The progression-free-survival ratio is a popular endpoint in oncology trials, which is frequently applied to evaluate the efficacy of molecularly targeted treatments in late-stage patients. Using elementary calculations and simulations, numerous shortcomings of the current methodology are pointed out. As a remedy to these shortcomings, an alternative methodology is proposed, using a marginal Cox model or a marginal accelerated failure time model for clustered time-to-event data. Using comprehensive simulations, it is shown that this methodology outperforms existing methods in settings where the intrapatient correlation is low to moderate. The performance of the model is further demonstrated in a real data example from a molecularly aided tumor trial. Sample size considerations are discussed.
Keywords: Weibull distribution; accelerated failure time model; growth modulation index; marginal model; paired time‐to‐event data; progression‐free‐survival ratio.
© 2024 The Author(s). Biometrical Journal published by Wiley‐VCH GmbH.