Low birth weight and preterm birth are important risk factors for perinatal morbidity and mortality (¹). Furthermore, lower birth weight is associated with an increased risk of chronic diseases in adulthood, including cardiovascular disease, high blood pressure, coronary heart disease, type 2 diabetes and adult mortality (²–⁵). Maternal smoking during pregnancy is strongly associated with low birth weight (⁶,⁷). Babies whose mothers smoked during pregnancy are on average 150 to 200 g lighter at birth than babies with non-smoking mothers (⁶,⁸). The 2006 legislation in Scotland, which banned smoking in public areas, has been associated with a reduced number of preterm births and small for gestational age babies (⁹).

Genetic variation at the CHRNA5–CHRNA3–CHRNB4 locus on chromosome 15q25 is robustly associated with the quantity of smoking in those who smoke (¹⁰–¹⁴). Each additional copy of the risk allele (T) of the single-nucleotide polymorphism (SNP), rs1051730, is associated with an increase in smoking quantity of approximately one cigarette per day (¹²). The variant is not associated with smoking initiation, since it is equally prevalent in smokers and never-smokers (¹²). The rs1051730 SNP is also strongly associated with a reduced ability of women to quit smoking during pregnancy, and with a higher number of cigarettes per day in pregnant women who smoke (¹⁵,¹⁶).

Studies of women from the UK (¹⁵) and the Netherlands (¹⁷) have independently investigated the association between the 15q25 variant and birth weight. Both studies observed reductions in birth weight with each additional copy of the maternal risk allele in women who smoked during pregnancy [−28 g (95% confidence interval—95% CI: −59 to 2 g) per T-allele in n = 1829 UK women; −38 g (95% CI: −89 to 13 g) per T-allele in n = 610 Dutch women]. However, statistical power was limited by the available sample sizes, and neither association achieved P < 0.05. The association between the maternal rs1051730 variant and birth weight has not been investigated in a well-powered sample of mothers and offspring, and furthermore, any potential associations with fetal genotype have not been examined.

We therefore performed a meta-analysis of 26 241 women from 14 studies to assess the evidence for association between maternal rs1051730 and birth weight, stratified by maternal smoking status in pregnancy. We hypothesized that, due to its strong association with smoking quantity, there would be an association between the maternal risk allele and lower offspring birth weight in women who smoked during pregnancy, but no association with birth weight in the non-smokers. Such an interaction would be interpreted within a Mendelian randomization context as providing evidence of the causal nature of an association between maternal smoking and offspring birth weight (¹⁸).

The 15q25 variant is strongly associated with lung cancer in smokers. Several analyses have shown that the association with self-reported number of cigarettes per day is insufficient to explain the lung cancer association (¹²,¹⁹,²⁰). These results initially raised the possibility that the 15q25 variant might increase vulnerability to the harmful effects of tobacco smoke in those who are exposed, independently of smoking behaviour. If this were the case, we might expect to see an association between fetal rs1051730 genotype and birth weight, which is independent of maternal genotype, in the offspring of mothers who smoked. However, a recent study showed a stronger association with objectively measured cotinine levels than with self-reported number of cigarettes per day. This suggested that the locus works predominantly by influencing tobacco exposure in smokers, and that although cigarettes reported per day have the advantage of enabling the necessarily large sample sizes for genetic association studies, they fail to capture smoking exposure fully (²¹). Based on this, and other recent evidence (²²), we would not expect to observe an association between fetal rs1051730 genotype and birth weight that is independent of maternal genotype. We examined evidence of association between fetal rs1051730 genotype and birth weight in 25 090 offspring from 11 studies. In a subset of 12 489 mother–offspring pairs from eight studies, we were able to test the association between fetal genotype and birth weight that was independent of maternal genotype.

The basic characteristics of study participants are presented in Table 1.

In a meta-analysis of 26 241 women from 14 studies, we have found strong evidence of a gene × environment interaction between maternal smoking behaviour and genotype of the 15q25 variant, rs1051730, in relation to offspring birth weight (P = 0.007). In women who smoked during pregnancy, there was evidence that each additional copy of the T-allele, which is associated with higher smoking quantity, was associated with a 20 g (95% CI: 4–36 g) reduction in offspring birth weight. In non-smoking mothers, the estimated change in birth weight was 5 g (95% CI: −4 to 14 g) per T-allele. Although the result is as expected, given prior knowledge of the role of the SNP in smoking behaviour and the observational link between smoking and low birth weight, the result provides an important proof-of-principle: when clearly defined, genetic and behavioural risk factors can interact to influence a phenotype.

