The exploration of treatment effects in patient subgroups has been controversial. Most trials are not designed with sample sizes that are large enough to detect moderate interactions in treatment effects among subgroups or to develop precise estimates of the effects within subgroups. Meta-analysis of completed clinical trials may be useful for exploring these questions with improved statistical power, but often there are no multiple studies that are sufficiently similar to support an informative meta-analysis.
Freedman and colleagues [12
], in their commentary on the 1993 legislation described above, stress the importance of using appropriate methods to compare intervention effects among gender and racial/ethnic subgroups. They stress the possibilities of finding clinically unimportant but statistically significant differences, and vice versa
. In fact, they go on to argue against designing trials with sufficient power to detect treatment by subgroup interactions in the absence of a priori
evidence that such subgroup differences might exist. (How one generates this a priori
evidence is left unclear.) Clinical trialists recognize that the requirement that all studies should be powered statistically to detect treatment X gender (or other) interactions would make trials infeasible. Freedman et al
. also suggest that meta-analysis of multiple trials is the best way to obtain reliable information about subgroup differences.
Differences in treatment effect across subgroups of patients do not all have the same implications. The type of difference that might cause the most concern would be a directional difference: one group benefits from the treatment while the other is harmed. Such differences are rare; when they have been occasionally suggested they have generally not been supported by other data [13
]. It would also be important to be aware of a difference in magnitude of effect if the difference was substantial, as it would affect risk-benefit considerations. There are a few examples of this phenomenon. One recent example, of a heart failure drug that appeared effective in African Americans but ineffective in whites, remains controversial but is supported by data from multiple studies [16
The NIH Office of Research in Women's Health maintains data on the inclusion of women in NIH-funded trials. The most recent report shows that women comprise approximately 55% of clinical trial participants [18
]. Overall numbers, however, may be less informative than they seem. For example, although the annual incidence of breast cancer is about the same as that for prostate cancer, enrollment in breast cancer trials is much higher than in prostate cancer trials [19
]. Many other diseases are not gender neutral; more women get autoimmune diseases; more boys are diagnosed with ADHD. When there are exciting new molecules to study in a given medical area, trials in this area may have a greater advantage, thereby contributing to an artifactual gender imbalance in trial participants. These confounding issues, which may have their own time trends, make it difficult to assess the extent to which men and women are proportionally represented in trials of disease affecting both genders.
In summary, controversy over the inclusion of women in clinical trials has been motivated, in part, by theoretical concerns about gender differences in the effect of the treatment and, in part, by legitimate fears of exposing fetuses to investigational drugs. There is no question that some treatments do work differently in men and women, but the proportion of treatments for which men and women respond very differently is unknown. The broader issue really centres on biological factors, possibly defined by genes or gene expression, that may directly or indirectly modify the effect of specific treatments on specific individuals. Whether the current explosive interest in genetic profiling will ultimately lead to the medical nirvana of personalized medicine, that many have predicted, remains to be seen.