Incorporation of sex and gender into research study design: a necessary and often underappreciated perspective in clinical research

The incorporation of sex and gender into research study design in both experimental and observational studies is often ignored leading to potential bias in results. In conducting statistical analysis, age is frequently considered an important predictor variable for a particular outcome of interest e.g. the effects of low dose aspirin for prevention of CVD, however, sex and gender identity are often excluded, resulting in an overgeneralization of the outcomes to the entire population at large while marginalizing the effects on other subpopulations such as women, children, and race-gender minorities. (1) This effect modification can even result in poor health outcomes for the excluded populations. Recent studies have utilized propensity score matching as a means of capturing variability in these subpopulations and Cox-proportional hazard modeling to control for confounding and investigate interactions. (2) However, more sophisticated study designs are still needed utilizing larger sample sizes sufficiently powered to detect differences in sex, which can require a fourfold increase in sample size compared to those to detect main effects alone. (3) Furthermore, the direct costs associated with recruiting larger sample sizes is a detrimental factor to the implementation of research designs with adequate statistical models to detect important differences in sex, gender, and race that may significantly impact health outcomes. The objective of this presentation will be to demonstrate through a series of case-studies from the literature the potential negative health outcomes of excluding sex and gender as covariates in summary statistical analysis.

Conclusions: The accommodation of sex and gender into research study design and statistical analysis may preclude higher costs, however, will diminish the risks of adverse health outcomes associated with their exclusion.