Original Research

It is thought that, by placing more emphasis on soft-tissue preservation, minimally invasive surgery total hip arthroplasty (MIS-THA) results in less soft-tissue trauma, less blood loss, and earlier recovery.^1-3 Despite these improvements over standard methods, there is a concern that the vigorous retraction needed for proper visualization through smaller incisions could injure soft tissues.^4-7 Single-incision direct anterior THA (DA-THA) has gained in popularity because of the true intermuscular/internervous plane through which the procedure can be performed with relatively minimal muscle dissection using MIS techniques.^8,9 This approach may offer quicker recovery and superior stability in comparison with nonintermuscular methods, which unavoidably cause more muscle damage.^10-12

Although the evidence of these early gains is encouraging, several studies have found high complication rates with DA-THA.^8,13-17 Noted disadvantages include a steep learning curve, lateral femoral cutaneous neurapraxia, need for a specialized table, and higher fracture and wound complication rates. Not surprisingly, with increased surgeon experience, the complication rate decreased substantially.^14,15 However, wound-related complications remained steady, with 2 recent large studies reporting rates of 4.6% and 2.1%.^14,15 The thin anterior skin, high tensional forces along the groin crease and perpendicular to the typical DA incision, and less resilient soft-tissue envelope are postulated reasons for wound-related issues, which are likely magnified in patients who are more obese.^15,16

A novel device designed to lessen tissue damage is the ring retractor (Figure 1). Used initially in general surgery and obstetrics, it consists of 2 semirigid polymer rings connected by a flexible cylindrical polymer membrane.^18-20 The lower ring is tucked and anchored underneath the wound edge, and then the upper ring is rolled down and cinched onto the skin. The resultant tension on the polymer sleeve—imparted by the rigidity of the ring—provides strong, evenly distributed wound-edge retraction. It also provides a physical barrier between the wound edge and the rest of the operative field. Proponents of the ring retractor claim increased wound-edge moisture, less bruising, and reduced local trauma compared with standard metal retractors alone.

Wound-edge retractor forces are doubled during MIS-THA compared with conventional THA.^14-20 This may explain reports of worse scar cosmesis with MIS-THA. Given the theoretical benefits of minimized wound-edge trauma, the ring retractor may improve scar appearance compared with standard retraction alone. Any clinically relevant effect on cosmesis should be readily apparent to justify use of the retractor in this regard. Although some surgeons routinely use the device for primary THA, it has not been the subject of any recent orthopedic studies.

In the present study, we prospectively investigated wound cosmesis with and without use of the ring retractor in patients undergoing DA-THA.

Materials and Methods

This prospective, single-center, randomized study was reviewed and approved by the institutional review board at our facility. Consent was obtained from all participating patients.

We evaluated 50 surgical incisions in 48 patients. Eligible participants were over age 18 years and undergoing primary DA-THA. Exclusion criteria included previous surgery on the affected hip, a pathological hip condition requiring an extensile exposure, systemic inflammatory illness, chronic corticosteroid use, and dermatologic abnormality of the incisional area. One patient was having simultaneous bilateral THAs, and another was having staged bilateral THAs. Each hip in these patients was given its own case number and treated separately. Of the 49 patients who met all the inclusion criteria, only 1 decided not to participate (Figure 2).

Stratified randomization with permuted block size (sex, body mass index [BMI]) was used to assign patients in a 1:1 ratio to either the treatment group or the control group. In the treatment group, the Protractor Incision Protector and Retractor (Gyrus ACMI, Southborough, Massachusetts) was used with standard metal retractors. In the control group, only standard metal retractors were used. Patients were blinded to their group assignments, and surgeons were informed about each assignment only after the initial incision was made.

Clinical research investigators were blinded to the groups’ prospectively collected data. Collection time points were preoperative clinical visit, day of surgery through discharge, and 2-, 6-, and 12-week postoperative follow-ups. Day-of-surgery data included estimated intraoperative blood loss, operative side, operative time, intraoperative complications, and American Society of Anesthesiologists (ASA) physical status classification. Total length of stay, pain scores (range, 0-10), estimated drain output, and blood-transfusion data were also recorded. To evaluate whether the device had any effect on short-term functional outcome, we collected Harris Hip Scores (HHS) and Short Form–12 (SF-12, Version 2) scores at the preoperative and 6-week postoperative visits. We also documented any wound-healing-related issues or complications that occurred up until the final visit.

To account for any effect of nutrition status on wound healing, we obtained pre-albumin and albumin levels and absolute lymphocyte counts from the preoperative electronic records. We used an albumin level under 3.5 g/dL and an absolute lymphocyte count under 1500/µL for our analysis, as these cutoffs have been associated with wound complications after primary THA.²¹ There is no similarly established threshold for pre-albumin level, so we used values under 20 mg/L based on comparable literature.^22,23

At each postoperative visit, standardized high-resolution images were obtained. At the 12-week visit, patients completed 2 Likert scales regarding their overall opinion of their scars and how their scars compared with their expectations. They also ranked 5 separate THA-related outcomes in order of importance (Appendix).

Photographs were evaluated by 2 blinded plastic surgeons (Dr. Friedman and Dr. Jack) using 2 grading systems—the Stony Brook Scar Evaluation Scale (SBSES)²⁴ (Table 1) and a modified Manchester Scar Scale (MSS)²⁵ (Table 2). We used these systems because they were photograph-based, psychometrically studied, and specifically designed to assess surgical incision healing with established validity and reliability.^24-27 A particular advantage, strictly related to cosmetic outcome, is their validity in scoring scars from high-definition photographs in a different place or time. The SBSES, an ordinal wound evaluation scale that measures short-term cosmetic outcomes, consists of 6 items, each receiving 1 or 0 point, yielding a total score between 0 (worst) and 5 (best). The modified MSS includes a visual analog scale (VAS), which has a vertical hash marked on a 10-cm line and is scored between 0 (excellent) and 10 (poor) to 1 decimal point.^26,28 This value is added to grades on color, surface appearance, contour, and distortion, resulting in a score between 4 (best) and 24 (worst). The primary outcome measures were Likert-scale responses obtained at final visit and SBSES/MSS scores for each visit; 12-week scores were the primary end point.

Operative Procedure

Experienced fellowship-trained orthopedic surgeons performed all procedures. A modified Hueter approach was used for exposure.⁹ Mean incision length was about 12 cm. For the treatment group, the ring retractor was inserted at the level of the tensor fascia, with the inferior ring resting between the fascia and the subcutaneous layer and the superior ring cinched over the skin (Figure 3). The device is made in 4 different sizes for incisions from 2.5 to 17 cm; our study population required only 1 size. Otherwise, the surgical protocol was based on that described by Matta and colleagues.⁸ Wound closure (over a drain) was performed according to a standardized protocol—running No. 1 Vicryl suture for the superficial tensor fascia, interrupted 2-0 Vicryl for the deep dermal layer, and subcutaneous 4-0 Monocryl for the skin followed by application of Dermabond (Ethicon, Somerville, New Jersey) and Tegaderm +Pad (3M, St. Paul, Minnesota) for outer dressing, which was replaced on postoperative day 2 and removed at the 2-week visit.

Statistical Methods

An a priori sample-size calculation was performed. Power performed in a base of a prior study that evaluated anterolateral and posterolateral THA scars using a VAS, a component of the MSS, suggested a sample size of 16 per group to detect the minimal clinically important difference of 1.5 cm: SD (σ) = 1.5 cm, α = 0.05, β = 0.20.^29,30 In addition, a general estimate for detecting a 1-unit change on an ordinal scale (σ = 1.0, α = 0.05, β = 0.20) resulted in the same number. We conservatively decided to enroll 25 patients per arm in case of larger true variance.

The Wilcoxon rank sum test was used for comparisons of continuous data between groups. Differences between means were analyzed with 2-sided t tests. Categorical data were compared with the Pearson χ² test or the Fisher exact test, as indicated. Ordinal ranking scores were compared with the Mantel-Haenszel test. Multivariate logistic regression was applied to identify the significant independent predictors of better scar grades for each surgeon by considering candidate variables with Ps < .20 in the univariate analysis.

Results

We found no differences in demographic or perioperative characteristics between treatment and control groups (Tables 3, 4). The groups showed similar mean improvements in their respective 6-week HHS (38.7 and 36.4 points; P = .65), SF-12 physical component summary scores (11.8 and 14.5 points; P = .37), and SF-12 mental component summary scores (5.1 and 3.7; P = .70).

Patient questionnaire outcomes are listed in Table 5. For the control group, 25/25 image sets were obtained at the 2-week visit, 25/25 at the 6-week visit, and 24/25 at the 12-week visit. For the treatment group, there were 23/25, 24/25, and 23/25 images sets, respectively.

When surgeon scoring was analyzed separately, SBSES and MSS scores were similar between treatment and control groups, with 1 exception: 2-week MSS scores were better for the treatment group according to surgeon A (P = .026). When grades were averaged, SBSES scores were again similar at all time points (Figure 4A); MSS scores were better for the treatment group at 2 weeks (P = .036) and equivalent at all other time points (Figure 4B). For the SBSES, Spearman correlation coefficient ρ with 95% confidence interval (CI) was 0.37
(95% CI, 0.08-0.66) at 2 weeks, 0.48 (95% CI, 0.20-0.76) at 6 weeks, and 0.62 (95% CI, 0.33-0.91) at 12 weeks. Following the same pattern for the MSS, ρ was 0.20 (95% CI, –0.09 to 0.49), 0.51 (95% CI, 0.23-0.79), and 0.32 (95% CI, 0.03-0.61).

Independent multivariate analysis revealed that age over 65 years was a significant predictor of worse scores. On SBSES, the odds ratio (OR) was 1.15 (95% CI, 1.07-1.24) for surgeon A and 1.11 (95% CI, 1.05-1.18) for surgeon B. On MSS, the OR was 0.89 (95% CI, 0.84-0.94) for surgeon A and 0.95 (95% CI, 0.91-0.99) for surgeon B. The likelihood of having worse SBSES scores according to surgeon A was 4.72 times higher if the pre-albumin level was under 20 mg/L (95% CI, 1.15-19.36). Albumin level under 3.5 g/dL and absolute lymphocyte count under 1500 cells/µL were not found to be independent predictors of poorer scores.

Patients’ overall opinion (P = .63) and assessment of their scars relative to expectations (P = .25) on the Likert scales were not different between groups. More scars exceeded patients’ expectations and had more excellent ratings in the control group. The 2 groups were similar with regard to relative importance of various patient-related outcomes. Factors most important to overall outcome were relief of hip pain, followed by implant longevity and length of recovery. Least important were incision-related variables.

There were only 3 minor noninfectious wound complications (6%), 2 in the treatment group and 1 in the control group. In the treatment group, a 67-year-old man with diabetes (ASA class III; BMI, 32.1 kg/m²; received transfusion) had 2 small areas (<5 mm) of superficial ulceration at 6-week follow-up—one at the proximal aspect of the incision and the other near the midpoint along the flexion crease. Both lesions resolved by 12-week follow-up. Also in the treatment group, a 77-year-old woman (ASA class II; BMI, 24.9 kg/m²; received transfusion) at 6 weeks had a spitting suture, which was removed in clinic without further issue. In the control group, a 55-year-old woman (ASA class II; BMI, 27.4 kg/m²) had a suture reaction near the proximal aspect of her incision 3 weeks after surgery. This reaction, which presented as a mild, localized erythema without pain, tenderness, or drainage, resolved by 6-week follow-up. None of these wound complications required intervention beyond observation.

Discussion

This study was designed to provide a bipartisan measure of wound-healing cosmesis after DA-THA. Scar evaluation by blinded plastic surgeons served as a standardized, clinical assessment, whereas the patient questionnaire offered a more subjective appraisal. The modified MMS²⁵ and the SBSES²⁴ are the only 2 wound-grading systems designed and validated for photographic assessment of postsurgical scars. Most scar evaluation schemes pertain to burn or traumatic scars.^26,27,31 As a result, many earlier studies intending to compare incisional scars used poorly suited evaluation systems.

The current literature includes reports on 3 studies with scoring-based scar assessment in THA; all used grading systems designed for either burns or traumatic wounds, but 2 also used a VAS.^32-34 VASs have been validated for measuring wound cosmesis but are entirely subjective and without structure and provide no feedback as to why a scar was rated good or bad.²⁴ Mow and colleagues³² prospectively compared scars after standard posterior or MIS approaches and found no differences according to a scoring system intended for burn scars. In our study population, we found no group differences in patients’ cosmesis of their scars.

Although scars can take a year or longer to fully mature, researchers from the University of Michigan discovered that scar appearance at 1 year correlates highly with cosmesis 12 weeks after closure, though poorly with cosmesis 10 days after closure.³⁵ Therefore, any observed differences in scar cosmesis between groups at 12-week follow-up would likely persist, whereas differences at 2-week follow-up would have little bearing on ultimate appearance. For this reason, our primary outcome measure was healing process and cosmesis at 12 weeks. High wound complication rates have been reported for MIS-DA-THA.^8,14-16 Jewett and Collis¹⁵ noted a 4.6% wound complication rate (3% noninfectious ulcerative dehiscence, 1.6% superficial infection), which is comparable to the 6% rate found in this study. However, there likely is some variability across studies in what constitutes a wound complication or superficial infection. Of our 3 wound complications—stitch reaction, spitting suture, small noninfectious ulceration—only the ulceration was of a severity similar to that reported by Jewett and Collis.¹⁵ Matta and colleagues⁸ reported only 3 wound complications (in 494 patients), all severe enough to require operative intervention. One explanation for this low complication rate is use of a ring retractor, as it is routinely depicted in their technique paper. However, no specific reference is made to gauge how often the device was used.

Rates of superficial infection after DA-THA range from 0.6% to 1.6% in 3 large observational studies (combined deep infection rate, 0.43%).^8,14,15 In 2 of these studies, all patients with superficial infection underwent formal débridement, though none developed deep infection. A prospective randomized study of 221 patients who underwent colorectal surgery—where perioperative infectious morbidity ranges from 25% to 50%—found that ring retractor use significantly reduced superficial wound infection rates (8.1% vs 0%). A significant reduction in wound infection was shown in a similarly designed study involving 48 patients who had open appendectomy (14.6% vs 1.6%). The device had no effect on deep infection in either general surgery study. The wound infection rates reported in these general surgery studies are markedly higher than those in our study population. As a result, the effect of the ring retractor on wound infection in DA-THA may be less. Regardless of the effect on deep infection, fewer superficial infections, which often require operative intervention, would be of considerable benefit.

Below-threshold albumin level and absolute lymphocyte count have been associated with wound-healing complications after hip replacement.²¹ In the present study, pre-albumin level under 20 mg/L was the only nutritional marker predictive of poor wound appearance, but this finding was seen only in SBSES scores from surgeon A. Subgroup analysis did not reveal any relationship between wound appearance and any of the recorded demographic or perioperative variables, but for a small predictive influence with age over 65 years.

This study had some limitations. Our findings cannot be generalized to all patients who undergo THA, as only DA incisions were studied. Results also may not be generalizable to non-fellowship-trained orthopedists. In addition, selection bias likely resulted from including patients already selected for the DA approach. Using digital images for evaluation (vs real-life evaluation) may have affected reliability as well. Last, by not incorporating texture, we omitted a potentially informative feature from scoring.

It is paramount that surgeons undergo diligent training before undertaking this approach for minimizing unwanted results; furthermore, higher early complication rates level off with increased surgeon experience.^14,36,37 We recommend meticulous soft-tissue handling, cautious retraction, and careful patient selection (relative contraindication for patients with an abdominal pannus overlying the incision) as primary measures for minimizing incisional trauma and potential wound-healing complications.³⁸ Preservation of the tensor fascia is also crucial,³⁹ as it is the only closable layer separating deep and superficial compartments. Without good closure of the tensor fascia, there is no containment or tamponade of deep bleeding, which can facilitate hematoma formation.

In the population studied, we found no significant long-term differences in cosmetic appearance (based on clinician or patient evaluation) between wounds managed with and without the ring retractor. Our data do not support routine use of the ring retractor, during DA-THA, for improved wound cosmesis. Whether the device has any significant role in reducing the number of wound complications in THA is yet to be determined. Last, the ring retractor may have a role in other areas of orthopedic surgery, such as hip fractures in the elderly or orthopedic oncology. In situations like these, where adequate nutrition and immunocompetency may be lacking, the added protection provided by the device may translate into a more notable benefit than in elective THA.

