Effective Survival Prediction and Evaluation for Cancer Patients

Abstract

Background: As cancer is the leading global cause of death, an ongoing challenge is predicting an individual’s cancer progression accurately, to facilitate personalized treatment planning. In deploying individual survival prediction models, a pivotal question emerges: Are we striving to compare survival durations between patients (e.g., ‘Who survives longer between patients A and B?’) or are we endeavoring to estimate a specific patient’s survival time (e.g., ‘How long will patient A survive?’), among other scenarios. This paper addresses this fundamental inquiry and conducts a comprehensive evaluation of such predictive models. Materials and methods: We consider 9 common solid tumors (brain, breast, kidney, liver, lung, stomach, prostate, thyroid, and urinary bladder) using data from the Surveillance, Epidemiology, and End Results program. We employ both conventional and advanced machine learning models that predict individualized survival distributions. We consider several different possible goals of a survival prediction model and connect each goal to a specific evaluation metric. We propose modified versions of the mean absolute error tailored to address a query about a patient’s expected survival duration. Results: Our research involved training multiple models on various cancer types and rigorously evaluating those models using the proposed metrics. We demonstrate that a model might be effective for one goal but ineffective for another, and show that we can determine this based on the measure used. Our findings underscore the importance of selecting an appropriate evaluation measure that is aligned with the primary objective of a study. Conclusion: This work highlights the need for evaluation metrics that are relevant to the research objectives and identifies which objective leads to which evaluation metric. This research sets a path for future research that seeks to further refine predictive models for oncological prognostication. Keywords: survival analysis, cancer study, machine learning, effective evaluation