Prediction Models for Pathological Response to Neoadjuvant Chemoradiotherapy in Locally Advanced Rectal Cancer: A Systematic Review and Meta-Analysis

Abstract

Simple Summary Accurately predicting which patients with locally advanced rectal cancer (LARC) will achieve a pathological complete response (pCR) after preoperative chemoradiotherapy may help risk stratification and trial design. However, pCR is a postoperative pathological endpoint, whereas organ-preserving “watch-and-wait” decisions depend on clinical complete response (cCR). These endpoints are not interchangeable. We conducted a systematic review and meta-analysis of 43 studies covering 67,302 patients. Pooled estimates were numerically highest for deep learning models, followed by magnetic resonance imaging (MRI) radiomics, combined approaches, gene expression signatures, and clinical factors alone. However, between-study heterogeneity was substantial, publication bias was evident, and external validation performance was lower than internal validation performance. These findings should be interpreted cautiously because area under the receiver operating characteristic curve (AUC) alone does not establish clinical utility. Prospective, multicentre validation studies reporting sensitivity, specificity, predictive values, likelihood ratios, calibration, and decision curve analyses are urgently needed.