Psychosocial interventions for supporting women to stop smoking in pregnancy

  • Review
  • Intervention

Authors


Abstract

Background

Tobacco smoking in pregnancy remains one of the few preventable factors associated with complications in pregnancy, stillbirth, low birthweight and preterm birth and has serious long-term implications for women and babies. Smoking in pregnancy is decreasing in high-income countries, but is strongly associated with poverty and increasing in low- to middle-income countries.

Objectives

To assess the effects of smoking cessation interventions during pregnancy on smoking behaviour and perinatal health outcomes.

Search methods

In this fifth update, we searched the Cochrane Pregnancy and Childbirth Group's Trials Register (1 March 2013), checked reference lists of retrieved studies and contacted trial authors to locate additional unpublished data.

Selection criteria

Randomised controlled trials, cluster-randomised trials, randomised cross-over trials, and quasi-randomised controlled trials (with allocation by maternal birth date or hospital record number) of psychosocial smoking cessation interventions during pregnancy.

Data collection and analysis

Two review authors independently assessed trials for inclusion and trial quality, and extracted data. Direct comparisons were conducted in RevMan, and subgroup analyses and sensitivity analysis were conducted in SPSS.

Main results

Eighty-six trials were included in this updated review, with 77 trials (involving over 29,000 women) providing data on smoking abstinence in late pregnancy.

In separate comparisons, counselling interventions demonstrated a significant effect compared with usual care (27 studies; average risk ratio (RR) 1.44, 95% confidence interval (CI) 1.19 to 1.75), and a borderline effect compared with less intensive interventions (16 studies; average RR 1.35, 95% CI 1.00 to 1.82). However, a significant effect was only seen in subsets where counselling was provided in conjunction with other strategies. It was unclear whether any type of counselling strategy is more effective than others (one study; RR 1.15, 95% CI 0.86 to 1.53). In studies comparing counselling and usual care (the largest comparison), it was unclear whether interventions prevented smoking relapse among women who had stopped smoking spontaneously in early pregnancy (eight studies; average RR 1.06, 95% CI 0.93 to 1.21). However, a clear effect was seen in smoking abstinence at zero to five months postpartum (10 studies; average RR 1.76, 95% CI 1.05 to 2.95), a borderline effect at six to 11 months (six studies; average RR 1.33, 95% CI 1.00 to 1.77), and a significant effect at 12 to 17 months (two studies, average RR 2.20, 95% CI 1.23 to 3.96), but not in the longer term. In other comparisons, the effect was not significantly different from the null effect for most secondary outcomes, but sample sizes were small.

Incentive-based interventions had the largest effect size compared with a less intensive intervention (one study; RR 3.64, 95% CI 1.84 to 7.23) and an alternative intervention (one study; RR 4.05, 95% CI 1.48 to 11.11).

Feedback interventions demonstrated a significant effect only when compared with usual care and provided in conjunction with other strategies, such as counselling (two studies; average RR 4.39, 95% CI 1.89 to 10.21), but the effect was unclear when compared with a less intensive intervention (two studies; average RR 1.19, 95% CI 0.45 to 3.12).

The effect of health education was unclear when compared with usual care (three studies; average RR 1.51, 95% CI 0.64 to 3.59) or less intensive interventions (two studies; average RR 1.50, 95% CI 0.97 to 2.31).

Social support interventions appeared effective when provided by peers (five studies; average RR 1.49, 95% CI 1.01 to 2.19), but the effect was unclear in a single trial of support provided by partners.

The effects were mixed where the smoking interventions were provided as part of broader interventions to improve maternal health, rather than targeted smoking cessation interventions.

Subgroup analyses on primary outcome for all studies showed the intensity of interventions and comparisons has increased over time, with higher intensity interventions more likely to have higher intensity comparisons. While there was no significant difference, trials where the comparison group received usual care had the largest pooled effect size (37 studies; average RR 1.34, 95% CI 1.25 to 1.44), with lower effect sizes when the comparison group received less intensive interventions (30 studies; average RR 1.20, 95% CI 1.08 to 1.31), or alternative interventions (two studies; average RR 1.26, 95% CI 0.98 to 1.53). More recent studies included in this update had a lower effect size (20 studies; average RR 1.26, 95% CI 1.00 to 1.59), I2= 3%, compared to those in the previous version of the review (50 studies; average RR 1.50, 95% CI 1.30 to 1.73). There were similar effect sizes in trials with biochemically validated smoking abstinence (49 studies; average RR 1.43, 95% CI 1.22 to 1.67) and those with self-reported abstinence (20 studies; average RR 1.48, 95% CI 1.17 to 1.87). There was no significant difference between trials implemented by researchers (efficacy studies), and those implemented by routine pregnancy staff (effectiveness studies), however the effect was unclear in three dissemination trials of counselling interventions where the focus on the intervention was at an organisational level (average RR 0.96, 95% CI 0.37 to 2.50). The pooled effects were similar in interventions provided for women with predominantly low socio-economic status (44 studies; average RR 1.41, 95% CI 1.19 to 1.66), compared to other women (26 studies; average RR 1.47, 95% CI 1.21 to 1.79); though the effect was unclear in interventions among women from ethnic minority groups (five studies; average RR 1.08, 95% CI 0.83 to 1.40) and aboriginal women (two studies; average RR 0.40, 95% CI 0.06 to 2.67). Importantly, pooled results demonstrated that women who received psychosocial interventions had an 18% reduction in preterm births (14 studies; average RR 0.82, 95% CI 0.70 to 0.96), and infants born with low birthweight (14 studies; average RR 0.82, 95% CI 0.71 to 0.94). There did not appear to be any adverse effects from the psychosocial interventions, and three studies measured an improvement in women's psychological wellbeing.

Authors' conclusions

Psychosocial interventions to support women to stop smoking in pregnancy can increase the proportion of women who stop smoking in late pregnancy, and reduce low birthweight and preterm births.

Résumé scientifique

Interventions psychosociales pour encourager les femmes à arrêter de fumer pendant la grossesse

Contexte

Le tabagisme pendant la grossesse reste l'un des quelques facteurs pouvant être évité et qui est associé à des complications pendant la grossesse, à la mortinatalité, à l’insuffisance pondérale, à une naissance prématurée et à de graves conséquences à long terme pour les femmes et les bébés. Le tabagisme pendant la grossesse est en baisse dans les pays à revenu élevé, mais il est fortement associé à la pauvreté et s’accroît dans les pays à faible et moyen revenu.

Objectifs

Évaluer les effets des interventions de sevrage tabagique pendant la grossesse sur le comportement tabagique et les critères de jugement cliniques périnataux.

Stratégie de recherche documentaire

Dans cette cinquième mise à jour, nous avons effectué des recherches dans le registre des essais cliniques du groupe Cochrane sur la grossesse et la naissance (1er mars 2013), vérifié les références bibliographiques des études trouvées et contacté les auteurs des essais afin de localiser des données supplémentaires non publiées.

Critères de sélection

Essais contrôlés randomisés, essais randomisés par groupe, essais croisés randomisés et essais contrôlés quasi-randomisés (répartis par date de naissance ou par date d’admission à l'hôpital) des interventions psychosociales portant sur le sevrage tabagique pendant la grossesse.

Recueil et analyse des données

Deux auteurs de la revue ont indépendamment évalué les essais à inclure et la qualité des essais et extrait les données de manière indépendante. Des comparaisons directes ont été menées dans RevMan, les analyses en sous-groupe et l'analyse de sensibilité ont également été réalisées dans SPSS.

Résultats principaux

Quatre-vingt-six essais ont été inclus dans cette revue mise à jour, avec 77 essais (portant sur plus de 29 000 femmes) fournissant des données sur l'abstinence tabagique en fin de grossesse.

Dans des comparaisons séparées, les interventions d’aide psychologique ont démontré un effet significatif par rapport aux soins habituels (27 études; risque relatif moyen (RR) 1,44; intervalle de confiance (IC) à 95% 1,19 à 1,75) et un effet limité par rapport aux interventions moins intensives (16 études; RR moyen de 1,35, IC à 95% 1,00 à 1,82). Cependant, un effet significatif a été observé uniquement dans les sous-groupes où l’aide psychologique était administrée en association avec d'autres stratégies. Il était difficile de savoir si une quelconque stratégie d’aide psychologique était plus efficace que d'autres (une étude; RR 1,15, IC à 95% 0,86 à 1,53). Dans les études comparant l’aide psychologique à des soins habituels (la plus grande comparaison), il était difficile de savoir si les interventions prévenaient la rechute tabagique chez les femmes ayant arrêté de fumer spontanément en début de grossesse (huit études; RR moyen de 1,06, IC à 95% 0,93 à 1,21). Cependant, un effet notable était observé dans l'abstinence tabagique au bout de zéro à cinq mois après la naissance (10 études; RR moyen 1,76, IC à 95% 1,05 à 2,95), un effet limité au bout de six à 11 mois (six études; RR moyen de 1,33, IC à 95% 1,00 à 1,77) et un effet significatif au bout de 12 à 17 mois (deux études, RR moyenne 2,20, IC à 95% 1,23 à 3,96), mais pas à long terme. Dans d'autres comparaisons, l'effet n'était pas significativement différent de l'absence d'effet pour la plupart des critères de jugement secondaires, mais les échantillons étaient de petite taille.

Les interventions intensives avaient le plus d'effet par rapport à une intervention moins intensive (une étude; RR 3,64, IC à 95% 1,84 à 7,23) et une intervention alternative (une étude; RR 4,05, IC à 95% 1,48 à 11,11).

Les interventions commentées démontraient un effet significatif uniquement par rapport aux soins habituels et fournissaient des mesures de soutien à d'autres stratégies, tels que l’aide psychologique (deux études; RR moyen de 4,39, IC à 95% de 1,89 à 10.21), mais l'effet n'était pas clair par rapport à une intervention moins intensive (deux études; RR moyen de 1,19, IC à 95% 0,45 à 3,12).

L'effet de la formation sanitaire n'était pas clair en comparaison avec les soins habituels (trois études; RR moyen 1,51, IC à 95% 0,64 à 3,59) ou les interventions moins intensives (deux études; RR moyen 1,50, IC à 95% 0,97 à 2,31).

Les interventions de soutien social semblaient plus efficaces lorsqu'elles étaient fournies par des pairs (cinq études; RR moyen de 1,49, IC à 95% 1,01 à 2,19), mais l'effet n'était pas clair dans un unique essai de soutien apporté par le partenaire.

Les effets étaient mitigés lorsque l’intervention sur le tabagisme était fournie parmi des interventions plus vastes pour améliorer la santé maternelle, plutôt que les interventions ciblées uniquement sur le sevrage tabagique.

Les analyses en sous-groupe sur les critères de jugement de toutes les études ont montré que l'intensité des interventions et les comparaisons ont augmenté au fil du temps. Les interventions de plus grande intensité sont plus susceptibles d'avoir un nombre de comparaisons plus élevé. Lorsque il n'y avait aucune différence significative, les essais dans lesquels le groupe témoin recevait les soins habituels avaient le plus d’effets combinés (37 études; RR moyen de 1,34, IC à 95% 1,25 à 1,44), avec des effets plus faibles lorsque le groupe de comparaison avait reçu des interventions moins intensives (30 études; RR moyen 1,20, IC à 95% 1,08 à 1,31), ou d'autres interventions (deux études; RR moyen 1,26, IC à 95% 0,98 à 1,53). Des études plus récentes inclues dans cette mise à jour avaient un effet plus faible (20 études; RR moyen 1,26, IC à 95% 1,00 à 1,59),I2 =3%, par rapport à celles de la précédente version de la revue (50 études; RR moyen 1,50, IC à 95% 1,30 à 1,73). Il y avait des tailles d'effets similaires dans les essais avec une abstinence tabagique validée biochimiquement (49 études; RR moyen de 1,43, IC à 95% 1,22 à 1,67) et ceux avec une abstinence auto-déclarée (20 études; RR moyen de 1,48, IC à 95% 1,17 à 1,87). Il n'y avait aucune différence significative entre les essais appliqués par des chercheurs (études d’efficacité) et ceux appliqués par le personnel effectuant le suivi de la grossesse (études d’efficience). Cependant, l'effet n'était pas clair dans trois essais portant sur des interventions d’aide psychologique où l'intervention était à une échelle organisationnelle (RR moyen 0,96, IC à 95% 0,37 à 2,50). Les effets combinés étaient similaires dans les interventions proposées pour les femmes avec un statut socio-économique faible (44 études; RR moyen de 1,41, IC à 95% 1,19 à 1,66), par rapport à d'autres femmes (26 études; RR moyen de 1,47, IC à 95% 1,21 à 1,79); bien que l'effet n'était pas clair dans les interventions chez les femmes provenant de groupes minoritaires ethniques (cinq études; RR moyen 1,08, IC à 95% 0,83 à 1,40) et de groupes autochtones (deux études; RR moyen 0,40, IC à 95% 0,06 à 2,67). Mais surtout, les critères de jugement regroupés ont démontré une baisse de 18% des accouchements prématurés lorsque les femmes avaient reçu des interventions psychosociales (14 études; RR moyen 0,82, IC à 95% 0,70 à 0,96), de même qu’une baisse des nourrissons nés avec une insuffisance pondérale (14 études; RR moyen 0,82, IC à 95% 0,71 à 0,94). Les interventions psychosociales ne semblaient pas avoir d'effets indésirables et trois études mesuraient une amélioration du bien-être psychologique des femmes.

Conclusions des auteurs

Les interventions psychosociales visant à soutenir les femmes à arrêter de fumer pendant la grossesse peuvent augmenter la proportion de femmes qui arrêtent de fumer en fin de grossesse et réduire les naissances prématurées et les insuffisances pondérales des nourrissons.

Plain language summary

Psychosocial interventions for supporting women to stop smoking in pregnancy

Smoking during pregnancy increases the risk of the mother having complications during pregnancy and the baby being born with low birthweight and preterm (before 37 weeks). Tobacco smoking during pregnancy is relatively common, although the trend is towards it becoming less frequent in high-income countries and more frequent in low- to middle-income countries.

The review showed that psychosocial interventions to support women to stop smoking increased the proportion of women who stopped smoking in late pregnancy and reduced the number of low birthweight and preterm births. There did not appear to be any adverse effects from the psychosocial interventions, and three studies measured an improvement in women's psychological wellbeing.

The review includes 86 randomised controlled trials, with data from seventy-seven trials (involving over 29,000 women). Nearly all studies were in high-income countries. The intervention that supported the most women to stop smoking in pregnancy appeared to be providing incentives. However, these results are based on only four trials with a small number of women (all in the US), and they only seemed to help women stop smoking when provided intensively (three trials). Counselling also appeared to be effective in supporting women to quit, but only when combined with other strategies (27 trials). The effectiveness of counselling was less clear when women in the control group received a less intensive smoking intervention (16 trials). Feedback also appeared to help women quit, but only when compared with usual care and combined with other strategies (two studies). It was unclear whether health education alone helped women quit, but the numbers of women involved in these trials were comparatively small. The evidence for social support was mixed; for instance, targeted peer support appeared to help women quit (five trials) but in one trial partner support did not. Women also reported that peer and partner support could be both helpful and unhelpful.

Increasing the frequency and duration of the intervention did not appear to increase the effectiveness. Interventions appeared to be as effective for women who were poor, as those who were not; but there is insufficient evidence that the interventions were effective for ethnic (five trials) and aboriginal women (two trials). Trials where the interventions became part of routine pregnancy care did not appear to help more women to quit, which suggests there are challenges to translating this evidence into practice.

Résumé simplifié

Interventions psychosociales pour encourager les femmes à arrêter de fumer pendant la grossesse

Le tabagisme pendant la grossesse accroît le risque de complications pour la mère pendant la grossesse et le bébé pourrait naitre prématuré (avant 37 semaines) et avec une insuffisance pondérale. Le tabagisme pendant la grossesse est relativement courant, bien que la tendance soit en baisse dans les pays à revenu élevé, mais plus fréquente dans les pays à faible et moyen revenu.

La revue a montré que les interventions psychosociales pour encourager les femmes à arrêter de fumer permettait d’accroître le nombre de femmes ayant arrêté de fumer en fin de grossesse et réduisait le nombre d’insuffisance pondérale à la naissance et de naissances prématurées. Les interventions psychosociales ne semblaient pas avoir d'effets indésirables et trois études mesuraient une amélioration du bien-être psychologique des femmes.

La revue inclut 86 essais contrôlés randomisés, avec des données provenant de 77 essais (portant sur plus de 29 000 femmes). Presque toutes les études étaient menées dans les pays à revenu élevé. L'intervention, qui encourageait la plupart des femmes à arrêter de fumer pendant la grossesse, a semblé fournir des mesures incitatives. Cependant, ces critères de jugement sont seulement basés sur quatre essais portant sur un petit nombre de femmes (toutes aux États-Unis) et l’intervention semblait aider les femmes à arrêter de fumer uniquement sous suivi intensif (trois essais). L’aide psychologique paraissait également être efficace pour soutenir les femmes à arrêter de fumer, ceci seulement lorsqu' elle était combinée avec d'autres stratégies (27 essais). L'efficacité de l’aide psychologique était moins évidente lorsque les femmes dans le groupe témoin avaient reçu une intervention tabagique moins intensive (16 essais). Les commentaires semblaient également aider les femmes à arrêter de fumer, mais seulement lorsqu’ils étaient comparés avec les soins habituels et associés à d'autres stratégies (deux études). Il était difficile de savoir si la formation sanitaire seule aidait les femmes à arrêter de fumer, le nombre de femmes impliquées dans ces essais étant relativement faible. Les preuves de soutien social étaient contrastées; par exemple, l'entraide ciblée semblait aider les femmes à arrêter (cinq essais), mais dans un essai, le soutien des partenaires n'aidait pas. Les femmes ont également rapporté que l’entraide et le soutien des partenaires pourraient être à la fois utiles et inutiles.

Une hausse de la fréquence et de la durée de l'intervention ne semblait pas augmenter l'efficacité. Les interventions semblaient être aussi efficaces pour les femmes à faible revenu, que celles qui ne l'étaient pas; mais il n'existe pas suffisamment de preuves pour démontrer que les interventions étaient efficaces pour les femmes de groupes ethniques (cinq essais) et autochtones (deux essais). Les essais, dans lesquels les interventions sont devenues routinières dans le cadre des soins pendant la grossesse, ne semblaient pas aider davantage de femmes à arrêter de fumer, ce qui suggère qu’il existe des défis pour appliquer ces preuves dans la pratique.

Notes de traduction

Traduit par: French Cochrane Centre 14th January, 2014
Traduction financée par: Financeurs pour le Canada : Instituts de Recherche en Sant� du Canada, Minist�re de la Sant� et des Services Sociaux du Qu�bec, Fonds de recherche du Qu�bec-Sant� et Institut National d'Excellence en Sant� et en Services Sociaux; pour la France : Minist�re en charge de la Sant�

Background

Description of the condition

Risks associated with smoking in pregnancy

Tobacco smoking in pregnancy remains one of the few preventable factors associated with complications in pregnancy, such as placental abruption, miscarriage, low birthweight (Kramer 1987), preterm birth (US DHHS 2004; Hammoud 2005; Salihu 2007; Rogers 2009; Vardavas 2010; Baba 2012), stillbirth and neonatal death (Kallen 2001). Tobacco smoking also has serious long-term health implications for women and infants; 5.4 million people per year currently die from tobacco use, and this is expected to rise to eight million per year in the next 30 years (WHO 2008a). 

Nicotine and other harmful compounds in cigarettes are developmental toxicants (Rogers 2009), which impact on the brain at critical developmental periods (Dwyer 2008) restricting the supply of oxygen and other essential nutrients, fetal growth (Crawford 2008), development of organs (Morales-Suarez-Varela 2006), including the lungs (Maritz 2008) and neurological development (Herrmann 2008; Blood-Siegfried 2010). Growing evidence suggests these 'developmental origins of disease' have life-long implications (Gluckman 2008).

Young women start smoking for many reasons including: belief it is a rite of passage into adult life, a gesture against authority, trying to appear modern and affluent, or to fit in with social networks (Todd 2001). Tobacco addiction is then caused by nicotine in tobacco which produces a cascade of actions, including release of "pleasure enhancing" dopamine, which strengthens associations of positive feelings with smoking behaviour and appears to be involved in all addictive behaviours (Schmidt 2004). Some suggest the negative feelings of "nicotine hunger" and unpleasant symptoms associated with nicotine withdrawal (Balfour 2004; Hughes 2007) may be stronger for pregnant women due to the physiological adaptations in pregnancy which accelerate nicotine metabolism (Ebert 2009; Ussher 2012a), however a recent study reported less severe withdrawal symptoms among pregnant women in the first 24 hours of abstinence, compared to non-pregnant women (Ussher 2012b).

Epidemiology of smoking in pregnancy

In high-income countries, such as Australia, Canada, Denmark, New Zealand, Sweden, the United Kingdom (UK) and the United States (US), the prevalence of smoking in pregnancy has declined from between 20% to 35% in the 1980s to between 10% and 20% in the early 2000s (Cnattingius 2004; US DHHS 2004; Giovino 2007; Dixon 2009b; Tong 2009; Al-Sahab 2010; Tappin 2010), with significant declines in the last decade bringing the prevalence of smoking in pregnancy well below 10% by 2010 (Lanting 2012). However, the decline has not been consistent across all sectors of society, with lower rates of decline among women with lower socio-economic status (US DHHS 2004; Pickett 2009; Graham 2010; Johnston 2011b; Lanting 2012). Tobacco smoking in high-income countries is a marker of social disadvantage and has been cited as one of the principal causes of health inequality between rich and poor (Wanless 2004), and understanding these disparities are central to understanding the tobacco epidemic (Graham 2010).  In Scotland, 30% of women living in the most deprived areas continued to smoke during pregnancy in 2008, compared to 7% in the least deprived areas (Tappin 2010). Women who continue to smoke in pregnancy are more likely to: have a low income, higher parity, no partner, low levels of social support, limited education; access publicly funded maternity care; and feel criticised by society (Graham 1977; Frost 1994; Graham 1996; Tappin 1996; Wakschlag 2003; US DHHS 2004; Ebert 2007; Schneider 2008; Pickett 2009). The World Health Organization (WHO) report into the Social Determinants of Health recognises a paradigm whereby disadvantaged people are more likely to use substances in response to their circumstances (WHO 2008b). There is also a significantly higher prevalence of smoking in pregnancy in several ethnic and aboriginal minority groups (Wiemann 1994; Kaplan 1997; Chan 2001; US DHHS 2004; Wood 2008; Dixon 2009b; Johnston 2011b).  In Australia, smoking during pregnancy is three times more prevalent among Aboriginal and Torres Strait Islander women (53%) than among non-Aboriginal women (16%) (Johnston 2011b), and similar disparities are reported between Maori and non-Maori women in New Zealand (Dixon 2009b). These disparities are largely in accord with social and material deprivation. However, in some migrant groups, cultural differences may cut across this social gradient (Troe 2008), which suggests that there are aspects of smoking socialisation not entirely explained by material deprivation.  In the United States, the highest rates of pre-pregnancy smoking were reported among Alaskan Native women (55.6%), American Indian women (46.9%), and White women (46.4%), with significantly lower rates (less than 20%) reported among African American, Hispanic and Asian-Pacific women (Tong 2011; Watt 2012). Women who are migrants or refugees to Australia, Canada, New Zealand, Northern Europe, the UK, or the US or who originate from South East Asia also retain a lower prevalence of smoking, despite major social disadvantage (Potter 1996; Small 2000; Bush 2003; Dixon 2009b). However, second-generation migrant women are more likely to smoke during pregnancy than first-generation women (Troe 2008), reflecting movement between stages of 'the tobacco epidemic' (Lopez 1994).

In low- and middle-income countries there is marked variation in prevalence of smoking in pregnancy, which reflects the dynamic nature of the tobacco epidemic in these regions (Richmond 2003; Polanska 2004; Bloch 2008).  Smoking rates among pregnant women have been comparatively low (9%) compared to men (50%), due to historical cultural constraints on women's smoking in many low- to middle-income countries (Bloch 2008). However, the prevalence of tobacco smoking among women is increasing and is expected to rise to 20% by 2025, shifting the global tobacco smoking epidemic from high-income countries to low- and middle-income countries (Samet 2001; Richmond 2003). The highest rates of smoking during pregnancy were reported in Latin America (18.3% in Uruguay 2004 to 2005) (Bloch 2008) and Eastern Europe (15% in Romania 2005 to 2006) (Meghea 2010). Low rates were reported in Pakistan (3%) (Bloch 2008), South East Asia (1.3%) (Barraclough 1999; Ostrea 2008), and China (2% in 1999), though increasing rates among female school children are causing concern (Kong 2008).  In India and Africa, rates of cigarette smoking were low (1.7% and 6.1% pregnant women reporting smoking cigarettes, respectively), (Steyn 2006; Bloch 2008; Palipudi 2009), while use of smokeless tobacco products was high among Indian (4.9% to 33.5%) (Palipudi 2009; Bloch 2008) and African women (6% to 7.5%) (Steyn 2006; Bloch 2008).  The WHO has identified this rise of tobacco use in young females in low-income, high population countries as one of the most ominous developments of the tobacco epidemic (WHO 2008a), jeopardizing efforts to improve maternal and child health (Cnattingius 2004; Bloch 2008). This increase is being driven by aggressive marketing from tobacco companies, who are predicting high profits from sales in low- and middle-income countries (Kaufman 2001), along with increased tobacco production in these regions (FAO 2003), which further entrenches the countries' tobacco dependence. Marketing strategies are specifically targeted at women and weak regulation of tobacco company marketing has been linked to a rapid increase in smoking among women, particularly those who are vulnerable (Kaufman 2001; Gilmore 2004; Graham 2009). A survey of women's knowledge in two African countries suggests women's knowledge of the risks of tobacco products was extremely limited (Chomba 2010), making women more vulnerable to tobacco marketing.

Issues around smoking in pregnancy are complicated by the intersection of gender (Healton 2009), where a woman's role is seen primarily as a 'reproducer', and emphasis is placed on the rights of the unborn fetus (pxii; World Health Organization 2001). There is a risk these arguments may be used to impose authority over women's behaviour, 'blaming' women for their own plight and that of their children, and using guilt or other means to undermine self-confidence; further reducing the control women have in their lives (Greaves 2007a).

In addition to the socio-economic factors associated with continued smoking, there are strong psychological associations, especially with depression and stress (Blalock 2005; Aveyard 2007; Crittenden 2007; Orr 2012), including race-related stress (Heath 2006; Fernander 2010; Nguyen 2012a). Depressed women are up to four times more likely to smoke during pregnancy than non-depressed women (Blalock 2005). Despite these strong associations, there is limited information available about the effects of smoking and interventions in pregnant women with psychological symptoms, as they are often excluded from trials (Blalock 2005). Furthermore, while tobacco control initiatives in high-income countries have been effective in reducing smoking, the stigmatisations of smokers has been an unintended consequence (Burgess 2009; Wigginton 2012), which is being increasingly recognised by the tobacco control community (Farrimond 2006; Thompson 2007a; Burgess 2009). Anti-smoking campaigns strive to inform, shock or shame people into quitting smoking and rarely take into account low self-esteem, low self-efficacy, poverty, stress and increased caring responsibilities that are common among women who continue to smoke during pregnancy (Gilbert 2005). A systematic review of qualitative experiences of women describes how smoking in pregnancy triggered "intense feelings of personal responsibility and inadequacy" and that women's responses to social disapproval varied (Flemming 2013). For some, it provided an incentive to attempt to quit, while among others it resulted in increased smoking, either in response to the stress of social pressure or as an act of rebellion against it (Flemming 2013). Some argue that health risk narratives and the associated social stigma produced through anti-smoking campaigns contribute to oppression among marginalised people, and a consequence is that these strategies may inspire resistance and resentment rather than compliance (Bond 2012; Wigginton 2012; Flemming 2013).

Although commercial cigarettes are the most prevalent form of tobacco use worldwide, the use of other forms of tobacco (e.g. smokeless tobacco, cigars and pipes, and waterpipes) are becoming more popular in many parts of the world, especially low- and middle-income countries (England 2010). Of particular concern are increasing efforts by the tobacco industry to commercialise and market smokeless tobacco products to young adults (Lambe 2007). In high-income countries, the use of smokeless tobacco appears to be highly localised among some indigenous groups in Canada and the US, including Lumbee Indian, Navajo, and Alaskan Native communities (Strauss 1997; Spangler 2001; Patten 2009; Kim 2009a; Kim 2010). In India, one-third (33.5%) of all pregnant women reported using smokeless tobacco (Bloch 2008). In the Democratic Republic of Congo, 6% to 41.8% of pregnant women surveyed reported using other forms of tobacco, primarily snuff (Bloch 2008; Chomba 2010). In South Africa 7.5% of pregnant women surveyed reported using snuff (Steyn 2006). In Iran there has been concern over the 8% prevalence of local waterpipe tobacco smoking among pregnant women (Mirahmadizadeh 2008). These tobacco products may be cheaper and viewed as less harmful than cigarettes (England 2010). In some cases use may be a traditional cultural norm or a medicinal aid to reduce nausea in early pregnancy. However, these products can be high in nicotine content and cause nicotine addiction. Use of these products has been associated with increased oral and pancreatic cancer, and cardiovascular disease (England 2010). There is a paucity of research into the effect of these products on pregnancy outcomes and studies into the effects of these products can be challenging as the chemical content of various toxic compounds is variable and often poorly regulated.  However, limited evidence suggests smokeless tobacco use is associated with decreased birthweight and preterm birth (Verma 1983; Gupta 2004; Pratinidhi 2010), stillbirth (Gupta 2006; Gupta 2012), maternal anaemia (Subramoney 2008), degenerative placental changes (Ashfaq 2008), and adverse infant neurobehavioural outcomes (Hurt 2005). Smoking more than one waterpipe per day (Tamim 2008) or starting to smoke waterpipes during the first trimester (Mirahmadizadeh 2008) was also associated with an increased risk of having a low birthweight baby.

Exposure to environmental tobacco smoke (ETS) also poses risks to pregnant women and their infants (Yang 2010). Studies suggest the risk may be exacerbated in low-income countries where exposure to indoor cooking smoke is also common (Kadir 2010). In China, 75.1% of pregnant non-smoking women were regularly exposed to environmental tobacco smoke from their husbands’ smoking (Yang 2010). Studies in high-income countries demonstrate that eliminating smoking in the workplace and other public spaces significantly reduces environmental tobacco smoke exposure and improves health outcomes, including preterm births (Cox 2013). One study in Indonesia reported increased collective efficacy when environmental tobacco smoke exposure was addressed through a well-publicised community household smoking ban (Nichter 2010). However, as these measures do not extend to homes (Oncken 2009), some argue domestic environmental tobacco smoke exposure may be increasing as public health policies restrict smoking of partners in public places, and the social position of women may limit their ability to enforce smoke-free policies within their homes (Tong 2009).

A positive theme emerging from this literature is that a higher proportion of women stop smoking during pregnancy than at other times in their lives.  Up to 49% of women who smoked before pregnancy ‘spontaneously quit’ before their first antenatal visit (Quinn 1991; Woodby 1999; Hotham 2008), a quit rate substantially higher than reported in the general population (Ershoff 1999; McBride 2003; Tong 2008). However, these spontaneous quitting rates may be lower among women with lower socio-economic status (Mullen 1999). There are significant psychosocial differences between women who 'spontaneously quit' and women who continue to smoke in late pregnancy. Women who spontaneously quit usually smoke less, are more likely to have stopped smoking before, have a non-smoking partner, have more support and encouragement at home for quitting, are less seriously addicted, and have stronger beliefs about the dangers of smoking (Baric 1976; Ryan 1980; Cinciripini 2000; Passey 2012). Pregnant women are also more likely to use coping strategies to avoid relapse than non-pregnant women (Ortendahl 2007c; Ortendahl 2008a; Ortendahl 2009a), however less than a third of these women remain abstinent after one year postpartum (CDCP 2002; Fang 2004), supporting qualitative evidence that many women see pregnancy as a temporary period of abstinence for the sake of the baby (Stotts 1996; Lawrence 2005a; Flemming 2013). Despite high relapse rates, some studies suggest that the long-term effects of spontaneous quitting in pregnancy are significant (Rattan 2013), and others argue this success is important to recognise to avoid 'pathologising' smoking cessation and eroding confidence in human agency to overcome problems (Chapman 2010). 

Given the complexity of the health and social dimensions of smoking in pregnancy there are conflicting perspectives regarding the most appropriate approaches. A dominant theme is that smoking in pregnancy is a lifestyle choice, however, there is concern this can lead to 'victim blaming' (Bond 2005), that individualised, behaviourist approaches are unlikely to adequately address health inequalities alone (Baum 2009), and that drug dependence and addiction is best dealt with in the domain of social policy and public health (Ebert 2009). Nevertheless, some suggest there is a role for individual support which is positive, not punitive (Bond 2012), and others express a concern that framing smoking in pregnancy solely as a social problem may make health professionals reluctant to intervene and offer support (McLellan 2000).   

Description of the intervention

This review evaluates the effectiveness of individual psychosocial interventions that aim to motivate and support women to stop smoking in pregnancy, or prevent smoking relapse among women who have spontaneously quit. Psychosocial interventions are defined as non-pharmacological strategies that use cognitive-behavioural, motivational and supportive therapies to help women to quit, including counselling, health education, feedback, financial incentives, and social support from peers and/or partners (see Types of interventions), as well as dissemination trials.

Other smoking cessation intervention reviews

At the time of this update there were 73 other Cochrane reviews assessing the effectiveness of tobacco smoking cessation interventions for all populations (see Appendix 1). These include reviews on the following.

  • Population wide measures such as: legislative smoking bans, mass media campaigns, organisational interventions (workplace and school-based interventions), healthcare financing systems for increasing use of tobacco dependence treatment, advertising and promotion to reduce tobacco use, preventing tobacco smoking in public places, and impact of advertising on adolescent smoking.

  • Community interventions including family-based programmes, group behaviour interventions, family and carer interventions for reducing environmental tobacco smoke, school-based programmes, and school policies.

  • Individual psychosocial interventions, including aversive smoking, acupuncture, hypnotherapy, self-help, exercise, individual behavioural counselling, motivational interviewing, stage-based interventions, competitions and incentives, telephone counselling, mobile phone-based interventions, Internet-based interventions, nursing and physician advice, enhancing partner support, feedback, community pharmacy interventions, training health professionals in smoking cessation, use of electronic records, prevention of weight gain after smoking cessation, improving recruitment into cessation programs, harm reduction, reduction versus abrupt cessation, biomedical risk assessments, electronic cigarettes, incentives to prevent smoking in young people, relapse prevention, and interventions to reduce non-cigarette tobacco use, including waterpipe smoking cessation.

  • Individual pharmacological interventions, including antidepressants, anxiolytics, nicotine replacement therapy (NRT), clonidine, mecamylamine, nicobrevin, nicotine agonists, opioid agonists, cannabinoid type 1 receptor agonists, silver acetate, lobeline, and nicotine vaccines, increasing adherence to medications for tobacco dependence, behavioural interventions as adjuncts to pharmacotherapies, combined pharmacotherapy and behavioural interventions;and an ‘overview of pharmacological reviews’.

  •  Interventions in specific population groups, including people with: schizophrenia and serious mental illness, depression, substance abuse, cardiovascular and pulmonary disease; pre-operative and hospitalised patients; Indigenous populations and Indigenous youth; and people in dental settings.

  • Other reviews, assessing effectiveness of interventions to recruit patients into smoking cessation programs, and reduce harm from continued tobacco use.

How the intervention might work

Pregnancy has been described as a ‘window of opportunity’ for smoking cessation (McBride 2003). Pregnancy increases a woman’s perception of risk and personal outcomes, therefore strong affective or emotional responses are more likely to be prompted (Slade 2006; Ortendahl 2008b). It also redefines a woman's self-concept or social role (Ortendahl 2007b), especially when failure to comply with a social role results in social stigmatisation (Ortendahl 2007a; Ortendahl 2008c). Psychosocial interventions involve a range of social and psychological components which aim to increase motivation or affective or emotional responses to support pregnant women to stop smoking and support women to develop coping strategies to avoid relapse (Ortendahl 2007c; Pilling 2010). For example, counselling, feedback and financial incentives are all designed to enhance motivation to quit and move women closer towards the 'action' stage of change. Thirty-seven individual 'behaviour change techniques' or observable components used in interventions in the previous version of this review have been identified (Lorencatto 2012).

Psychosocial interventions to support women to stop smoking in pregnancy increasingly incorporate theoretical frameworks to inform, develop and evaluate strategies designed to influence behaviour (Green 2005b; Glanz 2008; Michie 2008; Bartholomew 2011). Using behaviour change theories in the context of addiction has been identified as a useful way to identify modifiable determinants and/or behaviour change techniques (Webb 2010).There are many theories of behaviour, which provide a summary of constructs, procedures and methods for understanding behaviour, and present hypothesised relationships or causal pathways that influence behaviour (Michie 2012). While some argue there is little apparent consensus about which theories are best to use in designing interventions (Noar 2005), most theories of behaviour change postulate a role for six broad classes of variables (Glanz 2008):

  1. attitudes and beliefs about the behaviours or the outcomes of change (used in health education and counselling strategies);

  2. beliefs about self-efficacy or perceived ability to enact and/or maintain the target behaviour change (used in counselling strategies such as motivational interviewing or cognitive behaviour therapy);

  3. the role of contextual factors, particularly social factors, either directly and/or mediated through people’s beliefs (used in social support strategies);

  4. previous experience with the behaviour either directly or indirectly through the processes of modelling (modelling can be seen as an element of social influence) (used in social support strategies);

  5. priority for action, a person can only pursue a limited number of goals of any one time; and

  6. the notion of a stage-based or systematic step-like progression towards behaviour change, which is incorporated into the assessment stage of many smoking cessation interventions (Prochaska 1992).