It is thought that, by placing more emphasis on soft-tissue preservation, minimally invasive surgery total hip arthroplasty (MIS-THA) results in less soft-tissue trauma, less blood loss, and earlier recovery.^1-3 Despite these improvements over standard methods, there is a concern that the vigorous retraction needed for proper visualization through smaller incisions could injure soft tissues.^4-7 Single-incision direct anterior THA (DA-THA) has gained in popularity because of the true intermuscular/internervous plane through which the procedure can be performed with relatively minimal muscle dissection using MIS techniques.^8,9 This approach may offer quicker recovery and superior stability in comparison with nonintermuscular methods, which unavoidably cause more muscle damage.^10-12

Although the evidence of these early gains is encouraging, several studies have found high complication rates with DA-THA.^8,13-17 Noted disadvantages include a steep learning curve, lateral femoral cutaneous neurapraxia, need for a specialized table, and higher fracture and wound complication rates. Not surprisingly, with increased surgeon experience, the complication rate decreased substantially.^14,15 However, wound-related complications remained steady, with 2 recent large studies reporting rates of 4.6% and 2.1%.^14,15 The thin anterior skin, high tensional forces along the groin crease and perpendicular to the typical DA incision, and less resilient soft-tissue envelope are postulated reasons for wound-related issues, which are likely magnified in patients who are more obese.^15,16

A novel device designed to lessen tissue damage is the ring retractor (Figure 1). Used initially in general surgery and obstetrics, it consists of 2 semirigid polymer rings connected by a flexible cylindrical polymer membrane.^18-20 The lower ring is tucked and anchored underneath the wound edge, and then the upper ring is rolled down and cinched onto the skin. The resultant tension on the polymer sleeve—imparted by the rigidity of the ring—provides strong, evenly distributed wound-edge retraction. It also provides a physical barrier between the wound edge and the rest of the operative field. Proponents of the ring retractor claim increased wound-edge moisture, less bruising, and reduced local trauma compared with standard metal retractors alone.

Wound-edge retractor forces are doubled during MIS-THA compared with conventional THA.^14-20 This may explain reports of worse scar cosmesis with MIS-THA. Given the theoretical benefits of minimized wound-edge trauma, the ring retractor may improve scar appearance compared with standard retraction alone. Any clinically relevant effect on cosmesis should be readily apparent to justify use of the retractor in this regard. Although some surgeons routinely use the device for primary THA, it has not been the subject of any recent orthopedic studies.

In the present study, we prospectively investigated wound cosmesis with and without use of the ring retractor in patients undergoing DA-THA.

Materials and Methods

This prospective, single-center, randomized study was reviewed and approved by the institutional review board at our facility. Consent was obtained from all participating patients.

We evaluated 50 surgical incisions in 48 patients. Eligible participants were over age 18 years and undergoing primary DA-THA. Exclusion criteria included previous surgery on the affected hip, a pathological hip condition requiring an extensile exposure, systemic inflammatory illness, chronic corticosteroid use, and dermatologic abnormality of the incisional area. One patient was having simultaneous bilateral THAs, and another was having staged bilateral THAs. Each hip in these patients was given its own case number and treated separately. Of the 49 patients who met all the inclusion criteria, only 1 decided not to participate (Figure 2).

Stratified randomization with permuted block size (sex, body mass index [BMI]) was used to assign patients in a 1:1 ratio to either the treatment group or the control group. In the treatment group, the Protractor Incision Protector and Retractor (Gyrus ACMI, Southborough, Massachusetts) was used with standard metal retractors. In the control group, only standard metal retractors were used. Patients were blinded to their group assignments, and surgeons were informed about each assignment only after the initial incision was made.

Clinical research investigators were blinded to the groups’ prospectively collected data. Collection time points were preoperative clinical visit, day of surgery through discharge, and 2-, 6-, and 12-week postoperative follow-ups. Day-of-surgery data included estimated intraoperative blood loss, operative side, operative time, intraoperative complications, and American Society of Anesthesiologists (ASA) physical status classification. Total length of stay, pain scores (range, 0-10), estimated drain output, and blood-transfusion data were also recorded. To evaluate whether the device had any effect on short-term functional outcome, we collected Harris Hip Scores (HHS) and Short Form–12 (SF-12, Version 2) scores at the preoperative and 6-week postoperative visits. We also documented any wound-healing-related issues or complications that occurred up until the final visit.

To account for any effect of nutrition status on wound healing, we obtained pre-albumin and albumin levels and absolute lymphocyte counts from the preoperative electronic records. We used an albumin level under 3.5 g/dL and an absolute lymphocyte count under 1500/µL for our analysis, as these cutoffs have been associated with wound complications after primary THA.²¹ There is no similarly established threshold for pre-albumin level, so we used values under 20 mg/L based on comparable literature.^22,23

At each postoperative visit, standardized high-resolution images were obtained. At the 12-week visit, patients completed 2 Likert scales regarding their overall opinion of their scars and how their scars compared with their expectations. They also ranked 5 separate THA-related outcomes in order of importance (Appendix).

Photographs were evaluated by 2 blinded plastic surgeons (Dr. Friedman and Dr. Jack) using 2 grading systems—the Stony Brook Scar Evaluation Scale (SBSES)²⁴ (Table 1) and a modified Manchester Scar Scale (MSS)²⁵ (Table 2). We used these systems because they were photograph-based, psychometrically studied, and specifically designed to assess surgical incision healing with established validity and reliability.^24-27 A particular advantage, strictly related to cosmetic outcome, is their validity in scoring scars from high-definition photographs in a different place or time. The SBSES, an ordinal wound evaluation scale that measures short-term cosmetic outcomes, consists of 6 items, each receiving 1 or 0 point, yielding a total score between 0 (worst) and 5 (best). The modified MSS includes a visual analog scale (VAS), which has a vertical hash marked on a 10-cm line and is scored between 0 (excellent) and 10 (poor) to 1 decimal point.^26,28 This value is added to grades on color, surface appearance, contour, and distortion, resulting in a score between 4 (best) and 24 (worst). The primary outcome measures were Likert-scale responses obtained at final visit and SBSES/MSS scores for each visit; 12-week scores were the primary end point.

Operative Procedure

Experienced fellowship-trained orthopedic surgeons performed all procedures. A modified Hueter approach was used for exposure.⁹ Mean incision length was about 12 cm. For the treatment group, the ring retractor was inserted at the level of the tensor fascia, with the inferior ring resting between the fascia and the subcutaneous layer and the superior ring cinched over the skin (Figure 3). The device is made in 4 different sizes for incisions from 2.5 to 17 cm; our study population required only 1 size. Otherwise, the surgical protocol was based on that described by Matta and colleagues.⁸ Wound closure (over a drain) was performed according to a standardized protocol—running No. 1 Vicryl suture for the superficial tensor fascia, interrupted 2-0 Vicryl for the deep dermal layer, and subcutaneous 4-0 Monocryl for the skin followed by application of Dermabond (Ethicon, Somerville, New Jersey) and Tegaderm +Pad (3M, St. Paul, Minnesota) for outer dressing, which was replaced on postoperative day 2 and removed at the 2-week visit.

Statistical Methods

An a priori sample-size calculation was performed. Power performed in a base of a prior study that evaluated anterolateral and posterolateral THA scars using a VAS, a component of the MSS, suggested a sample size of 16 per group to detect the minimal clinically important difference of 1.5 cm: SD (σ) = 1.5 cm, α = 0.05, β = 0.20.^29,30 In addition, a general estimate for detecting a 1-unit change on an ordinal scale (σ = 1.0, α = 0.05, β = 0.20) resulted in the same number. We conservatively decided to enroll 25 patients per arm in case of larger true variance.

The Wilcoxon rank sum test was used for comparisons of continuous data between groups. Differences between means were analyzed with 2-sided t tests. Categorical data were compared with the Pearson χ² test or the Fisher exact test, as indicated. Ordinal ranking scores were compared with the Mantel-Haenszel test. Multivariate logistic regression was applied to identify the significant independent predictors of better scar grades for each surgeon by considering candidate variables with Ps < .20 in the univariate analysis.

Results

We found no differences in demographic or perioperative characteristics between treatment and control groups (Tables 3, 4). The groups showed similar mean improvements in their respective 6-week HHS (38.7 and 36.4 points; P = .65), SF-12 physical component summary scores (11.8 and 14.5 points; P = .37), and SF-12 mental component summary scores (5.1 and 3.7; P = .70).

Patient questionnaire outcomes are listed in Table 5. For the control group, 25/25 image sets were obtained at the 2-week visit, 25/25 at the 6-week visit, and 24/25 at the 12-week visit. For the treatment group, there were 23/25, 24/25, and 23/25 images sets, respectively.

When surgeon scoring was analyzed separately, SBSES and MSS scores were similar between treatment and control groups, with 1 exception: 2-week MSS scores were better for the treatment group according to surgeon A (P = .026). When grades were averaged, SBSES scores were again similar at all time points (Figure 4A); MSS scores were better for the treatment group at 2 weeks (P = .036) and equivalent at all other time points (Figure 4B). For the SBSES, Spearman correlation coefficient ρ with 95% confidence interval (CI) was 0.37
(95% CI, 0.08-0.66) at 2 weeks, 0.48 (95% CI, 0.20-0.76) at 6 weeks, and 0.62 (95% CI, 0.33-0.91) at 12 weeks. Following the same pattern for the MSS, ρ was 0.20 (95% CI, –0.09 to 0.49), 0.51 (95% CI, 0.23-0.79), and 0.32 (95% CI, 0.03-0.61).

Independent multivariate analysis revealed that age over 65 years was a significant predictor of worse scores. On SBSES, the odds ratio (OR) was 1.15 (95% CI, 1.07-1.24) for surgeon A and 1.11 (95% CI, 1.05-1.18) for surgeon B. On MSS, the OR was 0.89 (95% CI, 0.84-0.94) for surgeon A and 0.95 (95% CI, 0.91-0.99) for surgeon B. The likelihood of having worse SBSES scores according to surgeon A was 4.72 times higher if the pre-albumin level was under 20 mg/L (95% CI, 1.15-19.36). Albumin level under 3.5 g/dL and absolute lymphocyte count under 1500 cells/µL were not found to be independent predictors of poorer scores.

Patients’ overall opinion (P = .63) and assessment of their scars relative to expectations (P = .25) on the Likert scales were not different between groups. More scars exceeded patients’ expectations and had more excellent ratings in the control group. The 2 groups were similar with regard to relative importance of various patient-related outcomes. Factors most important to overall outcome were relief of hip pain, followed by implant longevity and length of recovery. Least important were incision-related variables.

There were only 3 minor noninfectious wound complications (6%), 2 in the treatment group and 1 in the control group. In the treatment group, a 67-year-old man with diabetes (ASA class III; BMI, 32.1 kg/m²; received transfusion) had 2 small areas (<5 mm) of superficial ulceration at 6-week follow-up—one at the proximal aspect of the incision and the other near the midpoint along the flexion crease. Both lesions resolved by 12-week follow-up. Also in the treatment group, a 77-year-old woman (ASA class II; BMI, 24.9 kg/m²; received transfusion) at 6 weeks had a spitting suture, which was removed in clinic without further issue. In the control group, a 55-year-old woman (ASA class II; BMI, 27.4 kg/m²) had a suture reaction near the proximal aspect of her incision 3 weeks after surgery. This reaction, which presented as a mild, localized erythema without pain, tenderness, or drainage, resolved by 6-week follow-up. None of these wound complications required intervention beyond observation.

Discussion

This study was designed to provide a bipartisan measure of wound-healing cosmesis after DA-THA. Scar evaluation by blinded plastic surgeons served as a standardized, clinical assessment, whereas the patient questionnaire offered a more subjective appraisal. The modified MMS²⁵ and the SBSES²⁴ are the only 2 wound-grading systems designed and validated for photographic assessment of postsurgical scars. Most scar evaluation schemes pertain to burn or traumatic scars.^26,27,31 As a result, many earlier studies intending to compare incisional scars used poorly suited evaluation systems.

The current literature includes reports on 3 studies with scoring-based scar assessment in THA; all used grading systems designed for either burns or traumatic wounds, but 2 also used a VAS.^32-34 VASs have been validated for measuring wound cosmesis but are entirely subjective and without structure and provide no feedback as to why a scar was rated good or bad.²⁴ Mow and colleagues³² prospectively compared scars after standard posterior or MIS approaches and found no differences according to a scoring system intended for burn scars. In our study population, we found no group differences in patients’ cosmesis of their scars.

Although scars can take a year or longer to fully mature, researchers from the University of Michigan discovered that scar appearance at 1 year correlates highly with cosmesis 12 weeks after closure, though poorly with cosmesis 10 days after closure.³⁵ Therefore, any observed differences in scar cosmesis between groups at 12-week follow-up would likely persist, whereas differences at 2-week follow-up would have little bearing on ultimate appearance. For this reason, our primary outcome measure was healing process and cosmesis at 12 weeks. High wound complication rates have been reported for MIS-DA-THA.^8,14-16 Jewett and Collis¹⁵ noted a 4.6% wound complication rate (3% noninfectious ulcerative dehiscence, 1.6% superficial infection), which is comparable to the 6% rate found in this study. However, there likely is some variability across studies in what constitutes a wound complication or superficial infection. Of our 3 wound complications—stitch reaction, spitting suture, small noninfectious ulceration—only the ulceration was of a severity similar to that reported by Jewett and Collis.¹⁵ Matta and colleagues⁸ reported only 3 wound complications (in 494 patients), all severe enough to require operative intervention. One explanation for this low complication rate is use of a ring retractor, as it is routinely depicted in their technique paper. However, no specific reference is made to gauge how often the device was used.

Rates of superficial infection after DA-THA range from 0.6% to 1.6% in 3 large observational studies (combined deep infection rate, 0.43%).^8,14,15 In 2 of these studies, all patients with superficial infection underwent formal débridement, though none developed deep infection. A prospective randomized study of 221 patients who underwent colorectal surgery—where perioperative infectious morbidity ranges from 25% to 50%—found that ring retractor use significantly reduced superficial wound infection rates (8.1% vs 0%). A significant reduction in wound infection was shown in a similarly designed study involving 48 patients who had open appendectomy (14.6% vs 1.6%). The device had no effect on deep infection in either general surgery study. The wound infection rates reported in these general surgery studies are markedly higher than those in our study population. As a result, the effect of the ring retractor on wound infection in DA-THA may be less. Regardless of the effect on deep infection, fewer superficial infections, which often require operative intervention, would be of considerable benefit.

Below-threshold albumin level and absolute lymphocyte count have been associated with wound-healing complications after hip replacement.²¹ In the present study, pre-albumin level under 20 mg/L was the only nutritional marker predictive of poor wound appearance, but this finding was seen only in SBSES scores from surgeon A. Subgroup analysis did not reveal any relationship between wound appearance and any of the recorded demographic or perioperative variables, but for a small predictive influence with age over 65 years.

This study had some limitations. Our findings cannot be generalized to all patients who undergo THA, as only DA incisions were studied. Results also may not be generalizable to non-fellowship-trained orthopedists. In addition, selection bias likely resulted from including patients already selected for the DA approach. Using digital images for evaluation (vs real-life evaluation) may have affected reliability as well. Last, by not incorporating texture, we omitted a potentially informative feature from scoring.

It is paramount that surgeons undergo diligent training before undertaking this approach for minimizing unwanted results; furthermore, higher early complication rates level off with increased surgeon experience.^14,36,37 We recommend meticulous soft-tissue handling, cautious retraction, and careful patient selection (relative contraindication for patients with an abdominal pannus overlying the incision) as primary measures for minimizing incisional trauma and potential wound-healing complications.³⁸ Preservation of the tensor fascia is also crucial,³⁹ as it is the only closable layer separating deep and superficial compartments. Without good closure of the tensor fascia, there is no containment or tamponade of deep bleeding, which can facilitate hematoma formation.

In the population studied, we found no significant long-term differences in cosmetic appearance (based on clinician or patient evaluation) between wounds managed with and without the ring retractor. Our data do not support routine use of the ring retractor, during DA-THA, for improved wound cosmesis. Whether the device has any significant role in reducing the number of wound complications in THA is yet to be determined. Last, the ring retractor may have a role in other areas of orthopedic surgery, such as hip fractures in the elderly or orthopedic oncology. In situations like these, where adequate nutrition and immunocompetency may be lacking, the added protection provided by the device may translate into a more notable benefit than in elective THA.

