Abstract
There is a long history of underrepresentation of women and female animals being studied in scientific research, which has resulted in gaps in knowledge and at times, inaccurate clinical recommendations. There is a gradual shift in the mindset of the scientific community on this issue, in part related to policy changes enacted by the National Institute of Health (NIH). Sex must now be accounted for as a biological variable in both basic and clinical research. This review discusses the history of evolving policies on inclusion of sex-informed data in scientific research, and the ways in which epilepsy researchers can approach future studies in a manner that takes sex as a biological variable into account.
Introduction and Historic Context
For historic, cultural, and pragmatic reasons, there is a long history of underrepresentation of women and female animals in scientific research. Important sex differences have been overlooked, resulting in inaccurate clinical recommendations, due to failure to include and analyze scientific data taking sex into account. In the past few decades, there has been a shift in the mentality of researchers to begin to consider the importance of sex as a biological variable. This shift has occurred in part due to policy changes enacted by the National Institutes of Health (NIH). It is not enough to consider sex and gender when conducting research, but to pursue active integration of sex as a biological variable. Integration requires that data be analyzed by sex, and that data be reported in a manner that accounts for sex.
An influential example of the impact of sexual dimorphism on clinical disease presentation was revealed by the Women's Ischemia Syndrome Evaluation study, which found that roughly half of women with angina and cardiac ischemia lack obstruction of coronary arteries, and instead have disease in coronary microvessels. Cardiovascular disease (CVD) remains the leading cause of death in women, and women have higher rates of myocardial ischemia and mortality compared with men (1). Coronary heart disease had been previously thought to arise exclusively from obstruction of large coronary arteries, and this sex difference had not been identified during previous research. This finding was crucial for the improvement of diagnosis and treatment of CVD in women (2).
Historically, clinical research often excluded women as subjects for a multitude of reasons, including the complexity added by female menstrual cycling, the fear of teratogenic effects in women of childbearing age, and the mere fact that women were more difficult to recruit (3). Influenced in part by infamous historic examples of drugs taken that were not yet known to be teratogenic, including thalidomide and diethyl-stilbestrol, by pregnant women, the Food and Drug Administration (FDA) issued a statement in 1977 that “women of childbearing potential should be excluded from earliest dose-ranging studies” (4,5). Although the statement stipulated that women could be included in further studies once additional evidence established drug safety, the statement contributed to a scientific culture in which women were excluded or underrepresented as subjects. The FDA and NIH made statements reversing this practice in 1993, encouraging inclusion of women of childbearing potential in clinical research, and the NIH mandated that studies receiving funding include enough women to be able to detect sex differences (6,7). The statement applied to clinical trials, but did not address basic science research.
Continuing Insufficiencies
Improvements in inclusion of women in clinical research have been slow and incomplete. In 2011, the NIH sponsored a workshop entitled “Sex Specific Reporting of Scientific Research.” Concerns identified by the workshop included the persistent underrepresentation of women in clinical trials, and the paucity of studies that analyze results separately by sex (3,8). The NIH released a statement in 2015 requiring researchers to incorporate sex as a biological variable into research designs and analyses, in humans as well as in vertebrate animals. Required changes took effect impacting applications for NIH funding starting January 25, 2016 (9).
Gaps in scientific knowledge related to sex differences exist in basic science research, and improvements in equitable sex representation have lagged behind clinical research. Animal research has historically focused primarily on male animals for a variety of reasons, including the assumption that results from males will apply to females, that variability in female animals due to cyclical hormones will require larger sample sizes, and that research involving female animals will be more labor-intensive and expensive. A historic review of animal studies from five major science journals published from 1909 to 2009 showed that only one-third of articles included both sexes of animal subjects, and among the subgroup including data from both sexes, only one-third analyzed data separately by sex (10).
