Patients on maintenance hemodialysis exhibit a notably higher risk of cardio-cerebrovascular complications that constitute the major cause of death. Preceding studies have reported conflicting associations between traditional lipid measures and clinical outcome in dialysis patients. In this prospective longitudinal study, we utilized quantitative lipidomics to elucidate, at molecular resolution, changes in lipidome profiles of erythrocyte and plasma samples collected from maintenance hemodialysis patients followed up for 86 months (≈ 7 years). Primary outcome was defined as cardiovascular-related deaths or new-onset cardio-cerebrovascular events. Cox regression model uncovered plasma/erythrocyte lipids associated with incident cardio-cerebrovascular events in the erythrocyte cohort (n=117 patients, 37 events) and plasma cohort (n= 45 patients, 11 events), respectively. Both the erythrocyte lipid panel [PA 40:5, PI 34:2, PC 42:6, AUC= 0.83] and plasma lipid panel [PC O-34:1, GM3 18:1;O2/25:0, TG 44:1(16:1_28:0), AUC = 0.94] significantly improved the prediction of cardio-cerebrovascular-related outcome compared to the base model comprising age, sex and dialysis vintage alone. Our findings underscore the pathophysiological significance of anionic phospholipid accretion in erythrocytes in the development of cardio-cerebrovascular complications in dialysis patients. In particular, distorted membrane lipid asymmetry leads to compromised membrane deformability, aberrant cell-cell interactions and altered glutathione metabolism in the erythrocytes of high-risk individuals even at relatively early stage of hemodialysis. Our findings thus underscore the importance of maintaining the RBC pool to lower the risk of cardio-cerebrovascular complications in dialysis patients.