPRINT criteria and can tolerate intensive blood pressure lowering, it would be prudent to aim for a more conservative goal in elderly patients who are frail and at risk for falls, considering the higher incidence of specified adverse events in the intensive group.

Results of cognitive outcomes, as well as data related to quality of life, are still awaited. Long-term renal outcomes are also unclear.

As noted above, the question of generalizability of SPRINT results to patients with diabetes is open to debate. In our opinion, with currently available evidence, it is difficult to conclusively answer the question of whether a lower systolic target provides cardiovascular benefit in diabetes. It is also unclear whether similar beneficial results would be seen with intensive treatment in a population at low cardiovascular risk. The American Heart Association and the American College of Cardiology are in the process of formulating new hypertension guidelines, and evidence from  SPRINT will inform any new recommendations.

As more medications will likely be needed for intensive systolic blood pressure control, side effects and tolerability of medications with polypharmacy and potential nonadherence with increasing complexity of medication regimens should be kept in mind. Lifestyle modifications will need to be emphasized and reinforced, with greater use of combination antihypertensive therapy.

The data from SPRINT indicate that lower systolic pressure is better, as long as untoward clinical events can be monitored and avoided or easily managed. Careful monitoring will likely entail more frequent clinic visits and more frequent assessment of renal function and electrolyte levels (participants in the intensive group in the trial were seen every month until goal was achieved). A team approach that includes pharmacists and nurse practitioners, along with optimal use of best practice algorithms and remote monitoring technology, will need to be implemented for efficient and effective care.

In treating hypertension, lower systolic pressure is better than higher—with caveats. This is the message of the Systolic Blood Pressure Intervention Trial (SPRINT),¹ a large, federally funded study that was halted early when patients at high cardiovascular risk who were randomized to a goal systolic pressure of less than 120 mm Hg were found to have better outcomes, including lower rates of heart failure, death from cardiovascular causes, and death from any cause, than patients randomized to a goal of less than 140 mm Hg.

See related editorial

The caveats: the benefit came at a price of more adverse events. Also, the trial excluded patients who had diabetes mellitus or previous strokes, so it is uncertain if these subgroups would also benefit from intensive lowering of systolic pressure—and in earlier trials they did not.

This article reviews the trial design and protocol, summarizes the results, and briefly discusses the implications of these results.

BEFORE SPRINT

Hypertension is very common in adults in the United States, and is a risk factor for heart disease, stroke, heart failure, and kidney disease. The estimated prevalence of hypertension in the 2011–2014 National Health and Nutrition Examination Survey (NHANES) was 29%, and the prevalence increases with age (7.3% in those ages 18 to 39, 32.2% in those ages 40 to 59, and 64.9% in those ages 60 and older).² Isolated systolic hypertension (ie, systolic blood pressure > 140 mm Hg with diastolic pressure < 90 mm Hg) is the most common form of hypertension after age 50.³

Clinical trials have provided substantial evidence that treating hypertension reduces the incidence of stroke, myocardial infarction, and heart failure.^4,5 Although observational studies show a progressive and linear rise in cardiovascular risk as systolic blood pressure rises above 115 mm Hg,⁶ clinical trials in the general population have not documented benefits of lowering systolic pressure to this level.^7–11 However, clinical trials that directly evaluated two different blood pressure goals in the general population showed benefit with achieving systolic blood pressure less than 150 mm Hg,^7,9 with limited data on lower blood pressure targets.^10–12

No benefit found in intensive systolic lowering in diabetes or after stroke

The Action to Control Cardiovascular Risk in Diabetes-Blood Pressure (ACCORD BP) trial¹³ in patients with type 2 diabetes found no benefit in lowering systolic pressure to less than 120 mm Hg compared with less than 140 mm Hg in terms of the trial’s primary composite cardiovascular outcome (ie, nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes). However, the intensively treated group in this trial did enjoy a benefit in terms of fewer stroke events.

The Secondary Prevention of Small Subcortical Strokes (SPS3) trial¹⁴ in patients with stroke found no significant benefit in lowering systolic pressure to less than 130 mm Hg compared with less than 150 mm Hg for overall risk of another stroke, but a significant benefit was noted in reduced risk of intracerebral hemorrhage.

Current guidelines, based on available evidence, advocate treatment to a systolic goal of less than 140 mm Hg in most patients, and recommend relaxing this goal to less than 150 mm Hg in the elderly.^15,16

SPRINT was stopped early due to better outcomes in the intensive treatment group

Given the uncertainty surrounding optimal systolic targets, SPRINT was designed to test the hypothesis that a goal of less than 120 mm Hg would reduce the risk of cardiovascular events more than the generally accepted systolic goal of less than 140 mm Hg.17 Patients with diabetes and stroke were excluded because a similar hypothesis was tested in the ACCORD BP and SPS3 trials, which included patients with these conditions.

SPRINT DESIGN

SPRINT was a randomized, controlled, open-label trial sponsored by the National Institutes of Health and conducted at 102 US sites.

Inclusion criteria. Participants had to be at least 50 years old, with systolic pressure of 130 to 180 mm Hg, and had to have at least one cardiovascular risk factor, eg:

• Clinical or subclinical cardiovascular disease (other than stroke)
• Chronic kidney disease, defined as estimated glomerular filtration rate (eGFR), calculated by the Modification of Diet in Renal Disease (MDRD) study equation, of 20 to less than 60 mL/min/1.73 m²
• Framingham risk score of 15% of more
• Age 75 or older.

Major exclusion criteria included:

• Diabetes
• Stroke
• Polycystic kidney disease
• Chronic kidney disease with an eGFR less than 20 mL/min/1.73 m²
• Proteinuria (excretion > 1 g/day).

Intensive vs standard treatment

Participants were randomized to receive intensive treatment (systolic goal < 120 mm Hg) or standard treatment (systolic goal < 140 mm Hg). Baseline antihypertensive medications were adjusted to achieve blood pressure goals based on randomization assignment.

Doses of medications were adjusted on the basis of an average of three seated office blood pressure measurements after a 5-minute period of rest, taken with an automated monitor (Omron Healthcare Model 907); the same monitor was used and the same protocol was followed at all participating sites. Blood pressure was also measured after standing for 1 minute to assess orthostatic change.

Intensive treatment required, on average, one more medication than standard treatment

Lifestyle modifications were encouraged in both groups. There was no restriction on using any antihypertensive medication, and this was at the discretion of individual investigators. Thiazide-type diuretics were encouraged as first-line agents (with chlorthalidone encouraged as the primary thiazide-type diuretic).

Outcomes measured

The primary outcome was a composite of myocardial infarction, acute coronary syndrome not resulting in myocardial infarction, stroke, acute decompensated heart failure, and cardiovascular mortality.

Secondary outcomes included individual components of the primary composite outcome, all-cause mortality, and the composite of primary outcome and all-cause mortality.

Renal outcomes were assessed as:

• Incident albuminuria (doubling of the urinary albumin-to-creatinine ratio from less than 10 mg/g to more than 10 mg/g)
• Composite of a 50% decrease in eGFR or development of end-stage renal disease requiring long-term dialysis or kidney transplantation (in those with baseline chronic kidney disease)
• A 30% decrease in eGFR (in those without chronic kidney disease).^1,17

SPRINT also recruited participants to two nested substudies: SPRINT MIND and SPRINT MIND MRI, to study differences in cognitive outcomes and small-vessel ischemic disease between intensive treatment and standard treatment.

STUDY RESULTS

Older patients at risk, but without diabetes

Of 14,692 participants screened, 9,361 were enrolled in the study between 2010 and 2013. Baseline characteristics were comparable in both groups.

Demographics. The mean age of the participants was 67.9, and about 28% were 75 or older. About 36% were women, 58% white, 30% black, and 11% Hispanic.

Cardiovascular risk. The mean Framingham risk score was 20% (ie, they had a 20% risk of having a cardiovascular event within 10 years), and 61% of the participants had a risk score of at least 15%. Twenty percent already had cardiovascular disease.

Blood pressure. The average baseline blood pressure was 139.7/78.2 mm Hg. One-third of the participants had baseline systolic pressures of 132 mm Hg or less, another third had pressures in the range of 132 to 145, and the rest had 145 mm Hg or higher.

Renal function. The mean serum creatinine level was about 1.1 mg/dL. The mean eGFR was about 71 mL/min/1.73 m² as calculated by the MDRD equation, and about 28% had eGFRs less than 60. The mean ratio of urinary albumin to creatinine was 44.1 mg/g in the intensive treatment group and 41.1 in the standard treatment group.

Other. The mean total cholesterol level was 190 mg/dL, fasting plasma glucose 99 mg/dL, and body mass index nearly 30 kg/m².

Blood pressure during treatment

People in the intensive treatment group were taking a mean of 2.8 antihypertensive medications, and those in the standard treatment group were taking 1.8. Patients in the intensive group required greater use of all classes of medications to achieve goal systolic pressure (Table 1).

Study halted early due to efficacy

Throughout the 3.26 years of follow-up, the average difference in systolic pressure between the two groups was 13.1 mm Hg, with a mean systolic pressure of 121.5 mm Hg in the intensive treatment group and 134.6 mm Hg in the standard treatment group. The mean diastolic blood pressure was 68.7 mm Hg in the intensive treatment group and 76.3 mm Hg in the standard treatment group.

Although the study was planned to run for an average follow-up of 5 years, the National Heart, Lung, and Blood Institute terminated it early at a median of 3.26 years in view of lower rates of the primary outcome and of heart failure and death in the intensive treatment group (Table 2).

The effects on the primary outcome and mortality were consistent across the prespecified subgroups of age (< 75 vs ≥ 75), sex (female vs male), race (black vs nonblack), cardiovascular disease (presence or absence at baseline), prior chronic kidney disease (presence or absence at baseline), and across blood pressure tertiles (≤ 132 mm Hg, > 132 to < 145 mm Hg, ≥ 145 mm Hg).

Follow-up for assessment of cognitive outcomes (SPRINT MIND) and small-vessel ischemic disease (SPRINT MIND MRI) is ongoing.

WHAT DOES THIS MEAN?

SPRINT is the first large, adequately powered, randomized trial to demonstrate cardiovascular and mortality benefit from lowering the systolic blood pressure (goal < 120 mm Hg) in older patients at cardiovascular risk but without a history of diabetes mellitus or stroke.¹

Most SPRINT patients had reasonably controlled blood pressure at baseline (the mean systolic pressure was 139.7 mm Hg, and two-thirds of participants had systolic pressure < 145 mm Hg). Of note, however, this trial excluded patients with systolic pressure higher than 180 mm Hg. There was excellent separation of systolic pressure between the two groups beginning at 1 year, which was consistent through the course of the trial.

The cardiovascular benefit in the intensive treatment group was predominantly driven by lower rates of heart failure (a 38% reduction in the intensive treatment group, P = .0002) and cardiovascular mortality (a 43% reduction in the intensive treatment group, P = .005), while there was no significant difference between the two groups in myocardial infarction or stroke. The beneficial effect on heart failure events is consistent with results from other trials including the Systolic Hypertension in the Elderly Program,⁷ Systolic Hypertension in Europe,⁸ and Hypertension in the Very Elderly Trial,⁹ all of which showed greatest risk reduction for heart failure events with systolic pressure-lowering (although to higher systolic levels than SPRINT).^7–9 It is unclear why there was no beneficial effect on stroke events. The reduction in all-cause mortality in the intensive treatment group in SPRINT was greater than the reduction in cardiovascular deaths, which is also unexplained.

Although the study was terminated early due to efficacy (which introduces the possible bias that the estimated effect size will be too high), the number of primary end points  reached was large (562 in the two groups combined), providing reassurance that the findings are valid. There was no blinding in the study (both participants and study investigators were aware of treatment assignment and study medications), but there was a structured assessment of outcomes and adverse events, with adjudication done by blinded reviewers.

SPRINT used an automated device for blood pressure measurement, which is known to reduce the “white coat” effect and correlates tightly with average daytime blood pressure done by ambulatory blood pressure monitoring.¹⁸ However, in clinical practice automated devices may not be available and a strict protocol for correct measurement may not be followed, with the possible result that blood pressure may be overestimated and overtreated.

What about diastolic pressure?

The trial, by design, focused on lowering systolic pressure (given the greater prevalence of isolated systolic hypertension with age), and the implications of lowering diastolic pressure are unclear. The issue of a J-shaped relationship between diastolic pressure and cardiovascular risk is debated in the literature: patients with a diastolic pressure of 60 to 65 mm Hg, especially those with existing coronary artery disease, may not tolerate aggressive blood pressure-lowering.^19,20 Further analysis of this association (if any) from SPRINT will be helpful.

What about patients with diabetes?

Patients were excluded from SPRINT if they were under age 50, were at low cardiovascular risk, or had diabetes, raising the question of whether the results apply to these groups as well.

The question is particularly relevant in diabetes, as the ACCORD BP study, which used the same blood pressure targets as SPRINT, did not show a significant difference in the primary cardiovascular outcome between the intensive and standard treatments in patients with diabetes (Table 3).¹³ In ACCORD BP, the rate of the primary outcome was 12% lower in the intensive treatment group than in the standard treatment group, but the 95% confidence interval was –27% to +6%, so the finding was not statistically significant. However, the wide confidence interval does not exclude the possibility of a benefit that was comparable to that observed in SPRINT.

It has been speculated that ACCORD BP was underpowered to detect significant differences in the primary outcome.²¹ An analysis combining data from both trials indicated that effects on individual outcomes were generally consistent in both trials (with no significant heterogeneity noted).²² Also, the primary composite outcome in ACCORD did not include heart failure, which is particularly sensitive to blood pressure reduction.

Additionally, ACCORD BP had a 2 × 2 factorial design involving a simultaneous comparison of intensive vs standard glycemic control, which may have influenced the effects due to blood pressure. Indeed, a post hoc analysis showed that there was a significant 26% lower risk of the primary outcome in ACCORD BP patients who received intensive systolic pressure  control plus standard glycemic control than in those receiving standard systolic control plus standard glycemic control.²³

Are more adverse events an acceptable trade-off?

Adverse events, including acute kidney injury, were more frequent in the intensive therapy group in SPRINT.

Acute kidney injury was coded as an adverse event on the basis of this diagnosis being included in the hospital discharge summary (as a primary or main secondary diagnosis) and if considered by the safety officer to be one of the top three reasons for admission or continued hospitalization. Further analysis of renal events should be forthcoming.

People in the intensive treatment group, on average, needed one more medication than those in the standard treatment group. Some of the adverse events may be related to the antihypertensive medications taken (eg, electrolyte abnormalities such as hyponatremia and hypokalemia due to diuretic use), and others may be related to blood pressure-lowering (eg, acute kidney injury due to renal hypoperfusion).

At this point, the long-term effects of these adverse events, especially on kidney function, are not known. Patients enrolled in clinical trials tend to be healthier than patients seen in clinical practice; thus, the rate of adverse events reported in the trial may be lower than one would see in the real world.

Does lower systolic pressure protect or harm the kidneys?

SPRINT included patients with stage 3 and 4 chronic kidney disease (ie, with eGFR 20–50 mL/min/1.73 m²), but it was designed to assess cardiovascular outcomes, not the progression of chronic kidney disease. The trial excluded patients with diabetic nephropathy or high degrees of proteinuria.

Only about half of hypertensive adults have their blood pressure under control, ie, < 140/90

Earlier randomized trials that focused on chronic kidney disease progression, including the MDRD²⁴ and the African American Study of Kidney Disease and Hypertension,²⁵ did not show benefit with more aggressive blood pressure-lowering (except in patients with higher degrees of proteinuria), and these trials were not powered to assess effects on cardiovascular outcomes.^24,25

The Irbesartan Diabetic Nephropathy Trial,^26,27 which was done in patients with overt diabetic nephropathy, showed that a progressively lower achieved systolic pressure down to 120 mm Hg predicted lower rates of heart failure, cardiovascular mortality, and renal events (although the trial target was ≤ 130/85 mm Hg and few participants achieved systolic pressure lower than 120 mm Hg).

IMPLICATIONS FOR MANAGEMENT

The recent estimates of hypertension prevalence and control from NHANES show that only about 53% of hypertensive adults have their blood pressure under control (defined as systolic pressure < 140 mm Hg and diastolic pressure < 90 mm Hg).² Analysis of the NHANES 2007–2012 data showed that 16.7% or 8.2 million US adults with treated hypertension meet the eligibility criteria for SPRINT.²⁸

Although the SPRINT results support the notion that “lower is better,” the risks and benefits of intensive control will need to be balanced in individual patients. Table 4 shows the number needed to treat and number needed to harm in the trial.

More aggressive management of hypertension is challenging. The median systolic pressure achieved in the intensive group in SPRINT was just over 120 mm Hg, which implies that at least half of the participants in the intensive group did not achieve the goal of less than 120 mm Hg. While it may be reasonable to aim for systolic pressure of less than 120 or 125 mm Hg in patients who fit the SPRINT criteria and can tolerate intensive blood pressure lowering, it would be prudent to aim for a more conservative goal in elderly patients who are frail and at risk for falls, considering the higher incidence of specified adverse events in the intensive group.

Results of cognitive outcomes, as well as data related to quality of life, are still awaited. Long-term renal outcomes are also unclear.

As noted above, the question of generalizability of SPRINT results to patients with diabetes is open to debate. In our opinion, with currently available evidence, it is difficult to conclusively answer the question of whether a lower systolic target provides cardiovascular benefit in diabetes. It is also unclear whether similar beneficial results would be seen with intensive treatment in a population at low cardiovascular risk. The American Heart Association and the American College of Cardiology are in the process of formulating new hypertension guidelines, and evidence from  SPRINT will inform any new recommendations.

As more medications will likely be needed for intensive systolic blood pressure control, side effects and tolerability of medications with polypharmacy and potential nonadherence with increasing complexity of medication regimens should be kept in mind. Lifestyle modifications will need to be emphasized and reinforced, with greater use of combination antihypertensive therapy.

The data from SPRINT indicate that lower systolic pressure is better, as long as untoward clinical events can be monitored and avoided or easily managed. Careful monitoring will likely entail more frequent clinic visits and more frequent assessment of renal function and electrolyte levels (participants in the intensive group in the trial were seen every month until goal was achieved). A team approach that includes pharmacists and nurse practitioners, along with optimal use of best practice algorithms and remote monitoring technology, will need to be implemented for efficient and effective care.

In a large phase trial, a combination drug that contains the angiotensin II receptor blocker (ARB) valsartan and the neprilysin inhibitor sacubitril was found to be superior to the angiotensin-converting enzyme (ACE) inhibitor enalapril in terms of important end points, including death and hospitalization for heart failure, in patients with heart failure with reduced ejection fraction.¹

See related editorial

Recently approved by the US Food and Drug Administration, this combination drug, marketed under the brand name Entresto, represents a new drug class, angiotensin receptor-neprilysin inhibitors, or ARNIs.

This article is an overview of the Prospective Comparison of ARNI With ACEI to Determine Impact on Global Mortality and Morbidity in Heart Failure (PARADIGM-HF) trial¹ and the implications it may have on the care of patients with chronic heart failure.

NEED FOR NEW HEART FAILURE DRUGS

Heart failure is a major public health problem, and the care of patients with heart failure is challenging.

Almost 6 million US adults have heart failure, and the prevalence is projected to increase in the next few decades as the population continues to age.² Furthermore, the total healthcare cost for heart failure patients was almost $31 billion in 2012 and is projected to rise to $70 billion by 2030.²

The care of patients with severely decompensated heart failure has changed dramatically in the last few decades with advances in heart transplantation and mechanical support devices. But day-to-day management of patients with chronic mildly to moderately symptomatic heart failure continues to pose a clinical challenge.

The drugs currently available for these patients include beta-blockers, ACE inhibitors, ARBs, aldosterone antagonists, digoxin, diuretics, and vasodilators. But even with these drugs, the death and readmission rates of patients with heart failure with reduced ejection fraction remain high. More than 50% of patients with heart failure die within 5 years of diagnosis,³ and 25% of patients hospitalized with heart failure are readmitted within 30 days of discharge.² Furthermore, death rates are higher in those patients who have a history of heart failure hospitalization.⁴

Although heart failure with preserved ejection fraction encompasses an important group of heart failure patients with high morbidity, the focus of this article will be on patients with heart failure with reduced ejection fraction.

Available drugs to date

The cornerstone drugs that lower the odds of death in patients with heart failure with reduced ejection fraction are ACE inhibitors, ARBs, beta-blockers, and mineralocorticoid antagonists.

ACE inhibitors were the first class of drugs shown to reduce the death rate in patients with heart failure with reduced ejection fraction. The landmark CONSENSUS trial,⁵ published in 1987, found that the death rate in patients who received enalapril was 27% lower than in those receiving placebo, an effect driven entirely by a reduction in progressive heart failure. Similarly, the SOLVD trial,⁶ published in 1991, showed a 26% reduction in heart failure hospitalization and a 16% lower rate of death with enalapril compared with placebo, an effect driven predominantly by a decrease in the progression of heart failure.

ARBs have also been shown to decrease the rate of death, although not by as much as ACE inhibitors. In the CHARM trial,⁷ compared with placebo, candesartan significantly decreased the risk of death from any cause, of death from cardiovascular causes, and of hospitalization related to heart failure.⁷

Beta-blockers. The MERIT-HF trial,⁸ published in 1999, was stopped early because fewer patients were dying in the group receiving metoprolol succinate than in the group receiving placebo (relative risk 0.66). Similarly, in 2001, the COPERNICUS trial⁹ reported a 34% reduction in deaths in patients receiving carvedilol in addition to an ACE inhibitor compared with those receiving an ACE inhibitor alone.

Mineralocorticoid receptor antagonists were found to be beneficial when added to standard therapy for chronic symptomatic heart failure in the RALES¹⁰ and EMPHASIS-HF¹¹ trials.

Vasodilators (specifically, the combination of isosorbide dinitrate and hydralazine) were found to have benefit in terms of mortality when added to standard therapy in African American patients in the A-HeEFT trial.¹²

WHY INHIBIT BOTH ANGIOTENSIN AND NEPRILYSIN?

The renin-angiotensin-aldosterone system is a major focus in treating heart failure, as overactivity of this system plays a key role in the pathophysiology of this disease. Therefore, essential drugs for heart failure patients include those that inhibit overactivity of this system such as ACE inhibitors, ARBs, and aldosterone antagonists.

The natriuretic peptide system is another important pathway that can be targeted in patients with heart failure. Natriuretic peptides are key molecules that counteract heart failure, as they contribute to diuresis and vasodilation and protect against vascular remodeling.¹³ An increased understanding of the importance of this system in slowing the progression of heart failure has motivated evaluation of drugs such as nesiritide in patients with symptomatic heart failure. However, these drugs can cause hypotension and have limited bioavailability.¹⁴

Neprilysin is an endopeptidase—an endogenous enzyme that degrades vasoactive peptides such as bradykinin and natriuretic peptides.¹⁴ Drugs that inhibit neprilysin increase the levels of these peptides and thus counteract neurohormonal stimuli that lead to cardiac remodeling, sodium retention, and vasoconstriction.¹⁵

However, neprilysin also hydrolyzes angiotensin I to angiotensin (1–7), an inhibitor of angiotensin II. Thus, inhibition of neprilysin alone could lead to increased activity of angiotensin II and so have an overall neutral effect. To be beneficial, neprilysin inhibition needs to be combined with renin-angiotensin system inhibition. Furthermore, the benefit of renin-angiotensin-aldosterone system blockade may be amplified by up-regulation of the endogenous natriuretic peptide system.¹⁵

Omapatrilat, the most studied combination neprilysin inhibitor and ACE inhibitor, improved cardiac function and decreased cardiac mass in animal experiments.¹⁵ In addition, this drug showed promise in terms of blood pressure, heart failure readmissions, death, and preservation of renal function when compared with ACE inhibitors in patients with heart failure.^15–17 But in clinical trials this drug posed a greater risk of hypotension, dizziness, and, its major shortcoming, an unacceptably high incidence of angioedema compared with ACE inhibitors.^15,16,18 This higher risk of angioedema is thought to be from inhibition of three enzymes that break down bradykinin: ACE, neprilysin, and aminopeptidase P.¹⁹

ARNIs contain an angiotensin receptor blocker rather than an ACE inhibitor, and thus in theory they may be associated with a lower risk of angioedema.¹⁹ Sacubitril-valsartan, the first drug of this class, contains its two constitutive drugs in a one-to-one molecular ratio (Figure 1).

PARADIGM-HF investigated the benefit of this drug in patients with systolic heart failure.¹

STUDY DESIGN AND OBJECTIVES

PARADIGM-HF was a double-blind, randomized controlled trial comparing sacubitril-valsartan and enalapril in patients with chronic systolic heart failure. As such, it was the first trial in recent years to study a new drug in comparison with a well-established heart failure drug rather than as an add-on strategy.¹

Inclusion criteria

To be included in the PARADIGM-HF trial, patients had to have:

• A left ventricular ejection fraction of 40% or less (later changed to ≤ 35%)
• New York Heart Association class II, III, or IV symptoms
• A B-type natriuretic peptide (BNP) level of at least 150 pg/mL or an N-terminal proBNP (NT-proBNP) level of at least 600 pg/mL; for patients hospitalized for heart failure within the previous 12 months, the cut points were lower (BNP ≥ 100 pg/mL or NT-proBNP ≥ 400 pg/mL).

End points

The primary end point was the composite of cardiovascular death or first hospitalization for heart failure. Other outcomes assessed were time to death from any cause, the change from baseline in the Kansas City Cardiomyopathy Questionnaire (KCCQ) score at 8 months, time to new-onset atrial fibrillation, and the time to decline in renal function (defined as end-stage renal disease or a decrease in estimated glomerular filtration rate of at least 50% from randomization). All end points were blindly adjudicated by a clinical end points committee.

Two run-in periods

The study enrolled 10,521 patients from 1,043 centers in 47 countries, who entered the initial run-in period consisting of 2 weeks of treatment with enalapril at the study dosage (10 mg twice daily) in order to ensure no unacceptable side effects. At this point, 1,102 patients exited the study, leaving 9,419 who entered the second run-in period.

The second run-in period consisted of 4 weeks of treatment with sacubitril-valsartan, initially at half the study regimen (100 mg twice daily) and eventually at the full study dosage (200 mg twice daily). During the second run-in period, 977 participants left the study, leaving a total of 8,442 patients who underwent randomization. Forty-three patients were then excluded (6 because of invalid randomization and 37 because of four sites that closed because of major violations of good clinical practice).

Of those randomized, 4,187 patients were assigned to the sacubitril-valsartan treatment group and 4,212 were assigned to the enalapril group. The investigators used an intention-to-treat analysis for this study.

Most patients had NYHA class II symptoms

The randomized patients had a mean age of 64 years, 75% were men, 66% were white, and 58% were from Europe (only 7% were from North America). The mean left ventricular ejection fraction was about 30%, and 60% of the study participants had an ischemic cause for their cardiomyopathy. Although one of the inclusion criteria was New York Heart Association class II, III, or IV symptoms, about 5% of the patients had class I symptoms. Seventy percent had class II symptoms, 24% had class III, and less than 1% had class IV symptoms.

At the time of randomization, 78% of the patients were taking an ACE inhibitor and 93% were taking a beta-blocker, but only a little more than half were taking a mineralocorticoid antagonist and only about 15% had an implantable cardioverter-defibrillator.

STUDY OUTCOMES

This study was designed to detect a 15% lower risk of cardiovascular death in the sacubitril-valsartan group. It was projected to continue for at least 34 months but was stopped early because of an overwhelming benefit of the new drug at a median follow-up of 27 months.

Major findings

The primary composite outcome (cardiovascular death or first hospitalization for heart failure)¹ occurred in 21.8% of the patients in the sacubitril-valsartan group vs 26.5% of patients in the enalapril group (hazard ratio [HR] 0.80, 95% confidence interval [CI] 0.73–0.87, P < .001). The number of patients who needed to be treated to prevent one occurrence of the primary composite outcome (100/absolute risk reduction) was only 21. The benefit was strong and consistent across both of the individual components of the composite outcome:

• Cardiovascular death 13.3% vs 16.5%, HR 0.80 (95% CI 0.71–0.89), P < .001
• First hospitalization for worsening heart failure 12.8% vs 15.6%, HR 0.79 (95% CI 0.71–0.89), P < .001.

Secondary outcomes. The sacubitril-valsartan group had a significantly lower rate of death from any cause (17.0% vs 19.8%, HR 0.84, 95% CI 0.76–0.93, P < .001) and a lower mean decrease in KCCQ clinical summary scores at 8 months (2.99 points vs 4.63 points, mean difference 1.64, 95% CI 0.63–2.65, P = .001). The KCCQ score measures subjective symptoms and physical limitations caused by heart failure; possible scores range from 0 to 100, with a higher score indicating better functional status. Notably, sacubitril-valsartan did not increase the KCCQ score in these patients; rather, sacubitril-valsartan recipients had a lower decrease in their scores than those in the enalapril group.

The incidence of new-onset atrial fibrillation was the same in both groups (3.1% and 3.1%).¹ A decline in renal function (defined as end-stage renal disease, a decrease of 50% or more in estimated glomerular filtration rate from the value at randomization or a decrease in the estimated glomerular filtration rate of more than 30 mL/min/1.73 m² to less than 60 mL/min/1.73 m²) occurred in 2.2% of the valsartan-sacubitril group and 2.6% of the enalapril group (P = .28).

The effects of the study drug on the primary composite outcome and on death from a cardiovascular cause were similar in all prespecified subgroups except for NYHA class: the reduction in the risk of the composite outcome was lower in sacubitril-valsartan recipients with NYHA I or II symptoms than in those with NYHA III or IV symptoms (P for interaction .03). However, there were no differences in the other prespecified subgroups, defined by age, sex, race, region, estimated glomerular filtration rate, diabetes, systolic blood pressure, ejection fraction, atrial fibrillation, NT-proBNP, hypertension, previous use of an ACE inhibitor, previous use of an aldosterone antagonist, previous heart failure hospitalization, and time since diagnosis of heart failure.

SAFETY: ANGIOEDEMA, HYPOTENSION, AND RENAL DYSFUNCTION

Angioedema

As noted above, the combination of neprilysin inhibitors and ACE inhibitors has been associated with an increased risk of angioedema. That was an important consideration before starting this study, which used a combination of a neprilysin inhibitor and an ARB in an attempt to avoid this serious side effect.

As it happened, there was no increased risk of significant angioedema with sacubitril-valsartan use compared with enalapril. Rates were similar to those in other studies, which showed a less than 1% risk of angioedema caused by ACE inhibitors.^20,21 Only 19 patients (0.5%) in the sacubitril-valsartan group and 10 patients (0.2%) in the enalapril group experienced any angioedema. Of these, just three patients in the sacubitril-valsartan group and one patient in the enalapril group experienced angioedema that required hospitalization (P = .31). None of these patients had airway compromise due to angioedema.

Hypotension, cough, renal dysfunction, hyperkalemia

Other safety issues that were assessed included hypotension, worsening renal function, increase in potassium levels, and cough. Patients in the sacubitril-valsartan group were more likely to have symptomatic hypotension than patients in the enalapril group (14.0% vs 9.2%, P < .001); however, the authors noted that this was a rare cause of drug discontinuation.

Patients in the sacubitril-valsartan group were less likely to develop cough (11.3% vs 14.3%, P < .001), a serum creatinine level of 2.5 mg/dL or more (3.3% vs 4.5%, P = .007), or a serum potassium level of more than 6.0 mmol/L (11.3% vs 14.3%, P = .007).¹

During the two run-in periods combined, 12% of the patients in the study withdrew because of adverse events, including cough, renal dysfunction, hyperkalemia, and symptomatic hypotension. During the enalapril run-in period, 591 patients (5.6%) withdrew due to adverse events, and 547 patients (5.8%) withdrew due to these events during the sacubitril-valsartan run-in period. After adjusting for the shorter time on enalapril, there was a higher rate of withdrawal because of adverse events from enalapril than from sacubitril-valsartan.

LOWER RISK OF CLINICAL PROGRESSION

In a separate paper,²² the PARADIGM-HF investigators reported that, among the survivors in the study, those who received sacubitril-valsartan fared better in terms of a number of markers of progression of heart failure, with lower rates of:

• Intensification of medical treatment for heart failure
• Emergency department visits for worsening heart failure
• Hospitalization for worsening heart failure
• Need for intensive care
• Need for intravenous inotropic agents
• Need for cardiac devices or heart transplants
• Worsening symptom scores
• Elevation of biomarkers of myocardial injury.

QUESTIONS AND CONCERNS

Although this study, which was funded by the manufacturer, showed consistent benefit for sacubitril-valsartan over enalapril, questions remain.

Are the findings generalizable?

Despite the study’s rigorous run-in period, 12% of patients dropped out because of adverse events, and thus the patients who completed the study may not be representative of the general population of heart failure patients. The authors included this double-level wash-out to ensure patient tolerance of both drugs. But in everyday practice, a significant number of patients may be unable to tolerate one of these drugs.

Moreover, after adjusting for the difference in the run-in periods, patients actually withdrew more often during the enalapril run-in period than during the sacubitril-valsartan run-in period. However, there may be overlap in tolerability in these two drugs, which both affect the renin-angiotensin-aldosterone system. Thus, the enalapril run-in period may have contributed to the lower tolerability of this drug compared with sacubitril-valsartan.

Were patients receiving the best possible therapy?

Another important point when considering how we treat heart failure patients in the United States is how few patients in this study were using cardiac implantable electronic devices. Only 15% of the patients in this study had an implantable cardioverter-defibrillator despite a mean left ventricular ejection fraction less than 30%. This likely reflects differences in practice internationally; however, based on American College of Cardiology, American Heart Association, and Heart Rhythm Society guidelines, these patients would have a class I indication for an implantable cardioverter-defibrillator for primary prevention of sudden cardiac death.²³

Therefore, based on these recommendations, the patients in this study were not necessarily on optimal medical and device therapy and furthermore may not be representative of heart failure patients in the United States.

Was enalapril 10 mg twice a day a fair comparison?

Another concern about the results of this study relates to the dosages used in the two treatment groups. The sacubitril-valsartan formulation included full-dose valsartan, whereas the ACE inhibitor group received enalapril at less than a full dose.

Although the authors explained that the dose of enalapril chosen for the study was based on the one used in previous studies that showed survival benefit, this raises the question of whether the significant difference in outcomes in this trial was due to a greater inhibition of the renin-angiotensin-aldosterone system related to a higher dose of drug in the sacubitril-valsartan group.

What about black patients taking hydralazine-isosorbide?

