Interpreting Key Trials

The medical community has struggled with two important questions for the past 10 years: When it comes to the low-density lipoprotein cholesterol (LDL-C) level, how low should one go and at what cost? And are there other markers of risk that can identify a higher-risk subpopulation in relatively healthy people? The JUPITER trial (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) provided partial answers for these questions by finding that a highly potent statin lowered the risk of cardiovascular events in patients with “normal” LDL-C but elevated levels of high-sensitivity C-reactive protein (hs-CRP).¹

In this article, we will critically evaluate the methods, results, and conclusions of the JUPITER trial. Additionally, we will discuss its limitations and areas of uncertainty.

BEFORE JUPITER

The LDL-C-lowering drugs called statins have revolutionized cardiovascular medicine.² They are beneficial in both the primary prevention setting and in acute coronary syndromes, stable angina, and unstable angina and can halt the progression of coronary artery disease—in some cases even resulting in modest regression of plaque.^3–6

Many experts have credited the reduction in LDL-C as being the sole factor responsible for the decrease in major adverse events seen with statin therapy.⁷ However, statins have other, non-lipid-lowering properties, including anti-inflammatory and antioxidant effects, that may also contribute to their benefits.^8–15

One of the anti-inflammatory actions of statins is evidenced by lower levels of the acute-phase reactant CRP.^10,11,15,16 Measuring systemic CRP levels with a highly sensitive assay (yielding the so-called high-sensitivity or hs-CRP level) provides significant clinical prognostic value across a spectrum of clinical situations, ranging from risk screening in apparently healthy people to stable and unstable angina.^17–22 People with higher hs-CRP levels are, on average, at higher risk of adverse cardiovascular events. However, controversy remains as to whether hs-CRP plays a mechanistic role in plaque formation and acute complications. Indeed, recent genetic studies argue strongly that hs-CRP lies outside the mechanistic path of atherosclerosis.²³ Nonetheless, an overwhelming amount of data indicates that hs-CRP serves as a marker of disease.^17–21

Nissen et al¹⁰ showed that the rate of progression of atherosclerosis is lower when the levels of atherogenic lipoproteins and hs-CRP are both lowered with statin therapy. Simultaneously, Ridker et al¹¹ showed that patients who have lower hs-CRP levels after statin therapy have better clinical outcomes than those with higher hs-CRP levels, regardless of their achieved level of LDL-C.

Collectively, these studies and others have led some to believe that, in people with relatively low LDL-C but persistently elevated hs-CRP, statin therapy may reduce the rate of events.^15,24 The JUPITER trial was undertaken to test this hypothesis.

JUPITER STUDY DESIGN

JUPITER was designed to see whether highly potent statin therapy is beneficial in people with elevated hs-CRP who otherwise do not meet the criteria for lipid-lowering therapy. The study was conducted at 1,315 sites in 26 countries. It was sponsored by AstraZeneca, the maker of rosuvastatin (Crestor).

Inclusion and exclusion criteria

All participants had to be free of known cardiovascular disease, have an LDL-C level lower than 130 mg/dL, and have an hs-CRP level of 2.0 mg/L or greater. Patients were excluded if they were previous or current users of lipid-lowering drugs; had severe arthritis, lupus, or inflammatory bowel disease; or were taking immune-modulating drugs such as cyclosporine (Sandimmune, others), tacrolimus (Prograf), azathioprine (Azasan, Imuran), or long-term oral corticosteroids.

Rosuvastatin therapy

Participants were randomly assigned in a 1:1 ratio to receive rosuvastatin 20 mg daily or a matching placebo in a double-blind fashion.

End points

The primary end point was the composite of nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina, an arterial revascularization procedure, or confirmed death from cardiovascular causes. Secondary end points were the individual components of the primary end point.

Statistical analysis

The study was powered to detect a 25% reduction in the primary end point among those treated with rosuvastatin. The trial was designed to run until 520 end point events had occurred. However, on March 29, 2008, after the first prespecified interim analysis, the Data and Safety Monitoring Board stopped the trial due to a significant reduction in the primary end point in the rosuvastatin group. As in most randomized clinical trials, all analyses were done on an intention-to-treat basis. Prespecified subgroup analyses were also performed.

STUDY RESULTS

Patient recruitment and eligibility

Between February 4, 2003, and December 15, 2006, a total of 89,890 people were screened. Of these, 17,802 met the inclusion and exclusion criteria and were included in the study. Of the 72,088 people who were excluded, 25,993 (36.1%) had an hs-CRP level below 2 mg/L and 37,611 (52.2%) had an LDL-C level of 130 mg/dL or higher.

A not-so-healthy population

The aim of the investigators was to include relatively healthy people. The median age was 66 years, about 16% of participants were current smokers, about 11% had a family history of heart disease, and about 41% met the criteria for metabolic syndrome, all conditions that are associated with elevated hs-CRP.²⁵ Of note, the median hs-CRP level was 4.2 mg/L, a level indicating higher global risk according to the American College of Cardiology/American Heart Association consensus statement.²⁶

Reduction in lipid levels and hs-CRP

By 12 months, in the rosuvastatin group, the median LDL-C level had fallen by 50% (from 108 to 55 mg/dL), and the median hs-CRP level had fallen by 37% (from 4.2 to 2.2 mg/L). Additionally, the triglyceride level had fallen by 17%. The high-density lipoprotein cholesterol levels did not change significantly.

Impact on end points

Adverse events

Ridker PM, et al. Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein. N Engl J Med 2008; 359:2195-2207. Copyright 2008 Massachusetts Medical Society. All rights reserved.

WHAT DOES THIS MEAN?

Is lower LDL-C better?

The JUPITER trial is the latest of several statin trials that have shown significant reductions in major adverse cardiovascular events when LDL-C was lowered below what has been recommended by the current guidelines.^27,28

In 2002, the Heart Protection Study²⁹ showed a significant reduction in major adverse cardiovascular events in patients at high risk of coronary artery disease if they received simvastatin (Zocor), even if they had LDL-C levels lower than 100 mg/dL at baseline. Similarly, the Pravastatin or Atorvastatin Evaluation and Infection-Thrombolysis in Myocardial Infarction 22 (PROVE-IT TIMI 22) trial³⁰ showed a 16% relative risk reduction in a composite end point in patients presenting with acute coronary syndrome if they received intensive statin therapy.

These two studies led to an update by the National Cholesterol Education Program (Adult Treatment Panel III), suggesting an optimal LDL-C goal of less than 70 mg/dL in those with coronary artery disease or its risk equivalent (ie, diabetes mellitus, peripheral vascular disease). Furthermore, in support of the “lower is better” theory, a number of studies that used intravascular ultrasonography have shown regression of coronary plaque with aggressive LDL-C lowering. Notably, in a Study to Evaluate the Effect of Rosuvastatin on Intravascular Ultrasound-Derived Coronary Atheroma Burden (the ASTEROID trial),⁵ rosuvastatin 40 mg daily caused significant plaque regression while lowering LDL-C to 61 mg/dL over a 24-month period.

A number of high-dose statin trials have shown that lowering LDL-C to less than 70 mg/dL significantly reduces major adverse cardiovascular events.^31–39 The JUPITER trial was unique in that it extended these findings to people without known coronary disease (ie, primary prevention) or elevated cholesterol but with elevated levels of a marker of inflammation—hs-CRP. In view of the JUPITER results and of studies using intravascular ultrasonography in the primary prevention setting, it seems clear that lowering LDL-C to levels less than 70 mg/dL also reduces both atherosclerotic plaque progression and the rate of first major adverse cardiovascular events in primary prevention in patients at higher global risk.

Did the study prove that reducing hs-CRP lowers risk?

Measuring hs-CRP levels has been extensively studied in apparently healthy populations, stable angina, unstable angina, and other cardiovascular settings.^{18,21,40–43} It has been shown to have significant prognostic implications in a number of primary and secondary trials.⁴⁴ Additionally, those with elevated LDL-C and hs-CRP levels benefit the most from statin therapy.^16,45,46 Animal studies have also provided some evidence that CRP may play a role in atherogenesis.^47,48 However, recent clinical and genetic studies have raised doubt about the direct causal relationship between CRP and coronary artery disease,^23,49,50 and epidemiologic studies have questioned its usefulness as a marker of risk.^51,52

The JUPITER study adds little to clear up the controversy about whether hs-CRP is a mechanistic participant in atherosclerotic disease. However, it also shows that this issue is somewhat irrelevant, in that selection of patients for high-potency statin therapy solely on the basis of high hs-CRP without other indications for lipid-lowering therapy clearly reduces risk and improves survival.

JUPITER did not examine whether people with higher hs-CRP levels benefited more from statin therapy than those with lower levels. The hypothesis-generating data for JUPITER came from an analysis of changes in hs-CRP and LDL-C in the Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS).¹⁶ Thus, JUPITER did not include people with both low LDL-C and low hs-CRP because, in the AFCAPS/TexCAPS analysis, those with low LDL-C and low hs-CRP had extremely low event rates and no clinical efficacy of statin therapy, despite good LDL-C reduction. In marked contrast, those with low LDL-C but elevated hs-CRP had high event rates and large relative risk reductions— hence the need for JUPITER to prospectively test this hypothesis. Nevertheless, the initial results of JUPITER as presented do not yet make it clear that there is a dose-response relationship between higher levels of hs-CRP and a greater reduction in events, even in a cohort with elevated hs-CRP at baseline. This analysis will no doubt be forthcoming in another manuscript from Ridker and colleagues. Specifically, it will be of interest to examine whether those with the highest hs-CRP levels benefited the most from rosuvastatin on both an absolute and relative scale, and whether those with the greatest hs-CRP reduction also benefited more. With the present data available from JUPITER, a reasonable interpretation is that an elevated hs-CRP simply widens the inclusion criterion for those for whom high-potency statin therapy improves clinical outcomes.⁵³

Even with a nonspecific marker such as hs-CRP, patients at higher global risk and with LDL-C below the recommended levels could be identified and treated aggressively. This benefit, however, required that approximately 100 people be treated with rosuvastatin for 2 years to prevent one event. Additionally, only 20% of all patients screened were eligible for the trial. Therefore, one could argue that its generalizability is limited.

Markers of risk that are more specific and sensitive are needed to identify people at higher global risk who would otherwise be considered to be at low risk with the current risk assessment tools. A number of such inflammatory and oxidative markers are under development.^54–60

Absolute vs relative risk reduction and the public health burden

The 44% reduction in the number of primary end point events in the rosuvastatin group was considerable in relative terms. However, in absolute terms, 95 people had to be treated for up to 2 years in order to prevent one event.⁵³ In making recommendations, the United States Department of Health and Human Services has to consider the clinical benefit of a test or a drug in light of its cost. With health care costs increasing, many agencies are refusing to pay for therapies on the basis of cost or small absolute benefit.

While we do not have the answer as to whether treating 95 people for 2 years to see one benefit is cost-effective, one thing is clear: the field of medicine is in desperate need of a better way to identify individuals who may benefit from a test or therapy.⁶¹ Additionally, we think it is important to note that the “numbers-needed-to-treat” (95 at 2 years and 25 at 5 years) derived from JUPITER are actually smaller than the values observed in the AFCAPS/TexCAPS and the West of Scotland Coronary Prevention Study.^62,63 This suggests that statin therapy is at least as cost-effective in those with elevated hs-CRP as in those with elevated LDL-C. Even our most robust therapies are effective in only a minority of patients treated.⁶¹

Should ‘healthy’ people be tested for hs-CRP?

In 2003, we wrote in this journal²¹ that measuring hs-CRP may add to the current risk-prediction models by identifying people at increased risk who would otherwise not be considered as such by current risk models. The US Centers for Disease Control and Prevention and the American Heart Association have also stated that measuring hs-CRP in those at intermediate risk may be reasonable.²⁶

The JUPITER investigators intended to study a relatively healthy population, but, as we mentioned, a close look at the cohort’s baseline characteristics indicates a substantial proportion met the criteria for metabolic syndrome. Therefore, one could challenge whether we really need hs-CRP in such a population to identify who will benefit from statin therapy.

We agree with the recommendation from the Centers for Disease Control and Prevention and the American Heart Association that measuring hs-CRP in people at intermediate risk is a reasonable option.²⁶ We also believe that hs-CRP should be tested as a secondary risk factor, in combination with blood pressure, lipids, diabetes, smoking, serum creatinine, and fasting blood glucose. Factors such as obesity, sedentary lifestyle, family history of heart disease, and emotional and physical stress should also be considered.

Safety of high-dose statin therapy

High-dose statin therapy has been well tolerated in clinical trials, but rates of discontinuation have been higher (7%–10%) than with moderate-dose therapy (4%–5%).⁶⁴ Fortunately, the rates of serious adverse events have in general been low. For example, with simvastatin 80 mg, the rates of myopathy and rhabdomyolysis were quite low.³¹

Rates of elevations in serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) with high-dose statin therapy have been reported to be below 1.3%. Studies have shown that reducing LDL-C to below 100 mg/dL is associated with a higher incidence of ALT and AST elevations. However, these elevations have usually been benign and often return to normal when the drug is reduced in dose or withdrawn.

In previous studies of rosuvastatin,⁶⁵ the incidence of myopathy and liver function abnormalities was less than 0.1%. Rates of proteinuria were similarly low, and in many patients renal function actually improved on rosuvastatin.^66,67 Furthermore, rosuvastatin may have different pharmacokinetic properties than atorvastatin (Lipitor) and simvastatin, which may result in a lower incidence of musculoskeletal toxicity.^68,69

In general, the incidence of cancer has been similar in those treated with high-dose statins and those treated with placebo. The Treating to New Targets trial⁷⁰ suggested that the incidence of cancer was higher with atorvastatin 80 mg daily than with 20 mg daily. However, a meta-analysis of 14 trials of moderate-dose statin therapy did not show any evidence of increased cancer rates with these agents.⁷⁰ Indeed, in JUPITER, there was a reduction in cancer-related mortality rates, which could have been due to chance.

The JUPITER trial also showed an increase in the physician-reported incidence of diabetes mellitus with rosuvastatin. This is an important finding, and it may be a class effect because modest increases have similarly been reported with other statins in other major trials, eg, with pravastatin (Pravachol) in PROSPER, simvastatin in the Heart Protection Study, and atorvastatin in PROVE-IT. However, even in those with diabetes or impaired fasting glucose, the reduction in the rate of major adverse events is significant. For example, in JUPITER, almost all of the cases of “incident diabetes” were in those with impaired fasting glucose at baseline, and this group had nearly a 50% reduction in rates of myocardial infarction, stroke, and cardiovascular death. Therefore, on balance, the modest risk of earlier diagnosis of diabetes with statin therapy seems substantially offset by the marked reduction in rates of major adverse cardiovascular events in people with diabetes and impaired fasting glucose on statin therapy.

TAKE-HOME POINTS

The JUPITER trial, like previous high-dose statin trials, calls into question whether current LDL-C guidelines are appropriate for people at higher global risk with otherwise “normal” LDL-C levels.^27,28 This trial heralds a new era in preventive therapy because it extends beyond LDL-C as an indication for statin therapy within the primary prevention setting. Statins have revolutionized the therapy of cardiovascular disease, and they continue to show benefit even in the “healthy.”

Clearly, hs-CRP serves as a nonlipid marker to identify those who may benefit from statin therapy. Nonetheless, more specific and sensitive markers (or panels) of cardiovascular risk are necessary. In the future, we will need markers that not only identify people at higher global risk, but that also tell us who would benefit from certain medical or surgical therapies. Elevated hs-CRP in a patient who otherwise would not be a candidate for statin therapy should trigger a reassessment of the risks vs benefits of statin therapy—JUPITER teaches us that statin therapy will benefit these patients.

Aggressive lifestyle modification that encompasses a balanced diet, routine exercise, and smoking cessation should be applied in both primary and secondary prevention. Additionally, risk factors such as elevated blood pressure and hyperlipidemia should be aggressively treated with appropriate medications.

The medical community has struggled with two important questions for the past 10 years: When it comes to the low-density lipoprotein cholesterol (LDL-C) level, how low should one go and at what cost? And are there other markers of risk that can identify a higher-risk subpopulation in relatively healthy people? The JUPITER trial (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) provided partial answers for these questions by finding that a highly potent statin lowered the risk of cardiovascular events in patients with “normal” LDL-C but elevated levels of high-sensitivity C-reactive protein (hs-CRP).¹

In this article, we will critically evaluate the methods, results, and conclusions of the JUPITER trial. Additionally, we will discuss its limitations and areas of uncertainty.

BEFORE JUPITER

The LDL-C-lowering drugs called statins have revolutionized cardiovascular medicine.² They are beneficial in both the primary prevention setting and in acute coronary syndromes, stable angina, and unstable angina and can halt the progression of coronary artery disease—in some cases even resulting in modest regression of plaque.^3–6

Many experts have credited the reduction in LDL-C as being the sole factor responsible for the decrease in major adverse events seen with statin therapy.⁷ However, statins have other, non-lipid-lowering properties, including anti-inflammatory and antioxidant effects, that may also contribute to their benefits.^8–15

One of the anti-inflammatory actions of statins is evidenced by lower levels of the acute-phase reactant CRP.^10,11,15,16 Measuring systemic CRP levels with a highly sensitive assay (yielding the so-called high-sensitivity or hs-CRP level) provides significant clinical prognostic value across a spectrum of clinical situations, ranging from risk screening in apparently healthy people to stable and unstable angina.^17–22 People with higher hs-CRP levels are, on average, at higher risk of adverse cardiovascular events. However, controversy remains as to whether hs-CRP plays a mechanistic role in plaque formation and acute complications. Indeed, recent genetic studies argue strongly that hs-CRP lies outside the mechanistic path of atherosclerosis.²³ Nonetheless, an overwhelming amount of data indicates that hs-CRP serves as a marker of disease.^17–21

Nissen et al¹⁰ showed that the rate of progression of atherosclerosis is lower when the levels of atherogenic lipoproteins and hs-CRP are both lowered with statin therapy. Simultaneously, Ridker et al¹¹ showed that patients who have lower hs-CRP levels after statin therapy have better clinical outcomes than those with higher hs-CRP levels, regardless of their achieved level of LDL-C.

Collectively, these studies and others have led some to believe that, in people with relatively low LDL-C but persistently elevated hs-CRP, statin therapy may reduce the rate of events.^15,24 The JUPITER trial was undertaken to test this hypothesis.

JUPITER STUDY DESIGN

JUPITER was designed to see whether highly potent statin therapy is beneficial in people with elevated hs-CRP who otherwise do not meet the criteria for lipid-lowering therapy. The study was conducted at 1,315 sites in 26 countries. It was sponsored by AstraZeneca, the maker of rosuvastatin (Crestor).

Inclusion and exclusion criteria

All participants had to be free of known cardiovascular disease, have an LDL-C level lower than 130 mg/dL, and have an hs-CRP level of 2.0 mg/L or greater. Patients were excluded if they were previous or current users of lipid-lowering drugs; had severe arthritis, lupus, or inflammatory bowel disease; or were taking immune-modulating drugs such as cyclosporine (Sandimmune, others), tacrolimus (Prograf), azathioprine (Azasan, Imuran), or long-term oral corticosteroids.

Rosuvastatin therapy

Participants were randomly assigned in a 1:1 ratio to receive rosuvastatin 20 mg daily or a matching placebo in a double-blind fashion.

End points

The primary end point was the composite of nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina, an arterial revascularization procedure, or confirmed death from cardiovascular causes. Secondary end points were the individual components of the primary end point.

Statistical analysis

The study was powered to detect a 25% reduction in the primary end point among those treated with rosuvastatin. The trial was designed to run until 520 end point events had occurred. However, on March 29, 2008, after the first prespecified interim analysis, the Data and Safety Monitoring Board stopped the trial due to a significant reduction in the primary end point in the rosuvastatin group. As in most randomized clinical trials, all analyses were done on an intention-to-treat basis. Prespecified subgroup analyses were also performed.

STUDY RESULTS

Patient recruitment and eligibility

Between February 4, 2003, and December 15, 2006, a total of 89,890 people were screened. Of these, 17,802 met the inclusion and exclusion criteria and were included in the study. Of the 72,088 people who were excluded, 25,993 (36.1%) had an hs-CRP level below 2 mg/L and 37,611 (52.2%) had an LDL-C level of 130 mg/dL or higher.

A not-so-healthy population

The aim of the investigators was to include relatively healthy people. The median age was 66 years, about 16% of participants were current smokers, about 11% had a family history of heart disease, and about 41% met the criteria for metabolic syndrome, all conditions that are associated with elevated hs-CRP.²⁵ Of note, the median hs-CRP level was 4.2 mg/L, a level indicating higher global risk according to the American College of Cardiology/American Heart Association consensus statement.²⁶

Reduction in lipid levels and hs-CRP

By 12 months, in the rosuvastatin group, the median LDL-C level had fallen by 50% (from 108 to 55 mg/dL), and the median hs-CRP level had fallen by 37% (from 4.2 to 2.2 mg/L). Additionally, the triglyceride level had fallen by 17%. The high-density lipoprotein cholesterol levels did not change significantly.

