The aim of this study was to investigate the association between visit-to-visit variability in HbA and cognitive function decline in the elderly population.We performed a pooled analysis of two prospective population-based cohorts (the Health Retirement Study [HRS] and the English Longitudinal Study of Ageing [ELSA]). Cognitive function, including memory and executive function, were assessed at baseline and every 2 years, while HbA levels were assessed at baseline and every 4 years. Visit-to-visit variability (VVV) in HbA was calculated using the CV, SD and variation independent of the mean (VIM) during the follow-up period. Linear mixed models were used to evaluate the association between HbA variability and cognitive function decline with adjustment for demographics, mean HbA, education, smoking, alcohol consumption, BMI, baseline hypertension, baseline diabetes status and HDL-cholesterol.The study enrolled 6237 participants (58.23% women, mean age 63.38 ± 8.62 years) with at least three measurements of HbA. The median follow-up duration was 10.56 ± 1.86 years. In the overall sample, compared with the lowest quartile of HbA variability, participants in the highest quartile of HbA variability had a significantly worse memory decline rate (-0.094 SD/year, 95% CI -0.185, -0.003) and executive function decline rate (-0.083 SD/year, 95% CI -0.125, -0.041), irrespective of mean HbA values over time. Among individuals without diabetes, each 1-SD increment in HbA CV was associated with a significantly higher rate of memory z score decline (-0.029, 95% CI -0.052, -0.005) and executive function z score decline (-0.049, 95% CI -0.079, -0.018) in the fully adjusted model.We observed a significant association between long-term HbA variability and cognitive decline among the non-diabetic population in this study. The effect of maintaining steady glucose control on the rate of cognitive decline merits further investigation.