Only about 5% of patients in the PARADIGM-HF trial were black. Based on the A-HeFT study results, black patients can be prescribed an ACE inhibitor as well as hydralazine and isosorbide dinitrate as tolerated to decrease the risk of death. Does sacubitril-valsartan offer benefit to these patients compared with a regimen of an ACE inhibitor, hydralazine, and isosorbide dinitrate?

Another concern is that the incidence of angioedema observed with ACE inhibitors and omapatrilat is higher in black patients.^15,21 Thus, it would be prudent to investigate whether the risk of angioedema with sacubitril-valsartan would be higher if more black patients are studied.

IMPLICATIONS AND CONSIDERATIONS

In this study, sacubitril-valsartan showed impressive and consistent results, with an almost 20% decrease in the composite end point of heart failure hospitalization or cardiovascular death and a similar decrease in the composite outcomes with a very low number needed to treat (21 patients). It did not show a decrease in the incidence of new-onset atrial fibrillation; however, only 80 cases of atrial fibrillation were reported, so there may have been a lack of statistical power to detect a difference.

To avoid angioedema, wait at least 36 hours after stopping an ACE inhibitor. Sacubitril-valsartan was not associated with an increased risk of severe angioedema, and no patients experienced life-threatening angioedema. In the trial, the sacubitril-valsartan run-in period was started at least 24 hours after enalapril was stopped, and thus the authors recommended at least a 1-day washout period after discontinuing an ACE inhibitor to avoid angioedema in patients starting sacubitril-valsartan.

Hypotension is a concern. Although there was actually a decreased risk of renal dysfunction, hyperkalemia, and cough compared with enalapril, there was a significantly increased rate of symptomatic hypotension in the sacubitril-valsartan group, which raises the question of patient tolerance and physician comfort when prescribing and titrating this drug in clinical practice. This side effect will be an important consideration when attempting to titrate the drug to target doses.

Start treatment early. This trial largely consisted of patients with NYHA class II or III symptoms, with about 70% of patients with class II symptoms. Since this drug showed benefit in patients with mildly to moderately symptomatic heart failure, clinicians who are considering prescribing this drug should not wait until the patient is closer to end-stage disease. Patients with mildly symptomatic heart failure may be followed by a general cardiologist, internist, or both, and thus it is important to emphasize to the entire medical community the need to start this medication early on.

How much will it cost? Cost is a concern that could heavily weigh on the decision to prescribe this drug. Generic ACE inhibitors are relatively inexpensive, and it may difficult to switch from an affordable generic drug to a new drug that is likely to be much more expensive. Arguably, this drug may be cost-effective in the long run owing to a large decrease in heart failure readmissions. We await further analyses to evaluate this issue.

Will patients take a twice-a-day drug as prescribed? Most patients who are prescribed an ACE inhibitor take it just once a day, and switching from a daily to a twice-daily drug may present a challenge for some.

What about other outcomes? Based on this study, it is unclear what effect sacubitril-valsartan has on the incidence of fatal arrhythmias, sudden cardiac death, and pump failure. Furthermore, the effect on quality of life is still uncertain. Quality of life is an integral component in the evaluation of heart failure patients, and in this study the changes in KCCQ scores were not impressive. We hope to see further evaluations of this drug’s impact on quality of life of patients with heart failure. Furthermore, it would be interesting to study if this drug has any long-term effects on the need for advanced therapies such as left ventricular assist devices and orthotopic heart transplant.

What about patients with heart failure with preserved ejection fraction? This study included only patients with heart failure with reduced ejection fraction. However PARAMOUNT, a phase 2 study that evaluated the benefit of sacubitril-valsartan in patients with heart failure with preserved ejection fraction, has shown encouraging results.²⁴ We look forward to further investigation of this agent in patients with heart failure with preserved ejection fraction.

Sacubitril-valsartan, the first ARNI to be studied in humans, has a dual action in that it enhances the activity of the natriuretic peptide system and inhibits that of the renin-angiotensin-aldosterone system. It is the first drug in over a decade to show mortality benefit in patients with chronic systolic heart failure when compared with an already well-established heart failure medication. It appears to decrease rates of mortality and heart failure hospitalization without increasing the risk of severe angioedema in patients with mild or moderate chronic systolic heart failure. Symptomatic hypotension and high cost may pose the largest barriers to the use of this new drug. And we have yet to see how the clinical community and patients with heart failure will respond to it.

In a large phase trial, a combination drug that contains the angiotensin II receptor blocker (ARB) valsartan and the neprilysin inhibitor sacubitril was found to be superior to the angiotensin-converting enzyme (ACE) inhibitor enalapril in terms of important end points, including death and hospitalization for heart failure, in patients with heart failure with reduced ejection fraction.¹

See related editorial

Recently approved by the US Food and Drug Administration, this combination drug, marketed under the brand name Entresto, represents a new drug class, angiotensin receptor-neprilysin inhibitors, or ARNIs.

This article is an overview of the Prospective Comparison of ARNI With ACEI to Determine Impact on Global Mortality and Morbidity in Heart Failure (PARADIGM-HF) trial¹ and the implications it may have on the care of patients with chronic heart failure.

NEED FOR NEW HEART FAILURE DRUGS

Heart failure is a major public health problem, and the care of patients with heart failure is challenging.

Almost 6 million US adults have heart failure, and the prevalence is projected to increase in the next few decades as the population continues to age.² Furthermore, the total healthcare cost for heart failure patients was almost $31 billion in 2012 and is projected to rise to $70 billion by 2030.²

The care of patients with severely decompensated heart failure has changed dramatically in the last few decades with advances in heart transplantation and mechanical support devices. But day-to-day management of patients with chronic mildly to moderately symptomatic heart failure continues to pose a clinical challenge.

The drugs currently available for these patients include beta-blockers, ACE inhibitors, ARBs, aldosterone antagonists, digoxin, diuretics, and vasodilators. But even with these drugs, the death and readmission rates of patients with heart failure with reduced ejection fraction remain high. More than 50% of patients with heart failure die within 5 years of diagnosis,³ and 25% of patients hospitalized with heart failure are readmitted within 30 days of discharge.² Furthermore, death rates are higher in those patients who have a history of heart failure hospitalization.⁴

Although heart failure with preserved ejection fraction encompasses an important group of heart failure patients with high morbidity, the focus of this article will be on patients with heart failure with reduced ejection fraction.

Available drugs to date

The cornerstone drugs that lower the odds of death in patients with heart failure with reduced ejection fraction are ACE inhibitors, ARBs, beta-blockers, and mineralocorticoid antagonists.

ACE inhibitors were the first class of drugs shown to reduce the death rate in patients with heart failure with reduced ejection fraction. The landmark CONSENSUS trial,⁵ published in 1987, found that the death rate in patients who received enalapril was 27% lower than in those receiving placebo, an effect driven entirely by a reduction in progressive heart failure. Similarly, the SOLVD trial,⁶ published in 1991, showed a 26% reduction in heart failure hospitalization and a 16% lower rate of death with enalapril compared with placebo, an effect driven predominantly by a decrease in the progression of heart failure.

ARBs have also been shown to decrease the rate of death, although not by as much as ACE inhibitors. In the CHARM trial,⁷ compared with placebo, candesartan significantly decreased the risk of death from any cause, of death from cardiovascular causes, and of hospitalization related to heart failure.⁷

Beta-blockers. The MERIT-HF trial,⁸ published in 1999, was stopped early because fewer patients were dying in the group receiving metoprolol succinate than in the group receiving placebo (relative risk 0.66). Similarly, in 2001, the COPERNICUS trial⁹ reported a 34% reduction in deaths in patients receiving carvedilol in addition to an ACE inhibitor compared with those receiving an ACE inhibitor alone.

Mineralocorticoid receptor antagonists were found to be beneficial when added to standard therapy for chronic symptomatic heart failure in the RALES¹⁰ and EMPHASIS-HF¹¹ trials.

Vasodilators (specifically, the combination of isosorbide dinitrate and hydralazine) were found to have benefit in terms of mortality when added to standard therapy in African American patients in the A-HeEFT trial.¹²

WHY INHIBIT BOTH ANGIOTENSIN AND NEPRILYSIN?

The renin-angiotensin-aldosterone system is a major focus in treating heart failure, as overactivity of this system plays a key role in the pathophysiology of this disease. Therefore, essential drugs for heart failure patients include those that inhibit overactivity of this system such as ACE inhibitors, ARBs, and aldosterone antagonists.

The natriuretic peptide system is another important pathway that can be targeted in patients with heart failure. Natriuretic peptides are key molecules that counteract heart failure, as they contribute to diuresis and vasodilation and protect against vascular remodeling.¹³ An increased understanding of the importance of this system in slowing the progression of heart failure has motivated evaluation of drugs such as nesiritide in patients with symptomatic heart failure. However, these drugs can cause hypotension and have limited bioavailability.¹⁴

Neprilysin is an endopeptidase—an endogenous enzyme that degrades vasoactive peptides such as bradykinin and natriuretic peptides.¹⁴ Drugs that inhibit neprilysin increase the levels of these peptides and thus counteract neurohormonal stimuli that lead to cardiac remodeling, sodium retention, and vasoconstriction.¹⁵

However, neprilysin also hydrolyzes angiotensin I to angiotensin (1–7), an inhibitor of angiotensin II. Thus, inhibition of neprilysin alone could lead to increased activity of angiotensin II and so have an overall neutral effect. To be beneficial, neprilysin inhibition needs to be combined with renin-angiotensin system inhibition. Furthermore, the benefit of renin-angiotensin-aldosterone system blockade may be amplified by up-regulation of the endogenous natriuretic peptide system.¹⁵

Omapatrilat, the most studied combination neprilysin inhibitor and ACE inhibitor, improved cardiac function and decreased cardiac mass in animal experiments.¹⁵ In addition, this drug showed promise in terms of blood pressure, heart failure readmissions, death, and preservation of renal function when compared with ACE inhibitors in patients with heart failure.^15–17 But in clinical trials this drug posed a greater risk of hypotension, dizziness, and, its major shortcoming, an unacceptably high incidence of angioedema compared with ACE inhibitors.^15,16,18 This higher risk of angioedema is thought to be from inhibition of three enzymes that break down bradykinin: ACE, neprilysin, and aminopeptidase P.¹⁹

ARNIs contain an angiotensin receptor blocker rather than an ACE inhibitor, and thus in theory they may be associated with a lower risk of angioedema.¹⁹ Sacubitril-valsartan, the first drug of this class, contains its two constitutive drugs in a one-to-one molecular ratio (Figure 1).

PARADIGM-HF investigated the benefit of this drug in patients with systolic heart failure.¹

STUDY DESIGN AND OBJECTIVES

PARADIGM-HF was a double-blind, randomized controlled trial comparing sacubitril-valsartan and enalapril in patients with chronic systolic heart failure. As such, it was the first trial in recent years to study a new drug in comparison with a well-established heart failure drug rather than as an add-on strategy.¹

Inclusion criteria

To be included in the PARADIGM-HF trial, patients had to have:

• A left ventricular ejection fraction of 40% or less (later changed to ≤ 35%)
• New York Heart Association class II, III, or IV symptoms
• A B-type natriuretic peptide (BNP) level of at least 150 pg/mL or an N-terminal proBNP (NT-proBNP) level of at least 600 pg/mL; for patients hospitalized for heart failure within the previous 12 months, the cut points were lower (BNP ≥ 100 pg/mL or NT-proBNP ≥ 400 pg/mL).

End points

The primary end point was the composite of cardiovascular death or first hospitalization for heart failure. Other outcomes assessed were time to death from any cause, the change from baseline in the Kansas City Cardiomyopathy Questionnaire (KCCQ) score at 8 months, time to new-onset atrial fibrillation, and the time to decline in renal function (defined as end-stage renal disease or a decrease in estimated glomerular filtration rate of at least 50% from randomization). All end points were blindly adjudicated by a clinical end points committee.

Two run-in periods

The study enrolled 10,521 patients from 1,043 centers in 47 countries, who entered the initial run-in period consisting of 2 weeks of treatment with enalapril at the study dosage (10 mg twice daily) in order to ensure no unacceptable side effects. At this point, 1,102 patients exited the study, leaving 9,419 who entered the second run-in period.

The second run-in period consisted of 4 weeks of treatment with sacubitril-valsartan, initially at half the study regimen (100 mg twice daily) and eventually at the full study dosage (200 mg twice daily). During the second run-in period, 977 participants left the study, leaving a total of 8,442 patients who underwent randomization. Forty-three patients were then excluded (6 because of invalid randomization and 37 because of four sites that closed because of major violations of good clinical practice).

Of those randomized, 4,187 patients were assigned to the sacubitril-valsartan treatment group and 4,212 were assigned to the enalapril group. The investigators used an intention-to-treat analysis for this study.

Most patients had NYHA class II symptoms

The randomized patients had a mean age of 64 years, 75% were men, 66% were white, and 58% were from Europe (only 7% were from North America). The mean left ventricular ejection fraction was about 30%, and 60% of the study participants had an ischemic cause for their cardiomyopathy. Although one of the inclusion criteria was New York Heart Association class II, III, or IV symptoms, about 5% of the patients had class I symptoms. Seventy percent had class II symptoms, 24% had class III, and less than 1% had class IV symptoms.

At the time of randomization, 78% of the patients were taking an ACE inhibitor and 93% were taking a beta-blocker, but only a little more than half were taking a mineralocorticoid antagonist and only about 15% had an implantable cardioverter-defibrillator.

STUDY OUTCOMES

This study was designed to detect a 15% lower risk of cardiovascular death in the sacubitril-valsartan group. It was projected to continue for at least 34 months but was stopped early because of an overwhelming benefit of the new drug at a median follow-up of 27 months.

Major findings

The primary composite outcome (cardiovascular death or first hospitalization for heart failure)¹ occurred in 21.8% of the patients in the sacubitril-valsartan group vs 26.5% of patients in the enalapril group (hazard ratio [HR] 0.80, 95% confidence interval [CI] 0.73–0.87, P < .001). The number of patients who needed to be treated to prevent one occurrence of the primary composite outcome (100/absolute risk reduction) was only 21. The benefit was strong and consistent across both of the individual components of the composite outcome:

• Cardiovascular death 13.3% vs 16.5%, HR 0.80 (95% CI 0.71–0.89), P < .001
• First hospitalization for worsening heart failure 12.8% vs 15.6%, HR 0.79 (95% CI 0.71–0.89), P < .001.

Secondary outcomes. The sacubitril-valsartan group had a significantly lower rate of death from any cause (17.0% vs 19.8%, HR 0.84, 95% CI 0.76–0.93, P < .001) and a lower mean decrease in KCCQ clinical summary scores at 8 months (2.99 points vs 4.63 points, mean difference 1.64, 95% CI 0.63–2.65, P = .001). The KCCQ score measures subjective symptoms and physical limitations caused by heart failure; possible scores range from 0 to 100, with a higher score indicating better functional status. Notably, sacubitril-valsartan did not increase the KCCQ score in these patients; rather, sacubitril-valsartan recipients had a lower decrease in their scores than those in the enalapril group.

The incidence of new-onset atrial fibrillation was the same in both groups (3.1% and 3.1%).¹ A decline in renal function (defined as end-stage renal disease, a decrease of 50% or more in estimated glomerular filtration rate from the value at randomization or a decrease in the estimated glomerular filtration rate of more than 30 mL/min/1.73 m² to less than 60 mL/min/1.73 m²) occurred in 2.2% of the valsartan-sacubitril group and 2.6% of the enalapril group (P = .28).

The effects of the study drug on the primary composite outcome and on death from a cardiovascular cause were similar in all prespecified subgroups except for NYHA class: the reduction in the risk of the composite outcome was lower in sacubitril-valsartan recipients with NYHA I or II symptoms than in those with NYHA III or IV symptoms (P for interaction .03). However, there were no differences in the other prespecified subgroups, defined by age, sex, race, region, estimated glomerular filtration rate, diabetes, systolic blood pressure, ejection fraction, atrial fibrillation, NT-proBNP, hypertension, previous use of an ACE inhibitor, previous use of an aldosterone antagonist, previous heart failure hospitalization, and time since diagnosis of heart failure.

SAFETY: ANGIOEDEMA, HYPOTENSION, AND RENAL DYSFUNCTION

Angioedema

As noted above, the combination of neprilysin inhibitors and ACE inhibitors has been associated with an increased risk of angioedema. That was an important consideration before starting this study, which used a combination of a neprilysin inhibitor and an ARB in an attempt to avoid this serious side effect.

As it happened, there was no increased risk of significant angioedema with sacubitril-valsartan use compared with enalapril. Rates were similar to those in other studies, which showed a less than 1% risk of angioedema caused by ACE inhibitors.^20,21 Only 19 patients (0.5%) in the sacubitril-valsartan group and 10 patients (0.2%) in the enalapril group experienced any angioedema. Of these, just three patients in the sacubitril-valsartan group and one patient in the enalapril group experienced angioedema that required hospitalization (P = .31). None of these patients had airway compromise due to angioedema.

Hypotension, cough, renal dysfunction, hyperkalemia

Other safety issues that were assessed included hypotension, worsening renal function, increase in potassium levels, and cough. Patients in the sacubitril-valsartan group were more likely to have symptomatic hypotension than patients in the enalapril group (14.0% vs 9.2%, P < .001); however, the authors noted that this was a rare cause of drug discontinuation.

Patients in the sacubitril-valsartan group were less likely to develop cough (11.3% vs 14.3%, P < .001), a serum creatinine level of 2.5 mg/dL or more (3.3% vs 4.5%, P = .007), or a serum potassium level of more than 6.0 mmol/L (11.3% vs 14.3%, P = .007).¹

During the two run-in periods combined, 12% of the patients in the study withdrew because of adverse events, including cough, renal dysfunction, hyperkalemia, and symptomatic hypotension. During the enalapril run-in period, 591 patients (5.6%) withdrew due to adverse events, and 547 patients (5.8%) withdrew due to these events during the sacubitril-valsartan run-in period. After adjusting for the shorter time on enalapril, there was a higher rate of withdrawal because of adverse events from enalapril than from sacubitril-valsartan.

LOWER RISK OF CLINICAL PROGRESSION

In a separate paper,²² the PARADIGM-HF investigators reported that, among the survivors in the study, those who received sacubitril-valsartan fared better in terms of a number of markers of progression of heart failure, with lower rates of:

• Intensification of medical treatment for heart failure
• Emergency department visits for worsening heart failure
• Hospitalization for worsening heart failure
• Need for intensive care
• Need for intravenous inotropic agents
• Need for cardiac devices or heart transplants
• Worsening symptom scores
• Elevation of biomarkers of myocardial injury.

QUESTIONS AND CONCERNS

Although this study, which was funded by the manufacturer, showed consistent benefit for sacubitril-valsartan over enalapril, questions remain.

Are the findings generalizable?

Despite the study’s rigorous run-in period, 12% of patients dropped out because of adverse events, and thus the patients who completed the study may not be representative of the general population of heart failure patients. The authors included this double-level wash-out to ensure patient tolerance of both drugs. But in everyday practice, a significant number of patients may be unable to tolerate one of these drugs.

Moreover, after adjusting for the difference in the run-in periods, patients actually withdrew more often during the enalapril run-in period than during the sacubitril-valsartan run-in period. However, there may be overlap in tolerability in these two drugs, which both affect the renin-angiotensin-aldosterone system. Thus, the enalapril run-in period may have contributed to the lower tolerability of this drug compared with sacubitril-valsartan.

Were patients receiving the best possible therapy?

Another important point when considering how we treat heart failure patients in the United States is how few patients in this study were using cardiac implantable electronic devices. Only 15% of the patients in this study had an implantable cardioverter-defibrillator despite a mean left ventricular ejection fraction less than 30%. This likely reflects differences in practice internationally; however, based on American College of Cardiology, American Heart Association, and Heart Rhythm Society guidelines, these patients would have a class I indication for an implantable cardioverter-defibrillator for primary prevention of sudden cardiac death.²³

Therefore, based on these recommendations, the patients in this study were not necessarily on optimal medical and device therapy and furthermore may not be representative of heart failure patients in the United States.

Was enalapril 10 mg twice a day a fair comparison?

Another concern about the results of this study relates to the dosages used in the two treatment groups. The sacubitril-valsartan formulation included full-dose valsartan, whereas the ACE inhibitor group received enalapril at less than a full dose.

Although the authors explained that the dose of enalapril chosen for the study was based on the one used in previous studies that showed survival benefit, this raises the question of whether the significant difference in outcomes in this trial was due to a greater inhibition of the renin-angiotensin-aldosterone system related to a higher dose of drug in the sacubitril-valsartan group.

What about black patients taking hydralazine-isosorbide?

Only about 5% of patients in the PARADIGM-HF trial were black. Based on the A-HeFT study results, black patients can be prescribed an ACE inhibitor as well as hydralazine and isosorbide dinitrate as tolerated to decrease the risk of death. Does sacubitril-valsartan offer benefit to these patients compared with a regimen of an ACE inhibitor, hydralazine, and isosorbide dinitrate?

Another concern is that the incidence of angioedema observed with ACE inhibitors and omapatrilat is higher in black patients.^15,21 Thus, it would be prudent to investigate whether the risk of angioedema with sacubitril-valsartan would be higher if more black patients are studied.

IMPLICATIONS AND CONSIDERATIONS

In this study, sacubitril-valsartan showed impressive and consistent results, with an almost 20% decrease in the composite end point of heart failure hospitalization or cardiovascular death and a similar decrease in the composite outcomes with a very low number needed to treat (21 patients). It did not show a decrease in the incidence of new-onset atrial fibrillation; however, only 80 cases of atrial fibrillation were reported, so there may have been a lack of statistical power to detect a difference.

To avoid angioedema, wait at least 36 hours after stopping an ACE inhibitor. Sacubitril-valsartan was not associated with an increased risk of severe angioedema, and no patients experienced life-threatening angioedema. In the trial, the sacubitril-valsartan run-in period was started at least 24 hours after enalapril was stopped, and thus the authors recommended at least a 1-day washout period after discontinuing an ACE inhibitor to avoid angioedema in patients starting sacubitril-valsartan.

Hypotension is a concern. Although there was actually a decreased risk of renal dysfunction, hyperkalemia, and cough compared with enalapril, there was a significantly increased rate of symptomatic hypotension in the sacubitril-valsartan group, which raises the question of patient tolerance and physician comfort when prescribing and titrating this drug in clinical practice. This side effect will be an important consideration when attempting to titrate the drug to target doses.

Start treatment early. This trial largely consisted of patients with NYHA class II or III symptoms, with about 70% of patients with class II symptoms. Since this drug showed benefit in patients with mildly to moderately symptomatic heart failure, clinicians who are considering prescribing this drug should not wait until the patient is closer to end-stage disease. Patients with mildly symptomatic heart failure may be followed by a general cardiologist, internist, or both, and thus it is important to emphasize to the entire medical community the need to start this medication early on.

How much will it cost? Cost is a concern that could heavily weigh on the decision to prescribe this drug. Generic ACE inhibitors are relatively inexpensive, and it may difficult to switch from an affordable generic drug to a new drug that is likely to be much more expensive. Arguably, this drug may be cost-effective in the long run owing to a large decrease in heart failure readmissions. We await further analyses to evaluate this issue.

Will patients take a twice-a-day drug as prescribed? Most patients who are prescribed an ACE inhibitor take it just once a day, and switching from a daily to a twice-daily drug may present a challenge for some.

What about other outcomes? Based on this study, it is unclear what effect sacubitril-valsartan has on the incidence of fatal arrhythmias, sudden cardiac death, and pump failure. Furthermore, the effect on quality of life is still uncertain. Quality of life is an integral component in the evaluation of heart failure patients, and in this study the changes in KCCQ scores were not impressive. We hope to see further evaluations of this drug’s impact on quality of life of patients with heart failure. Furthermore, it would be interesting to study if this drug has any long-term effects on the need for advanced therapies such as left ventricular assist devices and orthotopic heart transplant.

What about patients with heart failure with preserved ejection fraction? This study included only patients with heart failure with reduced ejection fraction. However PARAMOUNT, a phase 2 study that evaluated the benefit of sacubitril-valsartan in patients with heart failure with preserved ejection fraction, has shown encouraging results.²⁴ We look forward to further investigation of this agent in patients with heart failure with preserved ejection fraction.

Sacubitril-valsartan, the first ARNI to be studied in humans, has a dual action in that it enhances the activity of the natriuretic peptide system and inhibits that of the renin-angiotensin-aldosterone system. It is the first drug in over a decade to show mortality benefit in patients with chronic systolic heart failure when compared with an already well-established heart failure medication. It appears to decrease rates of mortality and heart failure hospitalization without increasing the risk of severe angioedema in patients with mild or moderate chronic systolic heart failure. Symptomatic hypotension and high cost may pose the largest barriers to the use of this new drug. And we have yet to see how the clinical community and patients with heart failure will respond to it.

In a large phase trial, a combination drug that contains the angiotensin II receptor blocker (ARB) valsartan and the neprilysin inhibitor sacubitril was found to be superior to the angiotensin-converting enzyme (ACE) inhibitor enalapril in terms of important end points, including death and hospitalization for heart failure, in patients with heart failure with reduced ejection fraction.¹

See related editorial

Recently approved by the US Food and Drug Administration, this combination drug, marketed under the brand name Entresto, represents a new drug class, angiotensin receptor-neprilysin inhibitors, or ARNIs.

This article is an overview of the Prospective Comparison of ARNI With ACEI to Determine Impact on Global Mortality and Morbidity in Heart Failure (PARADIGM-HF) trial¹ and the implications it may have on the care of patients with chronic heart failure.

NEED FOR NEW HEART FAILURE DRUGS

Heart failure is a major public health problem, and the care of patients with heart failure is challenging.

Almost 6 million US adults have heart failure, and the prevalence is projected to increase in the next few decades as the population continues to age.² Furthermore, the total healthcare cost for heart failure patients was almost $31 billion in 2012 and is projected to rise to $70 billion by 2030.²

The care of patients with severely decompensated heart failure has changed dramatically in the last few decades with advances in heart transplantation and mechanical support devices. But day-to-day management of patients with chronic mildly to moderately symptomatic heart failure continues to pose a clinical challenge.

The drugs currently available for these patients include beta-blockers, ACE inhibitors, ARBs, aldosterone antagonists, digoxin, diuretics, and vasodilators. But even with these drugs, the death and readmission rates of patients with heart failure with reduced ejection fraction remain high. More than 50% of patients with heart failure die within 5 years of diagnosis,³ and 25% of patients hospitalized with heart failure are readmitted within 30 days of discharge.² Furthermore, death rates are higher in those patients who have a history of heart failure hospitalization.⁴

Although heart failure with preserved ejection fraction encompasses an important group of heart failure patients with high morbidity, the focus of this article will be on patients with heart failure with reduced ejection fraction.

Available drugs to date

The cornerstone drugs that lower the odds of death in patients with heart failure with reduced ejection fraction are ACE inhibitors, ARBs, beta-blockers, and mineralocorticoid antagonists.

ACE inhibitors were the first class of drugs shown to reduce the death rate in patients with heart failure with reduced ejection fraction. The landmark CONSENSUS trial,⁵ published in 1987, found that the death rate in patients who received enalapril was 27% lower than in those receiving placebo, an effect driven entirely by a reduction in progressive heart failure. Similarly, the SOLVD trial,⁶ published in 1991, showed a 26% reduction in heart failure hospitalization and a 16% lower rate of death with enalapril compared with placebo, an effect driven predominantly by a decrease in the progression of heart failure.

ARBs have also been shown to decrease the rate of death, although not by as much as ACE inhibitors. In the CHARM trial,⁷ compared with placebo, candesartan significantly decreased the risk of death from any cause, of death from cardiovascular causes, and of hospitalization related to heart failure.⁷

Beta-blockers. The MERIT-HF trial,⁸ published in 1999, was stopped early because fewer patients were dying in the group receiving metoprolol succinate than in the group receiving placebo (relative risk 0.66). Similarly, in 2001, the COPERNICUS trial⁹ reported a 34% reduction in deaths in patients receiving carvedilol in addition to an ACE inhibitor compared with those receiving an ACE inhibitor alone.

Mineralocorticoid receptor antagonists were found to be beneficial when added to standard therapy for chronic symptomatic heart failure in the RALES¹⁰ and EMPHASIS-HF¹¹ trials.

Vasodilators (specifically, the combination of isosorbide dinitrate and hydralazine) were found to have benefit in terms of mortality when added to standard therapy in African American patients in the A-HeEFT trial.¹²

WHY INHIBIT BOTH ANGIOTENSIN AND NEPRILYSIN?

The renin-angiotensin-aldosterone system is a major focus in treating heart failure, as overactivity of this system plays a key role in the pathophysiology of this disease. Therefore, essential drugs for heart failure patients include those that inhibit overactivity of this system such as ACE inhibitors, ARBs, and aldosterone antagonists.

The natriuretic peptide system is another important pathway that can be targeted in patients with heart failure. Natriuretic peptides are key molecules that counteract heart failure, as they contribute to diuresis and vasodilation and protect against vascular remodeling.¹³ An increased understanding of the importance of this system in slowing the progression of heart failure has motivated evaluation of drugs such as nesiritide in patients with symptomatic heart failure. However, these drugs can cause hypotension and have limited bioavailability.¹⁴

Neprilysin is an endopeptidase—an endogenous enzyme that degrades vasoactive peptides such as bradykinin and natriuretic peptides.¹⁴ Drugs that inhibit neprilysin increase the levels of these peptides and thus counteract neurohormonal stimuli that lead to cardiac remodeling, sodium retention, and vasoconstriction.¹⁵

However, neprilysin also hydrolyzes angiotensin I to angiotensin (1–7), an inhibitor of angiotensin II. Thus, inhibition of neprilysin alone could lead to increased activity of angiotensin II and so have an overall neutral effect. To be beneficial, neprilysin inhibition needs to be combined with renin-angiotensin system inhibition. Furthermore, the benefit of renin-angiotensin-aldosterone system blockade may be amplified by up-regulation of the endogenous natriuretic peptide system.¹⁵

Omapatrilat, the most studied combination neprilysin inhibitor and ACE inhibitor, improved cardiac function and decreased cardiac mass in animal experiments.¹⁵ In addition, this drug showed promise in terms of blood pressure, heart failure readmissions, death, and preservation of renal function when compared with ACE inhibitors in patients with heart failure.^15–17 But in clinical trials this drug posed a greater risk of hypotension, dizziness, and, its major shortcoming, an unacceptably high incidence of angioedema compared with ACE inhibitors.^15,16,18 This higher risk of angioedema is thought to be from inhibition of three enzymes that break down bradykinin: ACE, neprilysin, and aminopeptidase P.¹⁹

ARNIs contain an angiotensin receptor blocker rather than an ACE inhibitor, and thus in theory they may be associated with a lower risk of angioedema.¹⁹ Sacubitril-valsartan, the first drug of this class, contains its two constitutive drugs in a one-to-one molecular ratio (Figure 1).

PARADIGM-HF investigated the benefit of this drug in patients with systolic heart failure.¹

STUDY DESIGN AND OBJECTIVES

PARADIGM-HF was a double-blind, randomized controlled trial comparing sacubitril-valsartan and enalapril in patients with chronic systolic heart failure. As such, it was the first trial in recent years to study a new drug in comparison with a well-established heart failure drug rather than as an add-on strategy.¹

Inclusion criteria

To be included in the PARADIGM-HF trial, patients had to have:

• A left ventricular ejection fraction of 40% or less (later changed to ≤ 35%)
• New York Heart Association class II, III, or IV symptoms
• A B-type natriuretic peptide (BNP) level of at least 150 pg/mL or an N-terminal proBNP (NT-proBNP) level of at least 600 pg/mL; for patients hospitalized for heart failure within the previous 12 months, the cut points were lower (BNP ≥ 100 pg/mL or NT-proBNP ≥ 400 pg/mL).

End points

The primary end point was the composite of cardiovascular death or first hospitalization for heart failure. Other outcomes assessed were time to death from any cause, the change from baseline in the Kansas City Cardiomyopathy Questionnaire (KCCQ) score at 8 months, time to new-onset atrial fibrillation, and the time to decline in renal function (defined as end-stage renal disease or a decrease in estimated glomerular filtration rate of at least 50% from randomization). All end points were blindly adjudicated by a clinical end points committee.

Two run-in periods

The study enrolled 10,521 patients from 1,043 centers in 47 countries, who entered the initial run-in period consisting of 2 weeks of treatment with enalapril at the study dosage (10 mg twice daily) in order to ensure no unacceptable side effects. At this point, 1,102 patients exited the study, leaving 9,419 who entered the second run-in period.

The second run-in period consisted of 4 weeks of treatment with sacubitril-valsartan, initially at half the study regimen (100 mg twice daily) and eventually at the full study dosage (200 mg twice daily). During the second run-in period, 977 participants left the study, leaving a total of 8,442 patients who underwent randomization. Forty-three patients were then excluded (6 because of invalid randomization and 37 because of four sites that closed because of major violations of good clinical practice).

Of those randomized, 4,187 patients were assigned to the sacubitril-valsartan treatment group and 4,212 were assigned to the enalapril group. The investigators used an intention-to-treat analysis for this study.

Most patients had NYHA class II symptoms

The randomized patients had a mean age of 64 years, 75% were men, 66% were white, and 58% were from Europe (only 7% were from North America). The mean left ventricular ejection fraction was about 30%, and 60% of the study participants had an ischemic cause for their cardiomyopathy. Although one of the inclusion criteria was New York Heart Association class II, III, or IV symptoms, about 5% of the patients had class I symptoms. Seventy percent had class II symptoms, 24% had class III, and less than 1% had class IV symptoms.

At the time of randomization, 78% of the patients were taking an ACE inhibitor and 93% were taking a beta-blocker, but only a little more than half were taking a mineralocorticoid antagonist and only about 15% had an implantable cardioverter-defibrillator.

STUDY OUTCOMES

This study was designed to detect a 15% lower risk of cardiovascular death in the sacubitril-valsartan group. It was projected to continue for at least 34 months but was stopped early because of an overwhelming benefit of the new drug at a median follow-up of 27 months.

Major findings

The primary composite outcome (cardiovascular death or first hospitalization for heart failure)¹ occurred in 21.8% of the patients in the sacubitril-valsartan group vs 26.5% of patients in the enalapril group (hazard ratio [HR] 0.80, 95% confidence interval [CI] 0.73–0.87, P < .001). The number of patients who needed to be treated to prevent one occurrence of the primary composite outcome (100/absolute risk reduction) was only 21. The benefit was strong and consistent across both of the individual components of the composite outcome:

• Cardiovascular death 13.3% vs 16.5%, HR 0.80 (95% CI 0.71–0.89), P < .001
• First hospitalization for worsening heart failure 12.8% vs 15.6%, HR 0.79 (95% CI 0.71–0.89), P < .001.

