Limitations of Disease X vaccine efficacy and safety clinical trials
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
In a vaccine clinical trial, the candidate is typically accepted or rejected based on predefined efficacy and safety thresholds. An efficacy threshold is the minimum measured treatment effect deemed statistically significant and clinically relevant. A safety threshold is the maximum number of adverse events deemed acceptable in the vaccine arm of the trial. However, the uncertainties and changing conditions met during the emergence of an unknown infectious disease (Disease X) may hinder such simple approaches.
Method
We model the emergence of a Disease X with a SIR (susceptible-infectious-recovered) transmission model. A vaccine is available and the objective is to minimize the total cost over the course of the epidemic, which includes infection, vaccination and adverse event costs. Uncertainties regarding transmission, vaccine, and cost parameters are represented by prior distributions. We simulate placebo controlled efficacy and safety trials, and different vaccination policies. Under policies using trial results, the susceptible population is vaccinated if and only if the vaccine candidate passes both the efficacy and safety tests.
Results
In our baseline scenario and on average over uncertain parameters, using clinical trial outcomes to decide whether to vaccinate the population yields a lower expected total cost compared to indiscriminate emergency vaccination and to no intervention. However, testing both efficacy and safety yields a higher cost compared to testing safety alone. The difference counts in billions of dollar for a population of 10 8 individuals. In our scenario, there is an optimal intermediate safety threshold value that better discriminates vaccine candidates. By contrast there is no optimal intermediate efficacy threshold value. The incidence difference between the treated and control arms is even a misleading measure of vaccine performance.
Conclusion
We show with an example that simple threshold-based decision rules may not be appropriate to discriminate candidates in a Disease X vaccine clinical trial. Uncertainties and the disease dynamics need to be fully taken into account, which may require more advanced pattern recognition methods.
