It is generally accepted that maternal smoking during pregnancy is associated with reduced offspring weight, height and head circumference at birth.^{1, 2} However, the literature on prenatal smoking and postnatal growth is less consistent. Studies following up children born to women who smoked during pregnancy suggest that deficits in weight and height observed at birth may persist in childhood and possibly even into adulthood.³–⁵ Prenatal smoking has also been associated with height components, particularly with leg length,⁶ which is especially sensitive to environmental factors in early childhood.⁷ Not all studies, however, support the finding of long-term detrimental effects of prenatal smoking on offspring weight or height.^{8, 9}

Overweight and obesity have reached epidemic proportions worldwide. In many countries, including the USA, the UK, Australia, China and Brazil, child overweight is increasing at a faster rate than adult obesity.¹⁰ There is evidence of tracking of overweight, with overweight children having a high risk of being overweight later in life.¹¹ Furthermore, several studies showed an increased prevalence of overweight and obesity in childhood in offspring of women who smoked during pregnancy¹²–¹⁵ and a recent meta-analyses strengthened these findings.¹⁶

Most of the studies that investigated the effects of maternal smoking during pregnancy on offspring growth during childhood have used data from high-income countries where confounding structures may differ from populations of low and middle-income countries.¹⁷ In addition, few of these studies have examined associations with paternal smoking to explore whether maternal smoking is of specific importance.

The objective of this study is to explore the effects of maternal smoking during pregnancy on offspring growth using three approaches: (1) multiple adjustments for socioeconomic and parental factors, (2) maternal–paternal comparisons as a test of putative intrauterine effects and (3) comparisons between two middle-income country birth cohort studies, the 1993 and 2004 Pelotas cohorts, in Brazil.

In the 1993 cohort study, 33.5% of mothers and 44.8% of fathers smoked during pregnancy. In 2004, the prevalence of prenatal smoking was 28% among mothers and 31% among fathers. Over the 11-year period, the prevalence of maternal and paternal prenatal smoking declined, however, this reduction was larger among fathers than among mothers (31% vs 16% for fathers and mothers, respectively).

In both cohort studies, maternal smoking during pregnancy was more likely among poorer, less educated mothers. Smoking mothers were more likely to be shorter, heavier and to have offspring of lower gestational age at delivery (table 1). Maternal smoking was also more likely among mothers of black/mixed ethnic origin. In the 1993 Pelotas cohort, the prevalence of smoking was lower among adolescent mothers.

Paternal smoking during pregnancy was more frequently seen among the poorest women, among those less educated, shorter and multiparous (table 1). In the 1993 Pelotas cohort study, paternal smoking status was not associated with maternal skin colour or maternal prepregnancy BMI. There was no association between paternal smoking status and maternal age, prepregnancy BMI or pregnancy duration in the 2004 Pelotas cohort study (table 1).

In unadjusted analyses (tables 2 and 3), maternal smoking in pregnancy in both cohorts was associated with lower offspring weight, length and head circumference at every follow-up, and with lower BMI at birth only. In fully adjusted analyses (adjusted for maternal characteristics, pregnancy duration and paternal smoking), maternal smoking in pregnancy was associated with lower offspring length at every follow-up and was associated with lower offspring BMI at birth but higher BMI at follow-ups at 12 months and older. While maternal smoking in adjusted analyses was consistently (in both cohorts) associated with lower offspring weight and head circumference at birth, this was not observed at later follow-ups, except for head circumference at 3 months in the 2004 cohort.

Statistical differences between the adjusted maternal and paternal associations were found for offspring length at every follow-up, weight at birth, BMI at birth and at follow-ups at 12 months and older and head circumference at birth.

No evidence of interaction between maternal and paternal smoking on any offspring outcome was found in the 1993 and 2004 cohort studies.

Maternal prenatal smoking was associated with a reduction in offspring trunk length, leg length and leg-to-trunk ratio at 48 months in the 2004 cohort, after adjustment for the child's age at the time of measurement and sex (table 4). Although additional adjustment for maternal factors, pregnancy duration and paternal smoking during pregnancy weakened associations, maternal prenatal smoking was still related to leg length and leg-to-trunk ratio. The associations between paternal smoking and trunk length, leg length and leg-to-trunk ratio were considerably weaker than those for maternal smoking and disappeared after adjustment for confounders. No differences were found between the sexes.

The present study found evidence of maternal-specific effects of smoking during pregnancy on child's growth. Consistent results were observed between the 1993 and 2004 Pelotas cohort studies, showing that even after adjustment for potential confounding factors, offspring of women who smoked during pregnancy had persistent reductions in length/height at repeated follow-ups from birth to 48 months, had shorter leg length at 48 months, and had higher BMI-for-age z scores at follow-ups at 12 months and older than children of non-smokers. The magnitude of maternal smoking effects on child's growth was greater than and significantly different from paternal smoking effects.

Several prenatal and postnatal factors influence childhood height.²⁴–²⁷ Childhood height is the best predictor of adult height, which is associated with adult health and human capital.²⁸ Leg length growth contributes markedly to stature attainment and is highly influenced by environmental factors in early childhood.⁷ In our study, consistent reductions in offspring height and height components (in particular leg length and leg-to-trunk ratio) were seen among offspring of women who smoked during pregnancy, findings that have also been reported by Leary et al⁶ at 7.5 years in the Avon Longitudinal Study of Parents and Children. Although there is some evidence in the literature that the effects of maternal smoking during pregnancy on offspring height decrease with greater offspring age,²⁹ this was not observed in the present study. However, follow-up was limited here to 48 months after birth. Therefore, analyses of future follow-ups in the 2004 cohort will enable further exploration of the long-term effects of maternal prenatal smoking on offspring height.