Secondary outcomes. The sacubitril-valsartan group had a significantly lower rate of death from any cause (17.0% vs 19.8%, HR 0.84, 95% CI 0.76–0.93, P < .001) and a lower mean decrease in KCCQ clinical summary scores at 8 months (2.99 points vs 4.63 points, mean difference 1.64, 95% CI 0.63–2.65, P = .001). The KCCQ score measures subjective symptoms and physical limitations caused by heart failure; possible scores range from 0 to 100, with a higher score indicating better functional status. Notably, sacubitril-valsartan did not increase the KCCQ score in these patients; rather, sacubitril-valsartan recipients had a lower decrease in their scores than those in the enalapril group.

The incidence of new-onset atrial fibrillation was the same in both groups (3.1% and 3.1%).¹ A decline in renal function (defined as end-stage renal disease, a decrease of 50% or more in estimated glomerular filtration rate from the value at randomization or a decrease in the estimated glomerular filtration rate of more than 30 mL/min/1.73 m² to less than 60 mL/min/1.73 m²) occurred in 2.2% of the valsartan-sacubitril group and 2.6% of the enalapril group (P = .28).

The effects of the study drug on the primary composite outcome and on death from a cardiovascular cause were similar in all prespecified subgroups except for NYHA class: the reduction in the risk of the composite outcome was lower in sacubitril-valsartan recipients with NYHA I or II symptoms than in those with NYHA III or IV symptoms (P for interaction .03). However, there were no differences in the other prespecified subgroups, defined by age, sex, race, region, estimated glomerular filtration rate, diabetes, systolic blood pressure, ejection fraction, atrial fibrillation, NT-proBNP, hypertension, previous use of an ACE inhibitor, previous use of an aldosterone antagonist, previous heart failure hospitalization, and time since diagnosis of heart failure.

SAFETY: ANGIOEDEMA, HYPOTENSION, AND RENAL DYSFUNCTION

Angioedema

As noted above, the combination of neprilysin inhibitors and ACE inhibitors has been associated with an increased risk of angioedema. That was an important consideration before starting this study, which used a combination of a neprilysin inhibitor and an ARB in an attempt to avoid this serious side effect.

As it happened, there was no increased risk of significant angioedema with sacubitril-valsartan use compared with enalapril. Rates were similar to those in other studies, which showed a less than 1% risk of angioedema caused by ACE inhibitors.^20,21 Only 19 patients (0.5%) in the sacubitril-valsartan group and 10 patients (0.2%) in the enalapril group experienced any angioedema. Of these, just three patients in the sacubitril-valsartan group and one patient in the enalapril group experienced angioedema that required hospitalization (P = .31). None of these patients had airway compromise due to angioedema.

Hypotension, cough, renal dysfunction, hyperkalemia

Other safety issues that were assessed included hypotension, worsening renal function, increase in potassium levels, and cough. Patients in the sacubitril-valsartan group were more likely to have symptomatic hypotension than patients in the enalapril group (14.0% vs 9.2%, P < .001); however, the authors noted that this was a rare cause of drug discontinuation.

Patients in the sacubitril-valsartan group were less likely to develop cough (11.3% vs 14.3%, P < .001), a serum creatinine level of 2.5 mg/dL or more (3.3% vs 4.5%, P = .007), or a serum potassium level of more than 6.0 mmol/L (11.3% vs 14.3%, P = .007).¹

During the two run-in periods combined, 12% of the patients in the study withdrew because of adverse events, including cough, renal dysfunction, hyperkalemia, and symptomatic hypotension. During the enalapril run-in period, 591 patients (5.6%) withdrew due to adverse events, and 547 patients (5.8%) withdrew due to these events during the sacubitril-valsartan run-in period. After adjusting for the shorter time on enalapril, there was a higher rate of withdrawal because of adverse events from enalapril than from sacubitril-valsartan.

LOWER RISK OF CLINICAL PROGRESSION

In a separate paper,²² the PARADIGM-HF investigators reported that, among the survivors in the study, those who received sacubitril-valsartan fared better in terms of a number of markers of progression of heart failure, with lower rates of:

• Intensification of medical treatment for heart failure
• Emergency department visits for worsening heart failure
• Hospitalization for worsening heart failure
• Need for intensive care
• Need for intravenous inotropic agents
• Need for cardiac devices or heart transplants
• Worsening symptom scores
• Elevation of biomarkers of myocardial injury.

QUESTIONS AND CONCERNS

Although this study, which was funded by the manufacturer, showed consistent benefit for sacubitril-valsartan over enalapril, questions remain.

Are the findings generalizable?

Despite the study’s rigorous run-in period, 12% of patients dropped out because of adverse events, and thus the patients who completed the study may not be representative of the general population of heart failure patients. The authors included this double-level wash-out to ensure patient tolerance of both drugs. But in everyday practice, a significant number of patients may be unable to tolerate one of these drugs.

Moreover, after adjusting for the difference in the run-in periods, patients actually withdrew more often during the enalapril run-in period than during the sacubitril-valsartan run-in period. However, there may be overlap in tolerability in these two drugs, which both affect the renin-angiotensin-aldosterone system. Thus, the enalapril run-in period may have contributed to the lower tolerability of this drug compared with sacubitril-valsartan.

Were patients receiving the best possible therapy?

Another important point when considering how we treat heart failure patients in the United States is how few patients in this study were using cardiac implantable electronic devices. Only 15% of the patients in this study had an implantable cardioverter-defibrillator despite a mean left ventricular ejection fraction less than 30%. This likely reflects differences in practice internationally; however, based on American College of Cardiology, American Heart Association, and Heart Rhythm Society guidelines, these patients would have a class I indication for an implantable cardioverter-defibrillator for primary prevention of sudden cardiac death.²³

Therefore, based on these recommendations, the patients in this study were not necessarily on optimal medical and device therapy and furthermore may not be representative of heart failure patients in the United States.

Was enalapril 10 mg twice a day a fair comparison?

Another concern about the results of this study relates to the dosages used in the two treatment groups. The sacubitril-valsartan formulation included full-dose valsartan, whereas the ACE inhibitor group received enalapril at less than a full dose.

Although the authors explained that the dose of enalapril chosen for the study was based on the one used in previous studies that showed survival benefit, this raises the question of whether the significant difference in outcomes in this trial was due to a greater inhibition of the renin-angiotensin-aldosterone system related to a higher dose of drug in the sacubitril-valsartan group.

What about black patients taking hydralazine-isosorbide?

Only about 5% of patients in the PARADIGM-HF trial were black. Based on the A-HeFT study results, black patients can be prescribed an ACE inhibitor as well as hydralazine and isosorbide dinitrate as tolerated to decrease the risk of death. Does sacubitril-valsartan offer benefit to these patients compared with a regimen of an ACE inhibitor, hydralazine, and isosorbide dinitrate?

Another concern is that the incidence of angioedema observed with ACE inhibitors and omapatrilat is higher in black patients.^15,21 Thus, it would be prudent to investigate whether the risk of angioedema with sacubitril-valsartan would be higher if more black patients are studied.

IMPLICATIONS AND CONSIDERATIONS

In this study, sacubitril-valsartan showed impressive and consistent results, with an almost 20% decrease in the composite end point of heart failure hospitalization or cardiovascular death and a similar decrease in the composite outcomes with a very low number needed to treat (21 patients). It did not show a decrease in the incidence of new-onset atrial fibrillation; however, only 80 cases of atrial fibrillation were reported, so there may have been a lack of statistical power to detect a difference.

To avoid angioedema, wait at least 36 hours after stopping an ACE inhibitor. Sacubitril-valsartan was not associated with an increased risk of severe angioedema, and no patients experienced life-threatening angioedema. In the trial, the sacubitril-valsartan run-in period was started at least 24 hours after enalapril was stopped, and thus the authors recommended at least a 1-day washout period after discontinuing an ACE inhibitor to avoid angioedema in patients starting sacubitril-valsartan.

Hypotension is a concern. Although there was actually a decreased risk of renal dysfunction, hyperkalemia, and cough compared with enalapril, there was a significantly increased rate of symptomatic hypotension in the sacubitril-valsartan group, which raises the question of patient tolerance and physician comfort when prescribing and titrating this drug in clinical practice. This side effect will be an important consideration when attempting to titrate the drug to target doses.

Start treatment early. This trial largely consisted of patients with NYHA class II or III symptoms, with about 70% of patients with class II symptoms. Since this drug showed benefit in patients with mildly to moderately symptomatic heart failure, clinicians who are considering prescribing this drug should not wait until the patient is closer to end-stage disease. Patients with mildly symptomatic heart failure may be followed by a general cardiologist, internist, or both, and thus it is important to emphasize to the entire medical community the need to start this medication early on.

How much will it cost? Cost is a concern that could heavily weigh on the decision to prescribe this drug. Generic ACE inhibitors are relatively inexpensive, and it may difficult to switch from an affordable generic drug to a new drug that is likely to be much more expensive. Arguably, this drug may be cost-effective in the long run owing to a large decrease in heart failure readmissions. We await further analyses to evaluate this issue.

Will patients take a twice-a-day drug as prescribed? Most patients who are prescribed an ACE inhibitor take it just once a day, and switching from a daily to a twice-daily drug may present a challenge for some.

What about other outcomes? Based on this study, it is unclear what effect sacubitril-valsartan has on the incidence of fatal arrhythmias, sudden cardiac death, and pump failure. Furthermore, the effect on quality of life is still uncertain. Quality of life is an integral component in the evaluation of heart failure patients, and in this study the changes in KCCQ scores were not impressive. We hope to see further evaluations of this drug’s impact on quality of life of patients with heart failure. Furthermore, it would be interesting to study if this drug has any long-term effects on the need for advanced therapies such as left ventricular assist devices and orthotopic heart transplant.

What about patients with heart failure with preserved ejection fraction? This study included only patients with heart failure with reduced ejection fraction. However PARAMOUNT, a phase 2 study that evaluated the benefit of sacubitril-valsartan in patients with heart failure with preserved ejection fraction, has shown encouraging results.²⁴ We look forward to further investigation of this agent in patients with heart failure with preserved ejection fraction.

Sacubitril-valsartan, the first ARNI to be studied in humans, has a dual action in that it enhances the activity of the natriuretic peptide system and inhibits that of the renin-angiotensin-aldosterone system. It is the first drug in over a decade to show mortality benefit in patients with chronic systolic heart failure when compared with an already well-established heart failure medication. It appears to decrease rates of mortality and heart failure hospitalization without increasing the risk of severe angioedema in patients with mild or moderate chronic systolic heart failure. Symptomatic hypotension and high cost may pose the largest barriers to the use of this new drug. And we have yet to see how the clinical community and patients with heart failure will respond to it.

Many patients who have osteoarthritis of the knee and a torn meniscus can defer having the meniscus repaired and undergo physical therapy instead. If a trial of physical therapy does not help, they can opt for surgery later.

This seems to be the take-home message from the recent Meniscal Tear in Osteoarthritis Research (METEOR) trial,¹ which compared the efficacy of arthroscopic partial meniscectomy plus physical therapy vs physical therapy alone for patients with knee symptoms, a meniscal tear, and mild to moderate osteoarthritis of the knee.¹

In brief, patients improved to a roughly similar degree with either approach, and although many patients assigned to physical therapy eventually underwent surgery anyway by 6 months, the delay did not adversely affect outcomes.

In this article, we review the background, design, and findings of the METEOR trial, and their implications for clinical practice.

SURGERY: HIGH VOLUME, BUT LITTLE EVIDENCE

Magnetic resonance imaging (MRI) often incidentally reveals meniscal lesions in middle-aged and older patients who have osteoarthritis and knee pain.² Should these patients undergo arthroscopic meniscal repair? The decision is difficult, since it is hard to distinguish the symptoms of a meniscal tear from those of osteoarthritis.³

Current evidence suggests that, for symptomatic knee osteoarthritis by itself, arthroscopic surgery is no more effective than conservative management.^4,5 But what about surgery for a torn meniscus in addition to osteoarthritis?

Osteoarthritis is the most common joint disease, accounting for many physician visits.⁶ More than 26 million Americans over age 25 have some form of it, and the prevalence of symptomatic, radiographically confirmed osteoarthritis of the knee was 12.1% in the third National Health and Nutrition Examination Survey.⁷

We used to consider osteoarthritis a “wear-and-tear” disease—thus the term “degenerative joint disease.” But today, we know that it is an active response to injury, involving inflammatory and metabolic pathways.⁸ Moreover, the risk of osteoarthritis and its progression seems to be higher in those who have had meniscal injury and total or arthroscopic partial meniscectomy.^9,10

MRI is not commonly used in managing knee osteoarthritis, but it has been used diagnostically in patients with symptoms of a meniscal tear, such as clicking, locking, popping, giving way, and pain with pivoting or twisting. Traumatic meniscal tears (a longitudinal or radial tear pattern) most often occur in active younger people and often lead to meniscal surgery.^11,12 In contrast, degenerative meniscal tears (horizontal, oblique, or complex tear pattern or meniscal maceration) tend to occur in older people,^11,12 but how to manage them is not widely agreed upon.

Of note, most patients with osteoarthritis of the knee have torn, macerated, or heavily damaged menisci.^13,14 Meniscal lesions are also common in middle-aged people in the general population, with a higher prevalence in people who are older, heavier, or female, or who have a family history of osteoarthritis.¹⁵

These abnormalities are only weakly associated with symptoms.² However, when a patient has knee symptoms and a torn meniscus is detected on MRI, the tear is often assumed to be the source of the symptoms, and meniscal tears are the most common reason for arthroscopy.¹⁶

Since we have no way to prevent the progression of joint damage from osteoarthritis with drugs or by any other means, the goal is to alleviate the symptoms. Many patients report pain relief or functional improvement after arthroscopic surgery. But arthroscopic lavage or debridement for osteoarthritis has not been found to be better than conservative treatment or placebo in randomized controlled trials.^4,5

In contrast, the current standard treatment for a symptomatic degenerative meniscal tear is arthroscopic partial meniscectomy. Nearly 500,000 of these procedures are performed annually in the United States.¹⁶ But based on the best evidence, arthroscopic partial meniscectomy does not result in better pain relief and functional improvement than does physical therapy alone in patients who have a torn meniscus and knee osteoarthritis.^17,18

OVERVIEW OF THE METEOR TRIAL

The METEOR trial was a randomized controlled trial conducted at seven US tertiary referral centers. Its aim was to compare the short-term (6-month) and long-term (12-month) efficacy of arthroscopic partial meniscectomy and physical therapy in patients with symptomatic meniscal tear and osteoarthritis of the knee.¹⁹ It was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases.¹

Patients were age 45 and older

METEOR patients had to be at least 45 years old and have symptomatic meniscal tears and knee osteoarthritis detected on MRI or radiography.¹

Osteoarthritis was defined broadly, given that it begins well before the appearance of radiographic evidence such as an osteophyte or joint-space narrowing.¹⁹ Patients with cartilage defects on MRI were also enrolled, as were patients with radiographically documented osteoarthritis.¹⁹

Patients were considered to have a symptomatic meniscal tear if they had had at least 4 weeks of symptoms (such as episodic pain and pain that was acute and localized to one spot on the knee, as well as typical mechanical pain suggesting a meniscal tear, such as clicking, catching, popping, giving way, or pain with pivoting or twisting) in addition to evidence of a meniscal tear on MRI.¹⁹

Patients were excluded if they had a chronically locked knee (a clear-cut indication for arthroscopic partial meniscectomy), advanced osteoarthritis (Kellgren-Lawrence grade 4), inflammatory arthritis, clinically symptomatic chondrocalcinosis, or bilateral symptomatic meniscal tears.¹⁹ Patients who had undergone surgery or injection of a viscosupplement in the index knee during the past 4 weeks were also excluded.¹⁹

Of 1,330 eligible patients, 351 (26.4%) were enrolled and randomly assigned in a 1:1 ratio to a treatment group by means of a secure program on the trial website.^1,19 Of those who were eligible but did not enroll, 195 (14.6%) were not referred and 784 (58.9%) declined to participate. Of those who declined, more preferred surgery than physical therapy (36.1% vs 21%). No information is available on any differences in baseline characteristics between the enrolled patients and the eligible patients who declined.

Randomization was done in blocks of varying size within each site, stratified according to sex and the extent of osteoarthritis on baseline radiography. The extent of osteoarthritis was categorized either as Kellgren-Lawrence grade 0 (normal, no features of osteoarthritis) to grade 2 (definite osteoarthritis, a definite osteophyte without joint-space narrowing) or as Kellgren-Lawrence grade 3 (moderate osteoarthritis, < 50% joint-space narrowing).^1,19 The two treatment groups were similar with respect to age, sex, race or ethnicity, baseline Kellgren-Lawrence grade, and baseline Western Ontario and McMaster Universities Arthritis Index (WOMAC) physical function score.¹

The mean age of the participants was 58, and 85% were white. Sixty-three percent had Kellgren-Lawrence grade 0 to 2 osteoarthritis, and 27% had grade 3.¹

Surgery plus physical therapy vs physical therapy alone

The surgery group underwent arthroscopic partial meniscectomy, which involved trimming the damaged meniscus back to a stable rim^1,19 and trimming loose fragments of cartilage and bone.

After the procedure, patients were scheduled for physical therapy. Although there is no consensus on the need for or the effectiveness of postoperative physical therapy in this setting, the investigators believed that including it in both study groups would help to isolate the independent effects of surgery. The physical therapy regimen after surgery was similar to that provided in the nonoperative group.^1,19

Physical therapy was designed to address inflammation, range of motion, muscle strength, muscle-length restriction, functional mobility, and proprioception and balance.^1,19 There were three stages; criteria for advancing from one phase to the next included the level of self-reported pain, observed strength, range of knee motion, knee effusion, and functional mobility.^1,18

The duration of participation varied depending on the pace of improvement. Generally, the program lasted about 6 weeks.^1,19

Crossover and other therapies were allowed

Crossover from physical therapy alone to surgery was allowed during the trial if the patient and surgeon thought it was clinically indicated.

Participants in both groups were permitted to take acetaminophen and nonsteroidal anti-inflammatory drugs as needed. Intra-articular injections of glucocorticoids were also allowed during the trial.

OUTCOMES MEASURED

WOMAC physical function score

The primary outcome of the METEOR trial was the difference between the study groups in the change in WOMAC physical function score from baseline to 6 months, at which point participants were expected to have achieved maximum improvement.^1,19 Questionnaires were also administered at 3 months to assess the early response to surgery or physical therapy and again at 12 months.

The complete WOMAC also measures pain and stiffness in addition to physical function, with separate subscales for each. The change in WOMAC score is one of the most widely endorsed outcome measures in assessing interventions in osteoarthritis or other conditions of the lower extremities.²⁰ The METEOR trial authors considered the WOMAC scale to be highly valid and reliable, with a Cronbach alpha value of 0.97 (maximum value = 1; the higher the better).

No ceiling or floor effects were observed in the WOMAC physical function score in patients with osteoarthritis and a meniscal tear in a pilot study for METEOR.¹⁹

In the main METEOR study, WOMAC physical function was scored on a scale of 0 to 100, with a higher score indicating worse physical function.¹ Changes in the score were also measured as a yes-or-no question, defined a priori as whether the score declined by at least 8 points, which is considered the minimal clinically important difference in osteoarthritis patients.^1,19

KOOS and MOS SF-36 scores

Secondary outcomes were measured in several domains, including pain, generic functional status, quality of life, and health care utilization.^1,19

The KOOS (Knee Injury and OA Outcome Scale) is specific for knee pain, being designed to evaluate short-term and long-term symptoms and function in patients with knee injury and associated problems.²¹ It has five subscales, which are scored separately: pain, other symptoms, activities of daily living, sport and recreation, and knee-related quality of life.²¹ Since the WOMAC pain scale showed a ceiling effect in the pilot study in patients undergoing surgery, the authors chose the KOOS pain scale as a pain measure.¹⁹ Scores were transformed to a 0–100 scale, with a higher score indicating more pain.¹

The MOS SF-36 (Medical Outcomes Study 36-item short form) was used to measure general health status and function.^1,19

STATISTICAL ANALYSIS: INTENTION-TO-TREAT AND AS-TREATED

The study was powered to detect a 10-point difference in WOMAC physical function scores at 6 months of follow-up between the operative and nonoperative groups, anticipating losses to follow-up and crossover, with preplanned subgroup (Kellgren–Lawrence grade 0–2 vs grade 3) analysis.^1,19

The primary analysis used a modified intention-to-treat approach and was implemented with an analysis of covariance with changes in the WOMAC score from baseline to 6 months as the dependent variable, treatment as the independent variable of interest, and study site as a covariate. Other covariates, such as age, sex, and baseline Kellgren-Lawrence grade, were balanced across groups and were therefore not included in the analysis.^1,19

Secondary analyses used an “as-treated” approach, ie, according to the treatment actually received.^1,19 Secondary intention-to-treat analysis—using binary outcome measures in which treatment failure was defined as improvement in the WOMAC score of less than 8 points or crossing over to the other treatment—was also performed to estimate efficacy at the level of the patient rather than at the group level.^1,19

BOTH GROUPS IMPROVED

In the intention-to-treat analyses at 6 months and 12 months after randomization, both groups improved, with no clinically important or statistically significant differences between the groups in functional status (WOMAC score, MOS SF-36 score) or pain (KOOS score).¹ The mean improvement (decline) in the WOMAC score from baseline to 6 months was 20.9 points in the surgery group vs 18.5 points in the physical therapy group, a difference of 2.4 points (95% confidence interval [CI], −1.8 to 6.5).¹

35% of physical therapy patients underwent surgery by 12 months

Of the 177 patients randomized to physical therapy alone, by 6 months 1 had died, 1 had undergone total knee replacement, 4 had withdrawn, and 2 were lost to follow-up. Of the 169 remaining, 51 (30%) had undergone arthroscopic partial meniscectomy. An additional 8 patients who were assigned to physical therapy crossed over to surgery between 6 and 12 months.^1,19

Of the 174 patients randomized to surgery, by 6 months 1 had died, 3 had undergone total knee replacement, 7 had withdrawn, and 2 were ineligible. Of the 161 remaining, 9 (6%) had not undergone the procedure.

Other outcomes

Subgroup analysis based on the baseline radiographic grade (Kellgren-Lawrence grade 0 to 2 vs grade 3) did not show a difference between groups in functional improvement at 6 months (P = .13 for interaction).¹

No statistically significant difference was noted in rates of overall or specific adverse events between the two groups over the first 12 months.¹ Adverse events rated as mild or moderate in severity occurred in 15 participants in the surgery group and 13 participants in the physical therapy group.¹ Long-term risks associated with these interventions are being assessed, and longitudinal assessment of imaging studies is planned to address this question but is not yet available.^1,18

In the physical therapy group, 21 patients (12%) received intra-articular glucocorticoid injections, as did 9 patients (6%) in the surgery group.^1,19

TRANSLATING THE METEOR RESULTS TO EVERYDAY PRACTICE

There are many challenges in designing surgical trials. Indeed, by one estimate,²² only about 40% of treatment questions involving surgical procedures can be evaluated by a randomized controlled trial.

Although the METEOR trial was not blinded, it was the first large, multicenter, randomized controlled trial to compare arthroscopic partial meniscectomy vs standardized physical therapy by using high-quality methodology such as careful sample-size calculation, balancing the groups according to known prognostic factors with block randomization, and intention-to-treat analysis. Moreover, the outcome measures were obtained from validated self-reporting questionnaires (WOMAC for function and KOOS for pain), reducing the possibility of observer bias.¹⁹ In addition, analyses were performed with the analysts blinded to the randomization assignment.

Limitations of the trial

A few limitations of the study are worth noting.

Patients age 45 or older with both symptomatic meniscal tear and osteoarthritis were the target population of this study. However, it is important to distinguish between the study population and the target population in a physician’s practice.

The investigators adopted broad definitions of osteoarthritis and symptoms of meniscal tear. Twenty-one percent of participants had normal findings on plain radiography, with cartilage defects visible only on MRI. Further, episodic pain or acute pain localized to a joint line was regarded as a symptom consistent with a torn meniscus.

In practice, arthroscopic partial meniscectomy is usually considered when a patient with a long history of tolerable osteoarthritis presents with a sudden onset of intolerable pain after a squatting or twisting injury.

In addition, the study population was predominantly white (85%), and the study was performed in tertiary referral academic medical centers. Therefore, the outcomes achieved with surgery or physical therapy may not translate to the community setting. Clinicians must be careful to account for these types of differences in extrapolating to patients in their own practice.

Potential enrollment bias

Although randomization is a rigorous method that eliminates selection bias in assigning individuals to study and control groups, selective enrollment could have created bias.¹ As the authors mentioned, only 26% of eligible patients were enrolled, possibly reflecting patients’ or surgeons’ strong preferences for one treatment or the other. Because the study and control groups were hardly random samples of eligible populations, we must be careful in generalizing the efficacy of physical therapy.¹

Crossover may have obscured the benefit of surgery

During the first 6 months, 30% of patients crossed over from physical therapy to surgery. High crossover rates in surgical trials are common, especially when comparing surgery with medical therapy.²³ Given that most of the patients assigned to only physical therapy who crossed over to surgery did not have substantial improvement in functional status, it seems that crossover occurred by nonrandom factors, potentially biasing the study results. With the high degree of crossover from the nonoperative group to the surgical group, intention-to-treat analysis may have given an inflated estimate of the effect of physical therapy.

To account for crossovers, researchers defined a binary outcome a priori: patients were considered to have had a successful treatment response if they improved by at least 8 points on the WOMAC scale (a clinically important difference) and did not cross over from their assigned treatment. At 6 months, 67.1% of patients assigned to surgery showed a successful treatment response, compared with 43.8% of patients assigned to physical therapy alone (P = .001).¹

In patients who crossed over, the last scores before crossover were carried over, and primary analysis of the WOMAC score at 6 months was repeated to estimate the effect of crossovers from the nonoperative to the surgery group. This exploratory analysis showed a 13.0-point improvement in WOMAC score at 6 months with physical therapy alone vs a 20.9-point improvement with surgery, suggesting that the similarity in outcomes between the two groups may be explained in part by additional improvements from surgery for those who crossed over from physical therapy alone.¹

Implications for functional improvement

Lacking a comparison group that underwent a sham surgical procedure, one cannot conclude that surgery after crossover improved functional status in those patients. However, there was no significant difference in WOMAC physical function scores at 12 months between the 30% of patients in the physical therapy group who crossed over and underwent surgery during the first 6 months and patients initially assigned to surgery. This finding suggests that physical therapy can be recommended as a first-line therapy, although we must be cautious, given that the physical therapy group required more background therapy (eg, intra-articular glucocorticoid injections), and that this study was not powered to detect such differences at 12 months.

Also, a patient may need to get better quickly, to get back to work, for example. Although the data were not definitive, at 3 months the patients in the surgery group seemed to have better pain control and function than those in the physical therapy group. A cost-benefit analysis of physical therapy compared with surgery for short-term outcomes may be helpful before generalizing these findings.

SURGERY VS SHAM PROCEDURE: THE FIDELITY GROUP RESULTS

In a later publication from the Finnish Degenerative Meniscal Lesion Study (FIDELITY) Group,²⁴ 146 patients with symptoms consistent with degenerative meniscal tear but no knee osteoarthritis were randomized to undergo arthroscopic partial meniscectomy or a sham procedure. At 12 months, no differences were noted between the groups in terms of change of symptoms from baseline to 12 months.

The authors concluded that the outcomes with meniscectomy were no better than with a sham procedure.²⁴

SURGERY FIRST, OR PHYSICAL THERAPY FIRST?

The use of knee arthroscopy has increased sharply in middle-aged patients in recent years. Indeed, this demographic group accounts for nearly half of the knee arthroscopic procedures performed for meniscal tears, although the increase may be due in part to issues with surgeons’ coding and insurance authorization.¹⁶

The METEOR trial showed that a structured physical therapy program can be as effective as surgery as a first-line therapy in many patients with symptomatic meniscal tears and mild to moderate osteoarthritis. These results should inform clinical practice in that most such patients need not be immediately referred for surgical intervention.

However, a subset of these patients may benefit from surgery rather than nonoperative therapy. Given the potential risks and public health implications of arthroscopic surgery for meniscal tears, further study is needed to better characterize these patients. A randomized sham-controlled trial is under way²⁵ with the goal of assessing the efficacy of arthroscopic partial meniscectomy for medial meniscus tears in patients with or without knee osteoarthritis, and it is hoped this study will shed further light on this issue.

Based on the results of the METEOR trial, the physical therapy regimen that was used may be reasonable before referring patients with knee osteoarthritis and symptomatic meniscal tears for surgery.

Many patients who have osteoarthritis of the knee and a torn meniscus can defer having the meniscus repaired and undergo physical therapy instead. If a trial of physical therapy does not help, they can opt for surgery later.

This seems to be the take-home message from the recent Meniscal Tear in Osteoarthritis Research (METEOR) trial,¹ which compared the efficacy of arthroscopic partial meniscectomy plus physical therapy vs physical therapy alone for patients with knee symptoms, a meniscal tear, and mild to moderate osteoarthritis of the knee.¹

In brief, patients improved to a roughly similar degree with either approach, and although many patients assigned to physical therapy eventually underwent surgery anyway by 6 months, the delay did not adversely affect outcomes.

In this article, we review the background, design, and findings of the METEOR trial, and their implications for clinical practice.

SURGERY: HIGH VOLUME, BUT LITTLE EVIDENCE

Magnetic resonance imaging (MRI) often incidentally reveals meniscal lesions in middle-aged and older patients who have osteoarthritis and knee pain.² Should these patients undergo arthroscopic meniscal repair? The decision is difficult, since it is hard to distinguish the symptoms of a meniscal tear from those of osteoarthritis.³

Current evidence suggests that, for symptomatic knee osteoarthritis by itself, arthroscopic surgery is no more effective than conservative management.^4,5 But what about surgery for a torn meniscus in addition to osteoarthritis?

Osteoarthritis is the most common joint disease, accounting for many physician visits.⁶ More than 26 million Americans over age 25 have some form of it, and the prevalence of symptomatic, radiographically confirmed osteoarthritis of the knee was 12.1% in the third National Health and Nutrition Examination Survey.⁷

We used to consider osteoarthritis a “wear-and-tear” disease—thus the term “degenerative joint disease.” But today, we know that it is an active response to injury, involving inflammatory and metabolic pathways.⁸ Moreover, the risk of osteoarthritis and its progression seems to be higher in those who have had meniscal injury and total or arthroscopic partial meniscectomy.^9,10

MRI is not commonly used in managing knee osteoarthritis, but it has been used diagnostically in patients with symptoms of a meniscal tear, such as clicking, locking, popping, giving way, and pain with pivoting or twisting. Traumatic meniscal tears (a longitudinal or radial tear pattern) most often occur in active younger people and often lead to meniscal surgery.^11,12 In contrast, degenerative meniscal tears (horizontal, oblique, or complex tear pattern or meniscal maceration) tend to occur in older people,^11,12 but how to manage them is not widely agreed upon.

Of note, most patients with osteoarthritis of the knee have torn, macerated, or heavily damaged menisci.^13,14 Meniscal lesions are also common in middle-aged people in the general population, with a higher prevalence in people who are older, heavier, or female, or who have a family history of osteoarthritis.¹⁵

These abnormalities are only weakly associated with symptoms.² However, when a patient has knee symptoms and a torn meniscus is detected on MRI, the tear is often assumed to be the source of the symptoms, and meniscal tears are the most common reason for arthroscopy.¹⁶

Since we have no way to prevent the progression of joint damage from osteoarthritis with drugs or by any other means, the goal is to alleviate the symptoms. Many patients report pain relief or functional improvement after arthroscopic surgery. But arthroscopic lavage or debridement for osteoarthritis has not been found to be better than conservative treatment or placebo in randomized controlled trials.^4,5

In contrast, the current standard treatment for a symptomatic degenerative meniscal tear is arthroscopic partial meniscectomy. Nearly 500,000 of these procedures are performed annually in the United States.¹⁶ But based on the best evidence, arthroscopic partial meniscectomy does not result in better pain relief and functional improvement than does physical therapy alone in patients who have a torn meniscus and knee osteoarthritis.^17,18

OVERVIEW OF THE METEOR TRIAL

The METEOR trial was a randomized controlled trial conducted at seven US tertiary referral centers. Its aim was to compare the short-term (6-month) and long-term (12-month) efficacy of arthroscopic partial meniscectomy and physical therapy in patients with symptomatic meniscal tear and osteoarthritis of the knee.¹⁹ It was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases.¹

Patients were age 45 and older

METEOR patients had to be at least 45 years old and have symptomatic meniscal tears and knee osteoarthritis detected on MRI or radiography.¹

Osteoarthritis was defined broadly, given that it begins well before the appearance of radiographic evidence such as an osteophyte or joint-space narrowing.¹⁹ Patients with cartilage defects on MRI were also enrolled, as were patients with radiographically documented osteoarthritis.¹⁹

Patients were considered to have a symptomatic meniscal tear if they had had at least 4 weeks of symptoms (such as episodic pain and pain that was acute and localized to one spot on the knee, as well as typical mechanical pain suggesting a meniscal tear, such as clicking, catching, popping, giving way, or pain with pivoting or twisting) in addition to evidence of a meniscal tear on MRI.¹⁹

Patients were excluded if they had a chronically locked knee (a clear-cut indication for arthroscopic partial meniscectomy), advanced osteoarthritis (Kellgren-Lawrence grade 4), inflammatory arthritis, clinically symptomatic chondrocalcinosis, or bilateral symptomatic meniscal tears.¹⁹ Patients who had undergone surgery or injection of a viscosupplement in the index knee during the past 4 weeks were also excluded.¹⁹

Of 1,330 eligible patients, 351 (26.4%) were enrolled and randomly assigned in a 1:1 ratio to a treatment group by means of a secure program on the trial website.^1,19 Of those who were eligible but did not enroll, 195 (14.6%) were not referred and 784 (58.9%) declined to participate. Of those who declined, more preferred surgery than physical therapy (36.1% vs 21%). No information is available on any differences in baseline characteristics between the enrolled patients and the eligible patients who declined.

Randomization was done in blocks of varying size within each site, stratified according to sex and the extent of osteoarthritis on baseline radiography. The extent of osteoarthritis was categorized either as Kellgren-Lawrence grade 0 (normal, no features of osteoarthritis) to grade 2 (definite osteoarthritis, a definite osteophyte without joint-space narrowing) or as Kellgren-Lawrence grade 3 (moderate osteoarthritis, < 50% joint-space narrowing).^1,19 The two treatment groups were similar with respect to age, sex, race or ethnicity, baseline Kellgren-Lawrence grade, and baseline Western Ontario and McMaster Universities Arthritis Index (WOMAC) physical function score.¹

The mean age of the participants was 58, and 85% were white. Sixty-three percent had Kellgren-Lawrence grade 0 to 2 osteoarthritis, and 27% had grade 3.¹

Surgery plus physical therapy vs physical therapy alone

The surgery group underwent arthroscopic partial meniscectomy, which involved trimming the damaged meniscus back to a stable rim^1,19 and trimming loose fragments of cartilage and bone.