It is thought that, by placing more emphasis on soft-tissue preservation, minimally invasive surgery total hip arthroplasty (MIS-THA) results in less soft-tissue trauma, less blood loss, and earlier recovery.^1-3 Despite these improvements over standard methods, there is a concern that the vigorous retraction needed for proper visualization through smaller incisions could injure soft tissues.^4-7 Single-incision direct anterior THA (DA-THA) has gained in popularity because of the true intermuscular/internervous plane through which the procedure can be performed with relatively minimal muscle dissection using MIS techniques.^8,9 This approach may offer quicker recovery and superior stability in comparison with nonintermuscular methods, which unavoidably cause more muscle damage.^10-12

Although the evidence of these early gains is encouraging, several studies have found high complication rates with DA-THA.^8,13-17 Noted disadvantages include a steep learning curve, lateral femoral cutaneous neurapraxia, need for a specialized table, and higher fracture and wound complication rates. Not surprisingly, with increased surgeon experience, the complication rate decreased substantially.^14,15 However, wound-related complications remained steady, with 2 recent large studies reporting rates of 4.6% and 2.1%.^14,15 The thin anterior skin, high tensional forces along the groin crease and perpendicular to the typical DA incision, and less resilient soft-tissue envelope are postulated reasons for wound-related issues, which are likely magnified in patients who are more obese.^15,16

A novel device designed to lessen tissue damage is the ring retractor (Figure 1). Used initially in general surgery and obstetrics, it consists of 2 semirigid polymer rings connected by a flexible cylindrical polymer membrane.^18-20 The lower ring is tucked and anchored underneath the wound edge, and then the upper ring is rolled down and cinched onto the skin. The resultant tension on the polymer sleeve—imparted by the rigidity of the ring—provides strong, evenly distributed wound-edge retraction. It also provides a physical barrier between the wound edge and the rest of the operative field. Proponents of the ring retractor claim increased wound-edge moisture, less bruising, and reduced local trauma compared with standard metal retractors alone.

Wound-edge retractor forces are doubled during MIS-THA compared with conventional THA.^14-20 This may explain reports of worse scar cosmesis with MIS-THA. Given the theoretical benefits of minimized wound-edge trauma, the ring retractor may improve scar appearance compared with standard retraction alone. Any clinically relevant effect on cosmesis should be readily apparent to justify use of the retractor in this regard. Although some surgeons routinely use the device for primary THA, it has not been the subject of any recent orthopedic studies.

In the present study, we prospectively investigated wound cosmesis with and without use of the ring retractor in patients undergoing DA-THA.

Materials and Methods

This prospective, single-center, randomized study was reviewed and approved by the institutional review board at our facility. Consent was obtained from all participating patients.

We evaluated 50 surgical incisions in 48 patients. Eligible participants were over age 18 years and undergoing primary DA-THA. Exclusion criteria included previous surgery on the affected hip, a pathological hip condition requiring an extensile exposure, systemic inflammatory illness, chronic corticosteroid use, and dermatologic abnormality of the incisional area. One patient was having simultaneous bilateral THAs, and another was having staged bilateral THAs. Each hip in these patients was given its own case number and treated separately. Of the 49 patients who met all the inclusion criteria, only 1 decided not to participate (Figure 2).

Stratified randomization with permuted block size (sex, body mass index [BMI]) was used to assign patients in a 1:1 ratio to either the treatment group or the control group. In the treatment group, the Protractor Incision Protector and Retractor (Gyrus ACMI, Southborough, Massachusetts) was used with standard metal retractors. In the control group, only standard metal retractors were used. Patients were blinded to their group assignments, and surgeons were informed about each assignment only after the initial incision was made.

Clinical research investigators were blinded to the groups’ prospectively collected data. Collection time points were preoperative clinical visit, day of surgery through discharge, and 2-, 6-, and 12-week postoperative follow-ups. Day-of-surgery data included estimated intraoperative blood loss, operative side, operative time, intraoperative complications, and American Society of Anesthesiologists (ASA) physical status classification. Total length of stay, pain scores (range, 0-10), estimated drain output, and blood-transfusion data were also recorded. To evaluate whether the device had any effect on short-term functional outcome, we collected Harris Hip Scores (HHS) and Short Form–12 (SF-12, Version 2) scores at the preoperative and 6-week postoperative visits. We also documented any wound-healing-related issues or complications that occurred up until the final visit.

To account for any effect of nutrition status on wound healing, we obtained pre-albumin and albumin levels and absolute lymphocyte counts from the preoperative electronic records. We used an albumin level under 3.5 g/dL and an absolute lymphocyte count under 1500/µL for our analysis, as these cutoffs have been associated with wound complications after primary THA.²¹ There is no similarly established threshold for pre-albumin level, so we used values under 20 mg/L based on comparable literature.^22,23

At each postoperative visit, standardized high-resolution images were obtained. At the 12-week visit, patients completed 2 Likert scales regarding their overall opinion of their scars and how their scars compared with their expectations. They also ranked 5 separate THA-related outcomes in order of importance (Appendix).

Photographs were evaluated by 2 blinded plastic surgeons (Dr. Friedman and Dr. Jack) using 2 grading systems—the Stony Brook Scar Evaluation Scale (SBSES)²⁴ (Table 1) and a modified Manchester Scar Scale (MSS)²⁵ (Table 2). We used these systems because they were photograph-based, psychometrically studied, and specifically designed to assess surgical incision healing with established validity and reliability.^24-27 A particular advantage, strictly related to cosmetic outcome, is their validity in scoring scars from high-definition photographs in a different place or time. The SBSES, an ordinal wound evaluation scale that measures short-term cosmetic outcomes, consists of 6 items, each receiving 1 or 0 point, yielding a total score between 0 (worst) and 5 (best). The modified MSS includes a visual analog scale (VAS), which has a vertical hash marked on a 10-cm line and is scored between 0 (excellent) and 10 (poor) to 1 decimal point.^26,28 This value is added to grades on color, surface appearance, contour, and distortion, resulting in a score between 4 (best) and 24 (worst). The primary outcome measures were Likert-scale responses obtained at final visit and SBSES/MSS scores for each visit; 12-week scores were the primary end point.

Operative Procedure

Experienced fellowship-trained orthopedic surgeons performed all procedures. A modified Hueter approach was used for exposure.⁹ Mean incision length was about 12 cm. For the treatment group, the ring retractor was inserted at the level of the tensor fascia, with the inferior ring resting between the fascia and the subcutaneous layer and the superior ring cinched over the skin (Figure 3). The device is made in 4 different sizes for incisions from 2.5 to 17 cm; our study population required only 1 size. Otherwise, the surgical protocol was based on that described by Matta and colleagues.⁸ Wound closure (over a drain) was performed according to a standardized protocol—running No. 1 Vicryl suture for the superficial tensor fascia, interrupted 2-0 Vicryl for the deep dermal layer, and subcutaneous 4-0 Monocryl for the skin followed by application of Dermabond (Ethicon, Somerville, New Jersey) and Tegaderm +Pad (3M, St. Paul, Minnesota) for outer dressing, which was replaced on postoperative day 2 and removed at the 2-week visit.

Statistical Methods

An a priori sample-size calculation was performed. Power performed in a base of a prior study that evaluated anterolateral and posterolateral THA scars using a VAS, a component of the MSS, suggested a sample size of 16 per group to detect the minimal clinically important difference of 1.5 cm: SD (σ) = 1.5 cm, α = 0.05, β = 0.20.^29,30 In addition, a general estimate for detecting a 1-unit change on an ordinal scale (σ = 1.0, α = 0.05, β = 0.20) resulted in the same number. We conservatively decided to enroll 25 patients per arm in case of larger true variance.

The Wilcoxon rank sum test was used for comparisons of continuous data between groups. Differences between means were analyzed with 2-sided t tests. Categorical data were compared with the Pearson χ² test or the Fisher exact test, as indicated. Ordinal ranking scores were compared with the Mantel-Haenszel test. Multivariate logistic regression was applied to identify the significant independent predictors of better scar grades for each surgeon by considering candidate variables with Ps < .20 in the univariate analysis.

Results

We found no differences in demographic or perioperative characteristics between treatment and control groups (Tables 3, 4). The groups showed similar mean improvements in their respective 6-week HHS (38.7 and 36.4 points; P = .65), SF-12 physical component summary scores (11.8 and 14.5 points; P = .37), and SF-12 mental component summary scores (5.1 and 3.7; P = .70).

Patient questionnaire outcomes are listed in Table 5. For the control group, 25/25 image sets were obtained at the 2-week visit, 25/25 at the 6-week visit, and 24/25 at the 12-week visit. For the treatment group, there were 23/25, 24/25, and 23/25 images sets, respectively.

When surgeon scoring was analyzed separately, SBSES and MSS scores were similar between treatment and control groups, with 1 exception: 2-week MSS scores were better for the treatment group according to surgeon A (P = .026). When grades were averaged, SBSES scores were again similar at all time points (Figure 4A); MSS scores were better for the treatment group at 2 weeks (P = .036) and equivalent at all other time points (Figure 4B). For the SBSES, Spearman correlation coefficient ρ with 95% confidence interval (CI) was 0.37
(95% CI, 0.08-0.66) at 2 weeks, 0.48 (95% CI, 0.20-0.76) at 6 weeks, and 0.62 (95% CI, 0.33-0.91) at 12 weeks. Following the same pattern for the MSS, ρ was 0.20 (95% CI, –0.09 to 0.49), 0.51 (95% CI, 0.23-0.79), and 0.32 (95% CI, 0.03-0.61).

Independent multivariate analysis revealed that age over 65 years was a significant predictor of worse scores. On SBSES, the odds ratio (OR) was 1.15 (95% CI, 1.07-1.24) for surgeon A and 1.11 (95% CI, 1.05-1.18) for surgeon B. On MSS, the OR was 0.89 (95% CI, 0.84-0.94) for surgeon A and 0.95 (95% CI, 0.91-0.99) for surgeon B. The likelihood of having worse SBSES scores according to surgeon A was 4.72 times higher if the pre-albumin level was under 20 mg/L (95% CI, 1.15-19.36). Albumin level under 3.5 g/dL and absolute lymphocyte count under 1500 cells/µL were not found to be independent predictors of poorer scores.

Patients’ overall opinion (P = .63) and assessment of their scars relative to expectations (P = .25) on the Likert scales were not different between groups. More scars exceeded patients’ expectations and had more excellent ratings in the control group. The 2 groups were similar with regard to relative importance of various patient-related outcomes. Factors most important to overall outcome were relief of hip pain, followed by implant longevity and length of recovery. Least important were incision-related variables.

There were only 3 minor noninfectious wound complications (6%), 2 in the treatment group and 1 in the control group. In the treatment group, a 67-year-old man with diabetes (ASA class III; BMI, 32.1 kg/m²; received transfusion) had 2 small areas (<5 mm) of superficial ulceration at 6-week follow-up—one at the proximal aspect of the incision and the other near the midpoint along the flexion crease. Both lesions resolved by 12-week follow-up. Also in the treatment group, a 77-year-old woman (ASA class II; BMI, 24.9 kg/m²; received transfusion) at 6 weeks had a spitting suture, which was removed in clinic without further issue. In the control group, a 55-year-old woman (ASA class II; BMI, 27.4 kg/m²) had a suture reaction near the proximal aspect of her incision 3 weeks after surgery. This reaction, which presented as a mild, localized erythema without pain, tenderness, or drainage, resolved by 6-week follow-up. None of these wound complications required intervention beyond observation.

Discussion

This study was designed to provide a bipartisan measure of wound-healing cosmesis after DA-THA. Scar evaluation by blinded plastic surgeons served as a standardized, clinical assessment, whereas the patient questionnaire offered a more subjective appraisal. The modified MMS²⁵ and the SBSES²⁴ are the only 2 wound-grading systems designed and validated for photographic assessment of postsurgical scars. Most scar evaluation schemes pertain to burn or traumatic scars.^26,27,31 As a result, many earlier studies intending to compare incisional scars used poorly suited evaluation systems.

The current literature includes reports on 3 studies with scoring-based scar assessment in THA; all used grading systems designed for either burns or traumatic wounds, but 2 also used a VAS.^32-34 VASs have been validated for measuring wound cosmesis but are entirely subjective and without structure and provide no feedback as to why a scar was rated good or bad.²⁴ Mow and colleagues³² prospectively compared scars after standard posterior or MIS approaches and found no differences according to a scoring system intended for burn scars. In our study population, we found no group differences in patients’ cosmesis of their scars.

Although scars can take a year or longer to fully mature, researchers from the University of Michigan discovered that scar appearance at 1 year correlates highly with cosmesis 12 weeks after closure, though poorly with cosmesis 10 days after closure.³⁵ Therefore, any observed differences in scar cosmesis between groups at 12-week follow-up would likely persist, whereas differences at 2-week follow-up would have little bearing on ultimate appearance. For this reason, our primary outcome measure was healing process and cosmesis at 12 weeks. High wound complication rates have been reported for MIS-DA-THA.^8,14-16 Jewett and Collis¹⁵ noted a 4.6% wound complication rate (3% noninfectious ulcerative dehiscence, 1.6% superficial infection), which is comparable to the 6% rate found in this study. However, there likely is some variability across studies in what constitutes a wound complication or superficial infection. Of our 3 wound complications—stitch reaction, spitting suture, small noninfectious ulceration—only the ulceration was of a severity similar to that reported by Jewett and Collis.¹⁵ Matta and colleagues⁸ reported only 3 wound complications (in 494 patients), all severe enough to require operative intervention. One explanation for this low complication rate is use of a ring retractor, as it is routinely depicted in their technique paper. However, no specific reference is made to gauge how often the device was used.

Rates of superficial infection after DA-THA range from 0.6% to 1.6% in 3 large observational studies (combined deep infection rate, 0.43%).^8,14,15 In 2 of these studies, all patients with superficial infection underwent formal débridement, though none developed deep infection. A prospective randomized study of 221 patients who underwent colorectal surgery—where perioperative infectious morbidity ranges from 25% to 50%—found that ring retractor use significantly reduced superficial wound infection rates (8.1% vs 0%). A significant reduction in wound infection was shown in a similarly designed study involving 48 patients who had open appendectomy (14.6% vs 1.6%). The device had no effect on deep infection in either general surgery study. The wound infection rates reported in these general surgery studies are markedly higher than those in our study population. As a result, the effect of the ring retractor on wound infection in DA-THA may be less. Regardless of the effect on deep infection, fewer superficial infections, which often require operative intervention, would be of considerable benefit.

Below-threshold albumin level and absolute lymphocyte count have been associated with wound-healing complications after hip replacement.²¹ In the present study, pre-albumin level under 20 mg/L was the only nutritional marker predictive of poor wound appearance, but this finding was seen only in SBSES scores from surgeon A. Subgroup analysis did not reveal any relationship between wound appearance and any of the recorded demographic or perioperative variables, but for a small predictive influence with age over 65 years.

This study had some limitations. Our findings cannot be generalized to all patients who undergo THA, as only DA incisions were studied. Results also may not be generalizable to non-fellowship-trained orthopedists. In addition, selection bias likely resulted from including patients already selected for the DA approach. Using digital images for evaluation (vs real-life evaluation) may have affected reliability as well. Last, by not incorporating texture, we omitted a potentially informative feature from scoring.

It is paramount that surgeons undergo diligent training before undertaking this approach for minimizing unwanted results; furthermore, higher early complication rates level off with increased surgeon experience.^14,36,37 We recommend meticulous soft-tissue handling, cautious retraction, and careful patient selection (relative contraindication for patients with an abdominal pannus overlying the incision) as primary measures for minimizing incisional trauma and potential wound-healing complications.³⁸ Preservation of the tensor fascia is also crucial,³⁹ as it is the only closable layer separating deep and superficial compartments. Without good closure of the tensor fascia, there is no containment or tamponade of deep bleeding, which can facilitate hematoma formation.

In the population studied, we found no significant long-term differences in cosmetic appearance (based on clinician or patient evaluation) between wounds managed with and without the ring retractor. Our data do not support routine use of the ring retractor, during DA-THA, for improved wound cosmesis. Whether the device has any significant role in reducing the number of wound complications in THA is yet to be determined. Last, the ring retractor may have a role in other areas of orthopedic surgery, such as hip fractures in the elderly or orthopedic oncology. In situations like these, where adequate nutrition and immunocompetency may be lacking, the added protection provided by the device may translate into a more notable benefit than in elective THA.

More than 13 million units of blood are transfused each year. Although transfusion can certainly be lifesaving, numerous studies over the past 20 years have shown significant, dose-dependent increases in morbidity, mortality, and cost with each unit of packed red blood cells (pRBCs) transfused.¹ Transfusion is one of the most common interventions in the critically ill population; however the negative effects of transfusion-related infection are well documented in the recent literature.^1-7 There is no question that transfusion of blood products can be lifesaving to acutely ill trauma patients, but there is little evidence regarding when transfusions are indicated in asymptomatic anemic patients who are no longer in need of acute resuscitation.

Several studies have analyzed healthy individuals with an isovolemic reduction in hemoglobin (Hgb) level to 5.0 g/dL.^8,9 They have found no significant compromise in oxygen delivery to the tissues. Currently, there is a lack of clinical data to suggest adequate RBC transfusion endpoints in trauma surgery.¹⁰ Given the lack of evidence to support transfusion triggers for young, healthy, asymptomatic orthopedic trauma patients, we decided to investigate whether a more conservative transfusion strategy might be as safe as a more liberal strategy.

