The impact of pharmaceutical company funding on results of randomized trials of nicotine replacement therapy for smoking cessation: a meta-analysis.

AIMS To assess whether source of funding affected the results of trials of nicotine replacement therapy (NRT) for smoking cessation. METHODS We reviewed all randomized controlled trials included in the Cochrane review. There were insufficient non-industry trials of the newer products for these to be included. We included 90 trials of either the nicotine gum (52) or nicotine patch (38). They comprised 18 238 treatment and 16 235 control participants. Forty-nine showed evidence of industry support (18 gum, 31 patch). RESULTS Industry (31 of 49, 63%) compared with non-industry (seven of 41, 17%, P < 0.001) supported a higher proportion of nicotine patch studies and had larger sample sizes (479 versus 268, P = 0.04). Twenty-five (51%) industry trials reported statistically significant (P < 0.05) results, compared with nine (22%) non-industry trials (OR = 3.70, 95% CI = 1.46-9.35). This difference was not explained by trial characteristics. Industry-supported trials had a pooled odds ratio of 1.90 (1.67-2.16), compared with 1.61 (1.43-1.80) for other studies (chi(2) = 3.6, P = 0.058). There was evidence of funnel-plot asymmetry among industry trials (t = 4.35, P < 0.001), but not among other trials, indicating that several small null-effect industry trials may not have reached publication. After imputation adjustment, the odds ratio for industry trials reduced to 1.64 (1.43-1.89) and the overall NRT odds ratio reduced from 1.73 (1.60-1.90) to 1.62 (1.49-1.77). CONCLUSIONS Compared with independent trials, industry-supported trials were more likely to produce statistically significant results and larger odds ratios. These differences persisted after adjustment for basic trial characteristics. Although we had no data on the amount of funding for each trial, it is possible that more resources led to higher treatment compliance and therefore greater efficacy in industry-supported trials. Differences can also possibly be explained by publication bias with several small, null-effect industry studies not having reached publication. After adjustment for this possible bias, results for industry trials were lower and similar to non-industry results. Similarly, the overall estimate of the net effect for these products reduces to about 5% attributable 1-year successes. This remains of considerable public health benefit. Registration of clinical trials has become mandatory in many countries since most of the trials considered here were conducted, and this should reduce the potential for publication bias in future.