Personalized Hemoglobin A1c Shows Better Correlation with Mean Glucose than Laboratory Hemoglobin A1c in Ugandan Youth with Type 1 Diabetes, but Mean Glucose Is Not Clinically Useful in This Population Due to Extreme Glucose Variability

Introduction: Continuous glucose monitoring (CGM) is unaffordable in sub-Saharan Africa, and providers rely heavily on hemoglobin A1c (A1c) to guide insulin adjustment. The relationship between A1c and mean glucose (MG) varies between individuals and populations. We assessed this relationship in Ugandan youth of age 4-26 years with type 1 diabetes, and evaluated whether calculation of the personalized A1c (pA1c), which only requires a brief initial sensor wear, is clinically useful. Materials and Methods: CGM data were averaged across three blinded sensor wears (31-41 days). We calculated individual apparent glycation ratios using A1c after the second sensor, and applied these to A1cs collected after the third sensor to determine pA1c. Participants were evaluated for clinical factors that influence red blood cell (RBC) lifespan (malaria, G6PD deficiency, sickle-cell trait, hemolysis, iron deficiency). Results: Patients across the A1c spectrum experienced substantial time in both hyper- and hypoglycemia; average coefficient of variation was 44%. MG was >250 mg/dL (13.9 mmol/L) in 50% of participants, and 55% of participants spent ≥4% time with glucose <70 mg/dL (3.9 mmol/L). There was considerable variability in the A1c-MG relationship. The pA1c more accurately represented MG by significantly reducing variation in this relationship (R² = 0.84 vs. 0.40; r = 0.92 vs. 0.63), but MG is not useful in individuals with the wide glucose fluctuations seen in this population. Clinical factors did not impact the A1c-MG relationship. Conclusions: Neither the measured A1c nor the calculated pA1c provided reliable guidance for insulin adjustment in this population. No matter how accurately MG is measured or estimated, it is just an average, with limited clinical application in individuals with wide glycemic variation. These measures cannot replace the information available from CGM about glycemic excursion, daily glucose patterns, or percent time in various glucose ranges. Our data suggest that it is essential to find a way to make CGM at least periodically affordable in low-resource settings.