Abstract
<title>Abstract</title> <p> <bold>Background</bold> Routinely collected perioperative immune, inflammatory, and nutritional data may capture host-response information that is not fully represented by tumour-burden variables alone. This study aimed to develop and internally validate an interpretable host-response score for postoperative disease-free survival (DFS) risk stratification after curative-intent surgery for colorectal adenocarcinoma, with potential relevance to medical-informatics-enabled postoperative decision making. <bold>Methods</bold> We conducted a retrospective prognostic modelling study of 2,806 patients with colorectal adenocarcinoma treated at Yunnan Cancer Hospital. Patients were randomly allocated to training and internal validation cohorts in a 7:3 ratio. A lymphocyte-based immune-inflammatory/nutritional score (LINS) was derived in the training cohort using an XGBoost accelerated failure-time survival model based on 10 routinely available preoperative host-response indicators. Performance was assessed using Harrell's C-index, time-dependent area under the receiver operating characteristic curve (AUC), calibration, decision-curve analysis, Kaplan-Meier risk stratification, and SHAP-based interpretability. A public colorectal cancer transcriptomic meta-cohort was used for exploratory immune-risk proxy analysis and was not treated as direct external validation of the blood-based model. <bold>Results</bold> During a median follow-up of approximately 73 months, 868 DFS events occurred. LINS separated patients into low-, intermediate-, and high-risk groups with graded DFS differences in both cohorts. In the internal validation cohort, LINS alone showed modest discrimination, whereas the integrated LINS plus clinical model achieved the best performance, with a C-index of 0.706 and 1-, 3-, and 5-year AUCs of 0.735, 0.763, and 0.742, respectively. Calibration was acceptable at 1, 3, and 5 years, and decision-curve analysis suggested potential net benefit across clinically relevant threshold ranges. <bold>Conclusions</bold> LINS is a routinely measurable host-response signature that adds modest but clinically interpretable prognostic information to conventional clinicopathological predictors for postoperative DFS risk stratification in colorectal adenocarcinoma. Prospective multicentre validation and workflow-based clinical utility assessment are required before clinical deployment. <bold>Trial registration</bold> Not applicable. </p>