Baseline and early changes in CT body composition can predict recurrence-free survival after radical gastrectomy: A sex-specific study

Abstract

Objective: This study aimed to explore the predictive value of baseline CT body composition and its early changes on recurrence-free survival (RFS) following radical gastrectomy, while also assessing potential sex-related differences.

Methods: We conducted a retrospective analysis of gastric cancer (GC) patients with confirmed pathology from October 2019 to May 2023. All patients underwent preoperative and postoperative CT scans to assess visceral fat area (VFA), subcutaneous fat area (SFA), skeletal muscle area (SMA), and skeletal muscle density (SMD), along with calculating their respective rates of change. Multivariate Cox regression analyses were used to identify independent predictors of RFS in male and female patients separately, and nomogram models were subsequently developed. The models' predictive performance was assessed using calibration curves, receiver operating characteristic (ROC) curves, and decision curve analysis (DCA).

Results: The study included 287 patients, consisting of 185 males and 102 females. At baseline, males had a lower subcutaneous adipose tissue index (SATI) but higher skeletal muscle index (SMI) and SMD compared to females (p＜0.001). Postoperatively, both SATI and visceral adipose tissue index (VATI) were significantly lower in both males and females than their corresponding preoperative values (p＜0.005). In males, SMI (HR = 0.442, p = 0.002), VATI (HR = 1.843, p = 0.018), lymphocyte (LYM) (HR = 0.486, p = 0.040), pathological T stage (HR = 3.004, p = 0.003), and postoperative complication (POC) (HR = 1.893, p = 0.014) were found to be independent predictors of RFS. In females, independent predictors of RFS included SMI (HR = 0.361, p = 0.013), SATI change rate (δSATI) (HR = 0.428, p = 0.024), albumin (ALB) (HR = 0.242, p = 0.003), CEA (HR = 5.418, p < 0.001), and POC (HR = 3.425, p < 0.001). The male-specific nomogram model demonstrated predictive accuracy for recurrence-free survival (RFS), with areas under the ROC curve (AUC) of 0.621, 0.783, and 0.796 at 1, 2, and 3 years, respectively. Similarly, the female-specific nomogram model achieved AUCs of 0.796, 0.836, and 0.783 at the corresponding time points. Calibration curves indicated a strong concordance between predicted and observed outcomes, while DCA validated the clinical utility of both models. Additionally, the models effectively stratified patients into low-risk and high-risk groups.

Conclusion: Sex differences were observed in the predictive value of CT body composition for RFS after gastrectomy. By incorporating CT body composition parameters and clinical indicators, sex-specific nomogram models were developed, demonstrating effective prediction of RFS in gastric cancer patients post-surgery.

Keywords: Body composition; Computed tomography; Gastric cancer; Prediction; Sex.