Figure. Kaplain-Meier survival plot of primary total hip arthroplasty survival by sex. Log rank P = .005. Shaded areas indicate 95% CIs.
Inacio MCS, Ake CF, Paxton EW, Khatod M, Wang C, Gross TP, Kaczmarek RG, Marinac-Dabic D, Sedrakyan A. Sex and Risk of Hip Implant FailureAssessing Total Hip Arthroplasty Outcomes in the United States. JAMA Intern Med. 2013;173(6):435-441. doi:10.1001/jamainternmed.2013.3271
Author Affiliations: Surgical Outcomes and Analysis Department, Southern California Permanente Medical Group, San Diego (Mss Inacio and Paxton and Dr Ake); Department of Orthopaedic Surgery, Southern California Permanente Medical Group, West Los Angeles (Dr Khatod); Office of Surveillance and Biometrics, Center for Devices and Radiological Health, US Food and Drug Administration, Silver Spring, Maryland (Drs Wang, Gross, Kaczmarek, and Marinac-Dabic); and Weill Medical College of Cornell University, New York, New York (Dr Sedrakyan).
Importance The role of sex in relationship to implant failure after total hip arthroplasty (THA) is important for patient management and device innovation.
Objective To evaluate the association of sex with short-term risk of THA revision after adjusting for patient, implant, surgery, surgeon, and hospital confounders.
Design and Setting A prospective cohort of patients enrolled in a total joint replacement registry from April 1, 2001, through December 31, 2010.
Participants Patients undergoing primary, elective, unilateral THA.
Main Outcome Measures Failure of THA, defined as revision procedure for (1) any reason, (2) septic reason, or (3) aseptic reason after the index procedure.
Results A total of 35 140 THAs with 3.0 years of median follow-up were identified. Women constituted 57.5% of the study sample, and the mean (SD) patient age was 65.7 (11.6) years. A higher proportion of women received 28-mm femoral heads (28.2% vs 13.1%) and had metal on highly cross-linked polyethylene-bearing surfaces (60.6% vs 53.7%) than men. Men had a higher proportion of 36-mm or larger heads (55.4% vs 32.8%) and metal on metal-bearing surfaces (19.4% vs 9.6%). At 5-year follow-up, implant survival was 97.4% (95% CI, 97.2%-97.6%). Device survival for men (97.7%; 95% CI, 97.4%-98.0%) vs women (97.1%; 95% CI, 96.8%-97.4%) was significantly different (P = .01). After adjustments, the hazards ratios for women were 1.29 (95% CI, 1.11-1.51) for all-cause revision, 1.32 (95% CI, 1.10-1.58) for aseptic revision, and 1.17 (95% CI, 0.81-1.68) for septic revision.
Conclusions After considering patient-, surgery-, surgeon-, volume-, and implant-specific risk factors, women had a 29% higher risk of implant failure than men after THA in this community-based sample.
Total hip arthroplasty (THA) is a successful orthopedic procedure that results in significant improvement in function, quality of life, and pain reduction for patients.1- 3 Total hip arthroplasty procedures also demonstrate excellent results at 5 to 7 years.4- 6 Despite this success, debate concerning patient-, implant-, surgeon-, procedure-, and volume-related risk factors of THA outcomes has not been fully resolved. Most importantly, the contribution of each of these risk factors is difficult to evaluate independently.
Recently, the US Food and Drug Administration provided guidance for the enrollment of women and for conducting sex-specific analyses in device studies.7 A recent “Viewpoint” published in JAMA called for transparency and enforcement of these recommendations.8 This is a pressing issue in orthopedics, a field in which sex differences are substantial and sex-specific device development is occurring.9Moreover, THA is more often performed in women than men.4,10 In other major surgical procedures,11- 13 sex-specific risk factors and outcomes have been investigated. In theory, sex differences might be more important to study in THA because of anatomical differences, such as the location of the femoral head center, size and shape of the femoral canal, and trabecular patterns.14,15 However, how anatomical sex differences influence functional outcomes and implant survivorship is unclear. Although some studies16,17 suggest men have higher perioperative complication rates and failure rates, others15,18 suggest similar failure rates and functional outcomes among men and women. A recent study19 by the National Joint Registry of England and Wales found higher occurrence of implant failure in women than men when using metal on metal (MOM) implants after adjusting for age and femoral head size.
