Differences Between Gynecologists and Primary Care Physicians in Hormone Therapy Prescribing: Why They Matter

In the 1990s, prescribing postmenopausal hormone therapy (HT) indefinitely for general health benefits—and not merely to treat vasomotor symptoms—became routine in many primary care and gynecology practices. Using observational and epidemiological studies as principal sources of supporting data, many authorities believed that the overall health benefits of HT outweighed the risks for most women, including those who were asymptomatic. In particular, HT was thought to reduce risk for cardiovascular disease (CVD). Clinicians were chided for ignoring the fact that CVD afflicted women as well as men and were encouraged to persuade women that long-term HT use was prudent. Prescribing HT for disease prevention became almost a badge of honor, showing that clinicians were serious about women's health.

These beliefs and practice recommendations were expressed in numerous textbooks and review articles. For example, the section on menopausal estrogen therapy in the 1994 edition of Harrison's Principles of Internal Medicine included the following passage, coauthored by an internist and a gynecologist:

…there is no documented evidence that low-dose estrogen therapy in the menopause increases the incidence or severity of thromboembolic disease, breast cancer, or hypertension…strong evidence suggests that, in fact, estrogens may decrease the incidence of death from myocardial infarction. ¹

The 1996 edition of a prominent gynecology textbook stated the following:

…because of the health risks associated with estrogen deficiency, replacement therapy should be offered to all postmenopausal women who do not have a contraindication. Women with known risk factors for cardiovascular disease … should be encouraged to take estrogen to minimize their risks.… estrogen replacement therapy protects against cardiovascular disease, including coronary artery disease and stroke.….there is no increased risk of either venous thrombosis or pulmonary embolism. ²

By the end of the 1990s, authors were buttressing their beliefs with quantitative estimates. For example, in the 2000 edition of the Cecil Textbook of Medicine, the author estimated a “net change of 5,561 lives saved per 100,000 estrogen users,” ³ although the time frame during which this alleged benefit accrued was not clearly stated. This analysis was dominated by prevention of ischemic heart disease (5250 lives saved) and osteoporotic fractures (563 lives saved) and was affected only minimally by a worst-case estimate of 187 lives lost from breast cancer. In a formal decision analysis, HT increased estimated life expectancy in almost all women, with gains of up to 3 years for certain subgroups of women. ⁴

The first warning shot that challenged the preventive efficacy of HT was fired in 1998 with publication of the Heart and Estrogen/progestin Replacement Study (HERS), a randomized trial of secondary prevention in women with established coronary disease; during 4 years of treatment, HT did not prevent recurrent coronary events but did increase the incidence of venous thromboembolic events. ⁵ Then, everything changed in July 2002 with publication of initial results from the randomized Women's Health Initiative (WHI) study, the first large primary prevention trial designed to assess the effect of HT on CVD and other health outcomes. ⁶ Nearly 17,000 women with an intact uterus received either placebo or combined HT with conjugated equine estrogen (CEE) plus medroxyprogesterone acetate (MPA). During average follow-up of 5 years, the harms of HT (increased risks for myocardial infarction [MI], stroke, thromboembolic events, and breast cancer) exceeded the benefits (reduced risks for fractures and colorectal cancer). Many patients heard about these results from intense media coverage and contacted their gynecologists and primary care physicians for advice. In our general internal medicine group, we reviewed the published findings, agreed that HT was not justified for asymptomatic women as a primary preventive intervention, and asked most of our patients to stop long-term use of HT. We found that very few patients had trouble doing so.

During the next few months of 2002, however, we observed something interesting: HT frequently remained on the medication list of our patients who had been prescribed long-term HT by gynecologists. Some of these patients related that they had phoned their gynecologists shortly after the WHI publication and were simply told “It's OK to keep taking your hormone therapy.” Other patients reported that their gynecologists made specific comments about the WHI, including “It's only one study,” “The study was flawed,” “Very few women were harmed,” and “The benefits still outweighed the risks.” I wondered if these local observations reflected practices and attitudes more generally across the United States. Thus, I collaborated with a gynecologist and distributed a survey to a geographically representative U.S. sample of general internists and gynecologists in October and November of 2002. We wanted to assess clinicians' attitudes in the immediate aftermath of the WHI publication before formal practice guidelines on HT were revised and published.

The focus of our survey was a case scenario: We asked clinicians how they would respond to a 60-year-old asymptomatic woman who had been taking CEE/MPA (Prempro, Wyeth) for 5 years for preventive reasons—not for treatment of ongoing vasomotor symptoms—and who expressed concern about the recently publicized WHI results. Our results revealed a striking interprofessional difference: gynecologists were significantly more likely to encourage or permit continuing HT use than were general internists (65% vs. 35%, p < 0.001). ⁷

As HT prescribing declined during the year after the initial WHI publication, ⁸ additional findings from the trial were published subsequently between 2003 and 2005. These analyses showed that combined HT did not improve general quality of life, ⁹ cognition, ^10,11 or urinary continence. ¹² Estrogen-only therapy (studied in a separate population of women with prior hysterectomy) had a somewhat different benefit-harm profile than did combined HT but still did not confer net benefits in healthy women. ^{13 –16} Thus, none of these additional reports provided compelling counterweight against the initial 2002 findings.

