The Food and Drug Administration and varenicline: should risk communication be improved?

On 11 May 2006, the US Food and Drug Administration (FDA) approved varenicline (Chantix, Champix) for use in smoking cessation; it was the first oral non-nicotine treatment licensed by the FDA since bupropion in 1997 [1]. However, it was not long before concernswere raised about the neuropsychiatric safety of the medicine. Anecdotal reports from popular press and internet sites, in addition to post-marketing case reports and reports to the FDA’s Adverse Event Reporting System (AERS), suggested that some patients prescribed varenicline had experienced suicidal behaviour [2]. As a result, in November 2007 the FDA announced that they were conducting two safety reviews of varenicline’s associations with suicidality and neuropsychiatric adverse events not related to suicidality [3]. Findings from these reviews were published in 2008, with the FDA concluding that ‘[varenicline] may cause worsening of a current psychiatric illness even if it is currently under control and may cause an old psychiatric illness to reoccur’ [4]. In July 2009 the FDA went further, mandating that varenicline carry a ‘Black Box warning’ [5]. These warnings highlighted ‘the risk of serious neuropsychiatric symptoms in patients using these products’. Symptoms included changes in behaviour, hostility, agitation, depressed mood, suicidal thoughts and behaviour and attempted suicide. The FDA subsequently commissioned two large observational studies to investigate the potential neuropsychiatric effects of varenicline using observational data from military veterans. Pfizer (the manufacturer of varenicline) was also instructed to undertake a large randomized trial: Evaluating Adverse Events in a Global Smoking Cessation Study (EAGLES) (NCT:01456936). Results from the observational studies were reported in October 2011. In the first, 14311 patients prescribed varenicline were matched to similar patients prescribed nicotine replacement therapy (NRT) using propensity scores. Patients prescribed varenicline were no more likely than those prescribed NRT to be hospitalized for psychiatric problems [hazard ratio (HR) = 0.76; 95% confidence interval (95% CI) = 0.40–1.46] [3]. In the second, 10814 patients prescribed varenicline were matched to patients prescribed NRT. This study also showed that patients prescribed varenicline were no more likely to be hospitalized for psychiatric problems (HR = 1.14; 95% CI = 0.56–2.34) [6]. Observational studies, which can suffer from residual confounding, could not prove that varenicline did not cause neuropsychiatric adverse events. Nevertheless, they provided a more robust form of evidence than the anecdotal and case report evidence used in the original safety review. The FDA ‘determined that the current warnings in the Chantix drug label, based on post marketing surveillance reports, remain appropriate’ and did not remove the Black Box warning [7]. Meanwhile, further evidence was emerging relating to the safety of varenicline. Gunnell and colleagues found no evidence of an increased risk of suicidal behaviour in a cohort of 80660 patients from the UK’s General Practice Research Database [8]. In 2013 we updated this analysis to include data from 119546 patients, using conventional and novel statistical methods to account for residual confounding [9]. The findings were the same: there was no evidence that patients prescribed varenicline had an increased risk of fatal or non-fatal self-harm. Other observational studies have reported similar findings [10]. In 2015 we published a meta-analysis of results from all available randomized trials, which found that smokers randomized to receive varenicline had similar risks of suicide or attempted suicide to those randomized to receive placebo (OR = 1.67, 95% CI = 0.33–8.57) [11]. In March 2015 the FDA issued a further Safety Communication, but did not remove the Black Box warning [12]. In 2016 the results of the EAGLES trial was published. This trial randomized 8144 smokers to receive varenicline, transdermal NRT patch, bupropion or placebo, and found that for every 100 smokers allocated to varenicline compared with NRT there were 1.07 (95% CI = –0.08 to 2.21) fewer moderate and severe neuropsychiatric adverse events [13].While it is possible for trials to suffer from bias, this provides the strongest evidence to date that varenicline does not cause neuropsychiatric adverse events [14]. Based on the results of this study, the European Medicines Agency (EMA)— Europe’s main drug regulator—lifted the warning about possible suicidal risks from varenicline in April 2016 [15].