To the Editor:
In their POEM review of the data of Jorenby and colleagues1 Drs Kane and Ellis2 prematurely conclude “it is clear…that using bupropion…achieves higher smoking cessation rates than using the [nicotine] patch alone.” Though the data does suggest that smoking cessation rates were higher in those patients randomized to bupropion than to a nicotine patch, that study is sufficiently flawed to render such a conclusion speculative.
When previously studied3 in more than 26,000 subjects, nicotine replacement therapy was shown to increase smoking cessation rates by 66% (95% confidence interval [CI], 58% - 74%). Yet, Jorenby and coworkers found that a nicotine patch worked no better than placebo. Something seems amiss.
Dropouts in a study are inevitable, but I am troubled that in a study requiring subjects be followed for only 6 to 12 months, 31% were lost to follow-up. Even more troubling is that the loss to follow-up was significantly different among the treatment groups. In the bupropion plus patch group, “only” 20% were lost to follow-up before the 6-month evaluation. This compares with 27% in the bupropion-alone group, 35% in the patch-alone group, and a whopping 46% in the placebo group. This means that 41% more patients randomized to nicotine alone were lost to follow-up than those randomized to bupropion plus patch (95% CI, 19% - 64%) . Fifty-six percent more placebo patients than bupropion plus patch patients were lost to follow-up (95% CI, 36% - 76%). These statistically significant differences reveal a bias or systematic error in Jorenby and colleagues’ study. I submit that the large loss to follow-up overall, and the systematically larger loss to follow-up in those not receiving bupropion, renders the conclusion that bupropion results in higher smoking cessation rates than nicotine replacement hypothetical. I hope that future trials with better, more consistent follow-up will clarify whether bupropion offers a real advantage in comparison with nicotine replacement.
Family physicians should understand that both forms of therapy increase smoking cessation rates; combining nicotine replacement with bupropion does not increase effectiveness; and it is unclear whether bupropion increases smoking cessation rates more than nicotine replacement.
Brian Budenholzer, MD
Group Health Northwest
- Jorenby DE, Leischow SJ, Nides MA, et al. A controlled trial of sustained-release bupropion, a nicotine patch, or both for smoking cessation. N Engl J Med 1999; 340:685-91.
- Kane KY, Ellis MR. Bupropion or patch for smoking cessation? J Fam Pract 1999; 48:419.
- Silagy C, Mant D, Fowler G, Lancaster T. Nicotine replacement therapy for smoking cessation. Cochrane library Oxford, England: Update Software, 1999:2.
The preceding letter was referred to Drs Kane and Ellis, who reply as follows:
We appreciate Dr Budenholzer’s comments regarding the attrition rate. The appropriate way to account for subjects lost to follow-up is through an intention-to-treat analysis in which the subjects are analyzed in the original groups to which they were assigned. Although it is true that 311 subjects (34.8%) discontinued treatment, 134 continued to participate in follow-up assessments. Jorenby and colleagues1 performed an intention-to-treat analysis and appropriately classified the 19.8% of subjects lost to follow-up as smokers, thus avoiding any inflation of the treatment efficacy. We agree that it is unlikely that those patients would have remained abstinent at 12 months. The disproportionate increased rate of attrition in the patch and placebo groups may reflect a relative failure of those treatments to produce sustained abstinence.
We agree with Dr Budenholzer that the recent Cochrane review found a 66% increase in smoking cessation with all forms of nicotine replacement therapy. However, subgroup analysis showed 12-month quit rates of 13% with the patch and 11% with placebo. These are similar to the 12-month quit rates of 16.4% with the patch and 15.6% with placebo found in the study by Jorenby and coworkers.
The lack of significant difference between patch and placebo may be explained by several factors. For instance, the placebo group may have a better than expected quit rate because of the intensive counseling sessions, the result of using a double placebo as mentioned in the article by Jorenby and colleagues, or random variation in a relatively small sample size. The sample size in the study we reviewed was determined by detecting a 14% difference in cessation rates after 4 weeks of treatment with bupropion compared with placebo. We suspect the lack of significance for the nicotine group was due to inadequate sample size to detect a difference of this magnitude.
This study shows that bupropion alone or in combination with the nicotine patch produces higher smoking cessation rates than either the nicotine patch alone or placebo, with some limits on generalizability.3 Those lost to follow-up were appropriately accounted for in an intention-to-treat analysis and correctly classified as smokers. Quit rates for the patch compared with placebo in this study are similar to those found in the Cochrane meta-analysis.