Summary of results
In this network meta-analysis, 15 studies covering 1565 patients were included. We evaluated the effectiveness and safety of some drugs (varenicline, naltrexone, baclofen, topiramate and bupropion) in quitting smoking for drinking smokers. The results of network meta-analysis showed that in all 15 pairwise comparisons (including direct and indirect comparison), only Varenicline showed smoking cessation effect compared with Placebo, Naltrexone, and Bupropion, no significant statistical difference was shown in the comparisons between other drugs or placebo. Meanwhile, the results of traditional meta-analysis showed that compared with Placebo, only Varenicline had obvious superiority in quitting smoking. Based on the above analysis, the value of Varenicline for smoking cessation is worth exploring. Judging from the current clinical evidence, many systematic reviews have explored the effect of smoking cessation on Varenicline. However, for the definition of the population, some reviews did not clearly divide it, or focused on other types of populations (such as cardiovascular smokers, schizophrenia smokers, etc.). For example, an updated meta-analysis performed in 2015 by Kishi, T et al. explored the effects of varenicline adjuvant therapy for smoking cessation in people with schizophrenia, 36 the results suggested that although varenicline adjuvant therapy was well tolerated, varenicline was not superior to placebo in smoking cessation. Another systematic review published by Wu, Q et al. determined the effectiveness and safety of varenicline in treating tobacco dependence in patients with severe mental illness, the author pointed out that Varenicline appeared to be significantly more effective than placebo in assisting with smoking cessation. 37 From the conclusions of the above two reviews, the characteristics of the population seem to play an important role in the effect of smoking cessation. Therefore, for drinking smokers, we still need to interpret the existing conclusions carefully.
As for the methodological quality of the included studies, although there are no ”high risk of bias ” studies, the risk of bias results of more than half of the studies are ”unclear risk of bias ”, so the methodological quality of these randomized controlled trials is low. Due to the random sequence generation, allocation concealment, and blinding for all people are important design steps of RCT, however, the description information of these three items is unclear, eventually an unclear risk bias is generated. There is no doubt that potential risk bias will reduce the quality of evidence. At the same time, we should consider the impact of other factors on the quality of evidence. For inconsistency, due to no closed loop was formed in the network diagram, thus we did not perform inconsistency test. However, precision, heterogeneity, and publication bias may reduce the level of evidence for some interventions. For example, in network meta-analysis, Varenicline showed statistical differences compared with Placebo, Naltrexone, and Bupropion, but they all had a wide confidence interval (imprecision). Moreover, in traditional meta-analysis, we found high statistical heterogeneity in Topiramate vs Placebo (I 2=69%). Due to few studies, we did not explore publication bias, but we cannot completely ignore the impact of this factor on the level of evidence.