Materials and Methods

After obtaining approval from our institutional review board, we performed a retrospective observational cohort analysis of patients treated at a level I trauma center between September 2006 and February 2009. The trauma registry included all patients who underwent surgery performed by a single orthopedic fellowship–trained trauma surgeon. All patients who had a recorded Hgb level of 9.0 g/dL or less at any time during their admission were included; they were considered no longer volume-depleted after initial resuscitation. Exclusion criteria were age under 18 years or over 50 years; pregnancy; head injury; and preexisting heart, pulmonary, or renal disease.

Initially, 963 patients were identified as orthopedic trauma patients treated by Dr. Mullis within the defined period. After inclusion and exclusion criteria were used to limit this database, the charts of the 109 patients who met the above criteria were reviewed. By chart review or telephone follow-up, 104 patients with 1-year follow-up were identified, and their cases became the basis for our analysis. Demographic information, length of hospital stay, surgeries performed, number of pRBC units transfused, Hgb level prompting transfusion, lowest recorded Hgb level, complications, and Injury Severity Score (ISS) were recorded for each patient. Seventy-two patients (69%) were male, 32 (31%) female. Mean age of the study population was 33 years.

Patients were divided into 2 groups by lowest Hgb level before first transfusion—under 7.0 g/dL and 7.0 g/dL or higher—and then by whether they had been transfused. General guidelines for erythrocyte transfusion on the orthopedic trauma service included patients who were symptomatic at rest (headache, dizziness, or shortness of breath) and asymptomatic patients with Hgb levels under 5.0 g/dL. For patients with varying (lesser) degrees of anemia, transfusion typically depended on clinical symptoms and overall decrease in Hgb level from that recorded on admission.

Patient charts were reviewed for complications extending through a 1-year period after initial discharge from the inpatient service. Patients who had not received follow-up treatment through a known outpatient clinic were contacted by telephone to ascertain outcome. Overall, 5 of the 109 patients were lost at 1-year follow-up, leaving 104 patients with 1-year follow-up (95%). Primary outcome of the study was postoperative complications. Superficial wound infection was defined as cellulitis near the surgical site within 1 year, requiring oral antibiotics; deep wound infection was defined as any related infection within 1 year of injury, requiring intravenous antibiotics or surgical débridement in the operating room. The review for complications included superficial infection, deep infection, urinary tract infection, pneumonia, pulmonary embolism, deep venous thrombosis, acute renal failure or insufficiency, nonunion, delayed union, compartment syndrome, osteomyelitis, nerve palsy, anoxic brain injury, cardiac ischemia or infarct, pancreatitis, and death.

Statistical Methods

The primary focus of this analysis was to determine if patients’ risk of complication at 1-year follow-up was affected by anemia—lowest recorded Hgb level before first transfusion for transfused patients, or lowest Hgb level during hospital stay for nontransfused patients—or whether transfusion itself might be a risk factor for complication. Multiple logistic regression models were used to determine the likelihood each group would have a complication. The dependent variable was complication rate; the explanatory variables included whether the patient was transfused, anemia/Hgb level (under 7 g/dL vs 7 g/dL or higher), and the 2-way interaction. Other possible explanatory variables entered into the model were age, sex, ISS, and whether the patient had had multiple surgeries. As the sample size was small, these variables were entered into the regression model one at a time. Results are presented as odds ratios (ORs) with corresponding 95% confidence intervals (CIs) and P values. The analysis was performed with SAS Version 9.1 (SAS Institute, Cary, North Carolina). Tests were considered statistically significant with P < .05 and marginally significant with P < .10. OR above 1 indicated that the odds of a complication occurring were higher in the exposed group (transfused patients) than in the unexposed group (nontransfused patients).

Results

The charts of 104 patients were reviewed and included in this analysis. Sixty-two patients (60%) had received a transfusion; 42 (40%) had not. Before first transfusion, 21 (34%) of the 62 transfused patients had Hgb levels under 7.0 g/dL, and the other 41 (66%) had Hgb levels of 7.0 g/dL or higher. Of the 42 nontransfused patients, 8 (19%) had lowest Hgb levels under 7.0 g/dL, and the other 34 (81%) had Hgb levels of 7.0 g/dL or higher (Table 1).

The transfused patients, considering all levels of anemia, had a mean ISS of 16.1 (range, 1-45), a mean of 2.0 operations (range, 1-6), a mean hospital stay of 18 days (range, 1-73 days), and a mean age of 34 years (range, 18-50 years). The nontransfused patients, considering all levels of anemia, had a mean ISS of 14.1 (range, 4-43), a mean of 1.4 operations (range, 1-5), a mean hospital stay of 10 days (range, 1-42 days), and a mean age of 33 years (range, 18-50 years). In the transfusion group, the mean number of transfused pRBC units was 6.9 (range, 1-31), or 7.8 units for patients with Hgb levels under 7 g/dL and 6.4 units for patients with Hgb levels of 7 g/dL or higher. At 1-year follow-up, complications were observed in 41 (66%) of the 62 transfused patients and in 17 (40%) of the 42 nontransfused patients (Table 1). The different types of complications seen in each group are listed in Table 2.

Statistical Analysis

Patients were divided into 2 groups by Hgb level—under 7.0 g/dL and 7.0 g/dL or higher—and then by whether they received pRBC transfusion. In addition, which patients had a complication over a 1-year period were identified.

For each group, we calculated sample size, number of complications, complication rate, and 95% CI for proportions. For transfused patients with Hgb level of 7.0 g/dL or higher, the complication rate was 71% (29/41). For nontransfused patients with Hgb of 7.0 g/dL or higher, the complication rate was 41% (14/34). Similarly, for transfused patients with Hgb under 7.0 g/dL, the complication rate was 57% (12/21). Last, for nontransfused patients with Hgb under 7.0 g/dL, the complication rate was 38% (3/8) (Table 3).

Transfused patients had a significantly higher risk of complication (OR, 3.1; 95% CI, 1.4-7.1; P < .01). Severity of anemia was not found to be independently associated with increased risk of complication (OR, 0.6; 95% CI, 0.3-1.6; P = .33) (Table 4). The interaction term was removed and eliminated from further analysis, as it was not found to be significant (P = .45).

Furthermore, the possibility of confounding variables (eg, age, sex, ISS, number of surgeries performed) was considered by including them in the model one at a time. From these logistic regression models, which included whether patients were transfused and level of anemia, an increased risk of complication (OR, 1.8; 95% CI, 1.1-2.9; P = .02) was found for each additional surgery, while receiving transfusion remained statistically significant (OR, 2.5; 95% CI, 1.0-5.8; P < .04). Age, sex, and ISS were not shown to be significantly associated with an increased complication rate (Ps = .71, .32, and .13, respectively).

We performed a subanalysis of the transfused patients to determine the impact of number of units transfused on complication rate. Each additional unit of pRBCs transfused increased the risk of complication, indicating a dose-dependent response (OR, 1.3; 95% CI, 1.04-1.51; P = .02).

Discussion

Transfusion is a generally accepted and common intervention both in the intensive care unit and in the perioperative period. However, there is little evidence to support routine transfusion of asymptomatic orthopedic trauma patients who are no longer within the initial resuscitative period after trauma. Nevertheless, the practice is routinely done based on expert opinion (level 5 evidence). The anemia protocol for our orthopedic trauma service routinely allowed the Hgb levels of asymptomatic healthy patients to drop to under 7.0 g/dL without transfusion; when other services were consulted or were primary, however, these asymptomatic patients were still routinely transfused based on practitioners’ practice patterns and anecdotal experiences.

In hemodynamically unstable patients, there is no acceptable substitute for blood transfusion. Blood replacement remains essential in the case of acute hemorrhage. However, the complications associated with transfusion should lead us to avoid, or at least minimize, unnecessary transfusion in young asymptomatic patients who are not actively bleeding in the postresuscitative period. In our study, we did not seek causation of increased complications with transfusion but assessed whether the risk of anemia outweighed the risk of transfusion in young, healthy, asymptomatic trauma patients who were no longer in the initial resuscitation period.

Our study was designed to evaluate a conservative transfusion strategy used in orthopedic trauma patients. We hypothesized that the risk of anemia in these asymptomatic patients would be lower than the risk of transfusing asymptomatic patients in the perioperative period. In addition, we thought the level of anemia would play a less significant role in the postoperative complication rate relative to transfusion itself. Our results suggest that a more conservative transfusion strategy of allowing asymptomatic patients to become and remain anemic even to a Hgb level of 5 g/dL may be as safe as a more liberal transfusion strategy of keeping patients at a Hgb level higher than 7 g/dL. In general, the complication rate was 66% for transfused patients and 40% for nontransfused patients. These results remain significant after correcting for possible confounding factors, including age, sex, ISS, and number of surgeries.

The results of this study do not suggest that there may not be complications associated with anemia; a 40% complication rate even in the nontransfused group is high. One might expect that patients who had isolated injuries and never developed anemia with an Hgb level under 9 g/dL might have an even lower complication rate. In the group used for inclusion in this study, however, there was not a significant increased risk for patients who tolerated a lower anemia (Hgb, <7 g/dL), whereas transfusion to keep the Hgb level above 7 g/dL appeared to correlate with a significant risk of complication and appeared to be dose-dependent. It should be noted that the complications in both the transfusion and anemia groups are not necessarily related to transfusion or anemia, as many factors in a retrospective study cannot be controlled. These findings simply suggest that it might be as safe to use a conservative transfusion strategy as a liberal transfusion strategy in this patient population.

Although our study is retrospective, prospective randomized studies in the elderly and in the critical care population have shown conservative transfusion guidelines are at least as safe as liberal transfusion strategies.^2,11 One study randomized intensive care unit patients with Hgb levels under 9.0 g/dL to 2 groups, one with liberal and the other with restrictive protocols for pRBC unit transfusion.² The liberal group maintained Hgb levels between 10.0 and 12.0 g/dL, and the restricted group kept Hgb levels between 7.0 and 9.0 g/dL. Thirty-day mortality was significantly lower in less acutely ill patients and younger patients (<55 years old) in the restrictive group than in the liberal group. It was concluded that a restrictive strategy of RBC transfusion is at least as effective as, and possibly superior to, a liberal transfusion strategy in the critically ill when considering short- and long-term outcomes. Another prospective study randomized elderly patients (N = 2016) with hip fractures and cardiovascular risk factors to a liberal transfusion strategy (if Hgb level fell under 10 g/dL) or a restrictive transfusion strategy (if Hgb level fell under 8 g/dL). The study found no difference between the 2 groups.¹¹

The deleterious effect of allogeneic blood transfusion on the immune system is complex and has been linked to the down-regulation of cellular immunity, including decreased function of natural killer cells, decreased function of macrophages and monocytes, and increased numbers of suppressor T cells.^12,13 This minimized immune response has been associated with a multitude of infectious morbidities in various patient populations.⁷ A meta-analysis of 20 studies reviewing outcomes of the effects of transfusion on postoperative bacterial infection found strong evidence supporting a correlation.⁵ Their analysis found an OR of 5.3 (range, 5.0-5.4) for infectious complication after allogeneic transfusion in the trauma population, and an OR of 3.5 (range, 1.4-15.2) considering all patient populations.

Similar results showing increased risk of infectious morbidities associated with transfusion were found in other studies involving the critically ill, patients after hip arthroplasty, and cardiothoracic surgery and general trauma populations.^1,3,4,14,15 Furthermore, these results were seen in a dose-dependent response leading to increased incidence of complication with each unit of blood transfused.

Our study did not focus only on infection but included other complications (eg, cardiac, renal, and brain ischemia) that might be associated with anemia or transfusion. It is intuitive that anemia can cause ischemic events but less intuitive that allogeneic transfusion can also cause ischemic events because of the poor deformability of the cells due to storage, which can lead to “sludging” in capillaries throughout the body.¹⁶ This has been shown to be important in animal models, but it is unclear what poses more risk in humans—anemia without transfusion or the initial insult from transfusion, before the body clears the “waste” from stored cells and the remaining viable cells gain oxygen-carrying capacity.

Our study has several limitations. The number of patients who had severe anemia (Hgb level, <7 g/dL) and were not transfused is relatively small compared with the numbers in the other groups used for comparison. Because our study was retrospective, we could only find associations and not prove causation. This is significant, as the higher complication rate seen with transfusions may only be caused by the transfusion as a predictor of a patient requiring more complex surgery with higher blood loss (and higher risk of complication) or other such risk factors that led to transfusion, but not the transfusion itself causing the complication. An attempt was made to remove this potential bias by controlling for age, sex, ISS, and whether the patient had multiple surgeries. However, there may have been other significant confounding variables not excluded. As complications were assessed by chart review, they may not include those that occurred at other institutions and that were never reported to the practitioners at our facility (though we did have the ability to search records of neighboring institutions electronically when electronic medical records were available). That no functional outcomes were included in this retrospective review might make the complication rate appear more or less sensitive than the patients’ own opinions regarding their outcomes. All these weaknesses could call into question whether the statistically significant higher risk associated with allogeneic transfusion found in this study is real, but the focus and reason for pursuing this study were to determine if permissive anemia was dangerous or would be associated with a higher risk of complications than routine allogeneic transfusion of asymptomatic patients to treat a laboratory value.

Strengths of the study include the review of a single surgeon’s practice with a written protocol in place for anemic orthopedic trauma patients. The 95% follow-up (104/109 patients) is good for this type of trauma population. Although this series is retrospective, it is reasonably large for a subgroup of young, healthy orthopedic trauma patients to study the effects of anemia or transfusion. Whether transfused or not, many of these patients tolerated Hgb levels under 7 g/dL, which gave a large enough comparison group to evaluate the independent effects of transfusion (or of using transfusion as a marker for complication risk) or anemia as a risk factor. As a result, it appears that a more conservative transfusion strategy may be as safe as a more liberal transfusion strategy. The results of this retrospective study were used to design a prospective multidisciplinary pilot study randomizing patients to either a liberal or a conservative transfusion strategy to determine which approach might carry higher risks of complications.

Conclusion

The results of this retrospective study suggest that a conservative transfusion strategy in a young, healthy, euvolemic asymptomatic patient who is not actively bleeding may be as safe as a liberal transfusion strategy and potentially may have fewer complications than does transfusion for a conventional laboratory value. Our study results do not suggest that transfusions should be held in patients who are symptomatic at rest or in patients who are being actively resuscitated, as transfusion can be lifesaving under these circumstances. A prospective randomized study has begun at our institution with enrollment expected to take 2 years with another year needed to complete 1-year follow-up of all patients.

More than 13 million units of blood are transfused each year. Although transfusion can certainly be lifesaving, numerous studies over the past 20 years have shown significant, dose-dependent increases in morbidity, mortality, and cost with each unit of packed red blood cells (pRBCs) transfused.¹ Transfusion is one of the most common interventions in the critically ill population; however the negative effects of transfusion-related infection are well documented in the recent literature.^1-7 There is no question that transfusion of blood products can be lifesaving to acutely ill trauma patients, but there is little evidence regarding when transfusions are indicated in asymptomatic anemic patients who are no longer in need of acute resuscitation.

Several studies have analyzed healthy individuals with an isovolemic reduction in hemoglobin (Hgb) level to 5.0 g/dL.^8,9 They have found no significant compromise in oxygen delivery to the tissues. Currently, there is a lack of clinical data to suggest adequate RBC transfusion endpoints in trauma surgery.¹⁰ Given the lack of evidence to support transfusion triggers for young, healthy, asymptomatic orthopedic trauma patients, we decided to investigate whether a more conservative transfusion strategy might be as safe as a more liberal strategy.

Materials and Methods

After obtaining approval from our institutional review board, we performed a retrospective observational cohort analysis of patients treated at a level I trauma center between September 2006 and February 2009. The trauma registry included all patients who underwent surgery performed by a single orthopedic fellowship–trained trauma surgeon. All patients who had a recorded Hgb level of 9.0 g/dL or less at any time during their admission were included; they were considered no longer volume-depleted after initial resuscitation. Exclusion criteria were age under 18 years or over 50 years; pregnancy; head injury; and preexisting heart, pulmonary, or renal disease.

Initially, 963 patients were identified as orthopedic trauma patients treated by Dr. Mullis within the defined period. After inclusion and exclusion criteria were used to limit this database, the charts of the 109 patients who met the above criteria were reviewed. By chart review or telephone follow-up, 104 patients with 1-year follow-up were identified, and their cases became the basis for our analysis. Demographic information, length of hospital stay, surgeries performed, number of pRBC units transfused, Hgb level prompting transfusion, lowest recorded Hgb level, complications, and Injury Severity Score (ISS) were recorded for each patient. Seventy-two patients (69%) were male, 32 (31%) female. Mean age of the study population was 33 years.

Patients were divided into 2 groups by lowest Hgb level before first transfusion—under 7.0 g/dL and 7.0 g/dL or higher—and then by whether they had been transfused. General guidelines for erythrocyte transfusion on the orthopedic trauma service included patients who were symptomatic at rest (headache, dizziness, or shortness of breath) and asymptomatic patients with Hgb levels under 5.0 g/dL. For patients with varying (lesser) degrees of anemia, transfusion typically depended on clinical symptoms and overall decrease in Hgb level from that recorded on admission.

