NEJM Editorial Women in Clinical Trials -- A Portfolio for Success Just a decade ago, in September 1990, public and political concern about the underrepresentation of women in major clinical trials, especially trials focusing on cardiovascular disease, resulted in the establishment of the Office of Research on Women's Health by the National Institutes of Health (NIH). The mission of the office was to set an agenda for research on women's health and to ensure that women and members of minority groups would be included in clinical research funded by the NIH. Three years later, the NIH Revitalization Act required that the NIH strengthen its policies to ensure not only that women and members of minority groups would be included in clinical trials but also that, if scientifically appropriate, they would be included in adequate numbers to allow valid analysis of differences according to sex and race or ethnic background in the effects of therapies. By many measures, these efforts have been successful, as indicated in the article by Harris and Douglas in this issue of the Journal. After reviewing all clinical trials on cardiovascular topics that were funded by the National Heart, Lung, and Blood Institute between 1965 and 1998, the authors found that women were enrolled in trials at rates that exceeded the prevalence of cardiovascular disease among women in the general population and that the proportion of women in such trials had increased over time. Even after excluding single-sex trials, the authors found that the enrollment of women in mixed-sex clinical trials was still consistent with the prevalence of cardiovascular disease among women. This finding is in accord with a recent report by the General Accounting Office on women in all NIH-funded studies. Women represented 61.9 percent of all subjects enrolled in NIH-funded extramural trials in 1997; after the exclusion of single-sex protocols, women accounted for 52 percent of the subjects. With regard to specific types of cardiovascular disease, Harris and Douglas found that the rates of enrollment of women in studies of coronary artery disease and hypertension were similar to the prevalence of these diseases in women. In contrast, women were consistently underrepresented in studies of heart failure, with little or no improvement over time. In describing these changes, which are encouraging overall, the authors note that much of the increase in the enrollment of women in clinical trials involving cardiovascular disease has been due to the initiation of two large single-sex trials, both investigating the primary prevention of cardiovascular disease, rather than to across-the-board increases in the enrollment of women in such trials. They conclude that single-sex trials should not be the solution to the systemic problem of the underrepresentation of women in cardiovascular trials. Their caution is appropriate. But the question remains, should single-sex trials be used as one of the strategies in the epidemiologic arsenal for addressing sex-specific issues? To answer this question, we must keep in mind the overarching purpose of increasing women's participation in clinical trials -- to gain better information about women's health -- while considering the two different but complementary goals that underlie the NIH guidelines: the representation of relevant populations and the assessment of possible differences in treatment effects. With respect to the first goal, it is important to ensure that a population at risk for a particular disease or receiving treatment for it is represented in the relevant clinical trials. This applies not only to women but also to elderly persons, members of racial or ethnic minorities, and other populations. Only when this goal is met will clinical decision making for women, for example, be based on direct findings in women rather than on an extrapolation from the findings in men. The second goal, assessing possible differences in the effects of treatment, is equally important from a scientific perspective. It is also more difficult to achieve than the goal of representation. To determine definitively whether an intervention or therapy affects women and men differently, adequate numbers of both men and women must be included in the trial. This issue is not always relevant to a therapy or intervention, as there are often more similarities than differences between the effects of treatment in women and in men. But if the sex-specific effects of an intervention are to be evaluated, then the mere representation of women will not be sufficient. The trial must include a large enough number of women to ensure that the study has the statistical power to evaluate the effect of the intervention in each sex and to compare the magnitude of these effects. The logistic problems in attaining this second goal are substantial. For example, in the late 1980s, an analysis of data from the all-male Physicians' Health Study showed a clear benefit of low-dose aspirin on the risk of a first myocardial infarction, but the numbers of strokes and deaths due to cardiovascular causes were too small to determine the effect on these outcomes. The Women's Health Study, one of the two large, single-sex, randomized studies of cardiovascular disease initiated in the past decade, was designed to evaluate the benefits and risks of low-dose aspirin and vitamin E in the primary prevention of cardiovascular disease (and cancer) in women. Because the base-line rate of cardiovascular disease is so much lower in women than in men, a preventive therapy that could be evaluated in a sample of 22,000 men over the age of 40 years in the Physicians' Health Study required a sample of 40,000 women over the age of 45 years in the Women's Health Study. Even if the Physicians' Health Study had included 20,000 women and 20,000 men, it would not have had adequate statistical power to determine whether low-dose aspirin reduced the risk of cardiovascular disease in women, let alone to determine whether the magnitude of the effect differed between women and men. This point illustrates one of the dangers of conducting a trial with a sample that is too small: it could have not just a null result but an uninformative null result. That is, the lack of a statistically significant effect may be due either to the ineffectiveness of the therapy or to the inability of the study to detect an effect because the sample is too small. Underpowered trials can do great scientific harm if their results are misinterpreted as demonstrating that an intervention has no effect; often, in fact, it may have an effect but the sample is simply too small to determine whether it does. When sex-specific differences are to be evaluated in a trial, the crucial question is how best to achieve this scientific goal. A single approach is unlikely to be effective in all situations, and a variety of approaches must be considered. For example, in the Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial, funded by the National Heart, Lung, and Blood Institute, the investigators enrolled a disproportionately high percentage of black participants in order to be able to evaluate the effect of each therapy in blacks, who are at particularly high risk for hypertension. On the other hand, the Women's Health Initiative combined two approaches, the use of a single-sex trial plus the inclusion of sites with large minority populations, in order to obtain adequate representation of women of different racial and ethnic backgrounds. The choice of the research design must always be driven by good science, and the inclusion of relevant populations must be addressed in developing the research design that is most appropriate to the scientific objectives of the study. The increased participation of women in clinical trials should not lull us into relaxing our efforts in this area. For example, as Harris and Douglas point out, women have historically been underrepresented in clinical studies of heart failure, with little improvement during the past decade. Since slightly more women than men have congestive heart failure, it is important to determine whether there are unique scientific issues or barriers to the enrollment of women that can be addressed in order to obtain direct evidence of the effects of therapies for heart failure in women. In addition, the recent report by the General Accounting Office noted that in the case of non-sex-specific studies, the investigators either did not analyze the data according to sex or omitted the results of such analyses from the findings published in medical journals. The authors of the report concluded, "NIH needs to expand its focus beyond simple inclusion and to ensure that, when it is scientifically appropriate, researchers conducting clinical trials enroll populations and analyze study data in ways that enable them to learn whether interventions affect women and men differently." No single study design is appropriate for answering all questions. Gathering the best information on women's health requires a portfolio of approaches. Both single-sex and mixed-sex trials, if well designed and conducted, will provide much-needed data for use in improving the cardiovascular, and general, health of women. Julie E. Buring, Sc.D. Brigham and Women's Hospital Boston, MA 02215 The New England Journal of Medicine August 17, 2000 Vol. 343, No. 7 Copyright 2000 by the Massachusetts Medical Society. All rights reserved. "http://www.nejm.org/content/2000/0343/0007/0505.asp" janet paterson 53 now / 44 dx cd / 43 onset cd / 41 dx pd / 37 onset pd tel: 613 256 8340 url: "http://www.geocities.com/janet313/" email: [log in to unmask] smail: POBox 171 Almonte Ontario K0A 1A0 Canada