After receiving hundreds of postmarketing adverse event reports of suicidal thoughts or behaviors (and 32 completed suicides) associated with smoking cessation drug varenicline (Chantix), the FDA slapped a black box warning on the med in 2009 and commissioned two large retrospective observational studies totaling more than 40,000 patients starting either varenicline (Chantix) or nicotine replacement. Those studies showed no difference in psychiatric hospitalizations in the 30-day window after therapy initiation with either treatment. But those events were rare -- only 72 hospitalizations -- and the study didn't capture suicide, depression, aggression, or assaults not resulting in hospitalization.
So Moore, Furberg et al looked into Chantix's record in the FDA's Adverse Event Reporting system from 1998-2010, and found 2,925 reports of "serious self-injury or depression" associated with varenicline use. That was 90% of the total reported for all smoking cessation products during that time (the others being nicotine replacement products and wellbutrin / Zyban), making Chantix 8 times more likely to be associated with such a report.
Pfizer points out that postmarketing reports do not prove cause and effect, and suggested that physicians and patients were primed by early reports of Chantix's psychiatric effects, and the excess adverse event reporting is all due to this so-called "notoriety bias."
Moore & Furberg acknowledge that the real risks from Chantix use remain unclear (from their discussion section):
While a growing body of research from multiple sources establishes that varenicline substantially increases the risk of psychiatric side effects, it remains uncertain how frequently these events occur. Estimating incidence rates would require a large cohort, a validated psychiatric symptoms checklist, and sufficient power over one year to detect event rates that may be as low as 1 or 2 per 1,000 but could be more than 1 in 100.
Chantix works better than other smoking cessation products, but to Furberg, the 10-20% quit rate at one year (they cite 10%) isn't worth the potentially dangerous side effects. He and this group were instrumental in convincing the FAA to ban pilots from using varenicline, after reports of unexplained blackouts while driving surfaced in postmarketing surveillance. Here, authors didn't pull punches in concluding "[t]he overall safety profile of varenicline makes it unsuitable for first-line use in smoking cessation," nor in speaking out to Reuters against Chantix:
"We found that Chantix is associated with more suicidal behavior reports than any other smoking-cessation drug on the U.S. market. The risks simply outweigh the benefits," Furberg said.
Other data out there: In the 2 randomized controlled trials on varenicline we reviewed recently (Ann Intern Med, Chest), there were no suicides and no excess suicidal thoughts or behavior reported in the Chantix arms (total n of a few hundred). However, as per Moore et al's point above, low event rates could hide any signal in trials that small.
Moore TJ, Furberg CD, et al. 2011 Suicidal Behavior and Depression in Smoking Cessation Treatments. PLoS ONE6(11): e27016.