Dr S-K Myung, 323 Ilsan-ro, Ilsandong-gu, Goyang, Gyeonggi-do 410-769, South Korea. Email firstname.lastname@example.org
Please cite this paper as: Myung S, Ju W, Jung H, Park C, Oh S, Seo H, Kim H, for the Korean Meta-Analysis (KORMA) Study Group. Efficacy and safety of pharmacotherapy for smoking cessation among pregnant smokers: a meta-analysis. BJOG 2012;119:1029–1039.
Background The efficacy and safety of pharmacotherapy for smoking cessation among pregnant smokers has not yet been established.
Objective To investigate the efficacy and safety of pharmacotherapy for smoking cessation among pregnant smokers.
Search strategy A search was made of PubMed, Embase and CENTRAL in June 2011.
Selection criteria Randomised controlled trials (RCTs), quasi-RCTs and retrospective or prospective controlled studies were included.
Data collection and analysis The main analyses were designed to examine the efficacy of pharmacotherapy for smoking cessation among pregnant smokers based on the longest follow-up data available and from data obtained at the latest available time-point in pregnancy in each study.
Main results Of 74 articles identified from the databases, seven studies (five RCTs, one quasi-RCT and one prospective study) involving a total of 1386 pregnant smokers, 732 in the intervention groups and 654 in the control groups, were included in the final analyses. In a fixed-effects meta-analysis of all seven studies based on the longest follow-up data available, pharmacotherapy had a significant effect on smoking cessation (relative risk [RR] 1.80; 95% confidence interval [CI] 1.32–2.44). Subgroup meta-analysis by type of study design also showed similar findings for RCTs (RR 1.48; 95% CI 1.04–2.09) and other types of studies (RR 3.25; 95% CI 1.65–6.39). The abstinence rate at late pregnancy in the intervention ranged from 7 to 22.6% (mean abstinence rate 13.0%; 95% CI 10.9–15.2%). A few minor adverse effects and serious adverse effects were reported in several studies.
Author’s conclusions This study indicates that there may be clinical evidence to support the use of pharmacotherapy for smoking cessation among pregnant smokers. Further RCTs are needed.
Smoking during pregnancy is one of the most important potentially modifiable causes of adverse pregnancy outcomes.1–3 It has been reported that about 45% of pregnant smokers ‘spontaneously quit’ before pregnancy or stop before their first antenatal visit to a clinic.4,5 However, two-thirds of women who stop smoking during pregnancy resume smoking 1 year after delivery.6 Data obtained in the general population have demonstrated the effectiveness of numerous medications available for tobacco dependence, such as nicotine replacement therapy (NRT), bupropion SR and varenicline.7 For example, in a meta-analysis of 83 studies, when compared with placebo, estimated odds ratios of pharmacotherapy for smoking cessation at 6 months were 1.9–2.3 for NRT, 2.0 for bupropion SR and 3.1 for varenicline, all of which were statistically significant.7
Pregnancy causes various physiological changes and can lead to important variations in the pharmacokinetic processes of absorption, distribution and elimination of drugs.8 The efficacy and safety of pharmacotherapy for smoking cessation has been established in the general population, but not in pregnant smokers. Although several studies9–15 have been published on this subject, their findings have been inconsistent. In 2009, a meta-analysis16 reported the effects of various interventions, such as counselling, hypnosis and pharmacotherapy, for smoking cessation during pregnancy and indicated that NRT was as effective as cognitive behavioural therapy. However, in that meta-analysis, detailed analyses were not performed to examine the efficacy of pharmacotherapy, its potential adverse side effects and its effects on birth outcomes.
The purpose of this meta-analysis was to examine the efficacy and safety of pharmacotherapy for smoking cessation among pregnant smokers by conducting a meta-analysis of studies with subgroup meta-analyses by various factors such as type of study design, validation, type of pharmacotherapy, follow-up period, methodological quality and use of placebo.
