LAUSR

Background For glycemic control in diabetic patients on dialysis it was unclear what level of glycated albumin (GA) was associated with the lowest mortality and GA's utility. Accordingly, we examined the difference in association between GA and hemoglobin A1c (HbA1c) with 1-year mortality in a cohort of the Japanese Society for Dialysis Therapy. Methods We examined 84 282 patients with prevalent diabetes who were on maintenance hemodialysis (HD) (female 30.3%; mean age 67.3 ± 11.2 years; mean dialysis vintage 6.4 ± 4.5 years). Of them, 22 441 had both GA and HbA1c. We followed these for a year, 2013-14, using Cox regression to calculate adjusted hazard ratios (HRs) and 95% confidence limits for 1-year mortality after adjusting for potential confounders such as baseline age, sex, smoking and diabetes type. Results One-year mortality was lowest in diabetic HD patients who had GA levels of 15.6-18.2% and HbA1c levels of 5.8-6.3%. The associations were linear or J-shaped for GA and U-shaped for HbA1c. Adjusted HRs were significantly higher in patients with GA <12.5% and GA ≥22.9%. This trend flattened in elderly patients, those with higher hemoglobin or those with prior cardiovascular disease. In addition, the C-statistics, Harrell's C and category-free net reclassification improvement to predict 1-year mortality were better when GA was added to the model than when HbA1c was added. Conclusions There was a linear or J-shaped association between GA and 1-year mortality, with the lowest mortality at GA 15.6-18.2%. Furthermore, our analyses suggest the potential superiority of GA over HbA1c in predicting mortality.