Why it is important to do this review

There are many psychosocial interventions that have been evaluated to support women to stop smoking during pregnancy. This review synthesises the evidence from these trials to generate evidence, which is of direct relevance for practitioners, policy-makers, and researchers. Synthesis enables comparison of whether interventions have been shown to be effective in individual studies and whether this effect has been replicated in other settings. Importantly, individual studies are unlikely to have sufficient power to evaluate the effect of interventions on perinatal outcomes or to conduct subgroup analyses to assess if there are differential effects among vulnerable subpopulations with high rates of smoking during pregnancy. Finally, collation of the body of evidence helps to identify any gaps for future research. 

This is the fifth update of this Cochrane review, previously entitled ‘Interventions to promote smoking cessation during pregnancy’. The first version was published in 1995 on CD Rom and previously updated in The Cochrane Library in 1999, 2004 and 2009.  Previous versions of this review have demonstrated the potential for individual interventions during pregnancy to have a modest but significant effect on reducing smoking, preterm births and infants born with low birthweight (Lumley 2009). This evidence has been instrumental in individual psychosocial interventions becoming a part of routine pregnancy care in many high-income countries in the past decade (Flenady 2005; Ministry of Health 2007; Fiore 2008; NICE 2010; Wong 2011). These guidelines generally incorporate a number of interventions, including identifying women who smoke during pregnancy, providing advice about risks, and supporting women to stop smoking.

In this review update, we have ‘split’ the previous version into two reviews: (1) this review focusing on psychosocial interventions to support women to stop smoking in pregnancy; and (2) a second review specifically focusing on pharmacological interventions to promote smoking cessation in pregnancy (Coleman 2012b). This split was necessary as there are different issues of concern for psychosocial and pharmacological interventions. Psychosocial interventions are now part of routine care in many high-income countries and contemporary issues focus on strategies to increase efficacy, and adaptation of psychosocial interventions to different contexts and settings, sometimes requiring different study designs (e.g. cluster trials of implementation). As many interventions involve multiple strategies or use of components which are tailored to individual women, it is very difficult to assess the independent effect of individual components of psychosocial interventions.  As the efficacy and safety of pharmacological treatment (e.g. Nicotine Replacemernt Therapy, Bupropion) during pregnancy (Slotkin 2008) remains uncertain, more rigid study designs (i.e. randomised double-blind placebo-controlled trials) are required to assess the risks and efficacy. 

To complement what is known from research literature about smoking in pregnancy, direct contributions to this review were sought from women who smoked before or during pregnancy in 1999. Women were identified through community networks, and their views emphasised the need to focus attention on potential adverse effects of smoking cessation programmes; in particular, the consequent guilt, anxiety and additional stress experienced by those who continue to smoke, especially through 'high-risk' pregnancies, and the detrimental effect on their relationships with their family and maternity care providers (Oliver 2001).

In this update, we indirectly considered women’s views reported in a systematic review of qualitative studies (Flemming 2013), which reinforce the previous contributions, identifying four main themes which have implications for interventions to support women to stop smoking in pregnancy.

  1. Smoking is an embedded part of the lives of many women living in disadvantaged circumstances.

  2. Women see smoking in pregnancy in terms of the risks it presents to their unborn baby, which can trigger guilt.

  3. Quitting was not seen in unambiguously positive terms and was seen to have downsides, disrupting relationships and removing a habit perceived as helping women cope.

  4. Partners play an important role in influencing women’s smoking behaviour in pregnancy, either as barriers or facilitators to quitting.

We also indirectly considered the views of pregnancy care providers reported in consultation for a Clinical Practice Guideline on Smoking Cessation in pregnancy (Williams 2010) in the UK; and the views of guideline developers requesting evidence for an international guideline on 'Management of Tobacco Use in Pregnancy' (CDCP 2013). Some of the major issues and gaps included:

  • whether psychological interventions are effective;

  • whether interventions are effective for pregnant teens and other hard-to-reach and vulnerable groups, including ethnic and minority populations;

  • whether interventions are effective for women who are mentally unwell or experiencing substance misuse;

  • whether interventions are effective in low- and middle-income countries.

In addition to consideration of women's views and feedback from guideline developers, we also considered thesis critiques of the previous version of this review (Gilligan 2008; Vilches 2009), health programme planning models (Green 2005b; Bartholomew 2011), various publications on factors affecting intervention efficacy (Greenhalgh 2004; Hoddinott 2010), descriptions of intervention components (Lorencatto 2012), and the 'critical factors' identified by authors of included studies reported in the results or discussion. As smoking in pregnancy has important impacts on health inequalities, we have introduced a focus on equity in this review, as recommended in the 'PRISM-Equity' guidelines for reporting interventions with a potential impact on equity (Welch 2012). We have synthesised this information into a logic model to identify key variables that may impact on intervention effectiveness (see Figure 1), to guide analysis and subgroup analyses planning 'a priori' (Petticrew 2012).

Figure 1.

Logic model for systematic review analysis of potential factors impacting on efficacy of interventions for supporting women to stop smoking in pregnancy.

Objectives

This review evaluated the effect of psychosocial interventions designed to support women to stop smoking in pregnancy and aimed to address the following questions.

Primary objectives

  • To identify whether psychosocial interventions can support women to stop smoking in pregnancy

  • To compare the effectiveness of the main psychosocial intervention strategies in supporting women to stop smoking in pregnancy (i.e. counselling, health education, feedback, social support, incentives)

Secondary objectives

  • To identify if the intensity of the intervention corresponds to an effect size

  • To identify any specific intervention components associated with an effect (e.g. telephone counselling, self-help manuals)

  • To identify if psychosocial interventions in pregnancy have an impact on health outcomes for the mother (i.e. caesarean section, breastfeeding) and infant (i.e. mean birthweight, low birthweight, preterm births, very preterm births, perinatal mortality)

  • To identify if there are any positive or negative psychological effects reported among women receiving psychosocial interventions in pregnancy

  • To identify participants (women and pregnancy care providers) views of the psychosocial interventions in this review

  • To identify if psychosocial interventions have an effect on family functioning or other relationships for the mother, including non-accidental injury

  • To identify if psychosocial interventions during pregnancy can reduce the proportion of women who start smoking postpartum

  • To identify whether any methods for training and implementing psychosocial interventions have an effect on the knowledge, attitudes and behaviour of pregnancy care providers

  • To identify whether psychosocial interventions provided for women who have spontaneously quit smoking in early pregnancy, can reduce the proportion of women who start smoking by late pregnancy (relapse)

  • To identify whether psychosocial interventions are effective for women in vulnerable subpopulation groups (including women categorised as having low socio-economic status, young women (less than 20 years), ethnic minority and aboriginal women, and women in low- and middle-income countries

  • To identify whether psychosocial interventions, which are shown to be effective when implemented under trial conditions by a dedicated research team (efficacy studies), are still effective when implemented in a routine pregnancy care setting by existing staff (effectiveness studies)

  • To identify if psychosocial interventions to support women to stop smoking in pregnancy are cost-effective

  • To identify if there are any adverse effects reported as a result of women receiving psychosocial interventions to support them to stop smoking in pregnancy

  • To identify whether recently included studies are as effective as studies included in previous versions of this review

  • To identify if any of the risk of bias assessments have a significant impact on the effect size of the intervention

Methods

Criteria for considering studies for this review

Types of studies

All randomised controlled trials, cluster-randomised controlled trials, and randomised cross-over trials of psychosocial interventions where a primary aim of the study was smoking cessation in pregnancy. Quasi-randomised studies were only considered for inclusion if there was a very low risk of interference with the sequence generation (e.g. allocation by odd or even maternal birth date or hospital record number).

Types of participants

  1. Women who are currently smoking or have recently quit smoking and are pregnant, in any care setting.

  2. Women who are currently smoking or have recently quit smoking and are seeking a pre-pregnancy consultation.          

  3. Health professionals in trials of implementation strategies of psychosocial interventions to support pregnant women to stop smoking.

Where possible, we have separated outcomes for women who spontaneously quit smoking when they become pregnant, and women who continue to smoke during pregnancy, as significant differences have been reported previously (Baric 1976; Ryan 1980; Cinciripini 2000; Passey 2012).

Types of interventions

  1. Counselling interventions are those which provide motivation to quit, support to increase problem solving and coping skills (Ortendahl 2007c; Ortendahl 2008a; Ortendahl 2009b), and may incorporate 'transtheoretical' models of change (Prochaska 1992; Prochaska 2007). This includes interventions such as motivational interviewing, cognitive behaviour therapy, psychotherapy, relaxation, problem solving facilitation, and other strategies. Counselling interventions may be provided face-to-face, by telephone, via interactive computer programs, or using audiovisual equipment. The duration of counselling may range from brief interventions (less than five minutes) to more intensive interventions, which can last for up to an hour and be repeated over multiple sessions. Counselling may be provided by a range of personnel, including pregnancy care providers, trained counsellors, or others, on-site or by referral to specialist stop smoking services. Interventions that involved provision of videos with personal stories were included as counselling in this review.

  2. Health education interventions are defined as those where women are provided with information about the risks of smoking and advice to quit, but are not given further support or advice about how to make this change. Interventions where the woman was provided with automated support such as self-help manuals or automated text messaging, but there was no personal interaction at all, were coded as health education in this review.

  3. Feedback interventions are those where the mother is provided with feedback with information about the fetal health status or measurement of by-products of tobacco smoking to the mother. This includes interventions such as ultrasound monitoring and carbon monoxide or urine cotinine measurements, with results fed back to the mother (does not include where measurements are used for confirming smoking abstinence in the study).      

  4. Incentive-based interventions include those interventions where women receive a financial incentive, contingent on their smoking cessation; these incentives may be gift vouchers. Interventions that provided a 'chance' of incentive (e.g. lottery tickets) were not included as 'incentives' in this update, but were included in counselling and subgroup analysis of trials incorporating use of lottery tickets will be reported. Gifts and other incentives to promote participation in the study (but were not contingent on smoking cessation), were not coded as incentive-based interventions in this review.

  5. Social support (peer and/or partner) includes those interventions where the intervention explicitly included provision of support from a peer (including self-nominated peers, 'lay' peers trained by project staff, or support from healthcare professionals), or partners, as a strategy to promote smoking cessation.  

  6. Other strategies, which could not be included in the categories listed above, including exercise, and dissemination interventions (where both intervention and control group received the same intervention, but the dissemination strategy differed). 

In this review we have categorised interventions according to the 'main' strategy used, however many interventions incorporate several components. Therefore, interventions are coded according to whether the strategy was a:

  • single intervention - with only one main strategy used;

  • multiple intervention - which included several strategies being offered to all women;

  • tailored intervention - where additional optional strategies were available for women.

Trials that combined strategies for smoking cessation with other interventions to promote maternal health in pregnancy were considered for the review for smoking cessation and reduction outcomes but not for infant outcome measures such as birthweight, preterm birth, breastfeeding and perinatal mortality, which might be attributable to other components of an intervention package. We have included interventions that offered pharmacological therapies as part of a tailored intervention where there were higher levels of psychosocial support provided to participants in the intervention arm, compared with the control arm. Trials were excluded where the sole aim was to reduce: smokeless tobacco use; environmental tobacco smoke exposure; where the primary population was not pregnant women (e.g. partners, non-pregnant women); or the intervention was not primarily aimed at cessation during pregnancy (e.g. postpartum interventions). Studies were included where smokeless tobacco use, environmental tobacco smoke exposure or partner smoking were targeted in conjunction with interventions addressing the primary aim of supporting pregnant women to stop smoking in pregnancy. We have included dissemination studies, where the primary intervention includes strategies to disseminate smoking cessation interventions in pregnancy care settings (e.g. training, audit and feedback).

Types of comparisons

Any type of comparison group was included and was coded according to the following.

  1. 'Usual care' or no additional intervention reported.

  2. Less intensive interventions where the control group received some of the intervention or an approximation of 'usual care' consistently provided by the research team.

  3. Alternative interventions, where the control group received different intervention components than the intervention group, of the same intensity.

Types of settings

Any setting, including residential and community settings, family planning clinics, pre-pregnancy planning clinics or general practitioner clinics, prenatal care clinics and hospitals.

The 'PROGRESS-Plus' criteria (Oliver 2008b; Ueffing 2009) were used to categorise interventions which were provided for vulnerable populations, including: social capital; place of residence; occupation; education; socio-economic status; ethnicity; age; or other factors which might impact on vulnerability. These categories are described in more detail in the methods.

Types of outcome measures

Primary outcomes
  1. Smoking abstinence in late pregnancy (point prevalence abstinence):

    1. self-reported or biochemically validated;

    2. biochemically validated only.

Secondary outcomes
  1. Continued abstinence in late pregnancy after spontaneous quitting (relapse prevention) in early pregnancy (self-reported or biochemically validated).

  2. Smoking abstinence in the postpartum period (self-reported or biochemically validated):

    1. zero to five months;

    2. six to 11 months;

    3. 12 to 17 months;

    4. 18 months or longer.

  3. Smoking reduction from the first antenatal visit to late pregnancy:

    1. numbers of women reducing smoking (any definition, > 50% self-reported, or biochemically validated);

    2. biochemical measures (mean cotinine and thiocynate);

    3. mean cigarettes per day (self-reported).

  4. Perinatal outcomes:

    1. mean birthweight;

    2. low birthweight (proportion less than 2500 g);

    3. very low birthweight (less than 1500 g);

    4. preterm births (proportion less than 37 weeks);

    5. stillbirths;

    6. neonatal deaths;

    7. all perinatal deaths.

  5. Mode of birth (caesarean section).

  6. Breastfeeding initiation and breastfeeding at three and six months after birth.

  7. Psychological effects: measures of anxiety, depression and maternal health status in late pregnancy and after birth.

  8. Impact on family functioning and other relationships in late pregnancy and postpartum.

  9. Participants' views of the interventions, both women’s and pregnancy care providers’ views.

  10. Measures of knowledge, attitudes and behaviour of health professionals (obstetricians, midwives and family physicians) with respect to facilitating smoking cessation in pregnancy.

  11. Cost-effectiveness.

  12. Adverse effects of smoking cessation programmes.

Search methods for identification of studies

This is the fifth update of this review and the details of previous searches are described in other published versions of this review (Lumley 1995a; Lumley 1995b; Lumley 1995c; Lumley 1995d; Lumley 1999; Lumley 2004; Lumley 2009).

Electronic searches

We searched the Cochrane Pregnancy and Childbirth Group’s Trials Register by contacting the Trials Search Co-ordinator (1 March 2013).

The Cochrane Pregnancy and Childbirth Group’s Trials Register is maintained by the Trials Search Co-ordinator and contains trials identified from:

  1. monthly searches of the Cochrane Central Register of Controlled Trials (CENTRAL);

  2. weekly searches of MEDLINE;

  3. weekly searches of Embase;

  4. handsearches of 30 journals and the proceedings of major conferences;

  5. weekly current awareness alerts for a further 44 journals plus monthly BioMed Central email alerts.

Details of the search strategies for CENTRAL, MEDLINE and Embase, the list of handsearched journals and conference proceedings, and the list of journals reviewed via the current awareness service can be found in the ‘Specialized Register’ section within the editorial information about the Cochrane Pregnancy and Childbirth Group.

Trials identified through the searching activities described above are each assigned to a review topic (or topics). The Trials Search Co-ordinator searches the register for each review using the topic list rather than keywords.

Searching other resources

We also checked cited studies while reviewing the trial reports and key reviews. Where necessary, we contacted trial authors to locate additional unpublished data.  

We did not apply any language restrictions.

[In addition, authors conducted a supplementary search for non-randomised studies, for the background and discussion, in MEDLINE, Embase, PsycLIT, and CINAHL (June 2008 to 1 March 2013) using the search strategy detailed in Appendix 2.]

Data collection and analysis

Selection of studies

Two review authors independently reviewed the full text of search results from the Cochrane Pregnancy and Childbirth Group and potential trials identified through other sources (CC/SP) to determine if they met the inclusion criteria for this review. Where there was disagreement, advice from co-authors was sought (SO/JC/AO/JT) and consensus reached by discussion.

Data extraction and management

Two review authors independently extracted data from the published reports without blinding as to journal, author, or research group. For each trial the following aspects were reported and coded into EPPI-Reviewer software (Thomas 2010). Independent data extraction was checked and areas of conflicting judgement were resolved by consensus, and where necessary discussion with co-authors. A summary of data collected is outlined in Appendix 3 and a summary reported for individual studies in the Characteristics of included studies table.

Assessment of risk of bias in included studies

We assessed the methodological quality of the included studies as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). The 'quality assessment' from previous reviews has been replaced with the 'Risk of bias' assessment.

(1) Sequence generation (checking for possible selection bias)

We have described for each included study the methods used to generate the allocation sequence, and have assessed the methods as:

  • low risk of bias (any truly random process, e.g. random number table; computer random number generator);

  • high risk of bias (any non random process, e.g. alternate clinic date; odd or even date of birth; hospital or clinic record number);

  • or unclear risk of bias.   

Studies where sequence generation was assessed as inadequate and there is a reasonable opportunity to interfere with random allocation (e.g. alternate clinic date) have been excluded in this update of the review. Studies randomised by odd or even date of birth or medical record number have continued to be included in this review as there is limited reasonable opportunity to manipulate the allocation.

(2) Equal baseline characteristics (checking for possible selection bias)

To further assess the risk of selection bias, we assessed whether the baseline characteristics were equal in each included study, and have assessed them as:

  • low risk of bias (baseline characteristics were assessed and equal in both study arms);

  • high risk of bias (where there were significant differences in baseline characteristics, suggesting possible bias in the selection of participants);

  • or unclear risk of bias.

(3) Allocation concealment (checking for possible selection bias)

We have described for each included study the method used to conceal the allocation sequence in sufficient detail to determine whether intervention allocation could have been foreseen in advance of, or during recruitment, or changed after assignment. We have assessed the methods as:

  • low risk of bias (e.g. telephone or central randomisation; consecutively numbered sealed opaque envelopes);

  • high risk of bias (e.g. open random allocation; unsealed or non-opaque envelopes; medical record number; date of birth);

  • or unclear risk of bias.   

(4) Blinding (checking for possible performance bias) of study participants and intervention providers

We have described for each included study the methods used, if any, to blind study participants and intervention providers from knowledge of which intervention a participant received. However, it is rarely feasible in psychosocial interventions to blind women or the intervention providers to group allocation. We have assessed the methods as:

  • low risk of bias;

  • high risk of bias;

  • or unclear risk of bias.

(5) Blinding (checking for possible performance bias) of outcome assessor

We have described for each included study the methods used, if any, to blind outcome assessors from knowledge of which intervention a participant received as recommended (West 2005). We have assessed the methods as:

  • low risk of bias;

  • high risk of bias;

  • or unclear risk of bias.

(6) Dealing with incomplete outcome data (checking for possible attrition bias through withdrawals, dropouts, protocol deviations, and intention-to-treat analysis)

We have described for each included study and for each outcome or class of outcomes the completeness of data including attrition and exclusions from the analysis. We have noted whether attritions and exclusions were reported, the numbers included in the analysis at each stage (compared with the total randomised participants), reasons for attrition or exclusion where reported, and whether missing data were balanced across groups. We considered it was reasonable to exclude women from the final analysis who had experienced miscarriage or fetal demise, developed serious medical conditions, moved out of the area, or changed to another provider of care. However, as there are also clear associations between these outcomes and smoking, we have categorised the risk of attrition bias as 'unclear'. Where possible, we included all other randomised women in the meta-analysis. Where data were not provided in such a way to enable inclusion of all other randomised participants, we have categorised these studies as high risk of attrition bias. We have assessed the methods as:

  • low risk of bias (outcomes for all randomised participants included in analysis);

  • high risk of bias (outcomes for all participants not reported, particularly if unequal attrition in both study arms);

  • or unclear risk of bias, which includes exclusions for medical conditions or moving.

(7) Reporting all outcomes (checking for possible selective reporting bias)

We have described for each included study how the possibility of selective outcome reporting bias was examined by us and what we found. We assessed the methods as:

  • low risk of bias (where it is clear that all of the studies' pre-specified primary outcomes and all expected outcomes of interest to the review have been reported);

  • high risk of bias (where not all the studies' pre-specified outcomes have been reported); one or more reported primary outcomes were not pre-specified; outcomes of interest are reported incompletely and so cannot be used; study fails to include results of a key outcome that would have been expected to have been reported);

  • or unclear risk of bias.

(8) Reliability of outcome measures used (checking for possible detection bias)

The unreliability of self-report as a measure of smoking status in healthcare settings, especially in maternity care (Pettiti 1981), was noted even in the first pregnancy trial (Donovan 1977). While this finding has not always been consistent (Fox 1989; Pickett 2009; Windsor 1985), the majority of other trials show substantial misclassification by self-report, with up to a quarter or a third of women who describe themselves as non-smokers having levels of salivary or urine cotinine (a metabolite of nicotine) incompatible with self-description (Mullen 1991; Petersen 1992; Kendrick 1995; Lillington 1995; Walsh 1997; Moore 2002; Tappin 2005; Parker 2007). A degree of misclassification is not surprising given the social stigma associated with smoking in pregnancy, and there appears to be less misclassification in non-pregnant populations (Patrick 1994). Some studies suggest that measurement of abstinence is reasonably accurate, but that there is greater inconsistency with reporting the amount of cigarettes smoked (Klebanoff 1998; Venditti 2012). Given this potential for bias, biochemical validation of smoking abstinence is now the standard for smoking cessation studies (West 2005; Shipton 2009). Use of cotinine concentration (saliva, urine or plasma) is the most sensitive and specific (saliva less than 15 ng/mL and urine less than 50 ng/mL). However, cotinine does not distinguish between smoking and use of nicotine replacement products, so expired air carbon monoxide is the preferred method for detecting recent smoking (less than 9 ppm) in many studies. Trials measuring cotinine need to ask participants about NRT use (available over the counter), ignore high levels in NRT users, and verify smoking abstinence with carbon monoxide levels (West 2005). However, several studies including use of NRT did use cotinine cut-offs to distinguish between smokers and non-smokers (Hegaard 2007). There may also be differential misclassification between intervention and control groups, though no investigations have published this effect. We have described for each included study whether the smoking outcome was biochemically validated (including measures used) or assessed by self-report only, and have included data on misclassification by self-report where they have been reported:

  • low risk of bias (biochemical validation);

  • high risk of bias (no biochemical validation);

  • or unclear risk of bias (including partial biochemical validation of a sample of the study population).

(9) Implementation of intervention

There are three main types of potential implementation problems trials (Walsh 2000):

  • not all participants in the intervention groups receiving the intervention;

  • intervention group participants not receiving all components of the intervention;

  • control groups receiving the intervention.

Failure to implement the intervention as planned limits the exposure of women to the intervention, and may negatively impact on the effectiveness of the intervention. Where possible, we included a description of any process evaluation reported. We have assessed the implementation of the intervention as:

  • low risk of bias (where process evaluation suggests the majority of participants received the intervention as planned);

  • high risk of bias (where process evaluation suggests a significant proportion of women did not receive the intervention as planned);

  • or unclear risk of bias (where process evaluation is not reported).

(10) Risk of control group contamination

Exposure of the control group to aspects of the intervention is a common challenge for intervention trials, particularly studies where healthcare providers are required to offer an intervention to some women, and not to others. Some trials use cluster-randomisation in order to reduce the risk of contamination, particularly when healthcare providers are involved in the intervention. The most likely impact is to increase the effect in the control arm, reducing the potential effect size between the intervention and control arms of the study. We have assessed the methods as:

  • low risk of bias, where the intervention providers are separate from the control group or strategies are employed to minimise the risk (such as cluster-randomisation);

  • high risk of bias, where the same provider is required to administer the intervention to both study arms, or there is specific reporting of suspected contamination in the trial report;

  • or unclear risk of bias.

(11) Other bias

We have considered any other potential sources of bias in the study, including whether recruitment was equal in both arms of cluster-randomised trials, and assessed these as:

  • low risk of bias;

  • high risk of bias;

  • or unclear risk of bias.

Measures of treatment effect

Dichotomous data

All data were entered into RevMan 5.2.5 and SPSS 20 for analysis. For dichotomous data, we have presented risk ratios (RR) with 95% confidence intervals. Analysis was conducted on the logged risk ratio, and then converted back to risk ratios for presentation purposes. In this update, smoking cessation outcomes have been converted from an 'odds ratio' for continued smoking, to a 'RR' for quitting, in line with other Cochrane Tobacco Group reviews. Therefore, an average RR > 1 in smoking cessation outcomes are positive in this review. Where less outcome events are desirable (e.g. preterm births, low birthweight infants, mean cigarettes per day), an average RR < 1 is a positive outcome. Analysis tables are labelled accordingly.

For two of the binary outcomes, abstinence in late pregnancy and perinatal deaths, zero cell counts for events in both the treatment and control groups were evident for one study each. The affected studies were Olds 1986 (abstinence in late pregnancy) and Valbo 1996 (perinatal deaths). This is problematic because the formula for calculating relative risk effect sizes requires non-zero cells (i.e., the numerator cannot be zero). Whilst RevMan 5.2.5 automatically corrects for zero events in one group, a manual 'fix' is required when both groups have zero events. The solution as recommended by the Cochrane statistician peer reviewer was to enter the values as zero in the analysis, which means the effect sizes are not estimable and those studies are effectively excluded from those analyses. The affected analyses are Analysis 9.1 for Olds 1986 and Analysis 1.16 and Analysis 11.15 for Valbo 1996. For all three of these affected analyses, the initial set of relevant studies was two; the result is that no pooled effect could be calculated because instead of two effect sizes we only have one effect size for each of these analyses. These instances are clearly marked in the results section.

Continuous data

For continuous data, we used the mean difference (MD) if outcomes were measured in the same way between trials (e.g. birthweight). We used the standardised mean difference (SMD) to combine trials that measured the same outcome, using different methods (e.g. biochemically-validated smoking reduction).  

Where standard errors (SE) were reported instead of standard deviations (SD), we used the RevMan calculator to calculate the effect size estimate. In one study, the SD was calculated from the SE. Where no SDs or SEs were reported, we estimated the mean SD from available studies, as recommended in the Cochrane Handbook 16.1.3.1 (Higgins 2008). The mean birthweight SD was calculated from 13 studies with available SDs (mean SD 578), and imputed for six studies. The mean cigarettes per day SD was calculated from 14 studies with available SDs (mean SD 6.5), and imputed for five studies.

Unit of analysis issues

There are good reasons for considering random allocation of midwives, clinics, health educators, hospitals, general practitioners, or antenatal classes to intervention or comparison group, rather than random allocation of pregnant women. It may be difficult for pregnancy care providers to treat women differentially according to the intervention or usual care protocol, and not to introduce co-interventions in one or other groups (contamination). As women within a cluster are more likely to be similar to one another, and less like the women in another cluster, outcomes from cluster-randomised trials were adjusted for the intra-cluster correlation for the data to be included in this review.  Adjusting for the clustering of studies means that cluster trials could be analysed in the same models as individual randomised trials.

Adjustment for cluster randomisation was conducted using a reported intra-cluster correlation (ICC) if available, and if not, a range of ICCs (from 0.003 to 0.20) was assumed and a sensitivity analysis conducted as recommended by (Merlo 2005). The results of the sensitivity analyses showed no substantial difference between the different ICCs (RRs were the same to at least three decimal places across ICC calculations). As such, for studies in which an ICC was not reported, an ICC value of 0.10 was used for the primary analysis and the cluster trials were included by adjusting the SEs (reported ICCs were used where available). The methods used for individual studies are reported in the Characteristics of included studies and Table 2. The adjustment involved reducing the size of each trial to its ‘effective sample size’ by dividing the sample size by the 'design effect', where the design effect is equal to 1 + (m – 1) × ICC,  and m is the average cluster size (see Section 16.3.4 of the Cochrane Handbook, Higgins 2008).

Table 2. Cluster-randomised trial adjustment details
ICCTrial IDTimingTiming codeOutcome descriptionOutcome codeMean cluster sizeNo. of clustersSample sizeCeased smoking %Continued smoking %Ceased smoking nContinued smoking nICCBetween cluster varIF(int)IF(comp)Effective sample size, denominatorEffective sample size, continue Effective sample size, ceased  
      mimcciccninci%c%i%c%icicrircs^2c  nincicORicORRR
0.003                                 
  Campbell 20062nd or subsequent visit0 171.062.5111178168810.56.489.593.682446996440.0030.003 1.211.186455815785440.58368371.721.641
  Hajek 2001Birth0 15.96.7928654557522.020.078.080.0120115425.14600.0030.003 1.011.017537.1565.4419452.30.8861181131.131.1
  Haug 1994 0 122125622529818.258.1681.7591.84468206900.0030.003 1.001.0025198205900.3984682.512.237
  Kendrick 199536/40 gest0 127.836.8323288811775.96.194.193.95272835.611050.0030.003 1.081.1182210637749981.03648650.970.967
  Lawrence 200330 wk gest0 117823413242895.61.7394.4498.271853062840.0030.003 1.051.023092832922780.2991753.343.211
  Lillington 1995late preg0 139.573.0227914643.0424.6656.9675.343436451100.0030.003 1.121.227112040900.43330302.311.745
  McLeod 200436/40 gest0 16.2734.6151113696020.313.379.7186.671485552   1.0961.100635550470.6041371.65 
  McLeod 200436/40 gest0 197.51281086017.610.082.41901968954   1.0751.1001005583490.521851.92 
  McLeod 200436/40 gest0 17.6965.714232117712018.611.781.3688.333314144106   *1.10163109133960.57730131.73 
  Messimer 198932-36 weeks' gest0 110.9110.365.55.5605725.014.07585.9615845490.0030.003 1.031.03585544480.491582.041.781
  Moore 200224-28/40 gest0 110.5911.83646467875716.719.083.3380.981131445656130.0310.031 1.301.345235674354591.175871080.850.876
  Pbert 200436/40 gest0 163.67100.53219120120.011.080.089.03822152.8178.90.0030.003 1.191.301611551291380.49332172.031.822
                                  
0.05                                 
  Campbell 20062nd or subsequent visit0 171.062.5111178168810.56.489.593.682446996440.0500.050 4.504.081741691551580.58318111.721.641
  Hajek 2001Birth0 15.96.7928654557522.020.078.080.0120115425.14600.0500.050 1.251.284437.3447.7341358.20.88696901.131.1
  Haug 1994 0 122125622529818.258.1681.7591.84468206900.0500.050 1.051.0324095196870.3984482.512.237
  Kendrick 199536/40 gest0 127.836.8323288811775.96.194.193.95272835.611050.0030.003 1.081.1182210637749981.03648650.970.967
  Lawrence 200330 wk gest0 117823413242895.61.7394.4498.271853062840.0030.003 1.051.023092832922780.2991753.343.211
  Lillington 1995late preg0 139.573.0227914643.0424.6656.9675.343436451100.0500.050 2.934.60273215240.4331282.311.745
  McLeod 200436/40 gest0 16.2734.6151113696020.313.379.7186.671485552   1.0961.100635550470.6041371.65 
  McLeod 200436/40 gest0 197.51281086017.610.082.41901968954   1.0751.1001005583490.521851.92 
  McLeod 200436/40 gest0 17.6965.714232117712018.611.781.3688.333314144106   *1.10163109133960.57730131.73 
  Messimer 198932-36 weeks' gest0 110.9110.365.55.5605725.014.07585.9615845490.0500.050 1.501.47403930330.491052.041.781
  Moore 200224-28/40 gest0 11112646467875716.719.083.3380.981131445656130.0310.031 1.301.345235674354591.175871080.850.876
  36/40 gest0 1641013219120120.011.080.089.03822152.8178.90.0500.050 4.135.98463437300.493942.031.822
                                  
0.1                                 
  Campbell 20062nd or subsequent visit0 17163111178168810.56.489.593.682446996440.1000.100 8.007.15989687900.5831061.721.641
  Hajek 2001Birth0 167928654557522.020.078.080.0120115425.14600.1000.100 1.491.569365.2366.6285293.30.88680731.131.1
  Haug 1994 0 122125622529818.258.1681.7591.84468206900.1000.100 1.101.0622993187850.3984282.512.237
  Kendrick 199536/40 gest0 12837323288811775.96.194.193.95272835.611050.0030.003 1.081.1182210637749981.03648650.970.967
  Lawrence 200330 wk gest0 117823413242895.61.7394.4498.271853062840.0030.003 1.051.023092832922780.2991753.343.211
  Lillington 1995late preg0 139.573.0227914643.0424.6656.9675.343436451100.1000.100 4.858.2016189130.433742.311.745
  McLeod 200436/40 gest0 16.2734.6151113696020.313.379.7186.671485552   1.0961.100635550470.6041371.65 
  McLeod 200436/40 gest0 197.51281086017.610.082.41901968954   1.0751.1001005583490.521851.92 
  McLeod 200436/40 gest0 17.6965.714232117712018.611.781.3688.333314144106   *1.10163109133960.57730131.73 
  Messimer 198932-36 weeks' gest0 110.9110.365.55.5605725.014.07585.9615845490.1000.100 1.991.94302923250.49842.041.781
  Moore 200224-28/40 gest0 11112646467875716.719.083.3380.981131445656130.0310.031 1.301.345235674354591.175871080.850.876
  Pbert 200436/40 gest0 1641013219120120.011.080.089.03822152.8178.90.1000.100 7.2710.95261821160.493522.031.822
                                  
0.2                                 
  Campbell 20062nd or subsequent visit0 17163111178168810.56.489.593.682446996440.2000.200 15.0013.31525247480.583531.721.641
  Hajek 2001Birth0 167928654557522.020.078.080.0120115425.14600.2000.200 1.982.1372752692142150.88660541.131.1
  Haug 1994 0 122125622529818.258.1681.7591.84468206900.2000.200 1.201.1221088172810.3983872.512.237
  Kendrick 199536/40 gest0 12837323288811775.96.194.193.95272835.611050.0030.003 1.081.1182210637749981.03648650.970.967
  Lawrence 200330 wk gest0 117823413242895.61.7394.4498.271853062840.0030.003 1.051.023092832922780.2991753.343.211
  Lillington 1995late preg0 139.573.0227914643.0424.6656.9675.343436451100.2000.200 8.7015.4099570.433422.311.745
  McLeod 200436/40 gest0 16.2734.6151113696020.313.379.7186.671485552   1.0961.100635550470.6041371.65 
  McLeod 200436/40 gest0 197.51281086017.610.082.41901968954   1.0751.1001005583490.521851.92 
  McLeod 200436/40 gest0 17.6965.714232117712018.611.781.3688.333314144106   *1.10163109133960.57730131.73 
  Messimer 198932-36 weeks' gest0 110.9110.365.55.5605725.014.07585.9615845490.2000.200 2.982.87202015170.49532.041.781
  Moore 200224-28/40 gest0 11112646467875716.719.083.3380.981131445656130.0310.031 1.301.345235674354591.175871080.850.876
  Pbert 200436/40 gest0 1641013219120120.011.080.089.03822152.8178.90.2000.200 13.5320.9014101190.493312.031.822
                                  
Key:Outcome    Data givenSensitivity analysis  From formula in Merlo        * wt'd ave of IF in 2 intv arms         
                                  
                                  
                      
