Authors’ replyBMJ 2009; 339 doi: https://doi.org/10.1136/bmj.b4969 (Published 01 December 2009) Cite this as: BMJ 2009;339:b4969
- David Gunnell, professor of epidemiology1,
- David Irvine, pharmacoepidemiologist2,
- Lesley Wise, senior pharmacoepidemiologist2,
- Claire Davies, senior pharmacovigilance assessor2,
- Richard M Martin, professor of clinical epidemiology1
- 1University of Bristol, Department of Social Medicine, University of Bristol, Bristol BS8 2PS
- 2Vigilance and Risk Management of Medicines, Medicines and Healthcare products Regulatory Agency, London SW8 5NQ
Moore and Furberg highlight the difference between our findings and the strong signal of suicide risk for varenicline seen in yellow card reporting.1 2 They use the high ratio of non-fatal self harm to suicide (83:1) in our study to suggest our data on incident self harm are unreliable and that “more realistic ratios of suicidal thoughts, acts, and deaths” are seen in the yellow card data. This is potentially misleading.
The ratio of non-fatal self harm to suicide in their table is 77:22 (3.5:1). This is 6-8 times lower than the ratio seen in our previous General Practice Research Database study (30:1),3 a ratio in keeping with national epidemiological data based on estimates of >140 000 cases of self harm presenting to hospital and 5000 suicides annually in the UK.
The 95% confidence intervals for the proportion of acts of self harm that were fatal in our study (2/168, 1.2%) were 0.1% to 4.2%. The upper limit is compatible with a ratio of fatal to non-fatal self harm of 1:25.
In contrast, the 95% confidence intervals for the proportion seen in Moore and Furberg’s yellow card data (22.2%) are 14.5% to 31%, suggesting a ratio of non-fatal to fatal self harm of at most 7:1. This is considerably less than epidemiological data suggest is reasonable. The differences between the ratios are most likely due to the small sample size of our study (low precision) and the greater likelihood that more serious potential side effects are reported to regulatory bodies (yellow card data). We acknowledge that suicidal thoughts were under-recorded in the General Practice Research Database.
Most pharmacoepidemiologists would agree that cohort studies provide stronger evidence of causality than adverse event reporting. Yellow card reporting is useful for identifying possible safety signals, but the next step is to investigate such concerns using randomised controlled trials or cohort or case-control studies, with care being taken to control for confounding factors when using observational (cohort and case-control) approaches.
Cite this as: BMJ 2009;339:b4969
Competing interests: DG is a member of the Pharmacovigilance Expert Advisory Group of the MHRA. DI, LW, and CD are employees of the MHRA. RMM is a member of Independent Scientific Advisory Committee for MHRA database research (ISAC).