Effect of Date of Drug Marketing on Disproportionality Measures in Pharmacovigilance

Background: Warnings concerning an increased risk of suicide in patients treated with selective serotonin reuptake inhibitors (SSRIs) re-emerged in early 2003, culminating in the broadcast of a television programme in the UK. In the following months, cumulated proportional reporting ratios showed that the most recently marketed drug, escitalopram, had a much higher proportion of reports of suicide to other adverse drug reactions (ADRs) than the other drugs in the class.

Objective: To study the reporting patterns over time concerning suicide with the six SSRIs marketed in the UK as of March 2003 and their potential effect on disproportionality signal detection.

Methods: Monthly cumulated numbers of reports were obtained from the UK Medicines and Healthcare products Regulatory Agency (MHRA), from the time of the first marketing of the drugs concerned and monthly for the 2 months prior to and the 9 months following the broadcast of the television programme (broadcast date: 11 May 2003), and the monthly ratio of suicide to other reports was computed for each SSRI individually and for all SSRIs combined.

Results: Of the six SSRIs studied, five (citalopram, paroxetine, fluoxetine, sertraline and venlafaxine) had been marketed for several years and escitalopram for only a few months. At the end of the analysis period, 1.42% (4/281) of all ADR reports for escitalopram were of suicide versus 0.58% for the other five drugs combined (146/25 197). For all SSRIs combined, suicide represented 0.5% (123/24 315) of reports before the broadcast of the television programme, and increased to 2.3% (27/1163) following the programme. For escitalopram, suicide represented 1.1% (1/89) of all ADR reports before the television programme and 1.6% (3/192) afterwards. For the five other drugs combined, suicide represented 0.5% (122/24226) of ADR reports before the television programme and 2.5% (24/971) afterwards (varying from 1.4% to 4.7% for the various drugs). The post-programme events represented 68% of all reports and 75% of suicides for escitalopram, whereas for older drugs they represented 3.6% of reports and 13% of suicides.

Conclusion: For older drugs, the events reported during the high-reporting post-television programme period were diluted by years of low reporting. For escitalopram, although the television programme had little absolute impact on the number of reports, because the drug had been on the market for such a short period of time, a large relative effect was observed. Differential effects related to time on market on cumulated reporting of adverse drug reactions should be taken into account when analysing spontaneous reporting databases with automated signal generation methods after an alert has changed the spontaneous reporting patterns. Proper use of measures of disproportionality requires thorough knowledge of potential biases and careful analysis of reporting patterns. We found no obvious differences between SSRIs once these were taken into account.