Systematic review protocol compares interpretable and explainable models for breast cancer recurrence

A systematic review led by Eiman Sahly and colleagues is comparing intrinsically interpretable statistical models, such as logistic regression and Cox proportional hazards models, with explainable artificial intelligence (XAI) approaches for predicting breast cancer recurrence after primary treatment.

The review, registered in PROSPERO under CRD420251145602, aims to critically appraise published models across the black-box, grey-box, and white-box spectrum. The team is currently completing data extraction and seeks feedback on the review question, search strategy, and appraisal tools like CHARMS, PROBAST+AI, and TRIPOD+AI before moving to risk-of-bias assessment.

The authors invite comments from colleagues in breast cancer prognosis, prediction modelling, oncology, or explainable/interpretable AI to help identify methodological gaps and outline future directions for trustworthy prognostic models.