To determine whether sex is a significant risk factor of THA, we had to account for other reported risk factors of poor outcomes related to this surgery. These other risk factors, which could confound the association of sex and failures of THA, include surgeon-, hospital-,20- 22 surgery-, and implant-related variables.23- 26
The objective of this study was to use the largest total joint replacement registry (TJRR) cohort of elective primary THA in the United States to determine whether sex is associated with short-term increased risk of revision after adjusting for potential confounders, such as patient, surgeon, hospital, surgery, and implant characteristics.
An integrated health care system's TJRR was used to identify a cohort of consecutive patients who underwent THA. Data capture mechanisms, validation processes, covered population, and participation rates of the TJRR have been previously published.4,27- 29 In brief, TJRR data are collected prospectively through standardized documentation by surgeons. These forms collect information on patient demographics, implant characteristics, surgical techniques, and outcomes (eg, revisions and subsequent operations). Registry forms are supplemented with data from electronic health records and other independent databases. Voluntary TJRR participation for 2010 was 90% and nondifferential among locations.27 Electronic screening algorithms are used to search administrative databases for additional complications, subsequent operations, and revisions. Independent electronic health record files are used to validate registry cases. All outcomes are adjudicated by trained personnel through medical record reviews following standard protocols.4,29,30
Primary elective THAs performed from April 1, 2001, through December 31, 2010, at 46 hospitals by 319 surgeons in California, Hawaii, Northwest, and Colorado with implants in 6 bearing-surface categories, namely, metal on highly cross-linked polyethylene (XLPE), ceramic on XLPE, MOM, metal on conventional ultrahigh-molecular-weight polyethylene (UHMWPE), ceramic on ceramic (COC), and ceramic on conventional UHMWPE, were included in the study. Patients undergoing same-day bilateral THAs and/or resurfacing were excluded from the study.
All data elements were extracted from the TJRR. The outcome variable was revision surgery defined as removal or exchange of at least one prosthetic component. Revision for all reasons, aseptic (noninfectious) reasons, and septic reasons were evaluated separately. The exposure of interest of this study was sex, with male used as the referent category. Patient covariates consisted of age, American Society of Anesthesiologists score (<3 vs ≥3), body mass index, diabetes status, primary diagnosis (osteoarthritis vs other), and race. Hospital and surgeon covariates included THA yearly mean volume (both primary and revision THAs were included in the computation) and surgeon total joint arthroplasty fellowship training (yes/no). Annual hospital case volume was categorized as fewer than 75 cases per year vs 75 cases per year or more. Annual surgeon volume was categorized into fewer than 30 cases per year vs 30 cases per year or more. The THA implant covariates included implant fixation (cemented, uncemented, or hybrid), femoral head size (<36 vs ≥36 mm), bearing surfaces (metal on XLPE, ceramic on XLPE, MOM, metal on conventional UHMWPE, COC, and ceramic on conventional UHMWPE), and whether a recalled Depuy ASR MOM monoblock system (Depuy Orthopaedics Inc) was used in the procedure (yes/no).31
Descriptive statistics, including means, SDs, frequencies, and proportions, were used to characterize the study sample, implants used in the procedures, surgeons, and hospitals. The χ2, Fisher exact, and independent t tests were applied to evaluate univariate sex differences in patient demographics, diagnosis, health status (American Society of Anesthesiologist score), anthropometric measures, implant characteristics, and surgeon and hospital characteristics. Crude revision rates of all-cause, septic, and aseptic revision THA were calculated with 95% CIs for the overall group and by sex. Revision rates per 100 years of follow-up were also calculated with 95% CIs. Kaplan-Meier survival curves with log-rank tests using revision as the end point were used to evaluate implant survival for the overall group and by sex. Before the end of the study period, cases could be censored by death or termination of membership with the integrated health care system. Multiple imputations (N = 10) were used to accommodate missing values. Cox proportional hazard regression models were used to assess the risk of all-cause revision, aseptic revision, and septic revision by sex (male was the reference group). Hazard ratios (HRs) and 95% CIs are reported. All-cause revision and aseptic revision models were built similarly, but some septic revision models included fewer covariates because of fewer events. Interactions between sex and femoral head size, as well as sex and bearing surface, were tested. Subgroup analyses exploring the association of sex and revision were performed in high-risk groups, such as MOM THAs and THAs with small femoral heads. The sandwich covariance estimator was used to handle clustering by surgeon. Sensitivity analyses were conducted to determine the effect of loss to follow-up due to mortality and, separately, due to membership attrition. SAS statistical software (versions 9.1.3 and 9.2, SAS Institute Inc) was used to analyze the data. Institutional review board approval by the institution was granted before study commencement.
