Smoking in pregnancy remains an international public health problem and interventions to combat this are required. Myung et al.'s[1] meta-analysis intends to investigate the efficacy and safety of pharmacotherapies for smoking cessation in pregnancy and includes most of the recent, relevant studies. However, we previously published a similar review[2] with more cautious findings and wish to clarify the reasons for any discrepancies.

Myung et al. include two studies that are not in our meta-analysis; one investigated bupropion and the other involved nicotine replacement therapy (NRT). Their review combines findings from both bupropion and NRT studies, providing a pooled estimate for the efficacy of both interventions. Consequently, the review is unlikely to help to guide treatment decisions about whether or not either drug actually works for smoking cessation in pregnancy. The NRT study which Myung et al. include was quasi-randomised and women in one group were offered, but not required to use, NRT. However, all women offered NRT received behavioural support of substantially greater intensity than those who were not.[3] As behavioural support helps pregnant women to stop smoking,[4] this study does not inform us about the efficacy of NRT but instead tells us about the effectiveness of a behavioural support package to which NRT was sometimes added and by including it, Myung et al., have probably overestimated the efficacy of ‘pharmacotherapy’. The authors do, though, report a subgroup analysis without these two problematic studies, which finds a more conservative effect for NRT only, similar to that which we found.[2]

From a different subgroup analysis, Myung et al. show, as we did, that placebo-controlled NRT trials provide no conclusive evidence for the efficacy of this therapy and, what evidence there is, comes principally from studies without placebos. We believe that the biases inherent in designs that do not include a placebo help to explain positive findings from some individual studies. As the evidence for efficacy of NRT in Myung et al.'s review derives mainly from studies without placebos, findings should be considered with caution. Our review, published in early 2011, excluded both randomised controlled trials with unequal levels of behavioural support in active and control arms and non-randomised controlled trials; this explains why the pooled estimate derived from our analysis is more conservative, indicating that there is insufficient evidence to believe that NRT is efficacious; (pooled risk ratio RR and 95% confidence interval for smoking cessation in later pregnancy after using NRT, 1.63, 0.85–3.14). Recently we have updated our earlier work as a Cochrane Systematic Review; this now includes findings from a placebo-controlled trial of NRT that recruited 1050 pregnant women.[5] This review has similar findings to our earlier work and will appear in the Cochrane Library in September. Importantly, like Myung et al.,[1] after including all randomised controlled trials published by March 2012, we also concluded that there was no evidence that fetal harm has been caused by previously tested doses of NRT. We agree with Myung et al. that further randomised controlled trials investigating the efficacy of NRT are now required, but these should use higher doses of NRT than have previously been trialled.


  1. Top of page
  2. References