Impact on end points

Adverse events

Ridker PM, et al. Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein. N Engl J Med 2008; 359:2195-2207. Copyright 2008 Massachusetts Medical Society. All rights reserved.

WHAT DOES THIS MEAN?

Is lower LDL-C better?

The JUPITER trial is the latest of several statin trials that have shown significant reductions in major adverse cardiovascular events when LDL-C was lowered below what has been recommended by the current guidelines.^27,28

In 2002, the Heart Protection Study²⁹ showed a significant reduction in major adverse cardiovascular events in patients at high risk of coronary artery disease if they received simvastatin (Zocor), even if they had LDL-C levels lower than 100 mg/dL at baseline. Similarly, the Pravastatin or Atorvastatin Evaluation and Infection-Thrombolysis in Myocardial Infarction 22 (PROVE-IT TIMI 22) trial³⁰ showed a 16% relative risk reduction in a composite end point in patients presenting with acute coronary syndrome if they received intensive statin therapy.

These two studies led to an update by the National Cholesterol Education Program (Adult Treatment Panel III), suggesting an optimal LDL-C goal of less than 70 mg/dL in those with coronary artery disease or its risk equivalent (ie, diabetes mellitus, peripheral vascular disease). Furthermore, in support of the “lower is better” theory, a number of studies that used intravascular ultrasonography have shown regression of coronary plaque with aggressive LDL-C lowering. Notably, in a Study to Evaluate the Effect of Rosuvastatin on Intravascular Ultrasound-Derived Coronary Atheroma Burden (the ASTEROID trial),⁵ rosuvastatin 40 mg daily caused significant plaque regression while lowering LDL-C to 61 mg/dL over a 24-month period.

A number of high-dose statin trials have shown that lowering LDL-C to less than 70 mg/dL significantly reduces major adverse cardiovascular events.^31–39 The JUPITER trial was unique in that it extended these findings to people without known coronary disease (ie, primary prevention) or elevated cholesterol but with elevated levels of a marker of inflammation—hs-CRP. In view of the JUPITER results and of studies using intravascular ultrasonography in the primary prevention setting, it seems clear that lowering LDL-C to levels less than 70 mg/dL also reduces both atherosclerotic plaque progression and the rate of first major adverse cardiovascular events in primary prevention in patients at higher global risk.

Did the study prove that reducing hs-CRP lowers risk?

Measuring hs-CRP levels has been extensively studied in apparently healthy populations, stable angina, unstable angina, and other cardiovascular settings.^{18,21,40–43} It has been shown to have significant prognostic implications in a number of primary and secondary trials.⁴⁴ Additionally, those with elevated LDL-C and hs-CRP levels benefit the most from statin therapy.^16,45,46 Animal studies have also provided some evidence that CRP may play a role in atherogenesis.^47,48 However, recent clinical and genetic studies have raised doubt about the direct causal relationship between CRP and coronary artery disease,^23,49,50 and epidemiologic studies have questioned its usefulness as a marker of risk.^51,52

The JUPITER study adds little to clear up the controversy about whether hs-CRP is a mechanistic participant in atherosclerotic disease. However, it also shows that this issue is somewhat irrelevant, in that selection of patients for high-potency statin therapy solely on the basis of high hs-CRP without other indications for lipid-lowering therapy clearly reduces risk and improves survival.

JUPITER did not examine whether people with higher hs-CRP levels benefited more from statin therapy than those with lower levels. The hypothesis-generating data for JUPITER came from an analysis of changes in hs-CRP and LDL-C in the Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS).¹⁶ Thus, JUPITER did not include people with both low LDL-C and low hs-CRP because, in the AFCAPS/TexCAPS analysis, those with low LDL-C and low hs-CRP had extremely low event rates and no clinical efficacy of statin therapy, despite good LDL-C reduction. In marked contrast, those with low LDL-C but elevated hs-CRP had high event rates and large relative risk reductions— hence the need for JUPITER to prospectively test this hypothesis. Nevertheless, the initial results of JUPITER as presented do not yet make it clear that there is a dose-response relationship between higher levels of hs-CRP and a greater reduction in events, even in a cohort with elevated hs-CRP at baseline. This analysis will no doubt be forthcoming in another manuscript from Ridker and colleagues. Specifically, it will be of interest to examine whether those with the highest hs-CRP levels benefited the most from rosuvastatin on both an absolute and relative scale, and whether those with the greatest hs-CRP reduction also benefited more. With the present data available from JUPITER, a reasonable interpretation is that an elevated hs-CRP simply widens the inclusion criterion for those for whom high-potency statin therapy improves clinical outcomes.⁵³

Even with a nonspecific marker such as hs-CRP, patients at higher global risk and with LDL-C below the recommended levels could be identified and treated aggressively. This benefit, however, required that approximately 100 people be treated with rosuvastatin for 2 years to prevent one event. Additionally, only 20% of all patients screened were eligible for the trial. Therefore, one could argue that its generalizability is limited.

Markers of risk that are more specific and sensitive are needed to identify people at higher global risk who would otherwise be considered to be at low risk with the current risk assessment tools. A number of such inflammatory and oxidative markers are under development.^54–60

Absolute vs relative risk reduction and the public health burden

The 44% reduction in the number of primary end point events in the rosuvastatin group was considerable in relative terms. However, in absolute terms, 95 people had to be treated for up to 2 years in order to prevent one event.⁵³ In making recommendations, the United States Department of Health and Human Services has to consider the clinical benefit of a test or a drug in light of its cost. With health care costs increasing, many agencies are refusing to pay for therapies on the basis of cost or small absolute benefit.

While we do not have the answer as to whether treating 95 people for 2 years to see one benefit is cost-effective, one thing is clear: the field of medicine is in desperate need of a better way to identify individuals who may benefit from a test or therapy.⁶¹ Additionally, we think it is important to note that the “numbers-needed-to-treat” (95 at 2 years and 25 at 5 years) derived from JUPITER are actually smaller than the values observed in the AFCAPS/TexCAPS and the West of Scotland Coronary Prevention Study.^62,63 This suggests that statin therapy is at least as cost-effective in those with elevated hs-CRP as in those with elevated LDL-C. Even our most robust therapies are effective in only a minority of patients treated.⁶¹

Should ‘healthy’ people be tested for hs-CRP?

In 2003, we wrote in this journal²¹ that measuring hs-CRP may add to the current risk-prediction models by identifying people at increased risk who would otherwise not be considered as such by current risk models. The US Centers for Disease Control and Prevention and the American Heart Association have also stated that measuring hs-CRP in those at intermediate risk may be reasonable.²⁶

The JUPITER investigators intended to study a relatively healthy population, but, as we mentioned, a close look at the cohort’s baseline characteristics indicates a substantial proportion met the criteria for metabolic syndrome. Therefore, one could challenge whether we really need hs-CRP in such a population to identify who will benefit from statin therapy.

We agree with the recommendation from the Centers for Disease Control and Prevention and the American Heart Association that measuring hs-CRP in people at intermediate risk is a reasonable option.²⁶ We also believe that hs-CRP should be tested as a secondary risk factor, in combination with blood pressure, lipids, diabetes, smoking, serum creatinine, and fasting blood glucose. Factors such as obesity, sedentary lifestyle, family history of heart disease, and emotional and physical stress should also be considered.

Safety of high-dose statin therapy

High-dose statin therapy has been well tolerated in clinical trials, but rates of discontinuation have been higher (7%–10%) than with moderate-dose therapy (4%–5%).⁶⁴ Fortunately, the rates of serious adverse events have in general been low. For example, with simvastatin 80 mg, the rates of myopathy and rhabdomyolysis were quite low.³¹

Rates of elevations in serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) with high-dose statin therapy have been reported to be below 1.3%. Studies have shown that reducing LDL-C to below 100 mg/dL is associated with a higher incidence of ALT and AST elevations. However, these elevations have usually been benign and often return to normal when the drug is reduced in dose or withdrawn.

In previous studies of rosuvastatin,⁶⁵ the incidence of myopathy and liver function abnormalities was less than 0.1%. Rates of proteinuria were similarly low, and in many patients renal function actually improved on rosuvastatin.^66,67 Furthermore, rosuvastatin may have different pharmacokinetic properties than atorvastatin (Lipitor) and simvastatin, which may result in a lower incidence of musculoskeletal toxicity.^68,69

In general, the incidence of cancer has been similar in those treated with high-dose statins and those treated with placebo. The Treating to New Targets trial⁷⁰ suggested that the incidence of cancer was higher with atorvastatin 80 mg daily than with 20 mg daily. However, a meta-analysis of 14 trials of moderate-dose statin therapy did not show any evidence of increased cancer rates with these agents.⁷⁰ Indeed, in JUPITER, there was a reduction in cancer-related mortality rates, which could have been due to chance.

The JUPITER trial also showed an increase in the physician-reported incidence of diabetes mellitus with rosuvastatin. This is an important finding, and it may be a class effect because modest increases have similarly been reported with other statins in other major trials, eg, with pravastatin (Pravachol) in PROSPER, simvastatin in the Heart Protection Study, and atorvastatin in PROVE-IT. However, even in those with diabetes or impaired fasting glucose, the reduction in the rate of major adverse events is significant. For example, in JUPITER, almost all of the cases of “incident diabetes” were in those with impaired fasting glucose at baseline, and this group had nearly a 50% reduction in rates of myocardial infarction, stroke, and cardiovascular death. Therefore, on balance, the modest risk of earlier diagnosis of diabetes with statin therapy seems substantially offset by the marked reduction in rates of major adverse cardiovascular events in people with diabetes and impaired fasting glucose on statin therapy.

TAKE-HOME POINTS

The JUPITER trial, like previous high-dose statin trials, calls into question whether current LDL-C guidelines are appropriate for people at higher global risk with otherwise “normal” LDL-C levels.^27,28 This trial heralds a new era in preventive therapy because it extends beyond LDL-C as an indication for statin therapy within the primary prevention setting. Statins have revolutionized the therapy of cardiovascular disease, and they continue to show benefit even in the “healthy.”

Clearly, hs-CRP serves as a nonlipid marker to identify those who may benefit from statin therapy. Nonetheless, more specific and sensitive markers (or panels) of cardiovascular risk are necessary. In the future, we will need markers that not only identify people at higher global risk, but that also tell us who would benefit from certain medical or surgical therapies. Elevated hs-CRP in a patient who otherwise would not be a candidate for statin therapy should trigger a reassessment of the risks vs benefits of statin therapy—JUPITER teaches us that statin therapy will benefit these patients.

Aggressive lifestyle modification that encompasses a balanced diet, routine exercise, and smoking cessation should be applied in both primary and secondary prevention. Additionally, risk factors such as elevated blood pressure and hyperlipidemia should be aggressively treated with appropriate medications.

The medical community has struggled with two important questions for the past 10 years: When it comes to the low-density lipoprotein cholesterol (LDL-C) level, how low should one go and at what cost? And are there other markers of risk that can identify a higher-risk subpopulation in relatively healthy people? The JUPITER trial (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) provided partial answers for these questions by finding that a highly potent statin lowered the risk of cardiovascular events in patients with “normal” LDL-C but elevated levels of high-sensitivity C-reactive protein (hs-CRP).¹

In this article, we will critically evaluate the methods, results, and conclusions of the JUPITER trial. Additionally, we will discuss its limitations and areas of uncertainty.

BEFORE JUPITER

The LDL-C-lowering drugs called statins have revolutionized cardiovascular medicine.² They are beneficial in both the primary prevention setting and in acute coronary syndromes, stable angina, and unstable angina and can halt the progression of coronary artery disease—in some cases even resulting in modest regression of plaque.^3–6

Many experts have credited the reduction in LDL-C as being the sole factor responsible for the decrease in major adverse events seen with statin therapy.⁷ However, statins have other, non-lipid-lowering properties, including anti-inflammatory and antioxidant effects, that may also contribute to their benefits.^8–15

One of the anti-inflammatory actions of statins is evidenced by lower levels of the acute-phase reactant CRP.^10,11,15,16 Measuring systemic CRP levels with a highly sensitive assay (yielding the so-called high-sensitivity or hs-CRP level) provides significant clinical prognostic value across a spectrum of clinical situations, ranging from risk screening in apparently healthy people to stable and unstable angina.^17–22 People with higher hs-CRP levels are, on average, at higher risk of adverse cardiovascular events. However, controversy remains as to whether hs-CRP plays a mechanistic role in plaque formation and acute complications. Indeed, recent genetic studies argue strongly that hs-CRP lies outside the mechanistic path of atherosclerosis.²³ Nonetheless, an overwhelming amount of data indicates that hs-CRP serves as a marker of disease.^17–21

Nissen et al¹⁰ showed that the rate of progression of atherosclerosis is lower when the levels of atherogenic lipoproteins and hs-CRP are both lowered with statin therapy. Simultaneously, Ridker et al¹¹ showed that patients who have lower hs-CRP levels after statin therapy have better clinical outcomes than those with higher hs-CRP levels, regardless of their achieved level of LDL-C.

Collectively, these studies and others have led some to believe that, in people with relatively low LDL-C but persistently elevated hs-CRP, statin therapy may reduce the rate of events.^15,24 The JUPITER trial was undertaken to test this hypothesis.

JUPITER STUDY DESIGN

JUPITER was designed to see whether highly potent statin therapy is beneficial in people with elevated hs-CRP who otherwise do not meet the criteria for lipid-lowering therapy. The study was conducted at 1,315 sites in 26 countries. It was sponsored by AstraZeneca, the maker of rosuvastatin (Crestor).

Inclusion and exclusion criteria

All participants had to be free of known cardiovascular disease, have an LDL-C level lower than 130 mg/dL, and have an hs-CRP level of 2.0 mg/L or greater. Patients were excluded if they were previous or current users of lipid-lowering drugs; had severe arthritis, lupus, or inflammatory bowel disease; or were taking immune-modulating drugs such as cyclosporine (Sandimmune, others), tacrolimus (Prograf), azathioprine (Azasan, Imuran), or long-term oral corticosteroids.

Rosuvastatin therapy

Participants were randomly assigned in a 1:1 ratio to receive rosuvastatin 20 mg daily or a matching placebo in a double-blind fashion.

End points

The primary end point was the composite of nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina, an arterial revascularization procedure, or confirmed death from cardiovascular causes. Secondary end points were the individual components of the primary end point.

Statistical analysis

The study was powered to detect a 25% reduction in the primary end point among those treated with rosuvastatin. The trial was designed to run until 520 end point events had occurred. However, on March 29, 2008, after the first prespecified interim analysis, the Data and Safety Monitoring Board stopped the trial due to a significant reduction in the primary end point in the rosuvastatin group. As in most randomized clinical trials, all analyses were done on an intention-to-treat basis. Prespecified subgroup analyses were also performed.

STUDY RESULTS

Patient recruitment and eligibility

Between February 4, 2003, and December 15, 2006, a total of 89,890 people were screened. Of these, 17,802 met the inclusion and exclusion criteria and were included in the study. Of the 72,088 people who were excluded, 25,993 (36.1%) had an hs-CRP level below 2 mg/L and 37,611 (52.2%) had an LDL-C level of 130 mg/dL or higher.

A not-so-healthy population

The aim of the investigators was to include relatively healthy people. The median age was 66 years, about 16% of participants were current smokers, about 11% had a family history of heart disease, and about 41% met the criteria for metabolic syndrome, all conditions that are associated with elevated hs-CRP.²⁵ Of note, the median hs-CRP level was 4.2 mg/L, a level indicating higher global risk according to the American College of Cardiology/American Heart Association consensus statement.²⁶

Reduction in lipid levels and hs-CRP

By 12 months, in the rosuvastatin group, the median LDL-C level had fallen by 50% (from 108 to 55 mg/dL), and the median hs-CRP level had fallen by 37% (from 4.2 to 2.2 mg/L). Additionally, the triglyceride level had fallen by 17%. The high-density lipoprotein cholesterol levels did not change significantly.

Impact on end points

Adverse events

Ridker PM, et al. Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein. N Engl J Med 2008; 359:2195-2207. Copyright 2008 Massachusetts Medical Society. All rights reserved.

WHAT DOES THIS MEAN?

Is lower LDL-C better?

The JUPITER trial is the latest of several statin trials that have shown significant reductions in major adverse cardiovascular events when LDL-C was lowered below what has been recommended by the current guidelines.^27,28

In 2002, the Heart Protection Study²⁹ showed a significant reduction in major adverse cardiovascular events in patients at high risk of coronary artery disease if they received simvastatin (Zocor), even if they had LDL-C levels lower than 100 mg/dL at baseline. Similarly, the Pravastatin or Atorvastatin Evaluation and Infection-Thrombolysis in Myocardial Infarction 22 (PROVE-IT TIMI 22) trial³⁰ showed a 16% relative risk reduction in a composite end point in patients presenting with acute coronary syndrome if they received intensive statin therapy.

These two studies led to an update by the National Cholesterol Education Program (Adult Treatment Panel III), suggesting an optimal LDL-C goal of less than 70 mg/dL in those with coronary artery disease or its risk equivalent (ie, diabetes mellitus, peripheral vascular disease). Furthermore, in support of the “lower is better” theory, a number of studies that used intravascular ultrasonography have shown regression of coronary plaque with aggressive LDL-C lowering. Notably, in a Study to Evaluate the Effect of Rosuvastatin on Intravascular Ultrasound-Derived Coronary Atheroma Burden (the ASTEROID trial),⁵ rosuvastatin 40 mg daily caused significant plaque regression while lowering LDL-C to 61 mg/dL over a 24-month period.

A number of high-dose statin trials have shown that lowering LDL-C to less than 70 mg/dL significantly reduces major adverse cardiovascular events.^31–39 The JUPITER trial was unique in that it extended these findings to people without known coronary disease (ie, primary prevention) or elevated cholesterol but with elevated levels of a marker of inflammation—hs-CRP. In view of the JUPITER results and of studies using intravascular ultrasonography in the primary prevention setting, it seems clear that lowering LDL-C to levels less than 70 mg/dL also reduces both atherosclerotic plaque progression and the rate of first major adverse cardiovascular events in primary prevention in patients at higher global risk.

Did the study prove that reducing hs-CRP lowers risk?

Measuring hs-CRP levels has been extensively studied in apparently healthy populations, stable angina, unstable angina, and other cardiovascular settings.^{18,21,40–43} It has been shown to have significant prognostic implications in a number of primary and secondary trials.⁴⁴ Additionally, those with elevated LDL-C and hs-CRP levels benefit the most from statin therapy.^16,45,46 Animal studies have also provided some evidence that CRP may play a role in atherogenesis.^47,48 However, recent clinical and genetic studies have raised doubt about the direct causal relationship between CRP and coronary artery disease,^23,49,50 and epidemiologic studies have questioned its usefulness as a marker of risk.^51,52

The JUPITER study adds little to clear up the controversy about whether hs-CRP is a mechanistic participant in atherosclerotic disease. However, it also shows that this issue is somewhat irrelevant, in that selection of patients for high-potency statin therapy solely on the basis of high hs-CRP without other indications for lipid-lowering therapy clearly reduces risk and improves survival.

JUPITER did not examine whether people with higher hs-CRP levels benefited more from statin therapy than those with lower levels. The hypothesis-generating data for JUPITER came from an analysis of changes in hs-CRP and LDL-C in the Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS).¹⁶ Thus, JUPITER did not include people with both low LDL-C and low hs-CRP because, in the AFCAPS/TexCAPS analysis, those with low LDL-C and low hs-CRP had extremely low event rates and no clinical efficacy of statin therapy, despite good LDL-C reduction. In marked contrast, those with low LDL-C but elevated hs-CRP had high event rates and large relative risk reductions— hence the need for JUPITER to prospectively test this hypothesis. Nevertheless, the initial results of JUPITER as presented do not yet make it clear that there is a dose-response relationship between higher levels of hs-CRP and a greater reduction in events, even in a cohort with elevated hs-CRP at baseline. This analysis will no doubt be forthcoming in another manuscript from Ridker and colleagues. Specifically, it will be of interest to examine whether those with the highest hs-CRP levels benefited the most from rosuvastatin on both an absolute and relative scale, and whether those with the greatest hs-CRP reduction also benefited more. With the present data available from JUPITER, a reasonable interpretation is that an elevated hs-CRP simply widens the inclusion criterion for those for whom high-potency statin therapy improves clinical outcomes.⁵³

Even with a nonspecific marker such as hs-CRP, patients at higher global risk and with LDL-C below the recommended levels could be identified and treated aggressively. This benefit, however, required that approximately 100 people be treated with rosuvastatin for 2 years to prevent one event. Additionally, only 20% of all patients screened were eligible for the trial. Therefore, one could argue that its generalizability is limited.

