The relationship between suicide and antidepressant use is marked by a sharp divergence of professional opinion. Based on evidence largely obtained from antidepressant licensing randomised controlled trials (RCTs), the United States and United Kingdom have issued government safety warnings about antidepressant use increasing suicide risk in young people. Conversely, toxicology and epidemiological studies consistently suggest that antidepressants lower suicide risk (1, 2). A possible reason for this contradictory evidence is that suicide is a rare event. To construct an RCT to detect an effect on suicide mortality of 20% would require around two million subjects (1). RCTs are therefore forced to rely on measures of non-lethal suicidality which may not be very informative about drug impact on mortality. Alternative models to study the link between suicide and antidepressants are needed. While these models have unavoidable weaknesses, they are all that is feasible.
In this issue of Acta Psychiatrica Scandinavica, Isacsson et al. (3) present a model that looks for differences between observed and expected number of suicides with antidepressants detected by toxicological examination. They assume that antidepressants would be more frequently detected i) in those who committed suicide than those who died of other causes; ii) in suicides who received hospital care than those who did not and iii) in suicides who received hospital care for depression vs. those who were hospitalised for other diagnoses. What they report is surprising and even, to use their term, remarkable. While those who committed suicide had higher rates of detection of antidepressants than those who had died of other causes (22.4% vs. 6.5%) and those hospitalised had higher rates than those who were not (33% vs. 14.8%) when hospitalised patients were divided into those with depression only, those with depression plus comorbidity and those with other diagnoses, those with depression only have a marginally higher rate of detected antidepressants (15.2%) than those who were not hospitalised. In other words, those with severe depression (i.e., requiring hospitalisation) have the same rate of antidepressant toxicology as those who may not even have had contact with the health care system.
Because antidepressant treatment is the treatment of choice for those hospitalised for depression, even a conservative estimate would suggest that such patients would have a rate at least equal to those patients hospitalised for other reasons i.e., 37.3%. The authors contend that the most reasonable explanation for the expected, but not observed cases, is that antidepressant medication prevents suicide among those hospitalised for depression. The assumptions this finding relies on appear conservative and credible and need to be taken seriously. Ideally, prescribing and compliance rates would help clarify whether the depressed inpatients actually received antidepressants but it seems reasonable to assume they do. The results, as most interesting results do, lead to a number of other questions. Why do depressed inpatients with comorbidity not display a similar protective effect of antidepressants? Do they not receive antidepressants or not take them or do they not work? Is the effect only found in patients with severe depression?
Regardless, the principle finding supports the hypothesis that antidepressants reduce suicide in depressed patients at a population level. These findings are supported by an analysis using mortality data from 26 countries for up to 25 years which concluded that increases in SSRI prescribing are associated with reduced rates of suicide (1). A recent clinical trial that included patients with significant suicidality (a weakness in industry RCTs) also reported a sustained reduction in suicidal ideation and attempts during 6 months of antidepressant treatment (4).
One problem however is that these studies cannot reassure us that an individual patient will not have an increase in suicidality when started on antidepressants. There are a number of reported cases suggesting that antidepressants are associated with increases in suicidality in a small subset of depressed patients. Hunter et al. (5) also in this issue of Acta Psychiatrica Scandinavica have attempted to specify these individuals using the neurophysiological measure of changes in midline-and-night-frontal (MRF) cordance. They report that subjects who experienced worsening suicidality during treatment showed a large and transient decrease in MRF cordance biomarkers 48 h after beginning the medication. While the study is small and should be seen as hypothesis generating it does offer a model to help deal with the difficulties around antidepressant prescribing.
The apparent divergent views around the relationship between antidepressants and suicide may both be correct. Some individuals are vulnerable to increased suicidality when prescribed antidepressants but these numbers are small and the overall population effect is that increased antidepressant prescribing rates are associated with a fall in rates of suicide. On a population level, it is desirable that antidepressants are prescribed to those who are depressed but in a few individual cases it may not be. Hunter et al. (5) offer a possible way to help identify such cases so that the potential harm is reduced while the benefits are maintained. Perhaps, it is possible to please everyone occasionally.