Article Text

Download PDFPDF

Active and passive smoking during pregnancy and ultrasound measures of fetal growth in a cohort of pregnant women
  1. Carmen Iñiguez1,2,
  2. Ferran Ballester1,2,
  3. Rubén Amorós1,3,
  4. Mario Murcia1,2,
  5. Alfredo Plana4,
  6. Marisa Rebagliato2,5
  1. 1Centre Superior d'Investigació en Salut Pública, Conselleria de Sanitat, Valencia, Spain
  2. 2Ciber de Epidemiología y Salud Pública, CIBERESP, Barcelona, Spain
  3. 3Universidad CEU-Cardenal Herrera, Valencia, Spain
  4. 4Hospital Materno Infantil “La Fe”, Valencia, Spain
  5. 5Universidad Miguel Hernández, Alicante, Spain
  1. Correspondence to Carmen Iñiguez, Centre Superior d'Investigació en Salut Pública, Avda Cataluña no 21, 46020 Valencia, Spain; inyiguez_car{at}gva.es

Abstract

Background In utero tobacco exposure has been associated with adverse pregnancy outcomes but few studies have used longitudinal ultrasound measurements to asses the effects on fetal growth. The aim of this study was to examine the impact of active and passive smoking during pregnancy on fetal biometry in a cohort of Spanish women.

Methods Biparietal diameter (BPD), abdominal circumference (AC), femur length (FL) and estimated fetal weight (EFW) were evaluated in each trimester of pregnancy. Detailed information on smoking and potential confounders was assessed by questionnaire. SD scores were calculated from longitudinal growth curves adjusted for gestational age and potential determinants of growth. Size was assessed by means of unconditional SD scores at 12, 20, 32 and 38 weeks of pregnancy, while growth between these points was assessed by means of conditional SD scores. The association between smoking and fetal growth was investigated by regression models and adjusted for sociodemographic and lifestyle-related variables.

Results Maternal smoking was inversely associated with size of all parameters at weeks 32 and 38 and with growth in 20–32, 12–32 and 12–38 week intervals. In 32–38 weeks the effect was significant for AC and EFW. Environmental tobacco smoke (ETS) exposure was inversely associated with growth in BPD in all the intervals except 32–38 weeks.

Conclusions Active smoking during pregnancy was associated with a reduction in BPD, AC, FL and EFW from mid-gestation. ETS adversely affected BPD from early pregnancy.

  • Fetal
  • longitudinal studies
  • multilevel models
  • passive smoking
  • smoking & pregnancy

Statistics from Altmetric.com

Request Permissions

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.

Introduction

Maternal smoking during pregnancy has been linked with preterm delivery, intrauterine growth restriction and reduced fetal survival.1 2 This exposure in utero is also associated with health problems in childhood and adult life, such as respiratory tract infections and impaired cognitive development.3 4 Environmental tobacco smoke (ETS) contains similar toxins that smokers inhale; hence, maternal ETS exposure during pregnancy is likely to have similar effects.5

Regarding the effects on fetal growth, most epidemiological evidence focuses on birth weight. It is known that active smoking leads to a decrease in birth weight of approximately 150–250 g6 and ETS exposure leads to a smaller but significant decrease of 25–40 g.7 8 However, birth weight may be inappropriate for assessing growth as the same size may represent both normal and abnormal growth in different subjects. Furthermore, birth weight, like any other neonatal measurement, is unable to identifying critical periods within gestation. In fact, it is still unclear whether the impact of smoke on fetal growth is a continuous process or at what moment of gestation it can first be detected. Finally, birth weight is a global measure and it could be useful to add evidence on the development of specific fetal characteristics.

Despite the great body of evidence regarding the impact of smoking on pregnancy outcomes, there are few studies examining the impact on fetal biometry9–18 and many of them are small studies or follow a cross-sectional design.

Here, our aim was to examine the effects of self-reported active and passive smoking on fetal biometry in different periods of pregnancy using serial measurements of the same individual and taking into account their growth potential.