Markers of risk that are more specific and sensitive are needed to identify people at higher global risk who would otherwise be considered to be at low risk with the current risk assessment tools. A number of such inflammatory and oxidative markers are under development.^54–60

Absolute vs relative risk reduction and the public health burden

The 44% reduction in the number of primary end point events in the rosuvastatin group was considerable in relative terms. However, in absolute terms, 95 people had to be treated for up to 2 years in order to prevent one event.⁵³ In making recommendations, the United States Department of Health and Human Services has to consider the clinical benefit of a test or a drug in light of its cost. With health care costs increasing, many agencies are refusing to pay for therapies on the basis of cost or small absolute benefit.

While we do not have the answer as to whether treating 95 people for 2 years to see one benefit is cost-effective, one thing is clear: the field of medicine is in desperate need of a better way to identify individuals who may benefit from a test or therapy.⁶¹ Additionally, we think it is important to note that the “numbers-needed-to-treat” (95 at 2 years and 25 at 5 years) derived from JUPITER are actually smaller than the values observed in the AFCAPS/TexCAPS and the West of Scotland Coronary Prevention Study.^62,63 This suggests that statin therapy is at least as cost-effective in those with elevated hs-CRP as in those with elevated LDL-C. Even our most robust therapies are effective in only a minority of patients treated.⁶¹

Should ‘healthy’ people be tested for hs-CRP?

In 2003, we wrote in this journal²¹ that measuring hs-CRP may add to the current risk-prediction models by identifying people at increased risk who would otherwise not be considered as such by current risk models. The US Centers for Disease Control and Prevention and the American Heart Association have also stated that measuring hs-CRP in those at intermediate risk may be reasonable.²⁶

The JUPITER investigators intended to study a relatively healthy population, but, as we mentioned, a close look at the cohort’s baseline characteristics indicates a substantial proportion met the criteria for metabolic syndrome. Therefore, one could challenge whether we really need hs-CRP in such a population to identify who will benefit from statin therapy.

We agree with the recommendation from the Centers for Disease Control and Prevention and the American Heart Association that measuring hs-CRP in people at intermediate risk is a reasonable option.²⁶ We also believe that hs-CRP should be tested as a secondary risk factor, in combination with blood pressure, lipids, diabetes, smoking, serum creatinine, and fasting blood glucose. Factors such as obesity, sedentary lifestyle, family history of heart disease, and emotional and physical stress should also be considered.

Safety of high-dose statin therapy

High-dose statin therapy has been well tolerated in clinical trials, but rates of discontinuation have been higher (7%–10%) than with moderate-dose therapy (4%–5%).⁶⁴ Fortunately, the rates of serious adverse events have in general been low. For example, with simvastatin 80 mg, the rates of myopathy and rhabdomyolysis were quite low.³¹

Rates of elevations in serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) with high-dose statin therapy have been reported to be below 1.3%. Studies have shown that reducing LDL-C to below 100 mg/dL is associated with a higher incidence of ALT and AST elevations. However, these elevations have usually been benign and often return to normal when the drug is reduced in dose or withdrawn.

In previous studies of rosuvastatin,⁶⁵ the incidence of myopathy and liver function abnormalities was less than 0.1%. Rates of proteinuria were similarly low, and in many patients renal function actually improved on rosuvastatin.^66,67 Furthermore, rosuvastatin may have different pharmacokinetic properties than atorvastatin (Lipitor) and simvastatin, which may result in a lower incidence of musculoskeletal toxicity.^68,69

In general, the incidence of cancer has been similar in those treated with high-dose statins and those treated with placebo. The Treating to New Targets trial⁷⁰ suggested that the incidence of cancer was higher with atorvastatin 80 mg daily than with 20 mg daily. However, a meta-analysis of 14 trials of moderate-dose statin therapy did not show any evidence of increased cancer rates with these agents.⁷⁰ Indeed, in JUPITER, there was a reduction in cancer-related mortality rates, which could have been due to chance.

The JUPITER trial also showed an increase in the physician-reported incidence of diabetes mellitus with rosuvastatin. This is an important finding, and it may be a class effect because modest increases have similarly been reported with other statins in other major trials, eg, with pravastatin (Pravachol) in PROSPER, simvastatin in the Heart Protection Study, and atorvastatin in PROVE-IT. However, even in those with diabetes or impaired fasting glucose, the reduction in the rate of major adverse events is significant. For example, in JUPITER, almost all of the cases of “incident diabetes” were in those with impaired fasting glucose at baseline, and this group had nearly a 50% reduction in rates of myocardial infarction, stroke, and cardiovascular death. Therefore, on balance, the modest risk of earlier diagnosis of diabetes with statin therapy seems substantially offset by the marked reduction in rates of major adverse cardiovascular events in people with diabetes and impaired fasting glucose on statin therapy.

TAKE-HOME POINTS

The JUPITER trial, like previous high-dose statin trials, calls into question whether current LDL-C guidelines are appropriate for people at higher global risk with otherwise “normal” LDL-C levels.^27,28 This trial heralds a new era in preventive therapy because it extends beyond LDL-C as an indication for statin therapy within the primary prevention setting. Statins have revolutionized the therapy of cardiovascular disease, and they continue to show benefit even in the “healthy.”

Clearly, hs-CRP serves as a nonlipid marker to identify those who may benefit from statin therapy. Nonetheless, more specific and sensitive markers (or panels) of cardiovascular risk are necessary. In the future, we will need markers that not only identify people at higher global risk, but that also tell us who would benefit from certain medical or surgical therapies. Elevated hs-CRP in a patient who otherwise would not be a candidate for statin therapy should trigger a reassessment of the risks vs benefits of statin therapy—JUPITER teaches us that statin therapy will benefit these patients.

Aggressive lifestyle modification that encompasses a balanced diet, routine exercise, and smoking cessation should be applied in both primary and secondary prevention. Additionally, risk factors such as elevated blood pressure and hyperlipidemia should be aggressively treated with appropriate medications.

The Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial^1–5 was designed primarily to address, in patients with type 2 diabetes at high risk of cardiovascular events, whether intensive glucose control would result in a lower risk of atherosclerotic disease events or death than would standard treatment.

It was widely expected that intensive treatment would confer either modest benefit or, at worst, no benefit. However, the glucose-lowering arm of the trial was terminated early because of a higher mortality rate in the intensively treated group. (The ACCORD trial has two other arms, which concern blood pressure and lipid-lowering, and these are continuing.)

In earlier trials in type 2 diabetes, concerns had been raised about an increased risk of cardiovascular events and possibly death associated with glucose-lowering drugs, hypoglycemia itself, or both, and these were well known when ACCORD was convened. ACCORD was very carefully designed and included careful adjudication of each cardiovascular event and death, including whether hypoglycemia might have been a proximate cause of some sudden deaths.⁵

Therefore, the surprising result of the higher mortality rate with intensive glycemic control in ACCORD will be fodder for discussion in many arenas over the next several years, and it poses some challenges for physicians and patients in determining treatment goals, as well as for organizations that write clinical practice guidelines (and perhaps organizations involved in pay-for-performance based on these guidelines).

Still, I believe that the ACCORD results should not substantially change our approach to treatment goals in type 2 diabetes, although hemoglobin A_1c targets below 6% may not have much added value for cardiovascular risk reduction. The low overall mortality rate in all the arms of the ACCORD trial emphasizes the importance of lifestyle modification, lipid and blood pressure therapy, and encouragement of aspirin use in all patients with type 2 diabetes.

This article reflects my views as a practicing diabetologist and clinical trialist (I was an investigator in the ACCORD trial) with a long-standing interest in clinical trials and in how the results influence clinical practice. The views I express herein may not reflect the views of other ACCORD investigators, the National Heart, Lung, and Blood Institute (NHLBI), the ACCORD trial coordinating center at Wake Forest University, or its data safety and monitoring board.

RISK OF CORONARY DISEASE INCREASES WITH GLUCOSE

Many observational studies^6–10 have shown that the risk of cardiovascular disease, especially coronary heart disease, is two to five times higher in people with diabetes mellitus than in people without diabetes. The risk appears to be continuous, so the higher one’s glucose or hemoglobin A_1c, the higher the risk.⁶ This risk even extends to glucose values well below the threshold values currently used to diagnose diabetes mellitus.⁶ Since there is no glucose threshold for coronary heart disease, the term dysglycemia (rather than hyperglycemia) has been proposed to note the relationship between glucose and coronary heart disease. (The glucose threshold for microvascular complications of diabetes, such as retinopathy and nephropathy, appears to be between 110 and 126 mg/dL).

The clustering of multiple coronary risk factors such as obesity, dyslipidemia, and hypertension has always raised the question of whether glucose is a culprit in coronary risk or whether it simply “runs in bad company.”

EARLIER CLINICAL TRIALS SUGGEST INTENSIVE TREATMENT RAISES RISK

Even though it has been widely believed that intensive glucose-lowering would reduce cardiovascular risk in type 2 diabetes, there have been hints in previous studies that some intensive-treatment regimens might increase risk.

Two large randomized clinical trials and one small one (discussed below) addressed whether glucose control would reduce the risk of atherosclerotic vascular disease events. In each of them, an increased risk of cardiovascular events and possibly of death was seen in at least one intensively treated group.

In the following discussion, I have calculated all of the death rates as the number of deaths per 1,000 patients per year, based on published study results. In this way, we can compare the rates in the various studies (including ACCORD), regardless of the trial duration.

The university group Diabetes Program: Controvery about tolbutamide therapy

The University Group Diabetes Program (UGDP)^11–16 included about 1,000 participants randomized to five treatments: tolbutamide (Orinase, a sulfonylurea), insulin in a fixed dose based on body weight, insulin in adjusted doses based on fasting glucose levels, placebo, and (later) phenformin.

In the 1970s, when the UGDP was carried out, randomized clinical trials were uncommon. Like other trials from that era, the UGDP was underpowered by today’s standards and did not have a data safety and monitoring board.

Rates of cardiovascular events and deaths (per 1,000 patient-years):

• 25 (tolbutamide group)
• 12 (placebo group).

The two insulin groups did not differ from the placebo group in their rates of cardiovascular events or death.¹⁵ The tolbutamide arm was stopped, and the ensuing controversy about how to interpret the trial results lasted for more than a decade. It also resulted in a black-box warning for tolbutamide and all subsequent sulfonylureas.

The United Kingdom Prospective Diabetes Study (UKPDS)^17–27 was launched in 1977. A cohort of 5,102 patients (mean age 54 years) with newly diagnosed type 2 diabetes mellitus followed a “prudent diet” for the first 3 to 4 months. Then, if their fasting glucose levels were in the range of 6.1 to 15 mmol/L (110–270 mg/dL), they were randomized to receive various treatments.

Patients who were not obese were randomized to receive either intensive treatment or conventional treatment. The intensive-treatment group received either insulin or a sulfonylurea (chlorpropamide [Diabinese], glibenclamide, or glipizide [Glucotrol]); the conventional-treatment group received diet therapy. The sulfonylurea arm was included partly to address the UGDP results.

Patients who were obese were randomized to receive one of three treatments: intensive treatment (with the agents listed above), conventional treatment, or metformin (Fortamet, Glucophage).

The mean in-trial hemoglobin A_1c level in the intensive-treatment group was 7.0%, compared with 7.9% in the conventional-treatment group.

After a mean follow-up of more than 10 years, the incidence of myocardial infarction was 16% lower in the intensive-treatment group, but the difference was not statistically significant (P = .052).

Rates of death from all causes among nonobese subjects (per 1,000 patient-years):

• 18.2–20.5 (intensive-treatment group)
• 19.9 (conventional-treatment group).

In the obese patients who received metformin, the incidence of myocardial infarction was lower than in the conventional-treatment group but not the intensive-treatment group.

Rates of death among obese patients (per 1,000 patient-years):

• 13.5 (metformin group)
• 18.9 (intensive-treatment group)
• 20.6 (conventional-treatment group).

However, a small subset (n = 587) of the original group assigned to sulfonylurea therapy whose glycemic control deteriorated during the trial were rerandomized to continue to receive a sulfonylurea alone or to have metformin added. There was a statistically significantly higher rate of cardiovascular events and a nonsignificantly higher rate of total mortality in the metformin-plus-sulfonylurea group (30.3 per 1,000 patient-years) than in the sulfonylurea-only group (19.1 per 1,000 patient-years).

These data suggested that the way glucose-lowering was achieved might be as important as the glucose levels actually achieved. However, no definite conclusions could be drawn.

In an editorial on the UKPDS, Nathan²⁶ made a comment that may have been prescient in terms of the ACCORD trial: “Professional organizations will now scramble to decide how to translate the UKPDS results … Whether the UKPDS firmly establishes the choice of any one therapy…or any combination of therapies for the long-term treatment of type 2 diabetes is more questionable.”²⁶

Veterans Administration feasibility study

A Veterans Administration feasibility study^28,29 included 153 men (mean age 60) with type 2 diabetes (mean duration 7.8 years) who received either conventional therapy (a single daily dose of insulin) or intensive therapy (multiple doses of insulin plus a sulfonylurea). Over a mean of 27 months, the intensive-therapy group achieved a hemoglobin A_1c level that was 2 percentage points lower than in the conventional-therapy group.

At 2.25 years of follow-up, cardiovascular events had occurred in 24 (24%) of the intensive-therapy group and in 16 (20%) of the standard-therapy group (P = .10).

Rates of death from all causes (per 1,000 patient-years):

• 28.9 (intensive-treatment group)
• 17.5 (conventional-treatment group).

ACCORD TRIAL DESIGN

The primary outcome measured was the combined incidence of nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes. Secondary outcomes included death from any cause. The study is also evaluating the effect of intensive treatment on microvascular disease, hypoglycemia, cognition, quality of life, and cost-effectiveness.

The ACCORD study was designed to have 89% power to detect a 15% treatment effect of intensive glycemic control compared with standard glycemic control for the primary end point.

ACCORD RESULTS

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545-2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545-2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545–2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

Rates of death from any cause (per 1,000 patient-years):

• 14 (intensive-treatment group)
• 11 (standard-treatment group).

In the analyses available at the time that this study arm closed, the excess mortality was not attributable to any particular treatment regimen. In particular, rosiglitazone (Avandia) use did not contribute to the excess mortality. (Of note, 91.2% of the intensive-treatment group and 57.5% of the conventional-treatment group had been treated with rosiglitazone, with more than 19,000 patient-years of rosiglitazone exposure). The excess mortality was also not attributable to hypoglycemia immediately proximate to the death.

The ACCORD trial’s data safety and monitoring board recommended that this arm of the study be discontinued for safety reasons, and this recommendation was accepted by the NHLBI project office. All participants were notified by letter before the trial results were announced publicly, and all intensive-therapy group participants are now being treated by the protocol used in the standard-therapy group.¹

FEWER DEATHS IN ACCORD THAN IN OTHER STUDIES IN DIABETES

The mortality rates in both arms of ACCORD were much lower than in other observational studies and clinical trials in type 2 diabetes.

The National Health and Nutrition Education Survey (NHANES),³⁰ conducted from 1971 to 1975, included 14,374 people with diabetes between the ages of 25 and 74. Many of them were younger than the ACCORD patients, but two NHANES age-groups overlapped the ACCORD cohort. Rates of death from any cause at 22 years (per 1,000 patient-years):

• 39.7 (ages 45–64)
• 89.7 (ages 65–74).

The NHANES cohort would not have been treated as vigorously for coronary risk and other common causes of death.

UGDP, UKPDS. Death rates in the glucose-lowering trials of type 2 diabetes mellitus cited above were typically in the range of 20 deaths per 1,000 patient-years but were as high as 30 deaths per 1,000 patient-years in the UGDP tolbutamide group¹⁶ and the UK-PDS sulfonylurea-plus-metformin group.^20,22,26

Steno-2.³¹ Half of 160 patients with type 2 diabetes were randomized to intensive strategies for controlling glucose, lipids, and blood pressure and for taking aspirin and angiotensin-converting enzyme inhibitors and following a healthy lifestyle. The other half received conventional therapy. Even in the intensive-treatment group, the mortality rate at 13 years was higher than in ACCORD. Rates of death from any cause (per 1,000 patient years):

• 22.5 (intensive-treatment group)
• 37.6 (conventional-treatment group).

After the ACCORD results were presented, two other trials addressing the question of whether lower hemoglobin A_1c would reduce cardiovascular risk in type 2 diabetes have reported their outcomes:

The ADVANCE trial (Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation),^32,33 with 11,140 patients, had a target hemoglobin A_1c of 6.5% in an intensive-treatment group and 7.3% in a usual-treatment group. The intensive-treatment group showed no difference in the rates of major macrovascular events (HR 0.94, 95% CI 0.84–1.06, P = .32) or all-cause mortality (HR 0.93, 95% CI 0.83–1.06, P = .32). The overall death rate in ADVANCE (about 18 deaths per 1,000 patient-years) was higher than in ACCORD.

The Veterans Administration Diabetes Trial included 1,791 patients.³⁴ Like the ADVANCE trial, it also found no difference in major cardiovascular outcomes (HR 0.868, P = .11) or cardiovascular mortality rates (HR 1.258, P = .36) with intensive therapy vs conventional therapy, ie, achieved hemoglobin A_1c levels of 6.9% vs 8.4% (presented at the American Diabetes Association 2008 Scientific Sessions). Hypoglycemia was associated with an increased risk of death in the standard-treatment group.

An analysis suggested that patients with a shorter duration of diabetes may have had cardiovascular benefit from intensive glucose-lowering, while those who had had it longer may have had increased risk associated with the more intensive therapy. The rate of death from all causes appears to have been higher than in ACCORD, but this could not be determined accurately from the presentations.

Comment. Thus, the ACCORD cohort as a whole has had strikingly lower death rates than in these other studies. The fact that all participants had lower glucose levels on therapy than at baseline may possibly contribute to these lower death rates. In addition, all ACCORD participants in the lipid arm received a statin; all participants in the blood pressure arm had their blood pressure lowered to levels below those commonly seen in clinical practice; participants were encouraged to exercise regularly; most participants were given diet instruction; and other healthy behaviors such as aspirin use, regular follow-up with primary care physicians, and recommendations about smoking were encouraged throughout the study. These comprehensive strategies may represent better care and thus result in lower death rates than in other studies.

POSSIBLE EXPLANATIONS FOR THE ACCORD OUTCOMES

The ACCORD trial has already stimulated fierce debate about the reasons for the higher mortality rate in the intensive-treatment group. With longer follow-up, some new risk factors for death may be identified that are not evident in the analyses of the current 460 deaths. What follows are some of my thoughts, with the caveat that they are not confirmed (supported statistically) by any currently available analyses from ACCORD.

It seems unlikely that lower glucose values as reflected by lower hemoglobin A_1c values in the intensive-treatment group are an a priori explanation for the observed differences in mortality rates—especially since the mortality rates were lower than in the NHANES and clinical trial data sets cited above. If we assume that a type 1 statistical error (finding a difference where no difference actually exists) does not explain the findings, then at least four reasonable postulates exist:

Hypoglycemia may have some adverse effect, either acutely or from recurrent events that trigger a catecholamine response with associated risk for arrhythmia or increased coronary heart disease risk. However, the investigators analyzed each death to determine whether hypoglycemia was a contributing cause, and they found no statistically significant relationship between hypoglycemia and death in the intensive-treatment group.

Weight gain is common with intensive therapy. Obesity may be associated with greater cytokine production, higher concentrations of clotting factors, higher levels of free fatty acids, and other potential contributors to the risk of coronary heart disease and death. Currently, the ACCORD analyses do not suggest that weight gain explains the higher death rate.

Medications such as rosiglitazone, sulfonylureas, and the combination of a sulfonylurea plus metformin have been previously associated with increased death rates in some observational and intervention trials. These studies had some serious methodologic limitations (eg, absence of risk adjustment, events not adjudicated, small study cohorts, wide variation in study cohort characteristics) and small numbers of events.^{11–13,16,26,35} ACCORD analyses have not shown that any single glucose-lowering agent—including rosiglitazone—or combination of agents explains the death rates.

The stress of maintaining glycemic control has been speculated to have in some way contributed to an increased risk. To achieve intensive control, patients had to have frequent contact with their health care providers, they were often told that their hemoglobin A_1c values were “too high” even when they were well below those in the American Diabetes Association guidelines, and they had to follow complex glucose-lowering regimens.

Semiquantitative measures of overall attitudes about health exist (eg, the “Feeling Thermometer” scale), but stress was not measured quantitatively in the ACCORD trial.

IMPLICATIONS OF ACCORD

In practice, most clinicians believe that the target glucose level in patients with type 2 diabetes should be as low as safely possible. This approach does not need to be modified on the basis of current information from ACCORD.