After the procedure, patients were scheduled for physical therapy. Although there is no consensus on the need for or the effectiveness of postoperative physical therapy in this setting, the investigators believed that including it in both study groups would help to isolate the independent effects of surgery. The physical therapy regimen after surgery was similar to that provided in the nonoperative group.^1,19

Physical therapy was designed to address inflammation, range of motion, muscle strength, muscle-length restriction, functional mobility, and proprioception and balance.^1,19 There were three stages; criteria for advancing from one phase to the next included the level of self-reported pain, observed strength, range of knee motion, knee effusion, and functional mobility.^1,18

The duration of participation varied depending on the pace of improvement. Generally, the program lasted about 6 weeks.^1,19

Crossover and other therapies were allowed

Crossover from physical therapy alone to surgery was allowed during the trial if the patient and surgeon thought it was clinically indicated.

Participants in both groups were permitted to take acetaminophen and nonsteroidal anti-inflammatory drugs as needed. Intra-articular injections of glucocorticoids were also allowed during the trial.

OUTCOMES MEASURED

WOMAC physical function score

The primary outcome of the METEOR trial was the difference between the study groups in the change in WOMAC physical function score from baseline to 6 months, at which point participants were expected to have achieved maximum improvement.^1,19 Questionnaires were also administered at 3 months to assess the early response to surgery or physical therapy and again at 12 months.

The complete WOMAC also measures pain and stiffness in addition to physical function, with separate subscales for each. The change in WOMAC score is one of the most widely endorsed outcome measures in assessing interventions in osteoarthritis or other conditions of the lower extremities.²⁰ The METEOR trial authors considered the WOMAC scale to be highly valid and reliable, with a Cronbach alpha value of 0.97 (maximum value = 1; the higher the better).

No ceiling or floor effects were observed in the WOMAC physical function score in patients with osteoarthritis and a meniscal tear in a pilot study for METEOR.¹⁹

In the main METEOR study, WOMAC physical function was scored on a scale of 0 to 100, with a higher score indicating worse physical function.¹ Changes in the score were also measured as a yes-or-no question, defined a priori as whether the score declined by at least 8 points, which is considered the minimal clinically important difference in osteoarthritis patients.^1,19

KOOS and MOS SF-36 scores

Secondary outcomes were measured in several domains, including pain, generic functional status, quality of life, and health care utilization.^1,19

The KOOS (Knee Injury and OA Outcome Scale) is specific for knee pain, being designed to evaluate short-term and long-term symptoms and function in patients with knee injury and associated problems.²¹ It has five subscales, which are scored separately: pain, other symptoms, activities of daily living, sport and recreation, and knee-related quality of life.²¹ Since the WOMAC pain scale showed a ceiling effect in the pilot study in patients undergoing surgery, the authors chose the KOOS pain scale as a pain measure.¹⁹ Scores were transformed to a 0–100 scale, with a higher score indicating more pain.¹

The MOS SF-36 (Medical Outcomes Study 36-item short form) was used to measure general health status and function.^1,19

STATISTICAL ANALYSIS: INTENTION-TO-TREAT AND AS-TREATED

The study was powered to detect a 10-point difference in WOMAC physical function scores at 6 months of follow-up between the operative and nonoperative groups, anticipating losses to follow-up and crossover, with preplanned subgroup (Kellgren–Lawrence grade 0–2 vs grade 3) analysis.^1,19

The primary analysis used a modified intention-to-treat approach and was implemented with an analysis of covariance with changes in the WOMAC score from baseline to 6 months as the dependent variable, treatment as the independent variable of interest, and study site as a covariate. Other covariates, such as age, sex, and baseline Kellgren-Lawrence grade, were balanced across groups and were therefore not included in the analysis.^1,19

Secondary analyses used an “as-treated” approach, ie, according to the treatment actually received.^1,19 Secondary intention-to-treat analysis—using binary outcome measures in which treatment failure was defined as improvement in the WOMAC score of less than 8 points or crossing over to the other treatment—was also performed to estimate efficacy at the level of the patient rather than at the group level.^1,19

BOTH GROUPS IMPROVED

In the intention-to-treat analyses at 6 months and 12 months after randomization, both groups improved, with no clinically important or statistically significant differences between the groups in functional status (WOMAC score, MOS SF-36 score) or pain (KOOS score).¹ The mean improvement (decline) in the WOMAC score from baseline to 6 months was 20.9 points in the surgery group vs 18.5 points in the physical therapy group, a difference of 2.4 points (95% confidence interval [CI], −1.8 to 6.5).¹

35% of physical therapy patients underwent surgery by 12 months

Of the 177 patients randomized to physical therapy alone, by 6 months 1 had died, 1 had undergone total knee replacement, 4 had withdrawn, and 2 were lost to follow-up. Of the 169 remaining, 51 (30%) had undergone arthroscopic partial meniscectomy. An additional 8 patients who were assigned to physical therapy crossed over to surgery between 6 and 12 months.^1,19

Of the 174 patients randomized to surgery, by 6 months 1 had died, 3 had undergone total knee replacement, 7 had withdrawn, and 2 were ineligible. Of the 161 remaining, 9 (6%) had not undergone the procedure.

Other outcomes

Subgroup analysis based on the baseline radiographic grade (Kellgren-Lawrence grade 0 to 2 vs grade 3) did not show a difference between groups in functional improvement at 6 months (P = .13 for interaction).¹

No statistically significant difference was noted in rates of overall or specific adverse events between the two groups over the first 12 months.¹ Adverse events rated as mild or moderate in severity occurred in 15 participants in the surgery group and 13 participants in the physical therapy group.¹ Long-term risks associated with these interventions are being assessed, and longitudinal assessment of imaging studies is planned to address this question but is not yet available.^1,18

In the physical therapy group, 21 patients (12%) received intra-articular glucocorticoid injections, as did 9 patients (6%) in the surgery group.^1,19

TRANSLATING THE METEOR RESULTS TO EVERYDAY PRACTICE

There are many challenges in designing surgical trials. Indeed, by one estimate,²² only about 40% of treatment questions involving surgical procedures can be evaluated by a randomized controlled trial.

Although the METEOR trial was not blinded, it was the first large, multicenter, randomized controlled trial to compare arthroscopic partial meniscectomy vs standardized physical therapy by using high-quality methodology such as careful sample-size calculation, balancing the groups according to known prognostic factors with block randomization, and intention-to-treat analysis. Moreover, the outcome measures were obtained from validated self-reporting questionnaires (WOMAC for function and KOOS for pain), reducing the possibility of observer bias.¹⁹ In addition, analyses were performed with the analysts blinded to the randomization assignment.

Limitations of the trial

A few limitations of the study are worth noting.

Patients age 45 or older with both symptomatic meniscal tear and osteoarthritis were the target population of this study. However, it is important to distinguish between the study population and the target population in a physician’s practice.

The investigators adopted broad definitions of osteoarthritis and symptoms of meniscal tear. Twenty-one percent of participants had normal findings on plain radiography, with cartilage defects visible only on MRI. Further, episodic pain or acute pain localized to a joint line was regarded as a symptom consistent with a torn meniscus.

In practice, arthroscopic partial meniscectomy is usually considered when a patient with a long history of tolerable osteoarthritis presents with a sudden onset of intolerable pain after a squatting or twisting injury.

In addition, the study population was predominantly white (85%), and the study was performed in tertiary referral academic medical centers. Therefore, the outcomes achieved with surgery or physical therapy may not translate to the community setting. Clinicians must be careful to account for these types of differences in extrapolating to patients in their own practice.

Potential enrollment bias

Although randomization is a rigorous method that eliminates selection bias in assigning individuals to study and control groups, selective enrollment could have created bias.¹ As the authors mentioned, only 26% of eligible patients were enrolled, possibly reflecting patients’ or surgeons’ strong preferences for one treatment or the other. Because the study and control groups were hardly random samples of eligible populations, we must be careful in generalizing the efficacy of physical therapy.¹

Crossover may have obscured the benefit of surgery

During the first 6 months, 30% of patients crossed over from physical therapy to surgery. High crossover rates in surgical trials are common, especially when comparing surgery with medical therapy.²³ Given that most of the patients assigned to only physical therapy who crossed over to surgery did not have substantial improvement in functional status, it seems that crossover occurred by nonrandom factors, potentially biasing the study results. With the high degree of crossover from the nonoperative group to the surgical group, intention-to-treat analysis may have given an inflated estimate of the effect of physical therapy.

To account for crossovers, researchers defined a binary outcome a priori: patients were considered to have had a successful treatment response if they improved by at least 8 points on the WOMAC scale (a clinically important difference) and did not cross over from their assigned treatment. At 6 months, 67.1% of patients assigned to surgery showed a successful treatment response, compared with 43.8% of patients assigned to physical therapy alone (P = .001).¹

In patients who crossed over, the last scores before crossover were carried over, and primary analysis of the WOMAC score at 6 months was repeated to estimate the effect of crossovers from the nonoperative to the surgery group. This exploratory analysis showed a 13.0-point improvement in WOMAC score at 6 months with physical therapy alone vs a 20.9-point improvement with surgery, suggesting that the similarity in outcomes between the two groups may be explained in part by additional improvements from surgery for those who crossed over from physical therapy alone.¹

Implications for functional improvement

Lacking a comparison group that underwent a sham surgical procedure, one cannot conclude that surgery after crossover improved functional status in those patients. However, there was no significant difference in WOMAC physical function scores at 12 months between the 30% of patients in the physical therapy group who crossed over and underwent surgery during the first 6 months and patients initially assigned to surgery. This finding suggests that physical therapy can be recommended as a first-line therapy, although we must be cautious, given that the physical therapy group required more background therapy (eg, intra-articular glucocorticoid injections), and that this study was not powered to detect such differences at 12 months.

Also, a patient may need to get better quickly, to get back to work, for example. Although the data were not definitive, at 3 months the patients in the surgery group seemed to have better pain control and function than those in the physical therapy group. A cost-benefit analysis of physical therapy compared with surgery for short-term outcomes may be helpful before generalizing these findings.

SURGERY VS SHAM PROCEDURE: THE FIDELITY GROUP RESULTS

In a later publication from the Finnish Degenerative Meniscal Lesion Study (FIDELITY) Group,²⁴ 146 patients with symptoms consistent with degenerative meniscal tear but no knee osteoarthritis were randomized to undergo arthroscopic partial meniscectomy or a sham procedure. At 12 months, no differences were noted between the groups in terms of change of symptoms from baseline to 12 months.

The authors concluded that the outcomes with meniscectomy were no better than with a sham procedure.²⁴

SURGERY FIRST, OR PHYSICAL THERAPY FIRST?

The use of knee arthroscopy has increased sharply in middle-aged patients in recent years. Indeed, this demographic group accounts for nearly half of the knee arthroscopic procedures performed for meniscal tears, although the increase may be due in part to issues with surgeons’ coding and insurance authorization.¹⁶

The METEOR trial showed that a structured physical therapy program can be as effective as surgery as a first-line therapy in many patients with symptomatic meniscal tears and mild to moderate osteoarthritis. These results should inform clinical practice in that most such patients need not be immediately referred for surgical intervention.

However, a subset of these patients may benefit from surgery rather than nonoperative therapy. Given the potential risks and public health implications of arthroscopic surgery for meniscal tears, further study is needed to better characterize these patients. A randomized sham-controlled trial is under way²⁵ with the goal of assessing the efficacy of arthroscopic partial meniscectomy for medial meniscus tears in patients with or without knee osteoarthritis, and it is hoped this study will shed further light on this issue.

Based on the results of the METEOR trial, the physical therapy regimen that was used may be reasonable before referring patients with knee osteoarthritis and symptomatic meniscal tears for surgery.

Many patients who have osteoarthritis of the knee and a torn meniscus can defer having the meniscus repaired and undergo physical therapy instead. If a trial of physical therapy does not help, they can opt for surgery later.

This seems to be the take-home message from the recent Meniscal Tear in Osteoarthritis Research (METEOR) trial,¹ which compared the efficacy of arthroscopic partial meniscectomy plus physical therapy vs physical therapy alone for patients with knee symptoms, a meniscal tear, and mild to moderate osteoarthritis of the knee.¹

In brief, patients improved to a roughly similar degree with either approach, and although many patients assigned to physical therapy eventually underwent surgery anyway by 6 months, the delay did not adversely affect outcomes.

In this article, we review the background, design, and findings of the METEOR trial, and their implications for clinical practice.

SURGERY: HIGH VOLUME, BUT LITTLE EVIDENCE

Magnetic resonance imaging (MRI) often incidentally reveals meniscal lesions in middle-aged and older patients who have osteoarthritis and knee pain.² Should these patients undergo arthroscopic meniscal repair? The decision is difficult, since it is hard to distinguish the symptoms of a meniscal tear from those of osteoarthritis.³

Current evidence suggests that, for symptomatic knee osteoarthritis by itself, arthroscopic surgery is no more effective than conservative management.^4,5 But what about surgery for a torn meniscus in addition to osteoarthritis?

Osteoarthritis is the most common joint disease, accounting for many physician visits.⁶ More than 26 million Americans over age 25 have some form of it, and the prevalence of symptomatic, radiographically confirmed osteoarthritis of the knee was 12.1% in the third National Health and Nutrition Examination Survey.⁷

We used to consider osteoarthritis a “wear-and-tear” disease—thus the term “degenerative joint disease.” But today, we know that it is an active response to injury, involving inflammatory and metabolic pathways.⁸ Moreover, the risk of osteoarthritis and its progression seems to be higher in those who have had meniscal injury and total or arthroscopic partial meniscectomy.^9,10

MRI is not commonly used in managing knee osteoarthritis, but it has been used diagnostically in patients with symptoms of a meniscal tear, such as clicking, locking, popping, giving way, and pain with pivoting or twisting. Traumatic meniscal tears (a longitudinal or radial tear pattern) most often occur in active younger people and often lead to meniscal surgery.^11,12 In contrast, degenerative meniscal tears (horizontal, oblique, or complex tear pattern or meniscal maceration) tend to occur in older people,^11,12 but how to manage them is not widely agreed upon.

Of note, most patients with osteoarthritis of the knee have torn, macerated, or heavily damaged menisci.^13,14 Meniscal lesions are also common in middle-aged people in the general population, with a higher prevalence in people who are older, heavier, or female, or who have a family history of osteoarthritis.¹⁵

These abnormalities are only weakly associated with symptoms.² However, when a patient has knee symptoms and a torn meniscus is detected on MRI, the tear is often assumed to be the source of the symptoms, and meniscal tears are the most common reason for arthroscopy.¹⁶

Since we have no way to prevent the progression of joint damage from osteoarthritis with drugs or by any other means, the goal is to alleviate the symptoms. Many patients report pain relief or functional improvement after arthroscopic surgery. But arthroscopic lavage or debridement for osteoarthritis has not been found to be better than conservative treatment or placebo in randomized controlled trials.^4,5

In contrast, the current standard treatment for a symptomatic degenerative meniscal tear is arthroscopic partial meniscectomy. Nearly 500,000 of these procedures are performed annually in the United States.¹⁶ But based on the best evidence, arthroscopic partial meniscectomy does not result in better pain relief and functional improvement than does physical therapy alone in patients who have a torn meniscus and knee osteoarthritis.^17,18

OVERVIEW OF THE METEOR TRIAL

The METEOR trial was a randomized controlled trial conducted at seven US tertiary referral centers. Its aim was to compare the short-term (6-month) and long-term (12-month) efficacy of arthroscopic partial meniscectomy and physical therapy in patients with symptomatic meniscal tear and osteoarthritis of the knee.¹⁹ It was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases.¹

Patients were age 45 and older

METEOR patients had to be at least 45 years old and have symptomatic meniscal tears and knee osteoarthritis detected on MRI or radiography.¹

Osteoarthritis was defined broadly, given that it begins well before the appearance of radiographic evidence such as an osteophyte or joint-space narrowing.¹⁹ Patients with cartilage defects on MRI were also enrolled, as were patients with radiographically documented osteoarthritis.¹⁹

Patients were considered to have a symptomatic meniscal tear if they had had at least 4 weeks of symptoms (such as episodic pain and pain that was acute and localized to one spot on the knee, as well as typical mechanical pain suggesting a meniscal tear, such as clicking, catching, popping, giving way, or pain with pivoting or twisting) in addition to evidence of a meniscal tear on MRI.¹⁹

Patients were excluded if they had a chronically locked knee (a clear-cut indication for arthroscopic partial meniscectomy), advanced osteoarthritis (Kellgren-Lawrence grade 4), inflammatory arthritis, clinically symptomatic chondrocalcinosis, or bilateral symptomatic meniscal tears.¹⁹ Patients who had undergone surgery or injection of a viscosupplement in the index knee during the past 4 weeks were also excluded.¹⁹

Of 1,330 eligible patients, 351 (26.4%) were enrolled and randomly assigned in a 1:1 ratio to a treatment group by means of a secure program on the trial website.^1,19 Of those who were eligible but did not enroll, 195 (14.6%) were not referred and 784 (58.9%) declined to participate. Of those who declined, more preferred surgery than physical therapy (36.1% vs 21%). No information is available on any differences in baseline characteristics between the enrolled patients and the eligible patients who declined.

Randomization was done in blocks of varying size within each site, stratified according to sex and the extent of osteoarthritis on baseline radiography. The extent of osteoarthritis was categorized either as Kellgren-Lawrence grade 0 (normal, no features of osteoarthritis) to grade 2 (definite osteoarthritis, a definite osteophyte without joint-space narrowing) or as Kellgren-Lawrence grade 3 (moderate osteoarthritis, < 50% joint-space narrowing).^1,19 The two treatment groups were similar with respect to age, sex, race or ethnicity, baseline Kellgren-Lawrence grade, and baseline Western Ontario and McMaster Universities Arthritis Index (WOMAC) physical function score.¹

The mean age of the participants was 58, and 85% were white. Sixty-three percent had Kellgren-Lawrence grade 0 to 2 osteoarthritis, and 27% had grade 3.¹

Surgery plus physical therapy vs physical therapy alone

The surgery group underwent arthroscopic partial meniscectomy, which involved trimming the damaged meniscus back to a stable rim^1,19 and trimming loose fragments of cartilage and bone.

After the procedure, patients were scheduled for physical therapy. Although there is no consensus on the need for or the effectiveness of postoperative physical therapy in this setting, the investigators believed that including it in both study groups would help to isolate the independent effects of surgery. The physical therapy regimen after surgery was similar to that provided in the nonoperative group.^1,19

Physical therapy was designed to address inflammation, range of motion, muscle strength, muscle-length restriction, functional mobility, and proprioception and balance.^1,19 There were three stages; criteria for advancing from one phase to the next included the level of self-reported pain, observed strength, range of knee motion, knee effusion, and functional mobility.^1,18

The duration of participation varied depending on the pace of improvement. Generally, the program lasted about 6 weeks.^1,19

Crossover and other therapies were allowed

Crossover from physical therapy alone to surgery was allowed during the trial if the patient and surgeon thought it was clinically indicated.

Participants in both groups were permitted to take acetaminophen and nonsteroidal anti-inflammatory drugs as needed. Intra-articular injections of glucocorticoids were also allowed during the trial.

OUTCOMES MEASURED

WOMAC physical function score

The primary outcome of the METEOR trial was the difference between the study groups in the change in WOMAC physical function score from baseline to 6 months, at which point participants were expected to have achieved maximum improvement.^1,19 Questionnaires were also administered at 3 months to assess the early response to surgery or physical therapy and again at 12 months.

The complete WOMAC also measures pain and stiffness in addition to physical function, with separate subscales for each. The change in WOMAC score is one of the most widely endorsed outcome measures in assessing interventions in osteoarthritis or other conditions of the lower extremities.²⁰ The METEOR trial authors considered the WOMAC scale to be highly valid and reliable, with a Cronbach alpha value of 0.97 (maximum value = 1; the higher the better).

No ceiling or floor effects were observed in the WOMAC physical function score in patients with osteoarthritis and a meniscal tear in a pilot study for METEOR.¹⁹

In the main METEOR study, WOMAC physical function was scored on a scale of 0 to 100, with a higher score indicating worse physical function.¹ Changes in the score were also measured as a yes-or-no question, defined a priori as whether the score declined by at least 8 points, which is considered the minimal clinically important difference in osteoarthritis patients.^1,19

KOOS and MOS SF-36 scores

Secondary outcomes were measured in several domains, including pain, generic functional status, quality of life, and health care utilization.^1,19

The KOOS (Knee Injury and OA Outcome Scale) is specific for knee pain, being designed to evaluate short-term and long-term symptoms and function in patients with knee injury and associated problems.²¹ It has five subscales, which are scored separately: pain, other symptoms, activities of daily living, sport and recreation, and knee-related quality of life.²¹ Since the WOMAC pain scale showed a ceiling effect in the pilot study in patients undergoing surgery, the authors chose the KOOS pain scale as a pain measure.¹⁹ Scores were transformed to a 0–100 scale, with a higher score indicating more pain.¹

The MOS SF-36 (Medical Outcomes Study 36-item short form) was used to measure general health status and function.^1,19

STATISTICAL ANALYSIS: INTENTION-TO-TREAT AND AS-TREATED

The study was powered to detect a 10-point difference in WOMAC physical function scores at 6 months of follow-up between the operative and nonoperative groups, anticipating losses to follow-up and crossover, with preplanned subgroup (Kellgren–Lawrence grade 0–2 vs grade 3) analysis.^1,19

The primary analysis used a modified intention-to-treat approach and was implemented with an analysis of covariance with changes in the WOMAC score from baseline to 6 months as the dependent variable, treatment as the independent variable of interest, and study site as a covariate. Other covariates, such as age, sex, and baseline Kellgren-Lawrence grade, were balanced across groups and were therefore not included in the analysis.^1,19

Secondary analyses used an “as-treated” approach, ie, according to the treatment actually received.^1,19 Secondary intention-to-treat analysis—using binary outcome measures in which treatment failure was defined as improvement in the WOMAC score of less than 8 points or crossing over to the other treatment—was also performed to estimate efficacy at the level of the patient rather than at the group level.^1,19

BOTH GROUPS IMPROVED

In the intention-to-treat analyses at 6 months and 12 months after randomization, both groups improved, with no clinically important or statistically significant differences between the groups in functional status (WOMAC score, MOS SF-36 score) or pain (KOOS score).¹ The mean improvement (decline) in the WOMAC score from baseline to 6 months was 20.9 points in the surgery group vs 18.5 points in the physical therapy group, a difference of 2.4 points (95% confidence interval [CI], −1.8 to 6.5).¹

35% of physical therapy patients underwent surgery by 12 months

Of the 177 patients randomized to physical therapy alone, by 6 months 1 had died, 1 had undergone total knee replacement, 4 had withdrawn, and 2 were lost to follow-up. Of the 169 remaining, 51 (30%) had undergone arthroscopic partial meniscectomy. An additional 8 patients who were assigned to physical therapy crossed over to surgery between 6 and 12 months.^1,19

Of the 174 patients randomized to surgery, by 6 months 1 had died, 3 had undergone total knee replacement, 7 had withdrawn, and 2 were ineligible. Of the 161 remaining, 9 (6%) had not undergone the procedure.

Other outcomes

Subgroup analysis based on the baseline radiographic grade (Kellgren-Lawrence grade 0 to 2 vs grade 3) did not show a difference between groups in functional improvement at 6 months (P = .13 for interaction).¹

No statistically significant difference was noted in rates of overall or specific adverse events between the two groups over the first 12 months.¹ Adverse events rated as mild or moderate in severity occurred in 15 participants in the surgery group and 13 participants in the physical therapy group.¹ Long-term risks associated with these interventions are being assessed, and longitudinal assessment of imaging studies is planned to address this question but is not yet available.^1,18

In the physical therapy group, 21 patients (12%) received intra-articular glucocorticoid injections, as did 9 patients (6%) in the surgery group.^1,19

TRANSLATING THE METEOR RESULTS TO EVERYDAY PRACTICE

There are many challenges in designing surgical trials. Indeed, by one estimate,²² only about 40% of treatment questions involving surgical procedures can be evaluated by a randomized controlled trial.

Although the METEOR trial was not blinded, it was the first large, multicenter, randomized controlled trial to compare arthroscopic partial meniscectomy vs standardized physical therapy by using high-quality methodology such as careful sample-size calculation, balancing the groups according to known prognostic factors with block randomization, and intention-to-treat analysis. Moreover, the outcome measures were obtained from validated self-reporting questionnaires (WOMAC for function and KOOS for pain), reducing the possibility of observer bias.¹⁹ In addition, analyses were performed with the analysts blinded to the randomization assignment.

Limitations of the trial

A few limitations of the study are worth noting.

Patients age 45 or older with both symptomatic meniscal tear and osteoarthritis were the target population of this study. However, it is important to distinguish between the study population and the target population in a physician’s practice.

The investigators adopted broad definitions of osteoarthritis and symptoms of meniscal tear. Twenty-one percent of participants had normal findings on plain radiography, with cartilage defects visible only on MRI. Further, episodic pain or acute pain localized to a joint line was regarded as a symptom consistent with a torn meniscus.

In practice, arthroscopic partial meniscectomy is usually considered when a patient with a long history of tolerable osteoarthritis presents with a sudden onset of intolerable pain after a squatting or twisting injury.

In addition, the study population was predominantly white (85%), and the study was performed in tertiary referral academic medical centers. Therefore, the outcomes achieved with surgery or physical therapy may not translate to the community setting. Clinicians must be careful to account for these types of differences in extrapolating to patients in their own practice.

Potential enrollment bias

Although randomization is a rigorous method that eliminates selection bias in assigning individuals to study and control groups, selective enrollment could have created bias.¹ As the authors mentioned, only 26% of eligible patients were enrolled, possibly reflecting patients’ or surgeons’ strong preferences for one treatment or the other. Because the study and control groups were hardly random samples of eligible populations, we must be careful in generalizing the efficacy of physical therapy.¹

Crossover may have obscured the benefit of surgery

During the first 6 months, 30% of patients crossed over from physical therapy to surgery. High crossover rates in surgical trials are common, especially when comparing surgery with medical therapy.²³ Given that most of the patients assigned to only physical therapy who crossed over to surgery did not have substantial improvement in functional status, it seems that crossover occurred by nonrandom factors, potentially biasing the study results. With the high degree of crossover from the nonoperative group to the surgical group, intention-to-treat analysis may have given an inflated estimate of the effect of physical therapy.

To account for crossovers, researchers defined a binary outcome a priori: patients were considered to have had a successful treatment response if they improved by at least 8 points on the WOMAC scale (a clinically important difference) and did not cross over from their assigned treatment. At 6 months, 67.1% of patients assigned to surgery showed a successful treatment response, compared with 43.8% of patients assigned to physical therapy alone (P = .001).¹

In patients who crossed over, the last scores before crossover were carried over, and primary analysis of the WOMAC score at 6 months was repeated to estimate the effect of crossovers from the nonoperative to the surgery group. This exploratory analysis showed a 13.0-point improvement in WOMAC score at 6 months with physical therapy alone vs a 20.9-point improvement with surgery, suggesting that the similarity in outcomes between the two groups may be explained in part by additional improvements from surgery for those who crossed over from physical therapy alone.¹

Implications for functional improvement

Lacking a comparison group that underwent a sham surgical procedure, one cannot conclude that surgery after crossover improved functional status in those patients. However, there was no significant difference in WOMAC physical function scores at 12 months between the 30% of patients in the physical therapy group who crossed over and underwent surgery during the first 6 months and patients initially assigned to surgery. This finding suggests that physical therapy can be recommended as a first-line therapy, although we must be cautious, given that the physical therapy group required more background therapy (eg, intra-articular glucocorticoid injections), and that this study was not powered to detect such differences at 12 months.

Also, a patient may need to get better quickly, to get back to work, for example. Although the data were not definitive, at 3 months the patients in the surgery group seemed to have better pain control and function than those in the physical therapy group. A cost-benefit analysis of physical therapy compared with surgery for short-term outcomes may be helpful before generalizing these findings.

SURGERY VS SHAM PROCEDURE: THE FIDELITY GROUP RESULTS

In a later publication from the Finnish Degenerative Meniscal Lesion Study (FIDELITY) Group,²⁴ 146 patients with symptoms consistent with degenerative meniscal tear but no knee osteoarthritis were randomized to undergo arthroscopic partial meniscectomy or a sham procedure. At 12 months, no differences were noted between the groups in terms of change of symptoms from baseline to 12 months.

The authors concluded that the outcomes with meniscectomy were no better than with a sham procedure.²⁴

SURGERY FIRST, OR PHYSICAL THERAPY FIRST?

The use of knee arthroscopy has increased sharply in middle-aged patients in recent years. Indeed, this demographic group accounts for nearly half of the knee arthroscopic procedures performed for meniscal tears, although the increase may be due in part to issues with surgeons’ coding and insurance authorization.¹⁶

The METEOR trial showed that a structured physical therapy program can be as effective as surgery as a first-line therapy in many patients with symptomatic meniscal tears and mild to moderate osteoarthritis. These results should inform clinical practice in that most such patients need not be immediately referred for surgical intervention.

However, a subset of these patients may benefit from surgery rather than nonoperative therapy. Given the potential risks and public health implications of arthroscopic surgery for meniscal tears, further study is needed to better characterize these patients. A randomized sham-controlled trial is under way²⁵ with the goal of assessing the efficacy of arthroscopic partial meniscectomy for medial meniscus tears in patients with or without knee osteoarthritis, and it is hoped this study will shed further light on this issue.

Based on the results of the METEOR trial, the physical therapy regimen that was used may be reasonable before referring patients with knee osteoarthritis and symptomatic meniscal tears for surgery.

Many patients with diabetes mellitus develop complex, accelerated, multifocal coronary artery disease. Moreover, if they undergo revascularization with either coronary artery bypass grafting (CABG) or percutaneous coronary intervention (PCI), their risk of morbidity and death afterward is higher than in those without diabetes.^1,2

Over the last 2 decades, CABG and PCI have advanced significantly, as have antithrombotic therapy and drug therapies to modify cardiovascular risk factors such as hyperlipidemia, hypertension, and diabetes.

Several earlier studies showed CABG to be more beneficial than PCI in diabetic patients with multivessel coronary artery disease.^3–5 However, the topic has been controversial, and a substantial proportion of these patients continue to undergo PCI rather than CABG.

There are two main reasons for the continued use of PCI in this population. First, PCI is evolving, with new adjuvant drugs and drugeluting stents. Many cardiologists believe that earlier trials, which did not use contemporary PCI techniques, are outdated and that current, state-of-the-art PCI may be equivalent to—if not superior to—CABG.

Second, PCI is often performed on an ad hoc basis immediately after diagnostic angiography, leaving little time for discussion with the patient about alternative treatments. In this scenario, patients are inclined to undergo PCI immediately, while they are already on the table in the catheterization suite, rather than CABG at a later date.⁶

In addition, although the current joint guide-lines of the American College of Cardiology and the American Heart Association state that CABG is preferable to PCI for patients with diabetes and multivessel coronary artery disease, they give it only a level IIa recommendation.⁷

The much-anticipated Future Revascularization Evaluation in Patients With Diabetes Mellitus: Optimal Management of Multivessel Disease (FREEDOM) trial⁸ was designed to settle the CABG-vs-PCI debate, thereby leading to a stronger guideline recommendation for the preferred revascularization strategy in this patient population.

WHY ARE DIABETIC PATIENTS DIFFERENT?

Diabetes mellitus is a major risk factor for premature and aggressive coronary artery disease. Several mechanisms have been proposed to explain this association.

Diabetic patients have higher concentrations of several inflammatory proteins than those without diabetes, including C-reactive protein, tumor necrosis factor, and platelet-derived soluble CD40 ligand. They also have higher levels of adhesion molecules such as vascular cell adhesion molecule-1 and intercellular adhesion molecule.^9,10 In addition, when blood sugar levels are high, platelets express more glycoprotein IIb/IIIa receptors and are therefore more prone to aggregate.¹¹

These prothrombotic and proinflammatory cytokines, in conjunction with endothelial dysfunction and metabolic disorders such as hyperglycemia, hyperlipidemia, obesity, insulin resistance, and oxidative stress, lead to accelerated atherosclerosis in patients with diabetes.¹² Also, because diabetes is a systemic disease, the atherosclerotic process is diffuse, and many patients with diabetes have left main coronary artery lesions and diffuse multivessel coronary artery disease.^13,14

Although the short-term outcomes of revascularization by any means are comparable in patients with and without diabetes, diabetic patients have lower long-term survival rates and higher rates of myocardial infarction and need for repeat procedures.¹⁵ Diabetic patients who undergo PCI have a high rate of stent thrombosis and restenosis.^16,17 Similarly, those undergoing CABG have higher rates of postoperative infection and renal and neurologic complications.^18,19

BEFORE THE FREEDOM TRIAL

The question of CABG vs PCI has plagued physicians ever since PCI came to the forefront in the 1980s. Before stents were widely used, PCI with balloon angioplasty was known to be comparable to CABG for single-vessel disease, but whether it was beneficial in patients with multivessel disease or left main disease was not entirely evident. Randomized clinical trials were launched to answer the question.

Studies of balloon angioplasty vs CABG

The BARI trial (Bypass Angioplasty Revascularization Investigation),^5,20 published in 1996, compared PCI (using balloon angioplasty without a stent) and CABG in patients with multivessel coronary artery disease (Table 1^20–29).

Between 1988 and 1991, the trial randomly assigned 1,829 patients with multivessel disease to receive either PCI or CABG and compared their long-term outcomes. Although there was no difference in mortality rates between the two groups overall, the diabetic subgroup had a significantly better survival rate with CABG than with PCI, which was sustained over a follow-up period of 10 years.⁵

BARI had a significant clinical impact at the time and led to a clinical alert by the National Heart, Lung, and Blood Institute recommending CABG over PCI for patients with diabetes. However, not everyone accepted the results, because they were based on a small number of patients (n = 353) in a retrospectively determined subgroup. Further, the BARI trial was conducted before the advent of coronary stents, which were later shown to improve outcomes after PCI. Also, optimal medical therapy after revascularization was not specified in the protocol, which likely affected outcomes.