Smoking is a strongly socially patterned behaviour, thus it is possible that the association between maternal prenatal smoking and child height is due to confounding by wider socioeconomic factors. Although associations attenuated after adjustment for possible confounders, persisting associations were still observed. Furthermore, the markedly discordant maternal–paternal associations observed suggests that maternal smoking during pregnancy may have specific intrauterine effects on offspring height and that residual confounding may be a less likely explanation. Indeed, studies have reported embryotoxic effects of nicotine or other toxic pollutants found in cigarette smoke that lead to delayed skeletal growth.³⁰

Our findings support previous work showing that reductions in weight at birth in children born to mothers who smoke during pregnancy are overcome in infancy.⁹ The combination of weight being recovered and persisting deficits in height could result in greater body mass in children of maternal smokers. In our study higher BMI z scores were seen among the offspring of women who smoked during pregnancy. Our results are in line with previous studies that reported the association between maternal smoking during pregnancy and childhood overweight and obesity,^{12, 16} and several hypotheses have been postulated to explain this relationship.³¹ The persistence of the association between maternal prenatal smoking and offspring BMI z scores after adjustment for possible confounders, the discordant maternal–paternal associations observed and the consistency of the findings between the 1993 and 2004 Pelotas cohort studies again make confounding a less likely explanation.

We found a specific negative effect of maternal smoking on head circumference at birth. Our results showed that this association is independent of the relationship between premature birth and maternal smoking. However, deficits in head circumference did not persist in infancy, as also shown in other studies.^{4, 8}

A major strength of the present study was the method of data collection (prospective information obtained among large unselected populations and comparable timescales) combined with the use of standardised anthropometric measurements performed by trained fieldworkers, high follow-up rates and low missing data (below 5%) for most variables in both cohort studies. However, some methodological difficulties of the study need to be discussed. First, it is possible that different results would have been obtained if all children whose mothers originally enrolled in the 1993 cohort study were included in the analysis instead of only subsamples. However, no differences in maternal smoking during pregnancy or in socioeconomic or maternal characteristics were found between those children who were followed-up in the 1993 cohort and those originally enrolled in the study (data not shown, available on request), providing some reassurance that attrition is unlikely to have biased the results. Second, both maternal and paternal smoking data rely on maternal self-report. Even though the validity of self-reported smoking has been demonstrated in previous studies and would be acceptable for maternal smoking status,³² no validation has been carried out in our study for paternal information. However, we might anticipate similar, if not even less, reporting bias for paternal smoking than for maternal smoking in pregnancy, as there is likely to be more social pressure on women than on men to not report smoking during pregnancy. Third, even though anthropometric measurements were performed by trained fieldworkers following standardised procedures, leg length (calculated as the difference between standing height and sitting height) is likely to have some measurement error which could have decreased the precision of estimated associations. Finally, it would have been interesting to study the association between maternal passive smoking during pregnancy and child growth. However, no data were available on whether mothers were exposed to passive smoking during pregnancy (ie, if their partner or other people smoked near the woman at home or at work during pregnancy).

Our findings further emphasise the importance of advising women on smoking cessation, particularly in low- and middle-income countries like Brazil. Women of reproductive age should be advised to stop smoking, preferably before attempting to become pregnant, in order to protect their offspring from impaired linear growth and obesity in childhood with potential permanent consequences in adult life.

Previous studies have consistently reported associations between maternal smoking in pregnancy and offspring reduced height as well as overweight/obesity, but have been unable to ascertain whether or not associations are causal. This study explored causality based on three approaches: comparisons between two cohort studies 11 years apart, multiple adjustments for confounders and maternal–paternal smoking comparisons. In addition, the effects of maternal smoking on growth at multiple time points during infancy were explored. Although residual confounding cannot be ruled out, the persisting associations following confounder adjustment, the stronger maternal versus paternal smoking associations, the specificity of the maternal smoking associations and the consistency between the 1993 and 2004 Pelotas results strongly suggest a possible causal effect of maternal smoking during pregnancy on offspring impaired linear growth and overweight in childhood via intrauterine mechanisms.

Funding This research was specifically funded by the Brazilian National Research Council (CNPq), grant no. 481010/2009-5. The 1993 and 2004 Pelotas birth cohort studies are currently supported by the Wellcome Trust Initiative entitled ‘Major Awards for Latin America on Health Consequences of Population Change’, grant entitled ‘Implications of early life and contemporary exposures on body composition, human capital, mental health and precursors of complex chronic diseases in three Brazilian cohorts’ (grant no. 086974/Z/08/Z). Previous phases of the 1993 cohort study were supported by the European Union and the Fundação de Amparo à Pesquisa do Rio Grande do Sul, Brazil. Previous phases of the 2004 cohort study were supported by WHO (Division of Child and Adolescent Health and Development), the Brazilian National Research Council (CNPq) and the Pastoral da Criança (Catholic non-governmental organisation, Curitiba, Brazil).

Competing interests None.

Ethics approval This study was conducted with the approval of the Medical Ethics Committee of the Federal University of Pelotas, affiliated with the Brazilian Federal Medical Council.

Provenance and peer review Not commissioned; externally peer reviewed.

The authors are extremely grateful to all families who took part in the 1993 and 2004 Pelotas birth cohort studies, and the entire Pelotas cohort team, including interviewers, data clerks, laboratory technicians and volunteers. The authors are grateful to Professors Debbie Lawlor and John Judkin for their helpful feedback in preliminary analyses of this study.