Prognostic value of quantitative multiparametric magnetic resonance imaging parameters and histopathologic features for disease-free survival in locally advanced non-mucinous rectal adenocarcinoma

Abstract

Objectives

To develop a postoperative risk stratification model integrating baseline T2 mapping and histopathologic factors for predicting disease-free survival (DFS) in patients with locally advanced non-mucinous rectal adenocarcinoma after neoadjuvant chemoradiotherapy (NCRT) and surgery.

Methods

This retrospective study included 116 patients with non-mucinous rectal adenocarcinoma who underwent NCRT followed by surgery. Baseline T2 values and apparent diffusion coefficient (ADC) values were measured. Prognostic factors for DFS were assessed using Cox proportional hazards regression. Given the limited number of DFS events, the multivariable model size was prespecified, with baseline pretreatment T2 forced into the model. To assess the incremental prognostic value of T2, a clinicopathologic model was compared with a full model additionally including baseline pretreatment T2. The full model was refitted in the entire cohort and presented as a nomogram. Internal validation was performed using 1000 bootstrap resamples. Model performance was assessed by Harrell’s concordance index (C-index), bootstrap-estimated calibration slope, bootstrap calibration plots, time-dependent receiver operating characteristic analysis, and decision curve analysis.

Results

During a median follow-up of 51 months after surgery, 35 of 116 patients (30.2%) experienced DFS events. In multivariable analysis, EMVI positivity (HR 2.42, 95% CI 1.21–4.68; p = 0.013), ypN-positive status (HR 4.37, 95% CI 2.07–9.92; p < 0.001), and perineural invasion (HR 3.97, 95% CI 1.90–8.29; p < 0.001) were independently associated with worse DFS, whereas a higher baseline pretreatment T2 value was independently associated with better DFS (HR 0.88, 95% CI 0.82–0.94; p < 0.001). The full model included baseline pretreatment T2, EMVI, ypN status, and perineural invasion. After refitting in the entire cohort, the model showed an apparent C-index of 0.796 and an optimism-corrected C-index of 0.780 after 1000 bootstrap resamples. The bootstrap-estimated calibration slope was 0.914, and calibration plots showed reasonable agreement between predicted and observed 3-year and 5-year DFS probabilities.

Conclusion

A postoperative risk stratification model integrating baseline quantitative T2 mapping with postoperative histopathologic factors showed potential value for estimating DFS risk in locally advanced non-mucinous rectal adenocarcinoma. Lower baseline T2 values may reflect a more aggressive tumor microenvironment characterized by greater cellularity and denser fibrotic stroma, which may partly explain their association with worse DFS. After bootstrap internal validation, the model demonstrated acceptable discrimination and reasonable calibration, although external validation is required before clinical application.