PubMed, EMBASE and the Cochrane Central Register of Controlled Trials (CENTRAL) to 19 June 2011 were searched using common keywords related to pharmacotherapy for smoking cessation among pregnant smokers. The bibliographies of relevant articles were also searched to identify additional studies. The keywords were as follows: pharmacological treatment or NRT or nicotine patch or nicotine gum or nicotine inhaler or nicotine spray or bupropion or varenicline; and pregnant smokers. We included randomised controlled trials (RCTs), quasi-RCTs, and retrospective or prospective controlled studies that reported the efficacy and safety of pharmacotherapy for smoking cessation among pregnant smokers. The principal outcome measures included point-prevalence abstinence and continuous abstinence self-reported or validated with biochemical markers such as exhaled carbon monoxide, and salivary or urinary nicotine. We did not restrict the language of publication, and excluded unpublished articles.
Two authors (MSK and JW) independently assessed the eligibility of all studies retrieved from the databases and bibliographies. Disagreements regarding study eligibility were resolved by discussion. From the studies included in the final analysis we extracted the following data: study name (along with the name of the first author and the year of publication), country, study design, recruitment period (in years), number of participants by condition, the details of the intervention and control conditions, outcome measures, abstinence rates, relative risk (RR) with 95% confidence interval (95% CI) calculated from the numbers of the four cells of the 2 × 2 tables in each of the studies, the number of women who stopped smoking compared with the number of participants, and the percentage lost to follow up in both the intervention and control groups.
The methodological quality of the studies was assessed based on a validated scale for RCTs developed by Jadad et al.17 This five-point quality scale uses points for randomisation (i.e. randomised, 1 point; table of random numbers or computer-generated randomisation, additional 1 point), double-blinding (i.e. double-blind, 1 point; using identical placebo, additional 1 point), and follow up (i.e. stating numbers and reasons for withdrawal in each group, 1 point).16 We considered scores of 2 or less to indicate ‘low quality’ and scores of 3–5 to indicate ‘high quality’ studies.
The main analyses were designed to examine the efficacy of pharmacotherapy for smoking cessation among pregnant smokers based on data available for the longest follow-up period and at the highest gestation, in each study. Another main aim of the study was to perform a meta-analysis by type of study design. We also performed subgroup meta-analyses by type of validation (self-report versus validation using carbon monoxide or cotinine), type of pharmacotherapy (nicotine patch versus nicotine gum versus bupropion), follow-up period (short [<12 weeks] versus intermediate [12–24 weeks] versus long [>24 weeks]), methodological quality (high quality [score ≥3] versus low quality [score <3]), and use of placebo (placebo RCTs versus non-placebo studies).
Statistical analyses were conducted based on 2 × 2 tables consisting of the number of participants and number who gave up smoking in each intervention and control group in each study. If there was no event (i.e. if there were no women who gave up smoking) in one of the intervention or control groups (i.e. a ‘zero cell’ in the 2 × 2 table),9,12 0.5 was added to each cell of the table to avoid the estimated RR being zero or infinity; in this way the standard error could be calculated. For the test for heterogeneity across studies, we used Higgins’I2, which measures the percentage of total variation across studies. I2 was calculated as follows:
where Q is Cochran’s heterogeneity statistic, and df is the degrees of freedom. Negative values of I2 are set to zero so that I2 lies between 0% (no observed heterogeneity) and 100% (maximal heterogeneity). We considered an I2 value of >50% to represent substantial heterogeneity.
The pooled RR with 95% CI was calculated on the basis of both the fixed-effects and random-effects models. When there was no substantial heterogeneity, the pooled RR with 95% CI on the basis of the fixed-effects model is reported because the summary estimates from the fixed-effects and random-effects models are similar. When there was substantial heterogeneity, the summary estimates from the random-effects model are presented because CIs tend to be larger in the random-effects model than in the fixed-effects model. The Mantel–Haenszel method was used in the fixed-effects model, and the DerSimonian and Laird method was used in the random-effects model.
We evaluated publication bias using a contour-enhanced funnel plot of each study’s effect size and standard error.18 Funnel plot asymmetry was assessed using Begg and Egger tests.19,20 If the funnel plot was asymmetrical or the P-value was found to be <0.10 by Egger’s test, then publication bias was considered to exist.
