The December issue of the Archives of Internal Medicine contains a paper based on the Women's Health Initiative dataset (a large cohort of community dwelling women aged 50-79 in the US) that describes a possible association between antidepressant use and stroke (link to abstract). Hemmorhagic stroke (strokes caused by bleeding in the brain) seemed to account for most of the association. The authors also looked for an association between antidepressant use and coronary heart disease - finding no association. This report may cause some alarm for people taking antidepressant medications, but it is important (in my view) to consider three factors when interpreting these results. First, depression itself is associated with cardiovascular disease, so studies such as this one are vulnerable to "confounding by indication" - a situation where the reason for use of a medication causes a problem that subsequently seems to be due to the medication itself. The study used sophisticated means to control for confounding by indication, but even the most sophisticated approaches are not always effective for controlling this type of bias. Second, this is not the first study to look at this possibility, and the literature as a whole has been inconsistent. On the other hand, some antidepressants are known to increase the risk of bleeding related complications, so what is being reported is biologically plausible. The risk-benefit trade-offs for antidepressant use need to be better delineated so that people can make informed decisions about depression treatment - this study is an example of one that is pursuing this goal. The risks and benefits may play out differently in people with severe depression and/or other risk factors for stroke - hopefully, research in this area will remain active until these issues are decisively resolved.