Accuracy, Variability, and Bias of Glycemic Biomarkers in Patients Treated With Maintenance Dialysis.

Abstract

Accurate assessment of glycemia in patients treated with maintenance dialysis is imperative and hampered by known biases of glycated hemoglobin (HbA1c) in kidney failure (KF). This study evaluated the accuracy, variability, and covariate bias of three glycemic biomarkers compared with glycemia measured by continuous glucose monitor (CGM) among people with and without diabetes treated with maintenance dialysis. In a prospective community-based cohort study, 251 participants treated with maintenance dialysis wore a Dexcom G6 Pro CGM for 10 days. We compared correlations of HbA1c, glycated albumin (GA), and fructosamine with CGM-derived mean glycemia and examined sources of bias. Participants (43% women; 63% with diabetes) had a median of 9.3 (interquartile range 8.5-9.4) valid days of CGM data. Mean (SD) HbA1c, GA, fructosamine, and mean CGM glucose were 6.2% (1.4%), 19.6% (6.3%), 351 (99) µmol/L, and 170 (63) mg/dL, respectively. HbA1c, GA, and fructosamine all strongly correlated with mean CGM blood glucose, with HbA1c and GA more correlated than fructosamine (overall, r = 0.85, r = 0.87, and r = 0.70, respectively; in diabetes, r = 0.84, r = 0.84, and r = 0.64, respectively). Compared with mean CGM glucose, HbA1c was significantly biased by erythropoiesis-stimulating agent dose, BMI, hemoglobin, and serum albumin; GA and fructosamine were biased by dialysis modality and vintage, residual kidney function, and BMI. HbA1c and GA were strongly correlated with mean CGM blood glucose, but all biomarkers had substantial bias by relevant clinical characteristics. HbA1c and GA may be useful assessments of average glycemia in patients treated with maintenance dialysis, if bias can be adequately addressed.