Effect of hormone therapy on risk of hip and knee joint replacement in the women's health initiative.код для вставкиСкачать
ARTHRITIS & RHEUMATISM Vol. 54, No. 10, October 2006, pp 3194–3204 DOI 10.1002/art.22138 © 2006, American College of Rheumatology Effect of Hormone Therapy on Risk of Hip and Knee Joint Replacement in the Women’s Health Initiative Dominic J. Cirillo,1 Robert B. Wallace,2 LieLing Wu,3 and Robert A. Yood4 arthroplasty (HR 0.84 [95% CI 0.70–1.00], P ⴝ 0.05). However, this effect was borderline statistically significant for hip arthroplasty (HR 0.73 [95% CI 0.52–1.03], P ⴝ 0.07), and not significant for knee arthroplasty (HR 0.87 [95% CI 0.71–1.07], P ⴝ 0.19). In the estrogen-plusprogestin trial, there was no association for total arthroplasty (HR 0.99 [95% CI 0.82–1.20], P ⴝ 0.92) or for individual hip (HR 1.14 [95% CI 0.83–1.57], P ⴝ 0.41) or knee (HR 0.91 [95% CI 0.72–1.15], P ⴝ 0.41) arthroplasties. Conclusion. These data suggest that hormone therapy may influence joint health, but this observed decrease in risk may be limited to unopposed estrogen and may possibly be more important in hip than in knee osteoarthritis. Objective. To determine the effect of hormone therapy on arthroplasty rates. Methods. We examined data from the Women’s Health Initiative placebo-controlled, double-blind, randomized trials. Community-dwelling women ages 50–79 years were enrolled at 40 US clinics. Women with prior arthroplasty were excluded, yielding a sample size of 26,321 subjects. Women who had had hysterectomies (n ⴝ 10,272) were randomly assigned to receive 0.625 mg/day conjugated equine estrogens (n ⴝ 5,076), or placebo (n ⴝ 5,196), with a mean followup of 7.1 years. Those who had not had hysterectomies (n ⴝ 16,049) were randomly assigned to receive estrogen plus progestin (n ⴝ 8,240), given as 0.625 mg/day conjugated equine estrogens plus 2.5 mg/day medroxyprogesterone acetate, or placebo (n ⴝ 7,809), with a mean followup of 5.6 years. Participants reported hospitalizations, and arthroplasties were identified by procedure codes. Arthroplasties due to hip fracture were censored. Cox proportional hazards regression was used to assess hazard ratios (HRs) and 95% confidence intervals (95% CIs) using intent-to-treat methods and outcome of time to first procedure. Results. In the estrogen-alone trial, women receiving hormone therapy had significantly lower rates of any Osteoarthritis (OA) affects approximately 1 in 8 Americans over age 25 years and is the leading chronic medical condition among the elderly (1), reaching 80% prevalence by age 75 years (2). Costs related to OA include medication, direct and indirect costs of disability, and surgical procedures performed when conservative management fails to control pain and dysfunction (1,3,4). In 1999 in the US, there were ⬃249,000 total knee replacement surgeries for OA, and another 200,000 patients had total hip replacements (5). The literature on hormone therapy and OA onset and progression is complex due to varying definitions of disease, different classifications of hormone therapy use, joint differences, and study design effects. In addition, many studies have not differentiated the possible effects of various types of hormone therapy, such as unopposed estrogen or estrogen plus progestin. Hormone therapy has been shown to help relieve menopausal symptoms, including joint pain (6). Some cross-sectional studies suggest that current hormone therapy use protects against radiographic evidence of OA (7,8). Cohort studies of radiographic OA have tended to show nonsignif- The Women’s Health Initiative is supported by the NIH (National Heart, Lung, and Blood Institute) and by contract N01-WH2-2110 from the Department of Health and Human Services. 1 Dominic J. Cirillo, BS: University of Iowa College of Public Health, Iowa City; 2Robert B. Wallace, MD, MSc: University of Iowa College of Public Health, and University of Iowa Carver College of Medicine, Iowa City; 3LieLing Wu, MS: Fred Hutchinson Cancer Research Center, Seattle, Washington; 4Robert A. Yood, MD: Fallon Clinic, Worcester, Massachusetts, and University of Massachusetts Medical School, Worcester. Address correspondence and reprint requests to Robert B. Wallace, MD, MSc, Department of Epidemiology C21-N GH, University of Iowa College of Public Health, 200 Hawkins Drive, Iowa City, IA 52242. E-mail: email@example.com. Submitted for publication March 15, 2006; accepted in revised form June 28, 2006. 3194 EFFECT OF ESTROGEN THERAPY ON JOINT REPLACEMENT icant protective effects (9–11). However, since radiographic and self-reported diagnoses may not reflect disabling disease, arthroplasty may be considered a viable proxy outcome for severe OA (12). Studies examining risk of joint replacement due to OA have demonstrated increased risk (13), no association (14), a protective effect of hormone therapy (15), or mixed results depending on length of use (16). These reports do not form a cohesive picture of the potential effect of hormone therapy on OA or arthroplasty. In this report, we present findings from the Women’s Health Initiative (WHI), which included 2 randomized hormone trials based on participant hysterectomy status. A partial list of WHI investigators is shown in Appendix A; a complete list is shown on the WHI Web site (http://www.whi.org/about/ investigators.php). Women who had had a hysterectomy were randomized to receive unopposed estrogen or placebo, and those whose uteri were intact received a combination of estrogen plus progestin or placebo (17). The major findings of these trials were previously reported (18,19). Although hip and knee arthroplasty were not primary end points, the WHI is the largest randomized, double-blind trial of postmenopausal hormone therapy to date and provides an important opportunity to examine the effect of hormone therapy on the need for arthroplastic procedures among healthy, communitydwelling postmenopausal women. PATIENTS AND METHODS Role of the funding source. The National Heart, Lung, and Blood Institute did not participate in the design and conduct of the study, in the collection, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript. Participants. The WHI consisted of randomized trials of hormone therapy as well as dietary modification and calcium/vitamin