 ADDITIONAL OUTCOMES found 21/11/08                                
0.003                                 
  Eades 2012late preg0continued smoking for spontaneous quitters in late pregnancy2                   248146     
  Hajek 2001late preg0continued smoking for spontaneous quitters in late pregnancy21.21.6928611413522.020.064.953.3406374720.0030.003 1.001.002113.9134.87471.881.61940630.620.752
  Lillington 1995late preg0continued smoking for spontaneous quitters in late pregnancy238127.022762545.26310.6394.7489.37427722270.0030.003 1.111.3868184651652.1414200.470.495
  Pbert 2004late preg0continued smoking for spontaneous quitters in late pregnancy2273932817770.477.929.622.1576024170.0030.003 1.081.11756922151.48653540.670.903
  Polanska 2004late preg0continued smoking for spontaneous quitters in late pregnancy25.67.41055637100100005637000.0030.003 1.011.02553600#####5536#DIV/0!1
  Haug 19940-5 mo pp1maintained cessation at 0-5 mo pp122125622529818.258.1681.7591.84468206900.0030.003 1.001.0025198205900.3984682.512.237
  Lawrence 200310 days pp110 days pp117823413242898.03.592.096.526102982790.0030.003 1.051.023092832842730.41125102.43 
  Lillington 19950-5 mo pp1maintained cessation at 0-5 mo pp139.573.0227914625.3211.6474.6888.362017591290.0030.003 1.121.2271120531060.38918142.572.174
  McLeod 20044/12pp1maintained cessation at 0-5 mo pp186232117712015.810.884.289.228131491070.0030.003 1.021.011741181461060.64727131.55 
  Messimer 19890-5 mo pp1maintained cessation at 0-5 mo pp110.9110.365.55.560578.310.591.6789.475655510.0030.003 1.031.03585553501.294560.770.792
  Pbert 20040-5 mo pp1maintained cessation at 0-5 mo pp164101321912014.23.095.897.0861831950.0030.003 1.191.301611551541500.704751.421.403
  Polanska 20040-5 mo pp1maintained cessation at 0-5 mo pp114.928.810514914444.316.755.783.3662482.991200.0030.003 1.041.08143133801110.25263223.972.653
  Hajek 20016 mo pp2maintained cessation at 6-11 mo pp14.75.1928643144022.020.097.097.013134184270.0030.003 1.011.012426.3434.6413421.80.97913131.021.021
  Haug 19946-11 mo pp2maintained cessation at 6-11 mo pp122125622529818.258.1684.5288.783911213870.0030.003 1.001.0025198212870.69139111.451.379
  Pbert 20046-11 mo pp2maintained cessation at 6-11 mo pp164101321912014.72.595.397.5951821960.0030.003 1.191.301611551531510.516841.941.894
  Haug 199412-17 mo pp3maintained cessation at 12-17 mo pp122125622529818.258.1685.3292.86377215910.0030.003 1.001.0025198214910.4473772.242.056
  Polanska 200412-17 mo pp3maintained cessation at 12-17 mo pp120.536.210520518131.712.768.2987.2965231401580.0030.003 1.061.111941641321430.31461213.192.495
  Lawrence 200318 mo pp418 mo pp117823413242894.62.495.497.61573092820.0030.003 1.051.023092832952760.5111471.96 
0.05                                 
  Eades 2012late preg0continued smoking for spontaneous quitters in late pregnancy2                   248146     
  Hajek 2001late preg0continued smoking for spontaneous quitters in late pregnancy21.21.6928611413522.020.064.953.3406374720.0500.050 1.011.028112.7131.37370.011.61940610.620.752
  Lillington 1995late preg0continued smoking for spontaneous quitters in late pregnancy238127.022762545.26310.6394.7489.37427722270.0500.050 2.857.30273525312.141140.470.495
  Pbert 2004late preg0continued smoking for spontaneous quitters in late pregnancy2273932817770.477.929.622.1576024170.0500.050 2.302.8835271061.48625210.670.903
  Polanska 2004late preg0continued smoking for spontaneous quitters in late pregnancy25.67.41055637100100005637000.0500.050 1.231.32462800#####4628#DIV/0!1
  Haug 19940-5 mo pp1maintained cessation at 0-5 mo pp122125622529818.258.1681.7591.84468206900.0500.050 1.051.0324095196870.3984482.512.237
  Lawrence 200310 days pp110 days pp117823413242898.03.592.096.526102982790.0030.003 1.051.023092832842730.41125102.43 
  Lillington 19950-5 mo pp1maintained cessation at 0-5 mo pp139.573.0227914625.3211.6474.6888.362017591290.0500.050 2.934.60273220280.389742.572.174
  McLeod 20044/12pp1maintained cessation at 0-5 mo pp186232117712015.810.884.289.228131491070.0500.050 1.331.2413397112870.64721111.55 
  Messimer 19890-5 mo pp1maintained cessation at 0-5 mo pp110.9110.365.55.560578.310.591.6789.475655510.0500.050 1.501.47403937351.294340.770.792
  Pbert 20040-5 mo pp1maintained cessation at 0-5 mo pp164101321912014.23.095.897.0861831950.0500.050 4.135.98463444330.704211.421.403
  Polanska 20040-5 mo pp1maintained cessation at 0-5 mo pp1152910514914444.316.755.783.3662482.991200.0500.050 1.702.39886049500.25239103.972.653
  Hajek 20016 mo pp2maintained cessation at 6-11 mo pp14.75.1928643144022.020.097.097.013134184270.0500.050 1.181.206363.9364.9353354.10.97911111.021.021
  Haug 19946-11 mo pp2maintained cessation at 6-11 mo pp122125622529818.258.1684.5288.783911213870.0500.050 1.051.0324095203850.69137111.451.379
  Pbert 20046-11 mo pp2maintained cessation at 6-11 mo pp164101321912014.72.595.397.5951821960.0500.050 4.135.98463444330.516211.941.894
  Haug 199412-17 mo pp3maintained cessation at 12-17 mo pp122125622529818.258.1685.3292.86377215910.0500.050 1.051.0324095205880.4473572.242.056
  Polanska 200412-17 mo pp3maintained cessation at 12-17 mo pp120.536.210520518131.712.768.2987.2965231401580.0500.050 1.982.761046671570.3143383.192.495
  Lawrence 200318 mo pp418 mo pp117823413242894.62.495.497.61573092820.0030.003 1.051.023092832952760.5111471.96 
0.1                                 
  Eades 2012late preg0continued smoking for spontaneous quitters in late pregnancy2                   248146     
  Hajek 2001late preg0continued smoking for spontaneous quitters in late pregnancy21.21.6928611413522.020.064.953.3406374720.1000.100 1.021.057111.3127.77268.121.61939600.620.752
  Lillington 1995late preg0continued smoking for spontaneous quitters in late pregnancy238127.022762545.26310.6394.7489.37427722270.1000.100 4.7013.60161915172.141120.470.495
  Pbert 2004late preg0continued smoking for spontaneous quitters in late pregnancy2273932817770.477.929.622.1576024170.1000.100 3.604.752316741.48616130.670.903
  Polanska 2004late preg0continued smoking for spontaneous quitters in late pregnancy25.67.41055637100100005637000.1000.100 1.461.64382300#####3823#DIV/0!1
  Haug 19940-5 mo pp1maintained cessation at 0-5 mo pp122125622529818.258.1681.7591.84468206900.1000.100 1.101.0622993187850.3984282.512.237
  Lawrence 200310 days pp110 days pp117823413242898.03.592.096.526102982790.0030.003 1.051.023092832842730.41125102.43 
  Lillington 19950-5 mo pp1maintained cessation at 0-5 mo pp139.573.0227914625.3211.6474.6888.362017591290.1000.100 4.858.20161812160.389422.572.174
  McLeod 20044/12pp1maintained cessation at 0-5 mo pp186232117712015.810.884.289.228131491070.1000.100 1.671.471068289730.6471791.55 
  Messimer 19890-5 mo pp1maintained cessation at 0-5 mo pp110.9110.365.55.560578.310.591.6789.475655510.1000.100 1.991.94302928261.294330.770.792
  Pbert 20040-5 mo pp1maintained cessation at 0-5 mo pp164101321912014.23.095.897.0861831950.1000.100 7.2710.95261825180.704111.421.403
  Polanska 20040-5 mo pp1maintained cessation at 0-5 mo pp1152910514914444.316.755.783.3662482.991200.1000.100 2.393.78623835320.2522863.972.653
  Hajek 20016 mo pp2maintained cessation at 6-11 mo pp14.75.1928643144022.020.097.097.013134184270.1000.100 1.371.412314.9311.7305302.50.979991.021.021
  Haug 19946-11 mo pp2maintained cessation at 6-11 mo pp122125622529818.258.1684.5288.783911213870.1000.100 1.101.0622993194820.69135101.451.379
  Pbert 20046-11 mo pp2maintained cessation at 6-11 mo pp164101321912014.72.595.397.5951821960.1000.100 7.2710.95261825180.516101.941.894
  Haug 199412-17 mo pp3maintained cessation at 12-17 mo pp122125622529818.258.1685.3292.86377215910.1000.100 1.101.0622993195860.4473472.242.056
  Polanska 200412-17 mo pp3maintained cessation at 12-17 mo pp120.536.210520518131.712.768.2987.2965231401580.1000.100 2.954.52694047350.3142253.192.495
  Lawrence 200318 mo pp418 mo pp117823413242894.62.495.497.61573092820.0030.003 1.051.023092832952760.5111471.96 
0.2                                 
  Eades 2012late preg0continued smoking for spontaneous quitters in late pregnancy2                   248146     
  Hajek 2001late preg0continued smoking for spontaneous quitters in late pregnancy21.21.6928611413522.020.064.953.3406374720.2000.200 1.051.114108.8121.27164.631.61938570.620.752
  Lillington 1995late preg0continued smoking for spontaneous quitters in late pregnancy238127.022762545.26310.6394.7489.37427722270.2000.200 8.4026.20910992.141010.470.495
  Pbert 2004late preg0continued smoking for spontaneous quitters in late pregnancy2273932817770.477.929.622.1576024170.2000.200 6.208.50139421.486970.670.903
  Polanska 2004late preg0continued smoking for spontaneous quitters in late pregnancy25.67.41055637100100005637000.2000.200 1.922.28291600#####2916#DIV/0!1
  Haug 19940-5 mo pp1maintained cessation at 0-5 mo pp122125622529818.258.1681.7591.84468206900.2000.200 1.201.1221088172810.3983872.512.237
  Lawrence 200310 days pp110 days pp117823413242898.03.592.096.526102982790.0030.003 1.051.023092832842730.41125102.43 
  Lillington 19950-5 mo pp1maintained cessation at 0-5 mo pp139.573.0227914625.3211.6474.6888.362017591290.2000.200 8.7015.4099780.389212.572.174
  McLeod 20044/12pp1maintained cessation at 0-5 mo pp186232117712015.810.884.289.228131491070.2000.200 2.341.94766264550.6471271.55 
  Messimer 19890-5 mo pp1maintained cessation at 0-5 mo pp110.9110.365.55.560578.310.591.6789.475655510.2000.200 2.982.87202018181.294220.770.792
  Pbert 20040-5 mo pp1maintained cessation at 0-5 mo pp164101321912014.23.095.897.0861831950.2000.200 13.5320.9014101490.704101.421.403
  Polanska 20040-5 mo pp1maintained cessation at 0-5 mo pp1152910514914444.316.755.783.3662482.991200.2000.200 3.786.56392222180.2521743.972.653
  Hajek 20016 mo pp2maintained cessation at 6-11 mo pp14.75.1928643144022.020.097.097.013134184270.2000.200 1.741.823248.1241.3241234.20.979771.021.021
  Haug 19946-11 mo pp2maintained cessation at 6-11 mo pp122125622529818.258.1684.5288.783911213870.2000.200 1.201.1221088178780.69133101.451.379
  Pbert 20046-11 mo pp2maintained cessation at 6-11 mo pp164101321912014.72.595.397.5951821960.2000.200 13.5320.9014101390.516101.941.894
  Haug 199412-17 mo pp3maintained cessation at 12-17 mo pp122125622529818.258.1685.3292.86377215910.2000.200 1.201.1221088179820.4473162.242.056
  Polanska 200412-17 mo pp3maintained cessation at 12-17 mo pp120.536.210520518131.712.768.2987.2965231401580.2000.200 4.908.04422329200.3141333.192.495
  Lawrence 200318 mo pp418 mo pp117823413242894.62.495.497.61573092820.0030.003 1.051.023092832952760.5111471.96 
Key:Outcome    Data givenSensitivity analysis  From formula in Merlo                   
                                  
                                  
                                  
   Timing codes                             
   0late pregnancy                    
   10-5 mo pp                            
   26-11 mo pp                            
   312-17 mo pp                            
   418 mo pp                            
   Outcome codes                            
   1abstinence                            
   2relapse prevention for spontaneous quitters                

Dealing with missing data

Due to the nature of the intervention, there is a high likelihood that women withdrawing from the study or not providing a biochemical sample for analysis, without a 'plausible explanation' (e.g. miscarriage/fetal demise, moving out of the area or changed to another provider of care) are likely to be continuing smokers. Where sufficient information has been reported or has been supplied by the trial authors, we have re-included missing data from each treatment group in the analyses to comply with recommended outcome criteria assessment for smoking cessation trials (West 2005). Only data which were excluded for medical reasons (e.g. miscarriage or preterm birth) or moving from study site were not re-included in this review. We have indicated where an intention-to-treat (ITT) (or available case) analysis was carried out for the smoking cessation outcome in the published report, or adjusted for this review. These assessments and any adjustments are reported in the 'Risk of bias' tables (see incomplete outcome data). Where data could not be re-included, we conducted sensitivity analysis to determine the effect of inclusion of trials assessed as 'high risk' of attrition bias.

Assessment of heterogeneity

We examined levels of heterogeneity in all pooled analyses (Cochran 1954). We used the I² statistic to quantify heterogeneity (i.e., inconsistency) among the trials in each analysis (Higgins 2008) and Chi2 tests to assess the presence of significant variation amongst effect sizes (i.e., whether the observed effects are significantly different from chance) (Lipsey 2001; Higgins 2008). For the Chi2 tests, in addition to the P value, we report the Q-statistic calculated by the test and the degrees of freedom of the test.

We expected to find a substantial degree of heterogeneity given the breadth of types of interventions, which are broadly categorised as 'psychosocial' and the differences in comparisons. Therefore, we attempted to minimise heterogeneity in this update by reporting separate comparisons for each main intervention strategy (counselling, health education, feedback, incentives, and social support; and whether the intervention was provided as a specific smoking intervention or as part of a broader intervention to improve maternal health) and comparison type (usual care, less intensive intervention, or alternative intervention). Further, we grouped studies within each comparison according to whether the intervention was provided as a single, multiple or tailored intervention.

To indicate considerable statistical heterogeneity, we set a threshold of inconsistency of I2 > 75% and a Chi2 significance level of P < 0.05. Where considerable heterogeneity was evident, we did not present pooled results. We further explored heterogeneity by pre-specified secondary analysis identified during development of a logic model (see Figure 1 and section on Subgroup analysis and investigation of heterogeneity for a description).

Assessment of reporting biases

Concerns about publication bias have been raised after observations that research evaluations showing beneficial and/or statistically significant findings are more likely to be published than those that have undesirable outcomes or non-significant findings (Higgins 2008). If this phenomenon does occur, then reviews of a biased evidence base will draw biased conclusions. Unfortunately, it is difficult to assess publication bias because there is no way of knowing the extent of what has not been published.

As a result of these concerns, researchers have developed ways of estimating the extent to which there may be some publication bias in the evidence base. Funnel plots (scatter plots in which the effect size from individual studies are plotted against a measure of study precision) are a common method for assessing the possibility of publication bias. Ideally, the spread of effect sizes should be such that there is more scattering of effect sizes at the bottom of the plot, where there is less precision, with a narrowing of the scattering towards the top, where there is greater precision.

Following guidance (Sterne 2001; Higgins 2008), we produced a funnel plot of the RR for the primary outcome on the x-axis, and the SE of the log RR on the y-axis, for each of the main comparisons (Analyses 1 through 10). Only the funnel plots for 'counselling versus usual care' (Analysis 1.1, Figure 2) and 'counselling versus less intensive intervention' (Analysis 2.1, Figure 3) are shown, because the remaining comparisons had too few effect sizes to reliably detect asymmetry in the funnel plot. In the figures, the vertical line indicates the random-effects pooled effect size estimate. In the absence of publication bias, we would expect a roughly symmetrical distribution of effect sizes in the inverted funnel shape. Two review authors examined the plot for publication bias; under the assumption that publication bias is detectable in these funnel plots, we conclude that it is unlikely that publication bias has biased the findings of this review.

Figure 2.

Funnel plot of comparison: 1 Smoking cessation interventions: counselling vs usual care, outcome: 1.1 Abstinence in late pregnancy.

Figure 3.

Funnel plot of comparison: 2 Smoking cessation interventions: counselling vs less intensive intervention, outcome: 2.1 Abstinence in late pregnancy.

Data synthesis

We used the statistical methods described in the Cochrane Handbook (Higgins 2008). We adopted a random-effects approach using method of moments estimators. The comparison analyses and forest plots were generated in RevMan 5.2.5, and meta-regressions and other subgroup analyses (using an analog to the ANOVA) were conducted in SPSS 20.0 using macros developed by Wilson 2005. When examining statistical significance, P values greater than 0.05 were considered non-significant. Where only one study was included in the comparison, the outcomes are not displayed in a separate comparison table and are reported in text only in the results, and data used is displayed in Comparison 11 of 'all outcomes by main intervention strategy' (see Analysis 11.1 for primary outcome and subsequent analyses for secondary outcomes).

Effect sizes that were included in the subgroup analyses for the primary outcome (reported in Section 1.2 of the results) were checked for outliers. First, skewness and SE of the skewness were calculated for the primary outcome in SPSS. Skewness was considered to be statistically significant at the 0.05 level when the skewness value divided by its SE was greater than 1.96. Second, given that skewness was detected, we checked for univariate outliers, which were defined as effect sizes greater than two SDs above or below the unweighted mean.

A sensitivity analysis was conducted to test whether Winsorising the outliers (i.e. changing the value of the effect size estimate to the mean ± 2 SDs), which is recommended in Lipsey 2001, affected the pooled effect size estimates. The analyses on the Winsorised datasets were conducted in SPSS, while the unchanged datasets were analysed in RevMan.

There was no substantial difference between pooled effect size estimate for the primary outcome when outliers unchanged (risk ratio (RR) 1.45, 95% confidence interval (CI) 1.27 to 1.64) and pooled effect size estimate with outliers Winsorised (RR 1.44, 95% CI 1.27 to 1.63).

Multivariate outliers of the primary outcome (i.e. abstinence in late pregnancy) were also explored using the predictor variables main intervention strategy (counselling, feedback, incentives, and social support, with health education and the one study with 'other' intervention type as the reference category). As recommended by Tabachnick 2001, the Mahalanobis distance of each study was compared to the Chi2 critical value of 18.47 (based on P < .001 and df =4). The Mahalanobis distance of none of the studies exceeded this value. Therefore, no multivariate outliers were identified for the primary outcome in terms of intervention strategy.

For the comparison analyses (conducted in RevMan and reported in Section 1.1 of the Results), we used the raw (i.e. not Winsorised) effect sizes in the analyses. This is because the subsets of studies are typically too small to reliably detect outliers.

The number needed to treat for benefit (NNTB) (Altman 1998) was calculated to give an approximation of how many women would need to receive the intervention for one of them to avoid an adverse outcome. We used the Visual Rx programme (Cates 2008) and based the computation on the random-effects pooled odds ratio effect size calculated in RevMan 5.2.5. We used the odds ratio rather than the risk ratio as this is invariant to whether the outcome is presented as a beneficial or adverse outcome (Cates 2002).

Subgroup analysis and investigation of heterogeneity

Investigation of heterogeneity is critical in such a large review that includes many different types of interventions and comparisons. It is possible that there are significant differences between subgroups of studies based on characteristics of the interventions, participants, comparisons, study bias etc, as outlined in Figure 1. In the section on Assessment of heterogeneity above, we described how we identified the presence or absence of heterogeneity; in the current section, we describe how we attempted to identify the main sources of variability in the effect size estimates, that is, to attempt to explain inconsistency across studies. We therefore explored how the observed effectiveness differs under different conditions.

Subgroup analyses

Where subgroup analyses were possible for the primary outcome, they were conducted on the whole dataset in SPSS 20 using an adapted ANOVA test. Ideally, the results of the subgroup analyses should produce a non-significant within-group heterogeneity statistic (i.e. the P value for QW should be > 0.05) to indicate that the effect sizes within a group are statistically similar to each other. If the subgroups are significantly different from each other, then the between-group heterogeneity statistic will be significant (i.e. the P value for QB will be < 0.05). If the between-group heterogeneity statistic QB is not statistically significant, then the proposed subgroup variable does not significantly explain differences between the effect sizes.

Two investigations of heterogeneity required meta-regression analyses. These were (1) a model that included two indicators of the difference in intensity of the intervention and control conditions and (2) a model that included both self-help manuals and telephone support as predictors. Meta-regressions were conducted in SPSS 20 using an adapted regression analysis. The overall fit of the regression model is indicated by two statistics: QM and QR QM is the variability associated with the regression model, while QR is the random error variability (that which is not accounted for by the model). A significant QM suggests that significant variation in the effect size distribution has been explained by the model, and is therefore desired. A significant QR , on the other hand, suggests that variability beyond that explained by the model remains, and is thus not ideal (Lipsey 2001).

Subgroup analyses for the primary outcome

We considered both clinical and statistical heterogeneity in the dataset. For the primary outcome, we did not calculate an overall pooled effect size for all intervention types versus all comparison types because clinical heterogeneity makes the overall effect size difficult to interpret. Instead, we focused our analysis of the primary outcome on subgroup analyses, which statistically test the significance of differences between groups, and trends in the pooled effects for different subgroups. The following variables were included in subgroup analyses conducted in SPSS 20 for the primary outcome of smoking abstinence in late pregnancy.

  1. Main intervention strategy (counselling, health education, incentives, feedback, social support, or other).

  2. Comparison type (usual care, less intensive interventions, or alternative interventions).

  3. Biochemically validated versus self-report outcomes.

  4. Intensity of the intervention (duration and frequency).

  5. Features of the intervention (self-help manuals and telephone support).

  6. Socio-economic status of the participants.

  7. Newly included studies in this review update.

It is important to note that the subgroup analyses described below do not take into account interactions in the data. For example, the models do not include both intervention type and comparison type in the same model, so we did not test how these factors might interact. Whilst this is a limitation of the analyses presented, we feel that there is still value in determining overall trends across the dataset. Firstly, this allows better comparison with previous versions of the review, for which the review had not separated the studies by comparison. Secondly, it allows us to consider whether what the corpus of studies looks like and whether there are trends across all of the studies. Throughout, we have distinguished between statistical heterogeneity and conceptual (or clinical) heterogeneity, and we hope that these subgroup analyses help to explore these different types of variation more thoroughly. We also note that in future updates of the review, we hope to be able to incorporate the increasingly popular methods of network meta-analysis to better address all of these issues.

Heterogeneity in the secondary outcomes

For most secondary outcomes, we did not calculate an overall pooled effect but instead focused on comparisons within clinically homogeneous subsets. However, for infant outcomes, we calculated overall pooled effect sizes for all intervention types versus all comparison types, for two reasons. Firstly, there was less extreme clinical heterogeneity in terms of intervention strategy in the infant outcomes. Secondly, as a primary objective of this review is to determine whether psychosocial interventions to support women to abstain from smoking in pregnancy have an impact on infant and maternal health outcomes, and large numbers are needed to detect relatively rare events, the pooled infant outcomes are informative. The overall pooled effect size estimates demonstrate the relationship between being randomised to a smoking cessation intervention and birth outcomes only, rather than the effectiveness of any particular intervention strategy.

Due to the small number of studies reporting the secondary outcomes, we were limited in the range of subgroup analyses (i.e. tests for statistical heterogeneity) that we could conduct. As such, comparisons for the secondary outcomes were limited to description of pooled effect sizes for the subgroups, rather than statistical tests of between-group differences.

Descriptions of trends across studies

To gain a greater understanding of key issues that we were not able to synthesise statistically, we present narrative summaries of the intervention effectiveness for dissemination trials; intervention effectiveness by ethnicity of the participants; and other participant characteristic analyses reported by study authors.

Sensitivity analysis

Concerns have been raised about whether clinical trial efficacy will translate to clinical effectiveness when implemented in healthcare practice (Walsh 2000). To determine whether effectiveness studies (defined as those assessing the implementation of an intervention that uses existing service providers) demonstrate a beneficial outcome in the absence of efficacy trials (those provided by dedicated research staff), we conducted a sensitivity analysis with efficacy trials excluded. The pooled effect size estimate, 95% confidence interval, and I2 value of the effectiveness-only studies was then compared with the overall pooled effect size estimate and its precision and I2 value.

A number of potentially significant factors were identified during data extraction and coding of the trials (e.g. where 'counselling' was provided by a video-tape rather than in person; where 'counselling' included optional provision of nicotine replacement therapy or incentives etc.). The studies with these characteristics were highlighted and sensitivity analyses conducted for these studies, and the effect that removing them had on the remaining studies in the comparison.

Assessment of risk of bias across studies

Assessment of the risk of bias across studies was conducted through subgroup analyses in SPSS 20 using an adapted ANOVA test. We used subgroup analyses rather than an elimination approach to sensitivity analysis for two reasons. Firstly, the subgroup analysis allows us to test whether high or low risk of bias studies have statistically different pooled effect sizes. Secondly, we included the 'unclear risk of bias' studies as a subgroup in the analyses, which allows us to check for missing data problems. For some of the risk of bias types, many of the studies did not report sufficient information to be able to assess the potential risk of bias. Through the subgroup analysis, we could test whether there was a systematic difference between poorly reported studies and those with assessable risk of bias.

We conducted risk of bias analyses for the following bias types on the primary outcome.

  • Random sequence generation selection bias.

  • Allocation concealment selection bias.

  • Incomplete outcome data attrition bias.

  • Selective reporting bias.

  • Detection bias (biochemical validation of abstinence).

  • Blinding of participants and personnel.

  • Blinding of outcome assessment.

  • Incomplete implementation.

  • Equal baseline characteristics in study arms.

  • Contamination of control group.

  • Other bias.

Due to the small numbers of effect size estimates for the 16 secondary outcomes for which we calculated effect size estimates, very few subgroup analyses by risk of bias type were possible. Only four of the outcomes had sufficient data to be analysed in terms of only one or two of the 12 possible risk of bias types. Given this, we did not conduct risk of bias analyses for the secondary outcomes. However, where possible we reported the average RR for studies assessed as having a high and low risk bias.

Results

Description of studies

Results of the search

The original version of this review included a total of 19 studies identified up until 1993 included as separate reports in the Pregnancy and Childbirth CD Rom: behavioural strategies for reducing smoking (n = 9) (Lumley 1995a); counselling for reducing smoking in pregnancy (n = 1) (Lumley 1995b); advice as a strategy for reducing smoking (n = 6) (Lumley 1995c); and feedback as a strategy for reducing smoking (n = 3) (Lumley 1995d).

Following publication of a protocol in 1998, a search was conducted by the Pregnancy and Childbirth Group for the second update of the review published in The Cochrane Library in 1999.  This update included a total of 44 trials: 37 trials including 16,916 women providing data on smoking cessation and over 800 women in five trials of relapse prevention (Lumley 1999). 

The third update in 2004 was based on a search until July 2003 conducted by the Pregnancy and Childbirth Group, the Tobacco Addiction Group Trials Register and a search of MEDLINE, Embase, PsycLIT and AustHealth.  A total of 65 trials were included involving over 20,000 women: 48 trials provided data on smoking cessation, six additional cluster trials involving over 7500 women were not included in the meta-analysis (Lumley 2004).

In the fourth update, published in 2009; a search from January 2003 to June 2008 identified 898 reports which were screened, the full text of 35 reports were reviewed and a total of 73 studies, involving over 20,000 women, were included (72 provided outcome data): 56 randomised and quasi-randomised trials and nine cluster-randomised trials provided primary outcome data for this update (Lumley 2009).

In this fifth update of the review, we screened 2030 abstracts (in addition to the search of the Pregnancy and Childbirth Group's Trials Register) and reviewed the full text of 64 reports.  We identified 16 new studies meeting the inclusion criteria.  As a result of a change in the inclusion criteria we excluded 13 studies from the previous version of the review, including nine quasi-randomised trials, as well as four randomised controlled trials of pharmacological interventions which are now included in a separate review (Coleman 2012b). These are listed in Characteristics of excluded studies. We also included four studies that had been previously excluded (three cluster trials and one abstract report of a trial), as well as nine studies that did not report any outcomes which could be used in meta-analyses, and which are reported in a separate table. We combined two reports of relapse prevention (Ershoff 1995; Secker-Walker 1995) as ‘Associated References’ to the primary papers reporting smoking cessation (Ershoff 1989; Secker-Walker 1994), and another paper which did not report any usable outcomes (Solomon 1996) as an 'Associated reference' to the primary report (Secker-Walker 1998).  A total of 77 randomised controlled trials, involving over 29,000 women with relevant outcome data, were included in the meta-analysis for this report (primary outcome data for 21,948 women participating in 70 trials and secondary outcome data only for a further 7404 women participating in seven trials). A further nine without outcomes are included but results summarised in Table 1, making a total of 86 studies included in this update. See Figure 4 for summary of search results.

Figure 4.

Search flow chart.

Included studies

Participants

Over 29,000 pregnant women participating in 77 trials with outcomes included in the meta-analysis were assessed as current or recent ‘smokers’ at recruitment. The criteria used to assess a woman as a ‘smoker’ varied substantially between trials, and are detailed for each study in the Characteristics of included studies table. There were 1740 women who reported they had 'spontaneously quit' smoking when they became pregnant, and had outcomes reported separately from women who continued to smoke. In one study only one third of the study population smoked commercial cigarettes, while two thirds chewed traditional or commercial smokeless tobacco (Patten 2009).

Participants were generally healthy pregnant adult women over 16 years of age, with 19 trials explicitly excluding women with medical or psychological complications. The majority of trials (n = 47) included women categorised as having low socio-economic status; 43 of these measured the primary outcome. Most trials included women over 16 years of age, with only two trials explicitly targeting young women under 20 years (Albrecht 1998; Albrecht 2006) and one study including women over 15 years of age (Donatelle 2000). Four trials were specifically targeted towards women with ‘psychosocial risk factors’ (Graham 1992; Belizan 1995; Albrecht 1998; El-Mohandes 2011) and two trials were conducted among women requiring methadone treatment for opioid addiction (Haug 2004; Tuten 2012). Most trials recruited women at the first antenatal clinic visit and during the second trimester of pregnancy, excluding women in the last trimester due to limited time remaining to receive the intervention. However, four trials were explicitly targeted towards women who continued to smoke in late pregnancy ('heavy smokers') (Valbo 1994; Valbo 1996; Stotts 2002; Stotts 2009). Seven studies included mainly women belonging to an ethnic minority population (Graham 1992; Lillington 1995; Gielen 1997; Manfredi 1999; Malchodi 2003; El-Mohandes 2011; Ondersma 2012). Two trials were conducted in aboriginal communities (Creative Spirits 2013) among Aboriginal women in Australia (Eades 2012) and Alaskan Native women the US (Patten 2009), and one trial included more than 40% Maori women in New Zealand (McLeod 2004). Twenty-eight studies explicitly excluded women who were not able to speak English (n = 26), Danish (Hegaard 2003) or Swedish (Hjalmarson 1991). In eight studies access to a telephone or video recorder was required for participation in the study. In two studies, women using nicotine replacement therapy were excluded (Malchodi 2003; Tuten 2012).

Interventions

Of the studies which had outcomes included in the meta-analysis (n = 77/86), the main intervention strategies were categorised as counselling (n = 48), health education (n = 7), feedback (n = 7), incentives (n = 4), and social support (n = 10). In one study the intervention was classified as 'intensive dissemination' as both arms received the same counselling intervention, with only the dissemination differing (Campbell 2006), and is therefore reported as a separate comparison. In seven studies, the primary aim of the study was to improve maternal health, which included a smoking cessation component of counselling (El-Mohandes 2011); feedback (Reading 1982; LeFevre 1995) and social support (Olds 1986; Belizan 1995; Bullock 1995; Bullock 2009). These studies are reported as separate comparisons and only smoking outcomes are included, as there is potential for other aspects of these interventions to impact on birth outcomes.

One trial was designed exclusively for women who had spontaneously quit smoking (Lowe 1997), and 11 trials included a relapse prevention component for women who had spontaneously quit. Interventions which were provided only during the postpartum period were excluded from this review, though many interventions during pregnancy continued support into the postpartum period and measured postpartum outcomes.

Smoking cessation interventions implemented during pregnancy differ substantially in their intensity, their duration, and the people involved in their implementation. In 31/77 studies the intervention was coded as a single intervention, therefore the 'main intervention strategy' most accurately reflects the type of intervention. However in 33 studies the intervention was coded as 'multiple', where other components of the intervention were offered to all women. In 12 studies the intervention was coded as 'tailored' whereby different intervention components were offered and tailored to women's needs. For example, two trials offered optional nicotine replacement therapy as part of a counselling intervention (Hegaard 2003; Eades 2012), and one trial offered nicotine replacement therapy to both intervention and control participants (Patten 2009). Most counselling studies involved face-to-face contact, using a variety of strategies either alone or in combination (such as motivational interviewing, cognitive behavioural therapy, stages of change). Three trials with the main intervention strategy coded as counselling included a lottery chance for women who reported quitting (Sexton 1984; Walsh 1997; Parker 2007); five included support for peers (Donatelle 2000; Solomon 2000; Hajek 2001; Vilches 2009; Eades 2012) and three included support for partners to quit (Thornton 1997; Vilches 2009; Eades 2012). The duration and frequency of the intervention also varied considerably, as illustrated in Figure 5 and Figure 6.

Figure 5.

Duration of contact for each condition by publication year.

Figure 6.

Frequency of contact for each condition by publication year.

Thirteen of the counselling interventions involved telephone counselling and in five of these studies all counselling was provided via telephone (Ershoff 1989; Bullock 1995; Solomon 2000; Stotts 2002; Rigotti 2006), and one had only brief additional face-to-face contact (Bullock 2009). Twenty-six studies included self-help manuals as part of the intervention, and in five studies there was a brief introduction to the manuals (less than five minutes) and the intervention was therefore coded as counselling (Ershoff 1989; Messimer 1989; Price 1991; Valbo 1994; Moore 2002), with sensitivity analysis conducted to assess the independent effect of these five studies. In six studies the intervention was provision of a video alone (Secker-Walker 1997; Cinciripini 2000), with a brief intervention (Price 1991) or as part of a counselling intervention (Walsh 1997; Manfredi 1999; Windsor 2011), and these were also coded as counselling as the videos included stories from women. Five studies included use of computers in the intervention, three of which were part of another main strategy (Lawrence 2003; Vilches 2009; Ondersma 2012); one which included interaction with a pregnancy care provider and was therefore coded as counselling (Tsoh 2010) and another in which the computer-generated messages were the only intervention and was therefore coded as health education (Strecher 2000). In one study the provision of the self-help manual was the only intervention (Hjalmarson 1991), and was therefore coded as health education only as there was no explicit personal component to the interaction. One study provided a mailed audiotape and self-help manual only (Petersen 1992) and one study provided only automated text-messaging (Naughton 2012); these were coded as health education, as there was no clear personal component. Three other studies that reported the intervention consisted of advice to quit only, either in person (Donovan 1977; Lilley 1986) or by post (Burling 1991) were coded as health education.

Five dissemination trials were identified, carried out in Australia (Lowe 2002; Campbell 2006) and the US (Manfredi 1999; Pbert 2004; Windsor 2011), two of which reported only dissemination outcomes (Manfredi 1999; Lowe 2002) and not the primary outcomes of abstinence in late pregnancy, therefore outcomes not able to be included in the meta-analysis are reported in Table 1. In 26 studies the intervention was provided by staff involved in routine pregnancy care (coded as effectiveness studies), and in 43 studies the intervention was provided by dedicated research project staff (coded as efficacy studies), or via automated technology (n = 8), (coded as unclear).

Comparisons

Women in the control arms in 44 of the 77 trials received information about the risks of smoking in pregnancy and were advised to quit as part of 'usual care'. In 16 of these 44 trials the comparison/control group was described as receiving 'usual care' without specifying further what constituted usual practice (at a particular time and in a particular setting) with respect to advice and assistance. In 31 trials the comparison group received some kind of 'less intensive' intervention, which included studies where a dedicated research team consistently provided what they considered to be 'usual care' for women in the comparison group. In two studies the comparison group received an 'alternative intervention', which was categorised as having the same intensity as the intervention group. One was a counselling intervention using cognitive behavioural therapy compared with traditional health education (Cinciripini 2010) and another compared provision of incentives, contingent or not contingent on smoking status (Heil 2008). As expected, the intensity of interventions and controls has increased over time, as indicated by the change in duration (Figure 5) and frequency of contact during the interventions (Figure 6).

Setting

Included trials were conducted between 1976 and 2012 and almost all trials were conducted in high-income countries. This includes the USA (57), Canada (1), the UK (13), Norway (3), Sweden (1), Holland (1), Spain (1), Australia (5), and New Zealand (2). Only two trials have been conducted in middle-income countries: one trial was conducted in four Latin American countries (Argentina, Brazil, Cuba and Mexico) (Belizan 1995), and the other in Poland (Polanska 2004). Neither trial had biochemically validated smoking outcomes. Most trials of interventions to support pregnant women were conducted in public hospitals or community antenatal clinics.

Outcomes reported
Primary outcomes

Sixty randomised controlled trials and 10 cluster-randomised trials reported the primary outcome measure of smoking abstinence in late pregnancy, up to and including the period of hospitalisation for birth (21,948 women), and in 49 trials (including seven cluster-randomised trials), the abstinence was biochemically validated. Nineteen studies reported whether there was a differential effect among women from different ethnic groups, socio-economic status, or other factors such as depression or partner smoking. Nine studies did not report any outcomes which could be included in meta-analysis and a summary table of outcomes for these studies is reported in Table 1.

Secondary outcomes included in meta-analysis

Fourteen trials reported continued abstinence in late pregnancy among women who had quit spontaneously before the intervention, one of which was a trial exclusively for women who had spontaneously quit, so did not also report the primary outcome (Lowe 1997).

Thirty-two trials reported continued abstinence in the postpartum period at zero to five months (n = 26), six to 11 months (n = 13), 12 to 17 months (n = 5) and 18 months and over (n = 2). Two of these trials did not have outcomes in late pregnancy as the assessment was undertaken at home after birth (Strecher 2000; Polanska 2004). Continued abstinence for baseline smokers and spontaneous quitters are combined in this outcome measure for some studies, with abstinence among baseline smokers only reported where available. The details of the outcomes for each study are reported in the Characteristics of included studies table.

