Incorporating Individual Early Response in the Dose-Effect Relationship of Complete Pathological Response Following Neoadjuvant Radiochemotherapy for Rectal Cancer

Purpose: To develop and externally validate a model predicting pathological complete response (pCR) in patients with rectal cancer treated with neoadjuvant radiochemotherapy. The approach combines the classical logit dose-effect curve with individual early response assessed by magnetic resonance imaging during radiochemotherapy. Methods and materials: The dose-response model incorporating the early regression index (ERI) was derived from 2 published data sets. A population-averaged dose-response curve was obtained by fitting data from a recent meta-analysis. Then, an ERI-based model for pCR prediction was fit to 95 patients treated at San Raffaele Hospital; ERI was incorporated as a dynamic dose-modifying factor in the Hill model, based on Equivalent Uniform Dose to 2 Gy and including the time factor. The model was validated on an external cohort of 132 patients treated at Policlinico Gemelli with standard and dose-escalated schedules. Calibration plots, area under the curve, and average precision were used to assess performance. Recalibration refined predictions for the external cohort. Results: The final Hill model showed best-fit values of TD50 = 52.2 Gy, dynamic dose-modifying factor = (0.89 + 0.02 × ERI), and steepness k = 5.91 + 0.11 × ERI. The validation cohort had a pCR rate of 35.6%. Agreement between prediction and observed rates was high (offset, 0; slope, 0.84). Discriminative ability was robust (area under the curve, 0.77; average precision, 0.65 vs 0.356 for baseline). ERI-stratified dose-pCR relationships confirmed predictive value across 4 ERI categories: highly responsive (ERI, 1-6.9; pCR, 65%), moderately responsive (ERI, 6.9-13.1; pCR, 55%), poorly responsive (ERI, 13.1-36; pCR, 14% and 35%), and nonresponsive (ERI, >36; pCR, 0%). Predictions aligned with results using median ERI values per group. Conclusions: The ERI-dose model was validated on an external cohort with distinct radiation therapy regimens. Dose escalation of 8.5 Equivalent Uniform Dose to 2 Gy in moderate-to-good responders corresponds to an increase of approximately 20% in pCR, whereas no benefit was reported in nonresponders. These findings highlight the model's potential for personalizing radiation therapy protocols by optimizing dose escalation based on ERI.