Background Long-term HbA1c (glycated haemoglobin) variability is associated with micro- and macro-vascular complications in Type 2 diabetes, but it is not well established if it also increases the risk of infections. We explored prospective associations between HbA1c variability and serious infections; how these relate to average HbA1c level, and whether these are modified by age, sex and ethnicity.

Methods A retrospective cohort study of 411,963 Type 2 diabetes patients in England, aged 18-90, alive on 01/01/2015 in the Clinical Practice Research Datalink with ≥4 HbA1c measurements during 2011-14. An HbA1c Variability Score (HVS) was estimated (% of successive measurements whose difference exceeds a pre-defined threshold), and patients were categorised (0-<20, 20-<50, 50-<80, 80-100). Poisson regression estimated incidence rate ratios (IRRs) for infections requiring hospitalisation during 2015-19 by HVS categories, adjusting for confounders, stratified by average HbA1c level and age, sex, ethnicity. To assess the potential overall impact of variability versus average level, attributable risk fractions (AFs) were calculated, using a hypothetical reference category for both variability (HVS 0-<20) and average level (42-<48mmol/mol).

Results While increasing HbA1c level and variability were both independently associated with infections, a greater infection risk (IRRs>1.2) was seen with any variability category (HVS≥20, 73% of all patients), whereas for average level it was observed only at higher values (≥64mmol/mol, 27% patients). Adjusting variability for average level, and vice versa, attenuated associations, but did not alter this pattern. Estimated AFs were markedly greater for variability than average level (17.1% vs. 4.1%). The positive association between variability and infection was more apparent among patients with the lowest average HbA1c levels (<48 mmol/mol) and observed at all average levels except the very highest (≥86 mmol/mol). Associations with variability were greater among older patients (≥60 years), similar by sex, but not observed among people from Black ethnicities. Associations were similar irrespective of the direction of the last recorded HbA1c change.

Conclusion While the greatest individual risk for serious infection was observed in patients with Type 2 diabetes at very high HbA1c levels, among all patients HbA1c variability between primary care visits appeared to account for more serious infections overall. Since increased risks associated with variability were clearly observed at lower average levels, greater clinician and patient awareness of HbA1c variability may be beneficial. However, only well-designed trials can establish whether there are long-term benefits of directly treating variability in HbA1c on infections, quality of life and other diabetes complications.