Objective: To determine the effects of environmental tobacco smoke (ETS) exposure on birth outcomes.
Design: A systematic review and meta-analysis was performed in accordance with MOOSE guidelines. MEDLINE, EMBASE, CINAHL and LILACS (up to October 2007), were searched and also reviews and reference lists from publications, with no language restrictions. Pooled mean differences and odds ratios (ORs) with 95% confidence intervals were estimated using data extracted from papers, based on random effect models.
Setting: Comparative epidemiological studies.
Patients: Pregnant women or women who have given birth.
Exposures: Maternal exposure to ETS during pregnancy.
Main outcome measures: Mean birth weight and proportion of premature infants.
Results: 58 studies were included; 53 used cohort designs, 23 ascertaining ETS exposure prospectively and 30 retrospectively; 5 used case–control designs. In prospective studies, ETS exposure was associated with a 33 g (95% CI 16 to 51) reduction in mean birth weight, and in retrospective studies a 40 g (95% CI 26 to 54) reduction. ETS exposure was also associated with an increased risk of low birth weight (birth weight <2500 g; prospective studies: OR 1.32, 95% CI 1.07 to 1.63; retrospective studies: OR 1.22, 95% CI 1.08 to 1.37). The risk of small for gestational age (<10th centile) birth was significantly associated with ETS exposure only in retrospective studies (OR 1.21, 95% CI 1.06 to 1.37). There was no effect of ETS exposure on gestational age.
Conclusions: Exposure of non-smoking pregnant women to ETS reduces mean birth weight by 33 g or more, and increases the risk of birth weight below 2500 g by 22%, but has no clear effect on gestation or the risk of being small for gestational age.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The public health impact of environmental tobacco smoke (ETS) exposure is currently an important issue for policy makers and clinicians. This is because ETS exposure has a clinically important and detrimental impact on adult and child health.1–4 However, the effects of maternal ETS exposure on the unborn fetus remain less clearly defined. Since active maternal smoking during pregnancy impairs fetal growth5–8 and shortens gestation,9 with important consequences for fetal and infant mortality and morbidity, it is likely that maternal ETS exposure during pregnancy has similar, if less marked, adverse effects.
Two previous systematic reviews of the effects of maternal ETS exposure during pregnancy on birth outcomes10 11 found evidence of a small reduction in mean birth weight, and an increased pooled risk of babies being small for gestational age (SGA) or low birth weight (LBW, birth weight <2500 g) at term, but no significant association with either SGA or LBW at term assessed individually. The conclusions of one of these reviews11 were also endorsed by the US Surgeon General’s report, published in 2006,12 which concluded that at the time of writing an updated meta-analysis was not warranted.12 However, both reviews are several years old, including papers published only up to 199511 and 1998,10 and identified using relatively restrictive inclusion criteria with search strategies limited to only one10 or two11 electronic databases. It is therefore likely that the total available evidence base has increased substantially since these analyses were reported.
We therefore now report a new systematic review and meta-analysis of the evidence available to 2007 to quantify the impact of maternal ETS exposure during pregnancy on a range of birth outcomes including birth weight, prematurity, LBW and SGA.
Methods of the systematic review
Comparative epidemiological studies of the effect of ETS exposure on non-smoking, pregnant women were included in the review. Randomised controlled trials of smoking cessation in either parent were excluded. ETS exposure included any record of maternal contact with environmental tobacco smoke from domestic, occupational or other sources.
Primary outcome measures were mean birth weight, mean gestation period and proportion of premature infants. Secondary outcome measures were proportions of infants categorised as LBW or SGA at term.
