Predicting acute diarrhoea in rectal cancer chemoradiotherapy: Secondary analysis of the phase III ARISTOTLE trial.

Abstract

Neoadjuvant chemoradiotherapy is a standard treatment for locally advanced rectal cancer, but acute diarrhoea remains a significant side effect, affecting the completion of chemoradiotherapy treatment. This study aimed to predict acute diarrhoea after neoadjuvant chemoradiotherapy for rectal cancer and further develop a strategic tool to individualise rectal cancer treatment. The ARISTOTLE trial is a phase III trial comparing capecitabine chemo-radiotherapy (CRT) versus capecitabine-irinotecan CRT as a pre-operative treatment for locally advanced rectal cancer. We included 589 trial patients across 73 institutions. The volume of the AI-segmented small bowel receiving at least 10 Gy (V_10Gy) was used alongside the treatment arm, patient age, and performance status in a logistic regression model to predict a more than 2-grade increase in acute diarrhoea toxicity from baseline (ΔG ≥ 2). Finally, based on the prediction, we identified a sub-cohort of patients for whom a viable dose decrease would result in a reduction of toxicity, and conversely, we also identified individuals for whom adding irinotecan may not cause toxicity. The average mean receiver operating characteristic curve (AUROC) for predicting ΔG ≥ 2 is 0.71 [95 % CI 0.58-0.82] on the independent test dataset. Based on the prediction, we identified 71 patients (14 %) who could potentially benefit from irinotecan addition without a dose decrease to maintain ΔG < 2, and 77 patients (15 %) who could potentially benefit from irinotecan addition but need a dose decrease to maintain ΔG < 2. The multi-institutional cohort of 73 centres strengthens the reliability of these findings, demonstrating the model's potential as a strategic tool to individualise rectal cancer treatment while mitigating severe diarrhoea.