Clinical Trial Designs that Utilize Historical Controls in the setting of Cystic Fibrosis

Abstract

In certain rare disease settings, traditional randomized controlled trials face ethical and feasibility challenges that can exclude their use. Adequate participant recruitment can be prohibitively difficult, and a large placebo control arm could unethically expose vulnerable participants to risk. Development of new therapeutics for these diseases still require regulatory grade evidence of a treatment effect, and researchers must seek alternative but robust trials designs to meet these demands. Historically controlled trials are one potentially statistically efficient approach that can reduce needed sample sizes, and their applicability can be well demonstrated in cystic fibrosis (CF) research. Statistical methods for historically controlled trials include, among others, inverse probability weighting (Rosenbaum 1983), propensity score-based power priors (Lin 2017), and hierarchical Bayesian modeling with commensurate priors (Hobbs 2012). Additionally, an adaptive study design method for incorporating historical controls was proposed by Psioda in 2018. We assessed these four approaches in a large simulation study to quantify bias, coverage, precision, and power. All approaches performed well in terms of type I error and power even when there were different covariate distributions between the simulated historical and active trials. However, notable strengths and weaknesses of each were observed. We then reanalyzed two sequentially completed CF trials of treatments for pulmonary exacerbations to test whether the later trial could have been historically controlled with data from the prior trial using these methods. Specifically, we assessed whether the earlier trial data could be used to reduce the size of the concurrent control group in the later trial. This hypothetical historically controlled trial achieved similar results to the real trial results, establishing proof of concept for the validity of these methods in CF. We conclude that the later trial could have been conducted with a smaller concurrent control group without sacrificing precision and that historically controlled trials have great potential for use in future CF clinical trials.