MRI-guided risk stratification for neoadjuvant immunotherapy in rectal cancer.

Background: Neoadjuvant therapy improves local control and tumor downstaging in locally advanced rectal cancer (LARC) and is now the standard of care. However, reliable tools for identifying patients who may show different response patterns to neoadjuvant immunotherapy-containing treatment, particularly MRI-based risk stratification systems, remain limited. Materials and methods: This retrospective study included 135 patients with locally advanced rectal cancer, who were classified into a neoadjuvant immunotherapy plus chemoradiotherapy group (nICRT, n = 43) and a neoadjuvant chemoradiotherapy group (nCRT, n = 92). The nICRT group received short-course radiotherapy plus CAPOX and sintilimab, whereas the nCRT group received either short-course radiotherapy plus CAPOX or long-course chemoradiotherapy plus CAPOX. All patients underwent baseline pelvic MRI, including high-resolution T2-weighted, diffusion-weighted, and T1-weighted sequences. Evaluated variables included mrT stage, mrN stage, mrEMVI status, mrMRF involvement, tumor length, and clinical laboratory indices. Interobserver agreement was assessed using Cohen's kappa statistics, and factors associated with pathological complete response (pCR) were analyzed using group comparisons and logistic regression analyses. Fisher's exact test was used for subgroup comparisons. Receiver operating characteristic analysis was performed to evaluate predictive performance, and bootstrap resampling was used for internal validation. Results: An MRI-based assessment using mrEMVI, mrMRF, and tumor length ≥ 5 cm yielded a three-factor risk score that predicted pCR with an AUC of 0.835. In the MRI high-risk group (2-3 points), pCR rates were higher with nICRT than with nCRT (7/19, 36.8% vs. 7/60, 11.7%, P = 0.033), whereas in the low-risk group (0-1 points) the difference was not significant (19/24, 79.2% vs. 21/32, 65.6%, P = 0.373). In a sensitivity analysis restricted to the short-course SPRING-01 cohort, the direction of the association remained similar, although statistical significance was not retained. Conclusion: In MRI-defined high-risk patients, nICRT was associated with a higher pCR rate than nCRT. These findings suggest that pretreatment MRI-based stratification may help identify clinically relevant subgroups with different response patterns to intensified neoadjuvant treatment. However, the results should be considered exploratory and require prospective validation before clinical application.