Improving Surrogate Endpoints for Survival Prediction Through Integration of Patient-Reported Outcomes

Overall survival (OS) remains the gold standard for oncology drug approval, but measuring it requires long follow-up and is impractical in certain oncology settings. To expedite drug evaluation, radiographic endpoints such as progression-free survival (PFS) and overall response rate (ORR) are commonly used as alternatives to OS. However, these measures do not always reflect meaningful survival benefits, limiting their reliability as surrogate endpoints for regulatory decisions. Patient-reported outcomes (PROs), including quality of life (QoL) measures, capture the patient’s perspective on efficacy and toxicity, may enhance prediction of OS when combined with radiographic endpoints. We reviewed FDA oncology New Drug Registry from 2017 to 2023. Trials were included if they reported OS, at least one radiographic endpoint, and patient-reported QoL. We trained machine learning models to predict whether OS is improved in the trial, first using radiographic endpoints alone, and then adding QoL. Model performance was compared between these approaches. Of 387 oncology trials screened, 89 met inclusion criteria. Among these, 57 (64%) showed OS benefit, 76 (85%) reported improvement in the radiographic measures, and 35 (39%) had QoL benefits. Incorporating PROs into the model improved the predictive, with a 20% relative increase in AUC-ROC, indicating substantially better discrimination between trials with and without OS benefit. These findings suggest that incorporating PROs into traditional radiographic endpoints enhances prediction of OS benefit. A composite surrogate endpoint that combines radiographic and QoL measures may provide a more reliable and patient-centered framework for regulatory decision-making and help accelerate drug development. Overall survival (OS) is the gold standard for evaluating cancer therapies, but its reliance on long-term follow-up delays trial readouts and drug approvals. Commonly used alternatives, such as progression-free survival, provide radiographic signals of tumor progression, but they do not consistently correlate with OS, limiting their value as surrogate endpoints. Patient-reported outcomes (PROs) capture quality of life and symptom burden directly from the patient’s perspective. In this study, we show that incorporating PROs with radiographic endpoints substantially enhances the prediction of OS benefits in clinical trials, with machine learning models demonstrating about 20% better predictive performance compared to using radiographic measures alone. These findings support a novel composite surrogate endpoint that integrates PROs with radiographic measures, providing more patient-centered framework for regulatory decision-making and potentially accelerating oncology drug development.