This study was conducted among the women participating in a cohort of pregnant women within the Spanish ‘Environment and Child’ (INMA) project. INMA is a collaborative network of seven prospective birth cohort studies in Spain, which was designed to explore the role of maternal diet and environmental pollution on fetal health.19

Methods

Subjects

The initial study population comprised 855 pregnant women resident in Valencia (Spain). They were enrolled at early pregnancy (before 13 weeks of gestation) in the main public hospital in the area. Criteria for inclusion in INMA were a wish to deliver in this hospital, to be at least 16 years of age, a singleton pregnancy, no chronic disease, no communication problems and not taken part in an assisted reproduction programme. Eligible women for this analysis also had to verify to have given a live birth (n=787), provide smoking information (n=780) and have had at least two valid ultrasounds (n=818); thus, 780 mothers were finally included. All participating women gave their written informed consent and the study was approved by the Ethics Committee of the Hospital.

Fetal ultrasonography

Ultrasound examinations for all women were scheduled in weeks 12, 20 and 32 of pregnancy. An additional 100 ultrasounds were scheduled in week 38. All measurements were performed at routinely scheduled antenatal care visits by specialised obstetricians. Apart from the scheduled ultrasounds, we had access to the records of any other ultrasound performed to women in the same hospital unit during their pregnancy allowing us to obtain 2–5 valid ultrasounds per woman between 10 and 41 weeks of gestation. An early crown-rump length (CRL) measurement was also obtained and it was used for pregnancy dating. Gestational age was established using CRL when the difference with the age based on the self-reported last menstrual period (LMP) was 7 days or more. Only in one case was this difference more than 3 weeks and it was excluded in order to avoid possible bias. Fetal characteristics examined in this study were biparietal diameter (BPD), femur length (FL) and abdominal circumference (AC). All measurements were performed in millimetres using transabdominal ultrasound examination (equipment: Voluson Voluson 730 Pro and 730 Expert, Siemens Sienna, Berlin, Germany) and followed standardised procedures.20 Sonographists conducted a validation study to determine the inter-observer reliability. Intra-class correlation coefficients were in the range of 0.80–0.91 and coefficients of variation were lower that 5% except in the case of FL (coefficients of variation=7.8%) for which additional consensus in the measurement procedure was required.

Smoking information

Active maternal smoking (AS) and ETS were assessed by a questionnaire administered at week 32 of pregnancy. Regarding AS, mothers were classified as ‘non-smokers during pregnancy’, ‘smokers who give up smoking before week 12’ and ‘smokers still at week 12’. The last category includes women who continued smoking at least until week 32 of pregnancy.

ETS exposure was defined as yes/no in several environments: home, workplace, restaurants and leisure areas such as pubs or bars. Mothers were classified as exposed at home if they reported that their partner, other people living in the same house or frequent visitors (>2 times per week) smoked at home. ETS at workplace was assessed in the following four categories: ‘nothing’, ‘little’, ‘quite a bit’ and ‘a lot’. Mothers who answered ‘quite a bit’ or ‘a lot’ were classified as exposed. Finally, mothers were classified as exposed to ETS at restaurants and leisure areas if they reported to be exposed in these places more than twice a week.

Questionnaire validity

To evaluate the validity of smoking information from the questionnaire, urine samples from 708 women at week 32 of pregnancy were analysed for total cotinine (enzyme immunoassay (EIA)—Cotinine micro-plate EIA, OraSure Tech, Inc., Bethlehem, Pennsylvania, USA). Cotinine levels were log-transformed due to the skewness of the distribution and the log cotinine levels were compared to self-reported levels. Smokers at week 32 had a geometric mean of cotinine levels extremely higher than non-smokers (1540.7 ng/ml vs 7.48 ng/ml; p<0.001; n=698). Concerning ETS, the mother's cumulative number of ETS exposure sources was determined by adding the number of different sources to which she was exposed (home+work+restaurants+leisure areas). This variable was categorised in 0, 1 and 2 or more. Those exposed to ETS from one source had approximately two times the geometric mean cotinine than non-exposed (6.23 ng/ml vs 3.42 ng/ml; p<0.001) and those exposed to ETS from at least two sources had approximately three times the geometric mean cotinine than those not exposed (9.78 ng/ml vs 3.42 ng/ml; p<0.001). For ETS, analyses were restricted to non-smoking women (n=413).

Covariates

Detailed information on sociodemographic characteristics, paternal anthropometry, lifestyle variables and environmental exposures was obtained from two questionnaires administered by trained interviewers at weeks 12 and 32 of pregnancy.

