the law and policy apply only to NIH-funded studies, not to studies done by or supported by other agencies or entities. In addition, NIH cannot require editors and journals to mandate inclusion of analysis by sex in reports of studies. As a result, key health data are not reaching other researchers and the public.
To begin to address that situation, ORWH established a working group of scientific-journal editors as an ad hoc subgroup of the Advisory Committee on Research on Women’s Health. In 2001, the group issued a statement calling on scientific journals to require that, where appropriate, clinical and epidemiologic studies be analyzed to see whether there is an effect of sex; if there is no effect, that should also be reported. Any statistical limitations of such analyses should be made clear. To date, however, very few journals have adopted such a policy. Clayton cited the Journal of the National Cancer Institute (JNCI) as an example of journals that address sex-specific analysis in their instructions for authors.
The continuing challenge, Clayton concluded, is to get sex-differences research accomplished and the results reported, from basic through applied research.
Ameeta Parekh, director of research and development in the Office of Women’s Health of the U.S. Food and Drug Administration (FDA), reminded participants that the severe birth defects associated with thalidomide use by pregnant women in the 1960s led to a conservative approach to testing of new drugs in women. In 1977, FDA issued General Considerations for the Clinical Evaluation of Drugs, which stated that “women of childbearing potential should be excluded from the earliest dose-ranging studies.” Although the guidance went on to state that such women could be included in further studies if additional evidence had been amassed on the safety or preclinical toxicity of a drug, that exclusion inadvertently led to the underrepresentation or exclusion of women from all clinical trials. The exclusion of women from clinical research was not generally questioned, because sex was not recognized as a variable in health research and was not considered to be a factor that could affect health and illness. In addition, investigators believed that women were more difficult to study because they introduced more variables (for example, hormonal cycles) and were difficult to recruit. The result of not studying women is gaps in our knowledge and understand-
ing of the differences between men and women with regard to treatments and response.
Carolyn Clancy, director of the Agency for Healthcare Research and Quality (AHRQ), added that the first randomized trial of estrogen to prevent heart disease was conducted in the early 1960s in men. The effect of estrogen on heart disease in women was not studied in a randomized trial until the Women’s Health Initiative, 35 years later.
Martha Nolan, vice president for public policy of the Society for Women’s Health Research (SWHR), said that there is a great need to identify biologic and physiologic differences between men and women and to understand the implications of the differences for diagnosis and treatment. She noted, for example, that more women than men take antidepressants; women respond more slowly and are less likely to achieve an optimal response to treatment for depression; and women are more likely to stop using the medication because of adverse events. There are many other examples of differences that are not fully understood. Female athletes, especially those in contact sports, sustain a higher percentage of concussions during play than male athletes do, but virtually all the literature and mass-media attention is on male football and ice-hockey players. Transplantation of donor organs from females is less successful than transplantation from males. Boys are more likely than girls to receive a diagnosis of peanut allergy early in life, but by the age of 24, more women than men are receiving the diagnosis.
Parekh provided further support for the need to study both sexes. Women make up more than 50% of the U.S. population (50.7% according to the 2010 U.S. Census) and on the average outlive men (80.7 years vs 74.8 years). Many diseases place a heavier burden on women than on men (consider, for example, heart disease, cancer, rheumatoid arthritis, lupus, and osteoporosis); however, treatment guidelines are based largely on data on men. Women also rely more on medical systems than men do and are likely to seek treatment sooner.
Jesse Berlin, vice president of epidemiology at Johnson & Johnson Pharmaceutical Research and Development, said that sex-specific reporting helps to define the most appropriate population for treatment and to determine whether benefits or harms differ by sex. Differences between the sexes are more than just pharmacokinetic, however. For example, Berlin cited a recent report that describes sex-specific differences in cell regulatory processes (Mittelstrass et al., 2011).
Sex-specific analysis and reporting are not just “women’s health” issues. Better data on women would be better data for everyone, Clancy said. Sex-specific data could allow guidelines to be more specific and allow clinicians to better tailor care to individuals.
Speakers also presented examples of the importance of sex-specific differences. One example of critical differences between males and females is drug-induced electrocardiographic changes. Parekh explained that several drugs withdrawn from the market were associated with prolongation of the QT interval (a measure of cardiac repolarization) and torsades de pointes (a potentially fatal form of polymorphic ventricular tachycardia). Women have a longer baseline QT interval and a higher propensity for drug-induced QT prolongation, and they are two to three times more likely to develop torsades than men. The effects of drugs being studied for cardiotoxicity, Parekh said, need to be looked at and understood in both men and women.
A more recent example of the importance of sex-based data is A Diabetes Outcome Progression Trial (ADOPT), a randomized controlled trial (RCT) that compared rosiglitazone with metformin and glyburide over several years. The overall fracture rate associated with rosiglitazone use was higher than that associated with glyburide and metformin, but analysis by sex showed that women had a rate of fractures twice that of men (Kahn et al., 2008). As a result, the label for rosiglitazone includes data on the increased fracture risk for women.
A 1992 GAO review of FDA policies and pharmaceutical-industry practices found that women were not adequately included in clinical studies and that data were not analyzed for sex differences with any consistency, and that consequently there was a lack of understanding of sex differences (GAO, 1992). As a result, Parekh said, FDA issued several new guidance documents and regulations. The 1977 policy that mentioned exclusion of women of childbearing potential was reversed through the 1993 guideline Study and Evaluation of Gender Differences in the Clinical Evaluation of Drugs, which recommended collection and analysis of data on sex differences in effectiveness, adverse effects, pharmacokinetics, and pharmacodynamics.