To be safe, regimens should be associated with a low risk of hypoglycemia and a low risk of weight gain. Use of combinations of medications that work by different mechanisms is still prudent. Agents should be used that may have favorable effects on other cardiovascular risk factors (eg, lipids, blood pressure, visceral fat).

Hemoglobin A_1c targets below 7% are not precluded in all patients on the basis of the ACCORD results, though values lower than 6% may not have much added benefit for cardiovascular risk reduction. We should note that hemoglobin A_1c was reduced in all ACCORD participants and that death rates were lower than in many other type 2 diabetic cohorts. Pending data on other outcomes in ACCORD (nephropathy, retinopathy, dementia, fracture risk), I believe it is premature for organizations to change their proposed hemoglobin A_1c targets,^36,37 as none have proposed values as low as the target in the ACCORD intensive-treatment group. At present, no class of glucose-lowering agents needs to be excluded from consideration on the basis of the ACCORD data.

The overall low rates of death in this population at high risk of coronary heart disease deserve comment. Not only are they lower than in other glucose-lowering trials, but they are also lower than in a number of studies of mortality in diabetes cohorts. As noted above, multiple risk factors for coronary heart disease and death were (and are) addressed in the ACCORD study participants, including repeated recommendation for lifestyle modification, intervention arms with lipid and blood pressure therapy, encouragement of aspirin use, and regular follow-up with health care providers for risk factors not managed by the ACCORD trial protocol. It is likely that multiple approaches to reducing the risk of cardiovascular disease contributed to this low mortality rate and that similar approaches will reduce the risk of coronary disease and death in regular clinical practice.

The ACCORD lipid and blood pressure arms are continuing, with results expected in 2010. The future results from ACCORD as well as from several glucose-lowering trials currently in progress (ADVANCE,^32,33 Veteran’s Administration,³⁴ Bypass Angioplasty Revascularization Investigation 2 Diabetes [BARI-2D]³⁸) will likely help refine our understanding of the effects of glucose-lowering, glucose-lowering strategies and targets, and multiple interventions on coronary events and all-cause mortality.

For now, any strategy that lowers glucose and is associated with a low risk of hypoglycemia and does not cause excessive weight gain should be considered appropriate in patients with type 2 diabetes.

The Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial^1–5 was designed primarily to address, in patients with type 2 diabetes at high risk of cardiovascular events, whether intensive glucose control would result in a lower risk of atherosclerotic disease events or death than would standard treatment.

It was widely expected that intensive treatment would confer either modest benefit or, at worst, no benefit. However, the glucose-lowering arm of the trial was terminated early because of a higher mortality rate in the intensively treated group. (The ACCORD trial has two other arms, which concern blood pressure and lipid-lowering, and these are continuing.)

In earlier trials in type 2 diabetes, concerns had been raised about an increased risk of cardiovascular events and possibly death associated with glucose-lowering drugs, hypoglycemia itself, or both, and these were well known when ACCORD was convened. ACCORD was very carefully designed and included careful adjudication of each cardiovascular event and death, including whether hypoglycemia might have been a proximate cause of some sudden deaths.⁵

Therefore, the surprising result of the higher mortality rate with intensive glycemic control in ACCORD will be fodder for discussion in many arenas over the next several years, and it poses some challenges for physicians and patients in determining treatment goals, as well as for organizations that write clinical practice guidelines (and perhaps organizations involved in pay-for-performance based on these guidelines).

Still, I believe that the ACCORD results should not substantially change our approach to treatment goals in type 2 diabetes, although hemoglobin A_1c targets below 6% may not have much added value for cardiovascular risk reduction. The low overall mortality rate in all the arms of the ACCORD trial emphasizes the importance of lifestyle modification, lipid and blood pressure therapy, and encouragement of aspirin use in all patients with type 2 diabetes.

This article reflects my views as a practicing diabetologist and clinical trialist (I was an investigator in the ACCORD trial) with a long-standing interest in clinical trials and in how the results influence clinical practice. The views I express herein may not reflect the views of other ACCORD investigators, the National Heart, Lung, and Blood Institute (NHLBI), the ACCORD trial coordinating center at Wake Forest University, or its data safety and monitoring board.

RISK OF CORONARY DISEASE INCREASES WITH GLUCOSE

Many observational studies^6–10 have shown that the risk of cardiovascular disease, especially coronary heart disease, is two to five times higher in people with diabetes mellitus than in people without diabetes. The risk appears to be continuous, so the higher one’s glucose or hemoglobin A_1c, the higher the risk.⁶ This risk even extends to glucose values well below the threshold values currently used to diagnose diabetes mellitus.⁶ Since there is no glucose threshold for coronary heart disease, the term dysglycemia (rather than hyperglycemia) has been proposed to note the relationship between glucose and coronary heart disease. (The glucose threshold for microvascular complications of diabetes, such as retinopathy and nephropathy, appears to be between 110 and 126 mg/dL).

The clustering of multiple coronary risk factors such as obesity, dyslipidemia, and hypertension has always raised the question of whether glucose is a culprit in coronary risk or whether it simply “runs in bad company.”

EARLIER CLINICAL TRIALS SUGGEST INTENSIVE TREATMENT RAISES RISK

Even though it has been widely believed that intensive glucose-lowering would reduce cardiovascular risk in type 2 diabetes, there have been hints in previous studies that some intensive-treatment regimens might increase risk.

Two large randomized clinical trials and one small one (discussed below) addressed whether glucose control would reduce the risk of atherosclerotic vascular disease events. In each of them, an increased risk of cardiovascular events and possibly of death was seen in at least one intensively treated group.

In the following discussion, I have calculated all of the death rates as the number of deaths per 1,000 patients per year, based on published study results. In this way, we can compare the rates in the various studies (including ACCORD), regardless of the trial duration.

The university group Diabetes Program: Controvery about tolbutamide therapy

The University Group Diabetes Program (UGDP)^11–16 included about 1,000 participants randomized to five treatments: tolbutamide (Orinase, a sulfonylurea), insulin in a fixed dose based on body weight, insulin in adjusted doses based on fasting glucose levels, placebo, and (later) phenformin.

In the 1970s, when the UGDP was carried out, randomized clinical trials were uncommon. Like other trials from that era, the UGDP was underpowered by today’s standards and did not have a data safety and monitoring board.

Rates of cardiovascular events and deaths (per 1,000 patient-years):

• 25 (tolbutamide group)
• 12 (placebo group).

The two insulin groups did not differ from the placebo group in their rates of cardiovascular events or death.¹⁵ The tolbutamide arm was stopped, and the ensuing controversy about how to interpret the trial results lasted for more than a decade. It also resulted in a black-box warning for tolbutamide and all subsequent sulfonylureas.

The United Kingdom Prospective Diabetes Study (UKPDS)^17–27 was launched in 1977. A cohort of 5,102 patients (mean age 54 years) with newly diagnosed type 2 diabetes mellitus followed a “prudent diet” for the first 3 to 4 months. Then, if their fasting glucose levels were in the range of 6.1 to 15 mmol/L (110–270 mg/dL), they were randomized to receive various treatments.

Patients who were not obese were randomized to receive either intensive treatment or conventional treatment. The intensive-treatment group received either insulin or a sulfonylurea (chlorpropamide [Diabinese], glibenclamide, or glipizide [Glucotrol]); the conventional-treatment group received diet therapy. The sulfonylurea arm was included partly to address the UGDP results.

Patients who were obese were randomized to receive one of three treatments: intensive treatment (with the agents listed above), conventional treatment, or metformin (Fortamet, Glucophage).

The mean in-trial hemoglobin A_1c level in the intensive-treatment group was 7.0%, compared with 7.9% in the conventional-treatment group.

After a mean follow-up of more than 10 years, the incidence of myocardial infarction was 16% lower in the intensive-treatment group, but the difference was not statistically significant (P = .052).

Rates of death from all causes among nonobese subjects (per 1,000 patient-years):

• 18.2–20.5 (intensive-treatment group)
• 19.9 (conventional-treatment group).

In the obese patients who received metformin, the incidence of myocardial infarction was lower than in the conventional-treatment group but not the intensive-treatment group.

Rates of death among obese patients (per 1,000 patient-years):

• 13.5 (metformin group)
• 18.9 (intensive-treatment group)
• 20.6 (conventional-treatment group).

However, a small subset (n = 587) of the original group assigned to sulfonylurea therapy whose glycemic control deteriorated during the trial were rerandomized to continue to receive a sulfonylurea alone or to have metformin added. There was a statistically significantly higher rate of cardiovascular events and a nonsignificantly higher rate of total mortality in the metformin-plus-sulfonylurea group (30.3 per 1,000 patient-years) than in the sulfonylurea-only group (19.1 per 1,000 patient-years).

These data suggested that the way glucose-lowering was achieved might be as important as the glucose levels actually achieved. However, no definite conclusions could be drawn.

In an editorial on the UKPDS, Nathan²⁶ made a comment that may have been prescient in terms of the ACCORD trial: “Professional organizations will now scramble to decide how to translate the UKPDS results … Whether the UKPDS firmly establishes the choice of any one therapy…or any combination of therapies for the long-term treatment of type 2 diabetes is more questionable.”²⁶

Veterans Administration feasibility study

A Veterans Administration feasibility study^28,29 included 153 men (mean age 60) with type 2 diabetes (mean duration 7.8 years) who received either conventional therapy (a single daily dose of insulin) or intensive therapy (multiple doses of insulin plus a sulfonylurea). Over a mean of 27 months, the intensive-therapy group achieved a hemoglobin A_1c level that was 2 percentage points lower than in the conventional-therapy group.

At 2.25 years of follow-up, cardiovascular events had occurred in 24 (24%) of the intensive-therapy group and in 16 (20%) of the standard-therapy group (P = .10).

Rates of death from all causes (per 1,000 patient-years):

• 28.9 (intensive-treatment group)
• 17.5 (conventional-treatment group).

ACCORD TRIAL DESIGN

The primary outcome measured was the combined incidence of nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes. Secondary outcomes included death from any cause. The study is also evaluating the effect of intensive treatment on microvascular disease, hypoglycemia, cognition, quality of life, and cost-effectiveness.

The ACCORD study was designed to have 89% power to detect a 15% treatment effect of intensive glycemic control compared with standard glycemic control for the primary end point.

ACCORD RESULTS

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545-2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545-2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545–2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

Rates of death from any cause (per 1,000 patient-years):

• 14 (intensive-treatment group)
• 11 (standard-treatment group).

In the analyses available at the time that this study arm closed, the excess mortality was not attributable to any particular treatment regimen. In particular, rosiglitazone (Avandia) use did not contribute to the excess mortality. (Of note, 91.2% of the intensive-treatment group and 57.5% of the conventional-treatment group had been treated with rosiglitazone, with more than 19,000 patient-years of rosiglitazone exposure). The excess mortality was also not attributable to hypoglycemia immediately proximate to the death.

The ACCORD trial’s data safety and monitoring board recommended that this arm of the study be discontinued for safety reasons, and this recommendation was accepted by the NHLBI project office. All participants were notified by letter before the trial results were announced publicly, and all intensive-therapy group participants are now being treated by the protocol used in the standard-therapy group.¹

FEWER DEATHS IN ACCORD THAN IN OTHER STUDIES IN DIABETES

The mortality rates in both arms of ACCORD were much lower than in other observational studies and clinical trials in type 2 diabetes.

The National Health and Nutrition Education Survey (NHANES),³⁰ conducted from 1971 to 1975, included 14,374 people with diabetes between the ages of 25 and 74. Many of them were younger than the ACCORD patients, but two NHANES age-groups overlapped the ACCORD cohort. Rates of death from any cause at 22 years (per 1,000 patient-years):

• 39.7 (ages 45–64)
• 89.7 (ages 65–74).

The NHANES cohort would not have been treated as vigorously for coronary risk and other common causes of death.

UGDP, UKPDS. Death rates in the glucose-lowering trials of type 2 diabetes mellitus cited above were typically in the range of 20 deaths per 1,000 patient-years but were as high as 30 deaths per 1,000 patient-years in the UGDP tolbutamide group¹⁶ and the UK-PDS sulfonylurea-plus-metformin group.^20,22,26

Steno-2.³¹ Half of 160 patients with type 2 diabetes were randomized to intensive strategies for controlling glucose, lipids, and blood pressure and for taking aspirin and angiotensin-converting enzyme inhibitors and following a healthy lifestyle. The other half received conventional therapy. Even in the intensive-treatment group, the mortality rate at 13 years was higher than in ACCORD. Rates of death from any cause (per 1,000 patient years):

• 22.5 (intensive-treatment group)
• 37.6 (conventional-treatment group).

After the ACCORD results were presented, two other trials addressing the question of whether lower hemoglobin A_1c would reduce cardiovascular risk in type 2 diabetes have reported their outcomes:

The ADVANCE trial (Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation),^32,33 with 11,140 patients, had a target hemoglobin A_1c of 6.5% in an intensive-treatment group and 7.3% in a usual-treatment group. The intensive-treatment group showed no difference in the rates of major macrovascular events (HR 0.94, 95% CI 0.84–1.06, P = .32) or all-cause mortality (HR 0.93, 95% CI 0.83–1.06, P = .32). The overall death rate in ADVANCE (about 18 deaths per 1,000 patient-years) was higher than in ACCORD.

The Veterans Administration Diabetes Trial included 1,791 patients.³⁴ Like the ADVANCE trial, it also found no difference in major cardiovascular outcomes (HR 0.868, P = .11) or cardiovascular mortality rates (HR 1.258, P = .36) with intensive therapy vs conventional therapy, ie, achieved hemoglobin A_1c levels of 6.9% vs 8.4% (presented at the American Diabetes Association 2008 Scientific Sessions). Hypoglycemia was associated with an increased risk of death in the standard-treatment group.

An analysis suggested that patients with a shorter duration of diabetes may have had cardiovascular benefit from intensive glucose-lowering, while those who had had it longer may have had increased risk associated with the more intensive therapy. The rate of death from all causes appears to have been higher than in ACCORD, but this could not be determined accurately from the presentations.

Comment. Thus, the ACCORD cohort as a whole has had strikingly lower death rates than in these other studies. The fact that all participants had lower glucose levels on therapy than at baseline may possibly contribute to these lower death rates. In addition, all ACCORD participants in the lipid arm received a statin; all participants in the blood pressure arm had their blood pressure lowered to levels below those commonly seen in clinical practice; participants were encouraged to exercise regularly; most participants were given diet instruction; and other healthy behaviors such as aspirin use, regular follow-up with primary care physicians, and recommendations about smoking were encouraged throughout the study. These comprehensive strategies may represent better care and thus result in lower death rates than in other studies.

POSSIBLE EXPLANATIONS FOR THE ACCORD OUTCOMES

The ACCORD trial has already stimulated fierce debate about the reasons for the higher mortality rate in the intensive-treatment group. With longer follow-up, some new risk factors for death may be identified that are not evident in the analyses of the current 460 deaths. What follows are some of my thoughts, with the caveat that they are not confirmed (supported statistically) by any currently available analyses from ACCORD.

It seems unlikely that lower glucose values as reflected by lower hemoglobin A_1c values in the intensive-treatment group are an a priori explanation for the observed differences in mortality rates—especially since the mortality rates were lower than in the NHANES and clinical trial data sets cited above. If we assume that a type 1 statistical error (finding a difference where no difference actually exists) does not explain the findings, then at least four reasonable postulates exist:

Hypoglycemia may have some adverse effect, either acutely or from recurrent events that trigger a catecholamine response with associated risk for arrhythmia or increased coronary heart disease risk. However, the investigators analyzed each death to determine whether hypoglycemia was a contributing cause, and they found no statistically significant relationship between hypoglycemia and death in the intensive-treatment group.

Weight gain is common with intensive therapy. Obesity may be associated with greater cytokine production, higher concentrations of clotting factors, higher levels of free fatty acids, and other potential contributors to the risk of coronary heart disease and death. Currently, the ACCORD analyses do not suggest that weight gain explains the higher death rate.

Medications such as rosiglitazone, sulfonylureas, and the combination of a sulfonylurea plus metformin have been previously associated with increased death rates in some observational and intervention trials. These studies had some serious methodologic limitations (eg, absence of risk adjustment, events not adjudicated, small study cohorts, wide variation in study cohort characteristics) and small numbers of events.^{11–13,16,26,35} ACCORD analyses have not shown that any single glucose-lowering agent—including rosiglitazone—or combination of agents explains the death rates.

The stress of maintaining glycemic control has been speculated to have in some way contributed to an increased risk. To achieve intensive control, patients had to have frequent contact with their health care providers, they were often told that their hemoglobin A_1c values were “too high” even when they were well below those in the American Diabetes Association guidelines, and they had to follow complex glucose-lowering regimens.

Semiquantitative measures of overall attitudes about health exist (eg, the “Feeling Thermometer” scale), but stress was not measured quantitatively in the ACCORD trial.

IMPLICATIONS OF ACCORD

In practice, most clinicians believe that the target glucose level in patients with type 2 diabetes should be as low as safely possible. This approach does not need to be modified on the basis of current information from ACCORD.

To be safe, regimens should be associated with a low risk of hypoglycemia and a low risk of weight gain. Use of combinations of medications that work by different mechanisms is still prudent. Agents should be used that may have favorable effects on other cardiovascular risk factors (eg, lipids, blood pressure, visceral fat).

Hemoglobin A_1c targets below 7% are not precluded in all patients on the basis of the ACCORD results, though values lower than 6% may not have much added benefit for cardiovascular risk reduction. We should note that hemoglobin A_1c was reduced in all ACCORD participants and that death rates were lower than in many other type 2 diabetic cohorts. Pending data on other outcomes in ACCORD (nephropathy, retinopathy, dementia, fracture risk), I believe it is premature for organizations to change their proposed hemoglobin A_1c targets,^36,37 as none have proposed values as low as the target in the ACCORD intensive-treatment group. At present, no class of glucose-lowering agents needs to be excluded from consideration on the basis of the ACCORD data.

The overall low rates of death in this population at high risk of coronary heart disease deserve comment. Not only are they lower than in other glucose-lowering trials, but they are also lower than in a number of studies of mortality in diabetes cohorts. As noted above, multiple risk factors for coronary heart disease and death were (and are) addressed in the ACCORD study participants, including repeated recommendation for lifestyle modification, intervention arms with lipid and blood pressure therapy, encouragement of aspirin use, and regular follow-up with health care providers for risk factors not managed by the ACCORD trial protocol. It is likely that multiple approaches to reducing the risk of cardiovascular disease contributed to this low mortality rate and that similar approaches will reduce the risk of coronary disease and death in regular clinical practice.

The ACCORD lipid and blood pressure arms are continuing, with results expected in 2010. The future results from ACCORD as well as from several glucose-lowering trials currently in progress (ADVANCE,^32,33 Veteran’s Administration,³⁴ Bypass Angioplasty Revascularization Investigation 2 Diabetes [BARI-2D]³⁸) will likely help refine our understanding of the effects of glucose-lowering, glucose-lowering strategies and targets, and multiple interventions on coronary events and all-cause mortality.

For now, any strategy that lowers glucose and is associated with a low risk of hypoglycemia and does not cause excessive weight gain should be considered appropriate in patients with type 2 diabetes.

The Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial^1–5 was designed primarily to address, in patients with type 2 diabetes at high risk of cardiovascular events, whether intensive glucose control would result in a lower risk of atherosclerotic disease events or death than would standard treatment.

It was widely expected that intensive treatment would confer either modest benefit or, at worst, no benefit. However, the glucose-lowering arm of the trial was terminated early because of a higher mortality rate in the intensively treated group. (The ACCORD trial has two other arms, which concern blood pressure and lipid-lowering, and these are continuing.)

In earlier trials in type 2 diabetes, concerns had been raised about an increased risk of cardiovascular events and possibly death associated with glucose-lowering drugs, hypoglycemia itself, or both, and these were well known when ACCORD was convened. ACCORD was very carefully designed and included careful adjudication of each cardiovascular event and death, including whether hypoglycemia might have been a proximate cause of some sudden deaths.⁵

Therefore, the surprising result of the higher mortality rate with intensive glycemic control in ACCORD will be fodder for discussion in many arenas over the next several years, and it poses some challenges for physicians and patients in determining treatment goals, as well as for organizations that write clinical practice guidelines (and perhaps organizations involved in pay-for-performance based on these guidelines).

Still, I believe that the ACCORD results should not substantially change our approach to treatment goals in type 2 diabetes, although hemoglobin A_1c targets below 6% may not have much added value for cardiovascular risk reduction. The low overall mortality rate in all the arms of the ACCORD trial emphasizes the importance of lifestyle modification, lipid and blood pressure therapy, and encouragement of aspirin use in all patients with type 2 diabetes.

This article reflects my views as a practicing diabetologist and clinical trialist (I was an investigator in the ACCORD trial) with a long-standing interest in clinical trials and in how the results influence clinical practice. The views I express herein may not reflect the views of other ACCORD investigators, the National Heart, Lung, and Blood Institute (NHLBI), the ACCORD trial coordinating center at Wake Forest University, or its data safety and monitoring board.

