Flaws in analysis lead to misleading conclusions about varenicline’s safety in smoking cessationBMJ 2012; 344 doi: https://doi.org/10.1136/bmj.e3873 (Published 12 June 2012) Cite this as: BMJ 2012;344:e3873
- 1Johns Hopkins University, 680 B, 624 N Broadway, Baltimore, MD 21205, USA
- 2Norwich Medical School, University of East Anglia, Norwich, UK
The meta-analysis by Prochaska and Hilton has several methodological limitations in data analysis and interpretation, which lead to misleading conclusions.1 Despite the removal of cardiovascular events from the trials and a statistical approach that has limited power to detect a significant effect, there is an excess risk of cardiovascular events with varenicline in all five measures reported.
The risk difference model Prochaska and Hilton used is statistically underpowered at low event rates and biases the estimates towards the null.2 This flawed approach is not recommended by the Cochrane Handbook, which states that the Peto odds ratio method was found to be “the least biased and most powerful method” and that risk difference analytical methods “tended to show conservative confidence interval coverage and low statistical power when risks of events were low.”3
They analyse data by treatment level and exclude events occurring in randomised patients. By contrast, we adhered to intention to treat analysis according to the regulations and established and generally accepted scientific principles of the US Food and Drug Administration.4 The higher dropout rate in the placebo group is irrelevant when the intention to treat principle is adhered to.
Their study does not have the optimal information size to detect a significant result. They conflate the lack of significance in an underpowered meta-analysis as clinically insignificant.
Adequately powered randomised controlled trials are needed because none of the trials evaluated cardiovascular events as a primary outcome or was powered to detect individual differences in cardiovascular outcomes between varenicline and placebo. The CATS study, a 52 week post-marketing study comparing varenicline, placebo, bupropion, and nicotine replacement therapy in around 8000 patients, should provide further information on the size of this risk.5
Clinicians need to consider the overall risks of varenicline noted in the prescribing information—serious cardiovascular risk and risks of suicide and depression6—and balance them against its benefits. The United States Veterans Administration does not recommend varenicline as first line treatment for smoking cessation.7
Cite this as: BMJ 2012;344:e3873
Competing interests: None declared.