Patient charts were reviewed for complications extending through a 1-year period after initial discharge from the inpatient service. Patients who had not received follow-up treatment through a known outpatient clinic were contacted by telephone to ascertain outcome. Overall, 5 of the 109 patients were lost at 1-year follow-up, leaving 104 patients with 1-year follow-up (95%). Primary outcome of the study was postoperative complications. Superficial wound infection was defined as cellulitis near the surgical site within 1 year, requiring oral antibiotics; deep wound infection was defined as any related infection within 1 year of injury, requiring intravenous antibiotics or surgical débridement in the operating room. The review for complications included superficial infection, deep infection, urinary tract infection, pneumonia, pulmonary embolism, deep venous thrombosis, acute renal failure or insufficiency, nonunion, delayed union, compartment syndrome, osteomyelitis, nerve palsy, anoxic brain injury, cardiac ischemia or infarct, pancreatitis, and death.

Statistical Methods

The primary focus of this analysis was to determine if patients’ risk of complication at 1-year follow-up was affected by anemia—lowest recorded Hgb level before first transfusion for transfused patients, or lowest Hgb level during hospital stay for nontransfused patients—or whether transfusion itself might be a risk factor for complication. Multiple logistic regression models were used to determine the likelihood each group would have a complication. The dependent variable was complication rate; the explanatory variables included whether the patient was transfused, anemia/Hgb level (under 7 g/dL vs 7 g/dL or higher), and the 2-way interaction. Other possible explanatory variables entered into the model were age, sex, ISS, and whether the patient had had multiple surgeries. As the sample size was small, these variables were entered into the regression model one at a time. Results are presented as odds ratios (ORs) with corresponding 95% confidence intervals (CIs) and P values. The analysis was performed with SAS Version 9.1 (SAS Institute, Cary, North Carolina). Tests were considered statistically significant with P < .05 and marginally significant with P < .10. OR above 1 indicated that the odds of a complication occurring were higher in the exposed group (transfused patients) than in the unexposed group (nontransfused patients).

Results

The charts of 104 patients were reviewed and included in this analysis. Sixty-two patients (60%) had received a transfusion; 42 (40%) had not. Before first transfusion, 21 (34%) of the 62 transfused patients had Hgb levels under 7.0 g/dL, and the other 41 (66%) had Hgb levels of 7.0 g/dL or higher. Of the 42 nontransfused patients, 8 (19%) had lowest Hgb levels under 7.0 g/dL, and the other 34 (81%) had Hgb levels of 7.0 g/dL or higher (Table 1).

The transfused patients, considering all levels of anemia, had a mean ISS of 16.1 (range, 1-45), a mean of 2.0 operations (range, 1-6), a mean hospital stay of 18 days (range, 1-73 days), and a mean age of 34 years (range, 18-50 years). The nontransfused patients, considering all levels of anemia, had a mean ISS of 14.1 (range, 4-43), a mean of 1.4 operations (range, 1-5), a mean hospital stay of 10 days (range, 1-42 days), and a mean age of 33 years (range, 18-50 years). In the transfusion group, the mean number of transfused pRBC units was 6.9 (range, 1-31), or 7.8 units for patients with Hgb levels under 7 g/dL and 6.4 units for patients with Hgb levels of 7 g/dL or higher. At 1-year follow-up, complications were observed in 41 (66%) of the 62 transfused patients and in 17 (40%) of the 42 nontransfused patients (Table 1). The different types of complications seen in each group are listed in Table 2.

Statistical Analysis

Patients were divided into 2 groups by Hgb level—under 7.0 g/dL and 7.0 g/dL or higher—and then by whether they received pRBC transfusion. In addition, which patients had a complication over a 1-year period were identified.

For each group, we calculated sample size, number of complications, complication rate, and 95% CI for proportions. For transfused patients with Hgb level of 7.0 g/dL or higher, the complication rate was 71% (29/41). For nontransfused patients with Hgb of 7.0 g/dL or higher, the complication rate was 41% (14/34). Similarly, for transfused patients with Hgb under 7.0 g/dL, the complication rate was 57% (12/21). Last, for nontransfused patients with Hgb under 7.0 g/dL, the complication rate was 38% (3/8) (Table 3).

Transfused patients had a significantly higher risk of complication (OR, 3.1; 95% CI, 1.4-7.1; P < .01). Severity of anemia was not found to be independently associated with increased risk of complication (OR, 0.6; 95% CI, 0.3-1.6; P = .33) (Table 4). The interaction term was removed and eliminated from further analysis, as it was not found to be significant (P = .45).

Furthermore, the possibility of confounding variables (eg, age, sex, ISS, number of surgeries performed) was considered by including them in the model one at a time. From these logistic regression models, which included whether patients were transfused and level of anemia, an increased risk of complication (OR, 1.8; 95% CI, 1.1-2.9; P = .02) was found for each additional surgery, while receiving transfusion remained statistically significant (OR, 2.5; 95% CI, 1.0-5.8; P < .04). Age, sex, and ISS were not shown to be significantly associated with an increased complication rate (Ps = .71, .32, and .13, respectively).

We performed a subanalysis of the transfused patients to determine the impact of number of units transfused on complication rate. Each additional unit of pRBCs transfused increased the risk of complication, indicating a dose-dependent response (OR, 1.3; 95% CI, 1.04-1.51; P = .02).

Discussion

Transfusion is a generally accepted and common intervention both in the intensive care unit and in the perioperative period. However, there is little evidence to support routine transfusion of asymptomatic orthopedic trauma patients who are no longer within the initial resuscitative period after trauma. Nevertheless, the practice is routinely done based on expert opinion (level 5 evidence). The anemia protocol for our orthopedic trauma service routinely allowed the Hgb levels of asymptomatic healthy patients to drop to under 7.0 g/dL without transfusion; when other services were consulted or were primary, however, these asymptomatic patients were still routinely transfused based on practitioners’ practice patterns and anecdotal experiences.

In hemodynamically unstable patients, there is no acceptable substitute for blood transfusion. Blood replacement remains essential in the case of acute hemorrhage. However, the complications associated with transfusion should lead us to avoid, or at least minimize, unnecessary transfusion in young asymptomatic patients who are not actively bleeding in the postresuscitative period. In our study, we did not seek causation of increased complications with transfusion but assessed whether the risk of anemia outweighed the risk of transfusion in young, healthy, asymptomatic trauma patients who were no longer in the initial resuscitation period.

Our study was designed to evaluate a conservative transfusion strategy used in orthopedic trauma patients. We hypothesized that the risk of anemia in these asymptomatic patients would be lower than the risk of transfusing asymptomatic patients in the perioperative period. In addition, we thought the level of anemia would play a less significant role in the postoperative complication rate relative to transfusion itself. Our results suggest that a more conservative transfusion strategy of allowing asymptomatic patients to become and remain anemic even to a Hgb level of 5 g/dL may be as safe as a more liberal transfusion strategy of keeping patients at a Hgb level higher than 7 g/dL. In general, the complication rate was 66% for transfused patients and 40% for nontransfused patients. These results remain significant after correcting for possible confounding factors, including age, sex, ISS, and number of surgeries.

The results of this study do not suggest that there may not be complications associated with anemia; a 40% complication rate even in the nontransfused group is high. One might expect that patients who had isolated injuries and never developed anemia with an Hgb level under 9 g/dL might have an even lower complication rate. In the group used for inclusion in this study, however, there was not a significant increased risk for patients who tolerated a lower anemia (Hgb, <7 g/dL), whereas transfusion to keep the Hgb level above 7 g/dL appeared to correlate with a significant risk of complication and appeared to be dose-dependent. It should be noted that the complications in both the transfusion and anemia groups are not necessarily related to transfusion or anemia, as many factors in a retrospective study cannot be controlled. These findings simply suggest that it might be as safe to use a conservative transfusion strategy as a liberal transfusion strategy in this patient population.

Although our study is retrospective, prospective randomized studies in the elderly and in the critical care population have shown conservative transfusion guidelines are at least as safe as liberal transfusion strategies.^2,11 One study randomized intensive care unit patients with Hgb levels under 9.0 g/dL to 2 groups, one with liberal and the other with restrictive protocols for pRBC unit transfusion.² The liberal group maintained Hgb levels between 10.0 and 12.0 g/dL, and the restricted group kept Hgb levels between 7.0 and 9.0 g/dL. Thirty-day mortality was significantly lower in less acutely ill patients and younger patients (<55 years old) in the restrictive group than in the liberal group. It was concluded that a restrictive strategy of RBC transfusion is at least as effective as, and possibly superior to, a liberal transfusion strategy in the critically ill when considering short- and long-term outcomes. Another prospective study randomized elderly patients (N = 2016) with hip fractures and cardiovascular risk factors to a liberal transfusion strategy (if Hgb level fell under 10 g/dL) or a restrictive transfusion strategy (if Hgb level fell under 8 g/dL). The study found no difference between the 2 groups.¹¹

The deleterious effect of allogeneic blood transfusion on the immune system is complex and has been linked to the down-regulation of cellular immunity, including decreased function of natural killer cells, decreased function of macrophages and monocytes, and increased numbers of suppressor T cells.^12,13 This minimized immune response has been associated with a multitude of infectious morbidities in various patient populations.⁷ A meta-analysis of 20 studies reviewing outcomes of the effects of transfusion on postoperative bacterial infection found strong evidence supporting a correlation.⁵ Their analysis found an OR of 5.3 (range, 5.0-5.4) for infectious complication after allogeneic transfusion in the trauma population, and an OR of 3.5 (range, 1.4-15.2) considering all patient populations.

Similar results showing increased risk of infectious morbidities associated with transfusion were found in other studies involving the critically ill, patients after hip arthroplasty, and cardiothoracic surgery and general trauma populations.^1,3,4,14,15 Furthermore, these results were seen in a dose-dependent response leading to increased incidence of complication with each unit of blood transfused.

Our study did not focus only on infection but included other complications (eg, cardiac, renal, and brain ischemia) that might be associated with anemia or transfusion. It is intuitive that anemia can cause ischemic events but less intuitive that allogeneic transfusion can also cause ischemic events because of the poor deformability of the cells due to storage, which can lead to “sludging” in capillaries throughout the body.¹⁶ This has been shown to be important in animal models, but it is unclear what poses more risk in humans—anemia without transfusion or the initial insult from transfusion, before the body clears the “waste” from stored cells and the remaining viable cells gain oxygen-carrying capacity.

Our study has several limitations. The number of patients who had severe anemia (Hgb level, <7 g/dL) and were not transfused is relatively small compared with the numbers in the other groups used for comparison. Because our study was retrospective, we could only find associations and not prove causation. This is significant, as the higher complication rate seen with transfusions may only be caused by the transfusion as a predictor of a patient requiring more complex surgery with higher blood loss (and higher risk of complication) or other such risk factors that led to transfusion, but not the transfusion itself causing the complication. An attempt was made to remove this potential bias by controlling for age, sex, ISS, and whether the patient had multiple surgeries. However, there may have been other significant confounding variables not excluded. As complications were assessed by chart review, they may not include those that occurred at other institutions and that were never reported to the practitioners at our facility (though we did have the ability to search records of neighboring institutions electronically when electronic medical records were available). That no functional outcomes were included in this retrospective review might make the complication rate appear more or less sensitive than the patients’ own opinions regarding their outcomes. All these weaknesses could call into question whether the statistically significant higher risk associated with allogeneic transfusion found in this study is real, but the focus and reason for pursuing this study were to determine if permissive anemia was dangerous or would be associated with a higher risk of complications than routine allogeneic transfusion of asymptomatic patients to treat a laboratory value.

Strengths of the study include the review of a single surgeon’s practice with a written protocol in place for anemic orthopedic trauma patients. The 95% follow-up (104/109 patients) is good for this type of trauma population. Although this series is retrospective, it is reasonably large for a subgroup of young, healthy orthopedic trauma patients to study the effects of anemia or transfusion. Whether transfused or not, many of these patients tolerated Hgb levels under 7 g/dL, which gave a large enough comparison group to evaluate the independent effects of transfusion (or of using transfusion as a marker for complication risk) or anemia as a risk factor. As a result, it appears that a more conservative transfusion strategy may be as safe as a more liberal transfusion strategy. The results of this retrospective study were used to design a prospective multidisciplinary pilot study randomizing patients to either a liberal or a conservative transfusion strategy to determine which approach might carry higher risks of complications.

Conclusion

The results of this retrospective study suggest that a conservative transfusion strategy in a young, healthy, euvolemic asymptomatic patient who is not actively bleeding may be as safe as a liberal transfusion strategy and potentially may have fewer complications than does transfusion for a conventional laboratory value. Our study results do not suggest that transfusions should be held in patients who are symptomatic at rest or in patients who are being actively resuscitated, as transfusion can be lifesaving under these circumstances. A prospective randomized study has begun at our institution with enrollment expected to take 2 years with another year needed to complete 1-year follow-up of all patients.

More than 13 million units of blood are transfused each year. Although transfusion can certainly be lifesaving, numerous studies over the past 20 years have shown significant, dose-dependent increases in morbidity, mortality, and cost with each unit of packed red blood cells (pRBCs) transfused.¹ Transfusion is one of the most common interventions in the critically ill population; however the negative effects of transfusion-related infection are well documented in the recent literature.^1-7 There is no question that transfusion of blood products can be lifesaving to acutely ill trauma patients, but there is little evidence regarding when transfusions are indicated in asymptomatic anemic patients who are no longer in need of acute resuscitation.

Several studies have analyzed healthy individuals with an isovolemic reduction in hemoglobin (Hgb) level to 5.0 g/dL.^8,9 They have found no significant compromise in oxygen delivery to the tissues. Currently, there is a lack of clinical data to suggest adequate RBC transfusion endpoints in trauma surgery.¹⁰ Given the lack of evidence to support transfusion triggers for young, healthy, asymptomatic orthopedic trauma patients, we decided to investigate whether a more conservative transfusion strategy might be as safe as a more liberal strategy.

Materials and Methods

After obtaining approval from our institutional review board, we performed a retrospective observational cohort analysis of patients treated at a level I trauma center between September 2006 and February 2009. The trauma registry included all patients who underwent surgery performed by a single orthopedic fellowship–trained trauma surgeon. All patients who had a recorded Hgb level of 9.0 g/dL or less at any time during their admission were included; they were considered no longer volume-depleted after initial resuscitation. Exclusion criteria were age under 18 years or over 50 years; pregnancy; head injury; and preexisting heart, pulmonary, or renal disease.

Initially, 963 patients were identified as orthopedic trauma patients treated by Dr. Mullis within the defined period. After inclusion and exclusion criteria were used to limit this database, the charts of the 109 patients who met the above criteria were reviewed. By chart review or telephone follow-up, 104 patients with 1-year follow-up were identified, and their cases became the basis for our analysis. Demographic information, length of hospital stay, surgeries performed, number of pRBC units transfused, Hgb level prompting transfusion, lowest recorded Hgb level, complications, and Injury Severity Score (ISS) were recorded for each patient. Seventy-two patients (69%) were male, 32 (31%) female. Mean age of the study population was 33 years.

Patients were divided into 2 groups by lowest Hgb level before first transfusion—under 7.0 g/dL and 7.0 g/dL or higher—and then by whether they had been transfused. General guidelines for erythrocyte transfusion on the orthopedic trauma service included patients who were symptomatic at rest (headache, dizziness, or shortness of breath) and asymptomatic patients with Hgb levels under 5.0 g/dL. For patients with varying (lesser) degrees of anemia, transfusion typically depended on clinical symptoms and overall decrease in Hgb level from that recorded on admission.

Patient charts were reviewed for complications extending through a 1-year period after initial discharge from the inpatient service. Patients who had not received follow-up treatment through a known outpatient clinic were contacted by telephone to ascertain outcome. Overall, 5 of the 109 patients were lost at 1-year follow-up, leaving 104 patients with 1-year follow-up (95%). Primary outcome of the study was postoperative complications. Superficial wound infection was defined as cellulitis near the surgical site within 1 year, requiring oral antibiotics; deep wound infection was defined as any related infection within 1 year of injury, requiring intravenous antibiotics or surgical débridement in the operating room. The review for complications included superficial infection, deep infection, urinary tract infection, pneumonia, pulmonary embolism, deep venous thrombosis, acute renal failure or insufficiency, nonunion, delayed union, compartment syndrome, osteomyelitis, nerve palsy, anoxic brain injury, cardiac ischemia or infarct, pancreatitis, and death.