Also, mean abstinence rates were calculated by computing a sample size-weighted summary. Stata SE version 10.0 software (StataCorp, College Station, TX, USA) was used for statistical analysis.
Figure 1 shows a flow diagram for the identification of relevant studies. A total of 74 articles were identified after searching the three databases (PubMed, EMBASE and CENTRAL), and hand-searching relevant bibliographies. After excluding 23 duplicated articles and 41 articles that did not satisfy the selection criteria mentioned in the Methods section, the full texts of ten articles were reviewed by two of the authors. Of those, three articles21–23 were excluded for the following reasons: not a comparison study (n = 1);21 a study not related to pharmacotherapy (n = 1);22 a study sharing an identical population with the study included in this analysis (n = 1).23 Seven studies9–15 were included in the final analysis.
The seven studies included a total of 1386 pregnant smokers, with 732 randomised to pharmacological intervention groups and 654 to placebo or control groups. In the studies that reported age and cigarettes smoked per day, the mean age of the intervention and control groups ranged from 25.5 to 33.4 years and from 24.7 to 31.4 years, respectively; the number of cigarettes smoked per day in the intervention and control groups ranged from 13.4 to 25.4 and from 14 to 24.5, respectively.
One prospective controlled study and one quasi-RCT showed that the use of either bupropion or a multimodal intervention regimen with counselling supplemented by NRT as a voluntary option had a significant effect on smoking cessation among pregnant smokers.11,12 The remaining five RCTs reported that there was no significant effect of NRT, including nicotine patch, gum and lozenge.9,10,13–15
Table 1 shows the general characteristics of the seven trials (five RCTs, one quasi-RCT and one prospective controlled observational study) included in the final analysis. The included trials were published from 2000 to 2008, spanning 8 years. The countries in which the studies were carried out were as follows: USA (n = 2), Denmark (n = 2), Canada (n = 2) and Australia (n = 1). The follow-up period ranged from 12 weeks to about 26 weeks; the overall period of recruitment was from 1995 to 2006. The number of participants in each study ranged from 30 to 250. Of the seven studies, five used a nicotine patch or gum, one used nicotine gum, and one used bupropion. The doses of pharmacotherapy used were generally 15 mg nicotine patch/day (a choice of 7, 14 or 21 mg nicotine patch in Hotham et al.’s study12), 2 mg nicotine gum or lozenge for every cigarette usually smoked per day, and 150 mg bupropion (150–300 mg/day). The treatment period ranged from 6 to 12 weeks. The abstinence rate at late pregnancy in the intervention and control groups ranged from 7 to 22.6% (mean abstinence rate 13.0%; 95% CI 10.9–15.2%; Cochrane’s Q = 0.062; n = 7; random-effects model) and from 0 to 19.8% (mean abstinence rate 8.2%; 95% CI 6.0–10.3%; Cochrane’s Q = 0.011; n = 7; random-effects model), respectively.
Table 1. Characteristics of studies included in the final analysis (n = 7)
Study name (Ref. no.)
Intervention versus control
RR (95% CI)*
Abstinence rate (%)**
No. of loss to follow-up (%)
Abstinence rate (%)**
No. of loss to follow-up (%)
NA, not available.
*A relative risk with 95% confidence interval was calculated from the 2 × 2 table data consisting of No. of women who stopped smoking / No. of participants in each intervention and control group of an individual study. If a group had no quitter, then ‘0.5’ was added to the value in each cell.
