STATS ARTICLES 2004
Antidepressants and Suicide
October 05, 2004
The New York Times perpetuates a myth.
The controversy over prescribing antidepressants to children and teenagers deepened last week, when a Food and Drug Administration (FDA) advisory committee recommended that the agency issue a “black box” warning to doctors about the increased risk of suicidal behavior from taking such drugs.
The panel tried to weigh the risk of suicide from taking antidepressants against the risk from not taking antidepressants. But the evidence under consideration was insufficient to draw easy or strong conclusions, thus making the issue a complicated one for the public to understand.
To make matters worse, the media did not always promote a consistent story about these risks. For instance, the New York Times reported that
“Children and teenagers who take antidepressants are twice as likely as those given placebos to become suicidal, according to studies presented to the committee,” adding that “If 100 patients are given the drugs, 2 or 3 more will become suicidal than would have had they been given placebos.”
This would imply that the rate of becoming suicidal while using placebos is about two to three percent, and that the actual suicide rate would be even lower (F.D.A. Panel Urges Stronger Warning on Antidepressants, September 15, 2004).
But the paper also quoted Dr. Matthew V. Rudorfer, of the National Institute of Mental Health saying that about 15 percent of teenagers with untreated depression commit suicide. (This would imply that even more exhibit suicidal behavior.)
This level of suicide ought to have raised some eyebrows, given that 9.5% of the American population suffers from a depressive illness according to the National Institute of Mental Health (NIMH) ( the data is derived from Robins LN, Regier DA (Eds). Psychiatric Disorders in America, The Epidemiologic Catchment Area Study, 1990; New York: The Free Press). But the Times made no attempt to explain the huge disparity between the two rates or to press Rudorfer on the source of his claim. As a result, the paper perpetuated a long-standing myth in the study of depression and suicide.
According to the National Institute of Mental Health (NIMH), the suicide rate among all deaths is approximately 1.2 percent. Broken down by age, the rates for 2000 (the most recent figures) are as follows:
In 2000, 10.6 out of every 100,000 persons died by suicide [.01 percent].
For children age 10 to 14, the rate was 1.5 per 100,000 [.0015 percent]
For children 15 to 19, the rate was 8.2 deaths per 100,000 [.0082 percent]
For young adults age 20 to 24 years of age, the suicide rate was 12.8 per 100,000 [.0128 percent].
These are death rates for the entire population; in other words, a “random” 12-year old has a .0015 percent chance of committing suicide.
These rates are higher when you look only at those who are depressed (as depressed people tend to be more suicidal than those who are not depressed); nevertheless, the rates are still significantly less than 15 percent.
Research by Dr. Eve Moscicki, of NIMH finds that over 90 percent of suicides are related to depression, other mental disorders, and substance abuse disorders (often in combination with mental disorders). While the rate of depression among those who commit suicide is extremely high, there is some controversy over the percentage of depressed individuals who commit suicide.
Depending on what group of depressed individuals one is using, the rate of suicide is much less, or much much less, than 15 percent. According to NIMH the most recent data on depression, based on looking at sufferers over long periods of time suggests the following:
Two percent of those treated for depression in an outpatient setting will die by suicide.
Four percent of those treated for depression in an inpatient hospital setting will commit suicide.
Six percent of those treated for depression as inpatients following a suicide attempt or “suicide ideation” will die by suicide.
In other words, depressed people are more likely to commit suicide than the general population, people hospitalized for depression are even more likely to commit suicide, and people hospitalized due to an attempt to commit suicide form a group at even greater risk for committing suicide.
The original source of the suicide myth is a 1970 review and analysis of 17 studies of depressed patients by Guze and Robins (Br. Journal of Psychiatry), in which the authors concluded that 15 percent of depressed patients commit suicide. Two decades later another analysis reviewed 13 additional studies and concluded an even more alarming rate – 18.9 percent – of depressed patients were projected to die by suicide (Goodwin and Jamison, Suicide, in Manic Depressive Illness, Oxford University Press, 1990).
These figures were debunked by Drs. John Michael Bostwick and V. Shane Pankratz. (Affective Disorders and Suicide Risk: A Reexamination, American Journal of Psychiatry, December, 2000). As pointed out by these authors, there were several methodological problems in both of these studies.
The first significant problem is that the original studies use a very strict definition of depression. Their samples consist of patients who were almost exclusively in the highest risk category, having been hospitalized for suicidal behavior. In 1970, patients with mild to moderate levels of depression (who may be treated for depression under today’s definition of depression) were not considered depressed. The rates for suicide are skewed because the sample is more suicidal than the general population diagnosed with depression today.
The second mistake which led to the faulty number was more subtle. The high suicide rates were actually percentages of the dead who died by suicide (and not of the whole sample). This is also called the proportionate mortality prevalence. Proportionate mortality prevalence is a measure of conditional probability; it is the probability someone will have died by suicide given he/she is dead within the follow-up period of the study.
To illustrate the statistical error, suppose that 500 severely depressed patients are observed over six years. At the end of the time, suppose 100 patients have died, 25 from committing suicide. Given the methodology used in the original study, the conclusion is that 25 out of a hundred, or 25 percent of the patients died by suicide. But it is not clear that 25 percent of the remaining 400 patients will also die by suicide, as they have already survived six years.
Thus the risk of suicide might be (and has in fact been observed to be) much lower than 25 percent. This effect is exacerbated by short follow-up periods (often just a few years). The percentage of suicides among those who have died is disproportionately high in a short period of time, because there is less time for non-suicide deaths to occur.
A more accurate picture of the rate of death due to suicide would be obtained by following the patients over their whole lives. A simpler method would be to divide the number of suicide by the whole population – in our example, 25/500 or 5 percent. According to Bostwick and Pankratz,
“The percentage of subjects dead due to suicide (case fatality prevalence) is a more appropriate estimate of suicide risk than the percentage of the dead who died by suicide (proportionate mortality prevalence).”
Using the case fatality prevalence requires a reasonable follow-up period and presumes that suicides are much more likely to occur immediately after a hospitalization (within the follow-up period) than many years later.
Lastly, it seems apt to mention another misleading message given both by the New York Times and by Reuters in Yahoo! Health. Although both articles were about the use of antidepressants in children and teenagers, they quoted parents of young people on antidepressants who committed suicide aged 25 and 21, respectively. The message regarding the risk for children and teenagers (up to age 19) may be confused when mixed with examples of young adults, for whom the effectiveness of antidepressants against suicide was not in question.