The Women Physicians' Health Study (WPHS) is the first comprehensive survey of the personal characteristics, attitudes, health-related behaviors, and counseling practices of a representative sample of U.S. women physicians, examining 716 variables in 4501 respondents [6]. The present report describes the prevalence and covariates of personal use of HRT by women physicians. A subsequent analysis will explore the relation between the personal characteristics of U.S. women physicians and their HRT counseling practices. Other WPHS analyses have included descriptions of basic demographic and professional characteristics, health-related behaviors, and experiences with harassment [7-9]. Women physicians are an important population to study because they can substantially influence the behaviors of their patients and other women, can accurately describe their personal medical history, and are among the groups best informed about the benefits and risks of HRT.

Methods

Top Methods Results Discussion Author & Article Info References

The design of WPHS [6] and the demographic and professional characteristics of WPHS respondents [8] are more fully described elsewhere. In brief, the WPHS mail survey was sent to a stratified random sample of U.S. women physicians selected from the American Medical Association's Physician Masterfile, a database designed to record all physicians living in the United States and its territories. Sampling was stratified by decade of graduation from medical school (graduating classes 1950 through 1989); 2500 women were selected from each stratum by using simple random sampling. Active, professionally inactive, part-time, and retired women physicians 30 to 70 years of age were included.

Nonrespondents were repeatedly asked to participate (a maximum of four mailings was sent) until enrollment ended in October 1994. The total number of respondents was 4501. An estimated 23% of potential respondents were ineligible because they were male, were deceased, were living outside of the United States, were currently in training, or had incorrect addresses. Among physicians eligible to respond, the response rate was 59%.

Respondents were compared with nonrespondents in three ways. First, a random sample of 200 nonrespondents was interviewed by telephone, and their responses were compared with those of respondents. Second, information from the American Medical Association's Physician Masterfile was used to compare all nonrespondents with respondents. Third, respondents to earlier mailings were compared with respondents to later mailings. Respondents and nonrespondents did not consistently or significantly differ in age, ethnicity, marital status, number of children, alcohol consumption, fat intake, exercise, smoking status, hours worked per week, frequency of being a primary care provider, personal income, time since the most recent routine check-up, their own assessment of their personal general health, total cholesterol level, height, weight, and active practice of medicine. The mailing wave to which respondents responded was not significantly associated with HRT use (P > 0.2). In these three investigations, only the proportion of persons with board certification differed significantly, with nonrespondents less likely to be board certified.

Sampling weights were assigned to the data to adjust for differences in the probability of selection by decade of graduation. All sampling weights were further adjusted for the higher response rates of the sampled physicians who were board certified. The analysis weights (within decade) for board-certified and non-board-certified respondents, respectively, were 3.4 and 5.5 (1950s), 9.3 and 17.7 (1960s), 17.9 and 36.5 (1970s), and 28.3 and 63.9 (1980s). This weighting scheme allows inferences to be drawn about the population of U.S. women physicians who graduated from medical school between 1950 and 1989.

Responses about menopausal status and personal current use of HRT were obtained, but we did not ask about past use of HRT. Women were defined as postmenopausal if they reported cessation of menses secondary to natural menopause, bilateral oophorectomy with or without hysterectomy, or chemotherapy. Women who reported cessation of menses secondary to hysterectomy without bilateral oophorectomy or for unspecified reasons were considered postmenopausal only if they were 55 years of age or older. Women were excluded from the present analysis if they were younger than 40 years of age (n = 1033). Of the 4501 respondents, 1466 were defined as postmenopausal and were included in the present analysis. Other survey questions asked about demographic, medical, and professional characteristics that have been reported [2, 3, 10-12] or characteristics that we hypothesized would correlate with use of HRT.

A specialized statistics program for the analysis of weighted survey data, SUDAAN [13], was used to estimate the prevalence of HRT use by personal, professional, and medical history characteristics and to perform chi-square tests to determine whether HRT use was related to these characteristics. In general, SUDAAN analyses give wider CIs than do weighted analyses in standard statistical packages. A weighted, adjusted logistic regression analysis in SUDAAN was used to model a modified version of backward and stepwise selection.

Results

Top Methods Results Discussion Author & Article Info References

Of the 1466 physicians who were defined as postmenopausal, 644 (47.4%) were using HRT at the time of the survey. The prevalence of use was 59.8% in the fifth decade of life, 49.4% in the sixth decade, and 36.4% in the seventh decade (P < 0.001). Women were postmenopausal secondary to natural menopause (n = 1081; 71.3%), bilateral oophorectomy with or without hysterectomy (n = 190; 16.4%), or chemotherapy (n = 22; 2.7%) or were older than 55 years of age and had had cessation of menses secondary to hysterectomy without bilateral oophorectomy or for unspecified reasons (n = 173; 9.7%).