RISK OF CORONARY DISEASE INCREASES WITH GLUCOSE

Many observational studies^6–10 have shown that the risk of cardiovascular disease, especially coronary heart disease, is two to five times higher in people with diabetes mellitus than in people without diabetes. The risk appears to be continuous, so the higher one’s glucose or hemoglobin A_1c, the higher the risk.⁶ This risk even extends to glucose values well below the threshold values currently used to diagnose diabetes mellitus.⁶ Since there is no glucose threshold for coronary heart disease, the term dysglycemia (rather than hyperglycemia) has been proposed to note the relationship between glucose and coronary heart disease. (The glucose threshold for microvascular complications of diabetes, such as retinopathy and nephropathy, appears to be between 110 and 126 mg/dL).

The clustering of multiple coronary risk factors such as obesity, dyslipidemia, and hypertension has always raised the question of whether glucose is a culprit in coronary risk or whether it simply “runs in bad company.”

EARLIER CLINICAL TRIALS SUGGEST INTENSIVE TREATMENT RAISES RISK

Even though it has been widely believed that intensive glucose-lowering would reduce cardiovascular risk in type 2 diabetes, there have been hints in previous studies that some intensive-treatment regimens might increase risk.

Two large randomized clinical trials and one small one (discussed below) addressed whether glucose control would reduce the risk of atherosclerotic vascular disease events. In each of them, an increased risk of cardiovascular events and possibly of death was seen in at least one intensively treated group.

In the following discussion, I have calculated all of the death rates as the number of deaths per 1,000 patients per year, based on published study results. In this way, we can compare the rates in the various studies (including ACCORD), regardless of the trial duration.

The university group Diabetes Program: Controvery about tolbutamide therapy

The University Group Diabetes Program (UGDP)^11–16 included about 1,000 participants randomized to five treatments: tolbutamide (Orinase, a sulfonylurea), insulin in a fixed dose based on body weight, insulin in adjusted doses based on fasting glucose levels, placebo, and (later) phenformin.

In the 1970s, when the UGDP was carried out, randomized clinical trials were uncommon. Like other trials from that era, the UGDP was underpowered by today’s standards and did not have a data safety and monitoring board.

Rates of cardiovascular events and deaths (per 1,000 patient-years):

• 25 (tolbutamide group)
• 12 (placebo group).

The two insulin groups did not differ from the placebo group in their rates of cardiovascular events or death.¹⁵ The tolbutamide arm was stopped, and the ensuing controversy about how to interpret the trial results lasted for more than a decade. It also resulted in a black-box warning for tolbutamide and all subsequent sulfonylureas.

The United Kingdom Prospective Diabetes Study (UKPDS)^17–27 was launched in 1977. A cohort of 5,102 patients (mean age 54 years) with newly diagnosed type 2 diabetes mellitus followed a “prudent diet” for the first 3 to 4 months. Then, if their fasting glucose levels were in the range of 6.1 to 15 mmol/L (110–270 mg/dL), they were randomized to receive various treatments.

Patients who were not obese were randomized to receive either intensive treatment or conventional treatment. The intensive-treatment group received either insulin or a sulfonylurea (chlorpropamide [Diabinese], glibenclamide, or glipizide [Glucotrol]); the conventional-treatment group received diet therapy. The sulfonylurea arm was included partly to address the UGDP results.

Patients who were obese were randomized to receive one of three treatments: intensive treatment (with the agents listed above), conventional treatment, or metformin (Fortamet, Glucophage).

The mean in-trial hemoglobin A_1c level in the intensive-treatment group was 7.0%, compared with 7.9% in the conventional-treatment group.

After a mean follow-up of more than 10 years, the incidence of myocardial infarction was 16% lower in the intensive-treatment group, but the difference was not statistically significant (P = .052).

Rates of death from all causes among nonobese subjects (per 1,000 patient-years):

• 18.2–20.5 (intensive-treatment group)
• 19.9 (conventional-treatment group).

In the obese patients who received metformin, the incidence of myocardial infarction was lower than in the conventional-treatment group but not the intensive-treatment group.

Rates of death among obese patients (per 1,000 patient-years):

• 13.5 (metformin group)
• 18.9 (intensive-treatment group)
• 20.6 (conventional-treatment group).

However, a small subset (n = 587) of the original group assigned to sulfonylurea therapy whose glycemic control deteriorated during the trial were rerandomized to continue to receive a sulfonylurea alone or to have metformin added. There was a statistically significantly higher rate of cardiovascular events and a nonsignificantly higher rate of total mortality in the metformin-plus-sulfonylurea group (30.3 per 1,000 patient-years) than in the sulfonylurea-only group (19.1 per 1,000 patient-years).

These data suggested that the way glucose-lowering was achieved might be as important as the glucose levels actually achieved. However, no definite conclusions could be drawn.

In an editorial on the UKPDS, Nathan²⁶ made a comment that may have been prescient in terms of the ACCORD trial: “Professional organizations will now scramble to decide how to translate the UKPDS results … Whether the UKPDS firmly establishes the choice of any one therapy…or any combination of therapies for the long-term treatment of type 2 diabetes is more questionable.”²⁶

Veterans Administration feasibility study

A Veterans Administration feasibility study^28,29 included 153 men (mean age 60) with type 2 diabetes (mean duration 7.8 years) who received either conventional therapy (a single daily dose of insulin) or intensive therapy (multiple doses of insulin plus a sulfonylurea). Over a mean of 27 months, the intensive-therapy group achieved a hemoglobin A_1c level that was 2 percentage points lower than in the conventional-therapy group.

At 2.25 years of follow-up, cardiovascular events had occurred in 24 (24%) of the intensive-therapy group and in 16 (20%) of the standard-therapy group (P = .10).

Rates of death from all causes (per 1,000 patient-years):

• 28.9 (intensive-treatment group)
• 17.5 (conventional-treatment group).

ACCORD TRIAL DESIGN

The primary outcome measured was the combined incidence of nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes. Secondary outcomes included death from any cause. The study is also evaluating the effect of intensive treatment on microvascular disease, hypoglycemia, cognition, quality of life, and cost-effectiveness.

The ACCORD study was designed to have 89% power to detect a 15% treatment effect of intensive glycemic control compared with standard glycemic control for the primary end point.

ACCORD RESULTS

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545-2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545-2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

From Gerstein HC, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545–2559. Copyright 2008, Massachusetts Medical Society. All rights reserved.

Rates of death from any cause (per 1,000 patient-years):

• 14 (intensive-treatment group)
• 11 (standard-treatment group).

In the analyses available at the time that this study arm closed, the excess mortality was not attributable to any particular treatment regimen. In particular, rosiglitazone (Avandia) use did not contribute to the excess mortality. (Of note, 91.2% of the intensive-treatment group and 57.5% of the conventional-treatment group had been treated with rosiglitazone, with more than 19,000 patient-years of rosiglitazone exposure). The excess mortality was also not attributable to hypoglycemia immediately proximate to the death.

The ACCORD trial’s data safety and monitoring board recommended that this arm of the study be discontinued for safety reasons, and this recommendation was accepted by the NHLBI project office. All participants were notified by letter before the trial results were announced publicly, and all intensive-therapy group participants are now being treated by the protocol used in the standard-therapy group.¹

FEWER DEATHS IN ACCORD THAN IN OTHER STUDIES IN DIABETES

The mortality rates in both arms of ACCORD were much lower than in other observational studies and clinical trials in type 2 diabetes.

The National Health and Nutrition Education Survey (NHANES),³⁰ conducted from 1971 to 1975, included 14,374 people with diabetes between the ages of 25 and 74. Many of them were younger than the ACCORD patients, but two NHANES age-groups overlapped the ACCORD cohort. Rates of death from any cause at 22 years (per 1,000 patient-years):

• 39.7 (ages 45–64)
• 89.7 (ages 65–74).

The NHANES cohort would not have been treated as vigorously for coronary risk and other common causes of death.

UGDP, UKPDS. Death rates in the glucose-lowering trials of type 2 diabetes mellitus cited above were typically in the range of 20 deaths per 1,000 patient-years but were as high as 30 deaths per 1,000 patient-years in the UGDP tolbutamide group¹⁶ and the UK-PDS sulfonylurea-plus-metformin group.^20,22,26

Steno-2.³¹ Half of 160 patients with type 2 diabetes were randomized to intensive strategies for controlling glucose, lipids, and blood pressure and for taking aspirin and angiotensin-converting enzyme inhibitors and following a healthy lifestyle. The other half received conventional therapy. Even in the intensive-treatment group, the mortality rate at 13 years was higher than in ACCORD. Rates of death from any cause (per 1,000 patient years):

• 22.5 (intensive-treatment group)
• 37.6 (conventional-treatment group).

After the ACCORD results were presented, two other trials addressing the question of whether lower hemoglobin A_1c would reduce cardiovascular risk in type 2 diabetes have reported their outcomes:

The ADVANCE trial (Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation),^32,33 with 11,140 patients, had a target hemoglobin A_1c of 6.5% in an intensive-treatment group and 7.3% in a usual-treatment group. The intensive-treatment group showed no difference in the rates of major macrovascular events (HR 0.94, 95% CI 0.84–1.06, P = .32) or all-cause mortality (HR 0.93, 95% CI 0.83–1.06, P = .32). The overall death rate in ADVANCE (about 18 deaths per 1,000 patient-years) was higher than in ACCORD.

The Veterans Administration Diabetes Trial included 1,791 patients.³⁴ Like the ADVANCE trial, it also found no difference in major cardiovascular outcomes (HR 0.868, P = .11) or cardiovascular mortality rates (HR 1.258, P = .36) with intensive therapy vs conventional therapy, ie, achieved hemoglobin A_1c levels of 6.9% vs 8.4% (presented at the American Diabetes Association 2008 Scientific Sessions). Hypoglycemia was associated with an increased risk of death in the standard-treatment group.

An analysis suggested that patients with a shorter duration of diabetes may have had cardiovascular benefit from intensive glucose-lowering, while those who had had it longer may have had increased risk associated with the more intensive therapy. The rate of death from all causes appears to have been higher than in ACCORD, but this could not be determined accurately from the presentations.

Comment. Thus, the ACCORD cohort as a whole has had strikingly lower death rates than in these other studies. The fact that all participants had lower glucose levels on therapy than at baseline may possibly contribute to these lower death rates. In addition, all ACCORD participants in the lipid arm received a statin; all participants in the blood pressure arm had their blood pressure lowered to levels below those commonly seen in clinical practice; participants were encouraged to exercise regularly; most participants were given diet instruction; and other healthy behaviors such as aspirin use, regular follow-up with primary care physicians, and recommendations about smoking were encouraged throughout the study. These comprehensive strategies may represent better care and thus result in lower death rates than in other studies.

POSSIBLE EXPLANATIONS FOR THE ACCORD OUTCOMES

The ACCORD trial has already stimulated fierce debate about the reasons for the higher mortality rate in the intensive-treatment group. With longer follow-up, some new risk factors for death may be identified that are not evident in the analyses of the current 460 deaths. What follows are some of my thoughts, with the caveat that they are not confirmed (supported statistically) by any currently available analyses from ACCORD.

It seems unlikely that lower glucose values as reflected by lower hemoglobin A_1c values in the intensive-treatment group are an a priori explanation for the observed differences in mortality rates—especially since the mortality rates were lower than in the NHANES and clinical trial data sets cited above. If we assume that a type 1 statistical error (finding a difference where no difference actually exists) does not explain the findings, then at least four reasonable postulates exist:

Hypoglycemia may have some adverse effect, either acutely or from recurrent events that trigger a catecholamine response with associated risk for arrhythmia or increased coronary heart disease risk. However, the investigators analyzed each death to determine whether hypoglycemia was a contributing cause, and they found no statistically significant relationship between hypoglycemia and death in the intensive-treatment group.

Weight gain is common with intensive therapy. Obesity may be associated with greater cytokine production, higher concentrations of clotting factors, higher levels of free fatty acids, and other potential contributors to the risk of coronary heart disease and death. Currently, the ACCORD analyses do not suggest that weight gain explains the higher death rate.

Medications such as rosiglitazone, sulfonylureas, and the combination of a sulfonylurea plus metformin have been previously associated with increased death rates in some observational and intervention trials. These studies had some serious methodologic limitations (eg, absence of risk adjustment, events not adjudicated, small study cohorts, wide variation in study cohort characteristics) and small numbers of events.^{11–13,16,26,35} ACCORD analyses have not shown that any single glucose-lowering agent—including rosiglitazone—or combination of agents explains the death rates.

The stress of maintaining glycemic control has been speculated to have in some way contributed to an increased risk. To achieve intensive control, patients had to have frequent contact with their health care providers, they were often told that their hemoglobin A_1c values were “too high” even when they were well below those in the American Diabetes Association guidelines, and they had to follow complex glucose-lowering regimens.

Semiquantitative measures of overall attitudes about health exist (eg, the “Feeling Thermometer” scale), but stress was not measured quantitatively in the ACCORD trial.

IMPLICATIONS OF ACCORD

In practice, most clinicians believe that the target glucose level in patients with type 2 diabetes should be as low as safely possible. This approach does not need to be modified on the basis of current information from ACCORD.

To be safe, regimens should be associated with a low risk of hypoglycemia and a low risk of weight gain. Use of combinations of medications that work by different mechanisms is still prudent. Agents should be used that may have favorable effects on other cardiovascular risk factors (eg, lipids, blood pressure, visceral fat).

Hemoglobin A_1c targets below 7% are not precluded in all patients on the basis of the ACCORD results, though values lower than 6% may not have much added benefit for cardiovascular risk reduction. We should note that hemoglobin A_1c was reduced in all ACCORD participants and that death rates were lower than in many other type 2 diabetic cohorts. Pending data on other outcomes in ACCORD (nephropathy, retinopathy, dementia, fracture risk), I believe it is premature for organizations to change their proposed hemoglobin A_1c targets,^36,37 as none have proposed values as low as the target in the ACCORD intensive-treatment group. At present, no class of glucose-lowering agents needs to be excluded from consideration on the basis of the ACCORD data.

The overall low rates of death in this population at high risk of coronary heart disease deserve comment. Not only are they lower than in other glucose-lowering trials, but they are also lower than in a number of studies of mortality in diabetes cohorts. As noted above, multiple risk factors for coronary heart disease and death were (and are) addressed in the ACCORD study participants, including repeated recommendation for lifestyle modification, intervention arms with lipid and blood pressure therapy, encouragement of aspirin use, and regular follow-up with health care providers for risk factors not managed by the ACCORD trial protocol. It is likely that multiple approaches to reducing the risk of cardiovascular disease contributed to this low mortality rate and that similar approaches will reduce the risk of coronary disease and death in regular clinical practice.

The ACCORD lipid and blood pressure arms are continuing, with results expected in 2010. The future results from ACCORD as well as from several glucose-lowering trials currently in progress (ADVANCE,^32,33 Veteran’s Administration,³⁴ Bypass Angioplasty Revascularization Investigation 2 Diabetes [BARI-2D]³⁸) will likely help refine our understanding of the effects of glucose-lowering, glucose-lowering strategies and targets, and multiple interventions on coronary events and all-cause mortality.

For now, any strategy that lowers glucose and is associated with a low risk of hypoglycemia and does not cause excessive weight gain should be considered appropriate in patients with type 2 diabetes.

More than 2 years have passed since we published the results of the Women’s Health Initiative (WHI), which caused a storm of information—and misinformation—about the effect of long-term dietary intervention on disease outcomes in postmenopausal women. Now that the dust has long settled, what have we learned from this landmark study?

The WHI results led to numerous additional analyses of all aspects of the study.^1–7 What are the implications of all the analyses to clinical practice?

In this article, we summarize key aspects of the clinical trial, including study design, interventions, main results, and future plans. We also discuss potential clinical applications and practical considerations for public health efforts.

WHO WAS ELIGIBLE, WHO WAS NOT

A total of 48,835 postmenopausal women were randomly assigned to either no dietary intervention (n = 29,294) or a dietary intervention (n = 19,541) (see below).⁷ Participants were followed at 40 clinical centers between 1993 and 2005.⁴ Their mean age was 62.3 years; 18.6% were members of minorities.

Women were eligible if they were post-menopausal and had a daily dietary fat intake of at least 32% of total calories, based on assessment via a food-frequency questionnaire. They were excluded from the study if they had any of the following: a history of breast cancer, colorectal cancer, or other cancer except skin cancer during the past 10 years; type 1 diabetes; a medical condition in which the predicted survival was less than 3 years; and a potential barrier to adherence to the study regimen, including alcoholism or a lifestyle that involved often eating meals away from home.

THE WHI DIET: LESS FAT, BUT MORE FRUITS, VEGETABLES, GRAINS

The WHI dietary intervention was designed to prevent breast cancer, based on the evidence available when the study was planned. The targets included a total fat intake of less than 20% of energy (in kilocalories), increasing the intake of fruits and vegetables to at least five servings per day, and increasing the intake of grains to at least six servings per day.

Although reduction in saturated fat intake per se was not part of the WHI protocol, we assumed from previous pilot studies⁸ that the reduction of total fat intake would simultaneously produce a reduction in saturated fat intake to 7% of total calories.

A simpler dietary intervention

Unlike the 2006 American Heart Association guidelines and the US Department of Agriculture’s Dietary Guidelines for Americans 2005, the WHI dietary intervention had no specifications for dietary fiber, specific fatty acids (trans-fatty acids, omega-3 fatty acids, conjugated linoleic acid), complex carbohydrates, whole grains, vegetable protein, or other factors that have emerged as potential risk factors for cardiovascular and other chronic diseases since the study began. The WHI intervention also included no specific recommendation for total calorie intake, nor were patients in the intervention group encouraged to lose weight, as this could have confounded the results of the dietary intervention.

Education and encouragement

Those in the intervention group were each assigned a fat-gram goal, calculated on the basis of height. They were taught how to monitor their intake of total fat, fruits, vegetables, and grains. They attended intensive behavioral modification sessions to encourage them to keep to the dietary program: 18 group sessions in the first year and quarterly maintenance sessions thereafter, touching on a wide variety of nutrition- and behavior-related topics.^7,9 Specially trained and certified nutritionists supervised the dietary intervention and the behavioral modification sessions according to the WHI study protocol.

Control-group participants received a copy of the US Department of Agriculture’s Dietary Guidelines for Americans¹⁰ and other health-related materials. They had no contact with the study nutritionists.

Other arms of the study

The WHI trial design included several arms,^4,11–13 and many participants joined more than one arm: 20,592 postmenopausal women (42.2% of the total enrollment) chose dietary modification only, 8,050 (16.5%) chose diet plus hormone replacement therapy, 25,210 (51.6%) chose diet plus calcium and vitamin D supplementation, and 5,017 (10.3%) enrolled in all three.

Length of follow-up

Participants were followed from enrollment until they died, were lost to follow-up, or requested no further contact, or until the trial’s planned completion date, regardless of adherence to the dietary intervention, according to intention-to-treat analysis. All participants were contacted by clinic staff at 6-month intervals to provide updates on their health outcomes.

Factors assessed

Height, weight, waist circumference, and blood pressure were measured at annual visits using standardized procedures. Fasting blood samples were collected at baseline and at year 1 from all participants and from a subsample of 2,816 women (5.8% of the study population) at years 3 and 6. This subsample was randomly chosen with oversampling of minority women, for whom the odds for selection were six times higher than for white women.

Physical activity was assessed at baseline and at years 1, 3, 6, and 9. Walking and participation in sports and hours of activity per week were calculated for each participant. Physical activity was expressed as metabolic equivalent tasks per week for the analyses.

A food-frequency questionnaire⁶ to assess average dietary intake in the past 3 months was given at baseline and at year 1 for all participants. A third of all participants completed the questionnaire each year in a rotating sample. Completion rates were 100% at baseline and 81% thereafter. Follow-up data were collected from years 5 through 7. Also, 4-day food records were provided by all women before randomization.

HOW OUTCOMES WERE ASSESSED

The primary assessments of clinical outcome^1–3 were mammographic screening, a self-reported medical history documented by a review of medical records, and electrocardiograms digitally obtained every 3 years. Mammograms and electrocardiograms were centrally adjudicated. The diagnosis of acute myocardial infarction was based on an algorithm that included cardiac pain, enzyme levels, and electrocardiographic readings.

OVERALL RESULTS

At 8.1 years, the incidence of breast cancer was 9% lower in the intervention group than in the comparison group (95% confidence interval [CI] = 0.83–1.01; P = .07, P = .09 weighted for length of follow-up).³ Subgroup analysis further showed that women who reported higher intakes of total dietary fat at baseline reduced their risk of breast cancer by 22% (95% CI = 0.64–0.96). Whether extended follow-up will show a significant association has yet to be determined.