Our results strengthen the evidence that maternal cigarette smoking during pregnancy is causally associated with lower offspring birth weight. Conventional epidemiological studies are at risk of confounding by lifestyle or socioeconomic factors. Genetic variants should not be related to the confounding factors that distort associations in conventional observational epidemiological studies (²³). We have shown previously that the rs1051730 variant is not associated with several confounding factors of smoking behaviour (e.g. age, education level and occupation) (¹⁵). Hence, we can be confident that the association we have observed between maternal rs1051370 genotype and offspring birth weight, together with the null association in never-smokers, reflects a causal relationship between smoking during pregnancy and birth weight when considered within a Mendelian randomization framework (¹⁸). These data add to the evidence from previous epidemiological studies and clinical trials (⁶,⁹,²⁴–³⁰).

Each additional risk allele of rs1051730 is associated with approximately one extra cigarette per day reported by smokers (¹²,²¹). Our effect size estimates of a 20 g (95% CI: 4–36 g) lower birth weight per allele in smokers and a 24 g (95% CI: 3–45 g) lower birth weight per allele in women who smoked beyond the first trimester represent unconfounded estimates of this exposure on birth weight. These estimates are consistent with evidence from a prospective study of maternal smoking during pregnancy which reported a 27 g lower birth weight per cigarette per day smoked in the third trimester (³¹). As a consequence of the relatively small effect sizes, genetic variation at the 15q25 locus would be expected to explain only a very small proportion of the association between low birth weight and adverse health outcomes in the offspring.

Subdivision of the pregnancy-smoker group showed evidence that the association between maternal genotype and birth weight may differ depending on whether women smoked during the first trimester only [21 g (95% CI: −15 to 57 g) increase in birth weight per T-allele] or continued to smoke beyond the first trimester [24 g (95% CI: 3–45 g) lower birth weight per T-allele]. Although power was limited, particularly in the women who smoked only in the first trimester, there was evidence of heterogeneity between the strata (P = 0.035). This suggests, in line with previous evidence (³¹), that if women quit smoking during the first trimester of pregnancy, the adverse effects of smoking on birth weight are reduced. This result should, however, be treated with some caution. We stratified the data on whether or not women gave up smoking, which may be influenced both by rs1051730 genotype (¹⁵) and by smoking quantity, which itself influences birth weight. Stratification on smoking cessation could therefore potentially induce distorted relationships between genotype and birth weight. Women who continue smoking are likely to be heavier smokers than those who do not, and this might apply especially to the T-allele carriers. Further studies are necessary to ascertain the extent of this problem, for example, using cotinine levels as a more reliable indicator of smoking quantity.

Studies of the association between the 15q25 variant and lung cancer have reported a stronger association with disease than can be accounted for by the quantity of cigarettes per day (¹²,¹⁹,²⁰). Possible explanations for this include the likely scenario that the quantity of cigarettes per day fails to capture smoking exposure fully, but also have raised the possibility that the 15q25 variant might increase vulnerability to the harmful effects of tobacco smoke independently of smoking behaviour. To test whether the effects of maternal smoking exposure could be modified by the fetal genotype of rs1051730, we additionally examined evidence of association between fetal rs1051730 genotype and birth weight in 25 090 offspring from 11 studies. Maternal and fetal genotypes are correlated (r ≈ 0.5), so if there is no true association between fetal genotype and birth weight, we would still expect to observe an effect size of approximately half the magnitude of the association between maternal genotype and birth weight in offspring of smoking pregnancies. Our observed effect size estimate of −14 g (95% CI: −29 to 5 g) is consistent with this expectation. On adjustment for maternal genotype in 12 489 mother–offspring pairs, the estimate of the association between fetal genotype and birth weight that is independent of maternal genotype was −9 g (95% CI: −38 to 19 g). Together our results suggest that any influence of the rs1051730 variant on birth weight is mediated by maternal smoking behaviour. This supports the hypothesis that the CHRNA5–CHRNA3–CHRNB4 locus works wholly through modulating smoking behaviour and not by an alternative mechanism.

We acknowledge some limitations to our study. First, in BWHHS, MIDSPAN and NCCGP, we made some assumptions regarding smoking status in pregnancy due to incomplete data. This may have resulted in a misclassification of pregnancy smokers versus non-smokers. However, a sensitivity analysis excluding these studies produced very similar effect size estimates of the associations within each stratum. Second, the individual studies in the meta-analysis used different collection and classification methods for the smoking data. We found little evidence of heterogeneity between studies within each stratum. Third, smoking status was self-reported, and previous studies using biochemical markers suggest that women may not admit to smoking during pregnancy (³²). Any such misreporting is likely to result in an underestimate, rather than an overestimate of the difference between smokers and non-smokers. In addition, a recent meta-analysis, which included the BWHHS and MIDSPAN studies, has shown strong agreement between associations of the rs1051730 variant and both self-reported cigarettes per day and cotinine levels (²¹). Fourth, our statistical evidence for association with birth weight in the pregnancy smokers (P = 0.014), and for the genotype × smoking status interaction (P = 0.007), is modest in relation to levels now considered necessary for robust genetic association studies (³³). However, the evidence for association between rs1051730 and smoking quantity is extremely robust, giving high prior odds of association with birth weight, and the strength of evidence we have observed is as expected, given the available power and effect size estimates.

