Improving clinical trial design using interpretable machine learning based prediction of early trial termination

This study proposes using a machine learning pipeline to optimise clinical trial design. The goal is to predict early termination probability of clinical trials using machine learning modelling, and to understand feature contributions driving early termination. This will inform further suggestions to the study protocol to reduce the risk of wasted resources. A dataset containing 420,268 clinical trial records and 24 fields was extracted from the ct.gov registry. In addition to study characteristics features, 12,864 eligibility criteria search features are used, generated using a public annotated eligibility criteria dataset, CHIA. Furthermore, disease categorization features are used allowing a study to belong more than one category specified by clinicaltrials.gov. Ensemble models including random forest and extreme gradient boosting classifiers were used to train and evaluate predictive performance. We achieved a Receiver Operator Characteristic Area under the Curve score of 0.80, and balanced accuracy of 0.70 on the test set using gradient boosting classification. We used Shapley Additive Explanations to interpret the termination predictions to flag feature contributions. The proposed pipeline will lead to an optimised clinical trial design and consequently help potentially life-saving treatments reach patients faster.