Development of a preoperative predictive classifier and tailored staging system for obstructive colorectal cancer

Le résumé fourni par la source

OBJECTIVE: Obstructive colorectal cancer (OCC) often has a poor prognosis. The traditional TNM stage does not effectively predict disease prognosis, so there is an urgent need to establish a more suitable prognostic staging system for patients with OCC. METHODS: This study analyzed 164 patients treated for OCC at a single center between February 1, 2010, and January 31, 2020. The integrative two-step cluster (TSC) stage was used to self-classify OCCs into two groups on the basis of postoperative clinicopathological features. The Kaplan‒Meier method was subsequently used to compare the 3-year overall survival (OS) and disease-free survival (DFS) of patients grouped by TNM stage and integrative TSC stage. Machine learning classifiers were then applied to predict the integrative TSC stage using preoperative clinical features and radiomics. To evaluate the performance of the classifiers, we used the receiver operating characteristic (ROC) curve and the area under the receiver operating characteristic curve (AUC). RESULTS: After the integrative TSC stage was constructed, 164 patients with OCC were divided into an integrative superior group (iSG, n = 99) and an integrative inferior group (iIG, n = 65). In the integrative TSC staging system, lymphovascular invasion (1.00) was identified as the most valuable risk factor. The iSG outperformed the iIG in both OS and DFS (OS-months: 31.42 (27.96-34.88) vs. 19.30 (15.43-23.17), p < 0.001; DFS-months: 31.15 (27.49-34.82) vs. 18.12 (14.32-21.92), p < 0.001). Compared with the TNM stage, the integrative TSC stage was more effective in discriminating OS (integrative TSC stage, p < 0.001; TNM stage, p = 0.036) and DFS (integrative TSC stage, p < 0.001; TNM stage, p = 0.011). To predict the integrative TSC stage, the linear discriminant analysis (LDA) classifier based on carcinoembryonic antigen (CEA) and 15 radiomics features performed best (testing dataset: AUC 0.85, 95% CI: 0.74-0.96). In the stratified analysis, patients who underwent BTS were reclassified into the iSG group. These patients had lower CEA levels [3.22 (interquartile range[IQR], 1.60-5.50) ng/ml, p = 0.036] and higher ct_wavelet-LHH_gldm_SmallDependenceHighGray-LevelEmphasis (HGLE) levels [0.26 (IQR, 0.09-0.41), p = 0.044]. CONCLUSION: The integrative TSC stage surpasses the TNM staging system in distinguishing OCC prognosis, with LVI serving as a vital component. The LDA classifier using 16-dimensional features can effectively predict the integrative TSC stage.

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