Abstract
PURPOSE: Diabetic nephropathy (DN) is a common complication of type 2 diabetes mellitus (T2DM) that significantly impacts the quality of life for affected patients. Dyslipidemia is a known risk factor for developing cardiovascular complications in T2DM patients. However, the association between serum lipoprotein(a) (Lp(a)) and high-density lipoprotein cholesterol (HDL-C) with DN requires further investigation. PATIENTS AND METHODS: For this cross-sectional study, we randomly selected T2DM patients with nephropathy (DN, n = 211) and T2DM patients without nephropathy (T2DM, n = 217) from a cohort of 142,611 patients based on predefined inclusion and exclusion criteria. We collected clinical data from the patients to identify potential risk factors for DN using binary logistic regression and machine learning. After obtaining the feature importance score of clinical indicators by building a random forest classifier, we examined the correlations between Lp(a), HDL-C and the top 10 indicators. Finally, we trained decision tree models with top 10 features using training data and evaluated their performance with independent testing data. RESULTS: Compared to the T2DM group, the DN group had significantly higher serum levels of Lp(a) (p < 0.001) and lower levels of HDL-C (p = 0.028). Lp(a) was identified as a risk factor for DN, while HDL-C was found to be protective. We identified the top 10 indicators that were associated with Lp(a) and/or HDL-C, including urinary albumin (uALB), uALB to creatinine ratio (uACR), cystatin C, creatinine, urinary ɑ1-microglobulin, estimated glomerular filtration rate (eGFR), urinary β2-microglobulin, urea nitrogen, superoxide dismutase and fibrinogen. The decision tree models trained using the top 10 features and with uALB at a cut-off value of 31.1 mg/L showed an average area under the receiver operating characteristic curve (AUC) of 0.874, with an AUC range of 0.870 to 0.890. CONCLUSION: Our findings indicate that serum Lp(a) and HDL-C are associated with DN and we have provided a decision tree model with uALB as a predictor for DN.