Proposed Solutions
In anticipation of changing NIH policies, the NIH Office of Research on Women's Health convened in October 2014 to develop guidelines for researchers on how to incorporate sex as a biological variable into all aspects of basic and clinical scientific studies. Recommendations included incorporating stratified randomization by sex into study design, and assessing for adequate statistical power to detect interaction effects within subgroups (11). Regarding the reporting of clinical trials, the title and abstract should indicate whether a study involved only men or only women. If the study design allows for identification of sex differences, journals should require authors to present these results. If unable to identify sex differences, this should be discussed as a limitation of the study. Studies underpowered to detect sex-stratified main effects should be permitted to make sex-stratified data available, particularly in supplemental materials, to allow for subsequent pooled and meta-analysis (3,11). This data can be provided without making inferences regarding sex differences.
One of the best examples of failure to account for sex differences in epilepsy research can be found in the clinical trials that led to the approval of lamotrigine as a treatment for epilepsy. The 1994 multicenter randomized controlled trial stipulated that women of childbearing age included in the trial use an “approved form of contraception.” Although the original paper did not specify numbers of women taking specific types of contraception, the most common form of contraception used at the time was the oral contraceptive pill (12). Despite the fact that over half of subjects were female, and many women were using exogenous hormones, the study failed to detect a sex difference in lamotrigine concentrations. It was later discovered through small case series that exogenous hormones containing estrogen reduce lamotrigine concentrations by half. When data are analyzed by sex post hoc, subgroup analyses of main effects stratified by sex are often underpowered, which increases the chances of false negative results (Type II error). In retrospect, if the study had been designed to account for sex differences with a priori stratified randomization by sex, it is likely that the difference in lamotrigine serum levels, as influenced by exogenous hormones, would have been detected.
There should be cautious interpretation of sex differences detected by studies that have low pretest expectation of sex differences, because statistically significant differences may arise by chance (Type I error). One could argue that there is a reasonable likelihood that sex-based associations might exist in all epilepsy drugs, and that stratified randomization by sex should be done a priori when designing all future epilepsy medication trials. In cases of inadequate power, raw data should be archived by sex for future pooling and meta-analysis. Drug approval requires multiple trials, creating an opportunity for pooled analyses. A recent example of successful use of pooled data from epilepsy medication trials analyzed effects of everolimus on menstrual irregularities when treating tuberous sclerosis. Approximately 38% of women subjects pooled from three clinical trials experienced menstrual irregularities, most commonly amenorrhea (24.1%) and irregular menses (17%), and most events self-resolved. By pooled analysis, an odds ratio of 8.3 was calculated for menstrual/hormonal events (13). These results contributed to scientific knowledge about safety of everolimus in female patients, while individual studies lacked power to adequately detect sex differences.
Special Considerations for Basic Science Research
It is not always feasible to include an equal number of male and female animals or cells in basic science studies. More appropriately, researchers should specify the sex of animals or cell lines studied. In animal research (most often involving mice/rats), it is additionally useful to stage and report the female rats by phase of the estrous cycle. There have been several studies demonstrating that differences in hormonal states of female rats have effects on behavior. Relevant to epilepsy research, multiple studies have shown that female rats in the proestrous phase (during which estrogen levels peak) have a decreased seizure threshold. Because the rat estrous cycle is so short, it lends itself well to study of behaviors during different hormonal phases to isolate hormonal effects on particular behaviors (14–17). There is also an opportunity in some studies to include classic gonadectomy with or without hormone replacement to assess effects of sex hormones on sex-dependent or -specific processes (15).
Conclusions
There are many ways in which sex differences have clinically relevant effects on disease presentation, diagnosis, and treatment. There is no doubt that sex differences matter. Failure to address sex differences in past epilepsy research has resulted in knowledge gaps and clinically relevant misinformation. As research policies change to promote accountability for presenting sex-informed data, there is a growing need for guidelines on thoughtful integration of sex as a biological variable in every aspect of scientific research, from study design to data collection, analysis, and reporting. Sex is a complicated variable that intersects with gender and culture in humans, but despite and because of its complexities, it needs to be adequately addressed in both basic and clinical research to gain insight and advance scientific discovery.