Background: A new class of immuno-oncology agents has recently been shown to induce long-term survival in a proportion of treated patients. This phenomenon poses unique challenges for the prediction of analysis time in event-driven studies. If the phenomenon of long-term survival is not accounted for properly, the accuracy of the prediction based on the existing methods may be substantially compromised.
Methods: Parametric mixture cure rate models with the best fit to empirical clinical trial data were proposed to predict analysis times in immuno-oncology studies during the course of the study. The proposed prediction procedure also accounts for the mechanism of action introduced by cancer immunotherapies, such as delayed and long-term survival effects.
Results: The proposed methodology was retrospectively applied to a randomized phase III immuno-oncology clinical trial. Among various parametric mixture cure rate models, the Weibull cure rate model was found to be the best-fitting model for this study. The unique survival kinetics of cancer immunotherapy was captured in the longitudinal predictions of the final analysis times.
Conclusions: Parametric mixture cure rate models, along with estimated long-term survival rates, probabilities of study incompletion, and expected statistical powers over time, provide immuno-oncology clinical trial researchers with a useful tool for continuous event monitoring and prediction of analysis times, such that informed decisions with quantifiable risks can be made for better resource and logistic planning.