Association of triglyceride-glucose index and lipid parameters with colorectal cancer in non-diabetic and non-obese subjects: a case-control study

Background Insulin-mediated pathways are a plausible explanation for the pathogenesis of colorectal cancer (CRC). The triglyceride-glucose index (TyG) and blood lipid parameters both act as alternative indicators for evaluating insulin resistance (IR). To address this gap, this study aimed to analyze how the TyG index relates to CRC risk in a defined low-risk cohort—i.e., non-diabetic and non-obese participants. Methods We analyzed data from 180 participants at the Fourth Affiliated Hospital of Anhui Medical University, collected between September 2024 and May 2025. Via the sample package in R software, the total dataset was randomly divided into a training set (N=125) and a validation set (N=55) in a 7:3 ratio. Univariate Logistic regression was used to screen variables in the training set, with those having P<0.1 selected as candidates. LASSO regression was performed on the candidates to control overfitting, and the final variables were used to build a multivariate Logistic regression model. ROC curves, calibration curves, and decision curves for both sets were drawn to validate the model’s discriminative ability, calibration, and overall predictive performance. For the 180 patients with available TyG index data at the time of initial adenoma diagnosis, the AUC was used to evaluate the TyG index’s discriminative power for late recurrence. Results Retrospective findings on the TyG index, total cholesterol, and colorectal cancer (CRC) risk in non-diabetic, non-obese populations confirm that the TyG index is a risk factor and total cholesterol is a protective factor for colorectal cancer (CRC) incidence. Collected nonlinear indicators were subjected to univariate Logistic regression and LASSO regression, thus screening four variables: age (0.0332), cardiovascular disease (1.1450), total cholesterol (-0.4750), and TyG index (1.8776). These variables were included in multivariate Logistic regression, which revealed the TyG index and total cholesterol as independent factors. For the prediction model built with these two factors, training set ROC analysis showed that the combined model’s AUC was 0.806 (95% CI: 0.723–0.889), with a sensitivity of 0.882 and a specificity of 0.616. The validation set’s ROC analysis yielded an AUC of 0.851, which demonstrated the model’s good predictive performance. Conclusion A raised TyG index is associated with heightened colorectal cancer (CRC) risk, whereas total cholesterol exhibits a significant negative correlation with colorectal cancer risk. We developed and validated a colorectal cancer (CRC) risk assessment model based on the TyG index and total cholesterol. This model effectively facilitates accurate screening of high-risk patients, enabling clinicians to implement targeted risk management and individualized care plans for non-diabetic, non-obese patients.