Development and external validation of a preoperative CT body composition-based model for predicting postoperative metastasis in colorectal cancer: a multicenter retrospective cohort study.

OBJECTIVES

Most postoperative metastases in nonmetastatic colorectal cancer (CRC) occur within 2 years after curative-intent surgery, but tools for quantifying early metastasis risk remain limited. We aimed to develop and externally validate a nomogram integrating preoperative CT body composition with clinicopathologic variables.

METHODS

This multicenter retrospective study included 451 patients with nonmetastatic CRC who underwent curative-intent surgery from January 2019 to December 2022. The training cohort comprised 250 patients from the primary center and the validation cohort 201 patients from 2 external centers. L3-level CT was used to quantify visceral adipose tissue index (VATI). Metastasis-free survival was the primary endpoint, and early metastasis was defined as distant metastasis within 24 months. Independent predictors were identified with Cox regression and used to build a nomogram.

RESULTS

Postoperative metastasis occurred in 61 of 250 patients (24.4%) in the training cohort and 50 of 201 patients (24.9%) in the validation cohort. Elevated CA19-9, positive N stage, lymphovascular invasion, and higher VATI were independent predictors. The nomogram yielded a C-index of 0.793 in the training cohort and 0.754 in the validation cohort; the 24-month area under the curve was 0.818 and 0.804, respectively.

CONCLUSIONS

A nomogram combining CT-derived VATI with N stage, CA19-9, and lymphovascular invasion showed good discrimination and calibration for predicting postoperative metastasis and 24-month early metastasis in CRC.