**No. of quitters/ No. of participants in each group.
194 pregnant women who were currently smoking at least one cigarette per day
2 mg nicotine gum for 6 weeks followed by a 6-week taper period versus placebo gum
Seven-day point prevalence abstinence rate validated with an exhaled carbon monoxide of less than 8 ppm at 32–34 weeks of gestation
Overall effect in the seven studies
In a fixed-effects meta-analysis of all seven studies, based on the longest follow-up data available in each study, pharmacotherapy had a significant effect on smoking cessation (RR 1.80; 95% CI 1.32–2.44; I2 = 41.5%) (Figure 2). Also, in subgroup meta-analyses by type of study design, a significant effect on smoking cessation was found in both RCTs (RR 1.48; 95% CI 1.04–2.09; I2 = 20.9%; n = 5) and other studies (RR 3.25; 95% CI 1.65–6.39; I2 = 0.0%; n = 2). In the five selected studies, Begg’s funnel plot was asymmetrical, and Egger’s test showed that P for bias was 0.036 (Figure 3) (two10,13 of the seven studies were excluded because of ‘zero’ cells in the 2 × 2 table). A contour-enhanced funnel plot indicated that the ‘missing’ studies were in areas of statistical significance (in darker shaded areas). This suggests that the observed asymmetry is more likely to be attributable to factors other than publication bias, such as variable study quality.18
Methodological quality of studies
Of the seven studies, five received a score of 3 or more, while two studies11,12 received a score of ‘zero’ (Table 2). Those two studies were a quasi-RCT and a prospective controlled observational study.
Table 2. Methodological quality of studies included in the final analysis based on the Jadad scale (n = 7)
Table 3 shows the effects of pharmacotherapy for smoking cessation in pregnant smokers in subgroup meta-analyses by various factors. When compared with control groups, the summary RR for groups receiving pharmacotherapy at late pregnancy for smoking cessation was 1.65 (95% CI 1.20–2.28; I2 = 41.5%; n = 5). Subgroup meta-analyses of the studies reporting abstinence validated with salivary cotinine or exhaled carbon monoxide showed a significant effect of pharmacotherapy, while those studies with self-reported abstinence showed no significant effect.
Table 3. Effects of pharmacotherapy for smoking cessation in pregnant smokers in subgroup meta-analyses
No. of studies
Summary RR (95% CI)
Heterogeneity, I2 (%)
NA, not applicable.
Longest follow-up data in each trial
Type of study design
Non-RCTs (quasi-randomised or prospective studies)
Abstinence validated with salivary cotinine or exhaled carbon monoxide
Type of pharmacotherapy
Short-term (<12 weeks)
Midterm (12–24 weeks)
Long-term (>24 weeks)
High quality (score ≥3)
Low quality (score <3)
Use of placebo
Regarding the type of pharmacotherapy, a meta-analysis of the four studies in which a nicotine patch was used showed a significant effect for smoking cessation (RR 1.60; 95% CI 1.05–2.43; I2 = 39.2%). One study using bupropion reported an RR of 1.21 (95% CI 1.06–10.49) for smoking cessation.
Subgroup meta-analyses for short-term (<12 weeks) and intermediate-term (12–24 weeks) follow up showed an increased abstinence rate in the pharmacotherapy groups, whereas the meta-analysis for long-term follow up (>24 weeks) did not show a difference between the pharmacotherapy and control groups. The significant effect of pharmacotherapy for smoking cessation was found regardless of methodological quality. In subgroup meta-analyses by use of placebo, no significant effect of pharmacotherapy was found in the three placebo RCTs, whereas a significant effect was found in the four non-placebo studies (RR 3.17; 95% CI 1.83–5.50).
Adverse effects of pharmacotherapy and main birth outcomes
Table 4 summarises adverse effects of pharmacotherapy and main birth outcomes. Adverse effects were reported in three9,13,15 of the seven studies: skin irritation, headache, palpitation, nausea, an arm feeling ‘dead’, worsening of morning sickness symptoms and exacerbation of postnatal depression among the nicotine patch and placebo users; headache, dizziness, fatigue, heartburn, nausea and vomiting among the nicotine gum and placebo users. The remaining studies did not mention any adverse effects.
