Reexamining the Link Between Antidepressants and Suicidality in Children and Adolescents

In October 2004, the FDA mandated that a black box warning be added for all antidepressants used in children and adolescentsóa move that sparked widespread concern about the use of these drugs.¹ The FDA's decision followed similar action by the British Medicine and Healthcare Products Regulatory Agency.

The FDA based its decision on a review of pooled analyses of 9 antidepressants (selective serotonin reuptake inhibitors [SSRIs] and other novel antidepressants) used in 24 short-term (4 weeks to 16 weeks), randomized, placebo-placebo- controlled trials, both published and unpublished.² The trials involved 4587 children and adolescents with major depression, obsessive-compulsive disorder, and other disorders. The initial metaanalysis revealed an average risk of broadly defined suicidal behavior of 4% as determined by adverse event reporting, which is twice the placebo rate.

The FDA also concluded that although there had been trials supporting the efficacy of citalopram(Drug information on citalopram), sertraline(Drug information on sertraline), and paroxetine(Drug information on paroxetine), only fluoxetine(Drug information on fluoxetine) had met the efficacy standard of showing a benefit in at least 2 randomized controlled trials. They noted that a third, multi-site, randomized controlled trial, the Treatment for Adolescents with Depression Study (TADS), confirmed that fluoxetine, alone or in conjunction with cognitive-behavioral therapy, was an effective treatment for depression.³ In the TADS, 61% of the participants responded to fluoxetine alone, and 71% responded to fluoxetine plus therapy, which was twice the response rate of 35% for participants who received placebo. Weighing the efficacy data against the risks of suicidal behavior and ideation, the FDA concluded that only fluoxetine showed significant advantage over placebo and, therefore, the benefit of this drug exceeded the risk.

Hammad and colleagues⁴ reported on a more detailed reanalysis of all the randomized placebo-controlled studies, including the TADS data. In this study, the investigators analyzed the suicidal behavior data separately from other adverse events; they also separated the data on major depressive disorder (MDD) trials from data from other disorders. The overall risk ratio of suicidal ideation or nonfatal self-injurious events compared with placebo was found to be 1.95 (95% confidence interval [CI], 1.28 to 2.98%) for all disorders and 1.66 (95% CI, 1.02 to 2.68) for MDD.

Table 1 shows the risk ratios for suicide attempts and suicide behavior for each drug within MDD trials as compared with all other trials. Although there were some differences in the degree of risk among the 9 antidepressants evaluated, the FDA did not exempt any antidepressant from the black box warning because of differential risk rates. The FDA has not contraindicated any of the medications for use in pediatric populations but encourages physicians to balance the risk with clinical indications.

The FDA's actions and related studies have caused public consternation and sparked an ongoing debate in the medical community. The profusion of commercially sponsored trials that were not required to be reported for analysis created the perception of a reporting bias.⁵ The FDA quickly moved to start requiring the reporting of all dataó positive and negative.

Psychotropic medication use in children and adolescents has come under even more scrutiny with concerns about direct-to-consumer marketing and the increase in medication prescriptions for children. These concerns were illustrated by the recent recommendation of an FDA advisory committee to add a black box warning and further investigate possible cardiovascular risks of medications used to treat attentiondeficit/ hyperactivity disorder. However, another FDA panel then reversed the recommendation to add a black box warning.

Since the FDA's first public health advisory on antidepressants, reports suggest that there has been a 20% drop in the number of antidepressant prescriptions dispensed for minors.⁶ The public health consequences of this are unclear. In June 2005, a report issued by the AMA Council on Scientific Affairs stated that SSRIs should continue to be available as a component of depression treatment for children and adolescents.⁷ The AMA requested an independent review of the current data and asked the FDA to evaluate the impact of its regulatory actions on treatment patterns, compliance, and access to care.

The significance of these data is far from clear, because there are a variety of alternative explanations as well as methodologic limitations that make the findings concerning suicide risk and clinical efficacy difficult to interpret. For instance, the trials were carried out in a pediatric population, in which actual suicide is rare, so the events reported were very small in number. The relationship of suicidal ideation and suicidal behavior to completed suicide in this age group is also unclear. Suicidal ideation is common among adolescents, occurring in approximately 19%.

The studies analyzed by the FDA involved spontaneously reported adverse events rather than explicitly assessed behaviors. They also lacked uniformity in how suicidal behaviors were defined and lacked adequately explicit measurements of suicidal behaviors or ideation, including measures of severity of intent. It is important to emphasize that there were no suicides among the nearly 5000 cases reviewed by the FDA and that behaviors considered by external consultants as probable suicide attempts were rare. Thus, the pooled data include behaviors that may not pose a significant risk of suicide. Finally, adherence to assigned treatment was not measured, and investigators did not account for the possibility that discontinuation symptoms contributed to adverse events.⁸

With respect to efficacy, the duration of the trials (less than 16 weeks) was, in general, too short to allow identification of the long-term beneficial effects of antidepressant medications. Simon and colleagues⁹ recently published a report on attempted and completed suicide in a large HMO sample for whom antidepressants were prescribed. They found that the highest risk of suicide attempts and completion was in the month before antidepressant use, and the risk went down steadily during the course of treatment over 6 months, even for patients younger than 18.

The large number of multi-site trials included in the FDA meta-analysis presented additional problems for efficacy measurement. These include high variance because of multiple sites with few patients in each site, probable variations in methodologic and recruiting strategies, and variable placebo response rates.⁸

Mann and colleagues¹⁰ noted other problems with these trial results. They found that the trials did not document past suicide attempts nor did they include, in general, patients with current suicidality. As a result, these authors recommended that future trials be designed with high-risk patients, using stratification to avoid confounding effects.

Various alternative explanations for the trial results have been advanced. Among these is the possibility that the patients receiving medications were more able to report suicidal behavior (either because they elicited attention for other reasons or they were more communicative).⁴ It has also been posited that some of the adverse effects of antidepressant treatmentósuch as insomnia, agitation, restlessness, and irritabilityóoccurred in patients at risk for bipolar disorder who may not have been adequately monitored or treated.¹¹

Epidemiologic data also fail to support the conclusions of the FDA meta-analysis, since the recent increase in SSRI prescriptions has been correlated with a decrease in the youth suicide rate.¹² Adult data also show a decrease in the suicide rate corresponding to the period of increased SSRI prescriptions (with the exception of the elderly population).

In addition, autopsy studies have failed to find antidepressant use in most suicide victims, which is also inconsistent with the FDA's findings.⁴ A Swedish toxicology study compared suicides with natural deaths and reported that no SSRIs were detected in 52 suicides in persons younger than 15 years. In the 15– to 19– year-old age group, persons with detectable concentrations of SSRIs had a lower relative risk of suicide over a period of 9 years than those with detectable levels of other antidepressants. 13 Data on regional trends in prescribing antidepressants correlated with suicide statistics suggest that antidepressant treatment is inversely related to suicide for older adolescents and for males in general, who are at highest risk for suicide.¹⁴