Predictive value of primary tumour MRI regression grade on outcomes in locally recurrent rectal cancer

This study aimed to investigate whether radiological response of the primary rectal tumour, assessed using magnetic resonance tumour regression grade (mrTRG) following neoadjuvant chemoradiotherapy, is associated with oncological outcomes in patients with locally recurrent rectal cancer (LRRC) undergoing neoadjuvant treatment.

All patients with LRRC who underwent neoadjuvant treatment for both the primary and recurrent tumour with complete MRI imaging between 2010-2025 were retrospectively analysed. All primary and recurrent MRI scans underwent central reassessment by expert radiologists. Primary outcomes were overall survival (OS), disease-free survival (DFS) and local re-recurrence-free survival (LRFS). Secondary outcomes included associations between primary-tumour mrTRG, resection margin (R0/R1) and LRRC mrTRG.

Among 121 patients, primary-tumour response was classified as good (mrTRG 1-2) in 50 (41%) patients, intermediate (mrTRG 3) in 43 (36%), and poor (mrTRG 4-5) in 28 (23%). Using a two-tier grading system (mrTRG 1-2 vs mrTRG 3-5), poor versus good primary-tumour response yielded HRs of 1.07 (95% CI 0.69-1.65) for OS, 1.17 (0.76-1.82) for DFS, and 1.17 (0.76-1.82) for LRFS. After multivariable adjustment, corresponding HRs were 1.00 (0.63-1.60), 0.92 (0.57-1.48), and 1.31 (0.77-2.25), respectively. Binary logistic regression yielded an OR of 1.46 (95% CI 0.63-3.38, p = 0.378) for R1 resection in patients with a poor primary-tumour response. Ordinal logistic regression demonstrated an association between primary and recurrent tumour mrTRG response (p = 0.012), although the predictive value of the model was limited (Nagelkerke R² = 0.085).

Primary-tumour response following neoadjuvant chemoradiotherapy did not demonstrate sufficient prognostic value to support its use for clinical decision-making after development of LRRC.