D supplementation. Detailed discussions of the recruitment methods and the principal findings of the WHI trials are available elsewhere (17,19). Briefly, women who were between the ages of 50 and 79 years at an initial screening visit to 1 of 40 clinical sites were assessed for eligibility and provided written informed consent to enroll in the hormone therapy trial. The National Institutes of Health (NIH) and Institutional Review Boards for all participating centers approved the WHI protocols and consent forms. The screening and enrollment algorithm is shown in Figure 1. Women were excluded from the hormone trials if they had a terminal illness or a contraindication to hormone therapy, or if they were considered at high risk of poor adherence. Furthermore, in this secondary analysis, we excluded women with a history of arthroplasty at baseline. Women who had had a hysterectomy were eligible for the estrogen-alone trial, and women who had not had a hysterec- 3195 Figure 1. Profile of the hormone therapy trials of the Women’s Health Initiative. tomy were enrolled in the estrogen-plus-progestin trial. Eligible women who agreed to participate were randomly assigned in equal proportions to receive hormone therapy or placebo using a stratified permuted block algorithm (17,20). For women currently receiving hormone therapy at study enrollment, a 3-month washout period was required before randomization assignment. The active study drug and placebo were kindly supplied by Wyeth (St. Davids, PA). Women in the estrogen-alone trial received either placebo or 0.625 mg/day conjugated equine estrogens (Premarin; Wyeth), using blinded dispensing. Women in the estrogen-plus-progestin trial received either placebo or 0.625 mg/day conjugated equine estrogens plus 2.5 mg/day medroxyprogesterone acetate, administered as a single tablet (Prempro; Wyeth). In order to reinforce adherence and to assess side effects, followup telephone calls were conducted 6 weeks postrandomization. If the participant developed side effects from the treatment, the dosage or frequency could be altered by study physicians, in most instances without unblinding the subject or the study clinicians to the treatment assignment. Infrequently, an unblinding officer could disclose the treatment assignment to study clinicians for symptom management or safety concerns. Study clinicians were unblinded at rates of 2% in each group in the estrogen-alone trial and at rates of 41% and 6%, respectively, in the estrogen-plus-progestin trial active and placebo groups; however, these clinicians were not involved in the assessment of study outcomes. Data collection and outcomes. Baseline measurements were collected in a standardized manner, using extensive medical history questionnaires, self-administered forms, interviews, and clinical examinations. Weight and height were measured using a calibrated balance beam scale and a wallmounted stadiometer, respectively. Body mass index (BMI) was calculated as weight (in kilograms) divided by height (in meters squared). Race and ethnicity were self reported using predefined categories, to assess treatment effect differences in subgroup analyses. Other clinical, demographic, social, and behavioral baseline characteristics were determined from selfadministered, structured questionnaires (17). Subjects were asked about their medical history (including arthroplasty) as 3196 CIRILLO ET AL Table 1. Baseline descriptive characteristics of the Women’s Health Initiative hormone therapy trials* Estrogen-alone trial Age group at screening, years 50–59 60–69 70–79 Ethnicity White Black Hispanic American Indian Asian/Pacific Islander Other/unspecified Education 0–8 years Some high school High school/GED School after high school College degree or higher Time since quitting estrogen Nonuser Past, ⬍5 years Past, 5 to ⬍10 years Past, ⱖ10 years Current† BMI, kg/m2 ⬍18.5 18.5–24.9 ⱖ25.0 Baseline joint pain/stiffness None Mild Moderate Severe NSAID use at baseline Aspirin use at baseline Estrogen-plus-progestin trial Hormone therapy (n ⫽ 5,076) Placebo (n ⫽ 5,196) Hormone therapy (n ⫽ 8,240) Placebo (n ⫽ 7,809) 1,611 (31.7) 2,287 (45.1) 1,178 (23.2) 1,650 (31.8) 2,368 (45.6) 1,178 (22.7) 2,815 (34.2) 3,734 (45.3) 1,691 (20.5) 2,652 (34.0) 3,522 (45.1) 1,635 (20.9) 3,817 (75.2) 751 (14.8) 315 (6.2) 40 (0.8) 84 (1.7) 69 (1.4) 3,889 (74.8) 802 (15.4) 322 (6.2) 34 (0.7) 76 (1.5) 73 (1.4) 6,901 (83.8) 535 (6.5) 466 (5.7) 26 (0.3) 190 (2.3) 122 (1.5) 6,532 (83.6) 563 (7.2) 411 (5.3) 29 (0.4) 169 (2.2) 105 (1.3) 176 (3.5) 330 (6.5) 1,184 (23.3) 2,179 (42.9) 1,154 (22.7) 140 (2.7) 344 (6.6) 1,136 (21.9) 2,247 (43.2) 1,284 (24.7) 195 (2.4) 356 (4.3) 1,562 (19.0) 3,250 (39.4) 2,831 (34.4) 172 (2.2) 347 (4.4) 1,547 (19.8) 2,942 (37.7) 2,743 (35.1) 2,639 (52.0) 547 (10.8) 247 (4.9) 994 (19.6) 648 (12.8) 2,636 (50.7) 564 (10.9) 263 (5.1) 1,046 (20.1) 684 (13.2) 6,073 (73.7) 714 (8.7) 325 (3.9) 584 (7.1) 540 (6.6) 5,794 (74.2) 662 (8.5) 297 (3.8) 569 (7.3) 484 (6.2) 1,080 (21.3) 1,727 (34.0) 2,241 (44.1) 1,073 (20.7) 1,840 (35.4) 2,248 (43.3) 2,526 (30.7) 2,912 (35.3) 2,768 (33.6) 2,416 (30.9) 2,759 (35.3) 2,585 (33.1) 1,178 (23.2) 2,301 (45.3) 1,151 (22.7) 386 (7.6) 1,777 (35.0) 1,044 (20.6) 1,234 (23.7) 2,343 (45.1) 1,193 (23.0) 375 (7.2) 1,846 (35.5) 1,064 (20.5) 2,437 (29.6) 3,951 (47.9) 1,428 (17.3) 360 (4.4) 2,725 (33.1) 1,668 (20.2) 2,400 (30.7) 3,663 (46.9) 1,330 (17.0) 343 (4.4) 2,650 (33.9) 1,632 (20.9) * Values are the number (%) of patients. GED ⫽ general educational development; BMI ⫽ body mass index; NSAID ⫽ nonsteroidal antiinflammatory drug. † A 3-month washout period was required prior to randomization. well as current symptoms (including joint pain/stiffness, as “mild,” “moderate,” “severe,” or “did not occur”). Main study outcomes. Clinical outcomes were ascertained semiannually via questionnaires and annual clinic visits. The semiannual followup included standardized questionnaires on specific symptoms and safety concerns, as well as major health events, including the primary outcomes of heart disease, stroke, hip fracture, and breast cancer as well as hospitalizations and significant medical procedures. Participants provided consent to have their medical and hospital records reviewed by study physicians. The completion rate for clinical record