Thirty-four trials reported various measures of smoking reduction in late pregnancy, including self-reported 'any reduction' (n = 7), self-reported reduction greater than 50% (n = 5), and biochemically validated reduction (n = 6). Two trials recorded both self-reported and biochemically validated reduction (Windsor 1985; Tappin 2005); in these cases we have included only the validated data in the analysis. Other reduction measures of reduced smoking included mean biochemical cotinine (n = 6) thiocyanate (n = 1), or mean cigarettes per day (n = 20). Three studies that reported smoking reduction did not include the primary outcomes of smoking abstinence (Donovan 1977; LeFevre 1995; Vilches 2009).

Nineteen trials reported mean birthweight, one of which had not reported any smoking cessation outcomes (Haddow 1991). Fourteen trials reported rates of low birthweight babies (less than 2500 g) and three reported rates of very low birthweight babies (less than 1500 g). Fourteen studies reported rates of preterm births less than 37 weeks' gestation (n = 14). Other trials reporting perinatal outcomes included: perinatal deaths (n = 4), stillbirths (n = 7), neonatal deaths (n = 4), and neonatal intensive care unit (NICU) admissions (4).

Other perinatal outcome measures reported included fetal growth (Cope 2003; Heil 2008), mean Apgar scores (Tuten 2012), and head circumference (Cope 2003).

Secondary outcomes included in narrative synthesis

Three trials measured mode of birth (Thornton 1997; Cope 2003; Tappin 2005).

Three trials measured breastfeeding initiation and/or duration (Panjari 1999; McLeod 2004 and an associated reference to Heil 2008) (Higgins 2010a).

Nineteen studies reported baseline psychological measures of interventions, three studies reported associations between smoking outcomes and psychological measures, and nine studies reported psychological outcomes.

No studies reported measures of family functioning. However three studies reported perceptions of partner (McBride 2004)) and peer support (Bullock 2009; Hennrikus 2010), and one study provided analysis of social networks (Stotts 2009).

Twenty-six trials addressed issues identified as important to women in a consultation for this review; with two associated references (Berg 2008; Washio 2011) to included studies (Rigotti 2006; Heil 2008), reporting effects of smoking cessation on maternal weight gain.

Seven studies explicitly included the views of women or community in development of the intervention; and 32 trials reported women’s views about the content or delivery of the intervention. Three studies reported measures of knowledge, attitudes or practice among pregnancy care providers (Haug 1994; Secker-Walker 1994; Lawrence 2003).

Five studies reported cost-effectiveness measures (Windsor 1985; Ershoff 1989; Dornelas 2006; Parker 2007; Heil 2008).

Two studies reported rates of women who reported an increase in smoking (adverse events) (Haug 1994; Tappin 2005).

Excluded studies

Seventy-five studies did not meet the eligibility criteria and were excluded from the review, for the following reasons:

  • design not adequately randomised (e.g. cohort studies, pre-post design, quasi-experimental designs);

  • primary population was not pregnant women or intervention was not primarily aimed at cessation during pregnancy (e.g. postpartum interventions, intervention for partners, non-pregnant women);

  • trial evaluated efficacy of pharmacological treatment with equal psychosocial support in both arms;

  • cluster-randomised trials with insufficient information (e.g. number of clusters) provided to enable adjustment for clustering.

See Characteristics of excluded studies for details.

Risk of bias in included studies

Allocation

Sequence generation was described and adequate in 35 trials. In 48 trials the sequence generation was not described or simply described as ‘randomised’ so it was unclear whether this was adequate or not. Three trials were included which had non-random sequence generation, such as allocation by medical record numbers and birthdate, as it was considered the risk of interference with this sequence is low. There are also many studies where the method of sequence generation was not reported. Quasi-randomised trials where there was a potential for interference, such as clinic attendance day or other quasi-randomised methods were excluded from this update of the review and the reasons are listed in the Characteristics of excluded studies table.

The method of randomisation was not described in sufficient detail to permit assessment of whether the allocation was concealed at the time of trial entry in 63 studies. In only 12 studies was the allocation adequately concealed and in 11 studies there was clearly no concealment of group allocation.

Equal baseline characteristics

As the sequence generation was not reported in the majority of trials, we assessed whether the baseline characteristics were equal and these were assessed as adequate in 37 studies, unclear (minor differences or not reported) in 33 studies, and inadequate or significant differences in 16 studies. Of the 48 trials with unclear sequence generation, 18 had equal baseline characteristics, seven had unequal baseline characteristics and in 23 there were some minor differences or the baseline characteristics were not reported.

Blinding

Very few trials had any blinding of participants or providers, as this is not practicable in delivering most psychosocial interventions. In 60 studies the participants and providers were clearly aware of group allocation, it was unclear in 15 studies, and in one study they were able to blind participants and/or providers to group allocation.

Blinding of the outcome assessment was rarely reported and was assessed as adequate in 11 studies, unclear in 74 studies, and inadequate in one study.

Incomplete outcome data

Withdrawals from the trials were common. When women were recruited at their first antenatal visit some participants had a miscarriage or a termination of pregnancy before the time when smoking behaviour was reassessed. These women were often excluded from outcome measurement, which means that important outcomes linked in observational studies to smoking exposure were not ascertained. Assessing smoking at 20 to 28 weeks instead of at 36 to 38 weeks would reduce the need to exclude women with particularly adverse outcomes, since their smoking status in mid-pregnancy would have been ascertained before preterm birth or a perinatal death had occurred. Others moved out of the area or changed to another provider of care. The latter was a common cause of attrition in those trials carried out among populations characterised by severe poverty and the receipt of special needs benefits such as Medicaid, or WIC (food program for women, infants and children) clinics.

In studies where there was longer-term follow-up, attrition was sometimes high; approximately half of the included studies had high levels of missing data (greater than 20%) for some outcomes.

All randomised women were included in analysis for the primary outcome (abstinence in late pregnancy) in 25 trials. In 41 trials, some women were excluded from the analysis due to miscarriage or pregnancy loss, or moving, and these were assessed as unclear risk of attrition bias as there are some associations with smoking. In 20 trials, primary outcome data were missing and were unable to be included in this review, and they were assessed as inadequate due to risk of attrition bias. Levels of attrition for each study and information about any intention-to-treat analysis have been reported in the 'Risk of bias' tables .

Selective reporting

It was not clear in many trials the extent of outcome data that were collected and therefore, unclear whether the outcomes were selectively reported in 42 studies. All primary outcomes were adequately reported in 30 studies, and 14 studies were assessed as inadequately reporting primary outcomes.

Other potential sources of bias

Detection bias from misclassification by self-report

Fifty-two trials reported biochemical validation of the primary outcome measure, smoking abstinence. In seven trials there was unclear or partial validation of smoking status. Twenty-seven trials measured smoking status by self-report and are included in this review as ‘high risk’ of bias. Later trials more often relied on a definition of smoking abstinence requiring biochemical validation. 

Implementation of intervention

Some studies reported process evaluation demonstrating challenges implementing the intervention and delivering it to all women (Walsh 2000). In 26 studies, process evaluation suggested that the majority of women received the intervention as planned, however 31 studies reported that many women had not received the intervention as planned and in 29 studies it was unclear or not reported.

Smoking cessation interventions implemented during pregnancy differ substantially in their intensity, their duration, and the people involved in their implementation. The timing of the final antenatal assessment of smoking status varied considerably between trials between the second and third trimester. This may have affected the amount of time the participants were exposed to the intervention (if it involved ongoing support), as well as the number of those lost to follow-up and measurement of perinatal outcomes.

Exposure of the control group to the intervention

Another problem with trials in this area can be 'contamination' or exposure of the control group to intervention components, particularly if the study is being implemented in a routine care setting. Fifty-eight trials were implemented by dedicated research staff or technology and were assessed as having a low risk of exposing the control group to the intervention. In 12 studies it was unclear, and in 16 studies the authors reported problems with exposure of the control group, or the intervention was provided by routine care providers and the study design was assessed as having a 'high risk' of control group exposure.

Other bias

No other risk of bias was suspected in 68 studies. However, in nine studies there were some other risks, such as unequal recruitment to study arms in cluster-randomised trials or financial conflicts of interest, and in nine studies it was unclear if there may be other risks of bias.

Change in 'usual care'

In many cases the comparison/control group was described as receiving 'usual care' without specifying further what constituted usual practice (at a particular time and in a particular setting) with respect to advice and assistance. It can be seen from Figure 5 and Figure 6 that current 'usual care' may be a more substantial intervention than the defined intervention in some of the earliest trials (for example, Baric 1976).

A summary of Risk of bias' assessments in the included trials is set out in Figure 7 and Figure 8.

Figure 7.

'Risk of bias' summary: review authors' judgments about each risk of bias item for each included study.

Figure 8.

'Risk of bias' graph: review authors' judgments about each risk of bias item presented as percentages across all included studies.

Effects of interventions

A total of 88 meta-analyses are reported in this review. Meta-analyses were conducted and are presented in data tables for a total of 11 comparisons involving 59 outcomes. Data for comparisons with only one study reporting an outcome are reported in text, but not displayed. In addition, eight non-prespecified meta-analyses conducted in Revman 5.2.5 were reported in text, to assess the effect of factors identified during data extraction and coding (e.g. where 'counselling' involved provision of a videotape only). The results of 21 meta-analyses conducted in SPSS 20 to assess risk of bias and sensitivity analyses are also reported in text and not reported in tables.

1. Primary outcome: Smoking abstinence in late pregnancy

1.1 Comparisons: Main intervention strategy compared with usual care, less intensive intervention, or an alternative intervention, and subgrouped by single, multiple or tailored components.

Table 3 presents a cross-tabulation of the main intervention strategies and comparison type, for studies that report the primary outcome. The large number of cells that have very few (i.e., n ≤ 2) or zero studies means that it is not appropriate to run an interaction analysis with these two variables. Therefore, the synthesis in this section was not achieved through meta-analytic subgroup analyses; rather, the synthesis is a description of trends in the weighted pooled effect size estimate for subsets of studies based on the intervention strategy, the comparison type, and the number of components in the intervention (single component, multiple components, and tailored components). As such, we cannot draw any conclusions about statistical differences between subsets of studies in this section.

Table 3. Cross-tabulation of main intervention strategy by comparison type, for studies reporting the primary outcome
 Comparison typeTotal
Usual careLess intensive interventionAlternative intervention
Main intervention strategyCounselling2816145
Health education3205
Feedback2204
Incentives2114
Social support28010
Other0101
Total3730269
1.1.1 Counselling versus usual care

In trials where the main intervention strategy was counselling and the control group received 'usual care', the difference between intervention and control groups was significantly different from zero (27 studies; average risk ratio (average RR) 1.44, 95% confidence interval (CI) 1.19 to 1.75), I2 = 55%, see Analysis 1.1.

In subsets of studies, the effect size estimate was significantly different from zero where counselling was combined with other strategies (11 studies; average RR 1.59, 95% CI 1.15 to 2.21), I2 = 45% or tailored to the needs of individual women (six studies; average RR 1.49, 95% CI 1.01 to 2.20), I2 = 75%, but the effect was unclear when counselling was provided as a single intervention (10 studies; average RR 1.12, 95% CI 0.89 to 1.42), I2 = 11%.

There was no significant difference in biochemically validated abstinence in late pregnancy in a single study where smoking cessation counselling was provided as part of a broader intervention to improve maternal health (El-Mohandes 2011) and the control group received usual care (RR 1.00, 95% CI 0.72 to 1.40). The analysis for this comparison is not displayed in a table as only one study met the criteria.

1.1.2 Counselling versus less intensive interventions

In trials where the main intervention strategy was counselling and the control group received a less intensive intervention, the effect size had borderline significance (16 studies; average RR 1.35, 95% CI 1.00 to 1.82), I2 = 74%, see Analysis 2.1. In subsets of studies, the effect size was significantly different from zero for the single trial (Walsh 1997) where counselling was tailored to individual needs (RR 2.39, 95% CI 1.03 to 5.56), and included lottery tickets for women who were abstinent from smoking, but there was no clear difference where counselling was provided alone (n = 5), or in combination with other strategies (n = 10).

1.1.3 Counselling versus alternative intervention

There was no significant effect in the single study (Cinciripini 2010) that compared one counselling strategy (CBT) to an alternative counselling intervention (traditional health education or motivational interviewing) (RR 1.15, 95% CI 0.86 to 1.53). The analysis for this comparison is not displayed in a table as only one study met the criteria.

Other counselling subset analyses (not displayed)

In two studies where counselling was provided as part of a tailored intervention that included optional nicotine replacement therapy and was compared with usual care (Eades 2012; Hegaard 2003), the effect was not significantly different from zero (average RR 1.63, 95% CI 0.25 to 10.50), I2 = 59%.

In two studies where 'counselling' involved only provision of a video tape (Secker-Walker 1997; Cinciripini 2000) compared with a less intensive intervention, the effect was unclear as it was not significantly different from zero and there was considerable heterogeneity (average RR 2.31, 95% CI 0.08 to 65.02), I2 = 78%, and the effect on the subgroup of 'single' counselling interventions compared with usual care continued to be borderline non-significant when these two studies were removed from the pooled results (average RR 1.52, 95% CI 0.99 to 2.34). The effect was not significantly different from zero in a single study (Price 1991), which provided brief advice (less than five minutes) in conjunction with provision of a video, compared with usual care (RR 3.94, 95% CI 0.45 to 34.41).

Five studies coded as counselling provided brief advice (less than five minutes) and a self-help manual (Ershoff 1989; Messimer 1989; Price 1991; Valbo 1994; Moore 2002). Four of these studies reported abstinence in late pregnancy and the combined effect was not significantly different from zero (average RR 1.28, 95% CI 0.79 to 2.07), I2 = 54%.

Four studies coded as counselling included peer and/or partner support as part of a tailored intervention (Solomon 2000; Hajek 2001; Vilches 2009; Eades 2012) compared with usual care, and the combined effect of two studies that reported abstinence in late pregnancy (Hajek 2001; Eades 2012) was not significantly different from zero (average RR 1.09, 95% CI 0.82 to 1.44), I2 = 0%.

Three studies coded as counselling (tailored) included support for partners to quit smoking (Thornton 1997; Vilches 2009; Eades 2012) compared with usual care, and two studies that reported abstinence in late pregnancy (Thornton 1997; Eades 2012) did not show a combined effect that was significantly different from zero (average RR 1.23, 95% CI 0.66 to 2.31), I2 = 0%.

Three studies coded as multiple or tailored counselling that included a lottery chance for women who reported abstinence (Sexton 1984; Walsh 1997; Parker 2007) had a combined effect that was significantly different from zero (average RR 1.98, 95% CI 1.61 to 2.42), I2 = 6%. Two studies that measured self-reported abstinence compared with usual care (Sexton 1984) and a less intensive intervention (Parker 2007) showed a significant effect (average RR 1.69, 95% CI 1.21 to 2.36), and the effect of the single study that reported biochemically validated abstinence (Walsh 1997) was also significantly different from zero (RR 2.39, 95% CI 1.03 to 5.56).

1.1.5 Health education versus usual care

For studies in which the main intervention strategy was health education and the control group received usual care, the pooled effect size estimate was not significantly different from zero (three studies; average RR 1.51, 95% CI 0.64 to 3.59), I 2= 28%, see Analysis 3.1. The effect size estimate was not significant in subsets of trials where health education was provided alone (n = 2) or in combination with other strategies (n = 1); or when the analysis was restricted to studies with biochemical validation of abstinence, see Analysis 3.2.

1.1.6 Health education versus less intensive interventions

The effect was not significantly different from zero in trials where health education was compared with a less intensive intervention (two studies; average RR 1.50, 95% CI 0.97 to 2.31), I2 = 0%, and there was little difference whether health education was provided alone (n = 1), or in combination with other strategies (n = 1), see Analysis 4.1.

Other health education subset analyses (not displayed)

Two studies coded as health education involved provision of self-help manuals with no additional advice (Hjalmarson 1991) or an audiotape (Petersen 1992) and the combined effect was not significantly different from zero (average RR 1.28, 95% CI 0.79 to 2.07), I2 = 7%. When these studies were removed from the health education subgroup, the combined effect of the remaining three studies (Lilley 1986; Burling 1991; Naughton 2012) was statistically significantly different from zero (average RR 1.93, 95% CI 1.01 to 3.69), I2 = 0%.

A single study coded as health education that provided advice via a computer (Strecher 2000), compared with a less intensive intervention reported an effect that was not significantly different from zero in abstinence at six weeks postpartum (RR 1.00, 95% CI 0.91 to 1.09).

The effect of a single study coded as health education that provided advice and motivational statements via text compared with a less intensive intervention (Naughton 2012), was not significantly different from zero (RR 1.59, 95% CI 0.68 to 3.73).

1.1.7 Feedback versus usual care

For the two trials where the main intervention was feedback, provided in combination with other strategies, and the control group received usual care (Valbo 1994; Cope 2003), the combined effect size estimate was significantly different from zero (average RR 4.39, 95% CI 1.89 to 10.21), I = 0%, see Analysis 5.1.

The effect of self-reported smoking abstinence in late pregnancy was not significantly different from zero in a single study that provided ultrasound feedback alone (with no smoking cessation advice) as part of a broader intervention to improve maternal health and usual care for the control group (Reading 1982) (RR 2.11, 95% CI 0.98 to 4.52). The analysis for this comparison is not displayed in a table as only one study met the criteria.

1.1.8 Feedback versus less intensive interventions

Two studies assessed the effectiveness of feedback compared with less intensive interventions. The effect size estimates of both studies - one in which feedback was provided alone (Bauman 1983) and one in which feedback was provided in combination with other strategies, for women still smoking in late pregnancy (Stotts 2009), were not significantly different from zero; (average RR 1.19, 95% CI 0.45 to 3.12), I = 49%, see Analysis 6.1.

1.1.9 Incentives versus usual care

There was no significant difference in rates of biochemically validated abstinence in the pooled results of two studies where the main intervention strategy was financial incentives and the control group received usual care (average RR 3.59, 95% CI 0.10 to 130.49). However, there was significant heterogeneity ( I2 = 82%) and interaction between the subgroups (Chi2 4.03, P = 0.04), so caution is needed considering the combined effect of these trials. The analysis included a trial of incentives (single intervention) (Tuten 2012) (RR 20.72, 95% CI 1.28 to 336.01) and a trial of 'low intensity' incentives (multiple intervention) provided with assistance of a computer program and counselling via a computerised program (Ondersma 2012) (RR 0.90, 95% CI 0.25 to 3.23), see Analysis 7.1.

1.1.10 Incentives versus less intensive or alternative interventions

The effect was significantly different from zero in the single trial where incentives were provided in combination with peer support and the control group received a less intensive intervention (Donatelle 2000) (RR 3.64, 95% CI 1.84 to 7.23). The analysis for this comparison is not displayed in a table as only one study met the criteria.

The effect was also significantly different from zero in the single study where the intervention group received incentives contingent on smoking status (single intervention), and the control group received an equally intensive alternative intervention of incentives which were not contingent on smoking status (Heil 2008) (RR 4.05, 95% CI 1.48 to 11.11). The analysis for this comparison is not displayed in a table as only one study met the criteria.

Another trial of incentives included a second comparison arm of non-contingent incentives (Tuten 2012), which demonstrated a significant effect (RR 18.21, 95% CI 1.33 to 294.43), although this effect size estimate was not included in the meta-analysis (only the comparison with the usual care condition was included in the meta-analyses in this review).

1.1.11 Social support versus less intensive interventions

The combined effect size estimate of six trials where the main intervention strategy included peer or partner (social) support and the control group received a less intensive intervention was not significantly different from zero (average RR 1.29, 95% CI 0.94 to 1.78), I = 18%, see Analysis 8.1. However, the effect was significantly different from zero in five trials which included peer support (average RR 1.49, 95% CI 1.01 to 2.19), I2 = 3%, see Analysis 8.2. In the single trial where the intervention involved partner support (McBride 2004), there was no significant effect in self-reported abstinence (RR 1.02, 95% CI 0.70 to 1.50).The analysis for this comparison is not displayed in a table as only one study met the criteria.

1.1.12 Social support as a component of a broader maternal health intervention versus usual care

The effect size was significantly different from zero in one study where tailored peer support was provided as part of a broader intervention to improve maternal health and compared with usual care (RR 1.83, 95% CI 1.22 to 2.73), see Analysis 9.1. A further study in which tailored peer support was provided as part of a broader intervention to improve maternal health and compared with usual care with biochemically validation smoking cessation (Olds 1986) had zero events in both study arms and the effect size estimate was therefore 'not estimable' in Revman 5.2.5. As such, we could not calculate a pooled effect for this comparison.

1.1.13 Social support as a component of a broader maternal health intervention versus less intensive intervention

There was no significant effect in two studies where telephone peer support was provided as part of a broader intervention to improve maternal health, and the control group received a less intensive intervention (average RR 0.80, 95% CI 0.46 to 1.39); see Analysis 10.1 and Analysis 10.2.

1.2 Subgroup analyses

The following subgroup analyses were conducted on the whole dataset using all studies for the primary outcome (smoking abstinence in late pregnancy) (see Analysis 11.1 for list of studies). These analyses were conducted in SPSS using Winsorised data.

1.2.1 Subgroup analysis 1: Main intervention strategy

Three of the main intervention strategy subgroups had pooled effect size estimates that were significantly different from a null effect, indicating that abstinence in late pregnancy was significantly greater in the treatment than in the control group for these strategies: incentives (four studies; average RR 2.95, 95% CI 1.55 to 5.63, I2 = 15%), feedback (five studies; average RR 2.08, 95% CI 1.23 to 3.50, I2 = 26%), and counselling (45 studies; RR 1.36, 95% CI 1.17 to 1.57, I2 = 0%). However, there was no significant difference between treatment and control groups in subgroup analyses of trials where the main intervention strategy was social support (10 studies; average RR 1.29, 95% CI 0.92 to 1.80, I2 = 0%), or health education (five studies; RR 1.50, 95% CI 0.90 to 2.51, I2 = 0%). There was not a significant between-group difference (QB (4) = 7.70, P = 0.10) and there was within-group homogeneity (as indicated by low I2 in each subgroup and non-significant Q-statistics for each subgroup; overall QW (64) = 57.86, P = 0.69). One study, Campbell 2006, was treated as missing from this analysis as the intervention type category was unclear.

1.2.2 Subgroup analysis 2: Comparison type

We conducted a subgroup analysis to test for differences in the pooled effect size estimate of studies grouped by their comparison type. As there were only two studies with alternative intervention comparators that also reported the primary outcome, we used a pooled estimate of the between-study variance (τ2) following the method described in Borenstein 2009. The results suggests that there is no statistically significant difference between effect size estimates grouped by comparison type (QB (2) = 1.53, P = 0.47). Studies with comparisons consisting of usual care comparisons had the highest pooled effect size estimate (37 studies; average RR 1.34, 95% CI 1.25 to 1.44), I2 = 53%, followed by less intensive interventions (30 studies, average RR 1.20, 95% CI 1.08 to 1.31), I2 = 64%, and the effect size estimate for studies with an alternative intervention comparisons was not statistically different from zero (two studies, average RR 1.26, 95% CI 0.98 to 1.53), I2 = 82%. Forest plot not shown. It should be noted that studies where the comparison group received only 'usual care' were also more likely to provide a low intensity intervention, as shown in Figure 5 and Figure 6, and discussed below.

1.2.3 Subgroup analysis 3: Biochemically validated versus self-report outcomes

Given concerns about the potential biases (e.g. social desirability bias) of self-report measures of smoking behaviours, we conducted a subgroup analysis comparing biochemically validated smoking abstinence and self-reported abstinence. The results suggest that there is no statistically significant difference between the two groups of effect sizes (QB (1) = 0.06, P = 0.80; QW (67) = 61.33, P = 0.67), and there was a similar pooled effect size estimate for biochemically validated outcomes (49 studies; average RR 1.43, 95% CI 1.22 to 1.67, I2 = 0%), compared to self-reported outcomes (20 studies; average RR 1.48, 95% CI 1.17 to 1.87, I2 = 11%). Although this does not help us to explain the significant heterogeneity in the dataset, it gives us greater confidence in combining self-report with biochemically validated outcomes in further analyses. One study, Thornton 1997, was treated as missing from this analysis as the use of biochemical validation was unclear.

1.2.4 Subgroup analysis 4: Intensity of the intervention

There was no significant difference between effect sizes estimates subgrouped according to the frequency of contact in the intervention (QB (5) = 8.88, P = 0.11); see Table 4 for the pooled effect size estimates by group. Moreover, there was no significant difference between effect sizes estimates subgrouped according to the duration of contact in the intervention (QB (5) = 5.43, P = 0.37); see Table 5 for the pooled effect size estimates by group.

Table 4. Intensity of intervention subgroup analysis - frequency of contact in intervention
  1. 1 = Single contact during/at time of routine pregnancy care visits (but not ‘usual care’) without strategies to quit; 2 = Single contact, outside of ‘routine’ pregnancy care with strategies to quit; 3 = 2-5 contacts to sustain motivation to stop smoking provided during/at time of routine pregnancy care visits; 4 = 2-5 contacts to sustain motivation to stop smoking provided outside of routine care; 5 = > 5 contacts to sustain motivation to stop smoking provided during/at time of routine care visits; 6 = > 5 contacts to sustain motivation to stop smoking provided outside of routine care. One study, Campbell 2006, was treated as missing from this analysis as the intervention frequency was unclear.

GroupMean ESSE-95%CI+95%CIPNI2 (%)
11.89.241.193.02.01619
20.94.180.661.35.75837
31.84.191.272.66.0080
41.31.140.991.73.05130
51.37.161.001.86.05100
61.48.091.231.77.00245
Table 5. Intensity of intervention subgroup analysis - duration of contact in intervention
  1. 1 = Less than 15 mins; 2 = 15-44 mins; 3 = 24 mins to less than 2 hours; 4 = 2 hours to less than 5 hours; 5 = 5 hours to less than 8 hours; 6 = 8 or more hours. Two studies, Campbell 2006 and Bauman 1983, were treated as missing from this analysis as the intervention duration was unclear.

GroupMean ESSE-95%CI+95%CIPNI2 (%)
11.42.161.051.93.02120
21.43.141.081.89.011718
31.48.121.171.88.00170
41.03.210.681.55.9080
51.29.200.871.91.2060
61.88.171.352.62.0085

To explore whether the difference in intensity between conditions was a significant predictor of the outcome, a meta-regression was conducted. The model included two predictor variables: the difference between the intervention and control group frequency of contact categorisations, and the difference between the intervention and control group duration of contact categorisations. The analyses indicated that neither the magnitude of the difference in duration nor frequency of contact significantly predicted the primary outcome (QM (2) = 0.17, P = 0.92; QR (65) = 63.14, P = 0.54;R 2 = 0.00).

1.2.5 Subgroup analysis 5: Features of the intervention (self-help manuals and telephone support)

A meta-regression with two dichotomous predictor variables - the use of self-help manuals and the availability of telephone support - was conducted. Of the studies that reported the primary outcome, 24 studies offered self-help materials to participants and 13 provided telephone support (three of these offered both). The analyses indicated that neither self-help materials (B = -0.14, SE = 0.13) nor telephone support (B = -0.14, SE = 0.15) significantly predicted the primary outcome (QM (2) = 1.83, P = 0.40; QR (67) = 63.54, P = 0.60;R 2 = 0.03).

1.2.6 Subgroup analysis 6: Socio-economic status (SES) of the participants

For the primary outcome of abstinence in late pregnancy, there was no significant difference between the two groups of studies with women categorised as 'low' or 'not low' SES (QB (1) = 0.11, P = 0.74). The pooled effect size estimate for interventions provided for women categorised as 'low' SES interventions was similar (44 studies; average RR 1.41, 95% CI 1.19 to 1.66, I2 = 1%), to those provided for women categorised as 'not low' SES (26 studies; average RR 1.47, 95% CI 1.21 to 1.79, I2 = 0%).

1.2.7 Subgroup analysis 7: Newly included studies in this review update

Of the 70 studies reporting smoking abstinence in late pregnancy outcomes, 50 came from studies in the previous review (Lumley 2009), while 20 were from new studies identified in the updated search. We conducted this subgroup analysis to address concerns that newer trials may have a reduced effect due to the increased information about the risks of smoking in pregnancy in the general population. Although effect sizes from the newly-included studies tended to be lower (20 studies; average RR 1.26, 95% CI 1.00 to 1.59, I2= 3%), than those from the previous version of the review (50 studies; average RR 1.50, 95% CI 1.30 to 1.73, I2= 0%), this difference was not statistically significant (QB (1) = 1.51, P = 0.22).

1.3 Description of trends in intervention effectiveness: dissemination trials (not displayed)

There were five dissemination trials, defined as trials where the intervention was provided at an organisational level and strategies were employed to influence the practice of pregnancy care providers (Manfredi 1999; Lowe 2002; Pbert 2004; Campbell 2006; Windsor 2011). The combined effect of three trials that reported abstinence in late pregnancy (Pbert 2004; Campbell 2006; Windsor 2011) was not significantly different from zero (average RR 0.96, 95% CI 0.37 to 2.50), I2 = 72%.

1.4 Description of trends in intervention effectiveness: ethnic and aboriginal participants (not displayed)

The synthesis in this section was not achieved through meta-analytic subgroup analyses; rather, the synthesis is a description of trends in the weighted pooled effect size estimate for subsets of studies based on ethnicity of the participants. As such, we cannot draw any conclusions about statistical differences between subsets of studies in this section.

The combined effect of five studies (four counselling trials, one incentives trial) among women predominantly from a minority ethnic group (African-American and/or Hispanic) that reported abstinence in late pregnancy was not significantly different from zero (average RR 1.08, 95% CI 0.83 to 1.40), I2 = 0%. Of those five trials, three were conducted with African-American women (Gielen 1997; El-Mohandes 2011; Ondersma 2012) (average RR 1.01, 95% CI 0.75 to 1.37), I2 = 0%. The effect size estimate in a single trial among African-American and Hispanic women (Lillington 1995) was not significantly different from zero (RR 1.97, 95% CI 0.70 to 5.50). A single trial of social support developed specifically for Hispanic women in this review (Malchodi 2003) did not demonstrate a significant effect size estimate (RR 1.12, 95% CI 0.61 to 2.06).

The combined effect for the two tailored counselling interventions provided for aboriginal women in Australia (Eades 2012) and Canada (Patten 2009) did not show a significant difference between treatment and control groups in rates of abstinence in late pregnancy (average RR 0.40, 95% CI 0.06 to 2.67), I2 = 0%.

1.5 Description of participant characteristic analyses reported by study authors

The following is a narrative synthesis of the findings of subgroup analyses reported by primary study authors.

Low socio-economic status (SES)

Of seven studies which reported sensitivity analysis by a measure of SES, four reported lower abstinence rates or a negative association with quitting among women with lower SES (Baric 1976; McLeod 2004; Pbert 2004; Rigotti 2006), two reported no significant difference (Ershoff 1989; Tappin 2005), and one study reported 4/5 successful quitters had not graduated from high school (Secker-Walker 1997).

Ethnicity or race

Of nine studies which reported outcomes or sensitivity analysis by ethnic status, one study reported the intervention was less effective among Hispanic and African-American women (Kendrick 1995), one study reported the intervention was less effective among Hispanic compared to African American women (Lillington 1995), three studies reported no difference in outcomes by race or ethnicity (Burling 1991; Strecher 2000; Dornelas 2006), and four studies reported higher quit rates among African-American and/or Hispanic women compared to other women (Petersen 1992; Windsor 1993; Pbert 2004; Parker 2007).

Depression

Two studies that reported outcomes by rates of depression reported a negative association between smoking abstinence and depression (Cinciripini 2000; Rigotti 2006).

Low social support

Three studies that reported measures of social support reported a negative association with low social support (e.g. single mothers) and quitting (Loeb 1983; Thornton 1997; Rigotti 2006).

Partner smoking

Of four studies reporting associations with partner smoking and abstinence in late pregnancy, two reported no significant difference (Rigotti 2006; Stotts 2009) and two reported a negative association (i.e. lower rates of quitting among women whose partners' smoked) (McLeod 2004; Polanska 2004).

1.6 Sensitivity analysis
1.6.1 Efficacy versus effectiveness trials

Given concerns about whether clinical trial efficacy will translate to clinical effectiveness when implemented in healthcare practice (Walsh 2000), we conducted a sensitivity analysis to determine whether effectiveness studies (defined as those assessing the implementation of an intervention that uses existing service providers) demonstrate a beneficial outcome. That is, efficacy trials (those provided by dedicated research staff, n = 43) were excluded from the analysis. The frequencies of key variables for the 26 effectiveness studies (three of which did not report the primary outcome and so were not included in the aforementioned analysis) are presented in Table 6. For the 23 effectiveness trials with primary outcome data, the pooled effect size estimate significantly favoured the intervention group (average RR 1.42, 95% CI 1.11 to 1.82). This group of studies, however, was substantially heterogeneous (I2 = 67%; Q (22) = 66.37, P < .001). The pooled effect size estimate for effectiveness studies is very similar to the overall pooled effect size estimate (average RR 1.44, 95% CI 1.27 to 1.63) of the full sample (n = 70), although the effectiveness studies have a wider confidence interval and slightly greater heterogeneity. We can therefore conclude that our overall pooled effect size estimate (n = 70 studies) is not likely to be an over-estimate, although the addition of the efficacy trials introduced greater precision to the estimate.

Table 6. Frequency and percentage of key variables for effectiveness studies (n = 26)
Variable Frequency Percentage (n = 26)
Main intervention strategyCounselling1869%
Health education415%
Feedback312%
Other14%
Comparison typeUsual care1765%
Less intensive intervention831%
Unclear14%
ComponentsSingle component1038%
Multicomponent1246%
Tailored312%
Other14%
1.6.2 Assessment of risk of bias across studies
Random sequence generation selection bias

Not calculable due to insufficient numbers of studies with high risk of bias. Twenty-seven studies were classified as low risk of bias, three were high risk of bias, and the remainder were unclear.

Allocation concealment selection bias

Ten studies were classified as low risk of bias, 11 were high risk of bias, and the remainder were unclear. There was no significant between-group heterogeneity (QB (2) = 5.22, P = 0.07), although high risk studies had a larger pooled effect size estimate (average RR 2.11, 95% CI 1.48 to 3.00, I2= 0%) compared to low-risk studies (average RR 1.33, 95% CI 0.99 to 1.79, I2= 0%), or unclear bias studies (average RR 1.36, 95% CI 1.17 to 1.58, I2= 1%).

Incomplete outcome data attrition bias

Twenty-two studies were classified as low risk of bias, 13 were high risk of bias, and the remainder were unclear. There was no significant between-group heterogeneity (QB (2) = 0.13, P = 0.94). The mean effect size was largest for studies rated as high on this type of bias (average RR 1.47, 95% CI 1.09 to 1.99, I2= 0%), followed by unclear risk of bias (average RR 1.45, 95% CI 1.22 to 1.73, I2= 0%), and low risk of bias (average RR 1.39, 95% CI 1.10 to 1.75, I2= 13%).

Selective reporting bias

Twenty-nine studies were classified as low risk of bias, eight were high risk of bias, and the remainder were unclear. There was no significant between-group heterogeneity (QB (2) = 3.56, P = 0.17). The mean effect size was largest for studies rated as low on this type of bias (average RR 1.67, 95% CI 1.34 to 2.06, I2= 0%), followed by high risk of bias (average RR 1.50, 95% CI 1.09 to 2.08, I2= 0%), and unclear risk of bias (average RR 1.28, 95% CI 1.08 to 1.52, I2= 0%).

Detection bias (biochemical validation of smoking abstinence)

Forty-nine studies were classified as low risk of bias, 20 were high risk of bias, and one was unclear. There was no significant between-group heterogeneity (QB (1) = 0.06, P = 0.80). The mean effect size was similar, but largest, for studies rated as high on this type of bias (average RR 1.48, 95% CI 1.17 to 1.87, I2= 11%), followed by low risk of bias (average RR 1.43, 95% CI 1.22 to 1.67, I2= 0%); the one unclear study was treated as missing in this analysis.

Blinding of participants and personnel performance bias

Not calculable due to insufficient numbers of studies with low risk of bias.

Blinding of outcome assessment detection bias

Not calculable due to insufficient numbers of studies with high or low risk of bias.

Other bias (such as unequal recruitment to study arms in cluster trials; potential conflict of interest)

Fifty-four studies were classified as low risk of bias, eight were high risk of bias, and the remainder were unclear. There was no significant between-group heterogeneity (QB (2) = 1.28, P = 0.53). The mean effect size was largest for studies rated as low on this type of bias (average RR 1.47, 95% CI 1.28 to 1.69, I2= 0%), followed by high risk of bias (average RR 1.38, 95% CI 0.96 to 1.99, I2= 0%), and unclear risk of bias (average RR 1.18, 95% CI 0.82 to 1.70, I2= 0%).

Incomplete implementation

Twenty-two studies were classified as low risk of bias, 27 were high risk of bias, and the remainder were unclear. There was a significant between-group difference for this type of bias (QB (2) = 7.07, P = 0.03), though this is due to the difference in studies coded as 'unclear' (average RR 1.87, 95% CI 1.47 to 2.38, I2= 0%). Low risk of bias studies, assessed as having good implementation, had a similar effect size (average RR 1.33, 95% CI 1.10 to 1.62, I2= 17%) to high risk of bias studies (average RR 1.27, 95% CI 1.06 to 1.51, I2= 0%).

