Smoking cessation at 1–year follow-up is the preferred outcome measure. However, with the exception of the varenicline systematic review (in which end-of-treatment, 6-month, and 1-year outcome measures were calculated), the meta-analyses included in this review calculated outcomes based on pooled data from studies with variable follow-up periods of 6 months or longer. Additionally, the number needed to treat (NNT) is the preferred comparison measure. Owing to validity issues related to NNTs calculated from meta-analyses, only the NNT for varenicline was calculated; this calculation was based on an individual trial (8,10,11
The NNT from pooled results can be misleading owing to variation between trial event rates; differences in outcomes, clinical settings, and baseline risks; and imbalances in the number of patients between arms of individual trials.35,36
Several limitations of the preliminary research on varenicline are noteworthy. The generalizability of the trials might be limited given the extensive exclusion criteria.37
With increased exclusion criteria, the probability that treatment groups represent the greater population decreases.38
For example, positive associations exist between smoking and alcohol abuse, mental distress such as depression and anxiety, and decreased physical activity; however, alcohol abuse or dependence and treatment for depression during the previous 12 months were exclusion criteria.39
Minimal justification for selecting these specific exclusion criteria is provided.
The study protocol represents an ideal intervention, which might lend itself to an increased treatment effect. With the exception of 1 trial, all participants were provided with brief counseling (= 10 minutes) during treatment and follow-up. For example, in 2 of the trials each patient received a cessation self-help booklet at the baseline visit and a telephone call 3 days after the selected quit date. During the 12-week treatment phase, participants received brief counseling during weekly clinic visits. In the follow-up periods of the trials (13–52 weeks), patients attended clinics during weeks 13, 24, 36, 44, and 52, and received telephone calls at weeks 16, 20, 28, 32, 40, and 48.11,21
The average completion rates across the 4 studies included in the meta-analysis were 62.2% for patients treated with varenicline, 57.9% for those treated with bupropion, and 49.7% for those receiving placebo. Losses to follow-up were substantially different among groups in 3 of the studies. Losses to follow-up and study withdrawal can result in a threat to the power needed to detect a treatment effect and can contribute to attrition bias, whereby characteristics of the baseline participants differ from those who were lost to follow-up or withdrew.40
If a difference in the number of people leaving each arm exists, it is likely that the groups are not balanced.41
Intention-to-treat analysis, which is preferable in randomized controlled trials, was used in the studies. All randomized participants were included in the efficacy analysis regardless of whether they received treatment or complied with study protocol.42
However, given that individuals who dropped out of the study or were lost to follow-up were deemed continuing smokers and that intention-to-treat analysis was used, the efficacy of bupropion and placebo might have been misrepresented relative to varenicline owing to the higher completion rates in the varenicline groups.
Studies to date have been funded by varenicline’s manufacturer, Pfizer Inc, and the company was involved in all elements of each study. Independently funded and administrated studies are needed to more confidently determine the efficacy and tolerability of varenicline. Head-to-head studies are also needed to help determine the relative efficacy of treatments. It is unclear whether the cost of varenicline, which is approximately $150 monthly, will have an effect on its use.