Maternal anthropometric factors, including height (cm), pre-pregnancy weight (kg), parity, age and ethnicity were used for modelling fetal growth, together with fetal gender, father's height and a dummy variable, which identified mothers who had ultrasounds too close in time (18, 21 and 30 days were examined as possible intervals to define ‘too close’). Body mass index (BMI), gestational weight gain, socio-occupational status, education, employment, country of origin, zone of residence, season of conception, alcohol and caffeine consumption, vegetable, fruit, and energy intake, and exposure to ambient NO221 were used as confounders of association between smoking and fetal growth. BMI and gestational weight gain were classified following the Institute of Medicine guidelines.22 Social class was classified in three occupational categories according to current or most recent occupation using an adaptation of the British classification system.23 Alcohol and caffeine consumption was considered in three categories: 0, 0–1 and >1 g/d, and 0–100, 100–200 and ≥200 mg/d, respectively. Vegetable, fruit and energy intake was estimated through a food questionnaire and considered in quintiles.24

Fetal growth modelling

Linear mixed models25 were used to obtain longitudinal growth curves for BPD, AC, FL and for estimated fetal weight (EFW) using the Hadlock algorithm.26 Box-Cox transformations were used on these outcomes in order to normalise them. Each transformed outcome was modelled as a polynomial of gestational age in days until degree 3. Physiological determinants of growth (described below) and their interactions with days of gestation were tested using the likelihood ratio (LR) test (p<0.05) through a forward selection procedure.

Models were adjusted for physiological factors known to affect fetal growth in order to get an individualised rather than a population-based growth standard. Significant variables considered were maternal height, age, parity, ethnicity and pre-pregnancy weight, father height and fetus gender. The use of individualised standards is expected to reduce misclassification in the ‘small for gestational age’ detection excluding constitutionally small babies and including those within normal population limits who should have reached a greater size.27

The running time between ultrasounds was used to model the correlation structure for within-subjects errors. Gestational age, sex, parity, ethnicity and the dummies identifying mothers with ultrasounds too close in time were considered to estimate their variance (heteroscedasticity). Random effects on intercept and/or slope (days of gestation) were allowed and tested using the LR test (p<0.05). Finally, the goodness of fit was assessed by consideration of the normality and independence of the residuals.

Mean, SDs and predictions for weeks 12, 20, 32 and 38 conditioned on the nearest measurement were obtained from fetal curves and were used for calculating unconditional and conditional SD scores.28 The unconditional SD score at a certain time point describe the size at that time. The SD score at the end of a time-interval conditioned on the value at the starting point describes the growth experienced in this interval.28–30 Unconditional z-scores were obtained at 12, 20, 32 and 38 weeks of gestation and conditional SD scores were obtained for the intervals: 12–20, 20–32, 32–38, 12–32 and 12–38 weeks.

Analysis of the association

The association between smoking and each SD score was analysed by means of multiple linear regression models adjusted for the potential confounders described above (p<0.1, LR test). Models relating to week 38 are restricted to cases with a valid ultrasound examination within 2 weeks of week 38. Models for assessing the impact of ETS exposure were restricted to non-smoking mothers.

All measures of association are expressed as the percentage of change with respect to the mean and its 95% CI.

Results

Study population characteristics

Study population characteristics according to active maternal smoking status are shown in table 1.

Table 1

Characteristics of mothers and pregnancies according to maternal smoking during pregnancy

Curves of fetal characteristics

A total of 2515 fetal ultrasound examinations for 818 women were used to construct longitudinal growth curves for all fetal characteristics and gestational age adjusted SD scores. Altogether, 626 women (76.5%) had three ultrasound examinations, 109 (13%) had four and 12 (1.5%) had five or six. Gestational ages were very close to the planned schedule for the three first ultrasound examinations (12, 20 and 32 weeks) and lower than expected for the fourth one (38 weeks). Median gestational ages were 12.4 (95% CI 11.4 to 13.7), 20.3 (95% CI 19.1 to 21.6), 32.1 (95% CI 26.6 to 33.1) and 35.7 (95% CI 30.3 to 39.4) weeks, respectively.

In all cases, after adjusting for covariates and modelling the variance-covariance structure for within-subject errors, random effects were not necessary; therefore, model estimation was performed by using the generalised least-squares approach.25 The best-fitting models always included a 2nd or 3rd degree polynomial of gestational age. Pre-pregnancy weight, age of mother, fetus sex and father's height were the covariates most frequently included. Models always incorporated an exponential variogram25 of the days between ultrasound examinations as correlation structure for within-subject errors. Heteroscedasticity was usually found according to the factor identifying two ultrasounds close in time. Curves for the fetal characteristics are shown in the figure 1.