During the study period, 35 140 primary THA cases were performed. The study contained 20 219 women (57.5%), and the mean (SD) age of the cohort was 65.7 (11.6) years. Osteoarthritis was the most common diagnosis (n = 32105 [91.4%]) (Table 1). The women were older than the men (67.1 vs 63.8 years, P < .001). A total of 4950 patients (14.1%) were either lost to follow-up or died during the study period (2209 men [14.8%] and 2741 women [13.6%]). Of the 1829 patients who died (5.2%), 823 (5.5%) were male and 1006 (5.0%) were female. The median follow-up of the cohort was 3.0 years (interquartile range, 1.3-5.1 years).
Women received a 28-mm or smaller femoral head more often (28.5% vs 13.1%) and a 36-mm or greater head (32.8% vs 55.4%) less often than men (P < .001). Women also had a higher proportion of cases with metal on XLPE implants (60.6% vs 53.7%) and a lower proportion of cases with MOM bearings (9.6% vs 19.4%) than men (P < .001). Hybrid (14.0% vs 7.5%) rather than cementless (77.8% vs 84.9%) fixation was also used more often in women (P < .001). Differences in the proportion of surgeons with fellowship training, as well as surgeon and hospital volume, were small when comparing the care received by men and women (Table 2).
The all-cause crude revision rate was 2.3% (95% CI, 2.1%-2.5%) for women and 1.9% (95% CI, 1.6%-2.1%) for men. Women had a higher aseptic revision rate (1.9%; 95% CI, 1.7%-2.1%) than men (1.5%; 95% CI, 1.3%-1.7%) and the same septic revision rate (0.4%; 95% CI, 0.3%-0.5%). At 5-year follow-up, the unadjusted cumulative implant survival was 97.4% (95% CI, 97.2%-97.6%). The THA survival for men and women was 97.7% (95% CI, 97.4%-98.0%) and 97.1% (95% CI, 96.8%-97.4%) (P = .005), respectively (Figure). Leading reasons for all-cause revisions were instability, infection, aseptic loosening, and periprosthetic fracture (Table 3).
In the unadjusted all-cause revision model (Table 4, model 1), women had an HR of 1.23 (95% CI, 1.06-1.44) compared with men. After adjusting for patient characteristics (model 2), the HR increased to 1.30 (95% CI, 1.11-1.53). Further adjustment for hospital volume (model 3), then surgeon training and surgeon volume (model 4), and additionally implant fixation (model 5) leaves the HR unchanged. Adding femoral head size, bearing surface, and whether an ASR implant was used in the model (model 6) had a minimal effect (HR, 1.29; 95% CI, 1.11-1.51). After adjustment for confounders (model 6), the association of sex and risk of aseptic revision was 1.32 (95% CI, 1.10-1.58), and the risk of septic revision was 1.17 (95% CI, 0.81-1.68) (Table 4).
There was an interaction of sex with femoral head size in a version of model 6 for all-cause revision. In the femoral head size group of 28 mm or less, the HR for women compared with men was 1.15 (95% CI, 0.86-1.55); in the femoral head size group of greater than 28 mm and less than 36 mm, the HR was 1.17 (95% CI, 0.90-1.53); and in the femoral head size group of 36 mm or greater, the HR was 1.49 (95% CI, 1.14-1.95). In a subgroup of THAs with femoral head sizes of 36 mm or less, we found an HR of 1.19 (95% CI, 1.00-1.42) for women when compared with men. Similarly, a statistically significant interaction for sex and the MOM vs metal on XLPE term was observed (model 6). In the model with the interaction term, the HR of all-cause revision for MOM compared with the metal on XLPE was 0.68 (95% CI, 0.45-1.02) for men and 1.07 (95% CI, 0.72-1.60) for women. In a subgroup analysis of THAs with MOM bearings, we found an HR of 1.97 (95% CI, 1.29-3.00) for women when compared with men for all-cause revision.