Colon cancer rates did not differ between groups, but the number of polyps and adenomas reported was significantly lower in the dietary intervention group.¹ The rate of colon cancer will also be included in the extended follow-up study of the WHI.

Risk factors for coronary heart disease in both groups—including levels of serum total cholesterol and serum low-density lipoprotein cholesterol, body weight, body mass index, diastolic blood pressure, and factor VIIc—improved slightly, but at year 3 of the trial, differences in overall rates of coronary heart disease and stroke in the two groups were not statistically significant.² In addition, the low-fat diet intervention was associated with a reduction in blood estradiol concentrations between baseline and year 1.³ At the end of the study, however, differences in rates of breast cancer, colorectal cancer, and heart disease between the two groups were not statistically significant.

RESULTS OF DIETARY MODIFICATIONS

Fat as a percentage of total calories

At the beginning of the WHI, all participants reported consuming an average of 35% of their caloric intake from fat (Table 1). At 1 year from baseline, the fat intake decreased to 24.3% in the intervention group (short of the study goal of 20%); this level had risen again to 26.7% by year 3 and to 28.8% at the end of the study. Stratified by quartile, women who achieved the greatest reductions in saturated and trans-fatty acids or the largest increases in their intake of fruits and vegetables appeared to have a moderate reduction in the risk of coronary heart disease.² Women in the comparison group also decreased their fat intake initially, but to a lesser degree, and gradually increased it again thereafter. The mean net difference in self-reported total fat intake between the intervention group and the comparison group at 6 years was 8.2% (P < .001) (study goal, 13%).^1–3

Intake of fruits, vegetables, and grains

At baseline, fruit and vegetable intake averaged 3.6 servings per day (Table 1). In the intervention group, this increased to 5.1 servings per day at year 1, and to 5.2 servings at year 3, but at the end of the study it had decreased to 4.9 servings.

Women in the intervention group were eating 4.7 servings of grains per day at baseline. This increased to 5.1 servings at year 1 and then decreased to 4.6 servings at year 3 and to 4.3 servings at the end of the study. It seems that as the women grew older their determination to increase servings of these foods diminished.

Proponents of some currently popular diets blame weight gain on a higher intake of carbohydrates, but the women following the WHI low-fat diet did not gain weight.²

Total fat vs saturated fat

Intake of total fat and saturated fat decreased in the intervention group during the study, but the difference between fat intake in the intervention group and that in the comparison group did not reach the degree expected.

At year 1, total fat as a percentage of total caloric intake was 10.8 percentage points below that of the comparison group, whereas the study expected difference was 13.0. At the end of the trial, the difference was only 8.2 percentage points, whereas the expected difference was 11.0.

Intake of all fatty acids (saturated and unsaturated) decreased at year 1, but then went back up slightly by the end of the trial but did not exceed baseline levels, and saturated fatty acids remained well below baseline levels: 9.5% vs 12.5% of caloric intake at baseline.⁴

INTERPRETING THE RESULTS

It might be tempting to dismiss the results of the WHI dietary intervention trial as not significant and therefore not meaningful. This would be unfortunate. The trial had some remarkable accomplishments and offers important lessons for future investigations.

The initial reductions in total fat intake were impressive, and women who had the highest total fat intake at baseline achieved the greatest reduction of total fat (to less than 22% of total calories).³ Nonetheless, the dietary intervention goal of less than 20% of calories from fat was not achieved despite intensive dietary counseling and a highly motivated study population. Thus, this dietary fat target may not be reasonable in the general population.

Also, despite the absence of targeted intervention on specific fatty acids, the observed blood cholesterol levels were as expected based on the well-known formula of Mensink and Katan,¹⁴ which incorporates information on changes in saturated fat, polyunsaturated fat, and dietary cholesterol intake. The predicted reduction in low-density lipoprotein cholesterol was 2.7 mg/dL; the observed reduction was 2.3 mg/dL.² This illustrates that with greater modifications in specific known dietary risk factors for cardiovascular disease, such as saturated fatty acids, cholesterol, and unsaturated fatty acids, blood cholesterol levels respond in a predictable fashion. This was presumably not observed in WHI precisely because no goals and objectives were provided to participants for intake of saturated or polyunsaturated fatty acids.

Recent findings from the Optimal Macronutrient Intake Trial to Prevent Heart Disease (OmniHeart)¹⁵ further highlight differences in the total cholesterol response to diets of varying macronutrient (carbohydrate, protein, fat) content compared with the WHI dietary intervention.¹⁵ Participants in OmniHeart had reductions in levels of low-density lipoprotein cholesterol that were predictable from the changes reported in intake of saturated fatty acids. Presumably, the results of the WHI intervention would have been similar if the study had included this level of detail.

QUESTIONS REMAIN

Questions from the WHI that need consideration for future clinical applications include whether the study population may have already been “too old” to achieve a benefit from dietary modification, and whether the best timing for dietary intervention might be earlier adulthood with sustained changes in saturated fat, cholesterol, and unsaturated fat intake throughout life. Future subgroup analyses based on age at baseline will need to address these questions. Likewise, a longer follow-up period may be needed for a definitive evaluation of the impact of a regular low-fat diet on different health outcomes.

As reported by Patterson et al,¹⁶ the major contributors to total dietary fat intake at baseline were “added fats” such as sauces, gravies, butter, and margarines (25.1% of fat intake), followed by meats (20.9% of fat intake), and desserts (12.8% of fat intake). These findings highlight target areas for future interventions in women of this age group.

Another issue is how to standardize the dietary intervention from one clinical center to another—ie, to minimize differences in how each clinical center manages the study patients. Such differences were noted in WHI and other studies.¹⁷ Despite standardized training in delivering the dietary intervention, nutritionists encountered regional and cultural differences that required tailoring the dietary intervention to their patients’ needs. Staff turnover, an unavoidable phenomenon in long-term studies, has previously been reported to negatively influence dietary adherence.¹⁸

LIMITATIONS

A major limitation of diet modification research in general is the self-reporting of dietary intake, primarily by a food-frequency questionnaire. Although the use of a questionnaire is the most practical way to obtain dietary data for large studies, systematic biases may exist that obscure true nutrient-outcome relationships.¹⁹ Biomarker studies of energy balance suggest that people who are overweight or obese may under-report energy intake to a greater degree than people who are not overweight.²⁰ Also, we still do not know how to get people to follow a healthy diet, although theories and models abound, such as social learning and cognitive-behavioral theory, and a lack of data limits our understanding of factors related to dietary adherence.^21,22

FUTURE DIRECTIONS IN WHI

The WHI Extension Study is under way and has been funded through the year 2010. Outcomes ascertainment is the primary focus with no ongoing intervention, although the intervention group participants continue to receive a WHI newsletter that simply reiterates the importance of the study and encourages ongoing participation. As of 2006, an estimated 84% of the cohort, including both observational study and clinical trial participants, are involved. Efforts continue to recruit the remaining 16%, but many of these participants now consider themselves too old or too feeble to respond reliably.

In regard to breast cancer, the results published in 2006 are promising, albeit not statistically significant, and definitive statements cannot yet be made. However, postmenopausal women who are eating the diets highest in fat may have the greatest benefit from reductions in total fat.

Other considerations regarding the lack of statistically significant differences between groups may include the possibility that women in the intervention group may have been at lower risk for breast cancer at baseline. Likewise, although the results of the WHI dietary intervention do not include a statistically significant impact on colorectal cancer outcomes, the significant reduction in polyps and adenomas may later translate into a reduction in invasive cancer risk.

Finally, although no significant reduction was seen in the rate of death due to cardiovascular causes, greater reductions in saturated and trans-fatty acid intake were associated with greater reductions in blood cholesterol and cardiovascular risk.

Numerous subgroup analyses and ongoing assessments of the long-term impact of the diet modification are planned. Further associations are expected to emerge. The current and future results will continue to provide new insights that may lead to new clinical and public health recommendations in the future.

The WHI has raised additional issues that warrant further investigation:

• Will earlier dietary intervention, eg, during premenopausal years or even childhood, alter these results?
• Does the low-fat, high-carbohydrate diet used in WHI facilitate weight maintenance or even weight loss, as proposed by Howard et al²³?
• Do quantitative changes in physical activity and weight control attenuate morbidity and mortality rates beyond changes in diet alone?
• Do vitamin and mineral supplements or hormone therapy alter disease outcomes or quality of life?
• Which behavioral approaches are best suited to the recruitment of patients for dietary intervention trials?

More than 2 years have passed since we published the results of the Women’s Health Initiative (WHI), which caused a storm of information—and misinformation—about the effect of long-term dietary intervention on disease outcomes in postmenopausal women. Now that the dust has long settled, what have we learned from this landmark study?

The WHI results led to numerous additional analyses of all aspects of the study.^1–7 What are the implications of all the analyses to clinical practice?

In this article, we summarize key aspects of the clinical trial, including study design, interventions, main results, and future plans. We also discuss potential clinical applications and practical considerations for public health efforts.

WHO WAS ELIGIBLE, WHO WAS NOT

A total of 48,835 postmenopausal women were randomly assigned to either no dietary intervention (n = 29,294) or a dietary intervention (n = 19,541) (see below).⁷ Participants were followed at 40 clinical centers between 1993 and 2005.⁴ Their mean age was 62.3 years; 18.6% were members of minorities.

Women were eligible if they were post-menopausal and had a daily dietary fat intake of at least 32% of total calories, based on assessment via a food-frequency questionnaire. They were excluded from the study if they had any of the following: a history of breast cancer, colorectal cancer, or other cancer except skin cancer during the past 10 years; type 1 diabetes; a medical condition in which the predicted survival was less than 3 years; and a potential barrier to adherence to the study regimen, including alcoholism or a lifestyle that involved often eating meals away from home.

THE WHI DIET: LESS FAT, BUT MORE FRUITS, VEGETABLES, GRAINS

The WHI dietary intervention was designed to prevent breast cancer, based on the evidence available when the study was planned. The targets included a total fat intake of less than 20% of energy (in kilocalories), increasing the intake of fruits and vegetables to at least five servings per day, and increasing the intake of grains to at least six servings per day.

Although reduction in saturated fat intake per se was not part of the WHI protocol, we assumed from previous pilot studies⁸ that the reduction of total fat intake would simultaneously produce a reduction in saturated fat intake to 7% of total calories.

A simpler dietary intervention

Unlike the 2006 American Heart Association guidelines and the US Department of Agriculture’s Dietary Guidelines for Americans 2005, the WHI dietary intervention had no specifications for dietary fiber, specific fatty acids (trans-fatty acids, omega-3 fatty acids, conjugated linoleic acid), complex carbohydrates, whole grains, vegetable protein, or other factors that have emerged as potential risk factors for cardiovascular and other chronic diseases since the study began. The WHI intervention also included no specific recommendation for total calorie intake, nor were patients in the intervention group encouraged to lose weight, as this could have confounded the results of the dietary intervention.

Education and encouragement

Those in the intervention group were each assigned a fat-gram goal, calculated on the basis of height. They were taught how to monitor their intake of total fat, fruits, vegetables, and grains. They attended intensive behavioral modification sessions to encourage them to keep to the dietary program: 18 group sessions in the first year and quarterly maintenance sessions thereafter, touching on a wide variety of nutrition- and behavior-related topics.^7,9 Specially trained and certified nutritionists supervised the dietary intervention and the behavioral modification sessions according to the WHI study protocol.

Control-group participants received a copy of the US Department of Agriculture’s Dietary Guidelines for Americans¹⁰ and other health-related materials. They had no contact with the study nutritionists.

Other arms of the study

The WHI trial design included several arms,^4,11–13 and many participants joined more than one arm: 20,592 postmenopausal women (42.2% of the total enrollment) chose dietary modification only, 8,050 (16.5%) chose diet plus hormone replacement therapy, 25,210 (51.6%) chose diet plus calcium and vitamin D supplementation, and 5,017 (10.3%) enrolled in all three.

Length of follow-up

Participants were followed from enrollment until they died, were lost to follow-up, or requested no further contact, or until the trial’s planned completion date, regardless of adherence to the dietary intervention, according to intention-to-treat analysis. All participants were contacted by clinic staff at 6-month intervals to provide updates on their health outcomes.

Factors assessed

Height, weight, waist circumference, and blood pressure were measured at annual visits using standardized procedures. Fasting blood samples were collected at baseline and at year 1 from all participants and from a subsample of 2,816 women (5.8% of the study population) at years 3 and 6. This subsample was randomly chosen with oversampling of minority women, for whom the odds for selection were six times higher than for white women.

Physical activity was assessed at baseline and at years 1, 3, 6, and 9. Walking and participation in sports and hours of activity per week were calculated for each participant. Physical activity was expressed as metabolic equivalent tasks per week for the analyses.

A food-frequency questionnaire⁶ to assess average dietary intake in the past 3 months was given at baseline and at year 1 for all participants. A third of all participants completed the questionnaire each year in a rotating sample. Completion rates were 100% at baseline and 81% thereafter. Follow-up data were collected from years 5 through 7. Also, 4-day food records were provided by all women before randomization.

HOW OUTCOMES WERE ASSESSED

The primary assessments of clinical outcome^1–3 were mammographic screening, a self-reported medical history documented by a review of medical records, and electrocardiograms digitally obtained every 3 years. Mammograms and electrocardiograms were centrally adjudicated. The diagnosis of acute myocardial infarction was based on an algorithm that included cardiac pain, enzyme levels, and electrocardiographic readings.

OVERALL RESULTS

At 8.1 years, the incidence of breast cancer was 9% lower in the intervention group than in the comparison group (95% confidence interval [CI] = 0.83–1.01; P = .07, P = .09 weighted for length of follow-up).³ Subgroup analysis further showed that women who reported higher intakes of total dietary fat at baseline reduced their risk of breast cancer by 22% (95% CI = 0.64–0.96). Whether extended follow-up will show a significant association has yet to be determined.

Colon cancer rates did not differ between groups, but the number of polyps and adenomas reported was significantly lower in the dietary intervention group.¹ The rate of colon cancer will also be included in the extended follow-up study of the WHI.

Risk factors for coronary heart disease in both groups—including levels of serum total cholesterol and serum low-density lipoprotein cholesterol, body weight, body mass index, diastolic blood pressure, and factor VIIc—improved slightly, but at year 3 of the trial, differences in overall rates of coronary heart disease and stroke in the two groups were not statistically significant.² In addition, the low-fat diet intervention was associated with a reduction in blood estradiol concentrations between baseline and year 1.³ At the end of the study, however, differences in rates of breast cancer, colorectal cancer, and heart disease between the two groups were not statistically significant.

RESULTS OF DIETARY MODIFICATIONS

Fat as a percentage of total calories

At the beginning of the WHI, all participants reported consuming an average of 35% of their caloric intake from fat (Table 1). At 1 year from baseline, the fat intake decreased to 24.3% in the intervention group (short of the study goal of 20%); this level had risen again to 26.7% by year 3 and to 28.8% at the end of the study. Stratified by quartile, women who achieved the greatest reductions in saturated and trans-fatty acids or the largest increases in their intake of fruits and vegetables appeared to have a moderate reduction in the risk of coronary heart disease.² Women in the comparison group also decreased their fat intake initially, but to a lesser degree, and gradually increased it again thereafter. The mean net difference in self-reported total fat intake between the intervention group and the comparison group at 6 years was 8.2% (P < .001) (study goal, 13%).^1–3

Intake of fruits, vegetables, and grains

At baseline, fruit and vegetable intake averaged 3.6 servings per day (Table 1). In the intervention group, this increased to 5.1 servings per day at year 1, and to 5.2 servings at year 3, but at the end of the study it had decreased to 4.9 servings.

Women in the intervention group were eating 4.7 servings of grains per day at baseline. This increased to 5.1 servings at year 1 and then decreased to 4.6 servings at year 3 and to 4.3 servings at the end of the study. It seems that as the women grew older their determination to increase servings of these foods diminished.

Proponents of some currently popular diets blame weight gain on a higher intake of carbohydrates, but the women following the WHI low-fat diet did not gain weight.²

Total fat vs saturated fat

Intake of total fat and saturated fat decreased in the intervention group during the study, but the difference between fat intake in the intervention group and that in the comparison group did not reach the degree expected.

At year 1, total fat as a percentage of total caloric intake was 10.8 percentage points below that of the comparison group, whereas the study expected difference was 13.0. At the end of the trial, the difference was only 8.2 percentage points, whereas the expected difference was 11.0.

Intake of all fatty acids (saturated and unsaturated) decreased at year 1, but then went back up slightly by the end of the trial but did not exceed baseline levels, and saturated fatty acids remained well below baseline levels: 9.5% vs 12.5% of caloric intake at baseline.⁴

INTERPRETING THE RESULTS

It might be tempting to dismiss the results of the WHI dietary intervention trial as not significant and therefore not meaningful. This would be unfortunate. The trial had some remarkable accomplishments and offers important lessons for future investigations.

The initial reductions in total fat intake were impressive, and women who had the highest total fat intake at baseline achieved the greatest reduction of total fat (to less than 22% of total calories).³ Nonetheless, the dietary intervention goal of less than 20% of calories from fat was not achieved despite intensive dietary counseling and a highly motivated study population. Thus, this dietary fat target may not be reasonable in the general population.

Also, despite the absence of targeted intervention on specific fatty acids, the observed blood cholesterol levels were as expected based on the well-known formula of Mensink and Katan,¹⁴ which incorporates information on changes in saturated fat, polyunsaturated fat, and dietary cholesterol intake. The predicted reduction in low-density lipoprotein cholesterol was 2.7 mg/dL; the observed reduction was 2.3 mg/dL.² This illustrates that with greater modifications in specific known dietary risk factors for cardiovascular disease, such as saturated fatty acids, cholesterol, and unsaturated fatty acids, blood cholesterol levels respond in a predictable fashion. This was presumably not observed in WHI precisely because no goals and objectives were provided to participants for intake of saturated or polyunsaturated fatty acids.

Recent findings from the Optimal Macronutrient Intake Trial to Prevent Heart Disease (OmniHeart)¹⁵ further highlight differences in the total cholesterol response to diets of varying macronutrient (carbohydrate, protein, fat) content compared with the WHI dietary intervention.¹⁵ Participants in OmniHeart had reductions in levels of low-density lipoprotein cholesterol that were predictable from the changes reported in intake of saturated fatty acids. Presumably, the results of the WHI intervention would have been similar if the study had included this level of detail.

QUESTIONS REMAIN

Questions from the WHI that need consideration for future clinical applications include whether the study population may have already been “too old” to achieve a benefit from dietary modification, and whether the best timing for dietary intervention might be earlier adulthood with sustained changes in saturated fat, cholesterol, and unsaturated fat intake throughout life. Future subgroup analyses based on age at baseline will need to address these questions. Likewise, a longer follow-up period may be needed for a definitive evaluation of the impact of a regular low-fat diet on different health outcomes.

As reported by Patterson et al,¹⁶ the major contributors to total dietary fat intake at baseline were “added fats” such as sauces, gravies, butter, and margarines (25.1% of fat intake), followed by meats (20.9% of fat intake), and desserts (12.8% of fat intake). These findings highlight target areas for future interventions in women of this age group.

Another issue is how to standardize the dietary intervention from one clinical center to another—ie, to minimize differences in how each clinical center manages the study patients. Such differences were noted in WHI and other studies.¹⁷ Despite standardized training in delivering the dietary intervention, nutritionists encountered regional and cultural differences that required tailoring the dietary intervention to their patients’ needs. Staff turnover, an unavoidable phenomenon in long-term studies, has previously been reported to negatively influence dietary adherence.¹⁸

LIMITATIONS

A major limitation of diet modification research in general is the self-reporting of dietary intake, primarily by a food-frequency questionnaire. Although the use of a questionnaire is the most practical way to obtain dietary data for large studies, systematic biases may exist that obscure true nutrient-outcome relationships.¹⁹ Biomarker studies of energy balance suggest that people who are overweight or obese may under-report energy intake to a greater degree than people who are not overweight.²⁰ Also, we still do not know how to get people to follow a healthy diet, although theories and models abound, such as social learning and cognitive-behavioral theory, and a lack of data limits our understanding of factors related to dietary adherence.^21,22

FUTURE DIRECTIONS IN WHI

The WHI Extension Study is under way and has been funded through the year 2010. Outcomes ascertainment is the primary focus with no ongoing intervention, although the intervention group participants continue to receive a WHI newsletter that simply reiterates the importance of the study and encourages ongoing participation. As of 2006, an estimated 84% of the cohort, including both observational study and clinical trial participants, are involved. Efforts continue to recruit the remaining 16%, but many of these participants now consider themselves too old or too feeble to respond reliably.

In regard to breast cancer, the results published in 2006 are promising, albeit not statistically significant, and definitive statements cannot yet be made. However, postmenopausal women who are eating the diets highest in fat may have the greatest benefit from reductions in total fat.