The characteristics of postmenopausal women physicians and their weighted univariate relations with HRT use at the time of survey are shown in Table 1. Characteristics significantly associated with HRT status in univariate analysis were entered into a weighted logistic regression model. Each odds ratio, 95% CI, and P value was calculated after adjustment for all other variables in the model (Table 2). The medical characteristics most strongly related to HRT status were having had a hysterectomy (odds ratio, 4.64 [95% CI, 3.14 to 6.87]) and having a personal history of breast cancer (odds ratio, 0.12 [CI, 0.05 to 0.29]). The logistic regression model had a good fit (Hosmer and Lemeshow goodness-of-fit statistic, 3.76 with 8 degrees of freedom; P > 0.2).

Characteristics hypothesized to be associated with HRT use but found not to be significantly associated with HRT use in univariate analysis were entered in a stepwise fashion. In the weighted, adjusted model, mean hours of exercise per week; a personal history of high cholesterol levels, osteoporosis, or coronary heart disease; and a family history of high cholesterol levels or coronary heart disease had no association with HRT use (P > 0.2) and were removed from the final model (Table 2).

Discussion

Top Methods Results Discussion Author & Article Info References

The risks and benefits of postmenopausal HRT have received considerable attention in both the public press and the medical literature. Our analysis is the first that we know of to provide information about postmenopausal U.S. women physicians' personal use of HRT. The prevalence of HRT use by women physicians ranges from 59.8% in women 40 to 49 years of age to 36.4% in women 60 to 70 years of age; thus, a significant percentage of these well-informed women choose to use HRT.

A 1993 sample of postmenopausal Stanford University graduates [3] reported a similar rate of HRT use (49%). In contrast, in a 1987 survey of middle-aged white women in Massachusetts, only 9.3% of women were currently using oral estrogens [2]. In the first nationally representative survey on women's health to include HRT use, 24% of postmenopausal women reported current use of HRT [4]. Differences in the prevalence of HRT use can be partly explained by secular trends [1]. In addition, studies have identified a significant association between HRT use and higher socioeconomic status [2, 10, 12].

In a 1993 survey of 900 postmenopausal British women physicians, current HRT use was 53% in women 40 to 49 years of age, 49% in women 50 to 59 years of age, and 23% in women 60 to 69 years of age [14]. This is strikingly similar to our findings for U.S. women physicians. Despite such high personal use of HRT by British physicians, only 7% of British women aged 40 to 69 years reported current use of HRT [15]. Higher rates of HRT use by women physicians and other well-educated women may presage greater use of HRT for U.S. women in the future [2, 3, 10, 12, 14]. Health-related behaviors of physicians and other groups with high socioeconomic status have been shown to change before behaviors of society as a whole change. An example of this is the manner in which rates of tobacco smoking among U.S. physicians have led declines in national rates for the past 30 years [16]. In addition, evidence suggests that physicians' personal health practices influence the ways in which they counsel patients [17]. Physicians who use HRT may be more likely to recommend HRT to their patients.

The only professional characteristic significantly associated with the personal use of HRT was physician specialty. This is consistent with surveys that have explored differences in prescribing patterns by gynecologists and other specialists; gynecologists were more likely to prescribe HRT for prevention of osteoporosis and relief of menopausal symptoms [5, 18]. In agreement with results from surveys of female patients, hysterectomy status was the medical characteristic most strongly associated with the use of HRT by women physicians [2, 10, 11] and women physicians' decisions to use HRT varied significantly by geographic location of residence [10].

Lack of significant association between such characteristics as a personal history of diabetes or coronary heart disease and HRT use may be due to the small sample size (n < 40) in our subgroup comparisons. A personal or family history of breast cancer was strongly associated with lack of HRT use, consistent with U.S. guideline recommendations [19, 20].

A methodologic strength of our study is the extensive database of demographic, professional, behavioral, and personal and family medical history variables. Although other surveys have examined many of these variables, ours is the first to combine them into a single logistic regression model. Limitations of our study include the suboptimal response rate of 59% and potential response bias; lack of information on the types of HRT regimens used, the duration of HRT use, previous HRT use, and specific reasons for HRT use; and small sample sizes for some subgroup analyses.

In summary, use of HRT by U.S. women physicians is significantly influenced by age, ethnicity, geographic location, and several behavioral and medical characteristics. Higher rates of HRT use by women physicians and other well-educated women may presage greater use of HRT by U.S. women in the future.

Dr. Wenger: Division of Cardiology, Emory University School of Medicine, 69 Butler Street, Atlanta, GA 30303.

Dr. Frank: Department of Family and Preventive Medicine, Emory University School of Medicine, 69 Butler Street, Atlanta, GA 30303.