To examine the effect of loss to follow-up, a randomly distributed 2%, 5%, and 10% of cases lost to follow-up were considered to have the outcome event on the date they were censored. Estimates were stable across scenarios (data not shown).
In our analyses of a large THA cohort, including a diverse sample within 46 hospitals, we found that at the median follow-up of 3.0 years women have a higher risk of all-cause (HR, 1.29; 95% CI, 1.11-1.51) and aseptic (HR, 1.32; 95% CI, 1.10-1.58) revision but not septic revision (HR, 1.17; 95% CI, 0.81-1.68). We clarified the effect of multiple confounders, such as patient, implant, hospital, and surgeon factors on the association of sex and THA risk of revision.
Prior evidence is inconsistent. A recent study19 from the National Joint Registry of England and Wales reports higher occurrence of revision in women, particularly when MOM articulations are used. Earlier studies32 have reported that men are at a greater risk of failures, but most of these studies were published in the late 1980s and early 1990s. More recent studies, such as the one by Röder et al,17 report that women had lower early cup failure than men independent of cup fixation. Conversely, Howard et al33 reported a protective association of male sex and the risk of cup revision for any reason in a single-center study. One older publication from a Scandinavian registry,18 a study based on an elderly cohort,6and a large single-center series15 did not find significant sex differences in THA revision rates. Some conflicting findings among these studies may be attributed to the representativeness, differential definitions of revision, different follow-up times, type of analysis performed, and different mechanisms of identifying outcomes. For example, although the revision definition in the study by Howard et al33 was similar to our definition, Röder et al17 defined failure as revision and radiographic signs of failure. The most recent study on the topic, by Smith et al,19 used death as a competing risk in their modeling and excluded subclasses of higher-risk cases (those with a nonosteoarthritis diagnosis and an American Society of Anesthesiologists score ≥3), resulting in a different type of analysis from what we performed.
We did not find a statistically significant association between sex and risk of septic revision, which does not support a recent report from the Norwegian TJA registry34,35 that found an increased risk (2.4 times higher) for men. Again, the intercountry definitions of the outcomes and mode of data collection are different and important to consider. Although the Norwegian registry relies on surgeon-reported revisions for infection alone, our TJRR actively monitors its registered cohort. This active surveillance includes quarterly review of all primary procedures and incidence of lower-extremity operations using both the registry forms and electronic health records of the institution, followed by manual review of cases, ensuring high internal validity.
Certain types of implants and their attributes have been recently receiving attention because of a reportedly high risk of revisions and other complications (ie, MOM articulations and small femoral head sizes). Our subgroup analyses support and advance recent findings related to high occurrence of failure in women receiving MOM-bearing surfaces.19,36 Smith et al19 analyzed 30 000 cases, using the National Joint Registry of England and Wales,and reported a revision rate of 5.1% in women and 3.7% in men at 5 years. Similarly, in a smaller US study of 1589 THAs with MOM bearings, women had a 2-year revision rate of 8.2% compared with 2.7% in men. We found that women have an almost 2 times higher risk of revision when compared with men.
We also explored the relationship of sex and femoral head size on risk of revision. There are articles26,37 that report smaller femoral head sizes are associated with higher risk of dislocation, the most common THA complication. Indeed, in our study the leading cause for revision is instability. The size of the implants is dictated by what the pelvis and acetabulum of a patient can accommodate, meaning that smaller bone structures will not accommodate larger implants, which can be used to reduce the risk of dislocation and possible revision. In our study we were able to determine that among those who received smaller femoral head sizes, women continue to have a 19% higher risk of revision than men. Because women are more likely to receive these smaller femoral head sizes, they might have even greater risk of experiencing revision.
