Early regression index-guided adaptive dose escalation in locally advanced rectal cancer: results of the prospective THUNDER-2 trial.

INTRODUCTION: Complete response (CR) after neoadjuvant chemoradiotherapy (nCRT) in locally advanced rectal cancer (LARC) is associated with improved outcomes and may enable organ-preserving strategies. The Early Regression Index (ERI) is an imaging biomarker developed to predict treatment response. THUNDER-2 trial investigated ERI-guided magnetic resonance imaging (MRI)-based adaptive dose escalation to improve CR rates in predicted non-responders. MATERIALS AND METHODS: Patients with LARC underwent MRI-guided nCRT with concurrent capecitabine or 5-fluorouracil. ERI was calculated between simulation and the 10th treatment fraction. Patients with ERI ≤ 13.1 continued standard radiotherapy, whereas those with ERI > 13.1 received an adaptive boost to 60.1 Gy using online adaptive radiotherapy. CR was defined as pathological complete response (pCR) or sustained clinical complete response (cCR) at 2 years. Patients with cCR were considered for non-operative management with intensive follow-up. RESULTS: Sixty-two patients were enrolled (48.4% responders, 51.6% non-responders). Adaptive dose escalation increased the CR rate in non-responders from the historically expected value of approximately 3% to 21.9%, while maintaining acceptable toxicity. Responders achieved significantly higher TME (total mesorectal excision)-free survival than non-responders (38.7% vs 8.3%; hazard ratio (HR) 2.35, 95% confidence interval (CI) 1.29-4.26; p = 0.004). Two local regrowths occurred among non-responders and were successfully salvaged. ERI was the only variable significantly associated with CR in both univariate and multivariable analyses (adjusted odds ratio (OR) 0.912, 95% CI 0.840-0.990; p = 0.028). CONCLUSION: ERI-guided MRI-based adaptive dose escalation was feasible and associated with higher-than-expected CR rates in predicted non-responders, supporting personalized treatment intensification and organ-preservation strategies in LARC (THUNDER-2; NCT04815694).