Equal baseline characteristics in study arms

Thirty studies were classified as low risk of bias, 15 were high risk of bias, and the remainder were unclear. There was no significant between-group heterogeneity for this type of bias (QB (2) = 4.79, P = 0.09). The mean effect size was largest for studies with unclear risk of this type of bias (average RR 1.67, 95% CI 1.33 to 2.10, I2= 20%), followed by low risk of bias (average RR 1.45, 95% CI 1.21 to 1.74, I2= 0%), and high risk of bias (average RR 1.13, 95% CI 0.86 to 1.47, I2= 0%).

Contamination of control group

Forty-nine studies were classified as low risk of bias, 13 were high risk of bias, and the remainder were unclear. There was no significant between-group heterogeneity (QB (2) = 2.12, P = 0.35). The mean effect size was largest for studies with unclear risk of this type of bias (average RR 1.50, 95% CI 1.07 to 2.11, I2= 0%), followed by low risk of bias (average RR 1.48, 95% CI 1.28 to 1.71, I2= 0%), and high risk of bias (average RR 1.19, 95% CI 0.90 to 1.56, I2= 29%), which were not significantly different from the null effect.

2. Secondary outcomes

2.1 Relapse prevention

In examining trends in separate comparisons of studies, the effect was not statistically different from zero in eight trials where the intervention was counselling and the control group received usual care (average RR 1.06, 95% CI 0.93 to 1.21; see Analysis 1.3) or four trials comparing counselling with a less intensive intervention (average RR 1.05, 95% CI 0.98 to 1.13; see Analysis 2.3). Single studies comparing health education with usual care (Petersen 1992) and social support with a less intensive intervention (McBride 2004) also did not show a significant difference between intervention and control groups (RR 0.97, 95% CI 0.71 to 1.31 and RR 1.02, 95% CI 0.89 to 1.16, respectively), figures not displayed as comparisons as only single studies.

2.2 Continued abstinence in the postnatal period
2.2.1 Zero to five months

In examining trends in separate comparisons of studies, a significant difference in abstinence at zero to five months was seen between intervention and control groups only in trials where counselling was compared with usual care (10 studies; average RR 1.76, 95% CI 1.05 to 2.95, see Analysis 1.4). However there was considerable heterogeneity between trials (I2 = 83%) and subgroups (Chi2 25.05 P < 0.0001), so these results should be considered with caution. Within this comparison, there was a significant effect in single interventions (average RR 1.52, 95% CI 1.13 to 2.05) and multiple interventions (average RR 2.32, 95% CI 1.44 to 3.72), but not in the single tailored intervention (average RR 0.88, 95% CI 0.80 to 0.97). There was also a significant difference in abstinence in a single trial where incentives were compared with an alternative intervention (Heil 2008) (RR 9.73, 95% CI 1.29 to 73.13, analysis not displayed in a table as only one study met the criteria).

However, the difference between intervention and control groups was not statistically significant in trials where: counselling was compared with a less intensive intervention (six studies; average RR 1.17, 95% CI 0.82 to 1.66; see Analysis 2.4); or where social support was compared with a less intensive intervention (two studies; average RR 1.36, 95% CI 0.46 to 4.07; see Analysis 8.3); There was also no clear effect where health education was compared with a less intensive intervention (two studies; average RR 1.29, 95% CI 0.52 to 3.22, see Analysis 4.2), but there is considerable heterogeneity in this comparison (I2 = 93%, Chi2 = 25.03, P < 0.0001), so these pooled results should be considered with caution. No significant difference between intervention and control groups was noted in single studies (analyses not displayed in a table as only one study met the criteria) comparing two alternative counselling interventions (Cinciripini 2010) (RR 1.05, 95% CI 0.63 to 1.76); health education versus usual care (Petersen 1992) (RR 1.02, 95% CI 0.75 to 1.38); or counselling as part of a broader intervention to improve maternal health (El-Mohandes 2011) (RR 1.46, 95% CI 0.97 to 2.19); or where social support was provided as part of a broader strategy to improve maternal health (Bullock 2009) (RR 0.96, 95% CI 0.51 to 1.81).

2.2.2 Six to 11 months

In examining trends in separate comparisons of studies, the effect bordered on a significant difference from zero between intervention and control groups in a separate comparison of counselling and usual care (six studies; average RR 1.33, 95% CI 1.00 to 1.77; Analysis 1.5), but not when counselling was compared with a less intensive intervention (three studies; average RR 1.08, 95% CI 0.83 to 1.40, see Analysis 2.5 . Additionally, there was not a significant difference between intervention and control groups when social support was compared with a less intensive intervention (two studies; average RR 1.09, 95% CI 0.83 to 1.42; see Analysis 8.4), or in single studies comparing two alternative counselling interventions (Cinciripini 2010) (RR 0.76, 95% CI 0.33 to 1.73) or contingent and non-contingent incentives (Heil 2008) (RR 3.24, 95% CI 3.24, 95% CI 0.35 to 29.82) (results not displayed as there was only one study in these comparisons).

2.2.3 12 to 17 months

In examining trends in separate comparisons of studies, there was a significant difference between the treatment and control in the two trials comparing counselling versus usual care (average RR 2.20, 95% CI 1.23 to 3.96, see Analysis 1.6), but not in two trials where counselling was compared with a less intensive intervention (RR 1.25, 95% CI 0.71 to 2.20, see Analysis 2.6); or a single trial (McBride 2004) where a multiple social support intervention was compared with a less intensive intervention (RR 1.22, 95% CI 0.92 to 1.64, analysis not displayed in a table as only one study met the criteria).

2.2.4 18+ months

Two trials of counselling combined with other strategies, and compared with usual care, measured self-reported continued abstinence beyond 17 months postpartum (Secker-Walker 1994; Lawrence 2003). However, no significant difference was reported between intervention and control groups (average RR 1.25, 95% CI 0.57 to 2.73, see Analysis 11.7).

2.3 Smoking reduction

No significant biochemically validated reductions were reported in any comparisons, including a comparison of counselling with usual care (three studies; RR 1.11, 95% CI 0.54 to 2.26, see Analysis 1.8) or counselling with less intensive interventions (two studies; RR 1.35, 95% CI 0.98 to 1.87, see Analysis 2.8). No significant difference in biochemically validated reduction was seen in single study by Tuten 2012 (analyses not displayed in a table as only one study met the criteria) comparing incentives with usual care (RR 7.62, 95% CI 1.92 to 30.25), which also demonstrated a significant difference between intervention and control groups in mean cotinine (standardised mean difference (SMD) -0.87, 95% CI -1.36 to -0.39). El-Mohandes 2011, comparing counselling as part of a broader maternal health strategy similarly did not report a significant difference between intervention and control groups in mean cotinine (SMD 0.11, 95% CI -0.17 to 0.39). The difference was also statistically different from zero for one study (Sexton 1984) measuring mean thiocynate (SMD -0.29, 95% CI -0.44 to -0.15), but not for mean cotinine (SMD -0.05, 95% CI -0.14 to 0.05), see Analysis 1.10.

There was also no statistically significant difference in self-reported reduction in smoking (mean cigarettes per day) seen in comparisons of: counselling and less intensive interventions (two studies; SMD -0.11, 95% CI -0.30 to 0.09, see Analysis 2.9); or health education compared with usual care (two studies, pooled effect not calculated due to considerable heterogeneity I2 = 76.8%, see Analysis 3.3). No difference in self-reported smoking (mean cigarettes per day) was also seen in several single studies (results not displayed as only one study met criteria), including: Hjalmarson 1991, which compared health education with a less intensive intervention (SMD 0.02, 95% CI -0.15 to 0.18); Tuten 2012 which compared incentives with usual care (SMD -0.23, 95% CI -0.69 to 0.23); LeFevre 1995 which compared feedback as part of a broader maternal health intervention with usual care (SMD 0.23, 95% CI 0.16 to 0.30); or Bullock 1995 which compared social support as part of a broader maternal health intervention with a less intensive intervention (SMD 0.15, 95% CI -0.34 to 0.64). The difference was not significantly different from zero in self-reported reduction (over 50%) in a single study (Hartmann 1996) which compared counselling and usual care (RR 1.59, 95% CI 0.98 to 2.57); or (Solomon 2000) which compared social support with a less intensive intervention (RR 0.96, 95% CI 0.64 to 1.44). Similarly, no difference in self-reported 'any' reduction in smoking was seen in a single study (Reading 1982) where feedback as part of a broader maternal intervention was compared with usual care (RR 0.95, 95% CI 0.42 to 2.18).

However, significant differences in self-reported reductions in smoking were seen in separate comparisons of: counselling and usual care for 'any self-reported reduction' (two studies; average RR 1.61, 95% CI 1.06 to 2.43, Analysis 1.9) and mean cigarettes per day (nine studies; SMD -0.25, 95% CI -0.46 to -0.03, Analysis 1.11); counselling and less intensive interventions (two studies; average RR 1.35, 95% CI 1.07 to 1.71, Analysis 2.7); feedback and usual care (two studies; average RR 1.69, 95% CI 1.24 to 2.31, see Analysis 5.2); and social support as part of a broader maternal health intervention with usual care in mean cigarettes per day (SMD -0.28, 95% CI -0.45 to -0.11, see Analysis 9.2). One single study comparing feedback and usual care (Valbo 1994) also reported a significant reduction in mean cigarettes per day (RR -0.63, 95% CI -1.03 to -0.24; results not displayed as only one study in comparison).

2.4 Infant outcomes

As a primary objective of this review is to determine if psychosocial interventions to support women to stop smoking in pregnancy have an impact on infant and maternal health outcomes, and large numbers are needed to detect relatively rare events, the pooled infant outcomes are included in this section of the review. These outcomes demonstrate the relationship between being randomised to a smoking cessation intervention and birth outcomes only, rather than the effectiveness of any particular intervention strategy.

2.4.1 Low birthweight

The pooled results of 14 trials which reported low birthweight (less than 2500 g) demonstrated a significant reduction (average RR 0.82, 95% CI 0.71 to 0.94; see Analysis 11.11). This pooled effect represents the following intervention strategies: eight counselling, two health education, one feedback, two incentives, and one social support. The number needed to treat for benefit (NNTB) in terms of low birthweight is 61, with a 95% CI of 38 to 204. Presented in a different way, nine out of every 100 participants in the control group experienced low birthweight births, compared to seven (95% CI six to eight) out of 100 for the intervention group. In contrast, there was no significant difference in three trials (two counselling and one feedback intervention) which reported infants born very low birthweight (less than 1500 g) (average RR 1.11, 95% CI 0.62 to 2.01, see Analysis 11.12).

In separate comparisons of studies, the effect was no longer significantly different from zero in smaller comparisons of counselling and usual care (six studies; average RR 0.87, 95% CI 0.70 to 1.08, see Analysis 1.12) or less intensive interventions (two studies; average RR 0.58, 95% CI 0.32 to 1.04, see Analysis 2.10), as large sample sizes are required to detect a significant difference in this outcome. There was no significant effect on the proportion of infants born low birthweight (less than 2500 g) in any of the single studies (results not displayed in tables) comparing: health education and usual care (Donovan 1977) (RR 1.10, 95% CI 0.66 to 1.84) or a less intensive intervention (Hjalmarson 1991) (RR 0.60, 95% CI 0.28 to 1.29); feedback and usual care (Haddow 1991) (RR 0.82, 95% CI 0.63 to 1.06); incentives and usual care (Tuten 2012) (RR 0.47, 95% CI 0.20 to 1.11) or an alternative intervention (Heil 2008) (RR 0.43, 95% CI 0.12 to 1.49); or social support and a less intensive intervention (Malchodi 2003) (RR 1.00, 95% CI 0.33 to 2.99). The effect remained non-significant in the three trials reporting very low birthweight infants (less than 1500 g) when separated into comparison of counselling and usual care (Analysis 1.13) and in a single study (Haddow 1991) comparing feedback and usual care (RR 0.90, 95% CI 0.35 to 2,32).

2.4.2 Preterm births

Pooled data from 14 studies reporting preterm births (less than 37 weeks' gestation) showed a statistically significant reduction in preterm births among women receiving psychosocial interventions (average RR 0.82, 95% CI 0.70 to 0.96; see Analysis 11.13), compared to women in the control groups. This pooled effect represents eight counselling, two health education, two feedback, and two incentives intervention strategies. The number needed to treat for benefit in terms of preterm births is 71, with a 95% CI of 42 to 341. Presented in a different way, eight out of every 100 participants in the control group experienced preterm births, compared to seven (95% CI six to eight) out of 100 for the intervention group.

In separate comparisons of studies, the effect was no longer significantly different from zero in comparisons of counselling and usual care (five studies; average RR 0.90, 95% CI 0.64 to 1.27, Analysis 1.14), counselling and less intensive interventions (three studies; average RR 0.82, 95% CI 0.47 to 1.42, Analysis 2.11), or feedback and usual care (two studies; average RR 0.60, 95% CI 0.28 to 1.29, Analysis 5.3), as large sample sizes are required to detect these relatively rare outcomes. Nor was a significant effect seen in comparisons which had only a single study (results not displayed in tables), including: health education and usual care (Donovan 1977) (RR 1.05, 95% CI 0.53 to 2.00) or a less intensive intervention (Hjalmarson 1991) (RR 0.76, 95% CI 0.32 to 1.80); or incentives compared with usual care (Tuten 2012) (RR 0.58, 95% CI 0.20 to 1.66) or an alternative intervention of non-contingent incentives (Heil 2008) (RR 0.38, 95% CI 0.11 to 1.30).

2.4.3 Mean birthweight

Pooled data from 19 studies reporting mean birthweight showed there was a statistically significant increase in mean birthweight of 40.78 g among women receiving the intervention (95% CI 18.45 to 63.10g, see Analysis 11.14), compared to women in the control group. The difference in mean birthweight was statistically significantly different from zero in subgroups of trials using counselling (n = 12) and incentives (n = 2) as the main intervention strategy, but was not significant in subgroups of trials using health education (n = 2), feedback (n = 2), or social support (n = 1) as a main intervention strategy.

In examining trends in separate comparisons of studies, the effect was borderline significant in comparisons of counselling and usual care (nine studies; MD 36.72, 95% CI 0.70 to 72.74, z = 2.00, P = 0.05, see Analysis 1.15), but not for comparisons of counselling and less intensive interventions (three studies; MD 56.02, 95% CI -31.46 to 143.50, see Analysis 2.12), or feedback and usual care (two studies; MD 79.43, 95% CI -53.05 to 211.91, see Analysis 5.4). There was no significant difference in mean birthweight in single studies (results not displayed in separate comparisons, only in comparison 1) comparing: health education and usual care (Donovan 1977) (MD -12.00, 95% CI -102.29 to 78.29) or less intensive interventions (Hjalmarson 1991) (MD 71, 95% CI -26.58 to 168.58); incentives and usual care (Tuten 2012) (MD 162, 95% CI -132.93 to 456.93) or non-contingent (alternative) incentives (Heil 2008) (MD 253, 95% CI-3.67 to 509.67); or social support provided as part of a broader maternal health intervention and a less intensive intervention (Malchodi 2003) (MD 28, 95% CI -152.48 to 208.48).

2.4.4 Perinatal deaths

Pooled data did not show a significant difference between intervention and control groups in perinatal deaths (four studies; average RR 1.13, 95% CI 0.72 to 1.77, see Analysis 11.15; although note that Valbo 1996 had a non-estimable effect), stillbirths (seven studies; average RR 1.22, 95% CI 0.76 to 1.95, see Analysis 11.16), neonatal deaths (four studies; average RR 1.15, 95% CI 0.44 to 3.06, see Analysis 11.17) or neonatal intensive care unit (NICU) admissions (four studies; average RR 0.78, 95% CI 0.59 to 1.04, see Analysis 11.18). These pooled effect size estimates, however, were based on small numbers of studies and had low power to detect clinically important differences. A number of trials also excluded women who had a perinatal death or a preterm birth from the study population.

In separate comparisons of studies, there was no significant effect seen in comparisons of counselling and usual care for: stillbirths (four studies; average RR 1.08, 95% CI 0.51 to 2.30, Analysis 1.17), neonatal deaths (three studies; average RR 2.06, 95% CI 0.61 to 6.92, Analysis 1.18), or NICU admissions (two studies; average RR 0.82, 95% CI 0.52 to 1.29, Analysis 1.19). There was unclear evidence in relation to counselling and usual care for perinatal deaths because the effect size for one of the two studies (Valbo 1996) was not estimable due to zero events in both groups, therefore pooled effect size not calculable (see Analysis 1.16). There was no significant effect observed for feedback and usual care in stillbirths (two studies; average RR 1.28, 95% CI 0.69 to 2.39, Analysis 5.5). There was no difference in single studies (results not displayed in comparison tables, only in comparison 1) comparing: counselling and a less intensive intervention (Ershoff 1989) in stillbirths (RR 1.84, 95% CI 0.17 to 20.04); health education and usual care (Donovan 1977) in perinatal deaths (RR 4.40, 95% CI 0.49 to 39.08); feedback and usual care (Haddow 1991) in perinatal deaths (RR 1.05, 95% CI 0.59 to 1.87) or neonatal deaths (RR 0.40, 95% CI 0.08 to 2.07); incentives and usual care (Tuten 2012) in NICU admissions (RR 0.75, 95% CI 0.45 to 1.25); or incentives and an alternative (non-contingent incentive) intervention (Heil 2008) in NICU admissions (RR 0.76, 95% CI 0.24 to 2.49).

NB. The following sections for outcomes 2.4.5 to 2.12 are narrative descriptions based on the findings reported in the studies, rather than on results of statistical synthesis

2.4.5 Other infant outcomes

Two trials (Cope 2003; Heil 2008) reported significant increases in fetal growth measures including fetal femur length and fetal abdominal circumference, and infant length, but no significant difference in head circumference between control and intervention groups. Two trials reported no significant difference in Apgar scores at one and five minutes post-birth (Cope 2003; Tuten 2012).

2.5 Mode of birth

None of the three trials measuring mode of birth by intervention group (Thornton 1997; Cope 2003; Tappin 2005) reported a significant difference in the rate of operative births by intervention group.

2.6 Breastfeeding

There were mixed results for the effect of interventions on breastfeeding. Two trials that measured breastfeeding initiation (Panjari 1999; McLeod 2004) showed no significant difference in initiation or duration of breastfeeding in control or intervention arms. One trial of contingency management measured a significant effect on breastfeeding duration (Heil 2008) at both eight weeks and 12 weeks postpartum.

2.7 Psychological effects

Nineteen studies reported baseline psychological measures of interventions, reinforcing the findings from observational studies that there are significant psychological symptoms among many pregnant women who smoke. Up to 75% of pregnant women who smoked had current or previous psychological symptoms (Belizan 1995; Ershoff 1999; Cinciripini 2010; Ondersma 2012) and approximately 20% to 25% of women reported major depression based on CES-D scale assessments (Blalock 2005; Dornelas 2006; Bullock 2009; Cinciripini 2010; El-Mohandes 2011). Four studies identified baseline depression or stress as a ‘mediator’ or ‘predictor’ of continued smoking at follow-up (Crittenden 2007; Linares 2009; Stotts 2009; El-Mohandes 2011), suggesting depressive symptoms may be an ‘independent contributor to the problem of continued smoking during pregnancy’ (Linares 2009).

Nine trials reported post-intervention psychological outcome measures and none reported any negative psychological effects. Six trials showed that smoking cessation interventions in pregnancy do not increase stress and psychological symptoms for women (Manfredi 1999; Panjari 1999; Aveyard 2004; Rigotti 2006; Solomon 2006; El-Mohandes 2011). Furthermore, three studies demonstrated that smoking cessation interventions have the potential to improve women's psychological wellbeing and self-esteem (Stotts 2004; Bullock 2009; Cinciripini 2010) and self-efficacy (Stotts 2004).

2.8 Impact on family functioning and other relationships

No studies reported measures of family functioning. Studies reporting analysis of social networks (Stotts 2009), suggest a significant interaction between smoking networks (household and other) or partner smoking (Bullock 2009) and continued smoking of participants in late pregnancy. Two studies reporting perceptions of partner (McBride 2004) and peer support (Hennrikus 2010) had mixed findings. Pregnant women reported less negative partner support through pregnancy, but this increased in the postpartum period (McBride 2004). Women in another study reported an increase in both positive and negative support from a peer including: comments about the woman’s lack of willpower, trying to make them feel guilty, expressing anger about smoking and trying to scare women about smoking (Hennrikus 2010).

2.9 Participants views

Twenty-six trials included women's views of the interventions, 12 studies reported providers’ views of the interventions and two studies reported measures of knowledge, attitudes or practice among pregnancy care providers.

Women’s views

Twenty-nine studies reported that they addressed in the intervention issues identified as concerns by women when consulted for this review (Oliver 2001); including ‘coping with stress and emotions’, misconceptions about smoking risks, and feelings of guilt. Two studies described using interactive discussions to address issues of concern to individual women (Sexton 1984; Hennrikus 2010).

Three studies reported outcomes related to maternal weight gain. One study (Sexton 1984) reported a slightly higher mean weight gain in the intervention group (12.9 kg) compared to the control group (11.9 kg). Two other studies did not report weight gain by intervention exposure but reported that women with a ‘high concern’ about weight gain were less likely to quit smoking during pregnancy or remain abstinent postpartum (Berg 2008), and another reported an increased weight gain of 2.8 kg in women who were abstinent compared to women who continued to smoke (P = 0.04), with an estimated 0.34 kg increase in weight gain for every 10% increase in smoking abstinence (Washio 2011).

Two studies explicitly mentioned consideration of women’s views in developing the intervention (Albrecht 1998; Cinciripini 2010), and six studies described the involvement of women or community members in the development of the intervention (Windsor 1985; Belizan 1995; Gielen 1997; Albrecht 2006; Patten 2009; Eades 2012).

Thirty-two studies reported women’s views about the content and delivery of the interventions. When asked, most women gave favourable feedback on the intervention and intervention materials (Baric 1976; Ershoff 1989; Belizan 1995; Bullock 1995; Lillington 1995; Secker-Walker 1997; Walsh 1997; Cinciripini 2000; Strecher 2000; Tappin 2000; Hajek 2001; Cope 2003; Tappin 2005; El-Mohandes 2011; Ondersma 2012), particularly audiovisual materials (Windsor 1993; Patten 2009; Ondersma 2012) and telephone support (Bullock 1995; Solomon 2000; Rigotti 2006; Bullock 2009). Women offered personal contact and a manual considered the personal contact the most important element and women appreciated printed materials much less if they were also offered a video, although the video combined with printed materials was no more effective than the printed materials alone (Secker-Walker 1997; Cinciripini 2000). Similarly, women offered motivational interviewing for relapse prevention were more likely to be satisfied than those offered a booklet, although the motivational interviewing was no more effective (Ershoff 1999. Women participating in a study in Ireland (Thornton 1997) reported the importance of providing the intervention in privacy, and suggested that telephone follow-up between visits and a video would have been helpful components in that intervention. Two studies reported that even if they did not like it, women expected to be asked about smoking from their care provider (Walsh 1997; McLeod 2004). Two trials using computer-assisted technology were rated positively (Strecher 2000; Ondersma 2012), but in an earlier trial women expressed concern about entering personal information into a computer (Ershoff 1999).

Despite positive feedback about the content of the intervention, several trials reported difficulty recruiting and retaining women’s participation in the intervention (Loeb 1983; Secker-Walker 1994; Cinciripini 2000; Stotts 2004; Patten 2009), and many studies had low participation rates. In a multimodal intervention including counselling and nicotine replacement therapy (NRT), only 87/327 women in the intervention group participated in counselling and only 75 women used NRT (Hegaard 2003).

Offering additional group sessions for smoking cessation was generally a poorly accepted intervention even in otherwise successful trials (Loeb 1983; Windsor 1985), though one study reported groups were well accepted (Sexton 1984). Hypnosis was also a poorly accepted intervention in two studies (Sexton 1984; Valbo 1996). Five studies reported women's negative views of intervention components, including: use of carbon monoxide monitoring and prompt cards (Thornton 1997); some peer support behaviours (Hennrikus 2010), limited perceived efficacy of booklets (Moore 2002), and phone messages (Ershoff 1999).

Providers’ views

Ten studies reported providers' views of the intervention. While providers’ views about the interventions were generally positive, a recurrent theme was their concern about the time taken by the intervention (Kendrick 1995; Hajek 2001; Moore 2002; Campbell 2006) and the impact on their relationship with women (Hajek 2001; Wood 2008). Sixty-five per cent of midwives asked to use a carbon monoxide monitor and provide 'stage of change'-based advice considered that this could not be achieved in the time available. This led to less than full implementation and variable motivation to promote smoking cessation counselling among staff in some studies (Kendrick 1995; Moore 2002), but not all (Windsor 2011). One of the reasons given for tailoring messages to ‘stages of change’ was to address providers' concerns that interventions may alienate women not ready to quit (Hajek 2001). A survey of general practitioners suggested the smoking status of the provider influenced participation in intervention delivery (Haug 1994). Despite these challenges, engagement and involvement of providers was identified as a critical element of implementation (Lowe 1997; McLeod 2004; Campbell 2006) and providers reported that they would like more involvement (Tappin 2000).

2.10 Measures of knowledge attitudes and behaviour of health professionals with respect to facilitating smoking cessation in pregnancy

Two trials reported positive effects of the interventions on midwives' understanding, confidence in delivering the intervention, optimism that the intervention may influence women’s smoking behaviour (Lawrence 2003) and obstetric knowledge and practice (Secker-Walker 1992).

2.11 Cost-effectiveness

Four studies reported that the interventions were cost-effective using a variety of measures. Pregnancy-specific, self-help materials were more cost-effective than standard smoking cessation information or self-help materials (Windsor 1985). Specific estimates include: a benefit-cost ratio of 2.8:1 (Ershoff 1990); 1 (non-smoker): $84 (Parker 2007); and an average cost of $56 per person for each smoking cessation intervention, and $299 to produce a non-smoker at the end of pregnancy (Dornelas 2006).

2.12 Adverse effects

Three studies that measured whether women increased their smoking following exposure to the intervention showed mixed results. One trial reported a slightly lower level of cotinine in the intervention group, compared to the control group (Tappin 2005), another reported no difference in self-reported smoking (Hjalmarson 1991), and another reported an increase in smoking among women who did not quit (Haug 1994).

Discussion

Summary of main results

Studies in this review demonstrate that psychosocial interventions can support women to stop smoking in pregnancy. Importantly, the interventions do not appear to have any negative physical or psychological effects, are positively received by most women, and may improve psychological wellbeing. Incentives had the largest effect size, but only when provided intensively. Counselling was effective when provided in conjunction with other strategies or tailored to individual women, but it is unclear whether any types of counselling are more effective than others. Peer support appeared to be effective, but only when provided as a targeted intervention and not as part of a broader intervention to improve maternal health. It is unclear whether partner-assisted support helps women to quit. Feedback appeared to be effective when combined with other strategies, such as counselling, and compared with usual care, but not less intensive interventions. Health education was not effective in separate comparisons, but the pooled effect was significantly different from zero in subgroup analyses. Among women who received psychosocial interventions there was a significant reduction (18%) in preterm births (less than 37 weeks' gestation), the proportion of babies born low birthweight (18%) (less than 2500 g), and a significant increase in mean birthweight of 41 g. Using data from this review, the NNTB to prevent one infant being born low birthweight is 61 (95% CI 38 to 204); and 71 interventions (95% CI 42 to 341) to prevent one infant being born preterm. These findings provide strong and clear evidence about the risks of smoking during pregnancy, supporting recommendations that it may be an integral part of strategies to reduce preterm births (Green 2005a). Given the benefits of stopping smoking in pregnancy for the woman and her infant, this would seem to be an important intervention, particularly when applied at a population level. However, it remains unclear from dissemination trials whether interventions are effective when implemented into routine pregnancy care.

Among the subgroups of 'main intervention strategies' categorised in this review, the four studies that included use of incentives had the strongest effect. Three trials that compared provision of intensive incentives with usual care (Tuten 2012), incentives and social support compared with a less intensive intervention (Donatelle 2000), and contingent incentives compared with non-contingent incentives (Heil 2008), were significantly different from zero. A three-armed trial, which included a non-contingent arm (Tuten 2012), also showed a significant effect. These non-contingent comparisons provide a 'time-matched' alternative comparison of similar intensity, which helps to identify if it is the 'additional assistance' or incentives which are effective (Mantzari 2012). The effect was also significantly different from zero in the pooled results of three counselling interventions that included lottery tickets (Sexton 1984; Walsh 1997; Parker 2007). These findings are consistent with other reviews of financial incentives in pregnancy (Higgins 2012) and the mechanisms for the effectiveness of incentives for reducing substance abuse more generally has been well documented (Higgins 2008b). However, the results of the incentives trials should be considered with caution as they are based on few trials with a very small number of women (less than 500), all of whom were in the US. Additionally, there was no effect from one trial of 'low intensity' incentives ('CM Lite') combined with an interactive computer-generated counselling program (Ondersma 2012), which relied on women initiating contact with the research team for urine cotinine testing, and provided a maximum of only five verification and 'incentive' interactions, with less than half the women in this arm submitting even one urine test. Interestingly, women in this four-armed trial who received the interactive computer-generated counselling program alone were more likely to quit than women who received the combined incentive and computer-counselling intervention (see Ondersma 2012).

Pooled results of interventions in which counselling was the main intervention strategy showed a significant effect in abstinence in late pregnancy. However, in separate comparisons, the effect of counselling was only significantly different from zero when combined with other strategies or tailored to individual needs. There was no significant difference seen when one type of counselling (cognitive behavioural therapy (CBT)) was compared with traditional health education (Cinciripini 2010), or when counselling was provided as part of a broader intervention to improve maternal health (El-Mohandes 2011). Group interventions were generally not well accepted in this population of pregnant women, despite being reported as a potentially well accepted intervention in the general population (Bauld 2010). Feedback was effective when combined with other strategies such as counselling, and only when compared with usual care. Findings from this review support recommendations that pregnant women may need more support than just brief advice or health education (Coleman 2004), as it was unclear whether health education alone helped women to quit. However, there was a significant pooled effect among the three trials of health education when two studies were removed providing only self-help materials or an audiotape with no additional personal advice, which is similar to findings in another review (Murthy 2010), and which concluded that apart from brief physician advice, there was limited clarity on the duration of interventions required by other professionals.

Social networks have been suggested as a major cause of relapse (Nguyen 2012b), and a systematic review of qualitative studies identified partners as one of the most important influences on women's smoking and relapse (Flemming 2013). In this review, peer support appeared to be effective when provided as a targeted intervention, and when social support was provided as part of a broader intervention to improve maternal health, but not when [telephone] support was compared with a less intensive intervention. It is unclear from the single trial of partner-assisted support (McBride 2004) that this strategy can help women to stop smoking. Furthermore, counselling interventions that included support for partners to quit also did not show a significant effect, and there were mixed results in the four studies reporting associations between quitting and partner smoking. Mixed results have similarly been reported in a systematic review of five randomised controlled trials (Duckworth 2012), and another review of seven studies reported a non-significant effect (Hemsing 2012), concluding that, "Despite the importance of partner smoking, there are very few effective smoking cessation interventions for pregnant/postpartum women that include or target male partners". This raises questions about arguments that a major reason for the modest effect of smoking interventions is the focus on individual behavioural change rather than acknowledging social factors and focusing on external motivation (Okoli 2010). Additionally, feedback from women demonstrates the support from both partners and peers can sometimes be negative, which raises concerns about the potential risks for vulnerable women in physically or emotionally violent relationships. Evidence from this review suggests that while partner and peer support may be important factors influencing smoking behaviour, eliciting peer and partner support that is positive and can actually support women to stop smoking in pregnancy is a challenge.

The lack of a clear difference in effect seen by increasing intervention intensity challenges the validity of the assumption that ever-increasing the intensity of support will increase quit rates, as has been reported by other commentators (Lando 2001), and supports views that there may be an upper limit of what women accept (Chapman 2012). Newly included studies in this review had lower effect sizes than older studies in the previous version, despite a general trend towards higher intensity interventions in more recent trials. It may be that women who continue to smoke are not getting 'more hard core' but that there are many options already available and additional strategies may not be offering a lot of extra benefit, as risks of smoking during pregnancy, due to health education campaigns, are well known in high-income countries (Campion 1994; Eriksson 1996; Eriksson 1998). One study found relapse within the first two weeks was predictive of continued abstinence, and suggested this indicates that intensive support during the earlier period of nicotine withdrawal may be an important component of interventions (Higgins 2006b).

Studies in this review suggest the effect during pregnancy continues into the postpartum period, up until approximately 18 months postpartum, though the smaller effect size shows many women who did quit during pregnancy relapse postpartum. Some suggest that many pregnant smokers simply suspend their smoking for the duration of pregnancy as opposed to quitting altogether or they commit to 'temporary abstinence' for pregnancy (Stotts 1996; Lawrence 2005a; Flemming 2013), but these relapse rates are similar for non-pregnant women (Bombard 2012). Rather than being disappointed by these limited effects, some authors suggest healthcare workers should focus on the positive aspects of these findings and reinforce the positive decisions many women are making when pregnant (Hotham 2008). High post-pregnancy relapse rates have led to some commentators calling for an extension of the period of support for women to stop smoking (Coleman-Cowger 2012). Hjalmarson 1991 reported a high proportion of women abstaining from smoking during their hospital stay for the birth, and suggests this may be an opportunity for intervention to reduce the risk of postpartum relapse. These findings suggest there may be a need for different approaches to promote continued abstinence postpartum, including focusing on the benefits for the mother, without excessive emphasis solely on the benefits for the baby.

While results are mixed, studies in this review suggest there is a reduction in self-reported smoking but not biochemically validated smoking. Continued nicotine and cigarette exposure may have effects on other outcomes not measured in this review. The level of reduction required to improve health outcomes remains unclear (Secker-Walker 2002a). One study analysing data from Kendrick 1995 suggested that reduction in smoking to fewer than eight cigarettes a day is necessary to avoid reduction in infant birthweight (England 2001), and estimated approximately a mean birthweight which was 200 g higher among women who quit smoking after enrolment, compared to women who continued to smoke during pregnancy. Therefore, extrapolating these data to this review, if all women in the intervention groups stopped smoking and none of those in the control group did, the expected mean birthweight difference would be about 200 g, rather than 41 g. With an absolute difference of six in every 100 women stopping smoking, the expected mean difference from the extent of smoking cessation alone would have been about 12 g. This suggests that smoking reduction is also happening to a greater extent in the intervention than comparison groups, in line with self-reported changes.

There was no evidence from studies in this review that smoking cessation increases the rate of caesarean section (Thornton 1997; Cope 2003; Tappin 2005), contrary to concerns raised by women about the effects of increased fetal size (Sexton 1984). One observational study modelled increases in birthweight (from 2450 g to 2550 g) in Guatemala and found an increased risk in caesarean section due to obstruction of eight in every 1000 cases, but this was outweighed by a reduction in caesarean section due to fetal distress of 34 per 1000 cases (Merchant 2001).

Women who smoke are less likely to initiate breastfeeding (Amir 2001a; Amir 2002a; Donath 2004; Einarson 2009; Disantis 2010b), and breastfeed for shorter duration (Sayers 1995; Horta 1997). Therefore, supporting women to initiate and maintain breastfeeding should be considered an important part of any intervention in this population group, and reported as an outcome in intervention studies. Studies in this review had mixed reports of the effect of smoking cessation interventions on breastfeeding (Panjari 1999; McLeod 2004; Higgins 2010b).

Studies in this review (Cinciripini 2000; Rigotti 2006) support a recent qualitative study that concluded "Pregnant women with mental disorders appear more motivated...yet find it more difficult, to stop smoking" (Howard 2013), and other studies that report higher rates of quitting among women with higher self-esteem and self-efficacy (Massey 2013). For these reasons, healthcare workers have reported difficulty addressing smoking with pregnant women (Wood 2008). Qualitative studies have identified concerns about adverse effects of quitting, or increased guilt over continued smoking, on women's psychological wellbeing and capacity to cope with adverse circumstances, with follow-on effects to the women's families (Oliver 2001; Wood 2008; Flemming 2013). In earlier versions of this review, it has been difficult to assess the effect of interventions on depression, as, despite the strong associations with poor mental health and smoking in pregnancy, women with mental illness were frequently excluded from trials. However, mental wellbeing has been addressed in more recent trials and, contrary to the above concerns, there is no evidence from studies in this review that there are any negative psychological consequences from delivery of individual smoking cessation interventions in pregnancy. Rather, feedback from women from studies in this review was positive with women feeling that "somebody cared" (Bullock 1995). Three studies have shown that provision of psychosocial support can in fact improve women’s psychological wellbeing, which has the potential to have enormous benefits for the mother, the infant, and the whole family (Bullock 1995; Stotts 2004; Cinciripini 2010).