Other considerations regarding the lack of statistically significant differences between groups may include the possibility that women in the intervention group may have been at lower risk for breast cancer at baseline. Likewise, although the results of the WHI dietary intervention do not include a statistically significant impact on colorectal cancer outcomes, the significant reduction in polyps and adenomas may later translate into a reduction in invasive cancer risk.

Finally, although no significant reduction was seen in the rate of death due to cardiovascular causes, greater reductions in saturated and trans-fatty acid intake were associated with greater reductions in blood cholesterol and cardiovascular risk.

Numerous subgroup analyses and ongoing assessments of the long-term impact of the diet modification are planned. Further associations are expected to emerge. The current and future results will continue to provide new insights that may lead to new clinical and public health recommendations in the future.

The WHI has raised additional issues that warrant further investigation:

• Will earlier dietary intervention, eg, during premenopausal years or even childhood, alter these results?
• Does the low-fat, high-carbohydrate diet used in WHI facilitate weight maintenance or even weight loss, as proposed by Howard et al²³?
• Do quantitative changes in physical activity and weight control attenuate morbidity and mortality rates beyond changes in diet alone?
• Do vitamin and mineral supplements or hormone therapy alter disease outcomes or quality of life?
• Which behavioral approaches are best suited to the recruitment of patients for dietary intervention trials?

More than 2 years have passed since we published the results of the Women’s Health Initiative (WHI), which caused a storm of information—and misinformation—about the effect of long-term dietary intervention on disease outcomes in postmenopausal women. Now that the dust has long settled, what have we learned from this landmark study?

The WHI results led to numerous additional analyses of all aspects of the study.^1–7 What are the implications of all the analyses to clinical practice?

In this article, we summarize key aspects of the clinical trial, including study design, interventions, main results, and future plans. We also discuss potential clinical applications and practical considerations for public health efforts.

WHO WAS ELIGIBLE, WHO WAS NOT

A total of 48,835 postmenopausal women were randomly assigned to either no dietary intervention (n = 29,294) or a dietary intervention (n = 19,541) (see below).⁷ Participants were followed at 40 clinical centers between 1993 and 2005.⁴ Their mean age was 62.3 years; 18.6% were members of minorities.

Women were eligible if they were post-menopausal and had a daily dietary fat intake of at least 32% of total calories, based on assessment via a food-frequency questionnaire. They were excluded from the study if they had any of the following: a history of breast cancer, colorectal cancer, or other cancer except skin cancer during the past 10 years; type 1 diabetes; a medical condition in which the predicted survival was less than 3 years; and a potential barrier to adherence to the study regimen, including alcoholism or a lifestyle that involved often eating meals away from home.

THE WHI DIET: LESS FAT, BUT MORE FRUITS, VEGETABLES, GRAINS

The WHI dietary intervention was designed to prevent breast cancer, based on the evidence available when the study was planned. The targets included a total fat intake of less than 20% of energy (in kilocalories), increasing the intake of fruits and vegetables to at least five servings per day, and increasing the intake of grains to at least six servings per day.

Although reduction in saturated fat intake per se was not part of the WHI protocol, we assumed from previous pilot studies⁸ that the reduction of total fat intake would simultaneously produce a reduction in saturated fat intake to 7% of total calories.

A simpler dietary intervention

Unlike the 2006 American Heart Association guidelines and the US Department of Agriculture’s Dietary Guidelines for Americans 2005, the WHI dietary intervention had no specifications for dietary fiber, specific fatty acids (trans-fatty acids, omega-3 fatty acids, conjugated linoleic acid), complex carbohydrates, whole grains, vegetable protein, or other factors that have emerged as potential risk factors for cardiovascular and other chronic diseases since the study began. The WHI intervention also included no specific recommendation for total calorie intake, nor were patients in the intervention group encouraged to lose weight, as this could have confounded the results of the dietary intervention.

Education and encouragement

Those in the intervention group were each assigned a fat-gram goal, calculated on the basis of height. They were taught how to monitor their intake of total fat, fruits, vegetables, and grains. They attended intensive behavioral modification sessions to encourage them to keep to the dietary program: 18 group sessions in the first year and quarterly maintenance sessions thereafter, touching on a wide variety of nutrition- and behavior-related topics.^7,9 Specially trained and certified nutritionists supervised the dietary intervention and the behavioral modification sessions according to the WHI study protocol.

Control-group participants received a copy of the US Department of Agriculture’s Dietary Guidelines for Americans¹⁰ and other health-related materials. They had no contact with the study nutritionists.

Other arms of the study

The WHI trial design included several arms,^4,11–13 and many participants joined more than one arm: 20,592 postmenopausal women (42.2% of the total enrollment) chose dietary modification only, 8,050 (16.5%) chose diet plus hormone replacement therapy, 25,210 (51.6%) chose diet plus calcium and vitamin D supplementation, and 5,017 (10.3%) enrolled in all three.

Length of follow-up

Participants were followed from enrollment until they died, were lost to follow-up, or requested no further contact, or until the trial’s planned completion date, regardless of adherence to the dietary intervention, according to intention-to-treat analysis. All participants were contacted by clinic staff at 6-month intervals to provide updates on their health outcomes.

Factors assessed

Height, weight, waist circumference, and blood pressure were measured at annual visits using standardized procedures. Fasting blood samples were collected at baseline and at year 1 from all participants and from a subsample of 2,816 women (5.8% of the study population) at years 3 and 6. This subsample was randomly chosen with oversampling of minority women, for whom the odds for selection were six times higher than for white women.

Physical activity was assessed at baseline and at years 1, 3, 6, and 9. Walking and participation in sports and hours of activity per week were calculated for each participant. Physical activity was expressed as metabolic equivalent tasks per week for the analyses.

A food-frequency questionnaire⁶ to assess average dietary intake in the past 3 months was given at baseline and at year 1 for all participants. A third of all participants completed the questionnaire each year in a rotating sample. Completion rates were 100% at baseline and 81% thereafter. Follow-up data were collected from years 5 through 7. Also, 4-day food records were provided by all women before randomization.

HOW OUTCOMES WERE ASSESSED

The primary assessments of clinical outcome^1–3 were mammographic screening, a self-reported medical history documented by a review of medical records, and electrocardiograms digitally obtained every 3 years. Mammograms and electrocardiograms were centrally adjudicated. The diagnosis of acute myocardial infarction was based on an algorithm that included cardiac pain, enzyme levels, and electrocardiographic readings.

OVERALL RESULTS

At 8.1 years, the incidence of breast cancer was 9% lower in the intervention group than in the comparison group (95% confidence interval [CI] = 0.83–1.01; P = .07, P = .09 weighted for length of follow-up).³ Subgroup analysis further showed that women who reported higher intakes of total dietary fat at baseline reduced their risk of breast cancer by 22% (95% CI = 0.64–0.96). Whether extended follow-up will show a significant association has yet to be determined.

Colon cancer rates did not differ between groups, but the number of polyps and adenomas reported was significantly lower in the dietary intervention group.¹ The rate of colon cancer will also be included in the extended follow-up study of the WHI.

Risk factors for coronary heart disease in both groups—including levels of serum total cholesterol and serum low-density lipoprotein cholesterol, body weight, body mass index, diastolic blood pressure, and factor VIIc—improved slightly, but at year 3 of the trial, differences in overall rates of coronary heart disease and stroke in the two groups were not statistically significant.² In addition, the low-fat diet intervention was associated with a reduction in blood estradiol concentrations between baseline and year 1.³ At the end of the study, however, differences in rates of breast cancer, colorectal cancer, and heart disease between the two groups were not statistically significant.

RESULTS OF DIETARY MODIFICATIONS

Fat as a percentage of total calories

At the beginning of the WHI, all participants reported consuming an average of 35% of their caloric intake from fat (Table 1). At 1 year from baseline, the fat intake decreased to 24.3% in the intervention group (short of the study goal of 20%); this level had risen again to 26.7% by year 3 and to 28.8% at the end of the study. Stratified by quartile, women who achieved the greatest reductions in saturated and trans-fatty acids or the largest increases in their intake of fruits and vegetables appeared to have a moderate reduction in the risk of coronary heart disease.² Women in the comparison group also decreased their fat intake initially, but to a lesser degree, and gradually increased it again thereafter. The mean net difference in self-reported total fat intake between the intervention group and the comparison group at 6 years was 8.2% (P < .001) (study goal, 13%).^1–3

Intake of fruits, vegetables, and grains

At baseline, fruit and vegetable intake averaged 3.6 servings per day (Table 1). In the intervention group, this increased to 5.1 servings per day at year 1, and to 5.2 servings at year 3, but at the end of the study it had decreased to 4.9 servings.

Women in the intervention group were eating 4.7 servings of grains per day at baseline. This increased to 5.1 servings at year 1 and then decreased to 4.6 servings at year 3 and to 4.3 servings at the end of the study. It seems that as the women grew older their determination to increase servings of these foods diminished.

Proponents of some currently popular diets blame weight gain on a higher intake of carbohydrates, but the women following the WHI low-fat diet did not gain weight.²

Total fat vs saturated fat

Intake of total fat and saturated fat decreased in the intervention group during the study, but the difference between fat intake in the intervention group and that in the comparison group did not reach the degree expected.

At year 1, total fat as a percentage of total caloric intake was 10.8 percentage points below that of the comparison group, whereas the study expected difference was 13.0. At the end of the trial, the difference was only 8.2 percentage points, whereas the expected difference was 11.0.

Intake of all fatty acids (saturated and unsaturated) decreased at year 1, but then went back up slightly by the end of the trial but did not exceed baseline levels, and saturated fatty acids remained well below baseline levels: 9.5% vs 12.5% of caloric intake at baseline.⁴

INTERPRETING THE RESULTS

It might be tempting to dismiss the results of the WHI dietary intervention trial as not significant and therefore not meaningful. This would be unfortunate. The trial had some remarkable accomplishments and offers important lessons for future investigations.

The initial reductions in total fat intake were impressive, and women who had the highest total fat intake at baseline achieved the greatest reduction of total fat (to less than 22% of total calories).³ Nonetheless, the dietary intervention goal of less than 20% of calories from fat was not achieved despite intensive dietary counseling and a highly motivated study population. Thus, this dietary fat target may not be reasonable in the general population.

Also, despite the absence of targeted intervention on specific fatty acids, the observed blood cholesterol levels were as expected based on the well-known formula of Mensink and Katan,¹⁴ which incorporates information on changes in saturated fat, polyunsaturated fat, and dietary cholesterol intake. The predicted reduction in low-density lipoprotein cholesterol was 2.7 mg/dL; the observed reduction was 2.3 mg/dL.² This illustrates that with greater modifications in specific known dietary risk factors for cardiovascular disease, such as saturated fatty acids, cholesterol, and unsaturated fatty acids, blood cholesterol levels respond in a predictable fashion. This was presumably not observed in WHI precisely because no goals and objectives were provided to participants for intake of saturated or polyunsaturated fatty acids.

Recent findings from the Optimal Macronutrient Intake Trial to Prevent Heart Disease (OmniHeart)¹⁵ further highlight differences in the total cholesterol response to diets of varying macronutrient (carbohydrate, protein, fat) content compared with the WHI dietary intervention.¹⁵ Participants in OmniHeart had reductions in levels of low-density lipoprotein cholesterol that were predictable from the changes reported in intake of saturated fatty acids. Presumably, the results of the WHI intervention would have been similar if the study had included this level of detail.

QUESTIONS REMAIN

Questions from the WHI that need consideration for future clinical applications include whether the study population may have already been “too old” to achieve a benefit from dietary modification, and whether the best timing for dietary intervention might be earlier adulthood with sustained changes in saturated fat, cholesterol, and unsaturated fat intake throughout life. Future subgroup analyses based on age at baseline will need to address these questions. Likewise, a longer follow-up period may be needed for a definitive evaluation of the impact of a regular low-fat diet on different health outcomes.

As reported by Patterson et al,¹⁶ the major contributors to total dietary fat intake at baseline were “added fats” such as sauces, gravies, butter, and margarines (25.1% of fat intake), followed by meats (20.9% of fat intake), and desserts (12.8% of fat intake). These findings highlight target areas for future interventions in women of this age group.

Another issue is how to standardize the dietary intervention from one clinical center to another—ie, to minimize differences in how each clinical center manages the study patients. Such differences were noted in WHI and other studies.¹⁷ Despite standardized training in delivering the dietary intervention, nutritionists encountered regional and cultural differences that required tailoring the dietary intervention to their patients’ needs. Staff turnover, an unavoidable phenomenon in long-term studies, has previously been reported to negatively influence dietary adherence.¹⁸

LIMITATIONS

A major limitation of diet modification research in general is the self-reporting of dietary intake, primarily by a food-frequency questionnaire. Although the use of a questionnaire is the most practical way to obtain dietary data for large studies, systematic biases may exist that obscure true nutrient-outcome relationships.¹⁹ Biomarker studies of energy balance suggest that people who are overweight or obese may under-report energy intake to a greater degree than people who are not overweight.²⁰ Also, we still do not know how to get people to follow a healthy diet, although theories and models abound, such as social learning and cognitive-behavioral theory, and a lack of data limits our understanding of factors related to dietary adherence.^21,22

FUTURE DIRECTIONS IN WHI

The WHI Extension Study is under way and has been funded through the year 2010. Outcomes ascertainment is the primary focus with no ongoing intervention, although the intervention group participants continue to receive a WHI newsletter that simply reiterates the importance of the study and encourages ongoing participation. As of 2006, an estimated 84% of the cohort, including both observational study and clinical trial participants, are involved. Efforts continue to recruit the remaining 16%, but many of these participants now consider themselves too old or too feeble to respond reliably.

In regard to breast cancer, the results published in 2006 are promising, albeit not statistically significant, and definitive statements cannot yet be made. However, postmenopausal women who are eating the diets highest in fat may have the greatest benefit from reductions in total fat.

Other considerations regarding the lack of statistically significant differences between groups may include the possibility that women in the intervention group may have been at lower risk for breast cancer at baseline. Likewise, although the results of the WHI dietary intervention do not include a statistically significant impact on colorectal cancer outcomes, the significant reduction in polyps and adenomas may later translate into a reduction in invasive cancer risk.

Finally, although no significant reduction was seen in the rate of death due to cardiovascular causes, greater reductions in saturated and trans-fatty acid intake were associated with greater reductions in blood cholesterol and cardiovascular risk.

Numerous subgroup analyses and ongoing assessments of the long-term impact of the diet modification are planned. Further associations are expected to emerge. The current and future results will continue to provide new insights that may lead to new clinical and public health recommendations in the future.

The WHI has raised additional issues that warrant further investigation:

• Will earlier dietary intervention, eg, during premenopausal years or even childhood, alter these results?
• Does the low-fat, high-carbohydrate diet used in WHI facilitate weight maintenance or even weight loss, as proposed by Howard et al²³?
• Do quantitative changes in physical activity and weight control attenuate morbidity and mortality rates beyond changes in diet alone?
• Do vitamin and mineral supplements or hormone therapy alter disease outcomes or quality of life?
• Which behavioral approaches are best suited to the recruitment of patients for dietary intervention trials?

In patients at risk of myocardial infarction or stroke, two antiplatelet drugs are not always better than one. In a large recent trial,^1,2 adding clopidogrel (Plavix) to aspirin therapy did not offer much benefit to a cohort of patients at risk of cardiovascular events, although a subgroup did appear to benefit: those at even higher risk because they already had a history of myocardial infarction, ischemic stroke, or peripheral arterial disease.

These were the principal findings in the Clopidogrel for High Atherothrombotic Risk and Ischemic Stabilization, Management, and Avoidance (CHARISMA) study,^1,2 in which one of us (D.L.B.) was principal investigator.

These findings further our understanding of who should receive dual antiplatelet therapy, and who would be better served with aspirin therapy alone. In this article, we discuss important studies that led up to the CHARISMA trial, review CHARISMA’s purpose and study design, and interpret its results.

PREVENTING ATHEROTHROMBOSIS BY BLOCKING PLATELETS

Platelets are key players in the atherothrom-botic process.^3–5 The Antiplatelet Trialists’ Collaboration,⁶ in a meta-analysis of trials performed up to 1997, calculated that antiplatelet therapy (mostly with aspirin) reduced the vascular mortality rate by 15% in patients with acute or previous vascular disease or some other predisposing condition. Thus, aspirin has already been shown to be effective as primary prevention (ie, in patients at risk but without established vascular disease) and as secondary prevention (ie, in those with established disease).^7,8

Yet many patients have significant vascular events in spite of taking aspirin.⁶ Aspirin failure is thought to be multifactorial, with causes that include weak platelet inhibition, noncompliance, discontinuation due to adverse effects (including severe bleeding), and drug interactions. In addition, aspirin resistance has been linked to worse prognosis and may prove to be another cause of aspirin failure.^9–11

Clopidogrel, an adenosine diphosphate (ADP) receptor antagonist, has also been studied extensively as an antiplatelet agent.^5,12 Several studies have indicated that clopidogrel and ticlopidine (Ticlid, a related drug) may be more potent than aspirin, both in the test tube and in real patients.^13–15

KEY TRIALS LEADING TO CHARISMA

• Clopidogrel is more effective and slightly safer than aspirin as secondary prevention, as shown in the Clopidogrel Versus Aspirin in Patients at Risk of Ischemic Events (CAPRIE) trial.^16–21
• The combination of clopidogrel plus aspirin is more beneficial than placebo plus aspirin in patients with acute coronary syndromes, as shown in the Clopidogrel in Unstable Angina to Prevent Recurrent Ischemic Events (CURE) trial,^22–24 the Clopidogrel as Adjunctive Reperfusion Therapy-Thrombolysis in Myo-car-dial Infarction (CLARITY-TIMI 28) trial,²⁵ and the Clopidogrel and Metoprolol in Myocardial Infarction Trial (COMMIT).²⁶
• The combination of clopidogrel plus aspirin is beneficial in patients undergoing percutaneous coronary interventions, with or without drug-eluting stent placement,^27–30 as shown in the Clopidogrel for the Reduction of Events During Observation (CREDO) trial,²⁸ the Effect of Clopidogrel Pretreatment Before Percutaneous Coronary Intervention in Patients With ST-Elevation Myocardial Infarction With Fibrinolytics (PCI-CLARITY) study,²⁹ and the Effects of Pre-treatment With Clopidogrel and Aspirin Followed by Long-term Therapy in Patients Undergoing Percutaneous Coronary Intervention (PCI-CURE) study.³⁰ In fact, most patients undergoing percutaneous interventions now receive a loading dose of clopidogrel before the procedure and continue to take it for up to 1 year afterward. However, the ideal long-term duration of clopidogrel treatment is still under debate.

In view of these previous studies, we wanted to test dual antiplatelet therapy in a broader population at high risk of atherothrombosis, ie, in patients with either established vascular disease or with multiple risk factors for it.

CHARISMA STUDY DESIGN

CHARISMA was a prospective, randomized, double-blind, placebo-controlled study of the efficacy and safety of clopidogrel plus aspirin vs placebo plus aspirin in patients at high risk of cardiovascular events.

A total of 15,603 patients, all older than 45 years, were randomly assigned to receive clopidogrel 75 mg/day plus aspirin 75 to 162 mg/day or placebo plus aspirin, in addition to standard therapy as directed by individual clinicians (eg, statins, beta-blockers). Patients were followed up at 1, 3, and 6 months and every 6 months thereafter until study completion, which occurred after 1,040 primary efficacy end points. The median duration of follow-up was 28 months.¹

Patients had to have one of the following to be included: multiple atherothrombotic risk factors, documented coronary disease, documented cerebrovascular disease, or documented peripheral arterial disease (Table 2). Specific exclusion criteria included the use of oral antithrombotic or chronic nonsteroidal anti-inflammatory medications.¹

End points

The primary end point was the combined incidence of the first episode of myocardial infarction or stroke, or death from cardiovascular causes.

The secondary end point was the combined incidence of myocardial infarction, stroke, death from cardiovascular causes, or hospitalization for unstable angina, a transient ischemic attack, or revascularization procedure.

The primary safety end point was severe bleeding, as defined in the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO) study³¹ as intracranial hemorrhage, fatal bleeding, or bleeding leading to hemody-namic compromise. Moderate bleeding was defined as bleeding that required transfusion but did not meet the GUSTO definition of severe bleeding.

OVERALL, NO BENEFIT

The rates of the secondary end point were 16.7% vs 17.9% (absolute risk reduction 1.2%; relative risk reduction 8%; P = .04).

Possible benefit in symptomatic patients

In a prespecified analysis, patients were classified as being “symptomatic” (having documented cardiovascular disease, ie, coronary, cerebrovascular, or symptomatic peripheral arterial disease) or “asymptomatic” (having multiple risk factors without established cardiovascular disease).¹

In the symptomatic group (n = 12,153), the primary end point was reached in 6.9% of patients treated with clopidogrel vs 7.9% with placebo (absolute risk reduction 1.0%; relative risk reduction 13%; P = .046). The 3,284 asymptomatic patients showed no benefit; the rate of the primary end point for the clopido-grel group was 6.6% vs 5.5% in the placebo group (P = .20).