acquisition and adjudication was estimated to be 96%. Outcomes used in this analysis were obtained from medical records and identified with International Classification of Diseases, Ninth Revision (ICD-9) codes. Total hip replacement (ICD-9 code 81.51) and total knee replacement (ICD-9 code 81.54) were analyzed separately, and an index of either hip or knee arthroplasty was also considered. Hip and femoral neck fractures, which were primary outcomes in the hormone therapy trials, were ascertained in a similar manner, and women were censored at the date of fracture during the followup period (n ⫽ 340). As described elsewhere in detail (19), the estrogen-plus-progestin trial was discontinued earlier than scheduled after a recommendation from the Data and Safety Monitoring Board that the overall harms outweighed the benefits of estrogen plus progestin. Close-out of the estrogen-alone clinical trial was originally planned for October 2004 through March 2005; early study termination (February 2004) resulted in a followup time ⬃1 year shorter than planned (18). Statistical analysis. All statistical analyses were conducted using the SAS System for Windows, version 9.0 (SAS Institute, Cary, NC). The time to the first arthroplasty event was the primary analytical outcome. Associations between baseline variables and randomization assignment were assessed using P values from chi-square tests for categorical variables and 2-sample t-tests for continuous variables. All outcomes analyses used time-to-event methods and were based EFFECT OF ESTROGEN THERAPY ON JOINT REPLACEMENT 3197 Table 2. Incidence of arthroplasty by randomization assignment in the unopposed-estrogen (estrogen-alone) trial of the Women’s Health Initiative* Followup time, mean ⫾ SD years Total joint replacement‡ Total hip replacement Total knee replacement Censoring after nonadherence§ Total joint replacement‡ Total hip replacement Total knee replacement Estrogen alone (n ⫽ 5,076) Placebo (n ⫽ 5,196) HR (95% CI)† P† 7.1 ⫾ 1.6 222 (62) 58 (16) 169 (47) 7.1 ⫾ 1.7 269 (74) 81 (22) 198 (54) 0.84 (0.70–1.00) 0.73 (0.52–1.03) 0.87 (0.71–1.07) 0.05 0.07 0.19 119 (56) 28 (13) 93 (44) 169 (78) 53 (24) 121 (56) 0.73 (0.58–0.93) 0.55 (0.35–0.88) 0.80 (0.61–1.05) 0.01 0.01 0.11 * Except where indicated otherwise, values are the number of patients undergoing arthroplasty (rate per 10,000 person-years). † Hazard ratios (HRs), 95% confidence intervals (95% CIs), and P values from Cox proportional hazards analyses stratified by age. ‡ Events include any hip or knee arthroplasty. § Analyzed per-protocol, with followup time of participants censored at 6 months post-nonadherence, which was predefined as consuming ⬍80% of assigned study pills. on the intent-to-treat (ITT) principle. Outcome comparisons are presented as hazard ratios (HRs) and 95% confidence intervals (95% CIs) from Cox proportional hazards models, stratified by age. Random assignment of subjects to calcium/ vitamin D supplementation occurred 1–2 years after randomization in the hormone trial, so the calcium/vitamin D randomization date was treated as the only time-dependent covariate in the analysis. This adjustment did not appreciably change the results, and therefore the findings are not presented. The proportionality assumption was tested by including a time– treatment interaction term in the Cox model, in addition to the treatment assignment indicator. None of the P values was significant at the 5% level, which indicated no evidence of departure from proportionality. Absolute risk reduction was calculated as the difference in incidence rates between groups. This study was a secondary analysis of a large trial that was designed and sufficiently powered to detect clinically relevant changes in the primary outcomes of coronary heart disease, hip fracture, and breast cancer, but the study was not specifically powered to detect differences in arthroplasty rates. Kaplan-Meier plots illustrated the risk of arthroplasty over the followup period, and log rank test statistics were calculated to test for differences between treatment assignment curves. Sensitivity analyses were conducted to examine the effect of nonadherence, by repeating these analyses and censoring followup 6 months after a woman ceased to adhere to the assigned treatment, which was prospectively defined either as consuming ⬍80% of study pills as determined by pill count or as starting non–study prescribed hormone therapy. RESULTS Baseline subject characteristics for the entire study cohort were presented elsewhere (19,21). The primary reports from these trials described the results for 5,310 women in the conjugated equine estrogens group and 5,429 women in the placebo group for the estrogen-alone trial (18), and for 8,506 women receiving estrogen plus progestin and 8,102 women receiving placebo in the estrogen-plus-progestin trial (19). For the present analysis, the study population included these women, but exclusions were made if the participant had a history of arthroplasty (n ⫽ 1,026; 353 with hip replacement, 598 with other joint arthroplasty, and 75 with both conditions). The final sample size for this study was 26,321 (10,272 in the estrogen-alone trial and 16,049 in the estrogen-plus-progestin trial) (Figure 1). The proportion of women who were excluded from the analysis based on history of arthroplasty was slightly higher for the estrogen-alone trial compared with the estrogen-plusprogestin trial (4.3% versus 3.4%; P ⬍ 0.001), but these subjects were balanced between treatment groups within each trial. The treatment and control groups were also balanced on key baseline demographic and disease risk factors (Table 1). The only variables that were significantly different were found in the estrogen-alone trial, in which the placebo group subjects had attained a significantly higher educational level (P ⫽ 0.02) and were more likely to have a history of cancer (P ⬍ 0.01), although the difference between the groups was small, and statistical significance was driven by the large number of women enrolled in the trial. The remaining variables were not significantly different (P ⬎ 0.10), and no significant differences at baseline were found in the estrogen-plus-progestin trial. Forty-nine percent of women with hysterectomies had been previous or