In earlier versions of this review, there appeared to be little evidence of the involvement of pregnant women who smoked or caregivers being involved in the design and evaluation of interventions (Oliver 2001). However, there has been increasing discussion of women's preferences for cessation support in recent years (Ussher 2004). Studies included in this review suggest women prefer individual personal contact, particularly by telephone, though studies inclusive of telephone support in this review did not appear to be significantly more effective. Rates of satisfaction with interventions delivered by computers or mobile phones were generally positive, but again there was no evidence in this review that the use of these technologies increased the rate of abstinence in late pregnancy. Nevertheless, acceptability of an intervention is an important aspect of population-based interventions.

Some evidence suggests that women in high-income countries are more likely to smoke to control their weight, and that female body image is extensively targeted by tobacco marketing campaigns (Pomerleau 2000; CDCP 2002; Levine 2006), although concerns about gaining weight through stopping smoking during pregnancy were not raised by any of the women consulted for this review (Oliver 2001). The systematic review of qualitative studies of women smoking in pregnancy (Flemming 2013) found two studies mentioning weight gain as a factor in considering smoking cessation. Hotham 2002 found that fear of weight gain was a barrier to smoking cessation for some women and Lawson 1994 found some women used smoking to cope with weight gain.Three studies in this update of the review (Sexton 1984; Berg 2008; Washio 2011) address weight gain. Only one study reported a small increase in weight gain among women in the intervention group (Sexton 1984). This concern should be considered in interventions, with interventions available to support women to avoid unwanted weight gain (Farley 2012). It should be noted that weight gain in pregnancy may not necessarily be a negative outcome for many women, particularly women in low- and middle-income countries. The association between smoking and glucose intolerance, a potential mechanism for these effects, remains unclear (Wendland 2008). A Cochrane systematic review of interventions for preventing weight gain after smoking cessation mentioned neither pregnancy nor breastfeeding (Parsons 2009) and therefore cannot be relied upon for evidence relevant to a population where weight may fluctuate for normal physiological reasons and where babies may be sensitive to drug treatments in utero or when breastfeeding.

Public health impact of the interventions

Importantly, psychosocial interventions to support women to stop smoking during pregnancy reduce the population-attributable risk of preterm birth (by 18%) and low birthweight (by 18%), with approximately 71 interventions required to prevent one preterm birth and 61 interventions to prevent one infant being born with low birthweight. As such, smoking cessation is recommended as a key recommendation for reducing the risk of recurrent preterm birth (Chang 2012; Cypher 2012). The number of interventions needed to treat for benefit is extraordinarily low, given the serious clinical consequences of these adverse outcomes. Based on the effectiveness published in the 2004 version of this Cochrane review, if 75% of pregnant women in the US disclosed their smoking status and all received the intervention, then it has been estimated that 31,573 (6%) 'new quitters’ would be gained and the prevalence of smoking in pregnancy would potentially decrease from 16.4% to 15.6% (Kim 2009b). While these effect size estimates may appear modest, the response to interventions is similar to that of psychosocial interventions to reduce type 2 diabetes mellitus, hypertension and asthma, all of which are conditions that involve a combination of medical illness, personal choice and environmental factors (McLellan 2000). Importantly, the high prevalence of these conditions in the community means that interventions with a modest effect size estimate can have a substantial impact on population health if widely implemented.

Economic costs

Studies in this review report variable cost-effectiveness measures and costs of interventions. Based on a NNTB of one quitter for each 19 interventions, our cost estimates ($US1,064) based on $US56 per interventions is significantly higher than the $US299 reported in Dornelas 2006. However, even with higher estimates, other studies that evaluated the cost-effectiveness of these interventions clearly show that there is a ‘rapid return on investment’ (Lightwood 1999). Early studies estimated the smoking-attributable maternal costs during pregnancy alone ranged from $US150 million to $US995 million in the early 1990s (Adams 1998), with 2004 estimates of $US122 million or $US279 per smoker (Adams 2011). Estimated birth and first year costs for both mothers and infants attributed to smoking were $1142 to $1358 per smoking woman over a decade ago (Aligne 1997; Miller 2001; Adams 2002). Infant costs are approximately 10 times maternal costs, accounting for 90% of costs in the first year. Low birthweight produces the highest economic burden as it is the most common adverse outcome (Hueston 1994; Miller 2001). A 1% drop in smoking prevalence was estimated to prevent approximately 1300 low birthweight live births and save $US21 million in direct medical costs (Lightwood 1999). Inclusion of smoking attributable and environmental tobacco smoke exposure costs in birth and childhood conditions, pushes estimates into the billions (Aligne 1997), and long-term costs due to chronic disease up to $US57 billion in 1997, in the US alone (Bartlett 1994). An economic evaluation of data provided in the 2009 version of this review estimated the societal benefits from these interventions could be in excess of 500 million pounds sterling per annum in the United Kingdom (Taylor 2009). In contrast with that finding, the quality of diet in pregnancy (in high-income countries) has not been shown to affect the mean birthweight of infants over 32 weeks' gestation (Rogers 1998). While there is variation in reported costs dependent on conditions included and changing healthcare costs (Ayadi 2006), it is clear that healthcare costs due to smoking in pregnancy are substantial.

Impact on health inequalities

In high-income countries, the reduction in rates of smoking has not been as substantial in women experiencing psychosocial disadvantage, as for the general population. Hence smoking has been identified as a major preventable cause of the health inequalities experienced by women who suffer psychosocial disadvantage, including psychological illness, low educational attainment, young early motherhood, lack of social support, and limited employment (Graham 2006). Some of the reasons may be that disadvantaged women are unable to change the environmental factors that increase the risk of smoking; population-based interventions may have the effect of being judgemental and alienate women; and women are unable to change generational patterns (Graham 2009). Several authors have suggested that women who continue to smoke in late pregnancy would be unlikely to benefit from the usual antenatal interventions, which rely on women's capacity for self-initiation, self-control and social resources, which they suggest helps to explain why it remains such an intractable problem (Wakschlag 2003; Pickett 2009) and that individual interventions alone are unlikely to impact on inequalities (Baum 2009). However, subgroup analysis of studies included in this review refutes these arguments and suggests that individual interventions provided during pregnancy have similar effectiveness among women with low socio-economic status (SES), as women who are not classified as having low SES, despite several studies reporting a lower effect among participants with lower SES (Baric 1976; McLeod 2004; Pbert 2004; Rigotti 2006). This supports qualitative studies that suggest individual support, which is positive rather than punitive, has an important role (Bond 2012). Therefore, individual psychosocial support should form a part of the tobacco control ‘package’ to reduce smoking during pregnancy, in conjunction with population-based measures, which have also been shown to have a significant impact on birth outcomes (Adams 2012; Cox 2013) and reducing smoking in disadvantaged populations (Thomas 2008).

The pooled results were not significantly different from zero in eight studies, which were developed predominantly or specifically for ethnic and aboriginal minority women, including African-American women (Gielen 1997; Manfredi 1999; El-Mohandes 2011; Ondersma 2012), African American and Hispanic women (Lillington 1995), Hispanic women (Malchodi 2003), Alaskan Native Women (Patten 2009) and Australian Aboriginal and Torres Strait Islander women (Eades 2012). This is despite primary authors in several studies reporting subgroup analysis of higher quitting rates among African-American and Hispanic women than other women (Petersen 1992; Windsor 1993; Pbert 2004; Parker 2007). These studies tended to involve women more in the development of the intervention and all used several recommended strategies to tailor the intervention (American Legacy Foundation 2012) for initiatives that aim to address the disparities in tobacco use; including hiring culturally competent staff, conducting formative research to identify community needs, piloting and field-testing programs, ‘cultural tailoring’ of smoking cessation resources, and collaborating with key stakeholders and community organisations. Three studies adapted ‘SCRIPT’ materials in the US (see Windsor 2011), which include: 'asking' about smoking status; 'advising' women to quit; 'assisting' women to quit by providing advice on skills and materials such as video's and self-help materials; and arranging for follow-up by referral at future appointments. Two studies developed audiovisual resources for African American (Ondersma 2012) and Alaskan Indian (Patten 2009) women, and these resources received positive feedback. Despite interventions being reported as feasible and acceptable to communities, there were challenges with implementation and few demonstrated an effect size estimate that was significantly different from zero. Further suggestions included trying to recruit from different settings and including elders to improve recruitment, and recognising the importance of broader social interventions for potentially reaching a larger proportion of pregnant women (Patten 2009). Other reviews of interventions in non-pregnant aboriginal peoples have demonstrated interventions can be effective (Carson 2012), and suggest mobile phone technology may be a feasible intervention strategy (Johnston 2013). Only one study included women using smokeless tobacco products, and identified conflicting beliefs about the effect of these products during pregnancy and the primary change recommended by participants in the study was to provide “more objective” information on the risks of Iqmik (smokeless tobacco) use for the infant (Patten 2009).

Most interventions have been developed in high-income countries and there is very limited information about the effectiveness of psychosocial interventions for individual women in low- to middle-income countries (Murthy 2010). The restrictions on tobacco marketing in high-income countries may result in an increase in tobacco marketing companies in low- and middle-income countries. Smoking has the potential to undermine health improvements in low- and middle-income countries and a range of interventions are needed to manage the emerging epidemic (Lopez 1994; Abdullah 2004). However, given the modest effect size estimate of individual interventions, population-based tobacco control strategies are an urgent priority, as there is now a brief 'window of opportunity' to prevent the increase of smoking among women in many low-income countries (Chomba 2010).

Translation of evidence into practice

The first trials of anti-smoking interventions during pregnancy were published more than 30 years ago (Baric 1976; Donovan 1977). The first trial to demonstrate the reversibility of the birthweight reduction associated with smoking by an intensive intervention during pregnancy was published in 1984 (Sexton 1984). Since then, attempts at widespread implementation of psychosocial interventions to support women to stop smoking in pregnancy have demonstrated many of the challenges of translating ‘evidence into practice’, particularly non-pharmacological evidence (Windsor 1998; Windsor 2000b; Lowe 2002; Moore 2002; NICS 2003; McLeod 2004; Herbert 2005; McDermott 2006; Abatemarco 2007; Manfredi 2011).

Studies in this review can be conveniently categorised within a framework for translation of research into practice (Nutbeam 2006), which suggests progression through several stages from; problem definition (descriptive studies) and formative research for intervention design; intervention efficacy research; to implementation in routine/normal settings (effectiveness research); dissemination across several settings; and institutionalisation (as interventions are provided as part of routine care). Many studies in this review clearly defined the problem and conducted formative research for intervention development (Katz 2008; Gilligan 2009), particularly interventions developed for vulnerable women, including young women (Albrecht 1998; Albrecht 2006). The modest but significant efficacy of psychosocial interventions provided by researchers has been well demonstrated by studies in this review, including counselling interventions.

The transfer of an intervention from one setting to another may reduce its effectiveness if elements are changed or aspects of the materials are culturally inappropriate. An example in these trials was the performance of the Windsor self-help manual. This was developed and shown to be effective in Birmingham, Alabama (Windsor 1985; Windsor 1993). However, when it was implemented into routine care (Windsor 2011), used in Baltimore with peer counsellors who received minimal training instead of trained health educators (Gielen 1997), adapted for Alaskan Native women (Patten 2009) and transferred to other countries (Lowe 1998a; Lowe 1998b), the effectiveness was much lower. An analysis of health promotion trials has concluded that where the providers are also the researchers (more likely in single centre studies than multicentre studies), they appear to be better providers for influencing behavioural outcomes and about the same as other providers for other outcome domains (Oliver 2008a). The larger, multicentre trials may therefore be a more accurate representation of implementing policy than smaller, single centre trials. In this review, interventions provided by usual care providers were as effective as interventions provided by researchers, including counselling interventions. However, there was substantial heterogeneity in sensitivity analyses of trials provided by usual care providers in this review, which supports the views that there are many variables to consider when implementing interventions in routine settings (Hoddinott 2010).

Despite evidence of efficacy and effectiveness, dissemination trials of counselling interventions into pregnancy care settings suggest challenges to translating this efficacy research into routine practice and policy. Data from the five dissemination trials that targeted the intervention at the organisational level, demonstrated significant effects in terms of increased implementation of interventions in routine practice, although challenges were reported and this did not translate into a significant reduction in rates of smoking among women in the intervention arms of these studies. One study that provided clinics with resources and referral options reported an increase in women’s recall of receiving interventions (Manfredi 1999). A significantly higher program implementation rate was reported when using an intervention based on Rogers' 'Diffusion of Innovation' theory (43% compared with only 9% implementation in the control group after one year), but there were no data on the impact on smoking outcomes (Lowe 2002). An increased uptake of the intervention by staff was demonstrated using ‘active’ dissemination compared to a simple mail-out of information (Cooke 2001), but not at levels sufficient to have a significant impact on smoking outcomes in women (Campbell 2006), which was similar to other dissemination trials reporting smoking outcomes (Pbert 2004; Windsor 2011). Another non-randomised study compared the use of the RE-AIM dissemination model to increase the reach, efficacy, adoption, implementation, maintenance of interventions (Lando 2001) and concluded that multi-faceted approaches using strategies from each intervention were most likely to improve implementation.

There are a number of possible explanations for the limited effect in dissemination trials. Firstly, many of the studies that recruited individual women did not provide information on the number of women who were eligible for inclusion or were approached to take part in trials. The 'participation rate' would have provided useful information about the general ‘acceptability’ of the intervention, as well as the degree of ‘selection bias’ in the study population (Sedgwick 2013). Among those studies that did report the proportion approached and recruited from the total ‘eligible’ population, low participation rates were often reported. Therefore, some of the evidence in this review is from selective samples of the population of women who smoke during pregnancy. Women participating in studies (Mullen 1997) were more likely to be in contemplative and preparation stages of change, be ‘recent quitters’ and have a lower gestational age, compared to women not participating studies (Ruggiero 2003). The majority of women categorised as ‘Black’, ‘White’ and ‘Native American’ did enrol in the study, while women categorised as ‘Hispanic’ were less likely (51.6%) to enrol and the majority of Asian women did not enrol (Ruggiero 2003). Dissemination trials and ‘cluster trials’ that randomise clinics or providers are therefore likely to provide a more accurate estimate of the likely effect in a non-selective population of pregnant women.

Secondly, the implementation of interventions under conditions less stringent than an individually-randomised controlled trial may be reduced, which may limit exposure of the intervention group to the intervention, or components of the interventions (Walsh 2000). Several trials implemented in routine care settings by midwives (Moore 2002; DeVries 2006), doctors (Valbo 1994; Walsh 1997), and routine clinic staff (Kendrick 1995) reported difficulties with implementation. Some of the issues included: variable perceptions of smoking cessation as part of the providers' role (DeVries 2006), stating they were too busy and did not have enough time to complete the intervention (Dunkley 1997; Haines 1998; Hajek 2001; Valanis 2001b; Leviton 2003), difficulty recruiting providers to the study (Lawrence 2003), providers reporting pessimism about the efficacy of the intervention (Moore 2002), and lack of acceptability of resources (Lowe 1998a; McBride 1999). Several studies reported positive 'facilitators or enabling factors' associated with implementation. Proposed criteria for interventions to be implemented into routine maternity care include: having program materials readily available; feasible provider time commitments; clear training requirements; minimal organisational and administrative barriers (Strand 2003); and program components that are acceptable to providers and women (Haynes 1998; Cabana 1999; Grol 1999; Walsh 2000; Cooke 2001a). Written resources, a written protocol to identify staff responsibilities, and reimbursement have also been suggested as other strategies to improve implementation (Hartmann 2007). A significant increase in both intervention delivery and smoking outcomes was seen in a cluster trial that supported staff with training based on national guidelines, a clinic management system, and establishment of program boards (Pbert 2004). Suggestions to overcome the barriers in a busy clinic setting included increasing the use of referral services and technology to reduce demand on clinicians’ time (Moore 2002). Subsequently, use of referral services such as ‘quitline’ (Williams 2010) and technology-driven interventions have gained popularity in the past five years (Tsoh 2010; Naughton 2012; Ondersma 2012). In the United Kingdom (UK), most services reported use of ‘quitline’ referral services (Williams 2010). One excluded (non-randomised) study in South Australia (Bowden 2010), describes positive experiences and perceptions of staff in implementing a 'Smoke-free Pregnancy' Project involving brief '5A's' intervention and referrals to ‘quitline’. While use of materials such as self-help materials and technological aids did not appear to significantly increase rates of smoking abstinence in this review, they may help to increase the feasibility and reduce the costs of delivering interventions.

A third possible explanation for the limited effect seen in implementation is that trials that involve broader implementation across the system and provision by usual care providers (effectiveness studies), may result in greater exposure of the comparison group to the intervention. While the difference was not significantly different, the pooled effect size was lower among trials that were assessed as having a high risk of contamination in this review. One study illustrated this effect by including a ‘historical control’ group, in which only 4% stopped smoking, compared to 10% who stopped in the randomised ‘concurrent control’ and 12% in the intervention group who stopped (Windsor 2011).

Institutionalisation, where interventions are part of routine care, is the final stage of the evidence-practice translation process. Australia, Canada, the UK and the United States (US) have developed guidelines recommending all pregnant women receive interventions to promote smoking cessation in pregnancy (Aveyard 2007; Fiore 2008). However, studies of clinicians practice in Canada, the US and Argentina suggest that while the majority (50% to 100%) ‘ask’ about smoking status, rates of assistance with effective strategies to support women to stop smoking are very low (11.5% to below 50%) (Floyd 2001; Hartmann 2007; Tong 2008; Mejia 2010; Okoli 2010). Strategies to address the deficiencies identified in these surveys are reported (Chapin 2004) and several studies in this review have trialled strategies to adapt these guidelines and improve implementation into routine settings (Tsoh 2010; Ondersma 2012). A recent survey suggests attitudes may be shifting in the UK about the provision of advice and support, but not the efficacy of the interventions (Beenstock 2012). A recent survey of women giving birth in Australia suggests there has been a significant increase in the provision of smoking advice and support in routine pregnancy care from 2000 to 2008, though half of smokers still did not receive the full complement of advice and support according to state guidelines, and there was marked variability according to where and from whom women received antenatal care (Perlen 2013).

Strategies to increase disclosure of smoking status

Barriers to implementation have been identified at each step of service provision in relation to support for smoking cessation in pregnancy. This includes detection of women who smoke so they can then be offered a supportive intervention (Tappin 2010). As previously noted, self-reported disclosure of smoking status can be variable. Disclosure is influenced by several factors, including the stigma and guilt associated with smoking in pregnancy, the relationship between the care provider and the way the woman is asked about smoking. In general, it appears that less direct questioning increases disclosure, for example, changing the question format from ‘yes’ or ‘no’ to a series of multiple choice questions and asking women to best describe their smoking status (Mullen 1991). There is some evidence from the literature around broader substance use in pregnancy, that asking about substance use of family members (e.g. secondhand smoke exposure) first (Chasnoff 2005; Chasnoff 2007), and leaving sensitive probing personal questions until later in the interview, when a rapport has been established. The rationale is that this provides an opportunity for the woman to gauge the response of the healthcare provider and feel more confident disclosing her smoking status. In the UK, ‘opt out’ carbon monoxide screening has been proposed to increase disclosure (Tappin 2010; Bauld 2012). Biochemical validation of smoking status is an understandable pre-requisite prior to receipt of contingent incentives, to provide feedback on cotinine levels as a motivational aid; or in the context of a smoking trial. However, the benefits and rationale for not accepting women’s disclosure outside these contexts is unclear and was not well received by women in this review (Thornton 1997). Furthermore, there are questions about the accuracy of carbon monoxide monitoring among women with high secondhand smoke exposure (McLaren 2010), and whether there are any adverse effects from routine screening, such as increased domestic violence or effects on mental health.

Adverse effects of interventions

While psychosocial interventions do not pose the same risks to fetal health as pharmacological agents in pregnancy, there are concerns about the potential unintended consequences of these interventions that aim to encourage pregnant women to stop smoking (Burgess 2009). The potential adverse effects identified in this review include: increased smoking; unhelpful peer or partner support; stigmatisation; and nicotine withdrawal.

Despite the number of studies reporting smoking reduction, only three studies reported rates of women who increased smoking by intervention group, and these showed mixed results (Hjalmarson 1991; Haug 1994; Tappin 2005). It would be helpful for studies to measure any increased smoking, particularly in light of recent qualitative evidence that suggests anti-smoking advice may increase resistance to smoking messages for some women (Bond 2012; Flemming 2013).

There has been an increasing focus on the partners and peers of pregnant women, with the additional aim of facilitating cessation by the women themselves (Stanton 2004; Gage 2007). In some cases this reflects cultural and demographic patterns of smoking, where smoking rates are still highest amongst men (Loke 2005; Kazemi 2012); in others, interest in environmental barriers that hinder smoking cessation has led to an understanding of the influence of a woman’s social networks on smoking behaviour (McBride 2004). Studies in this review suggest that there are both positive and negative aspects to partner and peer assistance with supporting women to stop smoking in pregnancy (McBride 2004; Hennrikus 2010). This legitimises concerns about the potential adverse effects on relationships and women’s position (Greaves 2007a). Therefore, these risks should be taken into consideration when developing interventions involving partners or peers, particularly in subpopulations or regions where protection for women’s rights are less than optimal. Pro-active measures to identify women at risk and ensure their safety should be implemented as part of interventions involving peer or partner support (Greaves 2007b).

No studies measured the impact of interventions on stigmatisation of women. However, studies of psychological impact do not suggest there are any negative effects, and individual psychological support may be beneficial (Stotts 2004; Bullock 2009; Cinciripini 2010). Nevertheless, public health professionals must remain ever vigilant when implementing population-based measures, as policies can disrupt highly complex systems and unintended consequences of tobacco policy may differentially impact on vulnerable population groups (Healton 2009). Stigmatisation research suggests that such policies may have unanticipated outcomes for vulnerable mothers, including decreased mental health; increased use of alcohol or cigarettes; avoidance or delay in seeking medical care; and poorer treatment by health professionals (Moore 2009). This stigmatisation may be compounded for some population groups, such as racial minority groups (Bond 2012; Flemming 2013).

Few studies reported the effect of nicotine withdrawal, which is a gap given that these withdrawal effects may be more acute during pregnancy (Ussher 2012a; Ussher 2012b).

Overall completeness and applicability of evidence

Most of the included studies were carried out in high-income countries and it is not clear whether the results are applicable in other contexts. Given the rapidly evolving nature of the smoking epidemic in low- to middle-income countries, this is a major gap in the current body of evidence.

Many of the studies that recruited individual women did not provide information on the number of women who were eligible for inclusion or were approached to take part in trials (i.e. the participation rate), which would have provided useful information about the general ‘acceptability’ of the intervention, as well as the degree of ‘selection bias’ in the study population (Sedgwick 2013). Among those studies that did report the proportion approached and recruited from the total ‘eligible’ population, low participation rates were often reported. Therefore, some of the evidence in this review is from selective samples of the population of women who smoke during pregnancy and may affect the applicability of the evidence into routine settings.

The review includes a relatively large number of studies focusing on educational and counselling interventions but relatively few focusing on other approaches, such as the use of incentives and peer support. Furthermore, there are limited data for some outcomes (e.g. some perinatal outcomes, family functioning).

Quality of the evidence

The studies included in the review were of mixed quality and there is a substantial level of heterogeneity amongst the trial results (I2 often greater than 50%); hence, we would emphasise the need to consider the Risk of bias' tables and urge caution when interpreting the combined effect of the interventions.

Potential biases in the review process

The timing of the final antenatal assessment of smoking status varied considerably among trials between the second and third trimester. This may affect the amount of time the participants were exposed to the intervention (if it involved ongoing support), as well as the number of those lost to follow-up and measurement of perinatal outcomes.

Agreements and disagreements with other studies or reviews

Agreements and disagreements with the previous review

There have been significant changes in the inclusion criteria for this update, with the ‘splitting’ of the previous review into pharmacological interventions (Coleman 2012b), and the exclusion of quasi-randomised trials. In this update we have changed the outcome from continued smoking (odds ratio), to quitting (risk ratio) so it is consistent with other Cochrane reviews from the Tobacco Addiction Group, and we have included ‘number needed to treat for benefit’ analyses, as this is likely to be of greater relevance to service providers. In this update we have also revised all data extraction to ensure that missing data and 'Risk of bias' assessments from all trials have been dealt with consistently across the five updates, so there are some minor amendments to some trial data from previous versions. However, the major findings from this review are similar to the previous review, with minor differences in effect size estimates, namely:

  • psychosocial interventions which include counselling, incentives and feedback support women to stop smoking in pregnancy are effective in supporting women to quit, reducing low birthweight infants and preterm births;

  • interventions including use of incentives continue to have the largest effect size estimate, but the sample size is very small so these results should be interpreted with caution.

The main differences from the previous review are that a significant effect was demonstrated in:·

  • continued abstinence in the postpartum period.

A significant effect was not demonstrated in:

  • a new subcategory of trials providing ‘health education’ only;

  • a new subcategory of trials using social support, although a significant effect was seen in the combined results of trials using targeted peer support, but not in the single trial using partner-assisted support.

Agreements and disagreements with other Cochrane reviews

See Appendix 1 for a full list of other reviews of smoking interventions.

Pharmacological interventions in pregnancy

A review of pharmacological interventions to support women to stop smoking in pregnancy (Coleman 2012b) did not report a significant effect (RR 1.33, 95% CI 0.93 to 1.91) http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD010078/abstract.

Effects of types of interventions for the general population
Relapse prevention

The findings in this review of a significant effect on relapse prevention in the early postpartum period contrast to findings in another Cochrane review of relapse prevention (Hajek 2009). However, relapse prevention interventions for women who had spontaneously quit in this review did not demonstrate a significant effect, which is similar to the findings of Hajek 2009. http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD003999.pub3/abstract.

Enhanced partner support

The findings in this review were similar to findings in a review of enhanced partner support in the general population (Park 2012), which did not demonstrate a significant effect (RR 0.99, 95% CI 0.84 to 1.15). See http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD002928.pub3/abstract.

Stages of change

A systematic review of stage-based interventions concluded they are no more effective in general than interventions that do not tailor the intervention according to the stage of change (Riemsma 2003). http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD004492.pub4/abstract This is similar to the findings in the previous version of this review.

Individual behavioural support

Our review findings for counselling interventions were similar to those reported by Lancaster 2005a in a review of individual interventions (RR 1.39, 95% CI 1.24 to 1.57), with little difference between intensive support and brief interventions. See http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD001292.pub2/abstract.

Self-help materials

Our review findings were different from a review of provision of self-help materials in the general population (Lancaster 2005b) that demonstrated a modest but significant effect (RR 1.21, 95% CI 1.05 to 1.39), particularly when the materials were tailored (RR 1.31, 95% CI 1.20 to 1.42). See http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD001118.pub2/abstract.

Competitions and incentives

The findings of our review contrast with findings of a review of incentives among the general population (Cahill 2011a) that showed no significant difference. See http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD004307.pub4/abstract. Given the subgroup analysis in our study is based on a very small number of studies and participants, our results should be viewed with caution.

Effects of interventions among other population groups
Psychosocial interventions among patients with coronary heart disease

The findings of this review are similar to findings of psychosocial interventions among patients with coronary heart disease (Barth 2008), another population with strong motivational factors to stop smoking (odds ratio (OR) 1.66, 95% CI 1.25 to 2.22), with high heterogeneity, and a reduced effect among validated smoking outcomes (OR 1.44, 95% CI 0.99 to 2.11).

Pre-operative interventions

The effect of brief smoking cessation interventions among the patients preparing for surgery was similar to our review (RR 1.41, 95% CI 1.22 to 1.63), although the effect of intensive interventions was significantly higher than in our review. See http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD002294.pub3/abstract.

Hospitalised patients

Our results were similar to those among hospitalised patients (RR 1.37, 95% CI 1.27 to 1.48). See http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD001837.pub3/abstract.

Interventions in Indigenous populations

The findings of our review were in contrast to a review of four studies of non-pregnant Indigenous communities (Carson 2012) in New Zealand (2), United States (1) and Australia (1) that reported a modest but significant effect using psychosocial interventions, two of which were supplemented with pharmacological therapy.

Authors' conclusions

Implications for practice

Psychosocial interventions can support women to stop smoking in pregnancy, and reduce preterm births and infants born low birthweight. Therefore, psychosocial support to stop smoking should be considered for women who are pregnant, or seeking to become pregnant. Contrary to concerns that women may be upset by offering support to stop smoking, studies in this review suggest women expect and appreciate the support, and interventions are more likely to improve women's psychological wellbeing than worsen it. Qualitative evidence suggests this support should be positive, not punitive (Bond 2012), and is sensitive to potential feelings of guilt and worry, and concerns about the impact of quitting on women's lives and their relationship with significant others (Flemming 2013). Burgess 2009 suggests it may help for healthcare providers to become aware of any of their own biases against mothers who smoke.

Evidence from this review suggests provision of health education and risk advice is not sufficient, and any psychosocial support should include multiple or tailored intervention components that provide help with strategies to quit, positive encouragement and other strategies, such as incentives, feedback or peer support. Partner support does not appear to be effective from the single study in this review, and care is needed when including peer or partner-support components, as some peer and/or partner-support behaviours may be unhelpful, and may potentially expose vulnerable women to increased risk. Inclusion of support for breastfeeding and prevention of weight gain should also be considered as part of smoking interventions for pregnant women, as obesity has overtaken smoking as a major cause of preterm births in high-income countries (Flenady 2011). Given the high co-morbidity with psychological symptoms and the potential to improve psychological wellbeing, interventions that include psychological support for women with symptoms should be considered. Studies in this review suggest many women resume smoking after pregnancy, so consideration should be given to messages that reinforce the benefits for the mother, rather than solely focusing on benefits for the infant.

There is limited evidence from this review that increasing the intensity of the intervention corresponds to an increased effect size. Therefore, consideration should be given to the quality of the intervention, and providing support that is convenient for women and does not unnecessarily overburden them. Consultation with women and local piloting of programs shown elsewhere to be effective may be a good place to begin to develop strategies suitable for each population. Additionally consultative processes that involve healthcare providers and organisational leaders should be another important consideration for implementation.

Given the clear difficulties which most women still smoking at the first antenatal visit have in stopping smoking, population-wide strategies for smoking control in the whole community are needed to reduce the initiation of smoking by young women: action to prevent sales of tobacco products to young people, prohibition of smoking in all public places, increases in tobacco taxation, workplace smoking cessation programs and bans on tobacco sponsorship (WHO 2008a). However, these interventions should incorporate strategies to reduce risks identified in this review, including stigmatisation, and negative effects on relationships; avoid singling out mothers and focus more broadly on 'parents'; avoid depicting mothers who smoke as 'harming' their infants, but as women who are important in their own right; and assisting vulnerable women to develop alternative 'coping' strategies to deal with living in difficult circumstances (Burgess 2009). Given the strong association between social inequality and continued smoking by pregnant women shown in this review, there is a rationale to support WHO recommendations to reduce social inequalities in the wider community (WHO 2008b).

Implications for research

There is little doubt about ‘whether’ psychosocial interventions are effective in reducing smoking, preterm births or infants born with low birthweight. What is not clear is ‘which’ interventions are effective, ‘how’ these interventions work, ‘who for’ and ‘how’ should these interventions should be implemented, disseminated and institutionalised. As smoking rates have decreased in the general population in high-income countries, it is becoming increasingly recognised that smoking has become more closely correlated with entrenched social disadvantage and psychological co-morbidity (Shoff 2013). Studies are needed that refine interventions to address the specific needs of these subpopulations, without compounding problems of social alienation and low self-efficacy. Given the shifting demographics and burden of diseases from tobacco smoking from high- to low- and middle-income countries, more research is needed to develop strategies which are appropriate for these settings. In reflecting on whether the objectives of this review have been addressed, the authors feel that further research is needed into:

  • the feasibility and effectiveness of interventions in low- and middle-income countries, particularly given the aggressive tobacco marketing in these regions;

  • how to implement and disseminate interventions into routine care, and measures of whether they are effective when implemented at a population level;

  • the feasibility and effectiveness of the use of incentives to support pregnant women to quit smoking, including evaluation of any adverse effects or negative unforeseen circumstances for pregnant women or the broader community;

  • demonstrating effective interventions, including descriptions of how these were developed, to support ethnic and aboriginal women, and young women to stop smoking;

  • interventions to support women with mental illness to stop smoking, and whether interventions that improve mental health can also help women to quit smoking;

  • developing strategies to ensure that smoking interventions do not have a negative impact on breastfeeding, which would counteract some of the health benefits of quitting smoking for both the mother and her infant;

  • whether the timing of the psychosocial support is important, for instance, is more frequent support required in the early stages of quitting and less frequent support required later?

A WHO expert working group (Hunt 2012) recently recommended research in three areas to help reduce smoking during pregnancy:

  • social and cultural factors influencing pregnant women’s use of tobacco and exposure to secondhand smoke;

  • interventions to promote tobacco cessation and reduce secondhand smoke exposure during pregnancy in high-, low- and middle-income countries;

  • describing non-cigarette tobacco use by women and characterising the resulting risks for adverse pregnancy outcomes.

In 2009 the National Institute of Clinical Excellence developed guidance on Quitting smoking in pregnancy and following childbirth. Background documents for this guidance (Bauld 2010a; Williams 2010) identified a number of gaps in existing evidence, including:

  • whether the way the intervention is delivered influences the effect;

  • whether the site or setting influence the effect;

  • evidence of effective interventions for vulnerable population groups, including teenage mothers, disabled mothers, women with mental illness, and other women.

Future trials need to include the following elements:

  • number of potentially eligible women and number agreeing to participate, as this can help to assess the degree of selection bias in the trial and the potential acceptability and generalisability if implemented at a population level;

  • strategies to minimise contamination, as this appears to have an impact on the effect size;

  • a description of the intervention in sufficient detail for its replication even if the detail requires a separate paper;

  • process data as evidence of implementation;

  • women’s views of the intervention, particularly if partner or peer support are incorporated;

  • biochemical validation of non-smoking status;

  • nicotine withdrawal and adverse effects such as increased smoking, or disengagement with services;

  • the collection of perinatal outcome data on birthweight, preterm birth and perinatal deaths, particularly for nicotine replacement therapy trials;

  • collection of outcome data on breastfeeding, weight gain, operative delivery, maternal psychological wellbeing, and the perceived impact of the intervention on family functioning or other significant relationships;

  • subgroup analysis by vulnerabilities (to enable an equity analysis);

  • the impact factor or intra-cluster correlation needs to be reported, in order to assess the effect of clustering and include cluster-randomised trials in meta-analysis.

Acknowledgements

This review represents decades of dedicated work led by Professor Judith Lumley to improve the health of women and children, up until her recent retirement. We thank all previous contributors (outlined below), and acknowledged in previous versions.

This update of the review was jointly funded by the World Health Organization, Australian Government Department of Health and Ageing, the National Health Service (UK), and the EPPI –Centre. This financial support has been greatly appreciated and enabled timely submission.

We thank Ms Josephine Kavanagh and Dr Katy Sutcliffe from the EPPI-Centre for extracting data for the update.

We thank the Cochrane Pregnancy and Childibirth Group's editorial team for conducting the search and facilitating funding support for this review update. We thank Ms Sharon Kramer, from the Australasian Cochrane Centre for all her advice and assistance, including setting up the main comparisons.

We are very grateful to Professor Guadalupe X Ayala from the San Diego State University and Mrs Lorena Fromberg for translating a PhD thesis written in Spanish (Vilches 2009).

The National Institute for Health Research (NIHR) is the largest single funder of the Cochrane Pregnancy and Childbirth Group. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the NIHR, NHS or the Department of Health.