We used the Centre for Dissemination and Reviews guidelines13 to search Ovid MEDLINE (from 1966 to October 2007), Ovid OLDMEDLINE (from 1950 to October 2007), Ovid EMBASE (from 1980 to October 2007), CINAHL (from 1982 to October 2007), LILACS (from 1982 to October 2007) and CAB Abstracts (from 2000 to October 2007). The following five terms were used as keywords to locate studies, where $ indicates truncation: environ$, tobacco, smok$, ETS, cigarette. Explosions of the following MeSH terms were also used: tobacco smoke pollution, gestational age, newborn infant, premature infant, fetal death, birthweight, low birthweight infant, fetal growth retardation, and small-for-gestational-age infant. Hand searching of references was also performed. No language restrictions were imposed and translations were sought where necessary.
Titles identified from searches were checked for eligibility by one author (JL-B) and abstracts of potentially eligible studies (as judged by titles) were checked for eligibility by three authors (TC, AS, JB). The full text of studies regarded as potentially eligible were obtained and assessed independently by two authors to decide whether the studies met the inclusion criteria. Disagreements were resolved by discussion. For included studies, two authors independently extracted data and assessed methodological quality using the Newcastle–Ottawa quality assessment scale14 based on the following: (a) selection of cases and controls, or cohort; (b) comparability of cases and controls, or cohort; and (c) ascertainment of exposure/outcome.
Statistical methods for the meta-analysis
Tabulated data, crude estimates or adjusted estimates were extracted from the included studies and we wrote to authors of studies published in the past 10 years for further information or data, where necessary. Data are reported using mean differences (MD) for continuous outcomes or odds ratios (OR) for dichotomous outcomes, with 95% confidence intervals. Separate pooled analyses were conducted for case–control, prospective cohort and retrospective studies. For each meta-analysis, data were pooled using random effect models15 to allow for heterogeneity between studies. Heterogeneity was assessed using I2.16 We compared the impact of maternal ETS exposure during pregnancy with no maternal tobacco smoke exposure on primary birth outcomes. Sensitivity analyses were conducted by excluding studies with poor levels of methodological quality (<6 on the Newcastle–Ottawa scale). Data were analysed using the software program Review Manager 4.3 (RevMan, Version 4.3 for Windows, Nordic Cochrane Centre, Copenhagen, 2005). Probability values below 0.05 were considered statistically significant. The work was carried out in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.17
Overview of the included studies
We found 207 studies with titles which suggested that they were potentially eligible for inclusion (fig 1). After scrutinising the abstracts of these studies, 71 were deemed potentially eligible and of these 58 remained in the review after their full text was considered (table 1). The reasons for exclusion of the 13 potentially eligible studies are presented in the appendix.
In the 58 included studies, ETS exposure was collected by self-report questionnaire in 46 studies,11 19–29 31–33 35 37 39–43 46–51 53–55 57–60 63 65–68 69–74 and biochemically in three studies using serum cotinine33 75 or nicotine levels in the hair.56 The remaining 10 studies used both self-report and biochemical markers.18 30 36 38 44 45 52 61 62 64 ETS was assessed either as domestic exposure (29 studies) or as domestic and occupational exposure (20 studies). The remaining studies assessed any ETS exposure (8 studies) (table 1).
In 23/58 studies, ETS exposure was assessed using prospective cohorts where the data were collected within the first or second trimester of pregnancy,19 20 28 34 36 37 44 50 55 66 71 at a single exposure point during the three trimesters43 47 54 61 72 73 or only in the final trimester.22 42 60 62 64 69 Thirty studies used retrospective ascertainment of ETS exposure, recorded usually within a month following birth.11 18 23 24 27 29 30 32 33 35 38–41 45 46 48 49 51–53 57–59 63 65 67 68 70 74 The remaining five studies used a case–control design in which cases were SGA25 26 56 or preterm infants31 or babies with congenital malformations.21 Nine studies included only full-term babies with gestational age ⩾37 weeks.19 25 37 40 46 47 59 65 74 Two studies included only babies with birth weights ⩾2000 g,40 46 and two studies only included babies who were >500 g.32 44
Methodological quality of the studies
The methodological quality of the 58 included studies is presented in table 1. Two translated studies could not be assessed due to lack of information.43 67 The median quality score was 6 (range 2–9), indicating that the overall methodological quality was generally good. Using the a priori chosen threshold of 6, we judged 37 (63%) studies to be of high methodological quality. The quality of other studies was lower primarily because they had used retrospective ascertainment of ETS exposure, contained unclear information on the adequacy of follow up of the cohort, or did not adjust analyses for confounding variables.