current users of hormone therapy at study enrollment, whereas only 26% of women in the estrogen-plus-progestin study used hormones prior to the start of the study. Among the past users of hormone therapy in the estrogen-alone trial, 41% had not received hormone therapy in at least 10 3198 CIRILLO ET AL Table 3. Incidence of arthroplasty by randomization assignment in the estrogen-plus-progestin trial of the Women’s Health Initiative* Followup time, mean ⫾ SD years Total joint replacement‡ Total hip replacement Total knee replacement Censoring after nonadherence§ Total joint replacement‡ Total hip replacement Total knee replacement Estrogen plus progestin (n ⫽ 8,240) Placebo (n ⫽ 7,809) HR (95% CI)† P† 5.7 ⫾ 1.3 222 (48) 87 (19) 138 (30) 5.6 ⫾ 1.3 206 (48) 70 (16) 140 (32) 0.99 (0.82–1.20) 1.14 (0.83–1.57) 0.91 (0.72–1.15) 0.92 0.41 0.41 136 (43) 49 (16) 87 (28) 138 (44) 48 (15) 94 (30) 1.02 (0.81–1.29) 1.08 (0.72–1.61) 0.95 (0.71–1.27) 0.77 0.71 0.72 * Except where indicated otherwise, values are the number of patients undergoing arthroplasty (rate per 10,000 person-years). † Hazard ratios (HRs), 95% confidence intervals (95% CIs), and P values from Cox proportional hazards analyses stratified by age. ‡ Events include any hip or knee arthroplasty. § Analyzed per-protocol, with followup time of participants censored at 6 months post-nonadherence, which was predefined as consuming ⬍80% of assigned study pills. years, and 27% had been currently receiving hormone therapy prior to randomization. In the estrogen-alone trial, 44% of participants were overweight (BMI ⬎25), compared with 33% in the estrogen-plus-progestin trial. The estrogen-alone trial was stopped on February 29, 2004, with a mean followup period of 7.1 years and a maximum followup period of 10.2 years (18), with comparable adherence between the estrogen and placebo groups (53.8% for both treatments at study termination). Since the estrogen-plus-progestin trial was halted early (on July 7, 2002), the mean followup time in this analysis was 5.6 years, with a maximum followup period of 8.6 years at the time of trial closure. The mean followup time in this report was 5.6 years for those who received placebo and 5.7 years for those who received estrogen plus progestin. Forty-two percent of subjects randomly assigned to receive estrogen plus progestin and 38% of subjects randomly assigned to receive placebo were nonadherent during the study. In the estrogen-alone trial, the incidence rate for total hip replacement events was 16 per 10,000 personyears in the estrogen-alone group compared with 22 per 10,000 person-years in the placebo group (Table 2). For total knee replacement events, the rates were 47 and 54 per 10,000 person-years in the estrogen-alone and placebo groups, respectively. A total of 222 women had at least 1 total joint replacement in the estrogen-alone group, compared with a total of 269 women in the placebo group, which yielded incidence rates for any total joint replacement of 62 per 10,000 person-years in the estrogen-alone group and 74 per 10,000 person-years in the placebo group. In Cox proportional hazards modeling, estrogen alone was found to be protective against any joint replacement compared with placebo (HR 0.84 [95% CI 0.70–1.00], P ⫽ 0.05). When hip and knee joints were considered separately, the protective effect approached statistical significance in predicting total hip replacement (HR 0.73 [95% CI 0.52–1.03], P ⫽ 0.07), but not in predicting total knee replacement (HR 0.87 [95% CI 0.71–1.07], P ⫽ 0.19). In the estrogen-plus-progestin trial, the estimated rate of any arthroplasty was 48 per 10,000 person-years in both the estrogen-plus-progestin and placebo groups (Table 3). For total hip replacement, the rates were 19 and 16 per 10,000 person-years in the estrogen-plusprogestin and placebo groups, respectively. For total knee replacement, the rates were 30 and 32 per 10,000 person-years, respectively. The HRs were not statistically significant for total hip replacement (HR 1.14 [95% CI 0.83–1.57], P ⫽ 0.41), total knee replacement (HR 0.91 [95% CI 0.72–1.15], P ⫽ 0.41), or for the index of any joint replacement (HR 0.99 [95% CI 0.82–1.20], P ⫽ 0.92). When women were censored 6 months after they became nonadherent, the protective effect of estrogen alone was strengthened (Table 2). The HRs became 0.73 for any arthroplasty (95% CI 0.58–0.93, P ⫽ 0.01), 0.55 for total hip replacement (95% CI 0.35–0.88, P ⫽ 0.01), and 0.80 for total knee replacement (95% CI 0.61–1.05, P ⫽ 0.11). However, censoring nonadherent participants had little effect on the estimates or significance in the estrogen-plus-progestin trial (Table 3). Adjustment for educational attainment did not appreciably change the findings (data not shown). EFFECT OF ESTROGEN THERAPY ON JOINT REPLACEMENT 3199 Figure 2. Kaplan-Meier estimate of cumulative hazards for arthroplasty in the unopposed-estrogen trial. Risks of hip (A) and knee (B) joint replacement are shown separately among women receiving conjugated equine estrogens (CEE) alone or placebo. Total joint replacement (C) includes time to either outcome. A further analysis was conducted with censoring occurring at 6 months after nonadherence to randomization assignment (D). HR ⫽ hazard ratio; 95% CI ⫽ 95% confidence interval. The possibility of differential hormone therapy effects in some subgroups, including subjects grouped according to age, ethnicity, past hormone therapy, and tobacco and alcohol consumption, was considered for each joint separately. When estrogen alone was considered as a predictor of total hip replacement (crude HR 0.73), there was no evidence of effect modification by these variables. However, when estrogen alone was considered with total knee replacement (crude HR 0.87), significantly different hormone therapy effects among race/ethnic groups were observed (P for interaction ⫽ 0.04). The protective effect was limited to whites (HR 0.79 [95% CI 0.63–0.99]). Hormone therapy showed no effect in blacks (HR 1.02 [95% CI 0.60– 1.73]). Among Hispanics, hormone therapy appeared to elevate risk (HR 3.51 [95% CI 0.97–12.76]). However, it 3200 CIRILLO ET AL should be noted that these findings are based on a small number of minority participants, and the probability of at least 1 interaction test being