The 1998 investigational new drug (IND) application and new drug application (NDA) regulation, also called the demographic rule, requires NDA submissions to provide safety and effectiveness data and IND submissions to tabulate numbers of participants according to age,
race, and sex. In 2000, FDA issued the clinical-hold rule, which permits FDA to stop IND studies of treatments for life-threatening diseases if women are excluded because of their reproductive potential.
Parekh noted that data reported in poster sessions at a recent Drug Information Association meeting indicated that analysis of safety and efficacy data by sex has been increasing—around 75% of clinical trials in 2007–2009 reported analysis by sex—and a review of approved product labels found that nearly all included pharmacokinetic information by sex.
With regard to reporting in the literature, Nolan said that a decade ago the NIH ORWH, in collaboration with SWHR, convened a meeting of scientific-journal editors to discuss the development of specific instructions for authors and reviewers about the analysis of clinical-trial data by sex. However, in an informal survey of 11 science journals2 conducted by SWHR in 2010, only JNCI and Circulation required reporting of sex differences; the others did not set any sex-specific requirements for authors.
Nolan cited several recent articles that draw attention to the need to consider sex differences. In March of 2010, an article in Science reported on sex bias in animal models and predicted that reporting would change if journals adopted a common set of guidelines for manuscripts to provide details on the sex of the animals used and required authors to state their rationale for studying only one sex and the implications of not studying the other (Wald and Wu, 2010). A June 2010 editorial in Nature suggested that funding agencies should require researchers to justify sex inequalities in grant proposals and should favor proposals that include both sexes; that FDA should ensure that physicians and the public are aware of sex differences in drug reactions and dosages; and that medical schools should train physicians in how diseases, symptoms, and drug responses can differ by sex (Putting gender on the agenda, 2010). The editorial also noted that Nature was considering whether to require authors to document the sex of animals in published papers. Finally, an article in the New England Journal of Medicine in June 2010 noted how the global H1N1 influenza pandemic disproportionately affected pregnant women and stressed the need for inclusion of pregnant women in clinical trials (Goldkind et al., 2010).
2Journal of the National Cancer Institute (JNCI), Circulation, JAMA, New England Journal of Medicine, Endocrinology, American Journal of Physical Medicine & Rehabilitation, BMJ, Lancet, Immunology, Gastroenterology, and Urology.
The Department of Defense and the Department of Veterans Affairs are beginning to examine sex differences, such as how psychologic and physical health conditions affect female soldiers and veterans, Nolan said. These agencies are also reporting when research shows no difference between the sexes.
Great strides have been made in raising public awareness about sex-based differences in cardiovascular, muscular, skeletal, and behavioral health and disease, but only rarely are medical-care options tailored to the patient’s sex, Nolan said. She suggested that it could take less time for research to be translated into medical practice if major journal publishers required analysis by sex and reporting of differences found or the lack thereof.
There are both technical and political barriers to advancing knowledge of sex differences. Clancy described an imbalance between the fear of not knowing what the health-related differences between men and women are and the fear that identifying such differences is somehow impolitic or inappropriate. When the fear and concern associated with not knowing overpower concerns about the influence of politics on science, studying sex differences will become straightforward, she said.
There are methodologic challenges to studying population subgroups, such as males and females. A primary issue in breaking down data by sex is sample size. Berlin asked, Are two separate, adequately powered studies, one in each sex, needed? Or can a single study have sufficient statistical power to detect interaction? Separate studies of men and women risk confounding. Separate studies of men and women might use different doses as in, for example, studies of aspirin and myocardial-infarction prevention. It could then be difficult to tell whether differences in outcome were due to different doses, sex, or other factors. Instead, conducting two studies, each with both men and women, might allow stratification of both studies by sex and provide replication for sex-specific findings. Alternatively, meta-analytic principles could be applied to a program of development and testing.
A barrier to meta-analysis is availability of data. Clancy noted that the opportunity to conduct meta-analyses often rests on the goodwill of investigators in sharing data from clinical trials. As data collection has moved from paper to electronic form, the technical barriers to data-sharing have diminished. The unanswered question is who owns the data, particularly when studies have been funded with taxpayer dollars. With
regard to sample size, meta-analysis of clinical data can be a valuable exercise before investment in a large clinical trial—it can help in designing trials strategically.
Although the focus of the workshop was on sex-specific analysis and reporting, some panelists pointed out that race and ethnicity may also be clinically relevant, as may other clinically, genetically, or socially defined characteristics. Berlin cited Freedman and colleagues (1995), who discussed the possibility of finding clinically unimportant but statistically significant differences or clinically important but statistically nonsignificant differences and argued against separate results in the absence of a priori evidence of subgroup differences. Berlin argued, however, that such clinical-trial results can point to basic science and the needs for further elucidation.
Clancy referred participants to a 2009 Kaiser Family Foundation report, Putting Women’s Health Care Disparities on the Map: Examining Racial and Ethnic Disparities at the State Level.3 The principles being discussed in the present workshop do not refer only to definitions of gender and sex but extend to other population groups as well, she stressed.
Parekh highlighted several current FDA initiatives, including one focused on standardizing the data that are electronically submitted to FDA so that analysis of data on women and other populations is easier.
Clancy raised the concept of a learning health care system whereby medical knowledge is advanced by making use of the substantial amounts of data and other information collected every day in the provision of health care. The implementation of electronic health-record systems is a key component of a learning health care system. Many professional societies and other organizations have created patient-level registries, which offer another method of collecting data. Clancy added that AHRQ is using American Recovery and Reinvestment Act funds designated for comparative-effectiveness research and patient-centered outcomes research to develop “a registry of registries” that will be