Our study limitations include its observational design and short-term follow-up. Because of the observational nature of this study, there may be residual confounding. However, we adjusted for all known confounders captured within the TJRR. The TJRR does not capture patient-reported functional outcomes and radiographic assessment of patients at this time, so this information was unavailable, and revision surgery was used as the end point. Although our definition of failure could miss early indications of failures, we believe it represents the accurate association of sex and risk of revision procedures. The study did not adjust for differences in specific implant designs. Further, there seems to be a relationship with the use of smaller femoral head size and female sex. We attempted to isolate the effect of sex by adjusting for this confounding variable, evaluating the possible interaction in the risk of revision, and evaluating the relationship for restricted samples. We did this in an attempt to evaluate sex as a factor for revision surgery for equivalent-sized women and men in whom similar femoral head sizes could be chosen. Finally, the attrition rate of our cohort could affect our estimations; we have addressed this by conducting sensitivity analyses covering several scenarios.
The strengths of this study are the use of the largest US TJRR cohort, detailed information on patient and implant characteristics, the active surveillance mechanism to ascertain outcomes used by the TJRR, and the community-based sample. With more than 35 000 primary THAs and 743 failures, we were able to investigate the effect of numerous variables in the relationship between sex and risk of revision. The ability to conduct this type of adjustment in orthopedic studies is typically difficult in single-surgeon or single-center studies. We also were able to investigate variables that studies dependent on administrative data sources could not address (ie, body mass index, complete diabetes information, and detailed implant descriptions). The TJRR active surveillance mechanism of outcomes is also a strength, increasing our findings' internal validity. Finally, the integrated health care system membership population, the sampling frame of the TJRR, has been reported to be of similar age, sex, and racial distribution to the overall population in the major geographic areas covered by the TJRR.38- 40 This, we believe, extends the external validity of our findings to the larger US THA population.
Overall the findings of this study suggest that women have a 29% higher risk of short-term implant failure following THA after considering patient-, surgery-, surgeon-, volume-, and implant-specific risk factors compared with men. The differences in prosthesis choices in men and women, as well as the follow-up of the presented study and nonmeasured possible confounders, are important considerations when interpreting these results. Finally, the increased risk of all-cause implant revision appears to be related to factors other than infection.
Correspondence: Maria C. S. Inacio, MS, Surgical Outcomes and Analysis Department, Southern California Permanente Medical Group, 8954 Rio San Diego Dr, Ste 406, San Diego, CA 92108 (email@example.com).
Accepted for Publication: November 14, 2012.
Published Online: February 18, 2013. doi:10.1001/jamainternmed.2013.3271
Author Contributions: Dr Ake and Mss Inacio and Paxton had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Paxton, Khatod, Wang, Marinac-Dabic, Gross, Kaczmarek, and Sedrakyan. Acquisition of data: Inacio and Paxton. Analysis and interpretation of the data: Inacio, Ake, Paxton, Khatod, Wang, Marinac-Dabic, Gross, Kaczmarek, and Sedrakyan. Drafting of the manuscript: Inacio and Ake. Critical revision of the manuscript for important intellectual content: Paxton, Khatod, Wang, Marinac-Dabic, Gross, Kaczmarek, and Sedrakyan. Statistical analysis: Ake. Obtaining funding: Paxton.Administrative, technical and material support: Inacio. Study supervision: Inacio, Paxton, Khatod, Wang, Marinac-Dabic, Gross, Kaczmarek, and Sedrakyan.
Conflict of Interest Disclosures: Alan L. Schepps, MS, is employed by the Surgical Outcomes and Analysis Department and received no additional compensation for contributing to the manuscript.
Funding/Support: This study was funded by contract HHSF22200860493P from the Division of Epidemiology, Office of Surveillance and Biometrics, Center for Devices and Radiological Health, US Food and Drug Administration, Silver Spring, Maryland.
Role of the Sponsors: The Division of Epidemiology of the US Food and Drug Administration participated in the design and conduct of the study, interpretation of the data, and review and approval of the manuscript.
Additional Contributions: Alan L. Schepps, MS, provided support with the statistical data programming, tables, and graphs prepared for the manuscript. We acknowledge all the Kaiser Permanente orthopaedic surgeons who contribute to the TJRR and the Surgical Outcomes and Analysis Department, which coordinates registry operations.