significant by chance alone was ⬎0.60 based on the number of tests conducted. The Kaplan-Meier estimates of cumulative hazards of total hip replacement, total knee replacement, and any joint replacement in the estrogen-alone trial showed the effects of hormone therapy since the time of study enrollment (Figures 2A–C). The estrogen-plusprogestin trial curves did not show a divergence in risk between the treatment and placebo groups based on Kaplan-Meier plots and log-likelihood test (data not shown). In the estrogen-alone trial, the risk for hip replacements appeared to diverge after 1 year of followup, whereas the apparent divergence in knee replacement risk began at 2 years. When the analysis was limited by censoring subjects 6 months after they became nonadherent to the assigned therapy (Figure 2D), the estrogen-alone trial showed more separation between the curves but the timing of hormone therapy effects was not substantially altered. DISCUSSION This study included 2 separate populations, that is, more than 26,000 healthy, community-dwelling postmenopausal women with and without hysterectomy. Those who had had a hysterectomy were randomly assigned to receive a common oral estrogen therapy, and those who had not had a hysterectomy were randomly assigned to receive estrogen plus progestin. We studied the rates, during followup, of hip and knee arthroplasties, which are important clinical end points of severely symptomatic OA. A statistically significant decrease in arthroplasty risk was seen in the estrogen-alone trial. Estrogen alone also appeared to be stronger in preventing hip arthroplasty (27% decrease) than in preventing knee arthroplasty (13% decrease), although the individual joint findings were not statistically significant in the ITT analysis. The estimated absolute risk reduction for total joint replacement was 12 per 10,000 person-years, and, similarly, was 6 per 10,000 person-years for total hip replacement and 7 per 10,000 person-years for total knee replacement. When we censored women whose adherence to their assigned therapy dropped below 80%, the protective effect was more striking. However, the estrogen-plus-progestin trial still showed no relationship between hormone use and arthroplasty risk. To our knowledge, this is the first randomized trial to show evidence of a protective effect attributed to estrogen- alone therapy, while demonstrating no effect of estrogen in the presence of progestin. Observational studies evaluating clinical symptoms, signs, and physician diagnosis have indicated that hormone therapy increases the risk of OA. Results of the cross-sectional Rancho Bernardo study suggested that hip and hand OA (but not knee OA), as defined by pain questionnaires and clinical examination, were more prevalent among hormone therapy users, but the investigators also found a nonsignificant protective effect of recent hormone therapy initiation among current users (22). Cohort studies have also tended to reveal an increased OA incidence among hormone therapy users. In the National Health and Nutrition Examination Survey Epidemiological Follow-up Study, investigators found a dose-dependent association of duration of hormone therapy use with risk of self-reported incident OA (23). A cohort study among Canadian women, the Population Health Survey, showed a 2-fold relative risk for self-reported incident OA over a 2-year period among current, long-term hormone therapy users (24). Oliveria and colleagues used data from a nested case– control study to show that new hormone therapy use was associated with new clinical diagnosis of arthritis (25). The investigators in that study found that OA diagnosis was associated with frequency of physician visits. Since OA may have an insidious onset, detection of incident disease may be subject to surveillance bias. Health care utilization may be a consequence of joint symptoms, but frequent visits for other reasons, such as the treatment of perimenopausal symptoms, could result in coincidental diagnosis of OA. Therefore, this bias may cause positive associations among users of hormone therapy due to study design artifacts, especially when study outcomes are new diagnosis or medical claims of OA. Unlike the studies of clinical manifestations, most radiographic studies suggested a protective effect of hormone therapy against radiographic OA incidence or progression (7–11). Nevitt and colleagues found radiographic evidence of protection against hip OA among hormone therapy users in the cross-sectional Study of Osteoporotic Fractures, and the protective effect was greater among long-term hormone therapy users (7). Furthermore, the protective effect appeared to be limited to unopposed estrogen (7). The Chingford initial cross-sectional study also showed a protective association with radiographic knee OA (8), although the later followup study showed a nonsignificant protective effect against incident knee osteophytosis (10). The Framingham Osteoarthritis Study initially showed a EFFECT OF ESTROGEN THERAPY ON JOINT REPLACEMENT modest, nonsignificant protective effect among women with longer hormone therapy use (9). After 14 years of followup, the Framingham Study showed a nonsignificant protective effect against progression of radiographic knee OA among postmenopausal women currently using hormone therapy (11). Previous findings of protective effects could also be affected by confounding effects, since estrogen users have been shown to have healthier behaviors and attitudes (26), but our randomized trial limits the likelihood of confounding. The radiographic findings suggest that hormone therapy may influence the detectable progression of OA, but it is not clear if this relates to symptom severity. Requirement for arthroplasty has been suggested as a sensitive and specific proxy outcome for OA progression, particularly in clinical trials (27). In the Nurses’ Health Study cohort, investigators found no association between hormone therapy and hip replacement due to arthritis (14). A Swedish case–control study showed a significant 80% increase in risk of knee arthroplasty due to OA among postmenopausal women (13). However, a second Swedish case–control study showed that hormone therapy provided a significant protective effect against hip replacement due to OA (15). A British case–control study of hip replacement showed a protective effect among long-term users of hormone therapy and a harmful effect among new users (16). Some difficulties may arise in generalizing findings on arthroplasty across health care systems, and reports of these studies did not distinguish hormone therapy formulation. Arthroplasty may also depend on access to and utilization of health services, but these factors would be expected to be balanced in this randomized study. Only 1 randomized trial, the Heart and Estrogen/ Progestin Replacement Study (HERS), evaluated estrogen-plus-progestin therapy and OA and found no effect of estrogen plus progestin on knee pain severity or disability (28). Our methods differed from those of the HERS, since we did not have a standardized measure of joint symptoms. However, the HERS cannot provide a full picture of the relationship between estrogen and OA because the HERS investigators did not examine OA of the hip, arthroplasty outcomes, women without cardiovascular disease, and women using unopposed estrogen. Our findings were among healthy postmenopausal women, and significant findings were found only in the estrogen-alone trial and were stronger for hip, rather than knee, arthroplasty. This large clinical trial provides the best evidence to date of a possible protective effect of estrogen against arthroplasty risk. One interpretation of the differential effect of 3201 hormone therapy between the estrogen-alone and estrogen-plus-progestin trials may involve population differences. That is, women in the estrogen-alone trial had hysterectomies, and therefore may have had surgically induced menopause and longer durations of “estrogen deficiency” prior to the study compared with women with natural menopause. Women in the estrogen-alone trial reported more hormone therapy use and had greater BMI at baseline than women in the estrogen-plus-progestin trial. Since the study protocols were virtually identical between the separate trials, the fact that these data show a relationship only in the estrogen-alone trial and not in the estrogen-plus-progestin trial suggests a true association. If these findings are biologically meaningful, oral estrogen, in the absence of progestin, must affect OA development. Based on animal and in vitro data, estrogen appears to have an influential role in the health of cartilage and large joints. Some studies suggest that estradiol at physiologic levels yields net beneficial effects on cartilage metabolism, but that higher doses become deleterious (29). Both known estrogen receptors (ERs), ER␣ and ER␤, are expressed in chondrocytes from human OA hip and knee cartilage (30,31), and polymorphisms in ERs have been associated with generalized OA (32). Estrogen is active in the cartilage and synovium. Estrogen can have harmful effects, since estradiol increases interleukin-6 production in the synovial fluid by chondrocytes, which leads to joint inflammation and degeneration (33). However, estrogen also protects against cartilage degeneration through inhibition of matrix metalloproteinases (MMPs) (34). Additionally, estrogen and progesterone alter the function of fibroblastlike synoviocytes, antagonizing the effects of MMPs (35). Estrogen also exerts anabolic effects on cartilage matrix via insulin-like growth factor (IGF) metabolism by increasing IGF and IGF binding proteins (IGFBPs) (31,34). IGFBP-2 favors the cartilage matrix by storing IGF-1 and also increasing proteoglycan production (31). Also, estrogen inhibits MMP-1, decreasing the breakdown of type II collagen in OA patients (34). Estrogen therapy reduced cartilage lesions and osteophyte formation in monkeys (36). Furthermore, progestin may modulate the activity of estrogen in the articular space, suggesting the possibility that local regulation could contribute to the relationships observed in our study. In a study of ovariectomized monkeys that compared unopposed estrogen or estrogen plus progestin with control treatment, the estrogen group had significantly greater levels of IGF-1, IGF-2, IGFBP-1, and IGFBP-3 3202 in the synovial fluid, and levels in the estrogen-plusprogestin group were intermediate between those in the estrogen and control groups (37). It is possible that estrogen could have effects outside of the cartilage itself, such as neurohumoral effects on pain sensation (38), mood, or weight gain, which could mediate symptom severity and necessity for arthroplasty. Estrogen may provide a gross anatomic effect on maintaining the structural integrity of the knee joint, as was shown in a study on ovariectomized ewes (39). In an imaging study of knee cartilage, long-term users of estrogen had more tibial cartilage than those who had never used estrogen (40). Estrogen also has known antiresorptive effects on bone metabolism, and the estrogenic effect may be analogous to those found in recent studies of other antiresorptive drugs. Bisphosphonates were found to alter the progression of subchondral sclerosis in animal models (41) and decreased joint space narrowing and OA symptoms in a randomized trial (42). Findings from these studies suggested that bone remodeling can significantly influence OA progression. However, these alternate explanations do not discount the observed estrogen association; instead, they provide insight about potential mechanisms underlying the association of oral conjugated equine estrogens with joint replacement risk. There are some limitations to this study. While it is likely that the vast majority of the joint replacements were due to OA, we could not exclude the possibility that some were due to other causes, such as osteonecrosis or rheumatoid arthritis. However, it is likely that these would have been distributed equally among all study groups. Also, women with a history of arthroplasty were excluded after randomization. These exclusions applied to a relatively small proportion of subjects and were balanced between the treatment groups (4.4% in the estrogen-alone group compared with 4.3% in the placebo group [P ⫽ 0.77] and 3.1% in the estrogen-plusprogestin group compared with 3.6% in the placebo group [P ⫽ 0.08]). However, unmeasured confounders could contribute to differences in the treatment groups. Unblinding of study staff due to safety concerns