Effects of ETS on mean birth weight
Forty-four of the 48 studies that reported birth weight as an outcome had data available for meta-analysis (figs 2 and 3). The remaining four studies either reported no standard error,22 27 inconsistent standard deviations45 or presented log-transformed outcomes.62
Seventeen of the 44 studies ascertained ETS exposure prospectively. The pooled results for these studies found that ETS exposure was significantly associated with a 33 g (95% CI 15.7 to 51.3) reduction in birth weight (fig 2). Moderate levels of heterogeneity were found between the study estimates (pooled I2 = 34%). Studies reporting adjusted birth weight showed greater reductions in birth weight associated with ETS exposure than crude estimates. The remaining studies all ascertained ETS exposure retrospectively. The pooled result for these studies was larger in magnitude compared with the results from the prospective studies (40 g reduction, 95% CI 25.8 to 54.4; fig 3) with moderate levels of heterogeneity (pooled I2 = 39%).
Results from a sensitivity analysis restricted to studies of high methodological quality (⩾6 on the Newcastle–Ottawa Scale) were similar, with the overall pooled mean reduction in birth weight associated with ETS exposure for the 14 high-quality prospective studies being 30 g (95% CI 12.9 to 46.2, pooled I2 = 28%) and 41 g (95% CI 21.1 to 60.6, pooled I2 = 46%) in the 16 high-quality retrospective studies.
Four studies also presented gestation-corrected birth weight either using birthweight z (SD) scores,70 a birth weight ratio22 61 or adjusted birth weight per centile.52 However, there were no differences between the ETS exposed and non-exposed groups in either birth weight z score (p = 0.18), birth weight ratio (p = 0.5622; p = 0.9661) or adjusted birth weight per centile (p = 0.36).
Effects of ETS on the risk of prematurity
Prematurity was assessed using mean gestation period (in weeks) in 11 studies. No significant associations were seen between ETS exposure and mean gestation period in either prospective studies or retrospective studies (MD −0.04 weeks, 95% CI −0.22 to 0.13, 5 studies; MD −0.03 weeks, 95% CI −0.28 to 0.22 weeks, 6 studies, respectively). High levels of heterogeneity were seen between the prospective (I2 = 65%) and retrospective studies (I2 = 76%).
A further 19 studies assessed prematurity using the proportion of premature births (<37 weeks). One of these was not included in the analyses because the risk of prematurity for passive smokers was adjusted for active smoking in the analysis.28 A significant 18% increase in the risk of prematurity with ETS exposure was seen in the retrospective studies (OR 1.18, 95% CI 1.03 to 1.35; 9 studies), however, the effect was not statistically significant in the pooled analysis of prospective studies (p = 0.24; 8 studies) or in the case–control study (OR 0.92, 95% CI 0.65 to 1.31).31
Effects of ETS on the risk of LBW (birth weight <2500 g)
Twenty-six of the 28 studies which measured LBW as an outcome had data available for meta-analysis. Two studies were excluded because they did not present a measure of variance34 or presented results which adjusted for active smokers in the analysis.28 All of the 26 studies defined LBW as <2500 grams, except one study which used the definition of <2000 grams.53
In a pooled analysis of prospective studies, exposure to ETS was found to be significantly associated with an increased risk of having a LBW birth (OR 1.32, 95% CI 1.07 to 1.63; 9 studies; I2 = 55%; fig 4). Similar results were seen from the pooled analysis of retrospective studies (OR 1.22, 95% CI 1.08 to 1.37; 17 studies; I2 = 0%; fig 4). Six studies only analysed data from full-term babies (⩾37 weeks’ gestation) and excluding these (which is appropriate for consideration of the impact of ETS on LBW) slightly increased the risk of LBW (prospective studies: OR 1.30, 95% CI 1.04 to 1.62; 7 studies; retrospective studies: OR 1.31, 95% CI 1.15 to 1.50; 13 studies).