In a post hoc analysis, we examined the data from 9,478 patients who were similar to those in the CAPRIE study (ie, with documented prior myocardial infarction, prior ischemic stroke, or symptomatic peripheral arterial disease). The rate of cardiovascular death, myocardial infarction, or stroke was 8.8% in the placebo-plus-aspirin group and 7.3% in the clopidogrel-plus-aspirin group (absolute risk reduction 1.5%; relative risk reduction 17%; P = .01; Figure 1).²

HOW SHOULD WE INTERPRET THESE FINDINGS?

CHARISMA was the first trial to evaluate whether adding clopidogrel to aspirin therapy would reduce the rates of vascular events and death from cardiovascular causes in stable patients at risk of ischemic events. As in other trials, the benefit of clopidogrel-plus-aspirin therapy was weighed against the risk of bleeding with this regimen. How are we to interpret the findings?

• In the group with multiple risk factors but without clearly documented cardiovascular disease, there was no benefit—and there was an increase in moderate bleeding. Given these findings, physicians should not prescribe dual antiplatelet therapy for primary prevention in patients without known vascular disease.
• A potential benefit was seen in a prespecified subgroup who had documented cardiovascular disease. Given the limitations of subgroup analysis, however, and given the increased risk of moderate bleeding, this positive result should be interpreted with some degree of caution.
• CHARISMA suggests that there may be benefit of protracted dual antiplatelet therapy in stable patients with documented prior ischemic events.

A possible reason for the observed lack of benefit in the overall cohort but the positive results in the subgroups with established vascular disease is that plaque rupture and thrombosis may be a precondition for dual antiplatelet therapy to work.

Another possibility is that, although we have been saying that diabetes mellitus (one of the possible entry criteria in CHARISMA) is a “coronary risk equivalent,” this may not be absolutely true. Although it had been demonstrated that patients with certain risk factors, such as diabetes, have an incidence of ischemic events similar to that in patients with prior MI and should be considered for antiplatelet therapy to prevent vascular events,³² more recent data have shown that patients with prior ischemic events are at much higher risk than patients without ischemic events, even if the latter have diabetes.^33,34

• The observation in CHARISMA that the incremental bleeding risk of dual antiplatelet therapy vs aspirin does not persist beyond a year in patients who have tolerated therapy for a year without a bleeding event may affect the decision to continue clopidogrel beyond 1 year, such as in patients with acute coronary syndromes or patients who have received drug-eluting stents.^35,36
• Another important consideration is cost-effectiveness. Several studies have analyzed the impact of cost and found clopidogrel to be cost-effective by preventing ischemic events and adding years of life.^37,38 A recent analysis from CHARISMA also shows cost-effectiveness in the subgroup of patients enrolled with established cardiovascular disease.³⁹ Once clopidogrel becomes generic, the cost-effectiveness will become even better.

Further studies should better define which stable patients with cardiovascular disease should be on more than aspirin alone.

In patients at risk of myocardial infarction or stroke, two antiplatelet drugs are not always better than one. In a large recent trial,^1,2 adding clopidogrel (Plavix) to aspirin therapy did not offer much benefit to a cohort of patients at risk of cardiovascular events, although a subgroup did appear to benefit: those at even higher risk because they already had a history of myocardial infarction, ischemic stroke, or peripheral arterial disease.

These were the principal findings in the Clopidogrel for High Atherothrombotic Risk and Ischemic Stabilization, Management, and Avoidance (CHARISMA) study,^1,2 in which one of us (D.L.B.) was principal investigator.

These findings further our understanding of who should receive dual antiplatelet therapy, and who would be better served with aspirin therapy alone. In this article, we discuss important studies that led up to the CHARISMA trial, review CHARISMA’s purpose and study design, and interpret its results.

PREVENTING ATHEROTHROMBOSIS BY BLOCKING PLATELETS

Platelets are key players in the atherothrom-botic process.^3–5 The Antiplatelet Trialists’ Collaboration,⁶ in a meta-analysis of trials performed up to 1997, calculated that antiplatelet therapy (mostly with aspirin) reduced the vascular mortality rate by 15% in patients with acute or previous vascular disease or some other predisposing condition. Thus, aspirin has already been shown to be effective as primary prevention (ie, in patients at risk but without established vascular disease) and as secondary prevention (ie, in those with established disease).^7,8

Yet many patients have significant vascular events in spite of taking aspirin.⁶ Aspirin failure is thought to be multifactorial, with causes that include weak platelet inhibition, noncompliance, discontinuation due to adverse effects (including severe bleeding), and drug interactions. In addition, aspirin resistance has been linked to worse prognosis and may prove to be another cause of aspirin failure.^9–11

Clopidogrel, an adenosine diphosphate (ADP) receptor antagonist, has also been studied extensively as an antiplatelet agent.^5,12 Several studies have indicated that clopidogrel and ticlopidine (Ticlid, a related drug) may be more potent than aspirin, both in the test tube and in real patients.^13–15

KEY TRIALS LEADING TO CHARISMA

• Clopidogrel is more effective and slightly safer than aspirin as secondary prevention, as shown in the Clopidogrel Versus Aspirin in Patients at Risk of Ischemic Events (CAPRIE) trial.^16–21
• The combination of clopidogrel plus aspirin is more beneficial than placebo plus aspirin in patients with acute coronary syndromes, as shown in the Clopidogrel in Unstable Angina to Prevent Recurrent Ischemic Events (CURE) trial,^22–24 the Clopidogrel as Adjunctive Reperfusion Therapy-Thrombolysis in Myo-car-dial Infarction (CLARITY-TIMI 28) trial,²⁵ and the Clopidogrel and Metoprolol in Myocardial Infarction Trial (COMMIT).²⁶
• The combination of clopidogrel plus aspirin is beneficial in patients undergoing percutaneous coronary interventions, with or without drug-eluting stent placement,^27–30 as shown in the Clopidogrel for the Reduction of Events During Observation (CREDO) trial,²⁸ the Effect of Clopidogrel Pretreatment Before Percutaneous Coronary Intervention in Patients With ST-Elevation Myocardial Infarction With Fibrinolytics (PCI-CLARITY) study,²⁹ and the Effects of Pre-treatment With Clopidogrel and Aspirin Followed by Long-term Therapy in Patients Undergoing Percutaneous Coronary Intervention (PCI-CURE) study.³⁰ In fact, most patients undergoing percutaneous interventions now receive a loading dose of clopidogrel before the procedure and continue to take it for up to 1 year afterward. However, the ideal long-term duration of clopidogrel treatment is still under debate.

In view of these previous studies, we wanted to test dual antiplatelet therapy in a broader population at high risk of atherothrombosis, ie, in patients with either established vascular disease or with multiple risk factors for it.

CHARISMA STUDY DESIGN

CHARISMA was a prospective, randomized, double-blind, placebo-controlled study of the efficacy and safety of clopidogrel plus aspirin vs placebo plus aspirin in patients at high risk of cardiovascular events.

A total of 15,603 patients, all older than 45 years, were randomly assigned to receive clopidogrel 75 mg/day plus aspirin 75 to 162 mg/day or placebo plus aspirin, in addition to standard therapy as directed by individual clinicians (eg, statins, beta-blockers). Patients were followed up at 1, 3, and 6 months and every 6 months thereafter until study completion, which occurred after 1,040 primary efficacy end points. The median duration of follow-up was 28 months.¹

Patients had to have one of the following to be included: multiple atherothrombotic risk factors, documented coronary disease, documented cerebrovascular disease, or documented peripheral arterial disease (Table 2). Specific exclusion criteria included the use of oral antithrombotic or chronic nonsteroidal anti-inflammatory medications.¹

End points

The primary end point was the combined incidence of the first episode of myocardial infarction or stroke, or death from cardiovascular causes.

The secondary end point was the combined incidence of myocardial infarction, stroke, death from cardiovascular causes, or hospitalization for unstable angina, a transient ischemic attack, or revascularization procedure.

The primary safety end point was severe bleeding, as defined in the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO) study³¹ as intracranial hemorrhage, fatal bleeding, or bleeding leading to hemody-namic compromise. Moderate bleeding was defined as bleeding that required transfusion but did not meet the GUSTO definition of severe bleeding.

OVERALL, NO BENEFIT

The rates of the secondary end point were 16.7% vs 17.9% (absolute risk reduction 1.2%; relative risk reduction 8%; P = .04).

Possible benefit in symptomatic patients

In a prespecified analysis, patients were classified as being “symptomatic” (having documented cardiovascular disease, ie, coronary, cerebrovascular, or symptomatic peripheral arterial disease) or “asymptomatic” (having multiple risk factors without established cardiovascular disease).¹

In the symptomatic group (n = 12,153), the primary end point was reached in 6.9% of patients treated with clopidogrel vs 7.9% with placebo (absolute risk reduction 1.0%; relative risk reduction 13%; P = .046). The 3,284 asymptomatic patients showed no benefit; the rate of the primary end point for the clopido-grel group was 6.6% vs 5.5% in the placebo group (P = .20).

In a post hoc analysis, we examined the data from 9,478 patients who were similar to those in the CAPRIE study (ie, with documented prior myocardial infarction, prior ischemic stroke, or symptomatic peripheral arterial disease). The rate of cardiovascular death, myocardial infarction, or stroke was 8.8% in the placebo-plus-aspirin group and 7.3% in the clopidogrel-plus-aspirin group (absolute risk reduction 1.5%; relative risk reduction 17%; P = .01; Figure 1).²

HOW SHOULD WE INTERPRET THESE FINDINGS?

CHARISMA was the first trial to evaluate whether adding clopidogrel to aspirin therapy would reduce the rates of vascular events and death from cardiovascular causes in stable patients at risk of ischemic events. As in other trials, the benefit of clopidogrel-plus-aspirin therapy was weighed against the risk of bleeding with this regimen. How are we to interpret the findings?

• In the group with multiple risk factors but without clearly documented cardiovascular disease, there was no benefit—and there was an increase in moderate bleeding. Given these findings, physicians should not prescribe dual antiplatelet therapy for primary prevention in patients without known vascular disease.
• A potential benefit was seen in a prespecified subgroup who had documented cardiovascular disease. Given the limitations of subgroup analysis, however, and given the increased risk of moderate bleeding, this positive result should be interpreted with some degree of caution.
• CHARISMA suggests that there may be benefit of protracted dual antiplatelet therapy in stable patients with documented prior ischemic events.

A possible reason for the observed lack of benefit in the overall cohort but the positive results in the subgroups with established vascular disease is that plaque rupture and thrombosis may be a precondition for dual antiplatelet therapy to work.

Another possibility is that, although we have been saying that diabetes mellitus (one of the possible entry criteria in CHARISMA) is a “coronary risk equivalent,” this may not be absolutely true. Although it had been demonstrated that patients with certain risk factors, such as diabetes, have an incidence of ischemic events similar to that in patients with prior MI and should be considered for antiplatelet therapy to prevent vascular events,³² more recent data have shown that patients with prior ischemic events are at much higher risk than patients without ischemic events, even if the latter have diabetes.^33,34

• The observation in CHARISMA that the incremental bleeding risk of dual antiplatelet therapy vs aspirin does not persist beyond a year in patients who have tolerated therapy for a year without a bleeding event may affect the decision to continue clopidogrel beyond 1 year, such as in patients with acute coronary syndromes or patients who have received drug-eluting stents.^35,36
• Another important consideration is cost-effectiveness. Several studies have analyzed the impact of cost and found clopidogrel to be cost-effective by preventing ischemic events and adding years of life.^37,38 A recent analysis from CHARISMA also shows cost-effectiveness in the subgroup of patients enrolled with established cardiovascular disease.³⁹ Once clopidogrel becomes generic, the cost-effectiveness will become even better.

Further studies should better define which stable patients with cardiovascular disease should be on more than aspirin alone.

In patients at risk of myocardial infarction or stroke, two antiplatelet drugs are not always better than one. In a large recent trial,^1,2 adding clopidogrel (Plavix) to aspirin therapy did not offer much benefit to a cohort of patients at risk of cardiovascular events, although a subgroup did appear to benefit: those at even higher risk because they already had a history of myocardial infarction, ischemic stroke, or peripheral arterial disease.

These were the principal findings in the Clopidogrel for High Atherothrombotic Risk and Ischemic Stabilization, Management, and Avoidance (CHARISMA) study,^1,2 in which one of us (D.L.B.) was principal investigator.

These findings further our understanding of who should receive dual antiplatelet therapy, and who would be better served with aspirin therapy alone. In this article, we discuss important studies that led up to the CHARISMA trial, review CHARISMA’s purpose and study design, and interpret its results.

PREVENTING ATHEROTHROMBOSIS BY BLOCKING PLATELETS

Platelets are key players in the atherothrom-botic process.^3–5 The Antiplatelet Trialists’ Collaboration,⁶ in a meta-analysis of trials performed up to 1997, calculated that antiplatelet therapy (mostly with aspirin) reduced the vascular mortality rate by 15% in patients with acute or previous vascular disease or some other predisposing condition. Thus, aspirin has already been shown to be effective as primary prevention (ie, in patients at risk but without established vascular disease) and as secondary prevention (ie, in those with established disease).^7,8

Yet many patients have significant vascular events in spite of taking aspirin.⁶ Aspirin failure is thought to be multifactorial, with causes that include weak platelet inhibition, noncompliance, discontinuation due to adverse effects (including severe bleeding), and drug interactions. In addition, aspirin resistance has been linked to worse prognosis and may prove to be another cause of aspirin failure.^9–11

Clopidogrel, an adenosine diphosphate (ADP) receptor antagonist, has also been studied extensively as an antiplatelet agent.^5,12 Several studies have indicated that clopidogrel and ticlopidine (Ticlid, a related drug) may be more potent than aspirin, both in the test tube and in real patients.^13–15

KEY TRIALS LEADING TO CHARISMA

• Clopidogrel is more effective and slightly safer than aspirin as secondary prevention, as shown in the Clopidogrel Versus Aspirin in Patients at Risk of Ischemic Events (CAPRIE) trial.^16–21
• The combination of clopidogrel plus aspirin is more beneficial than placebo plus aspirin in patients with acute coronary syndromes, as shown in the Clopidogrel in Unstable Angina to Prevent Recurrent Ischemic Events (CURE) trial,^22–24 the Clopidogrel as Adjunctive Reperfusion Therapy-Thrombolysis in Myo-car-dial Infarction (CLARITY-TIMI 28) trial,²⁵ and the Clopidogrel and Metoprolol in Myocardial Infarction Trial (COMMIT).²⁶
• The combination of clopidogrel plus aspirin is beneficial in patients undergoing percutaneous coronary interventions, with or without drug-eluting stent placement,^27–30 as shown in the Clopidogrel for the Reduction of Events During Observation (CREDO) trial,²⁸ the Effect of Clopidogrel Pretreatment Before Percutaneous Coronary Intervention in Patients With ST-Elevation Myocardial Infarction With Fibrinolytics (PCI-CLARITY) study,²⁹ and the Effects of Pre-treatment With Clopidogrel and Aspirin Followed by Long-term Therapy in Patients Undergoing Percutaneous Coronary Intervention (PCI-CURE) study.³⁰ In fact, most patients undergoing percutaneous interventions now receive a loading dose of clopidogrel before the procedure and continue to take it for up to 1 year afterward. However, the ideal long-term duration of clopidogrel treatment is still under debate.

In view of these previous studies, we wanted to test dual antiplatelet therapy in a broader population at high risk of atherothrombosis, ie, in patients with either established vascular disease or with multiple risk factors for it.

CHARISMA STUDY DESIGN

CHARISMA was a prospective, randomized, double-blind, placebo-controlled study of the efficacy and safety of clopidogrel plus aspirin vs placebo plus aspirin in patients at high risk of cardiovascular events.

A total of 15,603 patients, all older than 45 years, were randomly assigned to receive clopidogrel 75 mg/day plus aspirin 75 to 162 mg/day or placebo plus aspirin, in addition to standard therapy as directed by individual clinicians (eg, statins, beta-blockers). Patients were followed up at 1, 3, and 6 months and every 6 months thereafter until study completion, which occurred after 1,040 primary efficacy end points. The median duration of follow-up was 28 months.¹

Patients had to have one of the following to be included: multiple atherothrombotic risk factors, documented coronary disease, documented cerebrovascular disease, or documented peripheral arterial disease (Table 2). Specific exclusion criteria included the use of oral antithrombotic or chronic nonsteroidal anti-inflammatory medications.¹

End points

The primary end point was the combined incidence of the first episode of myocardial infarction or stroke, or death from cardiovascular causes.

The secondary end point was the combined incidence of myocardial infarction, stroke, death from cardiovascular causes, or hospitalization for unstable angina, a transient ischemic attack, or revascularization procedure.

The primary safety end point was severe bleeding, as defined in the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO) study³¹ as intracranial hemorrhage, fatal bleeding, or bleeding leading to hemody-namic compromise. Moderate bleeding was defined as bleeding that required transfusion but did not meet the GUSTO definition of severe bleeding.

OVERALL, NO BENEFIT

The rates of the secondary end point were 16.7% vs 17.9% (absolute risk reduction 1.2%; relative risk reduction 8%; P = .04).

Possible benefit in symptomatic patients

In a prespecified analysis, patients were classified as being “symptomatic” (having documented cardiovascular disease, ie, coronary, cerebrovascular, or symptomatic peripheral arterial disease) or “asymptomatic” (having multiple risk factors without established cardiovascular disease).¹

In the symptomatic group (n = 12,153), the primary end point was reached in 6.9% of patients treated with clopidogrel vs 7.9% with placebo (absolute risk reduction 1.0%; relative risk reduction 13%; P = .046). The 3,284 asymptomatic patients showed no benefit; the rate of the primary end point for the clopido-grel group was 6.6% vs 5.5% in the placebo group (P = .20).

In a post hoc analysis, we examined the data from 9,478 patients who were similar to those in the CAPRIE study (ie, with documented prior myocardial infarction, prior ischemic stroke, or symptomatic peripheral arterial disease). The rate of cardiovascular death, myocardial infarction, or stroke was 8.8% in the placebo-plus-aspirin group and 7.3% in the clopidogrel-plus-aspirin group (absolute risk reduction 1.5%; relative risk reduction 17%; P = .01; Figure 1).²

HOW SHOULD WE INTERPRET THESE FINDINGS?

CHARISMA was the first trial to evaluate whether adding clopidogrel to aspirin therapy would reduce the rates of vascular events and death from cardiovascular causes in stable patients at risk of ischemic events. As in other trials, the benefit of clopidogrel-plus-aspirin therapy was weighed against the risk of bleeding with this regimen. How are we to interpret the findings?

• In the group with multiple risk factors but without clearly documented cardiovascular disease, there was no benefit—and there was an increase in moderate bleeding. Given these findings, physicians should not prescribe dual antiplatelet therapy for primary prevention in patients without known vascular disease.
• A potential benefit was seen in a prespecified subgroup who had documented cardiovascular disease. Given the limitations of subgroup analysis, however, and given the increased risk of moderate bleeding, this positive result should be interpreted with some degree of caution.
• CHARISMA suggests that there may be benefit of protracted dual antiplatelet therapy in stable patients with documented prior ischemic events.

A possible reason for the observed lack of benefit in the overall cohort but the positive results in the subgroups with established vascular disease is that plaque rupture and thrombosis may be a precondition for dual antiplatelet therapy to work.

Another possibility is that, although we have been saying that diabetes mellitus (one of the possible entry criteria in CHARISMA) is a “coronary risk equivalent,” this may not be absolutely true. Although it had been demonstrated that patients with certain risk factors, such as diabetes, have an incidence of ischemic events similar to that in patients with prior MI and should be considered for antiplatelet therapy to prevent vascular events,³² more recent data have shown that patients with prior ischemic events are at much higher risk than patients without ischemic events, even if the latter have diabetes.^33,34

• The observation in CHARISMA that the incremental bleeding risk of dual antiplatelet therapy vs aspirin does not persist beyond a year in patients who have tolerated therapy for a year without a bleeding event may affect the decision to continue clopidogrel beyond 1 year, such as in patients with acute coronary syndromes or patients who have received drug-eluting stents.^35,36
• Another important consideration is cost-effectiveness. Several studies have analyzed the impact of cost and found clopidogrel to be cost-effective by preventing ischemic events and adding years of life.^37,38 A recent analysis from CHARISMA also shows cost-effectiveness in the subgroup of patients enrolled with established cardiovascular disease.³⁹ Once clopidogrel becomes generic, the cost-effectiveness will become even better.

Further studies should better define which stable patients with cardiovascular disease should be on more than aspirin alone.