The FDA sample finds that the risk for suicidality (suicidal preparation or actions) on SSRI versus placebo is increased to a non statistically significant degree. The odds ratio is 1.23, yielding a p-value of .31. The same analysis finds that suicide risk, when measured as suicidal thoughts or worse (See table 15 in the FDA report), is actually decreased to a nonsignificant degree by SSRIs compared to placebo (OR = .86, p = .16). So, case closed, eh? No evidence of increased suicide – let’s go play golf. Well, not quite. The FDA did find an enhanced risk for suicidality among individuals age 25 and under, but found a protective effect for medication among the elderly (see tables 22-27 starting on page 43).
There’s much more to this analysis, especially when placed in the context of prior research…
Based on an analysis on FDA’s database of trials, Khan et al. (2000) estimated that the risk of completed suicide was about twice as high in participants who took antidepressants versus those who took placebos. Healy (2003) noted that some of the suicides classified as occurring on placebo in trials actually occurred during a placebo washout phase, during which all participants were receiving placebo. Since nobody at the time was receiving antidepressant medication, counting placebo washout events as occurring on placebo is not justified. Healy eliminated suicide and suicidal acts that occurred during placebo washout from his calculations and determined that the rate of completed suicides on SSRIs compared to placebo was higher (Odds ratio = 3.1; 95% CI .4 – 23.1). The odds ratio for any suicidal act on SSRI versus placebo was 2.0 (95% CI 1.2 – 3.3). For readers unfamiliar with what an odds ratio means, please see here and here. For a slightly more advanced treatment of the topic, see here. For a longer post on what other meta-analyses of clinical trials comparing SSRIs to placebo have found, please go here. Suffice to say they have consistently found an elevated risk of suicidal acts and/or risk of completed suicide on SSRI compared to placebo.
Below is a discussion of problems related to the FDA’s method of data collection and how the FDA’s data differs, rather widely, from that found by another analysis.
The FDA study is based upon reports from sponsors. Thus, drug companies were trusted to provide data in an honest fashion. Given that in the cases of paroxetine, sertraline, and fluoxetine, sponsors misleadingly labeled placebo washout suicidal acts as occurring during the double-blind placebo phase, this is an issue. I am not certain if this happened in the current FDA report, but it cannot be ruled out as a possibility. I will be never be sold on the FDA data because I doubt all of these data will be released in a manner that would allow determination of when suicidal events happened in the selected clinical trials. I strongly encourage readers to examine table 14 in the FDA report (page 103) and compare it to Healy’s figures in Table 1 of this document. The rate of suicidal acts in placebo is much higher in the FDA database (0.72%) compared to Healy’s database (0.48%). In addition, the rate on SSRIs (citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, and sertraline) is 1.55% in Healy’s database versus .060% in the FDA database. In comparison to Healy, the FDA has more risk on placebo and much less on SSRIs. Yes, it is clear that the FDA database contained a larger sample of trials than did Healy’s earlier analysis. However, it is equally clear that the data in Healy’s database is likely more trustworthy, as he was able to sort out placebo versus placebo washout, whereas the FDA relied on sponsors to do the sorting out of this data. Healy’s study also appears to have utilized some uncontrolled trials and some active comparator controlled trials (trials in which a medication was compared only to another medication, not a placebo), whereas the FDA numbers are based only on placebo controlled trials. So, yes, I understand that these are different samples of studies, but it is important to note the large differences between them in terms of results. Let’s also not forget that the MHRA (roughly equivalent to the USA’s FDA) did their own analysis and found that the use of SSRIs (even excluding paroxetine, which seems related to a greatly elevated risk of suicidal behavior relative to placebo – see here and here) increased the risk of suicidality.
More will be said by people better informed than myself in the next few days and I hope to keep everyone apprised of what the FDA and others have to say publicly about these findings. The hearing is scheduled in two days (Dec. 13), so more should be forthcoming soon.