Antidepressant medication use and trajectories of fasting plasma glucose, glycated haemoglobin, β-cell function and insulin sensitivity: A 9-year longitudinal study of the D.E.S.I.R. cohort

Abstract

Background: Use of antidepressants is seen to be a risk factor for type 2 diabetes, even though the underlying mechanisms remain unclear. We examined whether antidepressant use was associated with change in fasting plasma glucose, glycated haemoglobin (HbA1c), β-cell function (HOMA2-%B) and insulin sensitivity (HOMA2-%S) over time. Methods: Participants in the French D.E.S.I.R. cohort study included over 4700 men (48.1%) and women, free of diabetes, aged 30-65 years at baseline in 1994-96 (D.E.S.I.R. 0), who were followed for 9 years at 3-yearly intervals (D.E.S.I.R. 3, 1997-99; 6, 2000-02; 9, 2003-05). Antidepressant use, fasting plasma glucose, HbA1c, HOMA2-%B and HOMA2-%S were assessed concurrently at four medical examinations. Linear mixed models were used to examine the cross-sectional and longitudinal associations of timedependent antidepressant use with changes in these four biological parameters. Results: Mean fasting plasma glucose and HbA1c increased whereas HOMA2-%B and HOMA2-%S decreased over the follow-up. In a fully adjusted model, there were no differences in: Mean fasting plasma glucose (β=0.01 mmol/l, P=0.702); HbA1c (β=0.01 %, P=0.738); HOMA2-%B (β=0.00, P=0.812); or HOMA2-%S (β=-0.01, P=0.791) at baseline (1994-96) between antidepressant users and non-users. The interaction term with time also suggested no differences in the annual change in: Fasting plasma glucose (β=0.00 mmol/l, P=0.322); HbA1c (β=0.00 %, P=0.496); HOMA2-%B (β=0.00, P=0.609); or HOMA2-%S (β=0.00, P=0.332) between antidepressant users and nonusers. Similar associations were observed in analyses of type and cumulative use of antidepressants over follow-up. Conclusion: Our longitudinal data show that use of antidepressants is not associated with altered glucose metabolism, suggesting that the association between antidepressant use and diabetes reported by previous studies may not be causal. Detection bias or clinical ascertainment bias may account for much of this apparent association.