Introduction and Purpose: Neoadjuvant chemoradiotherapy (NCRT) followed by surgery is the standard treatment for locally advanced rectal cancer (LARC). While this approach has significantly reduced the local recurrence rate to 5-10%^1,2, the occurrence of distant metastasis (DM) remains a challenge^3,4. This underscores the need for a risk stratification model for LARC patients before treatment to enable the development of personalized neoadjuvant therapy. This study aimed to develop and validate a nomogram based on radiomics, pathomics, and biopsy-adapted immunoscore (IS_B) for predicting distant metastasis-free survival (DMFS) in LARC patients after NCRT.
Materials and Methods: This retrospective study included 201 LARC patients (142 men, 59 women; mean age 54.5±10.6 years; range 23-79 years) who underwent NCRT and surgery. The median follow-up duration was 91 months (interquartile range, 82–113 months). Radiomics features were extracted from the gross tumor volume using T2-weighted images and apparent diffusion coefficient maps. Pathomics features including global pattern (features of the entire image) and local pattern (features of the tumor nuclei) were extracted from whole-slide images of hematoxylin and eosin-stained biopsy specimens using Cellprofiler software. IS_B was calculated from the densities of CD3+ and CD8+ T cells in the tumor region using immunohistochemistry on biopsy specimens. CD3+ and CD8+ T cell densities in the tumor region of each patient were compared to that obtained in all patients and converted into percentile. Then, the mean of the two percentiles (CD3+ and CD8+ T cells) was translated into one of the three IS_B categories: Low (0–25%), Intermediate (> 25–70%), and High (> 70–100%)⁵. Feature selection and signature construction were performed using the least absolute shrinkage and selection operator (LASSO)-Cox analysis. The nomogram's performance in predicting DMFS was assessed using the concordance index (C-index) and area under the receiver operating characteristic curve (AUC) of the time-independent receiver operating characteristic (ROC) curve.
Results: In the final feature selection with LASSO-Cox analysis, seven radiomics features and seven pathomics features were included to construct the radiomics and pathomics signatures, respectively. The radiomics signature achieved C-indexes of 0.811 (95% CI, 0.746–0.877) in the training cohort and 0.752 (95% CI, 0.615–0.881) in the validation cohort for predicting DMFS. The pathomics signature achieved C-indexes of 0.713 (95% CI, 0.625–0.800) in the training cohort and 0.702 (95% CI, 0.569–0.835) in the validation cohort. After multivariable Cox analysis, clinical N stage, IS_B, radiomics signature, and pathomics signature were identified as independent factors for DMFS prediction. The nomogram achieved C-indexes of 0.902 (95% CI, 0.870–0.933) in the training cohort and 0.848 (95% CI, 0.743–0.951) in the validation cohort, with corresponding AUCs of 0.950 (95% CI, 0.916–0.985) and 0.872 (95% CI, 0.769–0.976) for 5-year DMFS. Kaplan-Meier analysis demonstrated that a cutoff value of 0.188 effectively stratified patients into high- and low-risk DM groups in both the training and validation cohorts (both, p < 0.001).
Discussion and Conclusions: This study investigated a novel approach that integrates radiomics, pathomics, and IS_B for predicting DM before treatment in LARC patients. The nomogram serves as a valuable tool for clinicians to identify high-risk DM patients and develop personalized neoadjuvant treatment strategies.

Acknowledgements

Funding: This research is supported by the National Natural Science Foundation of China [grant number 81971589].