EAST (Emory Angioplasty Versus Surgery Trial)²¹ and CABRI (Coronary Angioplasty Versus Bypass Revascularization Investigation) ²² were similar randomized trials comparing angioplasty and CABG in patients with multivessel coronary artery disease. These showed better outcomes after CABG in patients with diabetes. However, lack of statistical significance because of small sample sizes limited their clinical impact.

The ARTS trial (Arterial Revascularization Therapy Study) compared PCI (with bare-metal stents) and CABG in 1,205 patients with multivessel coronary artery disease.²³ The mortality rate did not differ significantly between two treatment groups overall or in the diabetic subgroup. However, the repeat revascularization rate was higher with PCI than with CABG.

The SoS trial (Stenting or Surgery)²⁴ had similar results.

The ERACI II trial (Argentine Randomized Study: Coronary Angioplasty With Stenting Versus Coronary Bypass Surgery in Multi-Vessel Disease)²⁵ found no difference in mortality rates at 5 years with CABG vs PCI.

These trials were criticized, as none of them routinely used glycoprotein IIb/IIIa inhibitors with PCI, which by then had been shown to reduce mortality rates.³⁰ However, these trials made it clear that restenosis requiring repeat revascularization was a major disadvantage of PCI with bare-metal stents compared with CABG in patients with diabetes. Drug-eluting stents, which significantly reduced the rates of in-stent restenosis and target-lesion revascularization, were expected to overcome this major disadvantage.

Studies of PCI with drug-eluting stents vs CABG

ARTS II was the first trial to compare PCI with drug-eluting stents vs CABG. This was a nonrandomized single-arm study of 607 patients (including 159 with diabetes) who were treated with drug-eluting stents; the outcomes were compared with the CABG group from the earlier ARTS trial.³¹

At 3 years, in the diabetic subgroup, the rates of death, myocardial infarction, and stroke were not significantly different between treatments, although a trend favored PCI. However, this comparison was limited by selection bias, as ARTS II was a nonrandomized trial in which operators chose patients for drug-eluting stents in an attempt to match already known outcomes from the CABG cohort of ARTS.

SYNTAX (Synergy Between PCI With Taxus and Cardiac Surgery) was the first randomized trial comparing PCI with drug-eluting stents (in this trial, paclitaxel-eluting) vs CABG in patients with three-vessel or left main coronary artery disease.^26,27 Subgroup analysis in patients with diabetes mellitus revealed a higher rate of major adverse cardiac and cerebrovascular events (death, myocardial infarction, stroke, or repeat revascularization) in the PCI group than in the CABG patients, largely driven by higher rates of repeat revascularization after PCI.^32,33 SYNTAX was not designed to assess significant differences in rates of death.

The CARDIa trial (Coronary Artery Revascularization in Diabetes) randomized patients with diabetes and multivessel coronary artery disease to PCI (about one-third with bare-metal stents and two-thirds with drug-eluting stents) or CABG. Rates of major adverse cardiac and cerebrovascular events were higher in the PCI group, again largely driven by higher rates of repeat revascularization.⁴ CARDIa was stopped early because of a lack of enrollment and could not provide sufficient evidence to endorse one strategy over the other.

VA-CARDS (Veteran Affairs Coronary Artery Revascularization in Diabetes) randomized patients with diabetes and proximal left anterior descending artery or multivessel coronary artery disease to receive PCI with drug-eluting stents or CABG.²⁸ Although the rate of death was lower with CABG than with PCI at 2 years, the trial was underpowered and was terminated at 25% of the initial intended patient enrollment. In addition, only 9% of diabetic patients screened were angiographically eligible for the study.²⁹

Registry data. Analysis of a large data set from the National Cardiovascular Disease Registry and the Society of Thoracic Surgeons revealed a survival advantage of CABG over PCI for a follow-up period of 5 years.³⁴ However, this was a nonrandomized study, so its conclusions were not definitive.

THE FREEDOM TRIAL

Given the limitations of the trials described above, the National Heart, Lung, and Blood Institute sponsored the FREEDOM trial—an appropriately powered, randomized comparison of PCI (with drug-eluting stents) and CABG (using arterial grafting) in patients with diabetes and multivessel coronary artery disease using contemporary techniques and concomitant optimal medical therapy.⁸

FREEDOM study design

The FREEDOM trial enrolled 1,900 patients with diabetes and angiographically confirmed multivessel coronary artery disease (83% with three-vessel disease) with stenosis of more than 70% in two or more major epicardial vessels involving at least two separate coronary-artery territories. The main exclusion criteria were severe left main coronary artery stenosis (≥ 50% stenosis), class III or IV congestive heart failure, and previous CABG or valve surgery. For CABG surgery, arterial revascularization was encouraged.

Dual antiplatelet therapy was recommended for at least 12 months in patients receiving a drug-eluting stent, and optimal medical management for diabetes, hypertension, and hyperlipidemia was strongly advocated.

Between April 2005 and April 2010, 32,966 patients were screened, of whom 3,309 were eligible for the trial and 1,900 consented and were randomized (953 to the PCI group and 947 to the CABG group). The patients were followed for a minimum of 2 years and had a median follow-up time of 3.8 years. Outcomes were measured with an intention-to-treat analysis.

Study results

Patients. The groups were comparable with regard to baseline demographics and cardiac risk factors.

The mean age was 63; 29% of the patients were women, and 83% had three-vessel coronary artery disease. The mean hemoglobin A_1c was 7.8%, and the mean ejection fraction was 66%. The mean SYNTAX score, which defines the anatomic complexity of lesions, was 26 (≤ 22 is mild, 23–32 is intermediate, and ≥ 33 is high). The mean EURO score, which defines surgical risk, was 2.7 (a score ≥ 5 being associated with a lower rate of survival).

The primary composite outcome (death, nonfatal myocardial infarction, or nonfatal stroke) occurred less frequently in the CABG group than in the PCI group (Table 2). CABG was also associated with significantly lower rates of death from any cause and of myocardial infarction. Importantly, survival curves comparing the two groups diverged at 2-year follow-up. In contrast to other outcomes assessed, stroke occurred more often in the CABG group. The 5-year rates in the CABG group vs the PCI group were:

• Primary outcome—18.7% vs 26.6%, P = .005
• Death from any cause—10.9% vs 16.3%, P = .049
• Myocardial infarction—6% vs 13.9%, P < .0001
• Stroke—5.2% vs 2.4%, P = .03.

The secondary outcome (death, nonfatal myocardial infarction, nonfatal stroke, or repeat revascularization at 30 days or 12 months) had occurred significantly more often in the PCI group than in the CABG group at 1 year (16.8% vs 11.8%, P = .004), with most of the difference attributable to a higher repeat revascularization rate in the PCI group (12.6% vs 4.8%, P < .001).

Subgroup analysis. CABG was superior to PCI across all prespecified subgroups, covering the complexity of the coronary artery disease. Event rates with CABG vs PCI, by tertiles of the SYNTAX score:

• SYNTAX scores ≤ 22: 17.2% vs 23.2%
• SYNTAX scores 23–32: 17.7% vs 27.2%
• SYNTAX scores ≥ 33: 22.8% vs 30.6%.

Cost-effectiveness. Although up-front costs were higher with CABG, at $34,467 for the index hospitalization vs $25,845 for PCI (P < .001), when the in-trial results were extended to a lifetime horizon, CABG had an incremental cost-effectiveness ratio of $8,132 per quality-adjusted life-year gained vs PCI.³⁵ Traditionally, therapies are considered costeffective if the incremental cost-effectiveness ratio is less than $50,000 per quality-adjusted life-year gained.

WHY MAY CABG BE SUPERIOR IN DIABETIC PATIENTS?

The major advantage of CABG over PCI is the ability to achieve complete revascularization. Diabetic patients with coronary artery disease tend to have diffuse, multifocal disease with several stenotic lesions in multiple coronary arteries. While stents only treat the focal area of most significant occlusion, CABG may bypass all proximal vulnerable plaques that could potentially develop into culprit lesions over time, truly bypassing the diseased segments (Figure 1).

In addition, heavy calcification may not allow optimal stenting in these patients.

Use of multiple stents increases the risk of restenosis, which could lead to a higher incidence of myocardial infarction and need for repeat revascularization. This was evident in the FREEDOM trial, in which the mean number of stents per patient was 4.2. Also, some lesions need to be left untreated because of the complexity involved.

The major improvement in outcomes after CABG has resulted from using arterial conduits such as the internal mammary artery rather than the saphenous vein.³⁶ The patency rates of internal mammary artery grafts exceed 80% over 10 years.³⁷ Internal mammary artery grafting was done in 94% of patients receiving CABG in the FREEDOM trial.

WHAT DOES THIS MEAN?

FREEDOM was a landmark trial that confirmed that CABG provides significant benefit compared with contemporary PCI with drug-eluting stents in patients with diabetes mellitus and multivessel coronary artery disease. It was a large multicenter trial that was adequately powered, unlike most of the earlier trials of this topic.

Unlike previous trials in which the benefit of CABG was driven by reduction in repeat revascularizations alone, FREEDOM showed lower incidence rates of all-cause mortality and myocardial infarction with CABG than with PCI. CABG was better regardless of SYNTAX score, number of diseased vessels, ejection fraction, race, or sex of the patient, indicating that it leads to superior outcomes across a wide spectrum of patients.

An argument that cardiologists often cite when recommending PCI is that it can save money due to lower length of index hospital stay and lower procedure costs of with PCI than with CABG. However, in FREEDOM, CABG also appeared to be highly cost-effective.

FREEDOM had limitations

While FREEDOM provided robust data proving the superiority of CABG, the study had several limitations.

Although there was an overall survival benefit with CABG compared with PCI, the difference in incidence of cardiovascular deaths (which accounted for 64% of all deaths) was not statistically significant.

The trial included only patients who were eligible for both PCI and CABG. Hence, the results may not be generalizable to all diabetic patients with multivessel coronary artery disease—indeed, only 10% of those screened were considered eligible for the trial. However, it is likely that several patients screened in the FREEDOM trial may not have been eligible for PCI or CABG at the time of screening, since the revascularization decision was made by a multidisciplinary team and a more appropriate decision (either CABG or PCI) was then made.

Other factors limiting the general applicability of the results were low numbers of female patients (28.6%), black patients (6.3%), patients with an ejection fraction of 40% or less (2.5%), and patients with a low SYNTAX score (35%).

There were several unexplained observations as well. The difference in events between the treatment groups was much higher in North America than in other regions. The number of coronary lesions in the CABG group was high (mean = 5.74), but the average numbers of grafts used was only 2.9, and data were not provided regarding use of sequential grafting. Similarly, an average of only 3.5 of the six stenotic lesions per patient in the PCI group were revascularized; whether this was the result of procedural limitations with PCI was not entirely clear.

In addition, while the investigators mention that an average patient received four stents, a surprising finding was that the mean total length of the stents used was only 26 mm. This appears too small, as the usual length of one drug-eluting stent is about 20 to 30 mm.

Since only high-volume centers with good outcome data were included in the trial, the results may lack validity for patients undergoing revascularization at low-volume community centers.

It remains to be seen if the benefits of CABG will be sustained over 10 years and longer, when saphenous vein grafts tend to fail and require repeat revascularization, commonly performed with PCI. Previous data suggest that the longer the follow-up, the better the results with CABG. However, long-term results (> 10 years) in studies comparing drugeluting stents and CABG are not available.

Despite limitations, FREEDOM may change clinical practice

Despite these limitations, the FREEDOM trial has the potential to change clinical practice and strengthen current recommendations for CABG in these patients.

The trial underscored the importance of a multidisciplinary heart team approach in managing patients with complex coronary artery disease, similar to that being used in patients with severe aortic stenosis since transcatheter aortic valve replacement became available.

It should also bring an end to the practice of ad hoc PCI, especially in patients with diabetes and multivessel coronary artery disease. It is now imperative that physicians discuss current evidence for therapeutic options with the patients and their families before performing diagnostic angiography rather than immediately afterward, to give the patients ample time to make an informed decision. This is important, as most patients are likely to choose PCI in the same setting over CABG unless there is extensive discussion about the risks and benefits of both strategies done in an unbiased manner before angiography.

The fear of open heart surgery, a longer hospital stay, and a higher risk of stroke with CABG may lead some patients to choose PCI instead. In addition, factors that may preclude CABG in otherwise-eligible patients include anatomic considerations (diffuse distal vessel disease, poor conduits), individual factors (frailty, poor renal function, poor pulmonary function, patient preference), and local expertise.

Nevertheless, the patient should be presented with current evidence, and discussions regarding the optimal procedure should be held with a heart team, which should include an interventional cardiologist, a cardiothoracic surgeon, and a noninvasive cardiologist to facilitate an unbiased decision.

Regardless of the strategy chosen, the importance of compliance with optimal medical therapy (statins, antiplatelet agents, diabetes treatment) should be continuously emphasized to the patient.

WHAT DOES THE FUTURE HOLD?

Despite unequivocal evidence that CABG is superior to PCI in eligible patients with diabetes mellitus in the current era, PCI technologies continue to evolve rapidly. Newer second-generation drug-eluting stents have shown lower rates of restenosis^38,39 and may shorten the duration of post-PCI dual-antiplatelet therapy, a nuisance that has negatively affected outcomes with drug-eluting stents (because of problems of cost, poor compliance, and increased bleeding risk).

At the same time, CABG has also improved, with more extensive use of complete arterial conduits and use of an off-pump bypass technique that in theory poses a lower risk of stroke, although this has not yet been shown in a randomized trial.⁴⁰

Alternative approaches are being investigated. One of them is a hybrid procedure in which minimally invasive off-pump arterial grafting is combined with drug-eluting stents, which may reduce the risk of stroke and speed postoperative recovery.

Many patients with diabetes mellitus develop complex, accelerated, multifocal coronary artery disease. Moreover, if they undergo revascularization with either coronary artery bypass grafting (CABG) or percutaneous coronary intervention (PCI), their risk of morbidity and death afterward is higher than in those without diabetes.^1,2

Over the last 2 decades, CABG and PCI have advanced significantly, as have antithrombotic therapy and drug therapies to modify cardiovascular risk factors such as hyperlipidemia, hypertension, and diabetes.

Several earlier studies showed CABG to be more beneficial than PCI in diabetic patients with multivessel coronary artery disease.^3–5 However, the topic has been controversial, and a substantial proportion of these patients continue to undergo PCI rather than CABG.

There are two main reasons for the continued use of PCI in this population. First, PCI is evolving, with new adjuvant drugs and drugeluting stents. Many cardiologists believe that earlier trials, which did not use contemporary PCI techniques, are outdated and that current, state-of-the-art PCI may be equivalent to—if not superior to—CABG.

Second, PCI is often performed on an ad hoc basis immediately after diagnostic angiography, leaving little time for discussion with the patient about alternative treatments. In this scenario, patients are inclined to undergo PCI immediately, while they are already on the table in the catheterization suite, rather than CABG at a later date.⁶

In addition, although the current joint guide-lines of the American College of Cardiology and the American Heart Association state that CABG is preferable to PCI for patients with diabetes and multivessel coronary artery disease, they give it only a level IIa recommendation.⁷

The much-anticipated Future Revascularization Evaluation in Patients With Diabetes Mellitus: Optimal Management of Multivessel Disease (FREEDOM) trial⁸ was designed to settle the CABG-vs-PCI debate, thereby leading to a stronger guideline recommendation for the preferred revascularization strategy in this patient population.

WHY ARE DIABETIC PATIENTS DIFFERENT?

Diabetes mellitus is a major risk factor for premature and aggressive coronary artery disease. Several mechanisms have been proposed to explain this association.

Diabetic patients have higher concentrations of several inflammatory proteins than those without diabetes, including C-reactive protein, tumor necrosis factor, and platelet-derived soluble CD40 ligand. They also have higher levels of adhesion molecules such as vascular cell adhesion molecule-1 and intercellular adhesion molecule.^9,10 In addition, when blood sugar levels are high, platelets express more glycoprotein IIb/IIIa receptors and are therefore more prone to aggregate.¹¹

These prothrombotic and proinflammatory cytokines, in conjunction with endothelial dysfunction and metabolic disorders such as hyperglycemia, hyperlipidemia, obesity, insulin resistance, and oxidative stress, lead to accelerated atherosclerosis in patients with diabetes.¹² Also, because diabetes is a systemic disease, the atherosclerotic process is diffuse, and many patients with diabetes have left main coronary artery lesions and diffuse multivessel coronary artery disease.^13,14

Although the short-term outcomes of revascularization by any means are comparable in patients with and without diabetes, diabetic patients have lower long-term survival rates and higher rates of myocardial infarction and need for repeat procedures.¹⁵ Diabetic patients who undergo PCI have a high rate of stent thrombosis and restenosis.^16,17 Similarly, those undergoing CABG have higher rates of postoperative infection and renal and neurologic complications.^18,19

BEFORE THE FREEDOM TRIAL

The question of CABG vs PCI has plagued physicians ever since PCI came to the forefront in the 1980s. Before stents were widely used, PCI with balloon angioplasty was known to be comparable to CABG for single-vessel disease, but whether it was beneficial in patients with multivessel disease or left main disease was not entirely evident. Randomized clinical trials were launched to answer the question.

Studies of balloon angioplasty vs CABG

The BARI trial (Bypass Angioplasty Revascularization Investigation),^5,20 published in 1996, compared PCI (using balloon angioplasty without a stent) and CABG in patients with multivessel coronary artery disease (Table 1^20–29).

Between 1988 and 1991, the trial randomly assigned 1,829 patients with multivessel disease to receive either PCI or CABG and compared their long-term outcomes. Although there was no difference in mortality rates between the two groups overall, the diabetic subgroup had a significantly better survival rate with CABG than with PCI, which was sustained over a follow-up period of 10 years.⁵

BARI had a significant clinical impact at the time and led to a clinical alert by the National Heart, Lung, and Blood Institute recommending CABG over PCI for patients with diabetes. However, not everyone accepted the results, because they were based on a small number of patients (n = 353) in a retrospectively determined subgroup. Further, the BARI trial was conducted before the advent of coronary stents, which were later shown to improve outcomes after PCI. Also, optimal medical therapy after revascularization was not specified in the protocol, which likely affected outcomes.

EAST (Emory Angioplasty Versus Surgery Trial)²¹ and CABRI (Coronary Angioplasty Versus Bypass Revascularization Investigation) ²² were similar randomized trials comparing angioplasty and CABG in patients with multivessel coronary artery disease. These showed better outcomes after CABG in patients with diabetes. However, lack of statistical significance because of small sample sizes limited their clinical impact.

The ARTS trial (Arterial Revascularization Therapy Study) compared PCI (with bare-metal stents) and CABG in 1,205 patients with multivessel coronary artery disease.²³ The mortality rate did not differ significantly between two treatment groups overall or in the diabetic subgroup. However, the repeat revascularization rate was higher with PCI than with CABG.

The SoS trial (Stenting or Surgery)²⁴ had similar results.

The ERACI II trial (Argentine Randomized Study: Coronary Angioplasty With Stenting Versus Coronary Bypass Surgery in Multi-Vessel Disease)²⁵ found no difference in mortality rates at 5 years with CABG vs PCI.

These trials were criticized, as none of them routinely used glycoprotein IIb/IIIa inhibitors with PCI, which by then had been shown to reduce mortality rates.³⁰ However, these trials made it clear that restenosis requiring repeat revascularization was a major disadvantage of PCI with bare-metal stents compared with CABG in patients with diabetes. Drug-eluting stents, which significantly reduced the rates of in-stent restenosis and target-lesion revascularization, were expected to overcome this major disadvantage.

Studies of PCI with drug-eluting stents vs CABG

ARTS II was the first trial to compare PCI with drug-eluting stents vs CABG. This was a nonrandomized single-arm study of 607 patients (including 159 with diabetes) who were treated with drug-eluting stents; the outcomes were compared with the CABG group from the earlier ARTS trial.³¹

At 3 years, in the diabetic subgroup, the rates of death, myocardial infarction, and stroke were not significantly different between treatments, although a trend favored PCI. However, this comparison was limited by selection bias, as ARTS II was a nonrandomized trial in which operators chose patients for drug-eluting stents in an attempt to match already known outcomes from the CABG cohort of ARTS.

SYNTAX (Synergy Between PCI With Taxus and Cardiac Surgery) was the first randomized trial comparing PCI with drug-eluting stents (in this trial, paclitaxel-eluting) vs CABG in patients with three-vessel or left main coronary artery disease.^26,27 Subgroup analysis in patients with diabetes mellitus revealed a higher rate of major adverse cardiac and cerebrovascular events (death, myocardial infarction, stroke, or repeat revascularization) in the PCI group than in the CABG patients, largely driven by higher rates of repeat revascularization after PCI.^32,33 SYNTAX was not designed to assess significant differences in rates of death.

The CARDIa trial (Coronary Artery Revascularization in Diabetes) randomized patients with diabetes and multivessel coronary artery disease to PCI (about one-third with bare-metal stents and two-thirds with drug-eluting stents) or CABG. Rates of major adverse cardiac and cerebrovascular events were higher in the PCI group, again largely driven by higher rates of repeat revascularization.⁴ CARDIa was stopped early because of a lack of enrollment and could not provide sufficient evidence to endorse one strategy over the other.

VA-CARDS (Veteran Affairs Coronary Artery Revascularization in Diabetes) randomized patients with diabetes and proximal left anterior descending artery or multivessel coronary artery disease to receive PCI with drug-eluting stents or CABG.²⁸ Although the rate of death was lower with CABG than with PCI at 2 years, the trial was underpowered and was terminated at 25% of the initial intended patient enrollment. In addition, only 9% of diabetic patients screened were angiographically eligible for the study.²⁹

Registry data. Analysis of a large data set from the National Cardiovascular Disease Registry and the Society of Thoracic Surgeons revealed a survival advantage of CABG over PCI for a follow-up period of 5 years.³⁴ However, this was a nonrandomized study, so its conclusions were not definitive.

THE FREEDOM TRIAL

Given the limitations of the trials described above, the National Heart, Lung, and Blood Institute sponsored the FREEDOM trial—an appropriately powered, randomized comparison of PCI (with drug-eluting stents) and CABG (using arterial grafting) in patients with diabetes and multivessel coronary artery disease using contemporary techniques and concomitant optimal medical therapy.⁸

FREEDOM study design

The FREEDOM trial enrolled 1,900 patients with diabetes and angiographically confirmed multivessel coronary artery disease (83% with three-vessel disease) with stenosis of more than 70% in two or more major epicardial vessels involving at least two separate coronary-artery territories. The main exclusion criteria were severe left main coronary artery stenosis (≥ 50% stenosis), class III or IV congestive heart failure, and previous CABG or valve surgery. For CABG surgery, arterial revascularization was encouraged.

Dual antiplatelet therapy was recommended for at least 12 months in patients receiving a drug-eluting stent, and optimal medical management for diabetes, hypertension, and hyperlipidemia was strongly advocated.

Between April 2005 and April 2010, 32,966 patients were screened, of whom 3,309 were eligible for the trial and 1,900 consented and were randomized (953 to the PCI group and 947 to the CABG group). The patients were followed for a minimum of 2 years and had a median follow-up time of 3.8 years. Outcomes were measured with an intention-to-treat analysis.

Study results

Patients. The groups were comparable with regard to baseline demographics and cardiac risk factors.

The mean age was 63; 29% of the patients were women, and 83% had three-vessel coronary artery disease. The mean hemoglobin A_1c was 7.8%, and the mean ejection fraction was 66%. The mean SYNTAX score, which defines the anatomic complexity of lesions, was 26 (≤ 22 is mild, 23–32 is intermediate, and ≥ 33 is high). The mean EURO score, which defines surgical risk, was 2.7 (a score ≥ 5 being associated with a lower rate of survival).

The primary composite outcome (death, nonfatal myocardial infarction, or nonfatal stroke) occurred less frequently in the CABG group than in the PCI group (Table 2). CABG was also associated with significantly lower rates of death from any cause and of myocardial infarction. Importantly, survival curves comparing the two groups diverged at 2-year follow-up. In contrast to other outcomes assessed, stroke occurred more often in the CABG group. The 5-year rates in the CABG group vs the PCI group were:

• Primary outcome—18.7% vs 26.6%, P = .005
• Death from any cause—10.9% vs 16.3%, P = .049
• Myocardial infarction—6% vs 13.9%, P < .0001
• Stroke—5.2% vs 2.4%, P = .03.

The secondary outcome (death, nonfatal myocardial infarction, nonfatal stroke, or repeat revascularization at 30 days or 12 months) had occurred significantly more often in the PCI group than in the CABG group at 1 year (16.8% vs 11.8%, P = .004), with most of the difference attributable to a higher repeat revascularization rate in the PCI group (12.6% vs 4.8%, P < .001).

Subgroup analysis. CABG was superior to PCI across all prespecified subgroups, covering the complexity of the coronary artery disease. Event rates with CABG vs PCI, by tertiles of the SYNTAX score:

• SYNTAX scores ≤ 22: 17.2% vs 23.2%
• SYNTAX scores 23–32: 17.7% vs 27.2%
• SYNTAX scores ≥ 33: 22.8% vs 30.6%.

Cost-effectiveness. Although up-front costs were higher with CABG, at $34,467 for the index hospitalization vs $25,845 for PCI (P < .001), when the in-trial results were extended to a lifetime horizon, CABG had an incremental cost-effectiveness ratio of $8,132 per quality-adjusted life-year gained vs PCI.³⁵ Traditionally, therapies are considered costeffective if the incremental cost-effectiveness ratio is less than $50,000 per quality-adjusted life-year gained.

WHY MAY CABG BE SUPERIOR IN DIABETIC PATIENTS?

The major advantage of CABG over PCI is the ability to achieve complete revascularization. Diabetic patients with coronary artery disease tend to have diffuse, multifocal disease with several stenotic lesions in multiple coronary arteries. While stents only treat the focal area of most significant occlusion, CABG may bypass all proximal vulnerable plaques that could potentially develop into culprit lesions over time, truly bypassing the diseased segments (Figure 1).

In addition, heavy calcification may not allow optimal stenting in these patients.

Use of multiple stents increases the risk of restenosis, which could lead to a higher incidence of myocardial infarction and need for repeat revascularization. This was evident in the FREEDOM trial, in which the mean number of stents per patient was 4.2. Also, some lesions need to be left untreated because of the complexity involved.

The major improvement in outcomes after CABG has resulted from using arterial conduits such as the internal mammary artery rather than the saphenous vein.³⁶ The patency rates of internal mammary artery grafts exceed 80% over 10 years.³⁷ Internal mammary artery grafting was done in 94% of patients receiving CABG in the FREEDOM trial.

WHAT DOES THIS MEAN?

FREEDOM was a landmark trial that confirmed that CABG provides significant benefit compared with contemporary PCI with drug-eluting stents in patients with diabetes mellitus and multivessel coronary artery disease. It was a large multicenter trial that was adequately powered, unlike most of the earlier trials of this topic.

Unlike previous trials in which the benefit of CABG was driven by reduction in repeat revascularizations alone, FREEDOM showed lower incidence rates of all-cause mortality and myocardial infarction with CABG than with PCI. CABG was better regardless of SYNTAX score, number of diseased vessels, ejection fraction, race, or sex of the patient, indicating that it leads to superior outcomes across a wide spectrum of patients.

An argument that cardiologists often cite when recommending PCI is that it can save money due to lower length of index hospital stay and lower procedure costs of with PCI than with CABG. However, in FREEDOM, CABG also appeared to be highly cost-effective.

FREEDOM had limitations

While FREEDOM provided robust data proving the superiority of CABG, the study had several limitations.

Although there was an overall survival benefit with CABG compared with PCI, the difference in incidence of cardiovascular deaths (which accounted for 64% of all deaths) was not statistically significant.

The trial included only patients who were eligible for both PCI and CABG. Hence, the results may not be generalizable to all diabetic patients with multivessel coronary artery disease—indeed, only 10% of those screened were considered eligible for the trial. However, it is likely that several patients screened in the FREEDOM trial may not have been eligible for PCI or CABG at the time of screening, since the revascularization decision was made by a multidisciplinary team and a more appropriate decision (either CABG or PCI) was then made.

Other factors limiting the general applicability of the results were low numbers of female patients (28.6%), black patients (6.3%), patients with an ejection fraction of 40% or less (2.5%), and patients with a low SYNTAX score (35%).

There were several unexplained observations as well. The difference in events between the treatment groups was much higher in North America than in other regions. The number of coronary lesions in the CABG group was high (mean = 5.74), but the average numbers of grafts used was only 2.9, and data were not provided regarding use of sequential grafting. Similarly, an average of only 3.5 of the six stenotic lesions per patient in the PCI group were revascularized; whether this was the result of procedural limitations with PCI was not entirely clear.

In addition, while the investigators mention that an average patient received four stents, a surprising finding was that the mean total length of the stents used was only 26 mm. This appears too small, as the usual length of one drug-eluting stent is about 20 to 30 mm.

Since only high-volume centers with good outcome data were included in the trial, the results may lack validity for patients undergoing revascularization at low-volume community centers.

It remains to be seen if the benefits of CABG will be sustained over 10 years and longer, when saphenous vein grafts tend to fail and require repeat revascularization, commonly performed with PCI. Previous data suggest that the longer the follow-up, the better the results with CABG. However, long-term results (> 10 years) in studies comparing drugeluting stents and CABG are not available.

Despite limitations, FREEDOM may change clinical practice

Despite these limitations, the FREEDOM trial has the potential to change clinical practice and strengthen current recommendations for CABG in these patients.

The trial underscored the importance of a multidisciplinary heart team approach in managing patients with complex coronary artery disease, similar to that being used in patients with severe aortic stenosis since transcatheter aortic valve replacement became available.

It should also bring an end to the practice of ad hoc PCI, especially in patients with diabetes and multivessel coronary artery disease. It is now imperative that physicians discuss current evidence for therapeutic options with the patients and their families before performing diagnostic angiography rather than immediately afterward, to give the patients ample time to make an informed decision. This is important, as most patients are likely to choose PCI in the same setting over CABG unless there is extensive discussion about the risks and benefits of both strategies done in an unbiased manner before angiography.

The fear of open heart surgery, a longer hospital stay, and a higher risk of stroke with CABG may lead some patients to choose PCI instead. In addition, factors that may preclude CABG in otherwise-eligible patients include anatomic considerations (diffuse distal vessel disease, poor conduits), individual factors (frailty, poor renal function, poor pulmonary function, patient preference), and local expertise.

Nevertheless, the patient should be presented with current evidence, and discussions regarding the optimal procedure should be held with a heart team, which should include an interventional cardiologist, a cardiothoracic surgeon, and a noninvasive cardiologist to facilitate an unbiased decision.

Regardless of the strategy chosen, the importance of compliance with optimal medical therapy (statins, antiplatelet agents, diabetes treatment) should be continuously emphasized to the patient.

WHAT DOES THE FUTURE HOLD?

Despite unequivocal evidence that CABG is superior to PCI in eligible patients with diabetes mellitus in the current era, PCI technologies continue to evolve rapidly. Newer second-generation drug-eluting stents have shown lower rates of restenosis^38,39 and may shorten the duration of post-PCI dual-antiplatelet therapy, a nuisance that has negatively affected outcomes with drug-eluting stents (because of problems of cost, poor compliance, and increased bleeding risk).

At the same time, CABG has also improved, with more extensive use of complete arterial conduits and use of an off-pump bypass technique that in theory poses a lower risk of stroke, although this has not yet been shown in a randomized trial.⁴⁰

Alternative approaches are being investigated. One of them is a hybrid procedure in which minimally invasive off-pump arterial grafting is combined with drug-eluting stents, which may reduce the risk of stroke and speed postoperative recovery.

Many patients with diabetes mellitus develop complex, accelerated, multifocal coronary artery disease. Moreover, if they undergo revascularization with either coronary artery bypass grafting (CABG) or percutaneous coronary intervention (PCI), their risk of morbidity and death afterward is higher than in those without diabetes.^1,2

Over the last 2 decades, CABG and PCI have advanced significantly, as have antithrombotic therapy and drug therapies to modify cardiovascular risk factors such as hyperlipidemia, hypertension, and diabetes.

Several earlier studies showed CABG to be more beneficial than PCI in diabetic patients with multivessel coronary artery disease.^3–5 However, the topic has been controversial, and a substantial proportion of these patients continue to undergo PCI rather than CABG.

There are two main reasons for the continued use of PCI in this population. First, PCI is evolving, with new adjuvant drugs and drugeluting stents. Many cardiologists believe that earlier trials, which did not use contemporary PCI techniques, are outdated and that current, state-of-the-art PCI may be equivalent to—if not superior to—CABG.

Second, PCI is often performed on an ad hoc basis immediately after diagnostic angiography, leaving little time for discussion with the patient about alternative treatments. In this scenario, patients are inclined to undergo PCI immediately, while they are already on the table in the catheterization suite, rather than CABG at a later date.⁶

In addition, although the current joint guide-lines of the American College of Cardiology and the American Heart Association state that CABG is preferable to PCI for patients with diabetes and multivessel coronary artery disease, they give it only a level IIa recommendation.⁷

The much-anticipated Future Revascularization Evaluation in Patients With Diabetes Mellitus: Optimal Management of Multivessel Disease (FREEDOM) trial⁸ was designed to settle the CABG-vs-PCI debate, thereby leading to a stronger guideline recommendation for the preferred revascularization strategy in this patient population.

WHY ARE DIABETIC PATIENTS DIFFERENT?

Diabetes mellitus is a major risk factor for premature and aggressive coronary artery disease. Several mechanisms have been proposed to explain this association.

Diabetic patients have higher concentrations of several inflammatory proteins than those without diabetes, including C-reactive protein, tumor necrosis factor, and platelet-derived soluble CD40 ligand. They also have higher levels of adhesion molecules such as vascular cell adhesion molecule-1 and intercellular adhesion molecule.^9,10 In addition, when blood sugar levels are high, platelets express more glycoprotein IIb/IIIa receptors and are therefore more prone to aggregate.¹¹

These prothrombotic and proinflammatory cytokines, in conjunction with endothelial dysfunction and metabolic disorders such as hyperglycemia, hyperlipidemia, obesity, insulin resistance, and oxidative stress, lead to accelerated atherosclerosis in patients with diabetes.¹² Also, because diabetes is a systemic disease, the atherosclerotic process is diffuse, and many patients with diabetes have left main coronary artery lesions and diffuse multivessel coronary artery disease.^13,14

Although the short-term outcomes of revascularization by any means are comparable in patients with and without diabetes, diabetic patients have lower long-term survival rates and higher rates of myocardial infarction and need for repeat procedures.¹⁵ Diabetic patients who undergo PCI have a high rate of stent thrombosis and restenosis.^16,17 Similarly, those undergoing CABG have higher rates of postoperative infection and renal and neurologic complications.^18,19

BEFORE THE FREEDOM TRIAL

The question of CABG vs PCI has plagued physicians ever since PCI came to the forefront in the 1980s. Before stents were widely used, PCI with balloon angioplasty was known to be comparable to CABG for single-vessel disease, but whether it was beneficial in patients with multivessel disease or left main disease was not entirely evident. Randomized clinical trials were launched to answer the question.