Table 4. Adverse effects of pharmacotherapy and main birth outcomes
SAE rates (P = 0.07): a) CBT plus NRT group – 34/113 (30%) b) CBT-only – 10/58 (17%) *An imbalance between the arms was observed in the proportion of women with a previous preterm birth (12% in CBT-only versus 32% CBT plus NRT)
Headache, dizziness, fatigue, heartburn, nausea, and vomiting were reported by at least 10% of participants. The incidence of nausea increased more from baseline to treatment in the nicotine gum group than in the placebo group (P = 0.019)
SAE rates (P = 0.06) a) Nicotine gum group: 24/97 (24.7%) b) Placebo group: 33/87 (37.9%) *No significant differences in baseline characteristics including preterm delivery were reported between two groups
Low birthweight (<2500 g)
P < 0.001
Preterm delivery rate (<37 weeks)
P = 0.027
Other adverse effects were reported in three studies.10,14,15 Kapur et al.’s study10 reported the use of a placebo patch in one participant. After a night’s sleep, the woman stopped smoking, and 3 hours later she experienced severe signs of withdrawal and at the same time the fetus exhibited rapid, forceful movements. The woman resumed smoking and the fetal movements subsided within 20 minutes. In Pollak et al.’s study,14 in which a nicotine patch, gum or lozenge was used, serious adverse effects were reported, such as preterm birth (<37 weeks), neonatal intensive care unit (NICU) admission (the newborn being small for gestational age), placental abruption and fetal demise. Rates of serious adverse effects were 30% in the cognitive behavioural therapy plus NRT group and 17% in the cognitive behavioural therapy only group (P = 0.07). However, an imbalance between the arms was observed in the proportion of women with a previous preterm birth (12% in the cognitive behavioural therapy only group versus 32% in the cognitive behavioural therapy plus NRT group). In Oncken et al.’s study,14 in which nicotine gum was used, maternal hospitalisation, low birthweight (<2500 g), preterm delivery (<37 weeks), spontaneous abortion, intrauterine fetal demise, second-trimester pregnancy loss, newborn death and NICU admission were reported as serious adverse effects. There was no significant difference in the rate of serious adverse effects or in baseline characteristics between the nicotine gum and placebo groups (P = 0.06).
Five studies9,11,12,14,15 reported mean birthweight, low birthweight rate, preterm delivery rate and/or mean gestational age. Two studies9,15 reported that the pharmacotherapy group had a higher mean birthweight than the control group, while three studies11,12,14 reported that there was no significant difference in mean birthweight between the two comparison groups. No significant difference in low birthweight rate, mean gestational age or preterm delivery rate was observed between the two comparison groups in the three studies.9,11,14 Only one study,15 in which nicotine gum was used, showed that low birthweight and preterm delivery rates were lower in the pharmacotherapy group, who used nicotine gum, than in the control group.
The current meta-analysis of seven studies, including five RCTs and two others, showed that pharmacotherapy for smoking cessation among pregnant smokers yielded a significantly greater abstinence rate, about 1.8 times higher, than in the control group. When data were pooled, the mean abstinence rate at late pregnancy in the intervention and control groups was 13.6% (95% CI 11.0–16.1%) and 8.1% (95% CI 5.9–10.2%), respectively. Also, although a few minor adverse effects and serious adverse effects were reported in several studies, there was no evidence directly linking those serious adverse effects to the pharmacotherapy used in individual studies, and overall pharmacotherapy did not affect birth outcomes and was generally safe.
Our findings indicate that there may be clinical evidence to support the use of pharmacotherapy for smoking cessation among pregnant smokers in terms of its efficacy and safety. Pharmacotherapy for smoking cessation among pregnant smokers, to date, is not recommended because its efficacy and safety have not yet been established. For example, in the 2008 update to Treating Tobacco Use and Dependence, a Public Health Service-sponsored Clinical Practice Guideline of the US Department of Health and Human Services, the Panel did not make a recommendation regarding pharmacotherapy during pregnancy because at the time there was insufficient evidence to support its use.7 In the general population, two meta-analyses7,24 showed that the abstinence rate was 1.9–3.1 times higher in the pharmacotherapy group than in the placebo groups at 6 months7 and 1.8 times higher at 12 months.24 Our study found that pharmacotherapy (mainly NRT) had a similar effect on smoking cessation in pregnant smokers, with an RR of 1.8. Also, the quit rate for pharmacotherapy (mainly nicotine patch) at late pregnancy (at 4–5 months of follow-up since the beginning of the study) in pregnant smokers (13.0%; 95% CI 10.9–15.2%) was lower than that of pharmacotherapy (nicotine patch) at 6–12 months in the general population (15.4%; 95% CI 15.2–15.6%) (calculated by computing a sample-sized summary based on the figure ‘Analysis 1.1.’ of Stead et al.’s meta-analysis24). This finding may indicate that it is more difficult for pregnant smokers to stop smoking than for those in the general population, but further studies are required to confirm this, because the sample size of the current study was much lower than that of Stead et al.’s meta-analysis.
