3 The manufacturer's submission
3.1 The manufacturer's primary analysis compared the standard 12-week course of varenicline with bupropion and nicotine replacement therapy (NRT). The manufacturer identified four randomised controlled trials (RCTs). Two were three-arm trials that compared varenicline, bupropion and placebo (n = 1483 and 1413). Another trial compared maintenance treatment (24-week course of varenicline) with placebo (n = 2416). The manufacturer also presented data from an open-label trial (n = 957) that compared varenicline with NRT.
3.2 The two trials that compared varenicline and bupropion showed that the continuous quit rate for weeks 9–12 was statistically significantly greater for varenicline: odds ratio (OR) 1.93 (95% CI 1.40 to 2.68) and OR 1.90 (95% CI 1.40 to 2.68), respectively. Both trials also showed that the continuous quit rate for varenicline was statistically significantly greater than for placebo: OR 3.85 (95% CI 2.70 to 5.50) and OR 3.85 (95% CI 2.69 to 5.50), respectively. For the longer time horizon of weeks 9–52, the ORs for varenicline compared with bupropion, respectively for the two trials, were 1.46 (95% CI 0.99 to 2.17) and 1.77 (95% CI 1.19 to 2.63). The maintenance trial that compared 24-week varenicline with placebo showed that the continuous quit rate was statistically significantly greater with varenicline than with placebo: weeks 13–24 OR 2.47 (95% CI 1.95 to 3.15); weeks 13–52 OR 1.35 (95% CI 1.07 to 1.70). The results of the open-label trial were marked confidential by the manufacturer.
3.3 The manufacturer also submitted a meta-analysis of 70 NRT trials, 12 bupropion trials and 4 varenicline trials against control/placebo. The meta-analysis indirectly compared the efficacy of the treatments based on relative treatment effects. This indirect comparison showed that at 12 months varenicline was superior to NRT (OR 1.66 [CI 1.17 to 2.36]) and bupropion (OR 1.58 [95% CI 1.22 – 2.05]). Varenicline was also superior at 3 months to both NRT (OR 1.78 [95% CI 1.23 to 2.57]) and bupropion (OR 1.61 [95% CI 1.17 to 2.22]).
3.4 The manufacturer presented a cost-effectiveness analysis based on a Markov model. It assumes an individual makes a single quit attempt at the beginning of the model. The individual is followed from this initial quit attempt to various health states and potential comorbidities including lung cancer, asthma exacerbations, chronic obstructive pulmonary disease, stroke and cardiovascular disease. The probabilities of relapsing and developing comorbidities are assumed to decrease over time from smoking cessation. The efficacy rates for the treatments are calculated from the odds ratios derived from the results of the pooled direct clinical trials and the indirect comparison. The probabilities associated with relapse are derived from relative risks reported in US-based long-term longitudinal and cohort studies into smoking and abstinence. The costs and utilities are derived from several published sources. Some health-related utility estimates are based on US data, including baseline health-related utilities.
3.5 The base-case analysis showed that over a lifetime horizon varenicline dominated bupropion and NRT – that is, it was cheaper and more effective. Variation of the time horizon used in the analysis showed that, at 20 years and over, varenicline maintained its dominating position. Sensitivity analyses included altering baseline health-related utilities and costs of NRT, and the use of efficacy rates from the direct open-label trial that compared varenicline with NRT. Over a lifetime horizon varenicline dominated NRT and bupropion in all sensitivity analysis.
3.6 The ERG noted that the inclusion/exclusion criteria of the meta-analysis in the manufacturer's submission differed from existing analyses by the Cochrane collaboration and considered that they could overestimate the efficacy of varenicline. The ERG also conducted its own meta-analysis and indirect comparison which suggested that the odds ratio for varenicline in comparison with NRT was lower than in the manufacturer's model: OR 1.54 (95% CI 1.10 to 2.16).
3.7 The ERG noted that the assumptions included in the model could make external validity questionable. For example, it considered that the assumption of a single quit attempt was a limitation that did not allow consideration of the impact of subsequent quit attempts on costs, morbidity or mortality. In addition, the extrapolation of data on 1-year quit rates to a lifetime is associated with considerable uncertainty surrounding the long-term relapse or abstinence experience of the model cohort. The ERG noted that the use of indirect comparison in the base case was inappropriate given the availability of direct trial evidence. The ERG further identified some computational errors in the calculation of transition probabilities and population calculations. The ERG commented that the method used to convert odds ratios to efficacy rates was not validated and a model constructed around odds ratios would have been more appropriate. The ERG compared the number of life years gained in the model with the results of two published analyses and found no substantial differences.
3.8 Full details of all the evidence are in the manufacturer's submission and the ERG report.