Studies of balloon angioplasty vs CABG

The BARI trial (Bypass Angioplasty Revascularization Investigation),^5,20 published in 1996, compared PCI (using balloon angioplasty without a stent) and CABG in patients with multivessel coronary artery disease (Table 1^20–29).

Between 1988 and 1991, the trial randomly assigned 1,829 patients with multivessel disease to receive either PCI or CABG and compared their long-term outcomes. Although there was no difference in mortality rates between the two groups overall, the diabetic subgroup had a significantly better survival rate with CABG than with PCI, which was sustained over a follow-up period of 10 years.⁵

BARI had a significant clinical impact at the time and led to a clinical alert by the National Heart, Lung, and Blood Institute recommending CABG over PCI for patients with diabetes. However, not everyone accepted the results, because they were based on a small number of patients (n = 353) in a retrospectively determined subgroup. Further, the BARI trial was conducted before the advent of coronary stents, which were later shown to improve outcomes after PCI. Also, optimal medical therapy after revascularization was not specified in the protocol, which likely affected outcomes.

EAST (Emory Angioplasty Versus Surgery Trial)²¹ and CABRI (Coronary Angioplasty Versus Bypass Revascularization Investigation) ²² were similar randomized trials comparing angioplasty and CABG in patients with multivessel coronary artery disease. These showed better outcomes after CABG in patients with diabetes. However, lack of statistical significance because of small sample sizes limited their clinical impact.

The ARTS trial (Arterial Revascularization Therapy Study) compared PCI (with bare-metal stents) and CABG in 1,205 patients with multivessel coronary artery disease.²³ The mortality rate did not differ significantly between two treatment groups overall or in the diabetic subgroup. However, the repeat revascularization rate was higher with PCI than with CABG.

The SoS trial (Stenting or Surgery)²⁴ had similar results.

The ERACI II trial (Argentine Randomized Study: Coronary Angioplasty With Stenting Versus Coronary Bypass Surgery in Multi-Vessel Disease)²⁵ found no difference in mortality rates at 5 years with CABG vs PCI.

These trials were criticized, as none of them routinely used glycoprotein IIb/IIIa inhibitors with PCI, which by then had been shown to reduce mortality rates.³⁰ However, these trials made it clear that restenosis requiring repeat revascularization was a major disadvantage of PCI with bare-metal stents compared with CABG in patients with diabetes. Drug-eluting stents, which significantly reduced the rates of in-stent restenosis and target-lesion revascularization, were expected to overcome this major disadvantage.

Studies of PCI with drug-eluting stents vs CABG

ARTS II was the first trial to compare PCI with drug-eluting stents vs CABG. This was a nonrandomized single-arm study of 607 patients (including 159 with diabetes) who were treated with drug-eluting stents; the outcomes were compared with the CABG group from the earlier ARTS trial.³¹

At 3 years, in the diabetic subgroup, the rates of death, myocardial infarction, and stroke were not significantly different between treatments, although a trend favored PCI. However, this comparison was limited by selection bias, as ARTS II was a nonrandomized trial in which operators chose patients for drug-eluting stents in an attempt to match already known outcomes from the CABG cohort of ARTS.

SYNTAX (Synergy Between PCI With Taxus and Cardiac Surgery) was the first randomized trial comparing PCI with drug-eluting stents (in this trial, paclitaxel-eluting) vs CABG in patients with three-vessel or left main coronary artery disease.^26,27 Subgroup analysis in patients with diabetes mellitus revealed a higher rate of major adverse cardiac and cerebrovascular events (death, myocardial infarction, stroke, or repeat revascularization) in the PCI group than in the CABG patients, largely driven by higher rates of repeat revascularization after PCI.^32,33 SYNTAX was not designed to assess significant differences in rates of death.

The CARDIa trial (Coronary Artery Revascularization in Diabetes) randomized patients with diabetes and multivessel coronary artery disease to PCI (about one-third with bare-metal stents and two-thirds with drug-eluting stents) or CABG. Rates of major adverse cardiac and cerebrovascular events were higher in the PCI group, again largely driven by higher rates of repeat revascularization.⁴ CARDIa was stopped early because of a lack of enrollment and could not provide sufficient evidence to endorse one strategy over the other.

VA-CARDS (Veteran Affairs Coronary Artery Revascularization in Diabetes) randomized patients with diabetes and proximal left anterior descending artery or multivessel coronary artery disease to receive PCI with drug-eluting stents or CABG.²⁸ Although the rate of death was lower with CABG than with PCI at 2 years, the trial was underpowered and was terminated at 25% of the initial intended patient enrollment. In addition, only 9% of diabetic patients screened were angiographically eligible for the study.²⁹

Registry data. Analysis of a large data set from the National Cardiovascular Disease Registry and the Society of Thoracic Surgeons revealed a survival advantage of CABG over PCI for a follow-up period of 5 years.³⁴ However, this was a nonrandomized study, so its conclusions were not definitive.

THE FREEDOM TRIAL

Given the limitations of the trials described above, the National Heart, Lung, and Blood Institute sponsored the FREEDOM trial—an appropriately powered, randomized comparison of PCI (with drug-eluting stents) and CABG (using arterial grafting) in patients with diabetes and multivessel coronary artery disease using contemporary techniques and concomitant optimal medical therapy.⁸

FREEDOM study design

The FREEDOM trial enrolled 1,900 patients with diabetes and angiographically confirmed multivessel coronary artery disease (83% with three-vessel disease) with stenosis of more than 70% in two or more major epicardial vessels involving at least two separate coronary-artery territories. The main exclusion criteria were severe left main coronary artery stenosis (≥ 50% stenosis), class III or IV congestive heart failure, and previous CABG or valve surgery. For CABG surgery, arterial revascularization was encouraged.

Dual antiplatelet therapy was recommended for at least 12 months in patients receiving a drug-eluting stent, and optimal medical management for diabetes, hypertension, and hyperlipidemia was strongly advocated.

Between April 2005 and April 2010, 32,966 patients were screened, of whom 3,309 were eligible for the trial and 1,900 consented and were randomized (953 to the PCI group and 947 to the CABG group). The patients were followed for a minimum of 2 years and had a median follow-up time of 3.8 years. Outcomes were measured with an intention-to-treat analysis.

Study results

Patients. The groups were comparable with regard to baseline demographics and cardiac risk factors.

The mean age was 63; 29% of the patients were women, and 83% had three-vessel coronary artery disease. The mean hemoglobin A_1c was 7.8%, and the mean ejection fraction was 66%. The mean SYNTAX score, which defines the anatomic complexity of lesions, was 26 (≤ 22 is mild, 23–32 is intermediate, and ≥ 33 is high). The mean EURO score, which defines surgical risk, was 2.7 (a score ≥ 5 being associated with a lower rate of survival).

The primary composite outcome (death, nonfatal myocardial infarction, or nonfatal stroke) occurred less frequently in the CABG group than in the PCI group (Table 2). CABG was also associated with significantly lower rates of death from any cause and of myocardial infarction. Importantly, survival curves comparing the two groups diverged at 2-year follow-up. In contrast to other outcomes assessed, stroke occurred more often in the CABG group. The 5-year rates in the CABG group vs the PCI group were:

• Primary outcome—18.7% vs 26.6%, P = .005
• Death from any cause—10.9% vs 16.3%, P = .049
• Myocardial infarction—6% vs 13.9%, P < .0001
• Stroke—5.2% vs 2.4%, P = .03.

The secondary outcome (death, nonfatal myocardial infarction, nonfatal stroke, or repeat revascularization at 30 days or 12 months) had occurred significantly more often in the PCI group than in the CABG group at 1 year (16.8% vs 11.8%, P = .004), with most of the difference attributable to a higher repeat revascularization rate in the PCI group (12.6% vs 4.8%, P < .001).

Subgroup analysis. CABG was superior to PCI across all prespecified subgroups, covering the complexity of the coronary artery disease. Event rates with CABG vs PCI, by tertiles of the SYNTAX score:

• SYNTAX scores ≤ 22: 17.2% vs 23.2%
• SYNTAX scores 23–32: 17.7% vs 27.2%
• SYNTAX scores ≥ 33: 22.8% vs 30.6%.

Cost-effectiveness. Although up-front costs were higher with CABG, at $34,467 for the index hospitalization vs $25,845 for PCI (P < .001), when the in-trial results were extended to a lifetime horizon, CABG had an incremental cost-effectiveness ratio of $8,132 per quality-adjusted life-year gained vs PCI.³⁵ Traditionally, therapies are considered costeffective if the incremental cost-effectiveness ratio is less than $50,000 per quality-adjusted life-year gained.

WHY MAY CABG BE SUPERIOR IN DIABETIC PATIENTS?

The major advantage of CABG over PCI is the ability to achieve complete revascularization. Diabetic patients with coronary artery disease tend to have diffuse, multifocal disease with several stenotic lesions in multiple coronary arteries. While stents only treat the focal area of most significant occlusion, CABG may bypass all proximal vulnerable plaques that could potentially develop into culprit lesions over time, truly bypassing the diseased segments (Figure 1).

In addition, heavy calcification may not allow optimal stenting in these patients.

Use of multiple stents increases the risk of restenosis, which could lead to a higher incidence of myocardial infarction and need for repeat revascularization. This was evident in the FREEDOM trial, in which the mean number of stents per patient was 4.2. Also, some lesions need to be left untreated because of the complexity involved.

The major improvement in outcomes after CABG has resulted from using arterial conduits such as the internal mammary artery rather than the saphenous vein.³⁶ The patency rates of internal mammary artery grafts exceed 80% over 10 years.³⁷ Internal mammary artery grafting was done in 94% of patients receiving CABG in the FREEDOM trial.

WHAT DOES THIS MEAN?

FREEDOM was a landmark trial that confirmed that CABG provides significant benefit compared with contemporary PCI with drug-eluting stents in patients with diabetes mellitus and multivessel coronary artery disease. It was a large multicenter trial that was adequately powered, unlike most of the earlier trials of this topic.

Unlike previous trials in which the benefit of CABG was driven by reduction in repeat revascularizations alone, FREEDOM showed lower incidence rates of all-cause mortality and myocardial infarction with CABG than with PCI. CABG was better regardless of SYNTAX score, number of diseased vessels, ejection fraction, race, or sex of the patient, indicating that it leads to superior outcomes across a wide spectrum of patients.

An argument that cardiologists often cite when recommending PCI is that it can save money due to lower length of index hospital stay and lower procedure costs of with PCI than with CABG. However, in FREEDOM, CABG also appeared to be highly cost-effective.

FREEDOM had limitations

While FREEDOM provided robust data proving the superiority of CABG, the study had several limitations.

Although there was an overall survival benefit with CABG compared with PCI, the difference in incidence of cardiovascular deaths (which accounted for 64% of all deaths) was not statistically significant.

The trial included only patients who were eligible for both PCI and CABG. Hence, the results may not be generalizable to all diabetic patients with multivessel coronary artery disease—indeed, only 10% of those screened were considered eligible for the trial. However, it is likely that several patients screened in the FREEDOM trial may not have been eligible for PCI or CABG at the time of screening, since the revascularization decision was made by a multidisciplinary team and a more appropriate decision (either CABG or PCI) was then made.

Other factors limiting the general applicability of the results were low numbers of female patients (28.6%), black patients (6.3%), patients with an ejection fraction of 40% or less (2.5%), and patients with a low SYNTAX score (35%).

There were several unexplained observations as well. The difference in events between the treatment groups was much higher in North America than in other regions. The number of coronary lesions in the CABG group was high (mean = 5.74), but the average numbers of grafts used was only 2.9, and data were not provided regarding use of sequential grafting. Similarly, an average of only 3.5 of the six stenotic lesions per patient in the PCI group were revascularized; whether this was the result of procedural limitations with PCI was not entirely clear.

In addition, while the investigators mention that an average patient received four stents, a surprising finding was that the mean total length of the stents used was only 26 mm. This appears too small, as the usual length of one drug-eluting stent is about 20 to 30 mm.

Since only high-volume centers with good outcome data were included in the trial, the results may lack validity for patients undergoing revascularization at low-volume community centers.

It remains to be seen if the benefits of CABG will be sustained over 10 years and longer, when saphenous vein grafts tend to fail and require repeat revascularization, commonly performed with PCI. Previous data suggest that the longer the follow-up, the better the results with CABG. However, long-term results (> 10 years) in studies comparing drugeluting stents and CABG are not available.

Despite limitations, FREEDOM may change clinical practice

Despite these limitations, the FREEDOM trial has the potential to change clinical practice and strengthen current recommendations for CABG in these patients.

The trial underscored the importance of a multidisciplinary heart team approach in managing patients with complex coronary artery disease, similar to that being used in patients with severe aortic stenosis since transcatheter aortic valve replacement became available.

It should also bring an end to the practice of ad hoc PCI, especially in patients with diabetes and multivessel coronary artery disease. It is now imperative that physicians discuss current evidence for therapeutic options with the patients and their families before performing diagnostic angiography rather than immediately afterward, to give the patients ample time to make an informed decision. This is important, as most patients are likely to choose PCI in the same setting over CABG unless there is extensive discussion about the risks and benefits of both strategies done in an unbiased manner before angiography.

The fear of open heart surgery, a longer hospital stay, and a higher risk of stroke with CABG may lead some patients to choose PCI instead. In addition, factors that may preclude CABG in otherwise-eligible patients include anatomic considerations (diffuse distal vessel disease, poor conduits), individual factors (frailty, poor renal function, poor pulmonary function, patient preference), and local expertise.

Nevertheless, the patient should be presented with current evidence, and discussions regarding the optimal procedure should be held with a heart team, which should include an interventional cardiologist, a cardiothoracic surgeon, and a noninvasive cardiologist to facilitate an unbiased decision.

Regardless of the strategy chosen, the importance of compliance with optimal medical therapy (statins, antiplatelet agents, diabetes treatment) should be continuously emphasized to the patient.

WHAT DOES THE FUTURE HOLD?

Despite unequivocal evidence that CABG is superior to PCI in eligible patients with diabetes mellitus in the current era, PCI technologies continue to evolve rapidly. Newer second-generation drug-eluting stents have shown lower rates of restenosis^38,39 and may shorten the duration of post-PCI dual-antiplatelet therapy, a nuisance that has negatively affected outcomes with drug-eluting stents (because of problems of cost, poor compliance, and increased bleeding risk).

At the same time, CABG has also improved, with more extensive use of complete arterial conduits and use of an off-pump bypass technique that in theory poses a lower risk of stroke, although this has not yet been shown in a randomized trial.⁴⁰

Alternative approaches are being investigated. One of them is a hybrid procedure in which minimally invasive off-pump arterial grafting is combined with drug-eluting stents, which may reduce the risk of stroke and speed postoperative recovery.

Some members of the public may have noticed the conclusions of a recent study¹ that said that if an older postmenopausal woman has her bone mineral density measured to screen for osteoporosis and has a normal or only mildly low result, she does not need to come back for another measurement for approximately 15 years.

See related editorial

We believe this interpretation of the study’s findings is overly simplistic and may have the unfortunate result of causing some people to neglect their bone health. Moreover, the study looked mainly at baseline T scores as the determinant of the subsequent screening interval. However, clinicians must carefully consider a variety of clinical risk factors when deciding how often to obtain bone mineral density measurements. The ultimate goal is to not miss the window of opportunity for early detection and treatment when it would matter the most (ie, before fractures develop).

Here, we will review this recent study, its findings, and its implications.

OSTEOPOROSIS POSES AN ENORMOUS PUBLIC HEALTH PROBLEM

If we consider only the hip, an estimated 10 million people in the United States have osteoporosis (T score ≤ −2.5 or a preexisting fragility fracture), and 33.6 million have osteopenia (T score −1.01 to −2.49).² The number of people with osteopenia can be assumed to be much higher if other skeletal sites are considered.

By increasing the risk of fragility fractures, osteoporosis poses an enormous public health problem. The surgeon general’s report points out that one of every two white women over age 50 will experience an osteoporosis-related fracture in her lifetime.³ Of all osteoporosis-related fractures, those of the hip carry the worse clinical outcome. Approximately one in five elderly people who experience an osteoporosis-related hip fracture need long-term nursing home care, and as many as 20% die within 1 year.³

In recognition of the burden of osteoporosis, the US Preventive Services Task Force (USPSTF)⁴ and other scientific bodies^2,3 recommend an initial bone mineral density test for all women age 65 and older. Dual-energy x-ray absorptiometry (DXA) is considered the gold standard for bone mineral density testing. Although the patient population that should receive an initial bone mineral density test has been clearly identified (see below), guidelines on the optimal frequency of testing do not exist, as data have been lacking. Recognizing this knowledge gap, Gourlay et al¹ attempted to answer the question of how often elderly postmenopausal women should be retested.

WHEN DO 10% OF ELDERLY POSTMENOPAUSAL WOMEN REACH A T SCORE OF −2.5?

Gourlay et al¹ analyzed data from 4,957 women in the Study of Osteoporotic Fractures. These women were predominantly white, were at least 67 years old and ambulatory, and had normal bone mineral density or osteopenia and no history of hip or clinical vertebral fracture at baseline. They had been recruited between 1986 and 1988 at sites in Baltimore, MD, Minneapolis, MN, the Monongahela Valley near Pittsburgh, PA, and Portland, OR.

DXA of the hip had been performed at baseline and at multiple times thereafter. The average follow-up time was 8 years.

The primary outcome measured was how long it took for 10% of the patients to reach a T score of −2.5 or less at the femoral neck or total hip as they progressed from having normal bone mineral density to osteoporosis or from osteopenia to osteoporosis and before they developed a fracture or needed treatment for osteoporosis.

Clinical risk factors such as age, body mass index, estrogen use at baseline, fracture after age 50, current smoking, current or past use of glucocorticoids, and self-reported rheumatoid arthritis were included as covariates in time-to-event analyses.

ANSWER: 16.8 YEARS (IF NORMAL AT BASELINE)

The authors estimated that 10% of women would make the transition to osteoporosis before having a hip or clinical vertebral fracture in the following intervals:

• 16.8 years in women whose bone mineral density was normal at baseline (T score at femoral neck and total hip of −1.00 or higher)
• 17.3 years in women who had mild osteopenia at baseline (T score −1.01 to −1.49)
• 4.7 years in women with moderate osteopenia at baseline (T score −1.5 to −1.99)
• 1.1 years in women with advanced osteopenia at baseline (T score −2.00 to −2.49).

The authors also found that body mass index and current estrogen use were the only clinical risk factors that influenced these intervals; other clinical factors such as a fracture after age 50, current smoking, previous or current use of oral glucocorticoids, and self-reported rheumatoid arthritis did not.

They concluded that osteoporosis would develop in fewer than 10% of women if the rescreening interval was lengthened to 15 years for women with normal density or mild osteopenia, 5 years for women with moderate osteopenia, and 1 year for women with advanced osteopenia.

WHAT DOES THIS MEAN FOR THE PRACTICING CLINICIAN?

Who needs an initial DXA test according to current guidelines?

The USPSTF,⁴ the National Osteoporosis Foundation (NOF),⁵ the International Society for Clinical Densitometry (ISCD),⁶ and the American Association of Clinical Endocrinologists (AACE)⁷ propose that the following groups should undergo DXA:

• All women age 65 and older
• All postmenopausal women who have had a fragility fracture or who have one or more risk factors for osteoporosis (height loss, body mass index < 20 kg/m², family history of osteoporosis, active smoking, excessive alcohol consumption)
• Adults who have a condition (eg, rheumatoid arthritis) or are taking a medication (eg, glucocorticoids in a daily dose ≥ 5 mg of prednisone or its equivalent for ≥ 3 months) associated with low bone mass or bone loss
• Anyone being considered for drug therapy for osteoporosis, discontinuing therapy for osteoporosis (including estrogen), or being treated for osteoporosis, to monitor the effect of treatment.

Assessing fracture risk. Although clinicians have traditionally relied on bone mineral density obtained by DXA to estimate fracture risk, the World Health Organization has developed a computer-based algorithm that calculates an individual’s 10-year fracture probability from easily obtained clinical risk factors with or without adding femoral-neck bone mineral density. The Fracture Risk Assessment tool, or FRAX, has attracted intense interest since its introduction in 2007 and has been endorsed by the USPSTF⁴ and by other scientific societies, including the NOF⁵ and the ISCD.⁸ In fact, the most recent USPSTF guidelines,⁴ which recommend screening all women age 65 and older, call for using FRAX to identify younger women at higher risk of fracture.

According to FRAX, a 65-year-old white woman who has no risk factors has a 9.3% chance of developing a major osteoporotic fracture in the next 10 years. And if a younger woman (between the ages of 50 and 64) has a fracture risk as high or higher than a 65-year-old white woman who has no risk factors, then she too should be screened by DXA.

The FRAX calculator is available online at www.shef.ac.uk/FRAX.

What are the current recommendations about follow-up DXA testing?

In eligible patients, the Centers for Medicare and Medicaid Services will pay for a DXA scan every 2 years. This interval is based on the concept that in an otherwise healthy person, it takes a minimum of 2 years to see a significant change in bone mineral density that can be attributed to a biological change in the bone and not just chance. The USPSTF⁴ and scientific societies such as the NOF⁵ generally agree with the Medicare guidelines of retesting every 2 years but recognize certain clinical situations that may warrant more frequent retesting (see below).

But the real question is how long the DXA screening interval can be extended so that meaningful information can still be obtained to help make management decisions and before a complication such as a fracture occurs. While there is convincing evidence to support the recommendations for an initial DXA test, data to answer the question of how long the resting interval should be are lacking.

Before the study by Gourlay et al,¹ the only data on repeat DXA came from work by Hillier et al.⁹ But those investigators asked a different question. They were interested in how well repeated measurements predicted fractures. They used the same population that Gourlay et al did but evaluated fractures, not T scores. They concluded that in healthy, adult postmenopausal women, repeating the bone mineral density measurement up to 8 years later adds little value to initial measurement for predicting incident fractures.

Clinical factors also count

The T score should not be the only major factor determining the interval for bone mineral density testing in elderly women; clinical risk factors also should be kept in mind.

Gourlay et al concluded that age and T scores are the key predictive factors in determining the bone mineral density testing interval in elderly, postmenopausal women for screening purposes.¹ In their statistical model, clinical risk factors such as fracture after age 50, current smoking, previous or current use of glucocorticoids, and self-reported rheumatoid arthritis did not influence the testing interval. They say that clinicians should not feel compelled to shorten the testing interval when these risk factors are present.

Readers may take this to mean that if these results were strictly applied to a 70-year-old white woman receiving oral glucocorticoids for rheumatoid arthritis and who has a baseline T score of −1.45, then her next test may be postponed by 15 years (given that both these factors did not influence the testing interval). Readers may also conclude that if this patient’s T score were −1.51, then her screening interval would be 5 years and not 15 years.

However, Gourlay et al say¹ that clinicians can choose to shorten the testing interval if there is evidence of decreased activity or mobility, weight loss, or other risk factors not considered in their analysis.

Soon after this study1 was published, Lewiecki et al¹⁰ and others^11–13 published critical commentaries addressing controversial issues surrounding the study. They highlighted the importance of considering clinical risk factors for fracture in addition to the femoral neck and total hip T scores. In response to these comments, Gourlay et al clarified that their results were not generalizable to patients with secondary osteoporosis, such as those taking glucocorticoids or those who have rheumatic diseases.¹⁴

Readers should keep in mind that clinical risk factors make independent contributions to fracture risk (Figure 1).¹⁵

Readers should also recognize the following groups in whom the results of the study by Gourlay et al are not applicable since they were not included in their study:

• Men
• Women other than white women
• Women already diagnosed with osteoporosis and on bisphosphonates or any other osteoporosis treatment (except for estrogen). The findings also do not apply to:
• Patients who experience a significant decline in health status or who develop new clinical conditions (such as hyperparathyroidism, paraproteinemias, or type 2 diabetes) or who use medications such as glucocorticoids that cause rapid bone loss. Changes in clinical situations such as these may necessitate more frequent bone mineral density testing in spite of a “good” baseline T score.
• Perimenopausal women or women who received their first bone mineral density test before age 65. Perimenopause and menopause may trigger rapid bone loss, which may be as much as one T-score point (ie, 1 standard deviation) at the spine and femoral neck.¹⁶ Therefore, testing done during this time cannot be used as the basis of future monitoring.

Gourlay et al¹ did not take into account asymptomatic spinal fractures; they used only clinical vertebral fractures in their risk estimates of spinal fractures. Ascertainment of morphometric spinal fractures may be methodologically challenging, but if the study had included these fractures, the outcomes and conclusions could have been very different.

Vertebral fractures are present in as many as 14% to 33% of postmenopausal women¹⁷ and indicate osteoporosis (regardless of the bone mineral density). Moreover, most vertebral fractures are clinically silent and escape detection, and approximately only one in three radiographically defined vertebral fractures is reported clinically.^18,19 Given the prevalence of these fractures, we and others¹⁰ have noted that the results of the Gourlay study may be biased toward longer screening intervals because they did not account for morphometric vertebral fractures.

Gourlay et al used T scores only of the femoral neck and total hip and not those of the lumbar spine. Some studies have found that hip measurements may be superior to spine measurements for overall osteoporotic fracture prediction.^20,21 However, lumbar spine bone mineral density is predictive of fracture at other skeletal sites,^22,23 is a widely accepted skeletal site measurement, and is used to diagnose osteoporosis. Moreover, the lumbar spine T score can be −2.5 or higher even if the total hip or femoral neck T score is lower than −2.5.

More fractures occur in people with osteopenia than with osteoporosis

Osteoporosis imparts a much higher risk of fracture than does osteopenia. However, if one recognizes the much greater prevalence of osteopenia (33.6 million people) compared with osteoporosis (10 million),² it is not hard to appreciate that the number of fractures is higher in the osteopenic group than in those with osteoporosis based on T scores. Siris et al²⁴ point out that at least half of osteoporotic fractures are in patients with osteopenia, who comprise a larger segment of the population than those with osteoporosis.

Some clinical trials have shown that bisphosphonates are not effective in preventing clinical fractures in women who do not have osteoporosis.^25,26 However, clinicians must recognize that while bisphosphonates may not be as effective in preventing fractures in the osteopenic group with no other clinical risk factors, the presence of multiple clinical risk factors incrementally increases the fracture risk (which can be assessed via FRAX) and may require starting drug therapy earlier.

Women with vertebral fractures are considered to have clinical osteoporosis even if they have T scores in the osteopenic range, and must be considered for drug therapy.

The public health burden of fractures will not decrease unless individuals with low bone mineral density who are at an increased risk of fracture are identified and treated.²⁴

Is DXA testing overused or underused? does it decrease the rate of fractures?

The study of Gourlay et al¹ captured a lot of media attention, with many newspapers and blogs claiming that women may be getting tested too often.^27,28 However, in reality, this test is highly underutilized. The 2011 Healthcare Effectiveness Data and Information Set report noted that 71.0% of women in Medicare health maintenance organizations and 75.0% of women in Medicare preferred provider organizations ever had a bone mineral density test for osteoporosis.²⁹ While these numbers may not appear to be too far from the target, they are a poor gauge of DXA use as they include all types of bone mineral density tests in a woman’s lifetime, including even heel tests at health fairs.

Central DXA is used far less than one might expect. King and Fiorentino, in a recent analysis, showed that only about 14% of fee-for-service Medicare beneficiaries 65 years and older had one or more DXA tests in 2010.³⁰ DXA retesting also does not seem to be an issue, with only 1 in 10 elderly women reporting having had a repeat test at 2-year intervals, and fewer than 1 in 100 tested reported testing more frequently.³⁰

Some members of the public may have noticed the conclusions of a recent study¹ that said that if an older postmenopausal woman has her bone mineral density measured to screen for osteoporosis and has a normal or only mildly low result, she does not need to come back for another measurement for approximately 15 years.

See related editorial

We believe this interpretation of the study’s findings is overly simplistic and may have the unfortunate result of causing some people to neglect their bone health. Moreover, the study looked mainly at baseline T scores as the determinant of the subsequent screening interval. However, clinicians must carefully consider a variety of clinical risk factors when deciding how often to obtain bone mineral density measurements. The ultimate goal is to not miss the window of opportunity for early detection and treatment when it would matter the most (ie, before fractures develop).

Here, we will review this recent study, its findings, and its implications.

OSTEOPOROSIS POSES AN ENORMOUS PUBLIC HEALTH PROBLEM

If we consider only the hip, an estimated 10 million people in the United States have osteoporosis (T score ≤ −2.5 or a preexisting fragility fracture), and 33.6 million have osteopenia (T score −1.01 to −2.49).² The number of people with osteopenia can be assumed to be much higher if other skeletal sites are considered.

By increasing the risk of fragility fractures, osteoporosis poses an enormous public health problem. The surgeon general’s report points out that one of every two white women over age 50 will experience an osteoporosis-related fracture in her lifetime.³ Of all osteoporosis-related fractures, those of the hip carry the worse clinical outcome. Approximately one in five elderly people who experience an osteoporosis-related hip fracture need long-term nursing home care, and as many as 20% die within 1 year.³

In recognition of the burden of osteoporosis, the US Preventive Services Task Force (USPSTF)⁴ and other scientific bodies^2,3 recommend an initial bone mineral density test for all women age 65 and older. Dual-energy x-ray absorptiometry (DXA) is considered the gold standard for bone mineral density testing. Although the patient population that should receive an initial bone mineral density test has been clearly identified (see below), guidelines on the optimal frequency of testing do not exist, as data have been lacking. Recognizing this knowledge gap, Gourlay et al¹ attempted to answer the question of how often elderly postmenopausal women should be retested.

WHEN DO 10% OF ELDERLY POSTMENOPAUSAL WOMEN REACH A T SCORE OF −2.5?

Gourlay et al¹ analyzed data from 4,957 women in the Study of Osteoporotic Fractures. These women were predominantly white, were at least 67 years old and ambulatory, and had normal bone mineral density or osteopenia and no history of hip or clinical vertebral fracture at baseline. They had been recruited between 1986 and 1988 at sites in Baltimore, MD, Minneapolis, MN, the Monongahela Valley near Pittsburgh, PA, and Portland, OR.

DXA of the hip had been performed at baseline and at multiple times thereafter. The average follow-up time was 8 years.

The primary outcome measured was how long it took for 10% of the patients to reach a T score of −2.5 or less at the femoral neck or total hip as they progressed from having normal bone mineral density to osteoporosis or from osteopenia to osteoporosis and before they developed a fracture or needed treatment for osteoporosis.

Clinical risk factors such as age, body mass index, estrogen use at baseline, fracture after age 50, current smoking, current or past use of glucocorticoids, and self-reported rheumatoid arthritis were included as covariates in time-to-event analyses.

ANSWER: 16.8 YEARS (IF NORMAL AT BASELINE)

The authors estimated that 10% of women would make the transition to osteoporosis before having a hip or clinical vertebral fracture in the following intervals:

• 16.8 years in women whose bone mineral density was normal at baseline (T score at femoral neck and total hip of −1.00 or higher)
• 17.3 years in women who had mild osteopenia at baseline (T score −1.01 to −1.49)
• 4.7 years in women with moderate osteopenia at baseline (T score −1.5 to −1.99)
• 1.1 years in women with advanced osteopenia at baseline (T score −2.00 to −2.49).

The authors also found that body mass index and current estrogen use were the only clinical risk factors that influenced these intervals; other clinical factors such as a fracture after age 50, current smoking, previous or current use of oral glucocorticoids, and self-reported rheumatoid arthritis did not.

They concluded that osteoporosis would develop in fewer than 10% of women if the rescreening interval was lengthened to 15 years for women with normal density or mild osteopenia, 5 years for women with moderate osteopenia, and 1 year for women with advanced osteopenia.

WHAT DOES THIS MEAN FOR THE PRACTICING CLINICIAN?

Who needs an initial DXA test according to current guidelines?

The USPSTF,⁴ the National Osteoporosis Foundation (NOF),⁵ the International Society for Clinical Densitometry (ISCD),⁶ and the American Association of Clinical Endocrinologists (AACE)⁷ propose that the following groups should undergo DXA:

• All women age 65 and older
• All postmenopausal women who have had a fragility fracture or who have one or more risk factors for osteoporosis (height loss, body mass index < 20 kg/m², family history of osteoporosis, active smoking, excessive alcohol consumption)
• Adults who have a condition (eg, rheumatoid arthritis) or are taking a medication (eg, glucocorticoids in a daily dose ≥ 5 mg of prednisone or its equivalent for ≥ 3 months) associated with low bone mass or bone loss
• Anyone being considered for drug therapy for osteoporosis, discontinuing therapy for osteoporosis (including estrogen), or being treated for osteoporosis, to monitor the effect of treatment.

Assessing fracture risk. Although clinicians have traditionally relied on bone mineral density obtained by DXA to estimate fracture risk, the World Health Organization has developed a computer-based algorithm that calculates an individual’s 10-year fracture probability from easily obtained clinical risk factors with or without adding femoral-neck bone mineral density. The Fracture Risk Assessment tool, or FRAX, has attracted intense interest since its introduction in 2007 and has been endorsed by the USPSTF⁴ and by other scientific societies, including the NOF⁵ and the ISCD.⁸ In fact, the most recent USPSTF guidelines,⁴ which recommend screening all women age 65 and older, call for using FRAX to identify younger women at higher risk of fracture.

According to FRAX, a 65-year-old white woman who has no risk factors has a 9.3% chance of developing a major osteoporotic fracture in the next 10 years. And if a younger woman (between the ages of 50 and 64) has a fracture risk as high or higher than a 65-year-old white woman who has no risk factors, then she too should be screened by DXA.

The FRAX calculator is available online at www.shef.ac.uk/FRAX.

What are the current recommendations about follow-up DXA testing?

In eligible patients, the Centers for Medicare and Medicaid Services will pay for a DXA scan every 2 years. This interval is based on the concept that in an otherwise healthy person, it takes a minimum of 2 years to see a significant change in bone mineral density that can be attributed to a biological change in the bone and not just chance. The USPSTF⁴ and scientific societies such as the NOF⁵ generally agree with the Medicare guidelines of retesting every 2 years but recognize certain clinical situations that may warrant more frequent retesting (see below).

But the real question is how long the DXA screening interval can be extended so that meaningful information can still be obtained to help make management decisions and before a complication such as a fracture occurs. While there is convincing evidence to support the recommendations for an initial DXA test, data to answer the question of how long the resting interval should be are lacking.

Before the study by Gourlay et al,¹ the only data on repeat DXA came from work by Hillier et al.⁹ But those investigators asked a different question. They were interested in how well repeated measurements predicted fractures. They used the same population that Gourlay et al did but evaluated fractures, not T scores. They concluded that in healthy, adult postmenopausal women, repeating the bone mineral density measurement up to 8 years later adds little value to initial measurement for predicting incident fractures.

Clinical factors also count

The T score should not be the only major factor determining the interval for bone mineral density testing in elderly women; clinical risk factors also should be kept in mind.

Gourlay et al concluded that age and T scores are the key predictive factors in determining the bone mineral density testing interval in elderly, postmenopausal women for screening purposes.¹ In their statistical model, clinical risk factors such as fracture after age 50, current smoking, previous or current use of glucocorticoids, and self-reported rheumatoid arthritis did not influence the testing interval. They say that clinicians should not feel compelled to shorten the testing interval when these risk factors are present.

Readers may take this to mean that if these results were strictly applied to a 70-year-old white woman receiving oral glucocorticoids for rheumatoid arthritis and who has a baseline T score of −1.45, then her next test may be postponed by 15 years (given that both these factors did not influence the testing interval). Readers may also conclude that if this patient’s T score were −1.51, then her screening interval would be 5 years and not 15 years.

However, Gourlay et al say¹ that clinicians can choose to shorten the testing interval if there is evidence of decreased activity or mobility, weight loss, or other risk factors not considered in their analysis.

Soon after this study1 was published, Lewiecki et al¹⁰ and others^11–13 published critical commentaries addressing controversial issues surrounding the study. They highlighted the importance of considering clinical risk factors for fracture in addition to the femoral neck and total hip T scores. In response to these comments, Gourlay et al clarified that their results were not generalizable to patients with secondary osteoporosis, such as those taking glucocorticoids or those who have rheumatic diseases.¹⁴

Readers should keep in mind that clinical risk factors make independent contributions to fracture risk (Figure 1).¹⁵

Readers should also recognize the following groups in whom the results of the study by Gourlay et al are not applicable since they were not included in their study:

• Men
• Women other than white women
• Women already diagnosed with osteoporosis and on bisphosphonates or any other osteoporosis treatment (except for estrogen). The findings also do not apply to:
• Patients who experience a significant decline in health status or who develop new clinical conditions (such as hyperparathyroidism, paraproteinemias, or type 2 diabetes) or who use medications such as glucocorticoids that cause rapid bone loss. Changes in clinical situations such as these may necessitate more frequent bone mineral density testing in spite of a “good” baseline T score.
• Perimenopausal women or women who received their first bone mineral density test before age 65. Perimenopause and menopause may trigger rapid bone loss, which may be as much as one T-score point (ie, 1 standard deviation) at the spine and femoral neck.¹⁶ Therefore, testing done during this time cannot be used as the basis of future monitoring.

Gourlay et al¹ did not take into account asymptomatic spinal fractures; they used only clinical vertebral fractures in their risk estimates of spinal fractures. Ascertainment of morphometric spinal fractures may be methodologically challenging, but if the study had included these fractures, the outcomes and conclusions could have been very different.

Vertebral fractures are present in as many as 14% to 33% of postmenopausal women¹⁷ and indicate osteoporosis (regardless of the bone mineral density). Moreover, most vertebral fractures are clinically silent and escape detection, and approximately only one in three radiographically defined vertebral fractures is reported clinically.^18,19 Given the prevalence of these fractures, we and others¹⁰ have noted that the results of the Gourlay study may be biased toward longer screening intervals because they did not account for morphometric vertebral fractures.

Gourlay et al used T scores only of the femoral neck and total hip and not those of the lumbar spine. Some studies have found that hip measurements may be superior to spine measurements for overall osteoporotic fracture prediction.^20,21 However, lumbar spine bone mineral density is predictive of fracture at other skeletal sites,^22,23 is a widely accepted skeletal site measurement, and is used to diagnose osteoporosis. Moreover, the lumbar spine T score can be −2.5 or higher even if the total hip or femoral neck T score is lower than −2.5.

More fractures occur in people with osteopenia than with osteoporosis

Osteoporosis imparts a much higher risk of fracture than does osteopenia. However, if one recognizes the much greater prevalence of osteopenia (33.6 million people) compared with osteoporosis (10 million),² it is not hard to appreciate that the number of fractures is higher in the osteopenic group than in those with osteoporosis based on T scores. Siris et al²⁴ point out that at least half of osteoporotic fractures are in patients with osteopenia, who comprise a larger segment of the population than those with osteoporosis.

Some clinical trials have shown that bisphosphonates are not effective in preventing clinical fractures in women who do not have osteoporosis.^25,26 However, clinicians must recognize that while bisphosphonates may not be as effective in preventing fractures in the osteopenic group with no other clinical risk factors, the presence of multiple clinical risk factors incrementally increases the fracture risk (which can be assessed via FRAX) and may require starting drug therapy earlier.

Women with vertebral fractures are considered to have clinical osteoporosis even if they have T scores in the osteopenic range, and must be considered for drug therapy.

The public health burden of fractures will not decrease unless individuals with low bone mineral density who are at an increased risk of fracture are identified and treated.²⁴

Is DXA testing overused or underused? does it decrease the rate of fractures?

The study of Gourlay et al¹ captured a lot of media attention, with many newspapers and blogs claiming that women may be getting tested too often.^27,28 However, in reality, this test is highly underutilized. The 2011 Healthcare Effectiveness Data and Information Set report noted that 71.0% of women in Medicare health maintenance organizations and 75.0% of women in Medicare preferred provider organizations ever had a bone mineral density test for osteoporosis.²⁹ While these numbers may not appear to be too far from the target, they are a poor gauge of DXA use as they include all types of bone mineral density tests in a woman’s lifetime, including even heel tests at health fairs.

Central DXA is used far less than one might expect. King and Fiorentino, in a recent analysis, showed that only about 14% of fee-for-service Medicare beneficiaries 65 years and older had one or more DXA tests in 2010.³⁰ DXA retesting also does not seem to be an issue, with only 1 in 10 elderly women reporting having had a repeat test at 2-year intervals, and fewer than 1 in 100 tested reported testing more frequently.³⁰

Some members of the public may have noticed the conclusions of a recent study¹ that said that if an older postmenopausal woman has her bone mineral density measured to screen for osteoporosis and has a normal or only mildly low result, she does not need to come back for another measurement for approximately 15 years.

See related editorial

We believe this interpretation of the study’s findings is overly simplistic and may have the unfortunate result of causing some people to neglect their bone health. Moreover, the study looked mainly at baseline T scores as the determinant of the subsequent screening interval. However, clinicians must carefully consider a variety of clinical risk factors when deciding how often to obtain bone mineral density measurements. The ultimate goal is to not miss the window of opportunity for early detection and treatment when it would matter the most (ie, before fractures develop).

Here, we will review this recent study, its findings, and its implications.

OSTEOPOROSIS POSES AN ENORMOUS PUBLIC HEALTH PROBLEM

If we consider only the hip, an estimated 10 million people in the United States have osteoporosis (T score ≤ −2.5 or a preexisting fragility fracture), and 33.6 million have osteopenia (T score −1.01 to −2.49).² The number of people with osteopenia can be assumed to be much higher if other skeletal sites are considered.

By increasing the risk of fragility fractures, osteoporosis poses an enormous public health problem. The surgeon general’s report points out that one of every two white women over age 50 will experience an osteoporosis-related fracture in her lifetime.³ Of all osteoporosis-related fractures, those of the hip carry the worse clinical outcome. Approximately one in five elderly people who experience an osteoporosis-related hip fracture need long-term nursing home care, and as many as 20% die within 1 year.³

In recognition of the burden of osteoporosis, the US Preventive Services Task Force (USPSTF)⁴ and other scientific bodies^2,3 recommend an initial bone mineral density test for all women age 65 and older. Dual-energy x-ray absorptiometry (DXA) is considered the gold standard for bone mineral density testing. Although the patient population that should receive an initial bone mineral density test has been clearly identified (see below), guidelines on the optimal frequency of testing do not exist, as data have been lacking. Recognizing this knowledge gap, Gourlay et al¹ attempted to answer the question of how often elderly postmenopausal women should be retested.

WHEN DO 10% OF ELDERLY POSTMENOPAUSAL WOMEN REACH A T SCORE OF −2.5?

Gourlay et al¹ analyzed data from 4,957 women in the Study of Osteoporotic Fractures. These women were predominantly white, were at least 67 years old and ambulatory, and had normal bone mineral density or osteopenia and no history of hip or clinical vertebral fracture at baseline. They had been recruited between 1986 and 1988 at sites in Baltimore, MD, Minneapolis, MN, the Monongahela Valley near Pittsburgh, PA, and Portland, OR.

DXA of the hip had been performed at baseline and at multiple times thereafter. The average follow-up time was 8 years.

The primary outcome measured was how long it took for 10% of the patients to reach a T score of −2.5 or less at the femoral neck or total hip as they progressed from having normal bone mineral density to osteoporosis or from osteopenia to osteoporosis and before they developed a fracture or needed treatment for osteoporosis.

Clinical risk factors such as age, body mass index, estrogen use at baseline, fracture after age 50, current smoking, current or past use of glucocorticoids, and self-reported rheumatoid arthritis were included as covariates in time-to-event analyses.

ANSWER: 16.8 YEARS (IF NORMAL AT BASELINE)

The authors estimated that 10% of women would make the transition to osteoporosis before having a hip or clinical vertebral fracture in the following intervals:

• 16.8 years in women whose bone mineral density was normal at baseline (T score at femoral neck and total hip of −1.00 or higher)
• 17.3 years in women who had mild osteopenia at baseline (T score −1.01 to −1.49)
• 4.7 years in women with moderate osteopenia at baseline (T score −1.5 to −1.99)
• 1.1 years in women with advanced osteopenia at baseline (T score −2.00 to −2.49).

The authors also found that body mass index and current estrogen use were the only clinical risk factors that influenced these intervals; other clinical factors such as a fracture after age 50, current smoking, previous or current use of oral glucocorticoids, and self-reported rheumatoid arthritis did not.

They concluded that osteoporosis would develop in fewer than 10% of women if the rescreening interval was lengthened to 15 years for women with normal density or mild osteopenia, 5 years for women with moderate osteopenia, and 1 year for women with advanced osteopenia.

WHAT DOES THIS MEAN FOR THE PRACTICING CLINICIAN?

Who needs an initial DXA test according to current guidelines?

The USPSTF,⁴ the National Osteoporosis Foundation (NOF),⁵ the International Society for Clinical Densitometry (ISCD),⁶ and the American Association of Clinical Endocrinologists (AACE)⁷ propose that the following groups should undergo DXA:

• All women age 65 and older
• All postmenopausal women who have had a fragility fracture or who have one or more risk factors for osteoporosis (height loss, body mass index < 20 kg/m², family history of osteoporosis, active smoking, excessive alcohol consumption)
• Adults who have a condition (eg, rheumatoid arthritis) or are taking a medication (eg, glucocorticoids in a daily dose ≥ 5 mg of prednisone or its equivalent for ≥ 3 months) associated with low bone mass or bone loss
• Anyone being considered for drug therapy for osteoporosis, discontinuing therapy for osteoporosis (including estrogen), or being treated for osteoporosis, to monitor the effect of treatment.

Assessing fracture risk. Although clinicians have traditionally relied on bone mineral density obtained by DXA to estimate fracture risk, the World Health Organization has developed a computer-based algorithm that calculates an individual’s 10-year fracture probability from easily obtained clinical risk factors with or without adding femoral-neck bone mineral density. The Fracture Risk Assessment tool, or FRAX, has attracted intense interest since its introduction in 2007 and has been endorsed by the USPSTF⁴ and by other scientific societies, including the NOF⁵ and the ISCD.⁸ In fact, the most recent USPSTF guidelines,⁴ which recommend screening all women age 65 and older, call for using FRAX to identify younger women at higher risk of fracture.

According to FRAX, a 65-year-old white woman who has no risk factors has a 9.3% chance of developing a major osteoporotic fracture in the next 10 years. And if a younger woman (between the ages of 50 and 64) has a fracture risk as high or higher than a 65-year-old white woman who has no risk factors, then she too should be screened by DXA.

The FRAX calculator is available online at www.shef.ac.uk/FRAX.

What are the current recommendations about follow-up DXA testing?

In eligible patients, the Centers for Medicare and Medicaid Services will pay for a DXA scan every 2 years. This interval is based on the concept that in an otherwise healthy person, it takes a minimum of 2 years to see a significant change in bone mineral density that can be attributed to a biological change in the bone and not just chance. The USPSTF⁴ and scientific societies such as the NOF⁵ generally agree with the Medicare guidelines of retesting every 2 years but recognize certain clinical situations that may warrant more frequent retesting (see below).

But the real question is how long the DXA screening interval can be extended so that meaningful information can still be obtained to help make management decisions and before a complication such as a fracture occurs. While there is convincing evidence to support the recommendations for an initial DXA test, data to answer the question of how long the resting interval should be are lacking.

Before the study by Gourlay et al,¹ the only data on repeat DXA came from work by Hillier et al.⁹ But those investigators asked a different question. They were interested in how well repeated measurements predicted fractures. They used the same population that Gourlay et al did but evaluated fractures, not T scores. They concluded that in healthy, adult postmenopausal women, repeating the bone mineral density measurement up to 8 years later adds little value to initial measurement for predicting incident fractures.

Clinical factors also count

The T score should not be the only major factor determining the interval for bone mineral density testing in elderly women; clinical risk factors also should be kept in mind.

Gourlay et al concluded that age and T scores are the key predictive factors in determining the bone mineral density testing interval in elderly, postmenopausal women for screening purposes.¹ In their statistical model, clinical risk factors such as fracture after age 50, current smoking, previous or current use of glucocorticoids, and self-reported rheumatoid arthritis did not influence the testing interval. They say that clinicians should not feel compelled to shorten the testing interval when these risk factors are present.

Readers may take this to mean that if these results were strictly applied to a 70-year-old white woman receiving oral glucocorticoids for rheumatoid arthritis and who has a baseline T score of −1.45, then her next test may be postponed by 15 years (given that both these factors did not influence the testing interval). Readers may also conclude that if this patient’s T score were −1.51, then her screening interval would be 5 years and not 15 years.

However, Gourlay et al say¹ that clinicians can choose to shorten the testing interval if there is evidence of decreased activity or mobility, weight loss, or other risk factors not considered in their analysis.

Soon after this study1 was published, Lewiecki et al¹⁰ and others^11–13 published critical commentaries addressing controversial issues surrounding the study. They highlighted the importance of considering clinical risk factors for fracture in addition to the femoral neck and total hip T scores. In response to these comments, Gourlay et al clarified that their results were not generalizable to patients with secondary osteoporosis, such as those taking glucocorticoids or those who have rheumatic diseases.¹⁴

Readers should keep in mind that clinical risk factors make independent contributions to fracture risk (Figure 1).¹⁵

Readers should also recognize the following groups in whom the results of the study by Gourlay et al are not applicable since they were not included in their study:

• Men
• Women other than white women
• Women already diagnosed with osteoporosis and on bisphosphonates or any other osteoporosis treatment (except for estrogen). The findings also do not apply to:
• Patients who experience a significant decline in health status or who develop new clinical conditions (such as hyperparathyroidism, paraproteinemias, or type 2 diabetes) or who use medications such as glucocorticoids that cause rapid bone loss. Changes in clinical situations such as these may necessitate more frequent bone mineral density testing in spite of a “good” baseline T score.
• Perimenopausal women or women who received their first bone mineral density test before age 65. Perimenopause and menopause may trigger rapid bone loss, which may be as much as one T-score point (ie, 1 standard deviation) at the spine and femoral neck.¹⁶ Therefore, testing done during this time cannot be used as the basis of future monitoring.

Gourlay et al¹ did not take into account asymptomatic spinal fractures; they used only clinical vertebral fractures in their risk estimates of spinal fractures. Ascertainment of morphometric spinal fractures may be methodologically challenging, but if the study had included these fractures, the outcomes and conclusions could have been very different.

Vertebral fractures are present in as many as 14% to 33% of postmenopausal women¹⁷ and indicate osteoporosis (regardless of the bone mineral density). Moreover, most vertebral fractures are clinically silent and escape detection, and approximately only one in three radiographically defined vertebral fractures is reported clinically.^18,19 Given the prevalence of these fractures, we and others¹⁰ have noted that the results of the Gourlay study may be biased toward longer screening intervals because they did not account for morphometric vertebral fractures.

Gourlay et al used T scores only of the femoral neck and total hip and not those of the lumbar spine. Some studies have found that hip measurements may be superior to spine measurements for overall osteoporotic fracture prediction.^20,21 However, lumbar spine bone mineral density is predictive of fracture at other skeletal sites,^22,23 is a widely accepted skeletal site measurement, and is used to diagnose osteoporosis. Moreover, the lumbar spine T score can be −2.5 or higher even if the total hip or femoral neck T score is lower than −2.5.

More fractures occur in people with osteopenia than with osteoporosis

Osteoporosis imparts a much higher risk of fracture than does osteopenia. However, if one recognizes the much greater prevalence of osteopenia (33.6 million people) compared with osteoporosis (10 million),² it is not hard to appreciate that the number of fractures is higher in the osteopenic group than in those with osteoporosis based on T scores. Siris et al²⁴ point out that at least half of osteoporotic fractures are in patients with osteopenia, who comprise a larger segment of the population than those with osteoporosis.

Some clinical trials have shown that bisphosphonates are not effective in preventing clinical fractures in women who do not have osteoporosis.^25,26 However, clinicians must recognize that while bisphosphonates may not be as effective in preventing fractures in the osteopenic group with no other clinical risk factors, the presence of multiple clinical risk factors incrementally increases the fracture risk (which can be assessed via FRAX) and may require starting drug therapy earlier.

Women with vertebral fractures are considered to have clinical osteoporosis even if they have T scores in the osteopenic range, and must be considered for drug therapy.

The public health burden of fractures will not decrease unless individuals with low bone mineral density who are at an increased risk of fracture are identified and treated.²⁴

Is DXA testing overused or underused? does it decrease the rate of fractures?

The study of Gourlay et al¹ captured a lot of media attention, with many newspapers and blogs claiming that women may be getting tested too often.^27,28 However, in reality, this test is highly underutilized. The 2011 Healthcare Effectiveness Data and Information Set report noted that 71.0% of women in Medicare health maintenance organizations and 75.0% of women in Medicare preferred provider organizations ever had a bone mineral density test for osteoporosis.²⁹ While these numbers may not appear to be too far from the target, they are a poor gauge of DXA use as they include all types of bone mineral density tests in a woman’s lifetime, including even heel tests at health fairs.

Central DXA is used far less than one might expect. King and Fiorentino, in a recent analysis, showed that only about 14% of fee-for-service Medicare beneficiaries 65 years and older had one or more DXA tests in 2010.³⁰ DXA retesting also does not seem to be an issue, with only 1 in 10 elderly women reporting having had a repeat test at 2-year intervals, and fewer than 1 in 100 tested reported testing more frequently.³⁰

The recent publication of the AIM-HIGH trial (Atherothrombosis Intervention in Metabolic Syndrome With Low HDL/High Triglycerides: Impact on Global Health Outcomes)¹ has thrown the use of niacin as a lipid-modifying therapy into question. The trial was stopped early because an interim analysis found that the patients who took extended-release niacin had no clinical benefit. In addition, it found a trend toward more ischemic strokes, though this finding was later found not to be statistically significant.

Complicating the interpretation, while both the treatment group and the control group in the study received statin therapy, the researchers attempted to keep low-density lipoprotein cholesterol (LDL-C) levels equal, meaning that patients in the control group received more intensive statin therapy than those in the treatment group. And the placebo that the control patients received was actually a low dose of niacin, to induce flushing and thus to blind study participants and their physicians to which drug they were taking.

In the article that follows, I will explore the background, design, findings, and implications of this key trial and try to untangle the many questions about how to interpret it.

LOWERING LDL-C REDUCES RISK, BUT DOES NOT ELIMINATE IT

Large randomized controlled trials have consistently shown that lowering the level of LDL-C reduces cardiovascular event rates by 25% to 45% both in people who are known to have coronary artery disease and in those who are not.^2–4 As a result, guidelines for preventing cardiovascular disease have increasingly emphasized maintaining low LDL-C levels. This has led to a proliferation in the use of inhibitors of 3-hydroxy-3-methylglutaryl coenzyme A reductase (statins) in patients at high cardiovascular risk.

However, these agents only reduce the risk—they do not eliminate it. Needed are additional therapies to complement existing LDL-C-lowering approaches to lower the cardiovascular risk even further.

Raising HDL-C: The next frontier

One such strategy for further lowering cardiovascular risk that has received considerable interest is to promote the biological activity of the “good” cholesterol.

Studies have consistently shown that the higher the plasma level of high-density lipoprotein cholesterol (HDL-C), the lower the risk of cardiovascular events, suggesting that raising HDL-C may be beneficial.⁵ Studies in animals with atherosclerosis show that raising HDL-C via genetic modification of the animal or direct infusion of the molecule has a favorable impact on both the size and the structure of experimental plaque.^6,7

Accordingly, much activity has focused on developing new therapies that raise HDL-C more effectively than current ones.

Why niacin should protect the heart

For more than 50 years, niacin has been used to manage dyslipidemia.

In addition to raising HDL-C levels more effectively than any other agent available today, niacin also lowers the levels of LDL-C, triglycerides, and lipoprotein (a).⁸ Before statins were available, the Coronary Drug Project found that niacin reduced the rate of nonfatal myocardial infarction and the 15-year mortality rate.⁹ In addition, niacin has been shown to slow the progression of carotid intimal-medial thickness and coronary atherosclerosis, and even to reverse these processes in some trials.^10–12

However, a number of issues remain about using niacin to prevent cardiovascular events. Nearly all patients who take it experience flushing, which limits its tolerability and, thus, our ability to titrate doses to levels needed for adequate lipid changes. While a number of modifications of niacin administration have been developed (eg, extended-release formulations and products that inhibit flushing), no large study has tested the clinical efficacy of these strategies. Furthermore, until AIM-HIGH, no large-scale trial had directly evaluated the impact of niacin therapy on a background of statin therapy.

AIM-HIGH STUDY DESIGN

The intent of the AIM-HIGH trial was to determine whether extended-release niacin (Niaspan) would reduce the risk of cardiovascular events when added to therapy with a statin—in this case, simvastatin (Zocor) supplemented with ezetimibe (Zetia).¹

The trial was funded by the National Heart, Lung, and Blood Institute (NHLBI) and by Abbott Laboratories, which also supplied the extended-release niacin and the ezetimibe. Merck donated the simvastatin.

Patient characteristics

The patients were all at least 45 years of age with established, stable coronary heart disease, cerebrovascular or carotid arterial disease, or peripheral arterial disease. They also had to have low levels of HDL-C (< 40 mg/dL in men, < 50 mg/dL in women), elevated triglycerides (150–400 mg/dL), and LDL-C levels lower than 180 mg/dL if they were not taking a statin at entry.

The mean age of the patients was 64 years, 85% were men, and 92% were white. They had a high prevalence of cardiovascular risk factors: 34% had diabetes, 71% had hypertension, and 81% had metabolic syndrome. Nearly all (94%) of the patients were taking a statin at entry; 76% had been taking one for more than 1 year, and 40% had been taking one for more than 5 years.¹

Simvastatin, ezetimibe, and either niacin or placebo

All lipid-modifying agents except statins and ezetimibe were stopped for least 4 weeks after enrollment.

All patients then entered a 4- to 8-week open-label period, during which they took simvastatin 40 mg daily and extended-release niacin starting at 500 mg and increased weekly up to 2,000 mg daily. Patients who could tolerate at least 1,500 mg daily were randomly assigned to treatment with either niacin 1,500 to 2,000 mg or matching placebo. Both groups continued to receive simvastatin. The placebo contained a small dose of immediate-release niacin (50 mg) in each tablet to induce flushing and to maintain blinding of treatment.

Given that niacin also lowers LDL-C, an algorithm was used to try to keep LDL-C levels roughly the same in both treatment groups. This involved adjusting the simvastatin dose and permitting the use of ezetimibe 10 mg to keep the LDL-C level between 40 and 80 mg/dL. Accordingly, participating physicians were told their patients’ LDL-C levels but were blinded to their HDL-C and triglyceride levels throughout the study.

Every 6 months, patients had a follow-up visit in the clinic, and midway through each 6-month interval they received a phone call from the investigators.¹

AIM-HIGH end points

The primary end point was the composite of the first event of death due to coronary heart disease, nonfatal myocardial infarction, ischemic stroke, hospitalization for acute coronary syndrome, or symptom-driven revascularization of the coronary or cerebral arteries.

Secondary end points were:

• Death from coronary heart disease, nonfatal myocardial infarction, ischemic stroke, or hospitalization for acute coronary syndrome
• Death from coronary heart disease, nonfatal myocardial infarction, or ischemic stroke
• Death from cardiovascular causes.

Tertiary end points included:

• Death from any cause
• Individual components of the primary end point
• Prespecified subgroups according to sex, history or no history of diabetes, and presence or absence of the metabolic syndrome.¹

All clinical events were adjudicated by a central committee.

STUDY HALTED EARLY

The study was planned to run for a mean of 4.6 years, during which 800 primary end point events were expected. With these numbers, the investigators calculated that the study had 85% power to detect a 25% reduction in the primary end point, at a one-sided alpha level of 0.025.

The plan called for an interim analysis when 50% of the anticipated events had occurred, with prespecified stopping boundaries based on either efficacy or futility. The boundary for lack of efficacy required an observed hazard ratio of at least 1.02 with a probability of less than .001.

In the interim analysis, after a median follow-up of only 3 years, the data and safety monitoring board recommended stopping the study early because the boundary for futility had been crossed and, unexpectedly, the rate of ischemic stroke was higher in the niacin-treated patients than in those receiving placebo.

MAJOR FINDINGS OF AIM-HIGH

Of 4,273 patients who began open-label treatment with niacin, 3,414 were randomized to treatment with niacin or placebo.¹

HDL-C levels went up in both groups

At 2 years:

• HDL-C levels had increased by 25.0% (to 42 mg/dL) in the niacin group and by 9.8% (to 38 mg/dL) in the placebo group
• Triglycerides had decreased by 28.6% with niacin and by 8.1% with placebo
• LDL-C had decreased by 12.0% with niacin and by 5.5% with placebo.

Patients in the placebo group were more likely to have subsequently received the maximum dose of simvastatin, ie, 80 mg/day (24.7% vs 17.5%), and to have received ezetimibe (21.5% vs 9.5%). More patients in the niacin group required either dose reduction of the study drug (6.3% vs 3.4%) or drug discontinuation (25.4% vs 20.1%).¹

No difference in the primary end point

There was no difference between the two treatment groups in the rate of the primary end point, which occurred in 282 (16.4%) of the 1,718 patients in the niacin group and 272 (16.2%) of the 1,696 patients in the placebo group (P = .79; hazard ratio 1.02, 95% confidence interval 0.87–1.21).¹

However, more patients in the niacin group than in the placebo group who reached the primary end point did so by having a first ischemic stroke: 27 patients (1.6%) vs 15 patients (0.9%). Eight of these patients, all in the niacin group, had their stroke between 2 months and 4 years after they had stopped taking the study drug.

Further analysis that included all ischemic strokes revealed the same trend: 29 vs 18 patients (P = .11).¹

No benefit was observed for niacin-treated patients in terms of any of the secondary or tertiary end points.

Subgroup analysis revealed no evidence of statistical heterogeneity: ie, niacin seemed to lack efficacy in all the prespecified subgroups studied (age 65 and older vs younger, men vs women, and those with or without diabetes, metabolic syndrome, prior myocardial infarction, or statin use at entry).

In general, niacin was well tolerated in the active-treatment group, with a low incidence of liver and muscle abnormalities.

PUTTING AIM-HIGH IN CONTEXT

How should practicing clinicians interpret these outcomes?

Ever since the NHLBI reported (in an urgent press release) that it was stopping the study early due to futility and a potential excess of strokes,¹³ there has been considerable debate as to which factors contributed to these outcomes. In the wake of the publication of more detailed information about the trial,¹ this debate is likely to continue.

The AIM-HIGH results can be interpreted in several ways:

• Perhaps niacin is no good as a preventive agent
• Perhaps raising HDL-C is flawed as a preventive strategy
• Perhaps AIM-HIGH had methodologic flaws, such as looking at the wrong patient cohort or using a treatment protocol that set itself up for failure
• Perhaps statins are so good that, once you prescribe one, anything else you give provides no additional benefit.

Which of these is correct?

Is niacin no good?

In its most simple form, AIM-HIGH has always been seen as a clinical trial of niacin. While the early trials of immediate-release niacin were encouraging in terms of its effects on lipids, atherosclerotic plaque, and cardiovascular outcomes, using it in clinical practice has always been challenging, largely because many patients cannot tolerate it in doses high enough to be effective. A number of developments have improved niacin’s tolerability, but its clinical impact in the statin era has not been evaluated.

Niacin’s lack of efficacy in this trial will ultimately be viewed as a failure of the drug itself, but is this the case?

AIM-HIGH was not simply a direct comparison of niacin vs placebo on top of standard medical practice. The investigators recognized that niacin has additional effects—in particular, lowering levels of atherogenic lipids—and they attempted to control for these effects by titrating the other LDL-C-lowering therapies during the study. As a result, the trial was actually a comparison between niacin plus low-dose simvastatin on the one hand, and placebo plus high-dose simvastatin (and, more often, also ezetimibe) on the other.

Furthermore, the placebo-treated patients received small doses of immediate-release niacin to induce flushing and maintain blinding. It is therefore hard to conclude that this clinical trial was a direct evaluation of the impact of niacin.

In contrast, the Heart Protection Study 2-Treatment of HDL to Reduce the Incidence of Vascular Events (HPS2-THRIVE) study is currently evaluating extended-release niacin in combination with laropiprant, a prostaglandin receptor antagonist, vs placebo in more than 24,000 statin-treated patients.¹⁴ Without any in-trial titration of lipids, this study provides a more direct comparison of the effects of niacin in the statin era.

Niacin continues to attract interest, largely because it can raise HDL-C by 20% to 30% when given at doses of 1,500 mg or more. Also, consistent observations from population studies of an inverse relationship between HDL-C levels and cardiovascular risk⁵ have stimulated interest in developing novel agents that substantially raise HDL-C.

The failure of HDL-C-raising therapies in clinical trials^15,16 has fueled concern that HDL may not be the magic elixir that many have sought. Given that niacin is the most effective HDL-C-raising agent currently available, its lack of efficacy in AIM-HIGH could be perceived as another nail in the coffin of the hypothesis that raising the HDL-C level with pharmacologic agents is beneficial.

AIM-HIGH was designed to examine the effects of raising HDL-C. To this end, it was performed exclusively in patients with low HDL-C levels, and the investigators tried to isolate the potential effects of raising HDL-C by equalizing the LDL-C levels in the treatment groups.

However, the HDL-C changes observed in AIM-HIGH are likely to have undermined the study objective. While niacin predictably increased HDL-C levels by 25%, an unexpected increase in HDL-C of 9.8% in the placebo-treated patients resulted in a difference in achieved HDL-C levels of only 4 mg/dL between the groups. This was far less than anticipated, and it likely had a major impact on an already underpowered study.

AIM-HIGH was designed to have 85% power to demonstrate a 25% reduction in clinical events, which was an optimistic estimate. On the basis of population studies, a difference of 4 mg/dL in HDL-C would be anticipated to result in no more than a 10% lower rate of clinical events, far beyond AIM-HIGH’s limit of detection.

The reasons for the increase in HDL-C in the placebo group are unknown, but they likely reflect the use of higher doses of simvastatin, some regression to the mean, and, possibly, the small doses of immediate-release niacin that the placebo contained. (Contrary to the belief of the investigators, there have been some reports of lipid changes with such doses,¹⁷ which may have contributed to the observed HDL-C-raising.)

Given that the HDL-C difference between the groups was relatively small and that niacin has additional effects beyond raising HDL-C and lowering LDL-C, it is unlikely that the futility of AIM-HIGH reflects a major indictment of HDL-C-raising. For the time being, the jury is still out on this question.

Was AIM-HIGH methodologically flawed?

A number of methodologic issues may have affected AIM-HIGH’s ability to adequately address its objectives.

The wrong cohort? In planning a study such as AIM-HIGH, the need for a relatively small sample size and the need to detect the greatest relative risk reduction with niacin would require enrollment of patients at the highest risk of cardiovascular events despite the use of statins. These needs were satisfied by only including patients who had atherosclerotic cardiovascular disease and low HDL-C levels. The inclusion of patients with low levels of HDL-C was also expected to promote greater increases in this lipid, and potentially event reduction, with niacin.

But no benefit was observed. It remains to be determined whether the inclusion of a high proportion of patients with the metabolic syndrome adversely affected the ability to detect a benefit with niacin. While post hoc analyses of studies of carotid intimal-medial thickness demonstrated no relationship between raising HDL-C with niacin and slowing of disease progression in patients with the metabolic syndrome,¹⁸ it remains to be determined whether this would translate to any effect on cardiovascular event rates.

Inadequate statistical power? An underpowered study would leave very little room for error, a pertinent point given the variability in therapeutic response in both actively treated and placebo-treated patients typically encountered in clinical trials. Giving low doses of immediate-release niacin and titrating the simvastatin dose to control LDL-C, resulting in imbalances in lipid-modifying therapies, represent additional flaws in the study design.

Stopped too soon? The early cessation of the study was somewhat questionable. The study crossed the prespecified boundary for lack of efficacy at the time of the interim analysis, and initial review by the data and safety monitoring board suggested an excess rate of ischemic stroke with niacin. The inclusion of this latter finding in the press release prompted considerable speculation regarding potential mechanisms and also concern among patients currently taking niacin. The subsequent finding that this signal was not statistically significant serves as an important warning for those conducting clinical trials not to prematurely overstate preliminary observations.

The implications for agents used in clinical practice are considerable: negative findings should not be overemphasized without robust evidence.

Do statins make everything else irrelevant?

The final factor to consider is the relative modifiability of residual clinical risk in statin-treated patients.

While residual risk is often cited as the reason to develop new antiatherosclerotic therapies, it is unknown how many of these ongoing events can be prevented. Several nonmodifiable factors such as age and concomitant disease are likely to contribute to these clinical events, which may limit our ability to further reduce event rates in patients who have already achieved low LDL-C levels with statin therapy. This may underscore the observation that no major clinical trial has demonstrated clinical benefit of an antiatherosclerotic agent on top of background medical care that included statins.

