Feed

This study aimed to develop and validate an integrated model combining pretreatment hemoglobin, albumin, lymphocyte, platelet (HALP) score and T2-weighted magnetic resonance imaging radiomics for predicting pathological complete response (pCR) and overall survival in patients with locally advanced rectal cancer (LARC) undergoing neoadjuvant chemoradiotherapy. A total of 137 consecutive LARC patients from January 2019 to January 2022 were retrospectively included. Pretreatment 3.0T T2-weighted high-resolution magnetic resonance imaging scans were performed. Four stable features were selected using least absolute shrinkage and selection operator regression. The HALP score was calculated as (hemoglobin × albumin × lymphocytes)/platelets, with an optimal cutoff of 48.4 determined by X-tile analysis. Three predictive models were constructed: a radiomics-only model (RAD), a HALP-only model (HALP), and a combined model (RAD + HALP) using a random forest classifier. Model performance was assessed using the area under the curve (AUC), sensitivity, specificity, calibration curves, and decision curve analysis. The pCR rate was 13.1% (18/137). For pCR prediction, the combined model achieved an AUC of 0.808 (95% confidence interval: 0.73-0.88), compared to 0.788 for the RAD model. The HALP score alone did not show predictive capability (P = .59). For overall survival prediction, the combined model significantly outperformed the individual models, with an AUC of 0.957 (sensitivity: 0.945, specificity: 1.0; DeLong test, P = .002), the AUC was corrected to 0.852 (sensitivity: 0.788, specificity: 0.889; P < .001). Multivariate Cox analysis identified HALP as an independent prognostic factor (hazard ratio: 5.83, 95% confidence interval: 2.23-15.2). The combined RAD + HALP model demonstrated excellent and internally validated performance in predicting both short-term treatment response and long-term survival in LARC patients undergoing neoadjuvant chemoradiotherapy patients. Prospective multicenter studies are warranted prior to clinical application.