Mid-treatment MRI-based tumor volume reduction rate as a continuous prognostic factor after chemoradiation for...

This study aimed to quantify the continuous association of the mid-treatment MRI-based tumor volume reduction rate (TVRR) with oncologic outcomes after chemoradiation for cervical cancer and to develop prognostic models assessed by two-center internal-external cross-validation (IECV). A retrospective analysis was performed in patients with FIGO 2009 stage IB-IVA disease treated with definitive chemoradiation at two centers from 2010 to 2021 with both pre- and mid-treatment pelvic MRI during external beam radiotherapy (24-36 Gy). Pre- and mid-treatment volumes (Vpre, Vmid) were derived and TVRR was defined as (Vpre - Vmid)/Vpre × 100%. Cox models were used to evaluate overall survival (OS) and progression-free survival (PFS), and a Fine-Gray model was used for local control (LC), modeling TVRR with a penalized spline and adjusting for key covariates. IECV assessed discrimination and 60-month calibration for OS and PFS. The analysis included 281 patients (median age 59 years; squamous cell carcinoma 85%; FIGO 2009 stage III-IV 60%; node-positive 56%) with 65 deaths, 97 PFS events, and 27 local failures over a median follow-up period of 69.6 months. Higher TVRR was independently associated with better outcomes: per 10-percentage-point (pp) increase, OS hazard ratio (HR) 0.71 (95% CI, 0.64-0.79; P < 0.001), PFS HR 0.74 (0.68-0.81; P < 0.001), and LC subdistribution HR 0.68 (0.58-0.80; P < 0.001). Non-linear components were not significant. Discrimination was moderate-high (C 0.73-0.81 internally; 0.72-0.79 in IECV) with acceptable 60-month calibration. In conclusion, mid-treatment MRI-based TVRR provides a strong, near-linear prognostic signal and enables threshold-free, calibrated risk estimation and supports design of risk-adapted trials.