Statistical Methods

The primary focus of this analysis was to determine if patients’ risk of complication at 1-year follow-up was affected by anemia—lowest recorded Hgb level before first transfusion for transfused patients, or lowest Hgb level during hospital stay for nontransfused patients—or whether transfusion itself might be a risk factor for complication. Multiple logistic regression models were used to determine the likelihood each group would have a complication. The dependent variable was complication rate; the explanatory variables included whether the patient was transfused, anemia/Hgb level (under 7 g/dL vs 7 g/dL or higher), and the 2-way interaction. Other possible explanatory variables entered into the model were age, sex, ISS, and whether the patient had had multiple surgeries. As the sample size was small, these variables were entered into the regression model one at a time. Results are presented as odds ratios (ORs) with corresponding 95% confidence intervals (CIs) and P values. The analysis was performed with SAS Version 9.1 (SAS Institute, Cary, North Carolina). Tests were considered statistically significant with P < .05 and marginally significant with P < .10. OR above 1 indicated that the odds of a complication occurring were higher in the exposed group (transfused patients) than in the unexposed group (nontransfused patients).

Results

The charts of 104 patients were reviewed and included in this analysis. Sixty-two patients (60%) had received a transfusion; 42 (40%) had not. Before first transfusion, 21 (34%) of the 62 transfused patients had Hgb levels under 7.0 g/dL, and the other 41 (66%) had Hgb levels of 7.0 g/dL or higher. Of the 42 nontransfused patients, 8 (19%) had lowest Hgb levels under 7.0 g/dL, and the other 34 (81%) had Hgb levels of 7.0 g/dL or higher (Table 1).

The transfused patients, considering all levels of anemia, had a mean ISS of 16.1 (range, 1-45), a mean of 2.0 operations (range, 1-6), a mean hospital stay of 18 days (range, 1-73 days), and a mean age of 34 years (range, 18-50 years). The nontransfused patients, considering all levels of anemia, had a mean ISS of 14.1 (range, 4-43), a mean of 1.4 operations (range, 1-5), a mean hospital stay of 10 days (range, 1-42 days), and a mean age of 33 years (range, 18-50 years). In the transfusion group, the mean number of transfused pRBC units was 6.9 (range, 1-31), or 7.8 units for patients with Hgb levels under 7 g/dL and 6.4 units for patients with Hgb levels of 7 g/dL or higher. At 1-year follow-up, complications were observed in 41 (66%) of the 62 transfused patients and in 17 (40%) of the 42 nontransfused patients (Table 1). The different types of complications seen in each group are listed in Table 2.

Statistical Analysis

Patients were divided into 2 groups by Hgb level—under 7.0 g/dL and 7.0 g/dL or higher—and then by whether they received pRBC transfusion. In addition, which patients had a complication over a 1-year period were identified.

For each group, we calculated sample size, number of complications, complication rate, and 95% CI for proportions. For transfused patients with Hgb level of 7.0 g/dL or higher, the complication rate was 71% (29/41). For nontransfused patients with Hgb of 7.0 g/dL or higher, the complication rate was 41% (14/34). Similarly, for transfused patients with Hgb under 7.0 g/dL, the complication rate was 57% (12/21). Last, for nontransfused patients with Hgb under 7.0 g/dL, the complication rate was 38% (3/8) (Table 3).

Transfused patients had a significantly higher risk of complication (OR, 3.1; 95% CI, 1.4-7.1; P < .01). Severity of anemia was not found to be independently associated with increased risk of complication (OR, 0.6; 95% CI, 0.3-1.6; P = .33) (Table 4). The interaction term was removed and eliminated from further analysis, as it was not found to be significant (P = .45).

Furthermore, the possibility of confounding variables (eg, age, sex, ISS, number of surgeries performed) was considered by including them in the model one at a time. From these logistic regression models, which included whether patients were transfused and level of anemia, an increased risk of complication (OR, 1.8; 95% CI, 1.1-2.9; P = .02) was found for each additional surgery, while receiving transfusion remained statistically significant (OR, 2.5; 95% CI, 1.0-5.8; P < .04). Age, sex, and ISS were not shown to be significantly associated with an increased complication rate (Ps = .71, .32, and .13, respectively).

We performed a subanalysis of the transfused patients to determine the impact of number of units transfused on complication rate. Each additional unit of pRBCs transfused increased the risk of complication, indicating a dose-dependent response (OR, 1.3; 95% CI, 1.04-1.51; P = .02).

Discussion

Transfusion is a generally accepted and common intervention both in the intensive care unit and in the perioperative period. However, there is little evidence to support routine transfusion of asymptomatic orthopedic trauma patients who are no longer within the initial resuscitative period after trauma. Nevertheless, the practice is routinely done based on expert opinion (level 5 evidence). The anemia protocol for our orthopedic trauma service routinely allowed the Hgb levels of asymptomatic healthy patients to drop to under 7.0 g/dL without transfusion; when other services were consulted or were primary, however, these asymptomatic patients were still routinely transfused based on practitioners’ practice patterns and anecdotal experiences.

In hemodynamically unstable patients, there is no acceptable substitute for blood transfusion. Blood replacement remains essential in the case of acute hemorrhage. However, the complications associated with transfusion should lead us to avoid, or at least minimize, unnecessary transfusion in young asymptomatic patients who are not actively bleeding in the postresuscitative period. In our study, we did not seek causation of increased complications with transfusion but assessed whether the risk of anemia outweighed the risk of transfusion in young, healthy, asymptomatic trauma patients who were no longer in the initial resuscitation period.

Our study was designed to evaluate a conservative transfusion strategy used in orthopedic trauma patients. We hypothesized that the risk of anemia in these asymptomatic patients would be lower than the risk of transfusing asymptomatic patients in the perioperative period. In addition, we thought the level of anemia would play a less significant role in the postoperative complication rate relative to transfusion itself. Our results suggest that a more conservative transfusion strategy of allowing asymptomatic patients to become and remain anemic even to a Hgb level of 5 g/dL may be as safe as a more liberal transfusion strategy of keeping patients at a Hgb level higher than 7 g/dL. In general, the complication rate was 66% for transfused patients and 40% for nontransfused patients. These results remain significant after correcting for possible confounding factors, including age, sex, ISS, and number of surgeries.

The results of this study do not suggest that there may not be complications associated with anemia; a 40% complication rate even in the nontransfused group is high. One might expect that patients who had isolated injuries and never developed anemia with an Hgb level under 9 g/dL might have an even lower complication rate. In the group used for inclusion in this study, however, there was not a significant increased risk for patients who tolerated a lower anemia (Hgb, <7 g/dL), whereas transfusion to keep the Hgb level above 7 g/dL appeared to correlate with a significant risk of complication and appeared to be dose-dependent. It should be noted that the complications in both the transfusion and anemia groups are not necessarily related to transfusion or anemia, as many factors in a retrospective study cannot be controlled. These findings simply suggest that it might be as safe to use a conservative transfusion strategy as a liberal transfusion strategy in this patient population.

Although our study is retrospective, prospective randomized studies in the elderly and in the critical care population have shown conservative transfusion guidelines are at least as safe as liberal transfusion strategies.^2,11 One study randomized intensive care unit patients with Hgb levels under 9.0 g/dL to 2 groups, one with liberal and the other with restrictive protocols for pRBC unit transfusion.² The liberal group maintained Hgb levels between 10.0 and 12.0 g/dL, and the restricted group kept Hgb levels between 7.0 and 9.0 g/dL. Thirty-day mortality was significantly lower in less acutely ill patients and younger patients (<55 years old) in the restrictive group than in the liberal group. It was concluded that a restrictive strategy of RBC transfusion is at least as effective as, and possibly superior to, a liberal transfusion strategy in the critically ill when considering short- and long-term outcomes. Another prospective study randomized elderly patients (N = 2016) with hip fractures and cardiovascular risk factors to a liberal transfusion strategy (if Hgb level fell under 10 g/dL) or a restrictive transfusion strategy (if Hgb level fell under 8 g/dL). The study found no difference between the 2 groups.¹¹

The deleterious effect of allogeneic blood transfusion on the immune system is complex and has been linked to the down-regulation of cellular immunity, including decreased function of natural killer cells, decreased function of macrophages and monocytes, and increased numbers of suppressor T cells.^12,13 This minimized immune response has been associated with a multitude of infectious morbidities in various patient populations.⁷ A meta-analysis of 20 studies reviewing outcomes of the effects of transfusion on postoperative bacterial infection found strong evidence supporting a correlation.⁵ Their analysis found an OR of 5.3 (range, 5.0-5.4) for infectious complication after allogeneic transfusion in the trauma population, and an OR of 3.5 (range, 1.4-15.2) considering all patient populations.

Similar results showing increased risk of infectious morbidities associated with transfusion were found in other studies involving the critically ill, patients after hip arthroplasty, and cardiothoracic surgery and general trauma populations.^1,3,4,14,15 Furthermore, these results were seen in a dose-dependent response leading to increased incidence of complication with each unit of blood transfused.

Our study did not focus only on infection but included other complications (eg, cardiac, renal, and brain ischemia) that might be associated with anemia or transfusion. It is intuitive that anemia can cause ischemic events but less intuitive that allogeneic transfusion can also cause ischemic events because of the poor deformability of the cells due to storage, which can lead to “sludging” in capillaries throughout the body.¹⁶ This has been shown to be important in animal models, but it is unclear what poses more risk in humans—anemia without transfusion or the initial insult from transfusion, before the body clears the “waste” from stored cells and the remaining viable cells gain oxygen-carrying capacity.

Our study has several limitations. The number of patients who had severe anemia (Hgb level, <7 g/dL) and were not transfused is relatively small compared with the numbers in the other groups used for comparison. Because our study was retrospective, we could only find associations and not prove causation. This is significant, as the higher complication rate seen with transfusions may only be caused by the transfusion as a predictor of a patient requiring more complex surgery with higher blood loss (and higher risk of complication) or other such risk factors that led to transfusion, but not the transfusion itself causing the complication. An attempt was made to remove this potential bias by controlling for age, sex, ISS, and whether the patient had multiple surgeries. However, there may have been other significant confounding variables not excluded. As complications were assessed by chart review, they may not include those that occurred at other institutions and that were never reported to the practitioners at our facility (though we did have the ability to search records of neighboring institutions electronically when electronic medical records were available). That no functional outcomes were included in this retrospective review might make the complication rate appear more or less sensitive than the patients’ own opinions regarding their outcomes. All these weaknesses could call into question whether the statistically significant higher risk associated with allogeneic transfusion found in this study is real, but the focus and reason for pursuing this study were to determine if permissive anemia was dangerous or would be associated with a higher risk of complications than routine allogeneic transfusion of asymptomatic patients to treat a laboratory value.

Strengths of the study include the review of a single surgeon’s practice with a written protocol in place for anemic orthopedic trauma patients. The 95% follow-up (104/109 patients) is good for this type of trauma population. Although this series is retrospective, it is reasonably large for a subgroup of young, healthy orthopedic trauma patients to study the effects of anemia or transfusion. Whether transfused or not, many of these patients tolerated Hgb levels under 7 g/dL, which gave a large enough comparison group to evaluate the independent effects of transfusion (or of using transfusion as a marker for complication risk) or anemia as a risk factor. As a result, it appears that a more conservative transfusion strategy may be as safe as a more liberal transfusion strategy. The results of this retrospective study were used to design a prospective multidisciplinary pilot study randomizing patients to either a liberal or a conservative transfusion strategy to determine which approach might carry higher risks of complications.

Conclusion

The results of this retrospective study suggest that a conservative transfusion strategy in a young, healthy, euvolemic asymptomatic patient who is not actively bleeding may be as safe as a liberal transfusion strategy and potentially may have fewer complications than does transfusion for a conventional laboratory value. Our study results do not suggest that transfusions should be held in patients who are symptomatic at rest or in patients who are being actively resuscitated, as transfusion can be lifesaving under these circumstances. A prospective randomized study has begun at our institution with enrollment expected to take 2 years with another year needed to complete 1-year follow-up of all patients.

Anterior cruciate ligament (ACL) reconstruction remains one of the most common orthopedic procedures; almost 100,000 are performed in the United States each year, and they are among the procedures more commonly performed by surgeons specializing in sports medicine and by general orthopedists.^1,2 Recent years have seen a trend toward replacing the gold standard of bone–patellar tendon–bone autograft with autograft or allograft hamstring tendon in ACL reconstruction.³ This shift is being made to try to avoid the donor-site morbidity of patellar tendon autografts and decrease the incidence of postoperative anterior knee pain. With increased use of hamstring grafts in ACL reconstruction, graft fixation strength has become a priority in attempts to optimize recovery and rehabilitation.⁴

Rigid fixation of hamstring grafts is now recognized as a crucial factor in the long-term success of ACL reconstruction. Grafts must withstand both early rehabilitation forces as high as 500 N⁵ and stresses to the native ACL during healing, which may take up to 12 weeks for soft-tissue incorporation.⁶

The challenge has been to engineer devices that provide stable, rigid graft fixation that allows expeditious tendon-to-bone healing and increased construct stiffness. Many new fixation devices are being marketed, and there is controversy regarding which provides the best stability and strength.⁷ Several studies have tested various fixation devices,^8-16 but so far several devices have not been compared with one another.

We conducted a study to determine if femoral hamstring fixation devices used in ACL reconstruction differ in fixation strength. We hypothesized we would find no differences.

Materials and Methods

Fifty porcine femurs were harvested after the animals had been euthanized for other studies at our institution. Our study was approved by the institutional animal care and use committee. Specimens were stored at –25°C and, on day of testing, thawed to room temperature. Gracilis and semitendinosus tendon grafts were donated by a tissue bank (LifeNet Health, Virginia Beach, Virginia). The grafts were stored at –25°C; on day of testing, tendons were thawed to room temperature.

We evaluated 5 different femoral fixation devices (Figure 1): Delta screw and Bio-TransFix (Arthrex, Naples, Florida) and Bone Mulch screw, EZLoc, and Zip Loop (Arthrotek, Warsaw, Indiana). For each device, 10 ACL fixation constructs were tested.

Quadrupled human semitendinosus–gracilis tendon grafts were fixed into the femurs using the 5 femoral fixation devices. All fixations were done to manufacturer specifications.

Cyclic loading was followed by testing with the load-to-failure (LTF) protocol described by Kousa and colleagues.¹³ Specimens were tested in a custom load fixture (Figure 2). The base fixture used an adjustable angle vise mounted on a free rotary stage and a free x-y translation stage. This system allowed the load axis to be oriented to and aligned with the graft tunnel in the porcine femur, preventing off-axis or torsional loading of the grafts.

Pneumatic grips equipped with a custom pincer attachment allowed the graft to be grasped under a constant grip force during testing, regardless of graft thinning under tensile loads. Graft specimens were initially looped over a 3.8-mm horizontal metal shaft, and the 2 strands were double-looped at the graft insertion site. The 2 free strands were then drawn up around the metal shaft, and the shaft was placed above the serrated jaws. The metal shaft with enveloping tendon strands rested on a flat shelf at the top of the grip serrations. This configuration prevented the metal shaft and tendon strands from being pulled through the serrations when compressive force was applied to the jaws.

Before the study, the grip design was tested. There was no detectable relative motion of the strands at the grip end during graft testing to failure. The pincer attachment allowed close approach of the grips to the specimen at all femoral condyle orientations, so that a 25-mm length of exposed graft could be obtained for each specimen under initial conditions.

In the cyclic loading test, the load was applied parallel to the long axis of the femoral tunnel. A 50-N preload was initially applied to each specimen for 10 seconds, and the length of the exposed graft between grips and graft insertion was recorded. Subsequently, 1500 loading cycles between 50 N and 200 N at a rate of 1 cycle per 2 seconds (0.5 Hz) were performed. Standard force-displacement curves were then generated.

Specimens surviving the cyclic loading then underwent a single-cycle LTF test in which the load was applied parallel to the long axis of the drill hole at a rate of 50 mm per minute.

Residual displacement, stiffness, and ultimate LTF data were recorded from the force-displacement curves. Residual displacement data were generated from the cyclic loading test; residual displacement was determined by subtracting preload displacement from displacement at 1, 10, 50, 100, 250, 500, 1000, and 1500 cycles. Stiffness data were generated from the single-cycle LTF test; stiffness was defined as the linear region slope of the force-displacement curve corresponding to the steepest straight-line tangent to the loading curve. Ultimate LTF data were generated from the single-cycle LTF test; ultimate LTF was defined as the maximum load sustained by the specimen during a constant-displacement-rate tensile test for graft pullout.

Statistical analysis generated standard descriptive statistics: means, standard deviations, and proportions. One-way analysis of variance (ANOVA) was used to determine any statistically significant differences in stiffness, yield load, and residual displacement between the different fixation devices. Differences in force (load) between the single cycle and the cyclic loading test were determined by ANOVA. P < .05 was considered statistically significant for all tests.

Results

The modes of failure for the devices differed slightly (Table). Bone Mulch screw failed with a fracture through the femoral condyle extending to the bone tunnel. Zip Loop and EZLoc failed by pulling through their cortical attachment on the lateral femoral condyle. Bio-TransFix broke in the tunnel during LTF. Delta screw failed with slippage of the fixation device, and the tendons pulled out through the tunnel.