As described in the Results section, some studies reported minor adverse effects related to pharmacotherapy, such as skin irritation, headache, dizziness and nausea, which are also commonly observed in the general population. In addition, although several serious adverse effects were reported, such as preterm birth, NICU admissions, the infant being small for gestational age, placental abruption, fetal demise, spontaneous abortion and neonatal death, there was no evidence directly linking these serious adverse effects to the pharmacotherapy used in the individual studies. Overall, there was no significant difference in mean birthweight, low birthweight rate, mean gestational age, and preterm delivery rate between the pharmacotherapy and control groups. Of the seven studies, two reported a higher mean birthweight in the pharmacotherapy group than in the control group, and one study showed lower rates of low birthweight and preterm delivery in the pharmacotherapy group than in the control group. These findings may be attributable to a higher abstinence rate in the intervention group.
Our study has several possible limitations. First, because of the paucity of studies on this subject published to date, we were able to include only seven studies, involving 1386 pregnant smokers, in the analysis. In particular, only three of the seven studies were RCTs using identical placebos. This paucity of studies is associated with difficulties in recruiting pregnant smokers and receiving approval from institutional review boards. Second, most of the included studies included had follow-up times of at best 26 weeks. Only one study14 reported abstinence data at about 34 weeks and up to 3 months postpartum. Hence, we were unable to investigate the long-term efficacy of pharmacotherapy for smoking cessation among pregnant smokers. Third, publication bias was observed in our study. The asymmetry of the funnel plot in the current study may imply that studies that showed no significant effect or a negative effect of pharmacotherapy have not been published or have had their publication delayed. Therefore, our findings should be interpreted cautiously. Fourth, we included the only trial in which bupropion was used. Further trials are required to investigate the effects of oral medications such as bupropion or varenicline in addition to NRT. Lastly, only three RCTs using placebo as a control were included in our study because of a paucity of published data. As shown in the subgroup meta-analyses by use of placebo, no significant effect of pharmacotherapy was found in the three placebo RCTs, whereas a significant effect was found in the four non-placebo studies. This finding may attenuate the main finding of a positive effect of pharmacotherapy in pregnant smokers in the current study. However, the meta-analysis of the three placebo RCTs showed a non-significant ‘positive’ effect (RR = 1.25) and involved only 474 participants, much fewer than in the four non-placebo studies, which involved 1386 participants. Further larger, randomised, double-blind, placebo-controlled trials are required to confirm the findings of our study.
In summary, we found that there may be clinical evidence to support the use of pharmacotherapy for smoking cessation among pregnant smokers. Further, larger RCTs are needed to confirm the efficacy and safety of pharmacotherapy for these populations. Also, an economic evaluation of pharmacotherapy for smoking cessation during pregnancy should be performed based on a cost–benefit analysis of smoking cessation versus adverse effects. We believe that, although researchers have had difficulty in conducting such RCTs because of difficulties in recruiting pregnant smokers and receiving approval from ethics committees to date, our findings in the current study would form the basis for them to perform RCTs of pharmacotherapy for smoking cessation during pregnancy.
Disclosure of interest
Contribution to authorship
SKM, as principal investigator, had full access to all of the data in the study and is the guarantor for this paper, taking responsibility for the integrity of the data and the accuracy of the data analysis. SKM was responsible for the initial plan, study design and statistical analysis and for conducting the study. SKM and WJ were responsible for data collection, data extraction, data interpretation, and drafting of the manuscript. HSJ, CHP, SWO, HGS and HSK were responsible for data interpretation and drafting of the manuscript.