Changes in the Use of Postmenopausal Hormone Therapy after the Publication of Clinical Trial Results

Context Since 1998, 2 large trials have drastically changed the evidence for the preventive health benefits of postmenopausal hormone replacement therapy. However, changes in practice often lag behind changes in evidence. Contribution Among mammography recipients in San Francisco, California, the use of hormone replacement therapy decreased 1% per quarter after publication of the Heart and Estrogen/progestin Replacement Study and 18% per quarter after publication of results from the Women's Health Initiative (WHI). Reduction in use was unrelated to a woman's age, hysterectomy status, or race or ethnicity. Implications The WHI resulted in more dramatic changes in practice than are often associated with changes in evidence. The vigorous media coverage of the WHI may have contributed to rapid changes in practice. The Editors In 1995, approximately 38% of postmenopausal women in the United States were taking hormone therapy (1). At that time, several observational studies had suggested that hormone therapy offered women some protection against coronary heart disease and osteoporosis (2-5). A decision analysis published in 1997 concluded that the benefits of hormone therapy outweighed its risks for nearly all women (6). More recently, the results from 2 large randomized clinical trials, the Heart and Estrogen/progestin Replacement Study (HERS) (7) and the Women's Health Initiative (WHI), have been published (8). These clinical trials demonstrated that the risks associated with hormone therapy outweigh the benefits for women taking continuous estrogen and progestin regimens. As a result of these trial results, the U.S. Food and Drug Administration required new warning labels for all estrogen products (9), and the U.S. Preventive Services Task Force revised its assessment of hormone therapy to recommend against the routine use of estrogen and progestin for the prevention of chronic conditions in postmenopausal women (10). It is important to understand whether this new scientific evidence is changing the use of hormone therapy. Of note, the results of the WHI were widely disseminated. Despite this publicity, differential access to new information, varied interpretations of study findings, and individual perceptions of menopausal symptoms and hormone side effects may have resulted in different patterns of use. An understanding of how use is changing over time provides important information about the dissemination of clinical trial results to women. We designed our analysis to examine whether the use of hormone therapy has changed among postmenopausal women as a result of the publication of the results from HERS and the WHI. We were also interested in examining whether patterns of use differ by patient characteristics. Because HERS examined the outcomes of older women, we hypothesized that there would be earlier and more substantial declines in hormone therapy use among this group. We also expected that there would be variation in use by race or ethnicity because white women may have better access to new information (11). Finally, because the WHI study results were specific to women taking continuous estrogen plus progestin, we hypothesized that hormone use would be more stable among women who had had hysterectomies because such women typically take only estrogen and may believe that the findings do not apply to them. Methods Sample The San Francisco Mammography Registry is a population-based registry of women undergoing mammography in San Francisco, California. It is 1 of 7 registries participating in the National Cancer Institute Breast Cancer Surveillance Consortium (12). This registry began to prospectively collect patient data and mammography results in 1995 and currently captures about 90% of mammography examinations performed in San Francisco. Data from 11 mammography facilities are included in this analysis. Women were eligible for this analysis if they were between the ages of 50 to 74 years, were postmenopausal, did not report a personal history of breast cancer, and underwent screening or diagnostic mammography between January 1997 and 19 May 2003. Women 55 years of age and older were assumed to be menopausal. Women 50 to 54 years of age were considered to be menopausal if both ovaries had been removed or if they reported that their periods had stopped permanently. For women who had mammography more than once in any calendar year, we included only the first instance of mammography in that year to prevent overrepresentation of women undergoing an evaluation of an abnormal mammogram because this experience may influence use of hormone therapy. Our final sample included 15 1862 mammograms received by 71 219 women. Data At the facilities that participate in the San Francisco Mammography Registry, each woman completes a brief, scannable questionnaire at the time of mammography. This questionnaire collects information about current use of hormone therapy and