Background/Aims: Hypocholesterolemia is a common finding in hospitalized elderly people, critically ill surgical patients, septic patients and end-stage renal disease patients. The different effect of lipid subfractions on patients with end-stage renal disease has never been demonstrated. We aim to study the effect of lipid subfractions on hospitalization and mortality in maintenance hemodialysis (MHD) patients. Methods: Lipid subfractions, including total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C) were measured in 210 patients with MHD in a single dialysis center. Patients were stratified into three groups based on the tertiles of lipid levels, and differences in patient characteristics and survival were evaluated. Results: Of a total of 22 deceased patients in our MHD cohort, infection-related mortality (50%) was higher than cardiovascular-related mortality (18.2%). Significant differences (p < 0.05) in the duration and frequency of hospitalization and in mortality events were observed when patients were divided into different subgroups according to the tertiles of baseline TC and LDL-C levels. Patients with lower LDL had significantly lower levels of albumin, TC and TG. The LDL-C tertiles were similar in terms of age, hypertension, diabetes, biochemical results, hematocrit, adequacy of hemodialysis and normalized protein catabolism rate. Both TC and LDL-C predicted survival (p < 0.001), but not TG and HDL-C in the Kaplan-Meier model. The Cox proportional hazard model demonstrated that baseline serum LDL-C was the best lipid subfraction in predicting all-cause death with an adjusted hazard ratio (95% confidence interval) for each 10 mg/dl of 0.752 (0.631–0.898; p = 0.002). Conclusions: We firstly demonstrated that lipid subfractions, including TC and LDL-C, predict poor outcomes in a MHD cohort with high infection-related mortality.