In conclusion, the maternal rs1051730 variant is associated with lower birth weight when mothers smoke during pregnancy, but not in non-smoking mothers. In mothers who smoked beyond the first trimester of pregnancy, there was an association between the variant and lower birth weight, but not in mothers who smoked only in the first trimester, although this observation requires further careful evaluation. Our study strengthens the evidence that smoking during pregnancy is causally related to lower birth weight.

Supplementary Material is available at HMG online.

Researchers were funded by the European Centre for the Environment and Human Health (part of the Peninsula College of Medicine and Dentistry, which is a joint entity of the University of Exeter, the University of Plymouth and the NHS in the southwest) and is supported by investment from the European Regional Development Fund (ERDF) and the European Social Fund (ESF) Convergence Programme for Cornwall and the Isles of Scilly (J.T.); The Dutch Kidney Foundation (C08.2251) (H.R.T.); the Netherlands Organization for Health Research and Development (ZonMw 90700303, 916.10159) (V.W.V.J.); the Higher Education Funding Council for England (C.L.R.); National Institute for Health Research (A.T.H. and C.P.); the University of British Columbia, Clinician Investigator Program (J.W.Y.N.); UK Medical Research Council and University of Bristol (D.A.L. and G.D.S.); Wellcome Trust [WT 084762MA (L.P.), 085541/Z/08/Z (R.M.F.) and 085515 (M.-J.A.B.)]. Funding to pay the Open Access publication charges for this article was provided by the Wellcome Trust.

We are extremely grateful to the participants and families who contributed to all of the studies and the teams of investigators involved in each one. These include interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses.

ALSPAC: This publication is the work of the authors, and Rachel Freathy, Debbie Lawlor and George Davey Smith will serve as guarantors for the contents of this paper. The Avon Longitudinal Study of Parents and Children (ALSPAC) has core support provided by the UK Medical Research Council, the Wellcome Trust (grant ref: 092731) and the University of Bristol.

BWHHS: The British Women's Heart and Health Study (BWHHS) is funded by the Department of Health (England) Policy Research Programme and the British Heart Foundation.

DNBC: The DNBC was established with major funding from the Danish National Research Foundation. Additional support for the DNBC has been obtained from the Danish Pharmacists' Fund, the Egmont Foundation, the March of Dimes Birth Defects Foundation, the Augustinus Foundation and the Health Fund of the Danish Health Insurance Societies.

DNBC-GOYA: The DNBC-GOYA study was supported by the Wellcome Trust (grant WT084762).

DNBC-PTB: The GWAS data for the DNBC study of preterm birth was done within the Gene, Environment Association Studies (GENEVA) consortium, with funding provided through the US National Institutes of Health (NIH) Genes, Environment and Health Initiative (GEI; U01HG004423). Assistance with genotype cleaning and general study coordination for the preterm birth project were provided by the GENEVA Coordinating Center (U01HG004446). Genotyping was performed at the Johns Hopkins University Center for Inherited Disease Research, with support from the NIH GEI (U01HG004438).

EFSOCH: The Exeter Family Study of Childhood Health (EFSOCH) was supported by South West NHS Research and Development, Exeter NHS Research and Development, the Darlington Trust and the Peninsula National Institute of Health Research (NIHR) Clinical Research Facility at the University of Exeter. The opinions given in this paper do not necessarily represent those of NIHR, the NHS or the Department of Health.

Generation R: The Generation R Study is conducted by the Erasmus Medical Center in close collaboration with the School of Law and Faculty of Social Sciences of the Erasmus University Rotterdam, the Municipal Health Service Rotterdam area, Rotterdam, the Rotterdam Homecare Foundation, Rotterdam and the Stichting Trombosedienst and Artsenlaboratorium Rijnmond (STAR-MDC), Rotterdam. We gratefully acknowledge the contribution of children and parents, general practitioners, hospitals, midwives and pharmacies in Rotterdam. The generation and management of GWAS genotype data for the Generation R Study were done at the Genetic Laboratory of the Department of Internal Medicine, Erasmus MC, the Netherlands. We would like to thank Karol Estrada, Dr Tobias A. Knoch, Anis Abuseiris, Luc V. de Zeeuw and Rob de Graaf for their help in creating GRIMP, BigGRID, MediGRID and Services@MediGRID/D-Grid (funded by the German Bundesministerium fuer Forschung und Technology; grants 01 AK 803 A-H, 01 IG 07015 G) for access to their grid computing resources. We thank Mila Jhamai, Manoushka Ganesh, Pascal Arp, Marijn Verkerk, Lizbeth Herrera and Marjolein Peters for their help in creating, managing and QC of the GWAS database. Also, we thank Karol Estrada and Carolina Medina-Gomez for their support in the creation and analysis of imputed data. The Generation R Study is made possible by financial support from the Erasmus Medical Center, Rotterdam, the Erasmus University Rotterdam and the Netherlands Organization for Health Research and Development (ZonMw 21000074).