Data and analyses

Download statistical data

Comparison 1. Smoking cessation interventions: counselling vs usual care
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy2711979Risk Ratio (M-H, Random, 95% CI)1.44 [1.19, 1.75]
1.1 Single interventions103753Risk Ratio (M-H, Random, 95% CI)1.12 [0.89, 1.42]
1.2 Multiple interventions114407Risk Ratio (M-H, Random, 95% CI)1.59 [1.15, 2.21]
1.3 Tailored interventions63819Risk Ratio (M-H, Random, 95% CI)1.49 [1.01, 2.20]
2 Abstinence in late pregnancy: biochemically validated only189250Risk Ratio (M-H, Random, 95% CI)1.25 [1.03, 1.50]
2.1 Single interventions73413Risk Ratio (M-H, Random, 95% CI)1.03 [0.85, 1.25]
2.2 Multiple interventions73860Risk Ratio (M-H, Random, 95% CI)1.39 [0.94, 2.04]
2.3 Tailored interventions41977Risk Ratio (M-H, Random, 95% CI)1.42 [0.84, 2.41]
3 Continued abstinence (relapse prevention) in late pregnancy for spontaneous quitters8688Risk Ratio (M-H, Random, 95% CI)1.06 [0.93, 1.21]
3.1 Single interventions2100Risk Ratio (M-H, Random, 95% CI)1.00 [0.93, 1.07]
3.2 Multiple interventions3297Risk Ratio (M-H, Random, 95% CI)1.08 [0.93, 1.26]
3.3 Tailored interventions3291Risk Ratio (M-H, Random, 95% CI)1.19 [0.97, 1.46]
4 Abstinence at 0 to 5 months postpartum10 Risk Ratio (M-H, Random, 95% CI)Subtotals only
4.1 Single interventions51164Risk Ratio (M-H, Random, 95% CI)1.52 [1.13, 2.05]
4.2 Multiple interventions41097Risk Ratio (M-H, Random, 95% CI)2.32 [1.44, 3.72]
4.3 Tailored interventions1367Risk Ratio (M-H, Random, 95% CI)0.88 [0.80, 0.97]
5 Abstinence at 6 to 11 months postpartum62458Risk Ratio (M-H, Random, 95% CI)1.33 [1.00, 1.77]
5.1 Single interventions2776Risk Ratio (M-H, Random, 95% CI)1.34 [0.93, 1.92]
5.2 Multiple interventions31055Risk Ratio (M-H, Random, 95% CI)1.47 [0.86, 2.52]
5.3 Tailored interventions1627Risk Ratio (M-H, Random, 95% CI)0.99 [0.40, 2.46]
6 Abstinence at 12 to 17 months postpartum2431Risk Ratio (M-H, Random, 95% CI)2.20 [1.23, 3.96]
6.1 Single interventions1109Risk Ratio (M-H, Random, 95% CI)2.55 [1.05, 6.21]
6.2 Multiple interventions1322Risk Ratio (M-H, Random, 95% CI)1.97 [0.91, 4.29]
7 Abstinence at 18+ months postpartum2934Risk Ratio (M-H, Random, 95% CI)1.25 [0.57, 2.73]
7.1 Multiple interventions2934Risk Ratio (M-H, Random, 95% CI)1.25 [0.57, 2.73]
8 Reduction in late pregnancy: biochemically validated31311Risk Ratio (M-H, Random, 95% CI)1.11 [0.54, 2.26]
8.1 Single interventions1756Risk Ratio (M-H, Random, 95% CI)0.63 [0.34, 1.20]
8.2 Multiple interventions2555Risk Ratio (M-H, Random, 95% CI)1.50 [0.71, 3.20]
9 Reduction in late pregnancy: self reported (various definitions)2323Risk Ratio (M-H, Random, 95% CI)1.61 [1.06, 2.43]
9.1 Single interventions2323Risk Ratio (M-H, Random, 95% CI)1.61 [1.06, 2.43]
10 Biochemical measures in late pregnancy: mean cotinine31742Std. Mean Difference (IV, Random, 95% CI)-0.05 [-0.14, 0.05]
10.1 Single interventions21328Std. Mean Difference (IV, Random, 95% CI)-0.06 [-0.17, 0.05]
10.2 Multiple interventions1414Std. Mean Difference (IV, Random, 95% CI)-0.01 [-0.21, 0.18]
11 Mean cigarettes per day in late pregnancy93368Std. Mean Difference (IV, Random, 95% CI)-0.25 [-0.46, -0.03]
11.1 Single interventions51928Std. Mean Difference (IV, Random, 95% CI)-0.06 [-0.30, 0.18]
11.2 Multiple interventions2270Std. Mean Difference (IV, Random, 95% CI)-0.60 [-1.02, -0.18]
11.3 Tailored interventions21170Std. Mean Difference (IV, Random, 95% CI)-0.43 [-0.83, -0.03]
12 Low birthweight infants (< 2500 g)63836Risk Ratio (M-H, Random, 95% CI)0.87 [0.70, 1.08]
12.1 Single interventions21460Risk Ratio (M-H, Random, 95% CI)0.79 [0.56, 1.11]
12.2 Multiple interventions1414Risk Ratio (M-H, Random, 95% CI)1.08 [0.45, 2.61]
12.3 Tailored interventions31962Risk Ratio (M-H, Random, 95% CI)0.93 [0.66, 1.32]
13 Very low birthweight infants (< 1500 g)21666Risk Ratio (M-H, Random, 95% CI)1.27 [0.60, 2.71]
13.1 Single interventions1731Risk Ratio (M-H, Random, 95% CI)0.91 [0.32, 2.59]
13.2 Tailored interventions1935Risk Ratio (M-H, Random, 95% CI)1.83 [0.62, 5.43]
14 Preterm births52653Risk Ratio (M-H, Random, 95% CI)0.90 [0.64, 1.27]
14.1 Single interventions31571Risk Ratio (M-H, Random, 95% CI)0.83 [0.60, 1.17]
14.2 Tailored interventions21082Risk Ratio (M-H, Random, 95% CI)1.13 [0.46, 2.80]
15 Mean birthweight94846Mean Difference (IV, Random, 95% CI)36.72 [0.70, 72.74]
15.1 Single interventions41880Mean Difference (IV, Random, 95% CI)45.65 [-10.17, 101.48]
15.2 Multiple interventions2624Mean Difference (IV, Random, 95% CI)84.65 [-95.37, 264.67]
15.3 Tailored interventions32342Mean Difference (IV, Random, 95% CI)23.25 [-52.12, 98.62]
16 Perinatal deaths2 Risk Ratio (M-H, Random, 95% CI)Subtotals only
16.1 Single interventions1130Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
16.2 Tailored interventions1935Risk Ratio (M-H, Random, 95% CI)1.10 [0.52, 2.31]
17 Stillbirths42212Risk Ratio (M-H, Random, 95% CI)1.08 [0.51, 2.30]
17.1 Single interventions2859Risk Ratio (M-H, Random, 95% CI)2.58 [0.38, 17.48]
17.2 Tailored interventions21353Risk Ratio (M-H, Random, 95% CI)0.92 [0.41, 2.10]
18 Neonatal deaths32095Risk Ratio (M-H, Random, 95% CI)2.06 [0.61, 6.92]
18.1 Single interventions1762Risk Ratio (M-H, Random, 95% CI)1.17 [0.07, 18.65]
18.2 Tailored interventions21333Risk Ratio (M-H, Random, 95% CI)2.35 [0.61, 9.07]
19 NICU admissions21140Risk Ratio (M-H, Random, 95% CI)0.82 [0.52, 1.29]
19.1 Single interventions1762Risk Ratio (M-H, Random, 95% CI)0.71 [0.47, 1.07]
19.2 Tailored interventions1378Risk Ratio (M-H, Random, 95% CI)1.17 [0.55, 2.46]
Analysis 1.1.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 1 Abstinence in late pregnancy.

Analysis 1.2.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 2 Abstinence in late pregnancy: biochemically validated only.

Analysis 1.3.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 3 Continued abstinence (relapse prevention) in late pregnancy for spontaneous quitters.

Analysis 1.4.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 4 Abstinence at 0 to 5 months postpartum.

Analysis 1.5.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 5 Abstinence at 6 to 11 months postpartum.

Analysis 1.6.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 6 Abstinence at 12 to 17 months postpartum.

Analysis 1.7.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 7 Abstinence at 18+ months postpartum.

Analysis 1.8.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 8 Reduction in late pregnancy: biochemically validated.

Analysis 1.9.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 9 Reduction in late pregnancy: self reported (various definitions).

Analysis 1.10.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 10 Biochemical measures in late pregnancy: mean cotinine.

Analysis 1.11.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 11 Mean cigarettes per day in late pregnancy.

Analysis 1.12.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 12 Low birthweight infants (< 2500 g).

Analysis 1.13.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 13 Very low birthweight infants (< 1500 g).

Analysis 1.14.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 14 Preterm births.

Analysis 1.15.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 15 Mean birthweight.

Analysis 1.16.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 16 Perinatal deaths.

Analysis 1.17.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 17 Stillbirths.

Analysis 1.18.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 18 Neonatal deaths.

Analysis 1.19.

Comparison 1 Smoking cessation interventions: counselling vs usual care, Outcome 19 NICU admissions.

Comparison 2. Smoking cessation interventions: counselling vs less intensive intervention
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy165247Risk Ratio (M-H, Random, 95% CI)1.35 [1.00, 1.82]
1.1 Single interventions5735Risk Ratio (M-H, Random, 95% CI)1.51 [0.90, 2.54]
1.2 Multiple interventions104260Risk Ratio (M-H, Random, 95% CI)1.23 [0.84, 1.78]
1.3 Tailored interventions1252Risk Ratio (M-H, Random, 95% CI)2.39 [1.03, 5.56]
2 Abstinence in late pregnancy: biochemically validated only122858Risk Ratio (M-H, Random, 95% CI)1.46 [1.15, 1.85]
2.1 Single interventions5735Risk Ratio (M-H, Random, 95% CI)1.51 [0.90, 2.54]
2.2 Multiple interventions61871Risk Ratio (M-H, Random, 95% CI)1.38 [1.05, 1.80]
2.3 Tailored interventions1252Risk Ratio (M-H, Random, 95% CI)2.39 [1.03, 5.56]
3 Continued abstinence (relapse prevention) in late pregnancy (spontaneous quitters)4692Risk Ratio (M-H, Random, 95% CI)1.05 [0.98, 1.13]
3.1 Single interventions2204Risk Ratio (M-H, Random, 95% CI)1.02 [0.88, 1.18]
3.2 Multiple interventions2488Risk Ratio (M-H, Random, 95% CI)1.06 [0.96, 1.17]
3.3 Tailored interventions00Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
4 Abstinence at 0 to 5 months postpartum61980Risk Ratio (M-H, Random, 95% CI)1.17 [0.82, 1.66]
4.1 Single interventions182Risk Ratio (M-H, Random, 95% CI)0.63 [0.11, 3.60]
4.2 Multiple interventions41646Risk Ratio (M-H, Random, 95% CI)1.19 [0.99, 1.43]
4.3 Tailored interventions1252Risk Ratio (M-H, Random, 95% CI)12.80 [1.70, 96.35]
5 Abstinence at 6 to 11 months postpartum31271Risk Ratio (M-H, Random, 95% CI)1.08 [0.83, 1.40]
5.1 Single interventions1105Risk Ratio (M-H, Random, 95% CI)2.45 [0.50, 12.08]
5.2 Multiple interventions21166Risk Ratio (M-H, Random, 95% CI)1.05 [0.80, 1.38]
6 Abstinence at 12 to 17 months postpartum21188Risk Ratio (M-H, Random, 95% CI)1.25 [0.71, 2.20]
6.1 Multiple interventions21188Risk Ratio (M-H, Random, 95% CI)1.25 [0.71, 2.20]
7 Reduction in late pregnancy: self-reported > 50%21235Risk Ratio (M-H, Random, 95% CI)1.35 [1.07, 1.71]
7.1 Multiple interventions21235Risk Ratio (M-H, Random, 95% CI)1.35 [1.07, 1.71]
8 Reduction in late pregnancy: biochemically validated2857Risk Ratio (M-H, Random, 95% CI)1.35 [0.98, 1.87]
8.1 Multiple interventions2857Risk Ratio (M-H, Random, 95% CI)1.35 [0.98, 1.87]
9 Mean cigarettes per day in late pregnancy2397Std. Mean Difference (IV, Random, 95% CI)-0.11 [-0.30, 0.09]
9.1 Single interventions1121Std. Mean Difference (IV, Random, 95% CI)0.01 [-0.34, 0.37]
9.2 Multiple interventions1276Std. Mean Difference (IV, Random, 95% CI)-0.16 [-0.40, 0.08]
10 Low birthweight infants (< 2500 g)2503Risk Ratio (M-H, Random, 95% CI)0.58 [0.32, 1.04]
10.1 Single interventions1227Risk Ratio (M-H, Random, 95% CI)0.55 [0.25, 1.21]
10.2 Multiple interventions1276Risk Ratio (M-H, Random, 95% CI)0.61 [0.25, 1.50]
11 Preterm births3794Risk Ratio (M-H, Random, 95% CI)0.82 [0.47, 1.42]
11.1 Single interventions1227Risk Ratio (M-H, Random, 95% CI)0.55 [0.25, 1.21]
11.2 Multiple interventions1308Risk Ratio (M-H, Random, 95% CI)1.17 [0.46, 2.95]
11.3 Tailored interventions1259Risk Ratio (M-H, Random, 95% CI)1.30 [0.30, 5.71]
12 Mean birthweight3546Mean Difference (IV, Random, 95% CI)56.02 [-31.46, 143.50]
12.1 Single interventions1227Mean Difference (IV, Random, 95% CI)57.00 [-93.50, 207.50]
12.2 Multiple interventions2319Mean Difference (IV, Random, 95% CI)76.01 [-88.59, 240.61]
Analysis 2.1.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 1 Abstinence in late pregnancy.

Analysis 2.2.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 2 Abstinence in late pregnancy: biochemically validated only.

Analysis 2.3.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 3 Continued abstinence (relapse prevention) in late pregnancy (spontaneous quitters).

Analysis 2.4.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 4 Abstinence at 0 to 5 months postpartum.

Analysis 2.5.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 5 Abstinence at 6 to 11 months postpartum.

Analysis 2.6.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 6 Abstinence at 12 to 17 months postpartum.

Analysis 2.7.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 7 Reduction in late pregnancy: self-reported > 50%.

Analysis 2.8.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 8 Reduction in late pregnancy: biochemically validated.

Analysis 2.9.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 9 Mean cigarettes per day in late pregnancy.

Analysis 2.10.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 10 Low birthweight infants (< 2500 g).

Analysis 2.11.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 11 Preterm births.

Analysis 2.12.

Comparison 2 Smoking cessation interventions: counselling vs less intensive intervention, Outcome 12 Mean birthweight.

Comparison 3. Smoking cessation interventions: health education vs usual care
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy3374Risk Ratio (M-H, Random, 95% CI)1.51 [0.64, 3.59]
1.1 Single interventions2229Risk Ratio (M-H, Random, 95% CI)1.29 [0.49, 3.42]
1.2 Multiple interventions1145Risk Ratio (M-H, Random, 95% CI)4.06 [0.46, 35.41]
2 Abstinence in late pregnancy: biochemically validated only2229Risk Ratio (M-H, Random, 95% CI)1.29 [0.49, 3.42]
2.1 Single interventions2229Risk Ratio (M-H, Random, 95% CI)1.29 [0.49, 3.42]
3 Mean cigarettes per day in late pregnancy2 Std. Mean Difference (IV, Random, 95% CI)Subtotals only
3.1 Single interventions1552Std. Mean Difference (IV, Random, 95% CI)-0.72 [-0.89, -0.55]
3.2 Multiple interventions1135Std. Mean Difference (IV, Random, 95% CI)-0.32 [-0.66, 0.02]
Analysis 3.1.

Comparison 3 Smoking cessation interventions: health education vs usual care, Outcome 1 Abstinence in late pregnancy.

Analysis 3.2.

Comparison 3 Smoking cessation interventions: health education vs usual care, Outcome 2 Abstinence in late pregnancy: biochemically validated only.

Analysis 3.3.

Comparison 3 Smoking cessation interventions: health education vs usual care, Outcome 3 Mean cigarettes per day in late pregnancy.

Comparison 4. Smoking cessation interventions: health education vs less intensive intervention
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy: biochemically validated2851Risk Ratio (M-H, Random, 95% CI)1.50 [0.97, 2.31]
1.1 Single interventions1653Risk Ratio (M-H, Random, 95% CI)1.46 [0.88, 2.43]
1.2 Multiple interventions1198Risk Ratio (M-H, Random, 95% CI)1.59 [0.68, 3.73]
2 Abstinence at 0 to 5 months postpartum2 Risk Ratio (M-H, Random, 95% CI)Subtotals only
2.1 Single interventions2844Risk Ratio (M-H, Random, 95% CI)1.29 [0.52, 3.22]
Analysis 4.1.

Comparison 4 Smoking cessation interventions: health education vs less intensive intervention, Outcome 1 Abstinence in late pregnancy: biochemically validated.

Analysis 4.2.

Comparison 4 Smoking cessation interventions: health education vs less intensive intervention, Outcome 2 Abstinence at 0 to 5 months postpartum.

Comparison 5. Smoking cessation interventions: feedback vs usual care
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy2355Risk Ratio (M-H, Random, 95% CI)4.39 [1.89, 10.21]
1.1 Multiple interventions2355Risk Ratio (M-H, Random, 95% CI)4.39 [1.89, 10.21]
2 Reduction in late pregnancy: various definitions2355Risk Ratio (M-H, Random, 95% CI)1.69 [1.24, 2.31]
2.1 Multiple interventions2355Risk Ratio (M-H, Random, 95% CI)1.69 [1.24, 2.31]
3 Preterm births23111Risk Ratio (M-H, Random, 95% CI)0.60 [0.28, 1.29]
3.1 Multiple interventions23111Risk Ratio (M-H, Random, 95% CI)0.60 [0.28, 1.29]
4 Mean birthweight23006Mean Difference (IV, Random, 95% CI)79.43 [-53.05, 211.91]
4.1 Multiple interventions23006Mean Difference (IV, Random, 95% CI)79.43 [-53.05, 211.91]
5 Stillbirths22960Risk Ratio (M-H, Random, 95% CI)1.28 [0.69, 2.39]
5.1 Multiple interventions22960Risk Ratio (M-H, Random, 95% CI)1.28 [0.69, 2.39]
Analysis 5.1.

Comparison 5 Smoking cessation interventions: feedback vs usual care, Outcome 1 Abstinence in late pregnancy.

Analysis 5.2.

Comparison 5 Smoking cessation interventions: feedback vs usual care, Outcome 2 Reduction in late pregnancy: various definitions.

Analysis 5.3.

Comparison 5 Smoking cessation interventions: feedback vs usual care, Outcome 3 Preterm births.

Analysis 5.4.

Comparison 5 Smoking cessation interventions: feedback vs usual care, Outcome 4 Mean birthweight.

Analysis 5.5.

Comparison 5 Smoking cessation interventions: feedback vs usual care, Outcome 5 Stillbirths.

Comparison 6. Smoking cessation interventions: feedback vs less intensive intervention
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy: biochemically validated2319Risk Ratio (M-H, Random, 95% CI)1.19 [0.45, 3.12]
1.1 Single interventions179Risk Ratio (M-H, Random, 95% CI)0.60 [0.16, 2.22]
1.2 Multiple interventions1240Risk Ratio (M-H, Random, 95% CI)1.69 [0.89, 3.20]
Analysis 6.1.

Comparison 6 Smoking cessation interventions: feedback vs less intensive intervention, Outcome 1 Abstinence in late pregnancy: biochemically validated.

Comparison 7. Smoking cessation interventions: incentives vs usual care
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy:biochemically validated2129Risk Ratio (M-H, Random, 95% CI)3.59 [0.10, 130.49]
1.1 Single interventions174Risk Ratio (M-H, Random, 95% CI)20.72 [1.28, 336.01]
1.2 Tailored interventions155Risk Ratio (M-H, Random, 95% CI)0.90 [0.25, 3.23]
Analysis 7.1.

Comparison 7 Smoking cessation interventions: incentives vs usual care, Outcome 1 Abstinence in late pregnancy:biochemically validated.

Comparison 8. Smoking cessation interventions: social support vs less intensive intervention
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy (peer and partner support)6734Risk Ratio (M-H, Random, 95% CI)1.29 [0.94, 1.78]
1.1 Single interventions2224Risk Ratio (M-H, Random, 95% CI)1.35 [0.57, 3.18]
1.2 Multiple interventions3359Risk Ratio (M-H, Random, 95% CI)1.48 [0.74, 2.95]
1.3 Tailored interventions1151Risk Ratio (M-H, Random, 95% CI)1.22 [0.59, 2.52]
2 Abstinence in late pregnancy: biochemically validated (peer support only)5554Risk Ratio (M-H, Random, 95% CI)1.49 [1.01, 2.19]
2.1 Single interventions2224Risk Ratio (M-H, Random, 95% CI)1.35 [0.57, 3.18]
2.2 Multiple interventions2179Risk Ratio (M-H, Random, 95% CI)2.26 [1.15, 4.46]
2.3 Tailored interventions1151Risk Ratio (M-H, Random, 95% CI)1.22 [0.59, 2.52]
3 Abstinence at 0 to 5 months postpartum2473Risk Ratio (M-H, Random, 95% CI)1.36 [0.46, 4.07]
3.1 Single interventions182Risk Ratio (M-H, Random, 95% CI)5.8 [0.33, 101.27]
3.2 Multiple interventions1391Risk Ratio (M-H, Random, 95% CI)1.11 [0.87, 1.41]
4 Abstinence at 6 to 11 months postpartum2486Risk Ratio (M-H, Random, 95% CI)1.09 [0.83, 1.42]
4.1 Multiple interventions2486Risk Ratio (M-H, Random, 95% CI)1.09 [0.83, 1.42]
Analysis 8.1.

Comparison 8 Smoking cessation interventions: social support vs less intensive intervention, Outcome 1 Abstinence in late pregnancy (peer and partner support).

Analysis 8.2.

Comparison 8 Smoking cessation interventions: social support vs less intensive intervention, Outcome 2 Abstinence in late pregnancy: biochemically validated (peer support only).

Analysis 8.3.

Comparison 8 Smoking cessation interventions: social support vs less intensive intervention, Outcome 3 Abstinence at 0 to 5 months postpartum.

Analysis 8.4.

Comparison 8 Smoking cessation interventions: social support vs less intensive intervention, Outcome 4 Abstinence at 6 to 11 months postpartum.

Comparison 9. Maternal health intervention with smoking cessation component: social support (tailored) vs usual care
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy2 Risk Ratio (M-H, Random, 95% CI)Subtotals only
1.1 Self-reported1492Risk Ratio (M-H, Random, 95% CI)1.83 [1.22, 2.73]
1.2 Biochemically validated1141Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
2 Self-reported mean cigarettes per day in late pregnancy2542Std. Mean Difference (IV, Random, 95% CI)-0.28 [-0.45, -0.11]
2.1 Self-reported1401Std. Mean Difference (IV, Random, 95% CI)-0.24 [-0.43, -0.04]
2.2 Biochemically validated1141Std. Mean Difference (IV, Random, 95% CI)-0.40 [-0.73, -0.06]
Analysis 9.1.

Comparison 9 Maternal health intervention with smoking cessation component: social support (tailored) vs usual care, Outcome 1 Abstinence in late pregnancy.

Analysis 9.2.

Comparison 9 Maternal health intervention with smoking cessation component: social support (tailored) vs usual care, Outcome 2 Self-reported mean cigarettes per day in late pregnancy.

Comparison 10. Maternal health intervention with smoking cessation component: social support vs less intensive intervention
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy2316Risk Ratio (M-H, Random, 95% CI)0.80 [0.46, 1.39]
1.1 Single interventions166Risk Ratio (M-H, Random, 95% CI)0.45 [0.09, 2.16]
1.2 Tailored interventions1250Risk Ratio (M-H, Random, 95% CI)0.86 [0.48, 1.57]
2 Abstinence in late pregnancy: biochemically validated1250Risk Ratio (M-H, Random, 95% CI)0.86 [0.48, 1.57]
2.1 Tailored interventions1250Risk Ratio (M-H, Random, 95% CI)0.86 [0.48, 1.57]
Analysis 10.1.

Comparison 10 Maternal health intervention with smoking cessation component: social support vs less intensive intervention, Outcome 1 Abstinence in late pregnancy.

Analysis 10.2.

Comparison 10 Maternal health intervention with smoking cessation component: social support vs less intensive intervention, Outcome 2 Abstinence in late pregnancy: biochemically validated.

Comparison 11. Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Abstinence in late pregnancy: self-reported and biochemically validated (non-winsorised)7021948Risk Ratio (M-H, Random, 95% CI)1.45 [1.27, 1.64]
1.1 Counselling4517681Risk Ratio (M-H, Random, 95% CI)1.37 [1.17, 1.59]
1.2 Health education51225Risk Ratio (M-H, Random, 95% CI)1.47 [1.02, 2.13]
1.3 Feedback5739Risk Ratio (M-H, Random, 95% CI)2.09 [1.17, 3.72]
1.4 Incentives4426Risk Ratio (M-H, Random, 95% CI)3.09 [1.34, 7.15]
1.5 Social support101683Risk Ratio (M-H, Random, 95% CI)1.29 [0.97, 1.73]
1.6 Other1194Risk Ratio (M-H, Random, 95% CI)1.63 [0.62, 4.32]
2 Abstinence in late pregnancy: biochemically validated only (non-winsorised)49 Risk Ratio (M-H, Random, 95% CI)Subtotals only
2.1 Counselling3011924Risk Ratio (M-H, Random, 95% CI)1.27 [1.11, 1.47]
2.2 Health education41080Risk Ratio (M-H, Random, 95% CI)1.43 [0.98, 2.08]
2.3 Feedback3563Risk Ratio (M-H, Random, 95% CI)1.70 [0.71, 4.08]
2.4 Incentives4426Risk Ratio (M-H, Random, 95% CI)3.09 [1.34, 7.15]
2.5 Social support7945Risk Ratio (M-H, Random, 95% CI)1.31 [0.90, 1.91]
2.6 Other1194Risk Ratio (M-H, Random, 95% CI)1.63 [0.62, 4.32]
3 Continued abstinence (Relapse prevention) in late pregnancy for spontaneous quitters14 Risk Ratio (M-H, Random, 95% CI)Totals not selected
3.1 Counselling12 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
3.2 Health education1 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
3.3 Social support1 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
4 Abstinence at 0 to 5 months postpartum26 Risk Ratio (M-H, Random, 95% CI)Totals not selected
4.1 Counselling18 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
4.2 Health education3 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
4.3 Incentives2 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
4.4 Social support3 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
5 Abstinence at 6 to 11 months postpartum13 Risk Ratio (M-H, Random, 95% CI)Totals not selected
5.1 Counselling10 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
5.2 Incentives1 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
5.3 Social support2 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
6 Abstinence at 12 to 17 months postpartum5 Risk Ratio (M-H, Random, 95% CI)Totals not selected
6.1 Counselling4 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
6.2 Social support1 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
7 Abstinence at 18+ months postpartum2 Risk Ratio (M-H, Random, 95% CI)Totals not selected
7.1 Counselling2 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
8 Smoking reduction: numbers of women reducing smoking in late pregnancy15 Risk Ratio (M-H, Random, 95% CI)Totals not selected
8.1 Self-reported some reduction in smoking (various definitions)5 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
8.2 Self-reported > 50% reduction in smoking4 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
8.3 Biochemically validated reduction6 Risk Ratio (M-H, Random, 95% CI)0.0 [0.0, 0.0]
9 Smoking reduction: biochemical measures in late pregnancy6 Std. Mean Difference (IV, Random, 95% CI)Totals not selected
9.1 Mean cotinine levels5 Std. Mean Difference (IV, Random, 95% CI)0.0 [0.0, 0.0]
9.2 Mean thiocynate level1 Std. Mean Difference (IV, Random, 95% CI)0.0 [0.0, 0.0]
10 Smoking reduction: self-reported mean cigarettes per day measured in late pregnancy or at delivery20 Std. Mean Difference (IV, Random, 95% CI)Totals not selected
10.1 Counselling11 Std. Mean Difference (IV, Random, 95% CI)0.0 [0.0, 0.0]
10.2 Health education3 Std. Mean Difference (IV, Random, 95% CI)0.0 [0.0, 0.0]
10.3 Feedback2 Std. Mean Difference (IV, Random, 95% CI)0.0 [0.0, 0.0]
10.4 Incentives1 Std. Mean Difference (IV, Random, 95% CI)0.0 [0.0, 0.0]
10.5 Social support3 Std. Mean Difference (IV, Random, 95% CI)0.0 [0.0, 0.0]
11 Low birthweight (under 2500 g)148562Risk Ratio (M-H, Random, 95% CI)0.82 [0.71, 0.94]
11.1 Counselling84339Risk Ratio (M-H, Random, 95% CI)0.83 [0.68, 1.01]
11.2 Health education21172Risk Ratio (M-H, Random, 95% CI)0.87 [0.49, 1.55]
11.3 Feedback12848Risk Ratio (M-H, Random, 95% CI)0.82 [0.63, 1.06]
11.4 Incentives2124Risk Ratio (M-H, Random, 95% CI)0.45 [0.22, 0.93]
11.5 Social support179Risk Ratio (M-H, Random, 95% CI)1.00 [0.33, 2.99]
12 Very low birthweight (under 1500 g)34366Risk Ratio (M-H, Random, 95% CI)1.11 [0.62, 2.01]
12.1 Counselling21666Risk Ratio (M-H, Random, 95% CI)1.27 [0.60, 2.71]
12.2 Feedback12700Risk Ratio (M-H, Random, 95% CI)0.90 [0.35, 2.32]
13 Preterm birth (under 37 weeks)147852Risk Ratio (M-H, Random, 95% CI)0.82 [0.70, 0.96]
13.1 Counselling83447Risk Ratio (M-H, Random, 95% CI)0.93 [0.71, 1.20]
13.2 Health education21170Risk Ratio (M-H, Random, 95% CI)0.92 [0.55, 1.56]
13.3 Feedback23111Risk Ratio (M-H, Random, 95% CI)0.60 [0.28, 1.29]
13.4 Incentives2124Risk Ratio (M-H, Random, 95% CI)0.49 [0.22, 1.08]
14 Mean birthweight199859Mean Difference (IV, Random, 95% CI)40.78 [18.45, 63.10]
14.1 Counselling125392Mean Difference (IV, Random, 95% CI)39.93 [9.12, 70.74]
14.2 Health education21172Mean Difference (IV, Random, 95% CI)27.35 [-53.88, 108.58]
14.3 Feedback23006Mean Difference (IV, Random, 95% CI)79.43 [-53.05, 211.91]
14.4 Incentives2147Mean Difference (IV, Random, 95% CI)213.78 [20.16, 407.40]
14.5 Social support1142Mean Difference (IV, Random, 95% CI)28.0 [-152.48, 208.48]
15 Perinatal deaths44465Risk Ratio (M-H, Random, 95% CI)1.13 [0.72, 1.77]
15.1 Counselling21065Risk Ratio (M-H, Random, 95% CI)1.10 [0.52, 2.31]
15.2 Health education1552Risk Ratio (M-H, Random, 95% CI)4.40 [0.49, 39.08]
15.3 Feedback12848Risk Ratio (M-H, Random, 95% CI)1.05 [0.59, 1.87]
16 Stillbirths75414Risk Ratio (M-H, Random, 95% CI)1.22 [0.76, 1.95]
16.1 Counselling52454Risk Ratio (M-H, Random, 95% CI)1.14 [0.55, 2.33]
16.2 Feedback22960Risk Ratio (M-H, Random, 95% CI)1.28 [0.69, 2.39]
17 Neonatal deaths44905Risk Ratio (M-H, Random, 95% CI)1.15 [0.44, 3.06]
17.1 Counselling32095Risk Ratio (M-H, Random, 95% CI)2.06 [0.61, 6.92]
17.2 Feedback12810Risk Ratio (M-H, Random, 95% CI)0.40 [0.08, 2.07]
18 NICU admissions41264Risk Ratio (M-H, Random, 95% CI)0.78 [0.59, 1.04]
18.1 Counselling21140Risk Ratio (M-H, Random, 95% CI)0.82 [0.52, 1.29]
18.2 Incentives2124Risk Ratio (M-H, Random, 95% CI)0.76 [0.47, 1.21]
Analysis 11.1.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 1 Abstinence in late pregnancy: self-reported and biochemically validated (non-winsorised).

Analysis 11.2.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 2 Abstinence in late pregnancy: biochemically validated only (non-winsorised).

Analysis 11.3.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 3 Continued abstinence (Relapse prevention) in late pregnancy for spontaneous quitters.

Analysis 11.4.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 4 Abstinence at 0 to 5 months postpartum.

Analysis 11.5.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 5 Abstinence at 6 to 11 months postpartum.

Analysis 11.6.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 6 Abstinence at 12 to 17 months postpartum.

Analysis 11.7.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 7 Abstinence at 18+ months postpartum.

Analysis 11.8.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 8 Smoking reduction: numbers of women reducing smoking in late pregnancy.

Analysis 11.9.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 9 Smoking reduction: biochemical measures in late pregnancy.

Analysis 11.10.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 10 Smoking reduction: self-reported mean cigarettes per day measured in late pregnancy or at delivery.

Analysis 11.11.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 11 Low birthweight (under 2500 g).

Analysis 11.12.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 12 Very low birthweight (under 1500 g).

Analysis 11.13.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 13 Preterm birth (under 37 weeks).

Analysis 11.14.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 14 Mean birthweight.

Analysis 11.15.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 15 Perinatal deaths.

Analysis 11.16.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 16 Stillbirths.

Analysis 11.17.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 17 Neonatal deaths.

Analysis 11.18.

Comparison 11 Interventions for smoking cessation in pregnancy versus control: subgrouped by main intervention strategy, Outcome 18 NICU admissions.

Appendices

Appendix 1. Other Cochrane Reviews of smoking cessation interventions

Population-wide measures
Healthcare financing systems for increasing the use of tobacco dependence treatment (Reda 2012)
Impact of tobacco advertising and promotion on increasing adolescent smoking behaviours (Lovato 2011)
Interventions for preventing tobacco sales to minors (Stead 2005a)*
Interventions for preventing tobacco smoking in public places (Serra 2008)*
Legislative smoking bans for reducing secondhand smoke exposure, smoking prevalence and tobacco consumption (Callinan 2010)
Mass media interventions for preventing smoking in young people (Brinn 2010)
Mass media interventions for smoking cessation in adults (Bala 2008)*

Community interventions
Community interventions for preventing smoking in young people (Carson 2011)
Community interventions for reducing smoking among adults (Secker-Walker 2002b)*
Community pharmacy personnel interventions for smoking cessation (Sinclair 2004)*
Family and carer smoking control programmes for reducing children's exposure to environmental tobacco smoke (Priest 2008)*
Family-based programmes for preventing smoking by children and adolescents (Thomas 2007)*
Group behaviour therapy programmes for smoking cessation (Stead 2005b)*
School-based programmes for preventing smoking (Thomas 2013)
School policies for preventing smoking among young people (Coppo 2012)
Workplace interventions for smoking cessation (Cahill 2008)*

Individual strategies
Acupuncture and related interventions for smoking cessation (White 2011)
Aversive smoking for smoking cessation (Hajek 2001b)*
Biomedical risk assessment as an aid for smoking cessation (Bize 2012)
Competitions and incentives for smoking cessation (Cahill 2011a)
Electronic cigarettes for smoking cessation and reduction (McRobbie 2012)
Enhancing partner support to improve smoking cessation (Park 2012)
Exercise interventions for smoking cessation (Ussher 2012a)
Hypnotherapy for smoking cessation (Barnes 2010)
Incentives for preventing smoking in children and adolescents (Johnston 2012)
Individual behavioural counselling for smoking cessation (Lancaster 2005b)*
Internet-based interventions for smoking cessation (Civljak 2010)
Interventions for preventing weight gain after smoking cessation (Farley 2012)
Interventions for smokeless tobacco use cessation (Ebbert 2011)
Interventions for waterpipe smoking cessation (Maziak 2007)*
Mobile phone-based interventions for smoking cessation (Whittaker 2012)
Motivational interviewing for smoking cessation (Lai 2010)
Nursing interventions for smoking cessation (Rice 2008)*
Physician advice for smoking cessation (Stead 2008)
Quit and Win contests for smoking cessation (Cahill 2008c)*
Reduction versus abrupt cessation in smokers who want to quit (Lindson-Hawley 2012)
Relapse prevention interventions for smoking cessation (Hajek 2009)*
Self-help interventions for smoking cessation (Lancaster 2005a)
Stage-based interventions for smoking cessation (Cahill 2010)
Telephone counselling for smoking cessation (Stead 2006b)
Training health professionals in smoking cessation (Carson 2012c)
Use of electronic health records to support smoking cessation (Boyle 2011)

Pharmacotherapies
Antidepressants for smoking cessation (Hughes 2007)
Anxiolytics for smoking cessation (Hughes 2000b)*
Behavioural interventions as adjuncts to pharmacotherapy for smoking cessation (Stead 2012b)
Cannabinoid type 1 receptor antagonists (rimonabant) for smoking cessation (Cahill 2011)
Clonidine for smoking cessation (Gourlay 2004)*
Combined pharmacotherapy and behavioural interventions for smoking cessation (Stead 2012c)
Interventions to increase adherence to medications for tobacco dependence (Hollands 2011)
Lobeline for smoking cessation (Stead 2012d)
Mecamylamine (a nicotine antagonist) for smoking cessation (Lancaster 1998)*
Nicobrevin for smoking cessation (Stead 2006a)
Nicotine receptor partial agonists for smoking cessation (Cahill 2012)
Nicotine replacement therapy for smoking cessation (Stead 2012a)
Nicotine vaccines for smoking cessation (Hartmann-Boyce 2012)
Opioid antagonists for smoking cessation (David 2006)*
Pharmacological interventions for promoting smoking cessation during pregnancy (Coleman 2012b)
Pharmacological interventions for smoking cessation: an overview of reviews (Cahill 2013)
Silver acetate for smoking cessation (Lancaster 2012)

Specific population group interventions
Interventions for preoperative smoking cessation (Thomsen 2010)
Interventions for smoking cessation and reduction in individuals with schizophrenia (Tsoi 2013)
Interventions for smoking cessation in hospitalised patients (Rigotti 2012)
Interventions for smoking cessation in Indigenous populations (Carson 2012b)
Interventions for tobacco cessation in the dental setting (Carr 2012)
Interventions for tobacco use prevention in Indigenous youth (Carson 2012)
Psychosocial interventions for smoking cessation in patients with coronary heart disease (Barth 2008)
Smoking cessation advice for people with serious mental illness (Khanna 2012)
Smoking cessation interventions for smokers with current or past depression (van der Meer 2006)*
Smoking cessation for chronic obstructive pulmonary disease (van der Meer 2001)
Smoking cessation for the secondary prevention of coronary heart disease (Critchley 2012)
Tobacco cessation interventions for young people (Grimshaw 2006)*
Interventions for tobacco use cessation in people in treatment for or recovery from substance abuse (Apollonio 2012)

Other
Interventions for recruiting smokers into cessation programmes (Marcano 2012)
Interventions to reduce harm from continued tobacco use (Stead 2007)

*unchanged from previous update

Appendix 2. Supplementary search strategy

Authors conducted a supplementary search for non-randomised studies, for the background and discussion, in MEDLINE, Embase, PsycLIT, and CINAHL (June 2008 to 1 March 2013). The strategy below is the one used for Embase.