Effects of ETS on the risk of SGA
Twenty studies measured SGA as an outcome. Eighteen of these defined SGA as being below the <10th centile, the others as either being more than 1.5 standard deviations below the mean on the Japanese standard fetal growth curve54 or more than 2 standard deviations below the age-related mean.28 In a pooled analysis of nine retrospectively ascertained studies, exposure to ETS was found to be significantly associated with a 21% increase in the risk of having a SGA baby (95% CI 1.06 to 1.37; I2 = 0%; fig 5). However, the pooled result from the eight prospective studies was not statistically significant (OR 1.05, 95% CI 0.87 to 1.28; I2 = 45%; fig 5). The three remaining studies of SGA used case–control designs.25 26 56 The pooled results from these studies found no evidence of an association between ETS exposure and SGA (OR 1.02, 95% CI 0.65 to 1.61; 3 studies, I2 = 0).
What is already known on this topic
Previous reviews have found small reductions in mean birth weight associated with maternal exposure to environmental tobacco smoke (ETS).
The available evidence base has increased substantially since these reviews, such that the clinical implications of this impact can now be more clearly defined.
What this study adds
Exposure of non-smoking pregnant women to ETS reduces mean birth weight by 33 g or more, which is higher than previously reported; and increases the risk of higher morbidity, low birthweight births (<2500 g) by 22%.
These effects appear to be independent of any association between reduced length of gestation and maternal ETS exposure.
The main findings of this systematic review are that maternal ETS exposure during pregnancy is significantly associated with a mean reduction in birth weight of 33 g or greater and an approximately 22% or greater increase in the risk of LBW (<2500 g). However, ETS exposure did not appear to consistently alter the risk of SGA birth, and had little if any effect on the duration of gestation. Generally, low levels of heterogeneity were seen between the studies, indicating that the effect of ETS was similar irrespective of the source of ETS exposure. We have not attempted to distinguish the effects of ETS exposure at different stages of pregnancy.
Our findings suggest that the impact of ETS exposure on mean birth weight is between 25% and 40% higher than previous estimates, including that in the recent US Surgeon General’s Report on the health consequences of involuntary exposure to tobacco smoke.3 This is probably because our review includes 30 more studies published since the meta-analyses10 11 used as the basis of the Surgeon General’s report.12 Previous reviews reported a 20% increase in the combined outcome of LBW at term and SGA, but did not find consistently significant effects on these outcomes separately. We separated LBW at term and SGA, and found a significant 22–32% increase in LBW, but no consistent effect on SGA. Also, the increased risk of LBW in our analysis was still significantly apparent when we excluded studies which only assessed LBW in term babies. This present review is, to our knowledge, the first to quantify the effects on these separate outcomes.
Our findings for birth weight are based on data from 44 studies, whereas previous estimates were based on only 11 studies.11 Previous reviews may have been biased by their search strategies that were limited to one10 or two11 databases, and in one case by excluding foreign language papers unless an English translation was available.11 Our findings are more likely to be representative estimates of the true effects because they are based on results of a comprehensive search with no language restrictions, including data identified through hand searching of reference lists and previous reviews, and because we included studies in the meta-analysis that the previous systematic reviews did not, by successfully contacting authors who provided more information or data relating to their studies.
Our finding that ETS exposure was not consistently significantly associated with the risk of prematurity is in agreement with the findings of the Surgeon General’s report.3 Again our finding is based on rather more data (19 studies vs 8 in the Surgeon General’s report), but active maternal smoking is known to increase the risk of prematurity by over 27%76 and as we found the proportion of premature babies was significantly higher in retrospective studies, further research into this potential effect of ETS would be helpful.