HAPO: This study was funded by grants R01-HD34242 and R01-HD34243 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development and the National Institute of Diabetes, Digestive, and Kidney Diseases, U01 HG004415 from the National Human Genome Research Institute, M01-RR00048 and M01-RR00080 by the National Center for Research Resources and by the American Diabetes Association. Genotyping was funded by Diabetes UK grant RD08/0003692.

MoBa: The Norwegian Mother and Child Cohort Study is supported by the Norwegian Ministry of Health and the Ministry of Education and Research, NIH/NIEHS (contract no NO-ES-75558), NIH/NINDS (grant no. 1 UO1 NS 047537-01) and the Norwegian Research Council/FUGE (grant no. 151918/S10 and no 183220/S10), Norwegian Research Council, Oslo, Norway (FUGE 183220/S10). Swedish government grants to researchers in public health service (ALF) (ALFGBG-136431), Sahlgrenska University Hospital, Sahlgrenska Academy, Gothenburg, Sweden, Swedish Medical Society, Stockholm, Sweden (2008-21198) and Jane and Dan Olsson Research Foundation, Gothenburg, Sweden.

MIDSPAN: We thank Dr Mark Upton, who led the Midspan Offspring Study, and the members of the research team who collected the data used in this analysis. The offspring study in MIDSPAN was supported by grants from the Wellcome Trust and the NHS Cardiovascular Research and Development Programme.

NCCGP: We thank all of the participants of the North Cumbria Community Genetics Project and previous members of the study team, in particular Professor Sir John Burn and Mrs Pat Jonas. Funding for genotyping and sample analysis of NCCGP samples has been provided from a number of sources including the National Institute of Health Research.

NFBC1966: We thank Professor Paula Rantakallio (launch of NFBC1966 and initial data collection), Ms Sarianna Vaara (data collection), Ms Tuula Ylitalo (administration), Mr Markku Koiranen (data management), Ms Outi Tornwall and Ms Minttu Jussila (DNA biobanking). This work was supported by the Academy of Finland (project grants 104781, 120315, 129418 and Center of Excellence in Complex Disease Genetics and SALVE), University Hospital Oulu, Biocenter, University of Oulu, Finland (75617), the European Commission (EURO-BLCS, Framework 5 award QLG1-CT-2000-01643), NHLBI (5R01HL087679-02) through the STAMPEED programme (1RL1MH083268-01), NIH/NIMH (5R01MH63706:02), ENGAGE project and grant agreement (HEALTH-F4-2007-201413) and the Medical Research Council, UK (G0500539, G0600705, PrevMetSyn/SALVE).

Raine: We gratefully acknowledge the Raine Study participants and their families and the Raine Study Team for cohort co-ordination and data collection. We also thank the NHMRC for their long-term contribution to funding the study and the Telethon Institute for Child Health Research for long-term support of the study. Core management for the Raine study is funded by The University of Western Australia (UWA), Telethon Institute for Child Health Research, Raine Medical Research Foundation, UWA Faculty of Medicine, Dentistry and Health Sciences, Curtin University and the Women and Infants Research Foundation. Collection, extraction and genotyping of maternal and child DNA in Raine was funded by the Australian National Health and Medical Research Council and the Women and Infants Research Foundation. Antenatal data collection was supported by the Raine Medical Research Foundation.

1958BC_T1DGC and 1958BC_WTCCC2: Dr Sue Ring and Dr Wendy McArdle (University of Bristol) and Mr Jon Johnson (Centre for Longitudinal Studies, Institute of Education, London) are thanked for help with data linkage. The study was supported by the Medical Research Council (MRC G0601653 and SALVE/PrevMedsyn). Collection of DNA in the 1958 Birth Cohort was funded by the MRC grant G0000934 and the Wellcome Trust grant 068545/Z/02. This research used resources provided by the Type 1 Diabetes Genetics Consortium, a collaborative clinical study sponsored by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), National Institute of Allergy and Infectious Diseases, National Human Genome Research Institute, National Institute of Child Health and Human Development and Juvenile Diabetes Research Foundation International (JDRF) and supported by U01 DK062418. This study makes use of data generated by the Wellcome Trust Case-Control Consortium II (083948). The Medical Research Council provides funds for the MRC Centre of Epidemiology for Child Health.

Conflict of Interest statement. None declared.