CardioMetabolic Logit Index (CMLI): a composite clinical–laboratory model associated with coronary artery disease in a middle-aged population

Abstract Background Coronary artery disease (CAD) is closely associated with metabolic dysfunction and atherogenic lipoprotein remodeling. Conventional lipid parameters may not fully reflect the complex cardiometabolic alterations associated with atherosclerosis. This study aimed to develop the CardioMetabolic Logit Index (CMLI), an exploratory clinical–laboratory model integrating clinical factors, fasting glucose, and small dense LDL cholesterol (sdLDL-C), in relation to angiographically confirmed CAD. Methods This single-center case–control derivation study included 150 patients with angiographically confirmed CAD and 70 apparently healthy controls frequency matched to the CAD group for age and BMI. Serum lipid parameters, the triglyceride-glucose (TyG) index, the triglyceride/high-density lipoprotein cholesterol (TG/HDL-C) ratio, and LDL subfractions were assessed. Small dense LDL cholesterol (sdLDL-C) and LDL particle size were measured using the Lipoprint LDL System. The final CMLI model was derived in 218 participants (148 patients with CAD and 70 controls) using multivariable logistic regression with sex, hypertension, lnGlucose, and lnsdLDL, and was internally validated using bootstrap resampling. The CMLI equation was: $$\begin{aligned}\text{CMLI}\;\text{logit}=&-13.798+1.957\times\text{Hypertension}\\&+0.372\times\text{Sex}+2.630\times\text{InGlocuse}\\&+0.666\times\text{InsdLDL}.\end{aligned}$$ Results Patients with CAD had higher fasting glucose, triglyceride, TyG, and sdLDL-C levels and lower lbLDL and LDL particle size than controls (all p < 0.05). In the final multivariable model, hypertension, lnGlucose, and lnsdLDL were significantly associated with CAD, whereas sex was retained as a prespecified covariate but was not statistically significant. The basic clinical model comprising hypertension and sex had an AUC of 0.726. Adding lnGlucose increased the AUC to 0.792 (ΔAUC = 0.066, p = 0.004), and subsequent addition of lnsdLDL to form CMLI further increased the AUC to 0.843 (95% CI, 0.788–0.898; ΔAUC = 0.051, p = 0.013). Comparator models containing TyG or ln(TG/HDL-C) did not show significantly higher discrimination than the model containing lnGlucose ( p = 0.921 and p = 0.154, respectively). Bootstrap internal validation yielded an optimism-corrected AUC of 0.831. Conclusions CMLI, which combines sex, hypertension, fasting glucose, and sdLDL-C, was developed as an exploratory clinical–laboratory model associated with angiographically confirmed CAD. Within the derivation sample, adding fasting glucose to the basic clinical model improved discrimination, with a further improvement after the addition of sdLDL-C. CMLI should not be regarded as a validated diagnostic or clinical decision tool. External validation and appropriate recalibration in independent, clinically representative populations are required before clinical application can be considered.