HDL-C presents a suggestive negative moderating trend for the association between total cholesterol and colorectal polyps: model construction and bootstrap internal assessment of a predictive nomogram.
Colorectal polyps are classic precancerous lesions of colorectal cancer, closely linked to lipid metabolic disorders. Elevated total cholesterol does not consistently raise polyp risk, implying interactive effects among lipoprotein fractions. This study aimed to explore whether HDL-C exhibits a moderating trend in the TC-polyp association and construct a predictive nomogram with bootstrap internal performance evaluation.
A retrospective single-center case-control study enrolling 411 hospitalized colonoscopy patients. Our pre-defined primary multivariate logistic regression was adjusted for age and standardized objective baseline indicators (gender, BMI category, Helicobacter pylori, GSP, CST4). Normal weight was set as BMI reference; restricted cubic splines were applied to test BMI nonlinearity. Data on smoking, alcohol intake, diet, physical activity and family history of colorectal polyps were incompletely recorded in inpatient medical records and thus excluded from regression adjustment. Sensitivity analyses substituting non-HDL-C for TC were performed to eliminate compositional coupling bias between TC and HDL-C. Subgroup analyses stratified by polyp pathological subtypes were conducted. Nomogram discrimination and calibration were evaluated via ROC curves and 500 bootstrap resampling on the full cohort without train-test splitting. Random forest was used to rank variable importance; such rankings cannot independently corroborate the TC × HDL-C interaction term.
After adjustment for age and baseline clinical covariates, we detected a suggestive inverse TC × HDL-C interaction (OR = 0.787, 95% CI: 0.626-0.994, P = 0.040). Sensitivity analysis with non-HDL-C reproduced this consistent inverse trend (P = 0.023). Restricted cubic spline modeling indicated no nonlinear correlation between continuous BMI and polyp risk (P for non-linearity=0.348). The internally assessed nomogram achieved an AUC of 0.834 (95% CI: 0.789-0.879) with satisfactory bootstrap calibration. Consistent inverse moderating trends were observed across adenomatous and hyperplastic polyp subgroups at the descriptive level.
We identified a suggestive inverse TC-HDL-C moderating trend after covariate adjustment, which remained robust in non-HDL-C sensitivity analysis. HDL-C exhibited no independent main effect on colorectal polyps, yet it exerted a suggestive negative moderating trend that attenuated the positive association between total cholesterol and colorectal polyps. The internally assessed nomogram provides preliminary reference for individualized colorectal polyp risk stratification. Observations are constrained by inpatient-only recruitment and incomplete lifestyle covariate adjustment; further large community-based cohorts are needed to replicate these findings.
A retrospective single-center case-control study enrolling 411 hospitalized colonoscopy patients. Our pre-defined primary multivariate logistic regression was adjusted for age and standardized objective baseline indicators (gender, BMI category, Helicobacter pylori, GSP, CST4). Normal weight was set as BMI reference; restricted cubic splines were applied to test BMI nonlinearity. Data on smoking, alcohol intake, diet, physical activity and family history of colorectal polyps were incompletely recorded in inpatient medical records and thus excluded from regression adjustment. Sensitivity analyses substituting non-HDL-C for TC were performed to eliminate compositional coupling bias between TC and HDL-C. Subgroup analyses stratified by polyp pathological subtypes were conducted. Nomogram discrimination and calibration were evaluated via ROC curves and 500 bootstrap resampling on the full cohort without train-test splitting. Random forest was used to rank variable importance; such rankings cannot independently corroborate the TC × HDL-C interaction term.
After adjustment for age and baseline clinical covariates, we detected a suggestive inverse TC × HDL-C interaction (OR = 0.787, 95% CI: 0.626-0.994, P = 0.040). Sensitivity analysis with non-HDL-C reproduced this consistent inverse trend (P = 0.023). Restricted cubic spline modeling indicated no nonlinear correlation between continuous BMI and polyp risk (P for non-linearity=0.348). The internally assessed nomogram achieved an AUC of 0.834 (95% CI: 0.789-0.879) with satisfactory bootstrap calibration. Consistent inverse moderating trends were observed across adenomatous and hyperplastic polyp subgroups at the descriptive level.
We identified a suggestive inverse TC-HDL-C moderating trend after covariate adjustment, which remained robust in non-HDL-C sensitivity analysis. HDL-C exhibited no independent main effect on colorectal polyps, yet it exerted a suggestive negative moderating trend that attenuated the positive association between total cholesterol and colorectal polyps. The internally assessed nomogram provides preliminary reference for individualized colorectal polyp risk stratification. Observations are constrained by inpatient-only recruitment and incomplete lifestyle covariate adjustment; further large community-based cohorts are needed to replicate these findings.