For the cyclic loading tests, only 2 of the 10 Delta screws completed the 1500-cycle loading test before failure. Of the 8 Delta screws that did not complete this testing, the majority failed after about 100 cycles. All 10 tests of Bone Mulch, Zip Loop, EZLoc, and Bio-TransFix completed the 1500-cycle loading test.

Residual displacement data were calculated from cyclic loading tests (Table). Mean (SD) residual displacement was lowest for Bio-TransFix at 4.1 (0.4) mm, followed by Bone Mulch at 5.2 (1.0) mm, EZLoc at 6.4 (1.1) mm, and Zip Loop at 6.8 (1.3) mm. Delta screws at 8.2 (1.4) mm had the highest residual displacement, though only 2 completed the cyclic tests. Bio-TransFix had significantly (P < .001) less residual displacement compared with EZLoc, Zip Loop, and Delta. Bone Mulch had significantly less residual displacement compared with Zip Loop (P < .05) and Delta (P < .01).

Stiffness data were calculated from LTF tests (Table). Mean (SD) stiffness was highest for Bone Mulch at 218 (25.9) N/mm, followed by Bio-TransFix at 171 (24.2) N/mm, EZLoc at 122 (24.1) N/mm, Zip Loop at 105 (18.9) N/mm, and Delta at 84 (16.4) N/mm. Bone Mulch had significantly (P < .001) higher stiffness compared with Bio-TransFix, EZLoc, Zip Loop, and Delta. Bio-TransFix had significantly (P < .001) higher stiffness compared with EZLoc, Zip Loop, and Delta.

Mean (SD) ultimate LTF was highest for Bone Mulch at 867 (164) N, followed by Zip Loop at 615 (72.3) N, Bio-TransFix at 552 (141) N, EZLoc at 476 (89.7) N, and Delta at 410 (65.3) N (Table). Bone Mulch failed at a statistically significantly (P < .001) higher load compared with Zip Loop, Bio-TransFix, EZLoc, and Delta. There were no significant differences in mean LTF among Zip Loop, Bio-TransFix, EZLoc, and Delta.

Discussion

In this biomechanical comparison of 5 different femoral fixation devices, the Bone Mulch screw had results superior to those of the other implants. Bone Mulch failed at higher LTF and higher stiffness. Bio-TransFix performed well and had residual displacement similar to that of Bone Mulch, but significantly lower LTF. Overall, EZLoc and Zip Loop were similar to each other in performance. The Delta (interference) screw performed poorly with respect to LTF, residual displacement, and stiffness; a large proportion of these screws failed early into cyclic loading.

Bone Mulch and Bio-TransFix overall outperformed the other fixation devices. These 2 devices are cortical-cancellous suspension devices, which provide transcondylar fixation and resist tensile forces perpendicular to the pullout force. Multiple biomechanical studies have found superior performance for these types of devices compared with various implants.^10,13,15,16

Our results were similar to those of Kousa and colleagues,¹³ who found the Bone Mulch screw to provide highest LTF, highest stiffness, and lowest residual displacement. Another study found significantly higher stiffness for the Bone Mulch screw than for the Endobutton, a cortical suspensory fixation device.¹⁴ Bone Mulch failure modes differed, however. In the study by Kousa and colleagues,¹³ 3 specimens failed with bending of the screw tip, and 7 failed with rupture of the tendon loop. All specimens in our study failed with fractures through the condyle. It is unclear why the failure modes differed, as we followed similar manufacturer protocols for inserting the device. It is possible the bone mass density of the porcine femurs differed between studies. This was not reported by Kousa and colleagues,¹³ and we did not perform testing either. However, all the porcine femurs were about the same age for testing of each device in this study.

Bio-TransFix has also been compared with various implants, but not in the same study. Brown and colleagues⁸ found the TransFix device significantly stiffer than the Endobutton CL. Shen and colleagues¹⁶ determined that TransFix had significantly lower residual displacement compared with Endobutton CL. Milano and colleagues¹⁵ compared multiple cortical suspensory fixation devices, including Endobutton CL, with TransFix and Bio-TransFix, and concluded the cortical-cancellous devices (TransFix, Bio-TransFix) offered the best and most predictable results in terms of elongation, fixation strength, and stiffness. TransFix has also been shown to be superior to interference screw fixation in biomechanical studies.^10,15

Clinical outcomes of studies using TransFix for femoral fixation have been favorable, with improved Lysholm scores and improved laxity according to the KT-1000 test.¹⁷ However, multiple prospective studies have found no clinical difference in knee laxity between interference screw and Endobutton at 1- to 2-year follow-up^18-20 and no difference in clinical outcome scores, such as the International Knee Documentation Committee score.^11,18-20

Although these studies have shown no major clinical differences at short-term follow-up, the early aggressive rehabilitation period is the larger concern. Our study clearly demonstrated the biomechanical strength of transcondylar devices over other devices. The concern with transcondylar devices (vs other devices) is the increased difficulty that inexperienced surgeons have inserting them. In addition, when removed, transcondylar devices leave a large bone void.

In the present study, an important concern with femoral graft fixation is the poor performance of interference screws. Other authors recently expressed concern with using interference screws in soft-tissue ACL grafts—based on biomechanical study results of increased slippage, bone tunnel widening, and less strength.⁷ In the present study, Delta screws consistently performed poorest with respect to ultimate LTF, residual displacement, and stiffness. Only 20% of these screws completed 1500 cycles. Poor performance of interference screws has also been seen in other studies in tibial graft fixation^21,22 and femoral graft fixation.^13-15 Given their poor biomechanical properties, as seen in our study and these other studies, we think use of an interference screw alone is a poor choice for fixation.

Combined fixation techniques—interference screw plus other device(s)—may be used in clinical practice, but the present study did not evaluate any. In a biomechanical study, Yoo and colleagues²³ compared an interference screw; an interference screw plus a cortical screw and a spiked washer; and a cortical screw and a spiked washer used alone in the tibia. Stiffness nearly doubled, residual displacement was less, and ultimate LTF was significantly higher in the group with the interference screw plus the cortical screw and the spiked washer. In a similar study involving femoral fixation, Oh and colleagues²⁴ demonstrated improved stiffness, residual displacement, and LTF in cyclic testing with the combination of interference screw and Endobutton CL, compared with Endobutton CL alone. Further studies may include direct comparisons of additional femoral fixation techniques using more than 1 device.

The Zip Loop, or Toggle Loc with Zip Loop technology, is a suspensory cortical fixation device. It was initially designed for use in ACL fixation but has also been used in other surgeries, including distal biceps repair²⁵ and ulnar collateral ligament reconstruction.²⁶ The device itself is easy to use; more important, it allows for adjustment of graft length within the bone tunnel after deployment of the cortical fixation. Few biomechanical studies have been conducted with Zip Loop.^9,12 The present study is the first to compare Zip Loop with devices other than suspensory cortical fixation devices. Zip Loop performed very well in LTF testing but had lower stiffness and higher residual displacement compared with the transcondylar fixation devices. Despite these findings, we have continued to use this device for femoral fixation in ACL reconstruction because of its ease of insertion, the ability to adjust graft tension within the bone tunnel, and the difficulties encountered inserting and removing transcondylar fixation.

We recognize the limitations in our study design with respect to how axial and cyclical loading compares with the physiologic orientation of the ACL during ambulation and running activities. This biomechanical study was not able to replicate these types of activities. However, it did provide good data supporting early rehabilitation with various fixation devices, though concern with use of interference screws remains.

Conclusion

Superior strength in fixation of hamstring grafts in the femur was demonstrated by Bone Mulch screws, followed closely by Bio-TransFix. Delta screws demonstrated poor displacement, stiffness, and LTF. When used as the sole femoral fixation device, a device with low LTF, decreased stiffness, and high residual displacement should be used cautiously in patients undergoing aggressive rehabilitation.

Anterior cruciate ligament (ACL) reconstruction remains one of the most common orthopedic procedures; almost 100,000 are performed in the United States each year, and they are among the procedures more commonly performed by surgeons specializing in sports medicine and by general orthopedists.^1,2 Recent years have seen a trend toward replacing the gold standard of bone–patellar tendon–bone autograft with autograft or allograft hamstring tendon in ACL reconstruction.³ This shift is being made to try to avoid the donor-site morbidity of patellar tendon autografts and decrease the incidence of postoperative anterior knee pain. With increased use of hamstring grafts in ACL reconstruction, graft fixation strength has become a priority in attempts to optimize recovery and rehabilitation.⁴

Rigid fixation of hamstring grafts is now recognized as a crucial factor in the long-term success of ACL reconstruction. Grafts must withstand both early rehabilitation forces as high as 500 N⁵ and stresses to the native ACL during healing, which may take up to 12 weeks for soft-tissue incorporation.⁶

The challenge has been to engineer devices that provide stable, rigid graft fixation that allows expeditious tendon-to-bone healing and increased construct stiffness. Many new fixation devices are being marketed, and there is controversy regarding which provides the best stability and strength.⁷ Several studies have tested various fixation devices,^8-16 but so far several devices have not been compared with one another.

We conducted a study to determine if femoral hamstring fixation devices used in ACL reconstruction differ in fixation strength. We hypothesized we would find no differences.

Materials and Methods

Fifty porcine femurs were harvested after the animals had been euthanized for other studies at our institution. Our study was approved by the institutional animal care and use committee. Specimens were stored at –25°C and, on day of testing, thawed to room temperature. Gracilis and semitendinosus tendon grafts were donated by a tissue bank (LifeNet Health, Virginia Beach, Virginia). The grafts were stored at –25°C; on day of testing, tendons were thawed to room temperature.

We evaluated 5 different femoral fixation devices (Figure 1): Delta screw and Bio-TransFix (Arthrex, Naples, Florida) and Bone Mulch screw, EZLoc, and Zip Loop (Arthrotek, Warsaw, Indiana). For each device, 10 ACL fixation constructs were tested.

Quadrupled human semitendinosus–gracilis tendon grafts were fixed into the femurs using the 5 femoral fixation devices. All fixations were done to manufacturer specifications.

Cyclic loading was followed by testing with the load-to-failure (LTF) protocol described by Kousa and colleagues.¹³ Specimens were tested in a custom load fixture (Figure 2). The base fixture used an adjustable angle vise mounted on a free rotary stage and a free x-y translation stage. This system allowed the load axis to be oriented to and aligned with the graft tunnel in the porcine femur, preventing off-axis or torsional loading of the grafts.

Pneumatic grips equipped with a custom pincer attachment allowed the graft to be grasped under a constant grip force during testing, regardless of graft thinning under tensile loads. Graft specimens were initially looped over a 3.8-mm horizontal metal shaft, and the 2 strands were double-looped at the graft insertion site. The 2 free strands were then drawn up around the metal shaft, and the shaft was placed above the serrated jaws. The metal shaft with enveloping tendon strands rested on a flat shelf at the top of the grip serrations. This configuration prevented the metal shaft and tendon strands from being pulled through the serrations when compressive force was applied to the jaws.

Before the study, the grip design was tested. There was no detectable relative motion of the strands at the grip end during graft testing to failure. The pincer attachment allowed close approach of the grips to the specimen at all femoral condyle orientations, so that a 25-mm length of exposed graft could be obtained for each specimen under initial conditions.

In the cyclic loading test, the load was applied parallel to the long axis of the femoral tunnel. A 50-N preload was initially applied to each specimen for 10 seconds, and the length of the exposed graft between grips and graft insertion was recorded. Subsequently, 1500 loading cycles between 50 N and 200 N at a rate of 1 cycle per 2 seconds (0.5 Hz) were performed. Standard force-displacement curves were then generated.

Specimens surviving the cyclic loading then underwent a single-cycle LTF test in which the load was applied parallel to the long axis of the drill hole at a rate of 50 mm per minute.

Residual displacement, stiffness, and ultimate LTF data were recorded from the force-displacement curves. Residual displacement data were generated from the cyclic loading test; residual displacement was determined by subtracting preload displacement from displacement at 1, 10, 50, 100, 250, 500, 1000, and 1500 cycles. Stiffness data were generated from the single-cycle LTF test; stiffness was defined as the linear region slope of the force-displacement curve corresponding to the steepest straight-line tangent to the loading curve. Ultimate LTF data were generated from the single-cycle LTF test; ultimate LTF was defined as the maximum load sustained by the specimen during a constant-displacement-rate tensile test for graft pullout.

Statistical analysis generated standard descriptive statistics: means, standard deviations, and proportions. One-way analysis of variance (ANOVA) was used to determine any statistically significant differences in stiffness, yield load, and residual displacement between the different fixation devices. Differences in force (load) between the single cycle and the cyclic loading test were determined by ANOVA. P < .05 was considered statistically significant for all tests.

Results

The modes of failure for the devices differed slightly (Table). Bone Mulch screw failed with a fracture through the femoral condyle extending to the bone tunnel. Zip Loop and EZLoc failed by pulling through their cortical attachment on the lateral femoral condyle. Bio-TransFix broke in the tunnel during LTF. Delta screw failed with slippage of the fixation device, and the tendons pulled out through the tunnel.

For the cyclic loading tests, only 2 of the 10 Delta screws completed the 1500-cycle loading test before failure. Of the 8 Delta screws that did not complete this testing, the majority failed after about 100 cycles. All 10 tests of Bone Mulch, Zip Loop, EZLoc, and Bio-TransFix completed the 1500-cycle loading test.

Residual displacement data were calculated from cyclic loading tests (Table). Mean (SD) residual displacement was lowest for Bio-TransFix at 4.1 (0.4) mm, followed by Bone Mulch at 5.2 (1.0) mm, EZLoc at 6.4 (1.1) mm, and Zip Loop at 6.8 (1.3) mm. Delta screws at 8.2 (1.4) mm had the highest residual displacement, though only 2 completed the cyclic tests. Bio-TransFix had significantly (P < .001) less residual displacement compared with EZLoc, Zip Loop, and Delta. Bone Mulch had significantly less residual displacement compared with Zip Loop (P < .05) and Delta (P < .01).

Stiffness data were calculated from LTF tests (Table). Mean (SD) stiffness was highest for Bone Mulch at 218 (25.9) N/mm, followed by Bio-TransFix at 171 (24.2) N/mm, EZLoc at 122 (24.1) N/mm, Zip Loop at 105 (18.9) N/mm, and Delta at 84 (16.4) N/mm. Bone Mulch had significantly (P < .001) higher stiffness compared with Bio-TransFix, EZLoc, Zip Loop, and Delta. Bio-TransFix had significantly (P < .001) higher stiffness compared with EZLoc, Zip Loop, and Delta.

Mean (SD) ultimate LTF was highest for Bone Mulch at 867 (164) N, followed by Zip Loop at 615 (72.3) N, Bio-TransFix at 552 (141) N, EZLoc at 476 (89.7) N, and Delta at 410 (65.3) N (Table). Bone Mulch failed at a statistically significantly (P < .001) higher load compared with Zip Loop, Bio-TransFix, EZLoc, and Delta. There were no significant differences in mean LTF among Zip Loop, Bio-TransFix, EZLoc, and Delta.

Discussion

In this biomechanical comparison of 5 different femoral fixation devices, the Bone Mulch screw had results superior to those of the other implants. Bone Mulch failed at higher LTF and higher stiffness. Bio-TransFix performed well and had residual displacement similar to that of Bone Mulch, but significantly lower LTF. Overall, EZLoc and Zip Loop were similar to each other in performance. The Delta (interference) screw performed poorly with respect to LTF, residual displacement, and stiffness; a large proportion of these screws failed early into cyclic loading.

Bone Mulch and Bio-TransFix overall outperformed the other fixation devices. These 2 devices are cortical-cancellous suspension devices, which provide transcondylar fixation and resist tensile forces perpendicular to the pullout force. Multiple biomechanical studies have found superior performance for these types of devices compared with various implants.^10,13,15,16

Our results were similar to those of Kousa and colleagues,¹³ who found the Bone Mulch screw to provide highest LTF, highest stiffness, and lowest residual displacement. Another study found significantly higher stiffness for the Bone Mulch screw than for the Endobutton, a cortical suspensory fixation device.¹⁴ Bone Mulch failure modes differed, however. In the study by Kousa and colleagues,¹³ 3 specimens failed with bending of the screw tip, and 7 failed with rupture of the tendon loop. All specimens in our study failed with fractures through the condyle. It is unclear why the failure modes differed, as we followed similar manufacturer protocols for inserting the device. It is possible the bone mass density of the porcine femurs differed between studies. This was not reported by Kousa and colleagues,¹³ and we did not perform testing either. However, all the porcine femurs were about the same age for testing of each device in this study.