Figure 1

Longitudinal curves of fetal characteristics. INMA study, cohort of Valencia, 2004–2006. Each graph shows the mean and its 95% CI for a fetus with relevant covariates in their median values. BPD model: BPD0.66= GA+ GA2+ age+weight2+ GA×weight+ GA×sex+GA×height. Heteroscedasticity according to d21 femur length (FL) model: LF0.78=GA+GA2+GA3+ cage+ weight+ GA×weight+GA×weight2+GA×fheight+GA×parity+GA×height. Heteroscedasticity according to parity PA model: PA0.44=GA+GA2+GA3+sex+age+GA×weight+GA×weight2+GA×fheight. Heteroscedasticity according to GA and d30 EFW model: Ln(EFW)=GA+GA2+cage+weight+sex+GA×weight+GA×weight2+GA×fheight. Heteroscedasticity according to d30 where GA: gestational age; age: mother's age; cage: mother's age in three categories (≪Q25, Q25–Q75, ≫Q75); weight: pre-pregnancy mother's weight; sex: fetus gender; height: mother's height; fheight: father's height parity (0, 1, ≥2); d21 and d30: dummies of mothers with two ultrasounds less than 21 and 30 days apart, respectively. AC, abdominal circumference; BPD, biparietal diameter; EFW, estimated fetal weight; FL, femur length; GA, gestational age.

Smoking and fetal growth

Active smoking before pregnancy was reported by 41.1% of participants; only 44% of them (n=141) gave up smoking during pregnancy. Most of the women who stopped did so before week 12 (n=126), 15 women stopped between weeks 12 and 32, and 178 women continued smoking at least until week 32 of pregnancy. Passive smoking at home, the workplace, restaurants and leisure areas was reported by 35.0, 15.8, 30.0 and 23.1% of non-smoking mothers, respectively.

Maternal smoking was inversely associated with the size of all characteristics at weeks 32 and 38 and with growth in the periods 20–32, 12–32 and 12–38 weeks (table 2; figure 2). For AC and EFW a significant effect was also found in weeks 32–38, whereas BPD and FL were less affected at this stage. The size of observed effects was increasing with gestational age and similar in the two groups of smokers relating to non-smoking mothers. However, they were significant only for mothers who continued to smoke at week 12. The only exception was FL for which the difference between non-smoking mothers and ex-smokers before week 12 was also significant.

Table 2

Association between maternal smoking during pregnancy and fetal growth characteristics

Figure 2

% Change in fetal characteristics associated with maternal smoking during pregnancy. Non-smokers during pregnancy as reference category; smoking (1): ex-smokers before week 12; smoking (2): smokers at week 12.

Regarding passive smoking, effects were small and non-significant for all places except for BPD and exposure at restaurants. Size in BPD was inversely associated with exposure from week 20 and growth was inversely affected from week 12 but not significantly affected in the last interval (table 3; figure 3).

Table 3

Association between environmental tobacco smoke at restaurants during pregnancy and biparietal diameter

Figure 3

% Change in biparietal diameter (BPD) associated with environmental tobacco smoke (ETS) at restaurants during pregnancy.

Discussion

This study showed an association between maternal smoking during pregnancy and impaired fetal growth in all parameters as early as mid-pregnancy. In smokers at week 12, size was reduced around 2% at week 32 and around 10% at week 38 for all parameters. In ex-smokers before week 12, effects were of similar magnitude.

In the generation R Study17 maternal smoking was inversely associated with head circumference (HC), FL and AC in late pregnancy (≥25 weeks) and only with FL in mid-pregnancy (18–24 weeks). Mercer et al31 found maternal smoking to be associated with small fetal size from 10 to 19 weeks. However, the results of Bergsjo et al32 did not reveal any significant association at week 17.

There are several hypotheses about the mechanisms that could explain the negative impact of smoking on fetal growth: (1) a toxic component operates directly, which may affect growth from an early stage; (2) there is a nutritional deprivation due to slowed blood circulation—in this case, growth retardation may become more pronounced as pregnancy advances; and (3) maternal smoking in pregnancy occurs in the context of other risk factors such as co-abuse of other substances, dietary restriction, and so forth. Results of Jaddoe,17 Bergsjo32 and Lampl13 favour the second hypothesis while, in relation to effects on head size, Roza et al16 defend the first one. Our results concerning maternal smoking also favour the second hypothesis as growth in all fetal parameters was affected from mid-pregnancy leading to a reduced size from week 32. The fact that BPD and FL were less affected in late pregnancy than AC and EFW could be explained by their typical patterns through gestation (see figure 1).

