BACKGROUND: Concerns have arisen that treatment with specific antidepressants, including venlafaxine (VEN), can increase the risk of suicide. However, there is evidence that VEN prescribing is channeled towards patients at higher underlying suicide risk.
OBJECTIVES: To compare the risk of completed suicide in adults during treatment with VEN relative to citalopram (CIT), fluoxetine (FLU), and dothiepin (DOT), controlling for measurable confounders.
METHODS: We conducted a retrospective cohort study in the UK using 1995–2005 GPRD data. Patients were 18 years, treated with VEN, CIT, FLU, or DOT, had a diagnosis of depression or anxiety, and had more than one year of data prior to study. Follow-up started at incident study drug prescription, and patients could contribute to more than one exposure group. A time-dependent Cox model estimated the Hazard Ratio (HR) during VEN versus CIT, FLU, or DOT exposure, prior and after adjustment for measured confounders.
RESULTS: Of 219 088 study subjects, approximately 1/3 were male and 2/3 were 30–59 years old. VEN users had a higher burden of suicide risk factors, e.g., prior suicide attempts, family history of psychiatric morbidity, and proxies for severe and/or difficult to treat depression. There were 54 suicides in 173 452 person-years (PYs) at risk, 18 occurring during VEN treatment, 12 during CIT, and 15 and 9 during FLU and DOT treatment, respectively. Unadjusted HR (95% CI) for VEN versus CIT was 2.44 (1.12– 5.31); After adjustment for more than 20 covariates, HR (95% CI) was 1.70 (0.76–3.80); Unadjusted and adjusted HRs (95% CI) were 2.85 (1.37–5.94) and 1.63 (0.74–3.59) for VEN vs FLU and 2.54 (1.07–6.02) and 1.31 (0.53–3.25) for VEN vs DOT.
CONCLUSIONS: Higher risk of suicide was associated with VEN, compared to CIT, FLU, and DOT treatment. VEN users had a higher burden of suicide risk factors, and adjustment for measured confounders substantially reduced the excess risk. Because of potential misclassification of confounders resulting from the nature of the data, e.g. proxies of disease severity, one cannot rule out that residual confounding explains much or all of the remaining excess risk.