Bio-TransFix has also been compared with various implants, but not in the same study. Brown and colleagues⁸ found the TransFix device significantly stiffer than the Endobutton CL. Shen and colleagues¹⁶ determined that TransFix had significantly lower residual displacement compared with Endobutton CL. Milano and colleagues¹⁵ compared multiple cortical suspensory fixation devices, including Endobutton CL, with TransFix and Bio-TransFix, and concluded the cortical-cancellous devices (TransFix, Bio-TransFix) offered the best and most predictable results in terms of elongation, fixation strength, and stiffness. TransFix has also been shown to be superior to interference screw fixation in biomechanical studies.^10,15

Clinical outcomes of studies using TransFix for femoral fixation have been favorable, with improved Lysholm scores and improved laxity according to the KT-1000 test.¹⁷ However, multiple prospective studies have found no clinical difference in knee laxity between interference screw and Endobutton at 1- to 2-year follow-up^18-20 and no difference in clinical outcome scores, such as the International Knee Documentation Committee score.^11,18-20

Although these studies have shown no major clinical differences at short-term follow-up, the early aggressive rehabilitation period is the larger concern. Our study clearly demonstrated the biomechanical strength of transcondylar devices over other devices. The concern with transcondylar devices (vs other devices) is the increased difficulty that inexperienced surgeons have inserting them. In addition, when removed, transcondylar devices leave a large bone void.

In the present study, an important concern with femoral graft fixation is the poor performance of interference screws. Other authors recently expressed concern with using interference screws in soft-tissue ACL grafts—based on biomechanical study results of increased slippage, bone tunnel widening, and less strength.⁷ In the present study, Delta screws consistently performed poorest with respect to ultimate LTF, residual displacement, and stiffness. Only 20% of these screws completed 1500 cycles. Poor performance of interference screws has also been seen in other studies in tibial graft fixation^21,22 and femoral graft fixation.^13-15 Given their poor biomechanical properties, as seen in our study and these other studies, we think use of an interference screw alone is a poor choice for fixation.

Combined fixation techniques—interference screw plus other device(s)—may be used in clinical practice, but the present study did not evaluate any. In a biomechanical study, Yoo and colleagues²³ compared an interference screw; an interference screw plus a cortical screw and a spiked washer; and a cortical screw and a spiked washer used alone in the tibia. Stiffness nearly doubled, residual displacement was less, and ultimate LTF was significantly higher in the group with the interference screw plus the cortical screw and the spiked washer. In a similar study involving femoral fixation, Oh and colleagues²⁴ demonstrated improved stiffness, residual displacement, and LTF in cyclic testing with the combination of interference screw and Endobutton CL, compared with Endobutton CL alone. Further studies may include direct comparisons of additional femoral fixation techniques using more than 1 device.

The Zip Loop, or Toggle Loc with Zip Loop technology, is a suspensory cortical fixation device. It was initially designed for use in ACL fixation but has also been used in other surgeries, including distal biceps repair²⁵ and ulnar collateral ligament reconstruction.²⁶ The device itself is easy to use; more important, it allows for adjustment of graft length within the bone tunnel after deployment of the cortical fixation. Few biomechanical studies have been conducted with Zip Loop.^9,12 The present study is the first to compare Zip Loop with devices other than suspensory cortical fixation devices. Zip Loop performed very well in LTF testing but had lower stiffness and higher residual displacement compared with the transcondylar fixation devices. Despite these findings, we have continued to use this device for femoral fixation in ACL reconstruction because of its ease of insertion, the ability to adjust graft tension within the bone tunnel, and the difficulties encountered inserting and removing transcondylar fixation.

We recognize the limitations in our study design with respect to how axial and cyclical loading compares with the physiologic orientation of the ACL during ambulation and running activities. This biomechanical study was not able to replicate these types of activities. However, it did provide good data supporting early rehabilitation with various fixation devices, though concern with use of interference screws remains.

Conclusion

Superior strength in fixation of hamstring grafts in the femur was demonstrated by Bone Mulch screws, followed closely by Bio-TransFix. Delta screws demonstrated poor displacement, stiffness, and LTF. When used as the sole femoral fixation device, a device with low LTF, decreased stiffness, and high residual displacement should be used cautiously in patients undergoing aggressive rehabilitation.

Anterior cruciate ligament (ACL) reconstruction remains one of the most common orthopedic procedures; almost 100,000 are performed in the United States each year, and they are among the procedures more commonly performed by surgeons specializing in sports medicine and by general orthopedists.^1,2 Recent years have seen a trend toward replacing the gold standard of bone–patellar tendon–bone autograft with autograft or allograft hamstring tendon in ACL reconstruction.³ This shift is being made to try to avoid the donor-site morbidity of patellar tendon autografts and decrease the incidence of postoperative anterior knee pain. With increased use of hamstring grafts in ACL reconstruction, graft fixation strength has become a priority in attempts to optimize recovery and rehabilitation.⁴

Rigid fixation of hamstring grafts is now recognized as a crucial factor in the long-term success of ACL reconstruction. Grafts must withstand both early rehabilitation forces as high as 500 N⁵ and stresses to the native ACL during healing, which may take up to 12 weeks for soft-tissue incorporation.⁶

The challenge has been to engineer devices that provide stable, rigid graft fixation that allows expeditious tendon-to-bone healing and increased construct stiffness. Many new fixation devices are being marketed, and there is controversy regarding which provides the best stability and strength.⁷ Several studies have tested various fixation devices,^8-16 but so far several devices have not been compared with one another.

We conducted a study to determine if femoral hamstring fixation devices used in ACL reconstruction differ in fixation strength. We hypothesized we would find no differences.

Materials and Methods

Fifty porcine femurs were harvested after the animals had been euthanized for other studies at our institution. Our study was approved by the institutional animal care and use committee. Specimens were stored at –25°C and, on day of testing, thawed to room temperature. Gracilis and semitendinosus tendon grafts were donated by a tissue bank (LifeNet Health, Virginia Beach, Virginia). The grafts were stored at –25°C; on day of testing, tendons were thawed to room temperature.

We evaluated 5 different femoral fixation devices (Figure 1): Delta screw and Bio-TransFix (Arthrex, Naples, Florida) and Bone Mulch screw, EZLoc, and Zip Loop (Arthrotek, Warsaw, Indiana). For each device, 10 ACL fixation constructs were tested.

Quadrupled human semitendinosus–gracilis tendon grafts were fixed into the femurs using the 5 femoral fixation devices. All fixations were done to manufacturer specifications.

Cyclic loading was followed by testing with the load-to-failure (LTF) protocol described by Kousa and colleagues.¹³ Specimens were tested in a custom load fixture (Figure 2). The base fixture used an adjustable angle vise mounted on a free rotary stage and a free x-y translation stage. This system allowed the load axis to be oriented to and aligned with the graft tunnel in the porcine femur, preventing off-axis or torsional loading of the grafts.

Pneumatic grips equipped with a custom pincer attachment allowed the graft to be grasped under a constant grip force during testing, regardless of graft thinning under tensile loads. Graft specimens were initially looped over a 3.8-mm horizontal metal shaft, and the 2 strands were double-looped at the graft insertion site. The 2 free strands were then drawn up around the metal shaft, and the shaft was placed above the serrated jaws. The metal shaft with enveloping tendon strands rested on a flat shelf at the top of the grip serrations. This configuration prevented the metal shaft and tendon strands from being pulled through the serrations when compressive force was applied to the jaws.

Before the study, the grip design was tested. There was no detectable relative motion of the strands at the grip end during graft testing to failure. The pincer attachment allowed close approach of the grips to the specimen at all femoral condyle orientations, so that a 25-mm length of exposed graft could be obtained for each specimen under initial conditions.

In the cyclic loading test, the load was applied parallel to the long axis of the femoral tunnel. A 50-N preload was initially applied to each specimen for 10 seconds, and the length of the exposed graft between grips and graft insertion was recorded. Subsequently, 1500 loading cycles between 50 N and 200 N at a rate of 1 cycle per 2 seconds (0.5 Hz) were performed. Standard force-displacement curves were then generated.

Specimens surviving the cyclic loading then underwent a single-cycle LTF test in which the load was applied parallel to the long axis of the drill hole at a rate of 50 mm per minute.

Residual displacement, stiffness, and ultimate LTF data were recorded from the force-displacement curves. Residual displacement data were generated from the cyclic loading test; residual displacement was determined by subtracting preload displacement from displacement at 1, 10, 50, 100, 250, 500, 1000, and 1500 cycles. Stiffness data were generated from the single-cycle LTF test; stiffness was defined as the linear region slope of the force-displacement curve corresponding to the steepest straight-line tangent to the loading curve. Ultimate LTF data were generated from the single-cycle LTF test; ultimate LTF was defined as the maximum load sustained by the specimen during a constant-displacement-rate tensile test for graft pullout.

Statistical analysis generated standard descriptive statistics: means, standard deviations, and proportions. One-way analysis of variance (ANOVA) was used to determine any statistically significant differences in stiffness, yield load, and residual displacement between the different fixation devices. Differences in force (load) between the single cycle and the cyclic loading test were determined by ANOVA. P < .05 was considered statistically significant for all tests.

Results

The modes of failure for the devices differed slightly (Table). Bone Mulch screw failed with a fracture through the femoral condyle extending to the bone tunnel. Zip Loop and EZLoc failed by pulling through their cortical attachment on the lateral femoral condyle. Bio-TransFix broke in the tunnel during LTF. Delta screw failed with slippage of the fixation device, and the tendons pulled out through the tunnel.

For the cyclic loading tests, only 2 of the 10 Delta screws completed the 1500-cycle loading test before failure. Of the 8 Delta screws that did not complete this testing, the majority failed after about 100 cycles. All 10 tests of Bone Mulch, Zip Loop, EZLoc, and Bio-TransFix completed the 1500-cycle loading test.

Residual displacement data were calculated from cyclic loading tests (Table). Mean (SD) residual displacement was lowest for Bio-TransFix at 4.1 (0.4) mm, followed by Bone Mulch at 5.2 (1.0) mm, EZLoc at 6.4 (1.1) mm, and Zip Loop at 6.8 (1.3) mm. Delta screws at 8.2 (1.4) mm had the highest residual displacement, though only 2 completed the cyclic tests. Bio-TransFix had significantly (P < .001) less residual displacement compared with EZLoc, Zip Loop, and Delta. Bone Mulch had significantly less residual displacement compared with Zip Loop (P < .05) and Delta (P < .01).

Stiffness data were calculated from LTF tests (Table). Mean (SD) stiffness was highest for Bone Mulch at 218 (25.9) N/mm, followed by Bio-TransFix at 171 (24.2) N/mm, EZLoc at 122 (24.1) N/mm, Zip Loop at 105 (18.9) N/mm, and Delta at 84 (16.4) N/mm. Bone Mulch had significantly (P < .001) higher stiffness compared with Bio-TransFix, EZLoc, Zip Loop, and Delta. Bio-TransFix had significantly (P < .001) higher stiffness compared with EZLoc, Zip Loop, and Delta.

Mean (SD) ultimate LTF was highest for Bone Mulch at 867 (164) N, followed by Zip Loop at 615 (72.3) N, Bio-TransFix at 552 (141) N, EZLoc at 476 (89.7) N, and Delta at 410 (65.3) N (Table). Bone Mulch failed at a statistically significantly (P < .001) higher load compared with Zip Loop, Bio-TransFix, EZLoc, and Delta. There were no significant differences in mean LTF among Zip Loop, Bio-TransFix, EZLoc, and Delta.

Discussion

In this biomechanical comparison of 5 different femoral fixation devices, the Bone Mulch screw had results superior to those of the other implants. Bone Mulch failed at higher LTF and higher stiffness. Bio-TransFix performed well and had residual displacement similar to that of Bone Mulch, but significantly lower LTF. Overall, EZLoc and Zip Loop were similar to each other in performance. The Delta (interference) screw performed poorly with respect to LTF, residual displacement, and stiffness; a large proportion of these screws failed early into cyclic loading.

Bone Mulch and Bio-TransFix overall outperformed the other fixation devices. These 2 devices are cortical-cancellous suspension devices, which provide transcondylar fixation and resist tensile forces perpendicular to the pullout force. Multiple biomechanical studies have found superior performance for these types of devices compared with various implants.^10,13,15,16

Our results were similar to those of Kousa and colleagues,¹³ who found the Bone Mulch screw to provide highest LTF, highest stiffness, and lowest residual displacement. Another study found significantly higher stiffness for the Bone Mulch screw than for the Endobutton, a cortical suspensory fixation device.¹⁴ Bone Mulch failure modes differed, however. In the study by Kousa and colleagues,¹³ 3 specimens failed with bending of the screw tip, and 7 failed with rupture of the tendon loop. All specimens in our study failed with fractures through the condyle. It is unclear why the failure modes differed, as we followed similar manufacturer protocols for inserting the device. It is possible the bone mass density of the porcine femurs differed between studies. This was not reported by Kousa and colleagues,¹³ and we did not perform testing either. However, all the porcine femurs were about the same age for testing of each device in this study.

Bio-TransFix has also been compared with various implants, but not in the same study. Brown and colleagues⁸ found the TransFix device significantly stiffer than the Endobutton CL. Shen and colleagues¹⁶ determined that TransFix had significantly lower residual displacement compared with Endobutton CL. Milano and colleagues¹⁵ compared multiple cortical suspensory fixation devices, including Endobutton CL, with TransFix and Bio-TransFix, and concluded the cortical-cancellous devices (TransFix, Bio-TransFix) offered the best and most predictable results in terms of elongation, fixation strength, and stiffness. TransFix has also been shown to be superior to interference screw fixation in biomechanical studies.^10,15

Clinical outcomes of studies using TransFix for femoral fixation have been favorable, with improved Lysholm scores and improved laxity according to the KT-1000 test.¹⁷ However, multiple prospective studies have found no clinical difference in knee laxity between interference screw and Endobutton at 1- to 2-year follow-up^18-20 and no difference in clinical outcome scores, such as the International Knee Documentation Committee score.^11,18-20

Although these studies have shown no major clinical differences at short-term follow-up, the early aggressive rehabilitation period is the larger concern. Our study clearly demonstrated the biomechanical strength of transcondylar devices over other devices. The concern with transcondylar devices (vs other devices) is the increased difficulty that inexperienced surgeons have inserting them. In addition, when removed, transcondylar devices leave a large bone void.

In the present study, an important concern with femoral graft fixation is the poor performance of interference screws. Other authors recently expressed concern with using interference screws in soft-tissue ACL grafts—based on biomechanical study results of increased slippage, bone tunnel widening, and less strength.⁷ In the present study, Delta screws consistently performed poorest with respect to ultimate LTF, residual displacement, and stiffness. Only 20% of these screws completed 1500 cycles. Poor performance of interference screws has also been seen in other studies in tibial graft fixation^21,22 and femoral graft fixation.^13-15 Given their poor biomechanical properties, as seen in our study and these other studies, we think use of an interference screw alone is a poor choice for fixation.

Combined fixation techniques—interference screw plus other device(s)—may be used in clinical practice, but the present study did not evaluate any. In a biomechanical study, Yoo and colleagues²³ compared an interference screw; an interference screw plus a cortical screw and a spiked washer; and a cortical screw and a spiked washer used alone in the tibia. Stiffness nearly doubled, residual displacement was less, and ultimate LTF was significantly higher in the group with the interference screw plus the cortical screw and the spiked washer. In a similar study involving femoral fixation, Oh and colleagues²⁴ demonstrated improved stiffness, residual displacement, and LTF in cyclic testing with the combination of interference screw and Endobutton CL, compared with Endobutton CL alone. Further studies may include direct comparisons of additional femoral fixation techniques using more than 1 device.

The Zip Loop, or Toggle Loc with Zip Loop technology, is a suspensory cortical fixation device. It was initially designed for use in ACL fixation but has also been used in other surgeries, including distal biceps repair²⁵ and ulnar collateral ligament reconstruction.²⁶ The device itself is easy to use; more important, it allows for adjustment of graft length within the bone tunnel after deployment of the cortical fixation. Few biomechanical studies have been conducted with Zip Loop.^9,12 The present study is the first to compare Zip Loop with devices other than suspensory cortical fixation devices. Zip Loop performed very well in LTF testing but had lower stiffness and higher residual displacement compared with the transcondylar fixation devices. Despite these findings, we have continued to use this device for femoral fixation in ACL reconstruction because of its ease of insertion, the ability to adjust graft tension within the bone tunnel, and the difficulties encountered inserting and removing transcondylar fixation.

We recognize the limitations in our study design with respect to how axial and cyclical loading compares with the physiologic orientation of the ACL during ambulation and running activities. This biomechanical study was not able to replicate these types of activities. However, it did provide good data supporting early rehabilitation with various fixation devices, though concern with use of interference screws remains.

Conclusion

Superior strength in fixation of hamstring grafts in the femur was demonstrated by Bone Mulch screws, followed closely by Bio-TransFix. Delta screws demonstrated poor displacement, stiffness, and LTF. When used as the sole femoral fixation device, a device with low LTF, decreased stiffness, and high residual displacement should be used cautiously in patients undergoing aggressive rehabilitation.