Background: Chronic kidney disease (CKD) is an independent risk factor for atherosclerotic disease. We hypothesized that CKD promotes a proatherogenic lipid profile modifying lipoprotein composition and particle number. Methods: Cross-sectional study in 395 non-diabetic individuals (209 CKD patients and 186 controls) without statin therapy. Conventional lipid determinations were combined with advanced lipoprotein profiling by nuclear magnetic resonance, and their discrimination ability was assessed by machine learning. Results: CKD patients showed an increase of very-low-density (VLDL) particles and a reduction of LDL particle size. Cholesterol and triglyceride content of VLDLs and intermediate-density (IDL) particles increased. However, low-density (LDL) and high-density (HDL) lipoproteins gained triglycerides and lost cholesterol. Total-Cholesterol, HDL-Cholesterol, LDL-Cholesterol, non-HDL-Cholesterol and Proprotein convertase subtilisin-kexin type (PCSK9) were negatively associated with CKD stages, whereas triglycerides, lipoprotein(a), remnant cholesterol, and the PCSK9/LDL-Cholesterol ratio were positively associated. PCSK9 was positively associated with total-Cholesterol, LDL-Cholesterol, LDL-triglycerides, LDL particle number, IDL-Cholesterol, and remnant cholesterol. Machine learning analysis by random forest revealed that new parameters have a higher discrimination ability to classify patients into the CKD group, compared to traditional parameters alone: area under the ROC curve (95% CI), .789 (.711, .853) vs .687 (.611, .755). Conclusions: non-diabetic CKD patients have a hidden proatherogenic lipoprotein profile.
Keywords: Atherosclerosis; Lp(a); PCSK9; chronic kidney disease; dyslipidemia; lipoprotein subfractions.