was higher in the estrogen-plus-progestin trial, but these clinicians were not involved in the assessment of clinical outcomes and would not have had direct influence on orthopedic procedures. It is unlikely that the protective effect against hip arthroplasty seen in the estrogen-alone trial could be explained by differential surveillance in the treatment group, because, if anything, increased surveillance in the treatment group would tend to cause an apparent increase in risk. Finally, these trials have CIRILLO ET AL previously shown protective effects against osteoporotic hip fracture risk with both types of hormone therapy (18,19). Osteoporotic hip fracture is another significant cause of total hip replacement, and hormone therapy could protect against this. However, we censored women with hip fracture, and the protective effect was limited to one trial. More research is needed to elucidate the direct mechanism of the effect of hormone therapy, since this cannot be determined from the design of the present study. These reported findings suggest that estrogen therapy confers a reduction in the risk of joint degeneration, but since the effect was limited to unopposed estrogen and may be more important for hip replacement than for knee replacement, there is a need for further understanding of estrogen’s potential impact on joint health. REFERENCES 1. Beaupre GS, Stevens SS, Carter DR. Mechanobiology in the development, maintenance, and degeneration of articular cartilage. J Rehabil Res Dev 2000;37:145–51. 2. Cooper C, Campbell L, Byng P, Croft P, Coggon D. Occupational activity and the risk of hip osteoarthritis. Ann Rheum Dis 1996; 55:680–2. 3. Elders MJ. The increasing impact of arthritis on public health. J Rheumatol Suppl 2000;60:6–8. 4. Canale ST, editor. Campbell’s operative orthopaedics. Vol. 4. 9th ed. St. Louis: Mosby; 1998. 5. American Academy of Orthopaedic Surgeons. Improving musculoskeletal care in America. URL: http://www3.aaos.org/research/ imca/OA_knee_TOC.htm. 6. Barnabei VM, Cochrane BB, Aragaki AK, Nygaard I, Williams RS, McGovern PG, et al. Menopausal symptoms and treatmentrelated effects of estrogen and progestin in the Women’s Health Initiative. Obstet Gynecol 2005;105(5 Pt 1):1063–73. 7. Nevitt MC, Cummings SR, Lane NE, Hochberg MC, Scott JC, Pressman AR, et al, Study of Osteoporotic Fractures Research Group. Association of estrogen replacement therapy with the risk of osteoarthritis of the hip in elderly white women. Arch Intern Med 1996;156:2073–80. 8. Spector TD, Nandra D, Hart DJ, Doyle DV. Is hormone replacement therapy protective for hand and knee osteoarthritis in women? The Chingford Study. Ann Rheum Dis 1997;56: 432–4. 9. Hannan MT, Felson DT, Anderson JJ, Naimark A, Kannel WB. Estrogen use and radiographic osteoarthritis of the knee in women: the Framingham Osteoarthritis Study. Arthritis Rheum 1990;33:525–32. 10. Hart DJ, Doyle DV, Spector TD. Incidence and risk factors for radiographic knee osteoarthritis in middle-aged women: the Chingford Study. Arthritis Rheum 1999;42:17–24. 11. Zhang Y, McAlindon TE, Hannan MT, Chaisson CE, Klein R, Wilson PW, et al. Estrogen replacement therapy and worsening of radiographic knee osteoarthritis: the Framingham Study. Arthritis Rheum 1998;41:1867–73. 12. Dougados M, Gueguen A, Nguyen M, Berdah L, Lequesne M, Mazieres B, et al. Requirement for total hip arthroplasty: an EFFECT OF ESTROGEN THERAPY ON JOINT REPLACEMENT 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. outcome measure of hip osteoarthritis? J Rheumatol 1999;26: 855–61. Sandmark H, Hogstedt C, Lewold S, Vingard E. Osteoarthrosis of the knee in men and women in association with overweight, smoking, and hormone therapy. Ann Rheum Dis 1999;58:151–5. Karlson EW, Mandl LA, Aweh GN, Sangha O, Liang MH, Grodstein F. Total hip replacement due to osteoarthritis: the importance of age, obesity, and other modifiable risk factors. Am J Med 2003;114:93–8. Vingard E, Alfredsson L, Malchau H. Lifestyle factors and hip arthrosis: a case referent study of body mass index, smoking and hormone therapy in 503 Swedish women. Acta Orthop Scand 1997;68:216–20. Dennison EM, Arden NK, Kellingray S, Croft P, Coggon D, Cooper C. Hormone replacement therapy, other reproductive variables and symptomatic hip osteoarthritis in elderly white women: a case-control study. Br J Rheumatol 1998;37: 1198–202. The Women’s Health Initiative Study Group. Design of the Women’s Health Initiative clinical trial and observational study. Control Clin Trials 1998;19:61–109. Women’s Health Initiative Steering Committee. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy: the Women’s Health Initiative randomized controlled trial. JAMA 2004;291:1701–12. Writing Group for the Women’s Health Initiative. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results from the Women’s Health Initiative randomized controlled trial. JAMA 2002;288:321–33. Anderson GL, Manson J, Wallace R, Lund B, Hall D, Davis S, et al. Implementation of the Women’s Health Initiative study design. Ann Epidemiol 2003;13(9 Suppl):S5–17. Stefanick ML, Cochrane BB, Hsia J, Barad DH, Liu JH, Johnson SR. The Women’s Health Initiative postmenopausal hormone trials: overview and baseline characteristics of participants. Ann Epidemiol 2003;13(9 Suppl):S78–86. Von Muhlen D, Morton D, von Muhlen CA, Barrett-Connor E. Postmenopausal estrogen and increased risk of clinical osteoarthritis at the hip, hand, and knee in older women. J Womens Health Gend Based Med 2002;11:511–8. Sahyoun NR, Brett KM, Hochberg MC, Pamuk ER. Estrogen replacement therapy and incidence of self-reported physiciandiagnosed arthritis. Prev Med 1999;28:458–64. Wilkins K. Hormone replacement therapy and incident arthritis. Health Rep 1999;11:49–57(Eng); 57–66(Fre). Oliveria SA, Felson DT, Klein RA, Reed JI, Walker AM. Estrogen replacement therapy and the development of osteoarthritis. Epidemiology 1996;7:415–9. Posthuma WF, Westendorp RG, Vandenbroucke JP. Cardioprotective effect of hormone replacement therapy in postmenopausal women: is the evidence biased? BMJ 1994;308:1268–9. Dougados M. Monitoring osteoarthritis progression and therapy. Osteoarthritis Cartilage 2004;12 Suppl A:S55–60. Nevitt MC, Felson DT, Williams EN, Grady D. The effect of estrogen plus progestin on knee symptoms