Mean HbA1c and mortality in diabetic individuals with heart failure: a population cohort study

Aims Controversy exists regarding the importance of glycaemic control in patients with type 2 diabetes mellitus (T2DM) and chronic heart failure (CHF) based on conflicting reports using single baseline glycosyated haemoglobin (HbA(1c)). Using the time-weighted mean of serial HbA(1c) measurements has been found to be a better predictor of diabetic complications as it reflects the glycaemic burden for that individual over time. We therefore sought to confirm this in a large cohort of patients with T2DM and incident CHF. Methods and results A time-weighted mean HbA(1c) was calculated using all HbA(1c) measurements following CHF diagnosis. Patients were grouped into five categories of HbA(1c) (<= 6.0%, 6.1-7.0%, 7.1-8.0%, 8.1-9.0%, and >9.0%). The relationship between time-weighted mean HbA(1c) and all-cause death after CHF diagnosis was assessed. A total of 1447 patients with T2DM met the study criteria. During a median follow-up of 2.8 years, there were 826 (57.1%) deaths, with a crude death rate of 155 deaths per 1000 person-years [95% confidence interval (CI) 144-166]. A Cox regression model, adjusted for all significant predictors, with the middle HbA(1c) category (7.1-8.0%) as the reference, showed a U-shaped relationship between HbA(1c) and outcome [HbA(1c) <6.0%, hazard ratio (HR) 2.5, 95% CI 1.8-3.4; HbA(1c) 6.1-7.0%, HR 1.4, 95% 1.1-1.7; HbA(1c) 8.1-9.0%, HR 1.3, 95% CI 1.0-1.6; and HbA(1c) >9.0%, HR 1.8, 95% CI 1.4-2.3]. Further analysis revealed a protective effect of insulin sensitizers (i.e. metformin) (HR 0.7, 95% CI 0.61-0.93) but not other drug classes. Conclusions In patients with T2DM and CHF, our study shows a U-shaped relationship between HbA(1c) and mortality, with the lowest risk in patients with modest glycaemic control (HbA(1c) 7.1-8.0%) and those treated with insulin sensitizers.