Embase

 1. exp "PARAMETERS CONCERNING THE FETUS, NEWBORN AND PREGNANCY"/ or exp PREGNANCY/ or exp PREGNANCY OUTCOME/ or exp PREGNANCY COMPLICATION/

2. pregnan*.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

3. exp prenatal diagnosis/ or exp prenatal care/

4. antenatal.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

5. prenatal.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

6. exp PRENATAL GROWTH/ or exp PRENATAL STRESS/ or exp PRENATAL DISORDER/ or exp PRENATAL PERIOD/ or exp PRENATAL DIAGNOSIS/ or exp PRENATAL MORTALITY/ or exp PRENATAL DEVELOPMENT/ or exp PRENATAL CARE/ or exp PRENATAL SCREENING/ or exp PRENATAL EXPOSURE/

7. 1 or 2 or 3 or 4 or 5 or 6

8. exp maternal care/

9. matern*.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

10. exp family planning/

11. family planning.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

12. exp PRECONCEPTION INJURY/

13. preconcept*.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

14. 7 or 8 or 9 or 10 or 11 or 12 or 13

15. exp SMOKING/ or exp MATERNAL SMOKING/ or exp CIGARETTE SMOKING/

16. smok*.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

17. exp TOBACCO/

18. tobacco.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

19. 15 or 16 or 17 or 18

20. exp SMOKING CESSATION PROGRAM/ or exp SMOKING CESSATION/

21. cessation.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

22. exp nicotine gum/ or exp nicotine patch/

23. quit*.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

24. stop*.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

25. exp NICOTINE PATCH/ or exp NICOTINE VACCINE/ or exp NICOTINE GUM/ or exp NICOTINE REPLACEMENT THERAPY/

26. 20 or 21 or 22 or 23 or 24 or 25

27. exp NICOTINE/

28. nicotine.mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer]

29. 19 or 27 or 28

30. 14 and 26 and 29

Appendix 3. Data extraction summary

Methods

  • Type of randomisation and number of study arms.

  • Study aims, including whether the smoking cessation component was part of a broader study to improve maternal health, or it was a dissemination study.

  • Country, setting, and year of data collection.

Participants

  • Inclusion and exclusion criteria.

  • Participation rate of eligible study population, as an indicator of selection bias.

  • Description of participants/study population, including pre-pregnancy cigarettes per day, ETS exposure, and partner smoking.

  • PROGRESS-PLUS health equity assessment of participant characteristics and subgroup outcomes (including, socio-economic status, education, ethnicity, social support, place of residence, occupation, young women, or other groups vulnerable to health inequalities or social exclusion, such as women with mental illness).

  • Age range and dropout rate by age.

Interventions

  • A description of the control and intervention(s).

  • Intervention provider.

  • Main intervention strategies (including counselling, health education, feedback, incentives, social support, (as described in 'types of interventions').

  • Intervention coded as single, multiple, or tailored.

  • Comparison group coded as receiving usual care, less intensive intervention, or alternative intervention.

  • Description of any structural elements to support the intervention (material resources, provider education, administrative support or systems, alignment with other campaigns, including national guidelines, and other structural elements).

  • Whether the intervention was provided by a dedicated research team (efficacy study) or routine care providers (effectiveness study).

  • Intensity (frequency and duration) rating of intervention as controls (as detailed below).

Frequency of intervention contact (estimate):

0: ‘Usual Care’.
1: Single contact during/at time of routine pregnancy care visits (but not ‘usual care’) without strategies to quit such as information, education or advice to stop smoking in pregnancy, without any additional support to do so.
2: Single contact, outside of ‘routine’ pregnancy care with strategies to quit such as self-help materials and/or single counselling session with strategies to quit.
3: Two to five contacts to sustain motivation to stop smoking provided during/at time of routine pregnancy care visits (but not ‘usual care’).
4: Two to five contacts to sustain motivation to stop smoking provided outside of routine care (such as phone calls, letters, reminders, peer support, incentives, optional NRT, feedback).
5: Greater than five contacts to sustain motivation to stop smoking provided during/at time of routine pregnancy care visits (but not ‘usual care’).
6: Greater than five contacts to sustain motivation to stop smoking provided outside of routine care (such as phone calls, letters, reminders, peer support, incentives, optional NRT, feedback).

Total duration of intervention (estimate):

0: 'Usual Care'
1: Less than 15 minutes.
2: Fifteen to 44 minutes.
3: Forty-five minutes to less than two hours.
4: Two hours to less than five hours.
5: Five hours to less than eight hours.
6: More than eight hours.

Outcomes

  • Outcome measures including smoking cessation and reduction, birthweight, mode of birth, perinatal outcomes, breastfeeding, gestation, psychological measures, family functioning.

  • Timing within pregnancy of recruitment and outcome measurement.

  • Adverse outcomes reported.

  • Any supplementary outcomes not included in tables (e.g. subgroup analysis).

Outcomes included in meta-analysis are marked '*' in the Characteristics of included studies.

Notes

  • Sample size calculation details.

  • Any analysis or cluster adjustment details.

  • Other comments.

What's new

Last assessed as up-to-date: 25 June 2013.

DateEventDescription
1 March 2013New citation required but conclusions have not changed

This review has been updated.

The overall conclusions about the effectiveness of smoking cessation interventions have not changed.

1 March 2013New search has been performed

For this update the review has been split into two reviews. Pharmacological interventions (with equal psychosocial support in both arms) are now covered in a separate review by Coleman 2012b. For this update, the current review now covers all other interventions. The title has been changed from 'Interventions for promoting smoking cessation during pregnancy' to 'Psychosocial interventions for supporting women to stop smoking in pregnancy'.

Search updated. We identified 16 new studies meeting the inclusion criteria.

As a result of a change in the inclusion criteria we excluded 13 studies from the previous version of the review. We also included four studies that had been previously excluded (three cluster trials and one abstract report of a trial), as well as nine studies that did not report any outcomes which could be used in meta-analyses, and which are reported in a separate table. We combined two reports of relapse prevention (Ershoff 1995; Secker-Walker 1995) as ‘Associated References’ to the primary papers reporting smoking cessation (Ershoff 1989; Secker-Walker 1994), and another paper which did not report any usable outcomes (Solomon 1996) as an 'Associated reference' to the primary report (Secker-Walker 1998).

A total of 77 randomised controlled trials, involving over 29,000 women with relevant outcome data, were included in the meta-analysis for this review (primary outcome data for 21,948 women participating in 70 trials and secondary outcome data only for a further 7404 women participating in seven trials). A further nine trials without outcomes are included but results summarised in Table 1, making a total of 86 studies included in this updated review.

Table 1. Primary outcomes from studies which met inclusion criteria, however outcomes were not able to be included in meta-analysis
  1. CI: confidence interval
    LBW: low birthweight
    RR: risk ratio

Study ID Main findings Rationale for not including outcomes in meta-analysis
Byrd 1993There was no statistically significant difference in smoking status among those who received either type of media or nurse counselling.Results could not be included as smoking cessation rates were not reported by intervention group.
Graham 1992There was no decrease in the rate of low birthweight for women who received the intervention.Smoking outcomes were not reported. Birthweight outcomes were not included in this review, as aspects other than the smoking component of the intervention may have had an effect on birthweight, and it is unclear how many smokers were in each group, or what proportion quit.
Haug 2004There was no significant difference in smoking between the intervention (motivational enhancement therapy) and control groups on self-reported cigarettes per day, mean carbon monoxide or mean cotinine.Study reports actual outcome data for movement in stages of change only. Outcome data for smoking cessation, cigarettes per day, carbon monoxide and cotinine levels are not reported.
Hiett 2000Significantly more women were able to quit smoking when enrolled in the intervention.Actual cessation rates not reported (poster abstract only available).
Hughes 2000There was no difference between intervention and control groups in mean delta stage of change or 12-month rate of maintained cessation in pregnant women (-0.62 vs -0.65).

Data from intervention and control

Outcomes were combined for intervention and control groups in pregnant women. Unable to extract numbers.

Lowe 2002At 1 month, 65% of behaviourally-based intervention hospitals agreed to provide materials about smoking cessation, compared to 3% control hospitals. After 1 year, 43% intervention hospitals still provided materials, compared to 9% of control hospitals. McNemar’s Chi2 indicates a statistically meaningful difference between the proportion of intervention hospitals implementing the program and the proportion of control hospitals implementing the program (2 1 = 12, P = 0.0005).Implementation data only included. No smoking cessation data provided.
Manfredi 1999Compared to controls, smokers attending family planning, prenatal and well-child clinics, exposed to the intervention were more likely to have quit (14.5% vs 7.7%).It was not possible to separate out which data was related to pregnant women, as opposed to women recruited from family planning and well child clinics. Further, it was not clear at what stage in pregnancy women were recruited and what the post-partum time points were.
Moore 1998There was no significant difference in LBW were 10.9% in the intervention group and 14.0% in controls (RR = 0.75, 95% CI 0.55 to 1.03). Preterm births rates were 9.7 in the intervention group and 11.0 in the controls (RR = 0.87, 95% CI 0.62 to 1.22).Smoking outcomes were not reported. Birthweight and preterm birth outcomes were not included in this review, as aspects other than the smoking component of the intervention may have had an effect on birthweight and preterm births.
Olds 2002Significant reduction in mean cotinine among women who smoked at baseline. Mean reduction of 12.32 ng/mL in the control group, compared to as mean reduction of 259.00 ng/mL in nurse-home visiting group.Study reports the mean cotinine reduction only, not mean cotinine levels or smoking cessation rates. It is also unclear how many randomised women were included in this analysis.

History

Protocol first published: Issue 2, 1998
Review first published: Issue 3, 1998

DateEventDescription
17 November 2008New citation required but conclusions have not changedTwo new authors have joined the review team for this substantive update, which includes the addition of risk of bias assessments for all trials; additional outcomes tables for smoking reduction, continued cessation in the postnatal period, neonatal intensive care unit admissions, psychological impacts of smoking, views of participants and providers; inclusion of additional data from previously included cluster-randomised trials; and risk of bias sensitivity analysis.
17 November 2008New search has been performedSearch updated: 7 new randomised controlled trials (Cope 2003; Dornelas 2006; Heil 2008; Higgins 2004; Hotham 2005; Pollak 2007; Rigotti 2006) and 4 cluster randomised controlled trials (Campbell 2006; McLeod 2004; Pbert 2004; Polanska 2004) included. Mullen 1991 and Hughes 2000 previously included have now been excluded.
3 November 2008AmendedConverted to new review format.
31 July 2003New search has been performedWe have updated the Background and Results sections (comment on the differences between the interventions when trials are grouped by intervention).

Twenty new trials reporting smoking cessation were included with five additional cluster-randomised trials. Nine additional trials were excluded. Six trials provided new data on fetal and perinatal outcomes.

The overall conclusions about the effectiveness of smoking cessation interventions did not change. New analyses grouping interventions by strategies showed that the pooled cognitive-behavioural interventions were effective, nicotine replacement therapy was borderline, and trials using 'stages of change' approaches or feedback were not effective. The two trials using a combination of rewards and social support were significantly more effective than other strategies. The increased information on perinatal outcomes strengthened the findings of a reduction in preterm birth and low birthweight. One trial reported method of delivery and one reported breastfeeding: neither showed an effect of the intervention.

Contributions of authors

Update (2012)

Catherine Chamberlain (CC) planned and co-ordinated the review update, conducted the supplementary search and retrieved articles, co-reviewed articles for inclusion, co-extracted data, and drafted the report.

Alison O’Mara (AO) planned the review, addressed queries about included studies and data, conducted analysis and drafted the review.

Sandy Oliver (SO) conceived and planned the review, addressed queries about included studies and data, provided input into drafts.

Jenny Caird (JC) co-extracted data and provided feedback on drafts.

Susan Perlen (SP) planned the review, co-reviewed articles for inclusion and provided input into drafts.

Sandra Eades (SE) provided input into drafts.

James Thomas (JT) planned and co-ordinated the review update, addressed queries about included studies and data, assisted with data analysis, and provided input into drafts.

Update (2009)

Judith Lumley (JL) screened retrieved papers against eligibility criteria, provided general advice on the review and contributed to the final text.

CC co-ordinated and secured funding for the review, undertook searches, retrieved papers, extracted data, wrote to authors for additional information, entered and analysed data, and wrote the review.

Therese Dowswell (TD) completed 'Risk of bias' assessments for trials included prior to the 2009 update and revised data abstraction records into an electronic format. TD revised the 'Risk of bias' assessments for RevMan 5 format, extracted, entered and analysed data on reduction and postpartum outcomes, and provided general advice and a methodological perspective on the review.

SO and Laura Oakley (LO) extracted and analysed data on participant and provider views, provided general advice and contributed to the text of the review.

Lyn Watson (LW) provided expert statistical advice on including cluster trials, extracted data for cluster trials and adjusted the data.

Update (2004)

JL co-ordinated the review update, extracted data, conducted the analyses and interpretation of data and wrote the review.

CC searched and screened search results, retrieved papers, extracted data, wrote to authors for additional information and entered data.

SO and LO provided general advice and contributed to the final text.

Original review (1999)

JL and SO conceived and designed the original review, and together with Elizabeth Waters (EW) and LO completed data extraction and wrote the original review. JL carried out the analyses. EW was unable to contribute after 2002. All contributed to the final text.

Declarations of interest

Catherine Chamberlain is receiving an Australian National Health and Medical Research Council PhD scholarship (607247).

Sources of support

Internal sources

  • EPPI-Centre, Social Science Research Unit, Institute of Education, University of London, London, UK, Not specified.

External sources

  • National Institute for Health Research, UK.

    Cochrane Review Incentive Scheme Award 12/138/17

  • Australian Department of Health and Ageing, Australia.

  • World Health Organisation, Australia.

    Financial support for 2013 update

Differences between protocol and review

This review forms one part of the evidence for the published protocol 'Interventions to Promote Smoking Cessation in Pregnancy' (1999) and four previous versions. This fifth update has been split into two separate reviews: Pharmacological interventions (to address concerns over safety of pharmacological interventions in greater detail) and this review, which includes the majority of studies using a wide range of psychosocial interventions to support women to stop smoking during pregnancy.

The 1998 protocol was developed in consultation with pregnant women, with the aim of addressing issues that were identified as relevant to women. We have endeavoured to continue this principle of addressing current issues that are relevant in each update of the review. Prior to each update, we have considered the recommendations of peer reviewers from the previous updates, and other relevant information, as outlined in Why it is important to do this review, to identify factors for analysis 'a priori'.

While the review has evolved in each update, the major differences between the original 1998 protocol and this 2013 update are listed below.

Background:

  • Development of a logic model, in line with recent 'PRISM-Equity' reporting guidelines.

Objectives:

  • Including assessment of whether interventions are effective in preventing relapse among women who have spontaneously quit smoking in early pregnancy.

  • Including assessment of cost-effectiveness.

  • Including assessment of any reporting of 'other adverse effects' by trial authors.

  • 'Equity' analysis of the effect of interventions among populations categorised as 'low SES', ethnic, or other vulnerable category, in line with recent PRISM-Equity reporting guidelines (separate protocol is available on request).

  • Including assessment of whether interventions shown to be effective under trial conditions (efficacy studies) are effective if implemented in routine care by pregnancy care providers (effectiveness studies).

  • Assessment for additional risk of bias categories, including: whether baseline characteristics were equal (as sequence generation 'unclear' in many studies); whether the intervention was implemented as planned; and 'contamination' or exposure of the control group to the intervention.

  • Assessment of whether studies included in this recent update are as effective as studies included in previous versions of this review.

Types of studies:

  • Quasi-randomised studies have been excluded (e.g. allocation by clinic day or week), unless it was considered there was a relatively low risk of interference with the sequence generation (i.e. odd or even maternal date of birth or medical record number).

Types of interventions:

  • Pharmacotherapy interventions separated into another review, though sensitivity analyses has been conducted for psychosocial interventions where pharmacological agents are optional for participants.

  • Provision of Information about harmful effects of smoking and advice to stop smoking were combined into one 'main intervention strategy' of 'health education'.

  • Provision of feedback of smoking status (e.g. carbon monoxide), effects of smoking on fetus (e.g. fetal movements), and positive information about the fetus (e.g. ultrasound') were combined into one main intervention strategy of 'feedback'.

  • Rather than as separate interventions, provision of self-help manuals or telephone support was included in the main intervention strategy of 'health education' (if no personal contact) or 'counselling' (if provided in conjunction with personal contact), with sensitivity analyses reporting the effects of these interventions as a separate comparison (in text under each of the main strategies), and as subgroup analyses for the whole dataset.

  • Categorisation of primary outcomes (quitting in late pregnancy) and secondary outcomes (all other outcomes).

  • Interventions were categorised as: 'single, multiple or tailored' for subgroup analyses within each comparison, as many interventions included more than one component/strategy.

  • The 'intensity' (frequency and duration) of each intervention has been coded.

  • 'Stages of change'-based interventions have not been assessed separately.

Types of comparisons:

  • Comparisons have been categorised as 'usual care, less intensive or alternative'.

Types of outcomes:

  • Smoking abstinence in late pregnancy has been reported separately for women who had quit spontaneously in early pregnancy (relapse prevention).

  • Preterm births < 32 and < 30 weeks were not reported as often excluded from trials.

  • Caesarean section rates, cost-effectiveness, and adverse events have been included in narrative synthesis.

Risk of bias assessments:

  • Whether the baseline characteristics were equal, whether implementation difficulties were reported, and whether the control group was exposed to the intervention (contamination) have been added to the 'Risk of bias' assessments.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Albrecht 1998

Methods

3-armed randomised-controlled trial (pilot study) evaluated 2 different interventions provided to 'pregnant teens' to reduce smoking in pregnancy and relapse postpartum. The hypothesis was that an intervention including peer support would be more effective than the intervention alone.  

Study conducted in Pittsburgh, USA. Data collection dates not reported.

Participants

Inclusion criteria: 12 to 20 years of age; 4 to 28 weeks' gestation; reported smoking at least 1 cigarette a day; single marital status; no previous live birth; able to read and write English.
Exclusion criteria: Pregnancy complications preventing attendance at group sessions or participation in a home study program.
Recruitment: Participants were recruited through local prenatal clinics and public schools. 84 women recruited (not known how many were eligible or approached) and randomised (C = 29, I1 = 29, I2 =26).

Baseline characteristics: Mean cigarettes/day at first visit: C = 6.44; I1 (TFS) = 5.87; I2 (TFSB) = 6.81.

63% African-American heritage, 37% European-American heritage.

Progress+ coding: Coded as single (low social capital) and young age (less than 20).

Interventions

Control: 30 minutes individual educational session with project nurse including information about the risks of smoking to the mother and the fetus and brochures on smoking and pregnancy.
Intervention 1 (TFS): Cognitive behavioural group model designed specifically for adolescents based on problem-behaviour theory: eight modules to heighten awareness and attention to smoking messages; build and enhance smoking cessation skills; teach skills for maintenance of smoking control; includes experiential learning and round robin discussion. TFS was modified to include additional information on smoking and the fetus, body image changes and overall health. The intervention also included social activities, immediate rewards and adult modelling.
Intervention 2 - TFS plus peer support (TFSB): Utilised all the components of TFS plus 1-to-1 support through a non-smoking peer (buddy) chosen by the young woman. Buddies were asked to attend all 8 sessions and to be available at other times for reinforcement of techniques learned and encouragement for continued cessation.

Main intervention strategy: Social support (multiple intervention) compared to less intensive intervention. TFSB compared with TFS and control in this review as outcomes only reported as combined figures.

Intensity rating: Frequency (C = 2, I = 6); Duration (C = 2, I = 6).

Intervention provided by project staff:efficacy study.

Outcomes

Biochemically validated point prevalence abstinence at 4-6 weeks post baseline (late pregnancy*).

Reduction in exhaled CO and self-reported mean cigarettes per day are reported as 'reduction' but actual post-intervention measures weren't reported so are not included in this review. Baseline modified Fagerstrom Tolerance questionnaire for adolescents to assess nicotine dependence.

Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskDescribed as 'randomly assigned'.
Allocation concealment (selection bias)Unclear riskNo information.
Incomplete outcome data (attrition bias)
All outcomes
Low riskOnly 46/84 had complete outcome data (high attrition rate = 45%), UC = 12 (41%), TFS = 13 (46%), TFSB = 13 (50%). No explanation for attrition. ITT analysis not mentioned. All those lost to follow-up were included as continuing smokers in this review.
Selective reporting (reporting bias)High riskOnly smoking outcomes reported and outcomes not reported separately for each of the control arms.
Other biasLow riskNo other bias detected.
Biochemical validation of smoking abstinence (detection bias)Low riskCO level (>= 8 ppm) in exhaled air used to identify smokers.
Blinding of participants and personnel (performance bias)
All outcomes
High riskProvider and participants unable to be blinded to educational intervention.
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskNot reported.
Incomplete implementationHigh riskProcess evaluation showed there was a 'significant drop out rate' (45%).
Equal baseline characteristics in study armsUnclear riskBaseline smoking characteristics similar, but other baseline characteristics not reported.
Contamination of control groupLow riskIntervention provided by research project staff.

Albrecht 2006

Methods

3-armed randomised controlled trial evaluated the short- and long-term effects of 2 smoking cessation strategies tailored to support pregnant adolescents to attain abstinence in pregnancy and maintain abstinence postpartum.

The study was conducted in 5 hospital-based and 2 community-based prenatal clinics in Pittsburgh, Pennsylvania, USA. Years of data collection not reported.

Participants

Inclusion criteria: 'Pregnant teens' aged 14 to 19 years; 12 to 28 weeks' gestation; able to read, write, and understand English; smoking at least 1 cigarette per day; single marital status; having no previous live births; and capable of being reached by telephone.

Exclusion criteria: pregnancy complications (i.e., bleeding or preterm labor) or required confinement to home by their physician.

Recruitment: During prenatal assessment, adolescents self-reporting smoking were invited to participate in study. Those expressing interest signed a consent form to allow the research team to contact them. Expressions of interest also advertised through flyers and brochures.

470 screened; 142/224 (63%) eligible women randomised (C = 50; I1: (TFS) = 47; I2: (TFS + B) = 45. Baseline characteristics: Number of cigarettes per day before pregnancy: Control 15.75 (10.38); I1: (TFS) 14.08 (7.22); I2: (TFSB) 14.62 (9.72).

Fagerstrom dependence score: Control 3.38 (2.05); I1: (TFS) 3.44 (1.79); I2: (TFSB) 3.68 (1.89).

Progress + coding: Low SES, Low educational attainment, low social capital (single) and young age (< 20 years).

Interventions

Control: Usual care that all teens would typically receive from a healthcare provider throughout their pregnancy. Smoking during pregnancy was addressed in the clinic by giving the teens educational materials on this subject during the initial prenatal visit. In this study, this material was explained and distributed to the participants by a research team member during the initial assessment. The meetings lasted 45-60 minutes and occurred at 1 of the antenatal clinics or centrally located community site. During the meeting, addresses and telephone numbers of the control group participants were updated after completion of the assessment. Prior to leaving the meeting, participants were informed of the date and time of their next assessment. Participants also received an attendance incentive (e.g. lipstick, nail polish). If the participant had delivered, the attendance incentive was a baby item.

Intervention 1 (TFS): The TFS intervention consisted of an 8 week group program designed to promote and maintain smoking abstinence based on the Cognitive Behavioral Theory, with modification that incorporated developmental components of Jessor’s Problem Behavior Theory, including a peer buddy and a peer co-leader for peer modelling and sanctioning on smoking. Information pertinent to pregnancy and smoking was provided at the beginning of the 8-week program.
Intervention 2 (TFS-B): The TFS-B group received the same 8-week programming, but participants were required to bring a non-smoking female of a similar age as their buddy to the sessions. The role of the buddy was to reinforce smoking cessation strategies and to provide social support to the participant throughout the study.

Main intervention strategy: Social support (multiple intervention) compared to a less intensive intervention. The control group and TFS-B are compared in this review.

Intensity rating: Frequency (C = 2, I = 6); Duration (C = 3, I = 6).

Provided by dedicated project staff: efficacy study.

OutcomesBiochemically validated point prevalence abstinence 8 weeks (late pregnancy*) and 1 year (6-11 months post partum*) after the intervention.
Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskConsenting adolescents were assigned randomly to 1 of 3 group assignments (TFS, TFS-B, or control) by a computer algorithm with a permutated block design, stratified by entry site.
Allocation concealment (selection bias)Unclear riskNot reported.
Incomplete outcome data (attrition bias)
All outcomes
Low riskHigh attrition: C = 60% (i.e. 40% did not complete 1 yr follow-up), TFS = 55%, TFS-B = 53%. Participants included in primary aim analysis pertaining to randomised treatment assignment, regardless of adherence to study treatment (ITT analysis).
Selective reporting (reporting bias)Low riskPrimary outcomes reported.
Other biasLow riskNo other bias detected.
Biochemical validation of smoking abstinence (detection bias)Low riskBiochemical validation of self-reported smoking status (point prevalence abstinence) using salivary cotinine (> 10 ng). Women reporting less than 1 cigarette per day with salivary cotinine 10-15 ng had salivary nicotine assessment to rule out environmental exposure, and were classified as smokers if that test was > 5 ng.
Blinding of participants and personnel (performance bias)
All outcomes
High riskParticipants and providers unlikely to be blinded to this educational intervention.
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskBlinding of outcome assessor not reported.
Incomplete implementationHigh riskProcess evaluation showed poor implementation with almost 50% participants not completing study.
Equal baseline characteristics in study armsLow riskBaseline characteristics appear equal.
Contamination of control groupLow riskIntervention provided by research team.

Baric 1976

Methods

A randomised controlled pilot study to evaluate whether medical advice had a effect on smoking cessation in pregnancy.

Study conducted in Bolton, England. Years of data collection not reported.

Participants

Inclusion criteria: Pregnant smokers at their first antenatal visit, less than 20 weeks' gestation.

Exclusion criteria: Not reported.

Recruitment: Women recruited from public antenatal clinic at Bolton and District General Hospital. 510 women screened, 142 eligible, 8 moved house and could not be followed up, and 24 women had spontaneously quit. 110 women randomised: control = 47, intervention = 63.

Baseline characteristics: 89% heavy smokers and 75% had been smoking for 5 years or more.

72% 'working-class' (majority low SES) and 75% had no educational qualifications.

Progress+ coding: Low SES and low educational attainment.

Interventions

Control: Usual care, which was advice at the discretion of the doctor.
Intervention: 1 to 1 counselling ('a short interview') from a senior medical student which involved discussion of the disadvantages of smoking during pregnancy: risk to the fetus; long-term risks of physical and intellectual impairment and possible reasons for this; possible effects on the mother's own health; costs of smoking; special dangers of smoking in late pregnancy; various ways to help someone to stop smoking. Given strong encouragement to quit and to make a commitment to do so. If this was not agreed then reduction to less than 5 cigarettes a day. Half the intervention group were given a diary to record each cigarette smoked and a gift of a free smoking diary.

Main intervention strategy: Counselling (single intervention) compared with usual care.

Intensity: Frequency (C = 0, I = 1); Duration: (C = 0, I = 1).

Usual care intensity: Frequency = 1, duration = 1.
Intervention conducted by existing staff (medical student): effectiveness study.

Outcomes

Self-reported abstinence 11 weeks after baseline visit (late pregnancy*).

Smoking reduction reported for whole cohort, not by intervention group, therefore not included in this review.

Discusses participants' views of intervention.

Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided. Described as "randomly divided".
Allocation concealment (selection bias)Unclear riskNo information provided.
Incomplete outcome data (attrition bias)
All outcomes
Low riskThere are some missing data in the tables. It is not clear if there was any overall loss to follow-up or whether missing data relate to specific outcomes only. All randomised women included in this review and those lost to follow-up were included as continuing smokers in this review.
Selective reporting (reporting bias)Unclear riskNo other outcomes reported.
Other biasLow riskNo other bias detected.
Biochemical validation of smoking abstinence (detection bias)High riskSmoking outcomes were self-reported by participants during a visit at home. There was no biochemical validation.
Blinding of participants and personnel (performance bias)
All outcomes
High riskEducational intervention at first antenatal visit.
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskNot reported.
Incomplete implementationUnclear riskNot reported.
Equal baseline characteristics in study armsUnclear riskNot reported.
Contamination of control groupLow riskMedical student provided intervention (not usual care provider).

Bauman 1983

Methods

Randomised controlled trial of use of exhaled CO feedback for promoting smoking cessation in pregnancy.

Study conducted in Guildford County, North Carolina, USA over 6 months in 1981.

Participants

Inclusion criteria: Women currently or recently smoking, attending public clinics.

Exclusion criteria: Not reported.

All women attending antenatal care orientation sessions were randomly allocated to experimental or control groups.

Recruitment: 226 women entered prenatal program and 170 (75%) included in analyses. The authors compared those who did not participate and did not find any significant differences. 47% (79/170) were current smokers (C = 43, I = 36).

Baseline characteristics: 43% had completed high school education, 56% were black, 80% classified as having no pregnancy risks other than smoking. 38% in the first trimester and 46% in the second trimester of pregnancy.

Progress+ coding: Low SES as all attending public prenatal clinic.

Interventions

Control: Women were read a 135 script that described the relationship among cigarette smoking, CO, and the harmful consequences of smoking.

Intervention: Experimental group received same information as control group, and they provided breath specimen in which CO was measured, with feedback of the result.

Main intervention strategy: Feedback (single intervention) compared to a less intensive intervention.

Intensity: Frequency (C = 1, I = 1); Duration (C = 1, I = 1).

Implemented by regular health educators: effectiveness study.

Outcomes

Biochemically validated abstinence 6 weeks after intervention (late pregnancy*).

Exhaled CO (ppm), but no SD reported; unclear if 'quantity of cigarettes' is mean cigarettes per day; recency of smoking; depth of inhalation.

NotesNot clear whether this was a group intervention - in which case there was no adjustment for clustering.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskRandom number table.
Allocation concealment (selection bias)Unclear riskNo information provided.
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskUnclear exactly how many women were randomised to each group, however we assume that those reported as 'current smokers' in table 1 are the baseline numbers, which were all included in this review.
Selective reporting (reporting bias)Unclear riskNone apparent.
Other biasLow riskNo other bias detected.
Biochemical validation of smoking abstinence (detection bias)Low riskBiochemical validation of reported smoking behaviour for those followed up (CO >= 9 ppm in exhaled air).
Blinding of participants and personnel (performance bias)
All outcomes
High riskIntervention was carried out by clinical staff, no participant blinding reported.
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskNot reported.
Incomplete implementationLow riskAll women apparently received the intervention.
Equal baseline characteristics in study armsLow riskNo difference between experimental and control arms on 12 variables measured.
Contamination of control groupLow riskImplemented by regular health educators at the maternity clinics.

Belizan 1995

Methods

Randomised controlled trial of psychosocial support in pregnancy which aimed to improve maternal health, including reducing smoking during pregnancy.

Conducted in 4 countries in Latin America (Argentina, Brazil, Cuba, and Mexico) from January 1989 to March 1991.

Participants

Inclusion criteria: High-risk women whose antenatal care began at 15-22 weeks' gestation, singleton pregnancy, 1 or more of the following: prior LBW infant; preterm birth; perinatal/infant death; < 18 years; body weight <= 50 kg; height <= 150 cm; low family income (local definitions applied); < 3 years school; crowded household (4 or more persons/bedroom); smoking; not living with husband or partner.
Exclusion criteria: Heart or renal failure; diastolic BP > 100 mmHg; history of cervical cerclage; Rh negative; mental disease or any chronic disease that might interfere with pregnancy.

Recruitment: 2,235 women met eligibility criteria and gave consent (I = 1115-though 1110 in table, C = 1120).

Baseline characteristics: Smokers (I = 23.9%, C = 21.8%), with variation between countries - Argentina (I = 21.9%, C = 20.6%), Brazil (I = 40.7%, C = 33.1%), Cuba (I = 27.4%, C = 28.9%), Mexico (I = 9%, C = 6.8%). Mean cigarettes per day at randomisation: C = 7.9, I = 7.5.

Progress+ coding: Low SES based on place of residence (low family income 20% in Cuba, 52% in Mexico, 53% in Brazil and 100% in Argentina).

Interventions

Control: Routine antenatal care, otherwise unspecified.

Intervention: Flexible use of a standardised manual, based on site-specific ethnographic studies of needs, fears, expectations, social support networks, including detailed descriptions of situations likely to occur during home visits. 4 to 6 home visits of 1 to 2 hours with emphasis on psychosocial support, education on health habits including better nutrition, reducing smoking alcohol and other drugs, reducing their physical workload, recognition of alarm signs and symptoms, improved access to hospital facilities, reinforcement of health service utilisation. Additional components were a poster, a booklet, hotline to project office, guided tour of hospital, encouragement of family support and participation. Intervention was provided by specially trained female social workers or obstetric nurses with previous experience of childbirth.

Main intervention strategy: Social support (tailored) compared with usual care.

Intensity: Frequency (C = 0, I = 6), Duration (C = 0, I = 5).

Usual care frequency and duration = 0 (unclear).

Intervention provided by study team: efficacy study.

Outcomes

Self-reported point prevalence abstinence at 36 weeks' gestation (late pregnancy*); Mean cigarettes per day.*

Multiple perinatal and maternal health outcome data were collected, but not included in this review as other aspects of the intervention may have had an impact.

Baseline state anxiety score.

NotesSample size was planned for the primary trial objective.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskCentrally prepared, method not stated.
Allocation concealment (selection bias)Low riskAllocation was by opening sealed, opaque envelopes.
Incomplete outcome data (attrition bias)
All outcomes
High riskAttrition 202/2230 (9%): 101 in each arm. Unclear what attrition among smokers and no ITT analysis of drop-outs as continuing smokers, so not able to re-include smokers who dropped out in this review.
Selective reporting (reporting bias)Unclear riskNone apparent.
Other biasLow riskNo other bias detected.
Biochemical validation of smoking abstinence (detection bias)High riskNo biochemical validation of reported smoking behaviour.
Blinding of participants and personnel (performance bias)
All outcomes
High riskHome visitors were aware of group allocation. Social support intervention with home visits.
Blinding of outcome assessment (detection bias)
All outcomes
Low riskThe evaluation of the interventions was conducted by a team of independent professional interviewers who were not informed of the characteristics of the study.
Incomplete implementationLow riskMost (83%) of the women randomly assigned to the intervention group received the planned number of home visits, and 90% were visited at least once.
Equal baseline characteristics in study armsLow riskThe distribution of risk factors was similar in the 2 groups and the 2 groups had similar demographic, obstetric, and psychological characteristics at baseline.
Contamination of control groupLow riskThe clinic personnel were unaware of the identity of the women in the control group, and no attempts were made to inform them of which women were in the intervention group. Health educators providing intervention were separate from care providers.

Bullock 1995

Methods

Randomised controlled trial of telephone support for improving maternal health outcomes, including smoking cessation during pregnancy.

Study conducted in a metropolitan city in the south island of New Zealand from March to December 1993.

Participants

Inclusion criteria: Women with telephone access, who were either single or with an unemployed partner, less than 20 weeks' gestation.

Exclusion criteria: None stated.

Recruitment: Recruited in the outpatient department of a large maternity hospital, or its associated GP practices, or self-referral via an introductory letter, phone call, and full discussion of "Healthy Mothers/Healthy Babies".

The eligible population was 221 women of whom 49 were never located, 23 were not interested, 10 refused after explanation, and 8 moved away, did not speak English or had a miscarriage. 131 (59%) participated (103 OPD, 22 from GPs, 6 self-referred) (C = 66, I = 65 randomised). Just over 50% were smokers (C = 35, I = 31).
Baseline characteristics: Mean cigarettes per day at baseline = 6.

88% European, 10% Maori. 53% single.

Progress+ coding: Low SES.

Interventions

Control: Package of publicly available educational material on healthy behaviours during pregnancy.
Intervention: Package plus weekly telephone call from trained volunteer with the aim of providing minimal support until 12 weeks after birth; aim "to be a friend and a good listener"; to ask about symptoms; signs; alcohol; drugs; smoking and meals in every call; to encourage attendance at antenatal clinic appointments and to ask about "feeling stressed".
Intervention provided by 19 female volunteers, trained for the project with a "case load" of 2 to 6 women each.

Main intervention strategy: Social support (single intervention) compared to a less intensive intervention.

Intensity: Frequency (C = 2, I = 6); Duration (C = 1, I = 4).

Intervention provided by project staff: efficacy study.

Outcomes

Self-reported abstinence at 34/40 (late pregnancy*).

Mean cigarettes per day*.

Anxiety and depression scores at baseline and 34/40. There were other intervention components which might have influenced these outcomes.

Notes

No process evaluation is reported. No sample size justification.

SDs for mean cigarettes per day were not reported, therefore we calculated a mean SD from 14 studies with available mean cigarette SDs (6.5) to include in this review, as recommended by the cochrane handbook