Putting Numbers on Antidepressant Risk
(Revised April 2005)
The FDA black box warning on antidepressants' use in children has been a subject of considerable controversy. More data to go on would help. Here is one study that has attempted to put a number on the risk of antidepressants, quantifying the switch rate to bipolarity by looking at a database of 88,000 patients: Martin et al.
This is a very complex statistical analysis, and there is a risk that I may mislead you by taking their results out of the context of their article. However, I think I've got it right. Since I fear there are people out there who still underestimate the risk which antidepressants pose, I'm going to take the liberty of reproducing one of the figures from this research study, one which visually seems to demonstrate the problem rather well:
In this figure, the risk of manic conversion is a cumulative figure: as more kids "switched" to bipolar, the later ones were added to the earlier ones, so that the figure represents the total number who switched at any point in time. That's important to understand so that you can look at the total conversion rates by the end of the study period: in kids who were exposed, the total number of "switchers" by the end of the study period was nearly half the group, see that? By comparison, in those not exposed, the total number of switchers was less than 20% of the group.
Dr. Martin and colleagues were also able to create a rough estimate of "number needed to treat": how many kids or young people do you need to treat with antidepressants in order to see one "switch". Obviously if the switch rate was high, this number would be low, and vice versa. Granted that many switches are going to be spontaneous, not truly associated with the antidepressant; and granted that looking at pharmacy data and diagnoses reported to insurance companies is a very rough means of knowing what's really going on out there. But, some data may be better than none. This study may provide at least a "ballpark" estimate of switch rate risk. As you can see, there is difference by age:
|Age||Number needed to see one switch||(95% confidence
Obviously one of the main findings in this study is that kids are different than adults. There are two ways in which kids could differ. First, their brains could be different. Young kids/ might simply be more susceptible to the switch-generating effects of antidepressants than older kids or adults. This is definitely being considered as a mechanism for these results.
Alternatively, we must take into account the fact that children who present with depression severe enough to warrant use of an antidepressant are more likely to have bipolar disorder which just has not manifested itself as such yet. In other words, kids with bipolar disorder often have depression as the first manifestation of that illness. If a careful family history is not performed, one can miss the bipolarity entirely and treat with antidepressants. This can cause a "switch", as antidepressants are known to cause mania and hypomania in patients with bipolar disorder.
Big Picture of Antidepressant Prescribing, And Risk, In Adults
Here is what the trend in antidepressant use looked like in Britain up to 2002 (more recently we might see some decreases in the trends shown here):
This graph comes from a review which provides a nice overview of the risk story (getting just slightly dated now in late 2005, as it was published in mid-2004 so was written in late 2003). It includes a perspective I rarely see mentioned so wanted to include it here: "if the risks of SSRI-associated suicidal behavior seen in children were to apply to suicide in adults, the number of "antidepressant induced" suicides would be small enough to be masked by currently favorable suicide trends" (which are reviewed in the article). See Gunnell and Ashby.
Nevertheless, as they emphasize in a more recent analysis on this issue:
...the number of patients who need to be treated with an SSRI (fluoxetine) to get one response... is four to seven. Conversely, our data for non-fatal self harm indicate that the best estimate of the number needed to treat to harm is 759.
In other words, in adults, antidepressants will help many more people than they harm, if they do indeed harm anyone. The authors emphasize that it would take a huge study to directly assess that risk ("about 1.9 million subjects"). See Gunnell 2005.