Against this background, the study by Lakey et al. ¹⁷ in this issue of Journal of Women's Health is welcome and informative. They surveyed nearly 600 physicians in 2006, when virtually all key outcomes from the WHI had already been published. Obstetrician/gynecologists, compared with primary care physicians (family physicians and internists), were more likely to believe that they had expert-level knowledge of WHI (30% vs. 8%), that the “balance of HT benefits still outweigh the risks” (37% vs. 19%), and that the WHI findings are irrelevant for younger women with hot flashes (41% vs. 20%). Obstetrician/gynecologists also were more likely to “continue recommending HT to many women” (59% vs. 23%) and to recommend HT to treat menopause symptoms in women with coronary disease (20% vs. 10%). However, <5% of respondents in both groups would prescribe HT for primary prevention of heart disease. Finally, obstetrician/gynecologists reported a significantly greater tendency to use vaginal and transdermal estrogens than their primary care counterparts. Although respondents' self-reported prescribing habits were not confirmed by data on actual HT prescribing, there is little reason to suspect that such data would contradict these self-reports.

Most items in the survey by Lakey et al. ¹⁷ did not explore whether the divergent perspectives of obstetrician/gynecologists and primary care physicians apply both to recently menopausal patients and to older patient populations. In particular, although gynecology respondents were more likely to believe that “the balance of HT benefits still outweigh the risks,” this survey item did not obviously distinguish between short-term use in symptomatic women and long-term use for general preventive reasons in asymptomatic women. For some clinicians, an affirmative response to this item might have reflected a short-term quality of life judgment, applicable mainly to early menopausal women; that is, successful treatment of debilitating vasomotor symptoms might override small cardiovascular, thromboembolic, or breast cancer risks that some women might find acceptable. For other clinicians, an affirmative response might have reflected persistent belief that HT provides intangible health benefits in aging women that were not captured by the WHI. For still other clinicians, an affirmative response might simply reflect skepticism about the validity of the WHI study, which has been noted among gynecologists in several surveys.” ^18,19 A final possibility is that gynecologists are more likely to consider the WHI study as applicable only to CEE and MPA and to believe that alternative HT regimens—which they are now using preferentially—have more favorable benefit/harm ratios than oral CEE and MPA.

Undoubtedly, practical considerations, some of which are discussed by Lakey et al., ¹⁷ also explain the findings. For example, gynecologists might be more likely to attract patients who explicitly request or demand HT, in some cases, perhaps, after being referred by primary care physicians. Moreover, through sheer case volume, gynecologists are presumably more comfortable with HT than are primary care physicians and, thus, might have a lower threshold for prescribing it. The survey strongly suggests this possibility for use of transdermal and vaginal preparations. A different sort of case-volume argument might apply to internists and family physicians. These clinicians are more likely than gynecologists to manage patients with CVD and thromboembolism and, thus, might weigh the potential harms of HT more heavily.

Lingering beliefs about the value of HT play into what I would call “revisionist history” about the WHI. Because the mean age of WHI subjects was 63, some detractors claim that the results are irrelevant for women in their 50s. However, 63 was the average age, not the lower age limit. Fully one third of women in the WHI were aged 50–59, and no clear-cut age-related interactions were noted for the effects of CEE/MPA on cardiovascular or thrombotic events or on breast cancer. Some authorities have advanced a timing hypothesis—the idea that HT is safe and possibly beneficial for coronary prevention if started immediately after menopause ²⁰ —but others disagree. ^21,22 Recent analyses from the WHI study itself have reached conflicting conclusions about the validity of the timing hypothesis. ^23,24 Even if there is a small window of time after menopause during which estrogens favorably influence risk for coronary disease, it seems unlikely that randomized trials with clinical (and not surrogate) end points will ever identify this time window or the subgroup of women who clearly benefit from HT exposure during that interval. Moreover, risks for stroke, venous thromboembolism, and breast cancer enter the benefit/harm calculus at any age. Randomized trials, (for example, the Kronos Early Estrogen Prevention Study (KEEPS) ²⁵ and Early versus Late Intervention Trail with Estradiol (ELITE) ²⁶ trials, are currently examining surrogate end points (e.g., carotid intima-media thickness and coronary artery calcium) in recently postmenopausal women treated with various estrogen and progestin preparations. Recent history tells us, however, that favorable changes in surrogate end points are often unreliable substitutes for clinical outcomes.

Differences between the practice patterns of gynecologists and primary care physicians are not limited to HT. For example, in one recent survey of cervical cancer screening, internists were more likely to report guideline-consistent practices than were gynecologists (although both specialties tended to recommend excessive screening). ²⁷ In addition, I have observed a precipitous increase in bone density testing and serum vitamin D testing among low-risk or average-risk perimenopausal women. The bone density testing runs contrary to guidelines that recommend testing beginning at age 60 or 65 in average-risk women; ²⁸ and the appropriateness of routine vitamin D screening has not yet been established. My anecdotal impression—that these practices are more common among gynecologists than among internists—parallels my impression about HT prescribing shortly after the WHI publication in 2002. Empirical verification is obviously needed to confirm these anecdotal observations.

In the end, are studies like that of Lakey et al. ¹⁷ worthwhile? I think they are, for at least four reasons. First, variation in clinical perspectives and practice patterns, both across medical specialties and across geographic regions, ²⁹ are likely to become subjects of increasing attention in healthcare reform. Quality improvement and cost-containment initiatives should work in concert to promote more uniform application of clearly beneficial interventions and to discourage interventions that are unproven, useless, or harmful. Second, patients are entitled to know about systematic differences between the perspectives of medical professionals that might influence the treatment they receive; interprofessional comparative studies help to elucidate those differences. Third, medical educators could use research on variation in the practices and attitudes of physicians as teaching tools. A full appreciation of the possibilities and limits of evidence-based medicine comes not only from biostatistics and research methodology but also from understanding the value judgments and conflicts of interest that affect each clinical specialty. Finally, studies of interprofessional variability can help clinicians to compare their own practices and attitudes with those of their colleagues; hopefully, such self-reflection will result in better patient care.