Relating to secondhand smoke, ETS at restaurants was associated with reduced BPD from week 12 but no effect was found in late pregnancy suggesting the existence of a critical window of exposure in early pregnancy on fetal brain development.

The adverse effect found early in BPD only among passive smokers could be partially explained by differences in the placental enzyme activity between active and passive smokers. In active smokers, the induction of placental enzymes protects the fetus whereas the small amount of smoke in passive smokers is unable to induce the enzymes and the smoke remains toxic.33 Like us, Kalinka et al14 and Hanke et al15 also observed a significant adverse effect of ETS exposure in BPD at 20–24 weeks' gestation.

Regarding the clinical importance of this association, recent research has highlighted that the impaired neurodevelopment noted among children exposed to smoking in the uterus could be partially attributed to a reduction in brain size besides the neurotoxic effect of other components such as nicotine.34

We realise there are some limitations in our study. First, smoking information proceeds from a questionnaire, which could lead to under-reporting of maternal smoking.14 However, some authors35 36 consider that the use of biomarkers is not superior to the use of self-reports in studying the effect of maternal smoking in pregnancy. Second, the sample of ultrasounds at the last stage of pregnancy has a limited size leading to low statistical power of analyses relating to week 38. Third, this is not exactly a population-based study because it was restricted to women having a live birth baby and fulfilling the established inclusion criteria may impose some selection bias: no chronic diseases, greater homogeneity in education, age, and so forth.

In our study, LMP-based gestational dates were confirmed or corrected by an early CRL measurement. That was done because the use of self-reported LMP for gestational dating is prone to large random measurement error because of the large number of women who do not know the exact date of their last menstrual period or have irregular menstrual periods. But if smoking had had an adverse effect on fetal growth previous to this first scan (ie, before week 14), this procedure would erroneously shorten the age of affected fetuses leading to an underestimation of the impact.37 38 In our case, LMP- based ages were shortened to the same extent in smokers and non-smokers (median and Q5–Q95 of the shortening in days were 1.4 (−1.4 to 2.3) and 1.2 (−1.5 to 2.8), respectively; p=0.630). Furthermore, like Olsen et al,39 we think that large random errors probably affect estimates much more than smaller systematic errors.

Most of the published studies used covariate data collected from birth certificates. To this respect, a significant advantage of our study is the availability and quality of individual information on potential confounders. A major strength of this study is the repeated measurements of fetal biometry, which allowed us to observe effects in different stages of pregnancy. The other is the use of a longitudinal model adjusted for constitutional factors, which allowed us to properly assess fetal growth by considering each fetus as its own control. To our knowledge, there are really few studies following a similar approach.18 19 Finally, the calculated SD scores were confirmed as a valuable tool for examining the impact of other environmental exposures on fetal parameters in different stages of pregnancy.

In conclusion, our results on active smoking indicated that smoking cessation early during pregnancy (ie, before week 12) was found to keep differences with non-smokers reinforcing the need to encourage pregnant women to avoid smoking during pregnancy. Our results on passive smoking indicated the need to adequately restrict smoking in public places like bars or restaurants in order to protect the health of the most vulnerable.

Acknowledgments

We would like to thank all the parents who are taking part in the study for their generous collaboration and other members of the INMA study group in Valencia involved in this sub-study: Maria Andreu, Amparo Cases, Elena Creuá, Ana Esplugues, Marisa Estarlich, Virginia Fuentes, Francisco Garcia, Ana M García, M Carmen Gonzalez, Marina Lacasaña, Gemma Leon, M Jose López-Espinosa, Sabrina Llop, Alfredo Marco, Maria Monzonís, Alicia Moreno, Sandra Pérez, Amparo Quiles, Rosa Ramón, M Paz Rodriguez, Clara Rodríguez-Bernal, Elena Romero, Jesus Vioque.

References

Footnotes

  • Funding This study was funded by grants from the Instituto de Salud Carlos III (Red INMA G03/176, CB06/02/0041, FIS-FEDER 03/1615, 04/1509, 04/1112, 04/1931, 05/1079, 05/1052, 06/1213, 07/0314 and 09/02647) and the Conselleria de Sanitat Generalitat Valenciana, Spain, and Fundación Roger Torné.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval This study was conducted with the approval of the Hospital La Fe de Valencia.

  • Provenance and peer review Not commissioned; externally peer reviewed.