“Paradoxical” prognostic role of the TyG index and a novel machine learning-derived nomogram for colorectal cancer liver metastases

This is the authors' abstract. We don't add key points for this paper.

Background: The triglyceride-glucose (TyG) index is a reliable surrogate marker of insulin resistance. However, its prognostic significance in patients with colorectal cancer liver metastases (CRLM) remains unclear. This study aimed to investigate its prognostic value and develop a machine learning (ML) - based predictive model in CRLM. Methods: A total of 644 patients with synchronous CRLM who underwent curative-intent treatment were enrolled in this single-center retrospective study. The association between the TyG index and overall survival (OS) was assessed using Kaplan-Meier analysis and Cox regression models. Key predictive variables were selected via LASSO regression and six ML algorithms, which were then utilized to construct a nomogram. The model's performance was evaluated by the C-index and the area under the receiver operating characteristic curve (AUC), and compared against classical clinical risk scores (CRS). Results: The research revealed an inverse association between TyG index and CRLM. Multiple analyses confirmed that a low TyG index was an independent risk factor for shortened OS. ML methods identified nine key variables, including the TyG index, for nomogram construction. The nomogram demonstrated improved predictive performance in individualized prognosis prediction compared to the CRS: C-index 0.697 (95% CI: 0.627-0.767) vs. 0.631 (95% CI: 0.559-0.703); 5-year AUC 0.829 (95% CI: 0.755-0.904) vs. 0.741 (95% CI: 0.646-0.836). Conclusion: Unlike what was observed in previous studies, a low TyG index was associated with poor prognosis in CRLM. The ML-derived predictive model showed potential for OS prediction and clinical application, but further external validation is required.