several personal characteristics, including race or ethnicity (categorized as white, African American, Latina, Chinese, Filipina, other Asian, other), family history of breast cancer (including mother, sisters, and daughters), history of childbirth, whether the woman had undergone a hysterectomy, menopausal status, history of breast biopsy (including fine-needle aspiration, core biopsy, and surgical biopsy). Information about age at the time of mammography, date of mammography, and ZIP code of residence is reported by the facility. Data from the year 2000 U.S. Census was used to assign median income for each woman's ZIP code of residence as a proxy for socioeconomic status. Variables Our outcome variable for this analysis was the current use of hormone therapy. Date of mammography was represented as a linear term. Binary variables were created for the publication dates of HERS (before 19 August 1998 vs. that date or later) and the publication of the principal findings from the WHI (before 17 July 2002 vs. that date or later). Other independent variables examined were age at the time of mammography, race or ethnicity, median income of the ZIP code of residence, history of childbirth, family history of breast cancer, history of breast biopsy, and previous hysterectomy. Statistical Analysis Because some of the women in this sample had more than 1 mammogram represented in this data set, which spanned a 7-year period, we conducted a repeated-measures logistic regression to adjust the variance estimates for clustering of hormone therapy use over time for individual women and for the clustering of women within mammography facilities (13). Generalized estimating equations were implemented by using the SUDAAN statistical package, version 8.0.0 (Research Triangle Institute, Research Triangle Park, North Carolina) assuming an exchangeable correlation matrix. These models included a linear term indicating quarter from January 1997 to the first quarter of 2003 to control for temporal trends (the last quarter included mammograms through 19 May 2003), the variables specified above to indicate the dates of publication of HERS and the WHI, and an interaction term between each of these publication indicators and the time (in quarters) following each of these publications to measure changes in use after the publication of these clinical trials. These models also controlled for the individual characteristics described earlier (that is, age, race or ethnicity, history of childbirth, family history of breast cancer, history of breast biopsy, previous hysterectomy, median income for the ZIP code of residence). To specifically test our hypotheses about differential changes in the use of hormone therapy for subgroups of women on the basis of age, hysterectomy status, and race or ethnicity, we examined interaction terms to test for effect modification. For the main effects, a P value less than 0.05 was considered statistically significant, and for the interaction terms, a P value less than 0.01 was considered to be statistically significant. The likelihood ratio test compared the null model with the fitted model. Role of the Funding Source This work was supported by a National Cancer Institutefunded Breast Cancer Surveillance Consortium agreement. The funding source did not participate in the design, conduct, or reporting of this analysis or in the decision to submit the manuscript for publication. Results The Table shows the characteristics of the sample for each of the study years. Over the time period of the study, the median age decreased from 61 years to 59 years. The racial and ethnic composition of the sample also changed somewhat across the study years. Fewer women undergoing mammography in 2003 reported a history of childbirth (71.7% vs. 75.2%) or hysterectomy than did women undergoing mammography in 1997. Conversely, more women reported a family history of breast cancer (17.1% vs. 11.8%) or a personal history of a previous breast biopsy or aspiration. The average number of mammograms obtained for each woman in our sample across the 7-year study period was 2.1 (range, 1 to 7). Table. Description of the Sample (151 862 Mammograms) The Figure shows the unadjusted rates of current hormone therapy use by month for all of the women in the sample. Among menopausal women who had received mammography, we estimated that the average proportion reporting the current use of hormone therapy was 41% in 1997. In 1997, hormone use was highest among white women (52.6%) and lowest among African-American women (34.1%), Latina women (33.9%), Chinese women (32.2%), and Filipina women (29.6%). In 1997, hormone use was higher among younger women than older women (48.7% vs. 28.7%; P < 0.001) and among women who had had a hysterectomy compared with women who had not had a hysterectomy (60.0% vs. 36.4%; P < 0.001). Figure. Rates of hormone therapy use among postmenopausal women, 1997 to 2003 HERS WHI The adjusted multivariate model estimates that before the publication of HERS, the use of hormone thera