Selective serotonin reuptake inhibitors (SSRIs) and suicide in adults: meta-analysis of drug company data from placebo controlled, randomised controlled trials submitted to the MHRA's safety review

When the randomised control trial data submitted by pharmaceutical
companies is used for a meta-analysis to investigate whether SSRIs are
associated with increased risk of suicide a number of limitations are
there. First of all suicide is a very rare event even in patients with
severe depression. As correctly pointed out by the authors even after
pooling the data from hundreds of RCTs the study did not have sufficient
power and results should be interpreted cautiously. The RCTs had only few
cases to confirm or rule out an increased risk of suicide with SSRIs. To
complicate this further there is an obvious chance of underreporting of
suicide related outcomes in these clinical trials because the research
question in most of these RCTs would be focused on the efficacy of the
drug. Since the primary aim is not to check the association of SSRIs and
suicide, the reported adverse event may have been considered to be due to
the mental illness rather than medication by the researchers which would
lead to under-reporting. Reporting of events in subgroups is again
difficult because of the lesser number of patients. Even though there is a
less chance of differential reporting in placebo controlled trials it is
very important to make sure that allocation concealment was adequate,
which may influence the degree of effect. The selective inclusion of drug
company data would result in a location [publication] bias.

According to MHRA data [Incidence of possible suicide-related events
by treatment group and age group: adult active-control trials] data of
Paroxetine and a comparator in the age group above 70 shows an increase
incidence of possible suicide-related events in Paroxetine group.
[Paroxetine n/N (%) - 4/457 (0.9%); Comparator n/N (%) - 1/390 (0.3%) ;
Odds ratio (95% CI) - 3.4 (0.4, 30.8) ; p-value 0.38 ]

In spite of a wide confidence interval an large p –value this is an
interesting finding and we wonder whether it would be due to the
differences in the pharmaco-dynamics in the elderly population. In that
case this would be very significant because of the closeness to the real
life scenario than the research setting as observed by Alexander M
Ponizovsky in rapid responses to this article on 22 February 2005.

In the meta analysis the authors attempted to quantify heterogeneity
and recognized that heterogeneity between trials and individual products
is likely to have been masked. However we find that the paper is subject
to a high level of clinical heterogeneity because, the data is summarised
for all treatment indications, except for Citalopram. Moreover difference
in the length of the follow-up in different trials and, randomization
ratios are not adequately reported which could lead to over or under
estimation of risk.

Finally when the data provided by Prof. Healy is included in the
analysis there is a major shift of the study result in favour of placebo.
The call for urgent research by the authors to identify people at risk is
the most appropriate approach to clarify this problem. ,

Competing interests:
None declared