and related disability in postmenopausal women: the Heart and Estrogen/Progestin Replacement Study, a randomized, double-blind, placebo-controlled trial. Arthritis Rheum 2001;44:811–8. Richette P, Corvol M, Bardin T. Estrogens, cartilage, and osteoarthritis. Joint Bone Spine 2003;70:257–62. Ushiyama T, Ueyama H, Inoue K, Ohkubo I, Hukuda S. Expression of genes for estrogen receptors ␣ and ␤ in human articular chondrocytes. Osteoarthritis Cartilage 1999;7:560–6. Richmond RS, Carlson CS, Register TC, Shanker G, Loeser RF. Functional estrogen receptors in adult articular cartilage: estrogen replacement therapy increases chondrocyte synthesis of proteogly- 3203 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. cans and insulin-like growth factor binding protein 2. Arthritis Rheum 2000;43:2081–90. Ushiyama T, Ueyama H, Inoue K, Nishioka J, Ohkubo I, Hukuda S. Estrogen receptor gene polymorphism and generalized osteoarthritis. J Rheumatol 1998;25:134–7. Guerne PA, Carson DA, Lotz M. IL-6 production by human articular chondrocytes: modulation of its synthesis by cytokines, growth factors, and hormones in vitro. J Immunol 1990;144:499–505. Lee YJ, Lee EB, Kwon YE, Lee JJ, Cho WS, Kim HA, et al. Effect of estrogen on the expression of matrix metalloproteinase (MMP)-1, MMP-3, and MMP-13 and tissue inhibitor of metalloproteinase-1 in osteoarthritis chondrocytes. Rheumatol Int 2003; 23:282–8. Khalkhali-Ellis Z, Seftor EA, Nieva DR, Handa RJ, Price RH Jr, Kirschmann DA, et al. Estrogen and progesterone regulation of human fibroblast-like synoviocyte function in vitro: implications in rheumatoid arthritis. J Rheumatol 2000;27:1622–31. Ham KD, Loeser RF, Lindgren BR, Carlson CS. Effects of long-term estrogen replacement therapy on osteoarthritis severity in cynomolgus monkeys. Arthritis Rheum 2002;46:1956–64. Fernihough JK, Richmond RS, Carlson CS, Cherpes T, Holly JM, Loeser RF. Estrogen replacement therapy modulation of the insulin-like growth factor system in monkey knee joints. Arthritis Rheum 1999;42:2103–11. Blacklock AD, Johnson MS, Krizsan-Agbas D, Smith PG. Estrogen increases sensory nociceptor neuritogenesis in vitro by a direct, nerve growth factor-independent mechanism. Eur J Neurosci 2005;21:2320–8. Turner AS, Athanasiou KA, Zhu CF, Alvis MR, Bryant HU. Biochemical effects of estrogen on articular cartilage in ovariectomized sheep. Osteoarthritis Cartilage 1997;5:63–9. Wluka AE, Davis SR, Bailey M, Stuckey SL, Cicuttini FM. Users of oestrogen replacement therapy have more knee cartilage than non-users. Ann Rheum Dis 2001;60:332–6. Hayami T, Pickarski M, Wesolowski GA, McLane J, Bone A, Destefano J, et al. The role of subchondral bone remodeling in osteoarthritis: reduction of cartilage degeneration and prevention of osteophyte formation by alendronate in the rat anterior cruciate ligament transection model. Arthritis Rheum 2004;50:1193–206. Spector TD, Conaghan PG, Buckland-Wright JC, Garnero P, Cline GA, Beary JF, et al. Effect of risedronate on joint structure and symptoms of knee osteoarthritis: results of the BRISK randomized, controlled trial. Arthritis Res Ther 2005;7:R625–33. APPENDIX A: WHI INVESTIGATORS Program Office: Barbara Alving, Jacques Rossouw, Shari Ludlam, Linda Pottern, Joan McGowan, Leslie Ford, Nancy Geller (National Heart, Lung, and Blood Institute, Bethesda, MD). Clinical Coordinating Centers: Ross Prentice, Garnet Anderson, Andrea LaCroix, Charles L. Kooperberg, Ruth E. Patterson, Anne McTiernan (Fred Hutchinson Cancer Research Center, Seattle, WA); Sally Shumaker (Wake Forest University School of Medicine, Winston-Salem, NC); Evan Stein (Medical Research Labs, Highland Heights, KY); Steven Cummings (University of California at San Francisco). Clinical Centers: Sylvia Wassertheil-Smoller (Albert Einstein College of Medicine, Bronx, NY); Jennifer Hays (Baylor College of Medicine, Houston, TX); JoAnn Manson (Brigham and Women’s Hospital, Harvard Medical School, Boston, MA); Annlouise R. Assaf (Brown University, Providence, RI); Lawrence Phillips (Emory University, Atlanta, GA); Shirley Beresford (Fred Hutchinson Cancer Research Center, Seattle, WA); Judith Hsia (George Washington University Medical Center, Washington, DC); Rowan Chlebowski (Harbor–UCLA Research and Education Institute, Torrance, CA); Evelyn Whitlock (Kaiser Permanente Center for Health Research, 3204 Portland, OR); Bette Caan (Kaiser Permanente Division of Research, Oakland, CA); Jane Morley Kotchen (Medical College of Wisconsin, Milwaukee); Barbara V. Howard (MedStar Research Institute/ Howard University, Washington, DC); Linda Van Horn (Northwestern University, Chicago/Evanston, IL); Henry Black (Rush Medical Center, Chicago, IL); Marcia L. Stefanick (Stanford Prevention Research Center, Stanford, CA); Dorothy Lane (State University of New York at Stony Brook); Rebecca Jackson (The Ohio State University, Columbus); Cora E. Lewis (University of Alabama at Birmingham); Tamsen Bassford (University of Arizona, Tucson/Phoenix); Jean Wactawski-Wende (University at Buffalo, Buffalo, NY); John Robbins (University of California at Davis, Sacramento); F. Allan Hubbell (University of California at Irvine); Howard Judd (University of California at Los Angeles); Robert D. Langer (University of California at San Diego, La Jolla/Chula Vista); Margery Gass (University of CIRILLO ET AL Cincinnati, Cincinnati, OH); Marian Limacher (University of Florida, Gainesville/Jacksonville); David Curb (University of Hawaii, Honolulu); Robert B. Wallace (University of Iowa, Iowa City/Davenport); Judith Ockene (University of Massachusetts/Fallon Clinic, Worcester); Norman Lasser (University of Medicine and Dentistry of New Jersey, Newark); Mary Jo O’Sullivan (University of Miami, Miami, FL); Karen Margolis (University of Minnesota, Minneapolis); Robert Brunner (University of Nevada, Reno); Gerardo Heiss (University of North Carolina, Chapel Hill); Lewis Kuller (University of Pittsburgh, Pittsburgh, PA); Karen C. Johnson (University of Tennessee, Memphis); Robert Brzyski (University of Texas Health Science Center, San Antonio); Gloria E. Sarto (University of Wisconsin, Madison); Denise Bonds (Wake Forest University School of Medicine, Winston-Salem, NC); Susan Hendrix (Wayne State University School of Medicine/ Hutzel Hospital, Detroit, MI).