“Counseling is a must with this smoking cessation aid,” stated a PURL published in March 2012 (J Fam Pract. 2012;61:156, 176). “Varenicline [Chantix] is associated with a small but significant harmful effect on CV outcomes.” That statement, and the PURL itself, was based on a meta-analysis published in 2011 by Singh et al.1 The meta-analysis included 14 randomized controlled trials (RCTs) that compared varenicline with placebo for the occurrence of serious cardiovascular disease (CVD) events, including myocardial infarction, coronary artery disease, arrhythmias, stroke, sudden death, and any related coronary death. RCTs that reported no CVD events were excluded.
Using a Peto odds ratio [OR] for analysis, Singh et al reported that varenicline use increased the risk of CVD events compared with placebo (OR=1.72; 95% CI, 1.09-2.71). A more recent meta-analysis by Prochaska et al,2 however, challenges the validity of the Singh meta-analysis. As members of the Family Physicians Inquiries Network, which produces the PURLs, we would like to address the questions this newer study raises about varenicline’s actual risk.
The Prochaska meta-analysis included all 14 RCTs analyzed by Singh, and used the same CVD event outcome measures. In addition, Prochaska included 8 trials in which no CVD events were reported, some of which were published after the Singh meta-analysis. And rather than use the Peto OR to estimate the risk, Prochaska calculated the absolute risk (AR). The result? The researchers found no difference in CVD events in the varenicline group compared with placebo (AR=0.27; 95% CI, -0.10 to 0.63; P=.15).
This is a good example of how inclusion criteria, subsequently published clinical trials, and the choice of statistical methods can lead to conflicting conclusions from meta-analyses on the same topic. Including studies that showed no adverse CVD events is more likely to capture the true risk than excluding them, and reporting AR is more meaningful than estimating relative risk based on the Peto OR.
Therefore, the Prochaska findings are more convincing. Given the effectiveness of varenicline and the known benefits of successful smoking cessation, it is important for clinicians to understand that the true risk of CVD adverse events attributable to varenicline is extremely low or even nonexistent.
Dionna Brown, MD
Bernard Ewigman, MD, MSPH
Chicago