Article Text
PDF
Statistics from Altmetric.com
Question
Question: What is the risk of suicide death for people with depression on antidepressants?
People: 502 179 adults in the Veterans Affairs (VA) Health System with depression (identified using implantable cardioverter defibrillator (ICD) coding), who started taking any of the most commonly used antidepressants between April 1999 and September 2004. A new antidepressant start (index date) was defined as filling of a prescription for any of the antidepressants after at least 6 months with no observed antidepressant prescription fills. Exclusion criteria: diagnosis of bipolar I, schizophrenia or schizoaffective disorder.
Setting: Department of VA Health System, USA; 1999–2004.
Risk factors: New start on one of the antidepressants (27.0% started sertraline, 26.1% citalopram, 14.0% fluoxetine, 13.3% paroxetine, 10.9% bupropion, 4.4% venlafaxine and 4.2% mirtazapine). Days of exposure to the antidepressant were defined as a start date of the first new fill continuing until completion of the last days’ fill supply. To examine the relationships between suicide and antidepressant use, conventional Cox regression models, Cox models with inverse probability of treatment weighting (IPTW), propensity-stratified Cox models (PSCM), marginal structural models (MSM analysis of 12 month follow-up data reported here) and instrumental variable analyses were used. Analyses were adjusted for clinical and demographic factors.
Outcomes: Suicide deaths, identified using ICD coding in the National Death Index.
Methods
Design: Retrospective cohort study.
Follow-up period: Five years.
Main results
There were 247 suicides among the antidepressant users the study period. Crude suicide rates ranged from 88 to 247 per 100 000 person-years across the antidepressants. The highest rate per 100 000 person-years was among those prescribed mirtazapine (247, 95% CI 138 to 407), followed by venlafaxine (195), paroxetine (180), citalopram (165), sertraline (110), fluoxetine (102) and bupropion (88). In traditional Cox regression, IPTW and PSCM analyses buproprion was associated with a significantly lower risk of suicide death than citalopram (HRs 0.45 to 0.50; p<0.05) and paroxetine (HRs 0.34 to 0.47; p<0.05); the differences were not significant in other analyses. Buproprion was associated with a significantly lower risk of suicide death than mirtazapine in traditional Cox regression and MSM analysis (HRs 0.41 and 0.26, p<0.05), but not in the other analyses. Fluoxetine was associated with a significantly lower risk of suicide death than citalopram (HR 0.45), venlafaxine (HR 0.52) and mirtazapine (HR 0.21) in MSM analyses (p<0.005), but not in the other analyses. Sertraline was associated with a significantly lower risk of suicide death than citalopram in IPTW and MSM analyses (HRs 0.66 and 0.49, p<0.05), but not in the other analyses. Sertraline and fluoxetine were associated with significantly lower risks of suicide death than paroxetine in four of the five analysis methods (sertraline: HRs 0.57 to 0.64, p<0.05; fluoxetine: HRs 0.52 to 0.60, p<0.05). Both sertraline and paroxetine were associated with significantly lower risk of suicide death than mirtazapine in MSM analyses (sertraline: HR 0.23, p<0.005; paroxetine: HR 0.40, p<0.05), but not the other analyses. Instrumental variable analysis found no significant differences between the antidepressants in suicide risk.
Conclusions
In adults with depression, risk of suicide death does not differ among people taking most antidepressants. Several analysis methods, but not all, found lower risks associated with sertraline and fluoxetine than with paroxetine. Differences in findings between analysis methods suggest results should be interpreted with caution.
Commentary
There is increasing evidence suggesting that treatment with antidepressants increases the risk of suicide, particularly in young people and when treatment is initiated or the dose increased. However, methodologically different studies comparing different antidepressants or groups of antidepressants have reported mixed findings. Comparison and interpretation of these findings are problematic because of differing samples and sample sizes, bias in the choice of antidepressants based on symptom profile, differing measures of suicidality, treatment adherence, degree of response and possibility of tolerance.
The current study, a large observational study, was designed to build on this earlier work by comparing the impact of seven commonly used antidepressants on completed suicides among depressed patients using the US Department of Veterans Affairs (VA) Health System. The authors used several different methods of analysis to address the issue of selection bias for treatment and they looked for ‘convergence’ in the findings from the different analysis. This interesting innovation weakened the study because an ‘a priori’ study hypothesis (unidirectional or null) was not specified, and the concept of ‘convergence of findings’ was undefined. This approach can lead to the accusation of ‘a fishing expedition’ to search for the desired results.
It would be difficult to generalise the findings from the VA system to other healthcare systems in the USA and elsewhere because the diagnostic criteria, diagnostic classification and the availability of antidepressants may vary in different settings and countries. However, in the case of depressive illness and treatment with antidepressants, this issue and the actual study findings are almost redundant. Any patient with depressive illness receiving treatment should always be closely monitored, the clinician should always maintain a high vigilance for suicide risk, and the patient and family members should be made aware of the potential high risk of suicide because depressive illness is an important risk factor for suicide. The value of studies such as this is in identifying individual antidepressants that increase the risk of suicide, but clinical practice need not change if there is an appropriate model of service delivery already in place.
Footnotes
-
Sources of funding: The Department of Veteran Affairs, Health Services Research and Development Service and the National Institute of Mental Health.
Footnotes
Competing interests None.