The finding that atherosclerosis continues to progress in many patients even though they take statins in high doses or achieve low LDL-C levels suggests that there is still room for improvement.

WHAT FUTURE FOR NIACIN?

So what does the future hold for niacin? The ongoing HPS2-THRIVE study provides another opportunity to evaluate the potential clinical efficacy of niacin in statin-treated patients. For now, we must wait for the results of this study.

In the meantime, it would seem reasonable to continue treatment with niacin in patients who need it for its multiple lipid-modifying effects. Whether clinicians will be less likely to initiate niacin therapy until there is clear evidence of clinical benefit remains uncertain. As for HDL-C, it remains to be determined whether any therapy targeting either quantitative or qualitative changes will be beneficial.

Over the last 3 decades, clinical trials have provided important insights into the prevention of cardiovascular events and have had a profound impact on clinical practice. Such studies simply evaluate whether one strategy is better or worse than the existing standard of care. They do not provide mechanistic insights, and when attempts have been made to address mechanisms in the study design, the trial, as in the case of AIM-HIGH, leaves more questions than answers.

Future trials will provide more clarity as to the optimal way to treat patients, but they must be based on a robust design that permits the study question to be adequately addressed.

The recent publication of the AIM-HIGH trial (Atherothrombosis Intervention in Metabolic Syndrome With Low HDL/High Triglycerides: Impact on Global Health Outcomes)¹ has thrown the use of niacin as a lipid-modifying therapy into question. The trial was stopped early because an interim analysis found that the patients who took extended-release niacin had no clinical benefit. In addition, it found a trend toward more ischemic strokes, though this finding was later found not to be statistically significant.

Complicating the interpretation, while both the treatment group and the control group in the study received statin therapy, the researchers attempted to keep low-density lipoprotein cholesterol (LDL-C) levels equal, meaning that patients in the control group received more intensive statin therapy than those in the treatment group. And the placebo that the control patients received was actually a low dose of niacin, to induce flushing and thus to blind study participants and their physicians to which drug they were taking.

In the article that follows, I will explore the background, design, findings, and implications of this key trial and try to untangle the many questions about how to interpret it.

LOWERING LDL-C REDUCES RISK, BUT DOES NOT ELIMINATE IT

Large randomized controlled trials have consistently shown that lowering the level of LDL-C reduces cardiovascular event rates by 25% to 45% both in people who are known to have coronary artery disease and in those who are not.^2–4 As a result, guidelines for preventing cardiovascular disease have increasingly emphasized maintaining low LDL-C levels. This has led to a proliferation in the use of inhibitors of 3-hydroxy-3-methylglutaryl coenzyme A reductase (statins) in patients at high cardiovascular risk.

However, these agents only reduce the risk—they do not eliminate it. Needed are additional therapies to complement existing LDL-C-lowering approaches to lower the cardiovascular risk even further.

Raising HDL-C: The next frontier

One such strategy for further lowering cardiovascular risk that has received considerable interest is to promote the biological activity of the “good” cholesterol.

Studies have consistently shown that the higher the plasma level of high-density lipoprotein cholesterol (HDL-C), the lower the risk of cardiovascular events, suggesting that raising HDL-C may be beneficial.⁵ Studies in animals with atherosclerosis show that raising HDL-C via genetic modification of the animal or direct infusion of the molecule has a favorable impact on both the size and the structure of experimental plaque.^6,7

Accordingly, much activity has focused on developing new therapies that raise HDL-C more effectively than current ones.

Why niacin should protect the heart

For more than 50 years, niacin has been used to manage dyslipidemia.

In addition to raising HDL-C levels more effectively than any other agent available today, niacin also lowers the levels of LDL-C, triglycerides, and lipoprotein (a).⁸ Before statins were available, the Coronary Drug Project found that niacin reduced the rate of nonfatal myocardial infarction and the 15-year mortality rate.⁹ In addition, niacin has been shown to slow the progression of carotid intimal-medial thickness and coronary atherosclerosis, and even to reverse these processes in some trials.^10–12

However, a number of issues remain about using niacin to prevent cardiovascular events. Nearly all patients who take it experience flushing, which limits its tolerability and, thus, our ability to titrate doses to levels needed for adequate lipid changes. While a number of modifications of niacin administration have been developed (eg, extended-release formulations and products that inhibit flushing), no large study has tested the clinical efficacy of these strategies. Furthermore, until AIM-HIGH, no large-scale trial had directly evaluated the impact of niacin therapy on a background of statin therapy.

AIM-HIGH STUDY DESIGN

The intent of the AIM-HIGH trial was to determine whether extended-release niacin (Niaspan) would reduce the risk of cardiovascular events when added to therapy with a statin—in this case, simvastatin (Zocor) supplemented with ezetimibe (Zetia).¹

The trial was funded by the National Heart, Lung, and Blood Institute (NHLBI) and by Abbott Laboratories, which also supplied the extended-release niacin and the ezetimibe. Merck donated the simvastatin.

Patient characteristics

The patients were all at least 45 years of age with established, stable coronary heart disease, cerebrovascular or carotid arterial disease, or peripheral arterial disease. They also had to have low levels of HDL-C (< 40 mg/dL in men, < 50 mg/dL in women), elevated triglycerides (150–400 mg/dL), and LDL-C levels lower than 180 mg/dL if they were not taking a statin at entry.

The mean age of the patients was 64 years, 85% were men, and 92% were white. They had a high prevalence of cardiovascular risk factors: 34% had diabetes, 71% had hypertension, and 81% had metabolic syndrome. Nearly all (94%) of the patients were taking a statin at entry; 76% had been taking one for more than 1 year, and 40% had been taking one for more than 5 years.¹

Simvastatin, ezetimibe, and either niacin or placebo

All lipid-modifying agents except statins and ezetimibe were stopped for least 4 weeks after enrollment.

All patients then entered a 4- to 8-week open-label period, during which they took simvastatin 40 mg daily and extended-release niacin starting at 500 mg and increased weekly up to 2,000 mg daily. Patients who could tolerate at least 1,500 mg daily were randomly assigned to treatment with either niacin 1,500 to 2,000 mg or matching placebo. Both groups continued to receive simvastatin. The placebo contained a small dose of immediate-release niacin (50 mg) in each tablet to induce flushing and to maintain blinding of treatment.

Given that niacin also lowers LDL-C, an algorithm was used to try to keep LDL-C levels roughly the same in both treatment groups. This involved adjusting the simvastatin dose and permitting the use of ezetimibe 10 mg to keep the LDL-C level between 40 and 80 mg/dL. Accordingly, participating physicians were told their patients’ LDL-C levels but were blinded to their HDL-C and triglyceride levels throughout the study.

Every 6 months, patients had a follow-up visit in the clinic, and midway through each 6-month interval they received a phone call from the investigators.¹

AIM-HIGH end points

The primary end point was the composite of the first event of death due to coronary heart disease, nonfatal myocardial infarction, ischemic stroke, hospitalization for acute coronary syndrome, or symptom-driven revascularization of the coronary or cerebral arteries.

Secondary end points were:

• Death from coronary heart disease, nonfatal myocardial infarction, ischemic stroke, or hospitalization for acute coronary syndrome
• Death from coronary heart disease, nonfatal myocardial infarction, or ischemic stroke
• Death from cardiovascular causes.

Tertiary end points included:

• Death from any cause
• Individual components of the primary end point
• Prespecified subgroups according to sex, history or no history of diabetes, and presence or absence of the metabolic syndrome.¹

All clinical events were adjudicated by a central committee.

STUDY HALTED EARLY

The study was planned to run for a mean of 4.6 years, during which 800 primary end point events were expected. With these numbers, the investigators calculated that the study had 85% power to detect a 25% reduction in the primary end point, at a one-sided alpha level of 0.025.

The plan called for an interim analysis when 50% of the anticipated events had occurred, with prespecified stopping boundaries based on either efficacy or futility. The boundary for lack of efficacy required an observed hazard ratio of at least 1.02 with a probability of less than .001.

In the interim analysis, after a median follow-up of only 3 years, the data and safety monitoring board recommended stopping the study early because the boundary for futility had been crossed and, unexpectedly, the rate of ischemic stroke was higher in the niacin-treated patients than in those receiving placebo.

MAJOR FINDINGS OF AIM-HIGH

Of 4,273 patients who began open-label treatment with niacin, 3,414 were randomized to treatment with niacin or placebo.¹

HDL-C levels went up in both groups

At 2 years:

• HDL-C levels had increased by 25.0% (to 42 mg/dL) in the niacin group and by 9.8% (to 38 mg/dL) in the placebo group
• Triglycerides had decreased by 28.6% with niacin and by 8.1% with placebo
• LDL-C had decreased by 12.0% with niacin and by 5.5% with placebo.

Patients in the placebo group were more likely to have subsequently received the maximum dose of simvastatin, ie, 80 mg/day (24.7% vs 17.5%), and to have received ezetimibe (21.5% vs 9.5%). More patients in the niacin group required either dose reduction of the study drug (6.3% vs 3.4%) or drug discontinuation (25.4% vs 20.1%).¹

No difference in the primary end point

There was no difference between the two treatment groups in the rate of the primary end point, which occurred in 282 (16.4%) of the 1,718 patients in the niacin group and 272 (16.2%) of the 1,696 patients in the placebo group (P = .79; hazard ratio 1.02, 95% confidence interval 0.87–1.21).¹

However, more patients in the niacin group than in the placebo group who reached the primary end point did so by having a first ischemic stroke: 27 patients (1.6%) vs 15 patients (0.9%). Eight of these patients, all in the niacin group, had their stroke between 2 months and 4 years after they had stopped taking the study drug.

Further analysis that included all ischemic strokes revealed the same trend: 29 vs 18 patients (P = .11).¹

No benefit was observed for niacin-treated patients in terms of any of the secondary or tertiary end points.

Subgroup analysis revealed no evidence of statistical heterogeneity: ie, niacin seemed to lack efficacy in all the prespecified subgroups studied (age 65 and older vs younger, men vs women, and those with or without diabetes, metabolic syndrome, prior myocardial infarction, or statin use at entry).

In general, niacin was well tolerated in the active-treatment group, with a low incidence of liver and muscle abnormalities.

PUTTING AIM-HIGH IN CONTEXT

How should practicing clinicians interpret these outcomes?

Ever since the NHLBI reported (in an urgent press release) that it was stopping the study early due to futility and a potential excess of strokes,¹³ there has been considerable debate as to which factors contributed to these outcomes. In the wake of the publication of more detailed information about the trial,¹ this debate is likely to continue.

The AIM-HIGH results can be interpreted in several ways:

• Perhaps niacin is no good as a preventive agent
• Perhaps raising HDL-C is flawed as a preventive strategy
• Perhaps AIM-HIGH had methodologic flaws, such as looking at the wrong patient cohort or using a treatment protocol that set itself up for failure
• Perhaps statins are so good that, once you prescribe one, anything else you give provides no additional benefit.

Which of these is correct?

Is niacin no good?

In its most simple form, AIM-HIGH has always been seen as a clinical trial of niacin. While the early trials of immediate-release niacin were encouraging in terms of its effects on lipids, atherosclerotic plaque, and cardiovascular outcomes, using it in clinical practice has always been challenging, largely because many patients cannot tolerate it in doses high enough to be effective. A number of developments have improved niacin’s tolerability, but its clinical impact in the statin era has not been evaluated.

Niacin’s lack of efficacy in this trial will ultimately be viewed as a failure of the drug itself, but is this the case?

AIM-HIGH was not simply a direct comparison of niacin vs placebo on top of standard medical practice. The investigators recognized that niacin has additional effects—in particular, lowering levels of atherogenic lipids—and they attempted to control for these effects by titrating the other LDL-C-lowering therapies during the study. As a result, the trial was actually a comparison between niacin plus low-dose simvastatin on the one hand, and placebo plus high-dose simvastatin (and, more often, also ezetimibe) on the other.

Furthermore, the placebo-treated patients received small doses of immediate-release niacin to induce flushing and maintain blinding. It is therefore hard to conclude that this clinical trial was a direct evaluation of the impact of niacin.

In contrast, the Heart Protection Study 2-Treatment of HDL to Reduce the Incidence of Vascular Events (HPS2-THRIVE) study is currently evaluating extended-release niacin in combination with laropiprant, a prostaglandin receptor antagonist, vs placebo in more than 24,000 statin-treated patients.¹⁴ Without any in-trial titration of lipids, this study provides a more direct comparison of the effects of niacin in the statin era.

Niacin continues to attract interest, largely because it can raise HDL-C by 20% to 30% when given at doses of 1,500 mg or more. Also, consistent observations from population studies of an inverse relationship between HDL-C levels and cardiovascular risk⁵ have stimulated interest in developing novel agents that substantially raise HDL-C.

The failure of HDL-C-raising therapies in clinical trials^15,16 has fueled concern that HDL may not be the magic elixir that many have sought. Given that niacin is the most effective HDL-C-raising agent currently available, its lack of efficacy in AIM-HIGH could be perceived as another nail in the coffin of the hypothesis that raising the HDL-C level with pharmacologic agents is beneficial.

AIM-HIGH was designed to examine the effects of raising HDL-C. To this end, it was performed exclusively in patients with low HDL-C levels, and the investigators tried to isolate the potential effects of raising HDL-C by equalizing the LDL-C levels in the treatment groups.

However, the HDL-C changes observed in AIM-HIGH are likely to have undermined the study objective. While niacin predictably increased HDL-C levels by 25%, an unexpected increase in HDL-C of 9.8% in the placebo-treated patients resulted in a difference in achieved HDL-C levels of only 4 mg/dL between the groups. This was far less than anticipated, and it likely had a major impact on an already underpowered study.

AIM-HIGH was designed to have 85% power to demonstrate a 25% reduction in clinical events, which was an optimistic estimate. On the basis of population studies, a difference of 4 mg/dL in HDL-C would be anticipated to result in no more than a 10% lower rate of clinical events, far beyond AIM-HIGH’s limit of detection.

The reasons for the increase in HDL-C in the placebo group are unknown, but they likely reflect the use of higher doses of simvastatin, some regression to the mean, and, possibly, the small doses of immediate-release niacin that the placebo contained. (Contrary to the belief of the investigators, there have been some reports of lipid changes with such doses,¹⁷ which may have contributed to the observed HDL-C-raising.)

Given that the HDL-C difference between the groups was relatively small and that niacin has additional effects beyond raising HDL-C and lowering LDL-C, it is unlikely that the futility of AIM-HIGH reflects a major indictment of HDL-C-raising. For the time being, the jury is still out on this question.

Was AIM-HIGH methodologically flawed?

A number of methodologic issues may have affected AIM-HIGH’s ability to adequately address its objectives.

The wrong cohort? In planning a study such as AIM-HIGH, the need for a relatively small sample size and the need to detect the greatest relative risk reduction with niacin would require enrollment of patients at the highest risk of cardiovascular events despite the use of statins. These needs were satisfied by only including patients who had atherosclerotic cardiovascular disease and low HDL-C levels. The inclusion of patients with low levels of HDL-C was also expected to promote greater increases in this lipid, and potentially event reduction, with niacin.

But no benefit was observed. It remains to be determined whether the inclusion of a high proportion of patients with the metabolic syndrome adversely affected the ability to detect a benefit with niacin. While post hoc analyses of studies of carotid intimal-medial thickness demonstrated no relationship between raising HDL-C with niacin and slowing of disease progression in patients with the metabolic syndrome,¹⁸ it remains to be determined whether this would translate to any effect on cardiovascular event rates.

Inadequate statistical power? An underpowered study would leave very little room for error, a pertinent point given the variability in therapeutic response in both actively treated and placebo-treated patients typically encountered in clinical trials. Giving low doses of immediate-release niacin and titrating the simvastatin dose to control LDL-C, resulting in imbalances in lipid-modifying therapies, represent additional flaws in the study design.

Stopped too soon? The early cessation of the study was somewhat questionable. The study crossed the prespecified boundary for lack of efficacy at the time of the interim analysis, and initial review by the data and safety monitoring board suggested an excess rate of ischemic stroke with niacin. The inclusion of this latter finding in the press release prompted considerable speculation regarding potential mechanisms and also concern among patients currently taking niacin. The subsequent finding that this signal was not statistically significant serves as an important warning for those conducting clinical trials not to prematurely overstate preliminary observations.

The implications for agents used in clinical practice are considerable: negative findings should not be overemphasized without robust evidence.

Do statins make everything else irrelevant?

The final factor to consider is the relative modifiability of residual clinical risk in statin-treated patients.

While residual risk is often cited as the reason to develop new antiatherosclerotic therapies, it is unknown how many of these ongoing events can be prevented. Several nonmodifiable factors such as age and concomitant disease are likely to contribute to these clinical events, which may limit our ability to further reduce event rates in patients who have already achieved low LDL-C levels with statin therapy. This may underscore the observation that no major clinical trial has demonstrated clinical benefit of an antiatherosclerotic agent on top of background medical care that included statins.

The finding that atherosclerosis continues to progress in many patients even though they take statins in high doses or achieve low LDL-C levels suggests that there is still room for improvement.

WHAT FUTURE FOR NIACIN?

So what does the future hold for niacin? The ongoing HPS2-THRIVE study provides another opportunity to evaluate the potential clinical efficacy of niacin in statin-treated patients. For now, we must wait for the results of this study.

In the meantime, it would seem reasonable to continue treatment with niacin in patients who need it for its multiple lipid-modifying effects. Whether clinicians will be less likely to initiate niacin therapy until there is clear evidence of clinical benefit remains uncertain. As for HDL-C, it remains to be determined whether any therapy targeting either quantitative or qualitative changes will be beneficial.

Over the last 3 decades, clinical trials have provided important insights into the prevention of cardiovascular events and have had a profound impact on clinical practice. Such studies simply evaluate whether one strategy is better or worse than the existing standard of care. They do not provide mechanistic insights, and when attempts have been made to address mechanisms in the study design, the trial, as in the case of AIM-HIGH, leaves more questions than answers.

Future trials will provide more clarity as to the optimal way to treat patients, but they must be based on a robust design that permits the study question to be adequately addressed.

The recent publication of the AIM-HIGH trial (Atherothrombosis Intervention in Metabolic Syndrome With Low HDL/High Triglycerides: Impact on Global Health Outcomes)¹ has thrown the use of niacin as a lipid-modifying therapy into question. The trial was stopped early because an interim analysis found that the patients who took extended-release niacin had no clinical benefit. In addition, it found a trend toward more ischemic strokes, though this finding was later found not to be statistically significant.

Complicating the interpretation, while both the treatment group and the control group in the study received statin therapy, the researchers attempted to keep low-density lipoprotein cholesterol (LDL-C) levels equal, meaning that patients in the control group received more intensive statin therapy than those in the treatment group. And the placebo that the control patients received was actually a low dose of niacin, to induce flushing and thus to blind study participants and their physicians to which drug they were taking.

In the article that follows, I will explore the background, design, findings, and implications of this key trial and try to untangle the many questions about how to interpret it.

LOWERING LDL-C REDUCES RISK, BUT DOES NOT ELIMINATE IT

Large randomized controlled trials have consistently shown that lowering the level of LDL-C reduces cardiovascular event rates by 25% to 45% both in people who are known to have coronary artery disease and in those who are not.^2–4 As a result, guidelines for preventing cardiovascular disease have increasingly emphasized maintaining low LDL-C levels. This has led to a proliferation in the use of inhibitors of 3-hydroxy-3-methylglutaryl coenzyme A reductase (statins) in patients at high cardiovascular risk.

However, these agents only reduce the risk—they do not eliminate it. Needed are additional therapies to complement existing LDL-C-lowering approaches to lower the cardiovascular risk even further.

Raising HDL-C: The next frontier

One such strategy for further lowering cardiovascular risk that has received considerable interest is to promote the biological activity of the “good” cholesterol.

Studies have consistently shown that the higher the plasma level of high-density lipoprotein cholesterol (HDL-C), the lower the risk of cardiovascular events, suggesting that raising HDL-C may be beneficial.⁵ Studies in animals with atherosclerosis show that raising HDL-C via genetic modification of the animal or direct infusion of the molecule has a favorable impact on both the size and the structure of experimental plaque.^6,7

Accordingly, much activity has focused on developing new therapies that raise HDL-C more effectively than current ones.

Why niacin should protect the heart

For more than 50 years, niacin has been used to manage dyslipidemia.

In addition to raising HDL-C levels more effectively than any other agent available today, niacin also lowers the levels of LDL-C, triglycerides, and lipoprotein (a).⁸ Before statins were available, the Coronary Drug Project found that niacin reduced the rate of nonfatal myocardial infarction and the 15-year mortality rate.⁹ In addition, niacin has been shown to slow the progression of carotid intimal-medial thickness and coronary atherosclerosis, and even to reverse these processes in some trials.^10–12

However, a number of issues remain about using niacin to prevent cardiovascular events. Nearly all patients who take it experience flushing, which limits its tolerability and, thus, our ability to titrate doses to levels needed for adequate lipid changes. While a number of modifications of niacin administration have been developed (eg, extended-release formulations and products that inhibit flushing), no large study has tested the clinical efficacy of these strategies. Furthermore, until AIM-HIGH, no large-scale trial had directly evaluated the impact of niacin therapy on a background of statin therapy.

AIM-HIGH STUDY DESIGN

The intent of the AIM-HIGH trial was to determine whether extended-release niacin (Niaspan) would reduce the risk of cardiovascular events when added to therapy with a statin—in this case, simvastatin (Zocor) supplemented with ezetimibe (Zetia).¹

The trial was funded by the National Heart, Lung, and Blood Institute (NHLBI) and by Abbott Laboratories, which also supplied the extended-release niacin and the ezetimibe. Merck donated the simvastatin.

Patient characteristics

The patients were all at least 45 years of age with established, stable coronary heart disease, cerebrovascular or carotid arterial disease, or peripheral arterial disease. They also had to have low levels of HDL-C (< 40 mg/dL in men, < 50 mg/dL in women), elevated triglycerides (150–400 mg/dL), and LDL-C levels lower than 180 mg/dL if they were not taking a statin at entry.

The mean age of the patients was 64 years, 85% were men, and 92% were white. They had a high prevalence of cardiovascular risk factors: 34% had diabetes, 71% had hypertension, and 81% had metabolic syndrome. Nearly all (94%) of the patients were taking a statin at entry; 76% had been taking one for more than 1 year, and 40% had been taking one for more than 5 years.¹

Simvastatin, ezetimibe, and either niacin or placebo

All lipid-modifying agents except statins and ezetimibe were stopped for least 4 weeks after enrollment.

All patients then entered a 4- to 8-week open-label period, during which they took simvastatin 40 mg daily and extended-release niacin starting at 500 mg and increased weekly up to 2,000 mg daily. Patients who could tolerate at least 1,500 mg daily were randomly assigned to treatment with either niacin 1,500 to 2,000 mg or matching placebo. Both groups continued to receive simvastatin. The placebo contained a small dose of immediate-release niacin (50 mg) in each tablet to induce flushing and to maintain blinding of treatment.

Given that niacin also lowers LDL-C, an algorithm was used to try to keep LDL-C levels roughly the same in both treatment groups. This involved adjusting the simvastatin dose and permitting the use of ezetimibe 10 mg to keep the LDL-C level between 40 and 80 mg/dL. Accordingly, participating physicians were told their patients’ LDL-C levels but were blinded to their HDL-C and triglyceride levels throughout the study.

Every 6 months, patients had a follow-up visit in the clinic, and midway through each 6-month interval they received a phone call from the investigators.¹

AIM-HIGH end points

The primary end point was the composite of the first event of death due to coronary heart disease, nonfatal myocardial infarction, ischemic stroke, hospitalization for acute coronary syndrome, or symptom-driven revascularization of the coronary or cerebral arteries.

Secondary end points were:

• Death from coronary heart disease, nonfatal myocardial infarction, ischemic stroke, or hospitalization for acute coronary syndrome
• Death from coronary heart disease, nonfatal myocardial infarction, or ischemic stroke
• Death from cardiovascular causes.

Tertiary end points included:

• Death from any cause
• Individual components of the primary end point
• Prespecified subgroups according to sex, history or no history of diabetes, and presence or absence of the metabolic syndrome.¹

All clinical events were adjudicated by a central committee.

STUDY HALTED EARLY

The study was planned to run for a mean of 4.6 years, during which 800 primary end point events were expected. With these numbers, the investigators calculated that the study had 85% power to detect a 25% reduction in the primary end point, at a one-sided alpha level of 0.025.

The plan called for an interim analysis when 50% of the anticipated events had occurred, with prespecified stopping boundaries based on either efficacy or futility. The boundary for lack of efficacy required an observed hazard ratio of at least 1.02 with a probability of less than .001.

In the interim analysis, after a median follow-up of only 3 years, the data and safety monitoring board recommended stopping the study early because the boundary for futility had been crossed and, unexpectedly, the rate of ischemic stroke was higher in the niacin-treated patients than in those receiving placebo.

MAJOR FINDINGS OF AIM-HIGH

Of 4,273 patients who began open-label treatment with niacin, 3,414 were randomized to treatment with niacin or placebo.¹

HDL-C levels went up in both groups

At 2 years:

• HDL-C levels had increased by 25.0% (to 42 mg/dL) in the niacin group and by 9.8% (to 38 mg/dL) in the placebo group
• Triglycerides had decreased by 28.6% with niacin and by 8.1% with placebo
• LDL-C had decreased by 12.0% with niacin and by 5.5% with placebo.

Patients in the placebo group were more likely to have subsequently received the maximum dose of simvastatin, ie, 80 mg/day (24.7% vs 17.5%), and to have received ezetimibe (21.5% vs 9.5%). More patients in the niacin group required either dose reduction of the study drug (6.3% vs 3.4%) or drug discontinuation (25.4% vs 20.1%).¹

No difference in the primary end point

There was no difference between the two treatment groups in the rate of the primary end point, which occurred in 282 (16.4%) of the 1,718 patients in the niacin group and 272 (16.2%) of the 1,696 patients in the placebo group (P = .79; hazard ratio 1.02, 95% confidence interval 0.87–1.21).¹

However, more patients in the niacin group than in the placebo group who reached the primary end point did so by having a first ischemic stroke: 27 patients (1.6%) vs 15 patients (0.9%). Eight of these patients, all in the niacin group, had their stroke between 2 months and 4 years after they had stopped taking the study drug.

Further analysis that included all ischemic strokes revealed the same trend: 29 vs 18 patients (P = .11).¹

No benefit was observed for niacin-treated patients in terms of any of the secondary or tertiary end points.

Subgroup analysis revealed no evidence of statistical heterogeneity: ie, niacin seemed to lack efficacy in all the prespecified subgroups studied (age 65 and older vs younger, men vs women, and those with or without diabetes, metabolic syndrome, prior myocardial infarction, or statin use at entry).

In general, niacin was well tolerated in the active-treatment group, with a low incidence of liver and muscle abnormalities.

PUTTING AIM-HIGH IN CONTEXT

How should practicing clinicians interpret these outcomes?

Ever since the NHLBI reported (in an urgent press release) that it was stopping the study early due to futility and a potential excess of strokes,¹³ there has been considerable debate as to which factors contributed to these outcomes. In the wake of the publication of more detailed information about the trial,¹ this debate is likely to continue.

The AIM-HIGH results can be interpreted in several ways:

• Perhaps niacin is no good as a preventive agent
• Perhaps raising HDL-C is flawed as a preventive strategy
• Perhaps AIM-HIGH had methodologic flaws, such as looking at the wrong patient cohort or using a treatment protocol that set itself up for failure
• Perhaps statins are so good that, once you prescribe one, anything else you give provides no additional benefit.

Which of these is correct?

Is niacin no good?

In its most simple form, AIM-HIGH has always been seen as a clinical trial of niacin. While the early trials of immediate-release niacin were encouraging in terms of its effects on lipids, atherosclerotic plaque, and cardiovascular outcomes, using it in clinical practice has always been challenging, largely because many patients cannot tolerate it in doses high enough to be effective. A number of developments have improved niacin’s tolerability, but its clinical impact in the statin era has not been evaluated.

Niacin’s lack of efficacy in this trial will ultimately be viewed as a failure of the drug itself, but is this the case?

AIM-HIGH was not simply a direct comparison of niacin vs placebo on top of standard medical practice. The investigators recognized that niacin has additional effects—in particular, lowering levels of atherogenic lipids—and they attempted to control for these effects by titrating the other LDL-C-lowering therapies during the study. As a result, the trial was actually a comparison between niacin plus low-dose simvastatin on the one hand, and placebo plus high-dose simvastatin (and, more often, also ezetimibe) on the other.

Furthermore, the placebo-treated patients received small doses of immediate-release niacin to induce flushing and maintain blinding. It is therefore hard to conclude that this clinical trial was a direct evaluation of the impact of niacin.

In contrast, the Heart Protection Study 2-Treatment of HDL to Reduce the Incidence of Vascular Events (HPS2-THRIVE) study is currently evaluating extended-release niacin in combination with laropiprant, a prostaglandin receptor antagonist, vs placebo in more than 24,000 statin-treated patients.¹⁴ Without any in-trial titration of lipids, this study provides a more direct comparison of the effects of niacin in the statin era.

Niacin continues to attract interest, largely because it can raise HDL-C by 20% to 30% when given at doses of 1,500 mg or more. Also, consistent observations from population studies of an inverse relationship between HDL-C levels and cardiovascular risk⁵ have stimulated interest in developing novel agents that substantially raise HDL-C.

The failure of HDL-C-raising therapies in clinical trials^15,16 has fueled concern that HDL may not be the magic elixir that many have sought. Given that niacin is the most effective HDL-C-raising agent currently available, its lack of efficacy in AIM-HIGH could be perceived as another nail in the coffin of the hypothesis that raising the HDL-C level with pharmacologic agents is beneficial.

AIM-HIGH was designed to examine the effects of raising HDL-C. To this end, it was performed exclusively in patients with low HDL-C levels, and the investigators tried to isolate the potential effects of raising HDL-C by equalizing the LDL-C levels in the treatment groups.

However, the HDL-C changes observed in AIM-HIGH are likely to have undermined the study objective. While niacin predictably increased HDL-C levels by 25%, an unexpected increase in HDL-C of 9.8% in the placebo-treated patients resulted in a difference in achieved HDL-C levels of only 4 mg/dL between the groups. This was far less than anticipated, and it likely had a major impact on an already underpowered study.

AIM-HIGH was designed to have 85% power to demonstrate a 25% reduction in clinical events, which was an optimistic estimate. On the basis of population studies, a difference of 4 mg/dL in HDL-C would be anticipated to result in no more than a 10% lower rate of clinical events, far beyond AIM-HIGH’s limit of detection.

The reasons for the increase in HDL-C in the placebo group are unknown, but they likely reflect the use of higher doses of simvastatin, some regression to the mean, and, possibly, the small doses of immediate-release niacin that the placebo contained. (Contrary to the belief of the investigators, there have been some reports of lipid changes with such doses,¹⁷ which may have contributed to the observed HDL-C-raising.)

Given that the HDL-C difference between the groups was relatively small and that niacin has additional effects beyond raising HDL-C and lowering LDL-C, it is unlikely that the futility of AIM-HIGH reflects a major indictment of HDL-C-raising. For the time being, the jury is still out on this question.

Was AIM-HIGH methodologically flawed?

A number of methodologic issues may have affected AIM-HIGH’s ability to adequately address its objectives.

The wrong cohort? In planning a study such as AIM-HIGH, the need for a relatively small sample size and the need to detect the greatest relative risk reduction with niacin would require enrollment of patients at the highest risk of cardiovascular events despite the use of statins. These needs were satisfied by only including patients who had atherosclerotic cardiovascular disease and low HDL-C levels. The inclusion of patients with low levels of HDL-C was also expected to promote greater increases in this lipid, and potentially event reduction, with niacin.

But no benefit was observed. It remains to be determined whether the inclusion of a high proportion of patients with the metabolic syndrome adversely affected the ability to detect a benefit with niacin. While post hoc analyses of studies of carotid intimal-medial thickness demonstrated no relationship between raising HDL-C with niacin and slowing of disease progression in patients with the metabolic syndrome,¹⁸ it remains to be determined whether this would translate to any effect on cardiovascular event rates.

Inadequate statistical power? An underpowered study would leave very little room for error, a pertinent point given the variability in therapeutic response in both actively treated and placebo-treated patients typically encountered in clinical trials. Giving low doses of immediate-release niacin and titrating the simvastatin dose to control LDL-C, resulting in imbalances in lipid-modifying therapies, represent additional flaws in the study design.

Stopped too soon? The early cessation of the study was somewhat questionable. The study crossed the prespecified boundary for lack of efficacy at the time of the interim analysis, and initial review by the data and safety monitoring board suggested an excess rate of ischemic stroke with niacin. The inclusion of this latter finding in the press release prompted considerable speculation regarding potential mechanisms and also concern among patients currently taking niacin. The subsequent finding that this signal was not statistically significant serves as an important warning for those conducting clinical trials not to prematurely overstate preliminary observations.

The implications for agents used in clinical practice are considerable: negative findings should not be overemphasized without robust evidence.

Do statins make everything else irrelevant?

The final factor to consider is the relative modifiability of residual clinical risk in statin-treated patients.

While residual risk is often cited as the reason to develop new antiatherosclerotic therapies, it is unknown how many of these ongoing events can be prevented. Several nonmodifiable factors such as age and concomitant disease are likely to contribute to these clinical events, which may limit our ability to further reduce event rates in patients who have already achieved low LDL-C levels with statin therapy. This may underscore the observation that no major clinical trial has demonstrated clinical benefit of an antiatherosclerotic agent on top of background medical care that included statins.

The finding that atherosclerosis continues to progress in many patients even though they take statins in high doses or achieve low LDL-C levels suggests that there is still room for improvement.

WHAT FUTURE FOR NIACIN?

So what does the future hold for niacin? The ongoing HPS2-THRIVE study provides another opportunity to evaluate the potential clinical efficacy of niacin in statin-treated patients. For now, we must wait for the results of this study.

In the meantime, it would seem reasonable to continue treatment with niacin in patients who need it for its multiple lipid-modifying effects. Whether clinicians will be less likely to initiate niacin therapy until there is clear evidence of clinical benefit remains uncertain. As for HDL-C, it remains to be determined whether any therapy targeting either quantitative or qualitative changes will be beneficial.

Over the last 3 decades, clinical trials have provided important insights into the prevention of cardiovascular events and have had a profound impact on clinical practice. Such studies simply evaluate whether one strategy is better or worse than the existing standard of care. They do not provide mechanistic insights, and when attempts have been made to address mechanisms in the study design, the trial, as in the case of AIM-HIGH, leaves more questions than answers.

Future trials will provide more clarity as to the optimal way to treat patients, but they must be based on a robust design that permits the study question to be adequately addressed.