To investigate the association between mean fasting glucose (FG) and variability in visit-to-visit FG and risk of mortality in people with type 2 diabetes (T2D).

This retrospective cohort study included 48,077 Chinese men and women with T2D. The yearly mean and coefficient of variation for fasting glucose (mean-FG and FG-CV, respectively) were based on at least two measurements taken each year over a mean duration of 4.5 years. Deaths and their causes were identified using the Shanghai Vital Statistics registry. Mean-FG and FG-CV served as time-dependent variables in Cox models to estimate their associations with risk of mortality; hazard ratios (HRs) were adjusted for baseline risk factors. Potential non-linear associations were examined by restricted cubic splines.

During an average 4.5 years of follow-up, 2095 men and 1923 women died. Men with low mean-FG and women with low or high FG were at greater risk of death. Mean-FG was not associated with either cardiovascular disease (CVD) or cancer-specific mortality, whereas higher yearly FG-CV was associated with all-cause and CVD-/cancer-specific mortality in both genders. Compared with a yearly FG-CV of 1.76 (5th percentile), men and women with an FG-CV of 14.14 (75th percentile) had HRs (95% CI) of 1.41 (1.24–1.61) and 1.44 (1.26–1.65), respectively, for all-cause mortality.

Variability of visit-to-visit FG may be a more sensitive predictor of risk of death than mean-FG in people with T2D.