Performance of different clinical trial designs to evaluate treatments during an epidemic

In the 2013-2016 west Africa outbreak of Ebola Virus Disease (EVD), most of the planned clinical trials failed to reach a conclusion within the time frame of the epidemic. The performance of clinical trial designs for the evaluation of one or more experimental treatments in the specific context of an ongoing epidemic with changing case fatality rates (CFR) and unpredictable case numbers is unclear. We conduct a comprehensive evaluation of commonly used two- and multi-arm clinical trial designs based on real data, which was recorded during the 2013-16 EVD epidemic in west Africa. The primary endpoint is death within 14 days of hospitalization. The impact of the recruitment start times relative to the time course of the epidemic on the operating characteristics of the clinical trials is analysed. Designs with frequent interim analyses with the possibility of early stopping are shown to outperform designs with only a single analysis not only in terms of average time to conclusion and average sample size, but also in terms of the probability of reaching any conclusion at all. Historic control designs almost always result in substantially inflated false positive rates, when the case fatality rate changes over time. Response-adaptive randomization may be a compromise between the goal of scientific validity and the ethical goal of minimizing the number of patients allocated to ineffective treatments.