Thinking Outside the Black Box on Antidepressants


By Rob Mitchum

In 2004, the Food and Drug Administration placed their equivalent of a scarlet letter on the antidepressant fluoxetine. Acting on the compiled results of several clinical trials, the FDA affixed its foreboding “black box warning” on to the drug best known as Prozac, preaching caution about increased suicide risk in children and young adults who take the medication. The decision made headlines around the world, leading to similar warnings in other countries and widespread decreases in the amount of prescriptions issued for the drugs to both pediatric and adult patients.

The FDA’s decision, voted on by a 23-member scientific advisory panel, was not unanimous, passing 15-8. One of the panel’s dissenters was Robert Gibbons, a health statistician then at UIC who later joined the University of Chicago Medicine in 2010. Gibbons thought that the studies used by the FDA, based largely upon retrospective data and adverse event reports, left too much room for alternative explanations.

“I didn’t find the data compelling,” Gibbons said. “For example, kids randomized to drugs would have more side effects, more contact with their doctor, and would have more opportunity to talk about suicidal thoughts.”

Despite Gibbons’ misgivings, there wasn’t an alternative data set to replicate or dispute the conclusions of the FDA’s meta-analysis. But earlier this week, Gibbons and a group of collaborators published their own statistical analysis of the clinical trials assessing the relationship between antidepressants and suicide risk in Archives of General Psychiatry…and reached a very different conclusion.

The new analysis used longitudinal data collected in 40 pharmaceutical company clinical trials and one large NIH study of antidepressants in adults, senior citizens, and children. Instead of measuring suicide risk through reports of adverse events, these data sets contained the results of weekly planned assessments for each patient on a psychiatric scale that measures depressive and suicidal thoughts — allowing researchers to trace week-by-week the effects of drug vs. placebo on the symptoms.

Adult and geriatric clinical trials for two different antidepressant drugs, fluoxetine and venlafaxine, were analyzed, and both medications were found effective in reducing suicide risk and depression symptoms. Furthermore, the two effects were also found to be statistically associated, suggesting that the drugs reduced suicidal thoughts and behavior by alleviating depression. Therefore, Gibbons said, effective treatment of major depressive disorder and careful monitoring that the drug is actually working are both important for a patient’s safety.

“Basically, the results say that the mechanism by which the antidepressants affect suicide rates is by decreasing depression,” Gibbons said. “It follows that if a treatment is not working for an individual, the risk for suicidal behavior and perhaps worse remains high.”

The four pediatric trials included in the new analysis, which all tested fluoxetine, offered more mixed results. A significant reduction in depression symptoms was found across the four studies, suggesting that the drugs were working on their primary aim. But instead of supporting the justification for the black box warning, the new analysis found that the antidepressant did not increase suicidal thoughts and behaviors in children. Nor, however, did the antidepressant decrease suicidality in the population, suggesting that the riddle is not completely solved.

“I think that this paper supports the general idea that the effects of antidepressants in kids and adults are not really the same,” Gibbons said. “In kids, we don’t see a harmful effect, but we do see a disassociation between the beneficial effects on depression and the potential beneficial effect on suicide. Maybe children think about suicide in part because of depression, but also maybe due to other reasons not related to depression that are not affected by antidepressants.”

Despite this cloudy relationship, Gibbons thinks the new analysis should alleviate the fears of pediatricians or other clinicians who became reluctant to prescribe antidepressants — or even diagnose depression — for their patients, when the black box warning was added. While pharmacotherapies are only one way of treating depression, the new analysis supports that antidepressants can be an effective and safe option for patients who may be in serious danger.

“I hope that the warnings will not prevent depressed children and adults from getting treatment for depression,” Gibbons said. “The greatest cause of suicide is untreated or undiagnosed depression. It’s very important that this condition be recognized and appropriately treated and not discarded because doctors are afraid to be sued.”

More broadly, Gibbons argues that the new analysis exposes the flaws inherent in using traditional meta-analyses to determine clinical choices and policy.

“You can end up making really bad decisions by using traditional meta-analysis to pull information from studies. You end up with answers that are really occurring at the level of the individual analyzed at the level of a study, and it’s very easy to get the wrong answer,” Gibbons said. “It’s really important because so much medical decision making is based on these meta-analyses.”

[Read more coverage of the study at Los Angeles Times, NPR, and Forbes.]


Gibbons, R., Brown, C., Hur, K., Davis, J., & Mann, J. (2012). Suicidal Thoughts and Behavior With Antidepressant Treatment: Reanalysis of the Randomized Placebo-Controlled Studies of Fluoxetine and Venlafaxine Archives of General Psychiatry DOI: 10.1001/archgenpsychiatry.2011.2048

About Rob Mitchum (525 Articles)
Rob Mitchum is communications manager at the Computation Institute, a joint initiative between The University of Chicago and Argonne National Laboratory.
%d bloggers like this: