Diagnostic Performance and Misclassification Patterns of Preoperative MRI in Rectal Cancer: A Real-World Study

Le résumé fourni par la source

Introduction: Magnetic resonance imaging (MRI) is the reference imaging modality for locoregional staging and restaging of rectal cancer (RC). However, its agreement with surgical pathology in real-world practice is limited. We aimed to assess the agreement and diagnostic performance of preoperative MRI for dichotomized T and N staging in RC. Secondarily, we explored the direction of MRI misclassification and potential preoperative factors associated with discordance. Methods: We conducted a retrospective real-world study on 152 consecutive patients with pathologically confirmed RC who underwent surgery between September 2019 and June 2025 in our institution. Two cohorts were analyzed separately: patients treated without neoadjuvant therapy (non-NAT, n = 70) and patients treated with NAT followed by restaging MRI and surgery (NAT, n = 82). The main staging outcomes were dichotomized into T0-T2 vs. ≥T3 and N0 vs. N+, using final pathology as the reference standard. Agreement, Cohen’s kappa, sensitivity, specificity, predictive values, McNemar’s test, and exploratory regression analyses for misclassification were performed. Results: In the overall cohort, agreement was 72.4% for T staging and 73.0% for N staging, with moderate agreement for T (kappa = 0.452) and fair-to-moderate agreement for N (kappa = 0.349). Sensitivity and specificity were 80.3% and 67.0% for T staging and 54.5% and 80.6% for N staging, respectively. T-stage errors were mainly associated with overstaging. In NAT-treated patients, baseline MRI showed markedly poorer agreement with final pathology than restaging MRI, particularly for T stage (45.1% vs. 72.0%). Exploratory analyses did not identify strong or reproducible predictors of misclassification. Conclusions: This real-world study provides a contemporary estimate of MRI-pathology agreement for dichotomized T and N staging in routine RC care. Agreement was moderate, and performance was more consistent for advanced T-category assessment than for nodal staging. These findings support MRI as a practical tool for multidisciplinary risk stratification and highlight the need for continued monitoring of MRI usage and performance in clinical practice.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.