To assess the effect of methodological quality, we performed a sensitivity analysis which was restricted to high-quality studies, but the findings were similar to the overall analysis. We were inevitably limited in the range of confounding variables we could adjust for, and although we were able to correct for the effect of maternal age on birth weight,77 we could not adjust for socioeconomic status78 or ethnicity.79 We were also not always able to adjust mean birth weight for gestation in our analyses, so it is possible that the observed effect of ETS exposure on mean birth weight arises from more ETS-exposed babies being delivered earlier.10 However, consistent results were seen between the gestation-adjusted and gestation-unadjusted analyses, therefore the effect of ETS exposure on birth weight is likely to be real. Most studies used non-validated, self-reported measures to ascertain maternal ETS exposure, which could result in an underestimate of ETS exposure due to either a lack of awareness of maternal exposure or an unwillingness to declare exposure, and thus could results in biased reports. Biochemical markers of ETS exposure, such as cord serum or urinary cotinine, yield validated measures; however, ETS exposure earlier in pregnancy, which is pertinent to fetal outcomes, could be underestimated by such methods. Where individual studies reported both self-reported and biochemically validated exposures, no differences in outcomes were reported which suggests that it is reasonable to use self-report outcomes and our findings are likely to be valid.
A recent systematic review of active smoking studies found a 200 g reduction in birth weight.12 Since ETS exposure is typically equivalent to about 1% of the exposure from active maternal smoking,4 our estimate of 33–40 g reduction in birth weight equates to an effect of 17–20% of that of active smoking. This is consistent with findings from animal studies which suggest that sidestream smoke (the primary component of ETS) is more harmful than mainstream smoke (inhaled during active maternal smoking).80 Tobacco smoke contains many toxins81 and maternal exposure to these is thought to exert adverse fetal effects through a variety of mechanisms,82 which may therefore be disproportionately worse following ETS exposure than after active maternal smoking, potentially because ETS contains fetal toxins in greater concentrations.
Several mechanisms have been suggested to describe why active maternal smoking impacts on fetal birth weight.83 The adverse effects of maternal smoking may be due to the toxic effects of nicotine or other constituents, such as carbon monoxide, by affecting growth through impairing fetal metabolism of leptin84 or possibly by nicotine impairing nutrient and energy delivery to the fetus by suppressing appetite and inhibiting body weight gain in the mother.83 However, as we found that ETS reduces birth weight, but has no impact on the risk of prematurity, this is consistent with one or more constituent(s) of ETS (eg, nicotine or polycyclic aromatic hydrocarbons) exerting a direct effect on fetal growth. Nicotine may do this by constricting utero-placental arteries, reducing blood flow and thus oxygen delivery to the fetus85 and other toxins may operate by different, as yet unidentified, mechanisms.
Our study thus provides new estimates of the effect of ETS on mean birth weight and the risk of LBW, the former of which is higher than previous figures and the latter demonstrated to be independent of SGA for the first time. LBW is associated with an increased risk of infant mortality,86 developmental problems such as attention-deficit hyperactivity disorder87 and reduced IQ scores in childhood.88 There is also growing evidence that the adverse effects of LBW can extend throughout the lifespan by increasing the risks of chronic disease in later life.89 90 Therefore, the public health consequences of maternal ETS exposure are likely to be extensive and through effective tobacco control measures are potentially entirely avoidable.
The authors would like to thank the students at the University of Nottingham (J Chen, Y Nakanomori, G Marinov), M Pujades and K Szpakowska for their help with the translations of the non-English papers; and R Adamek, F Perera, F Matsubara, and W Hanke for providing us with further information or data relating to their studies.
Contributors: JL-B participated in the conception, design, identifying studies, data collection, study selection, data extraction, analysis, and interpretation of the data; in the writing of the protocol, drafting and revising the article. AS, JB and TC participated in the conception, design, study selection, data extraction, and interpretation of the data; in the writing the protocol, in drafting the article and revising it critically for important intellectual content, and approved the final version to be published.
Funding: This study was internally funded by the University of Nottingham. The sponsor had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication.
Competing interests: None declared.