Wang X, Zuckerman B, Pearson C, Kaufman G, Chen C, Wang G, Niu T, Wise PH, Bauchner H, Xu X. Maternal Cigarette Smoking, Metabolic Gene Polymorphism, and Infant
Birth Weight. JAMA. 2002;287(2):195–202. doi:10.1001/jama.287.2.195
Author Affiliations: Department of Pediatrics (Drs X. Wang, Zuckerman, Wise, and Bauchner and Ms Pearson) and Department of Obstetrics and Gynecology (Dr Kaufman), Boston University School of Medicine and Boston Medical Center, and Program for Population Genetics, Harvard School of Public Health (Drs Chen, Niu, and Xu), Boston, Mass; and Center for Ecogenetics and Reproductive Health, Beijing Medical University, Beijing, China (Dr G. Wang).
Context Little is known about genetic susceptibility to cigarette smoke in relation
to adverse pregnancy outcomes.
Objective To investigate whether the association between maternal cigarette smoking
and infant birth weight differs by polymorphisms of 2 maternal metabolic genes: CYP1A1 and GSTT1.
Design, Setting, and Participants Case-control study conducted in 1998-2000 among 741 mothers (174 ever
smokers and 567 never smokers) who delivered singleton live births at Boston
Medical Center. A total of 207 cases were preterm or low-birth-weight infants
and 534 were non–low-birth-weight, full-term infants (control).
Main Outcome Measure Birth weight, gestation, fetal growth by smoking status and CYP1A1 MspI (AA vs Aa
and aa, where Aa and aa were combined because of small numbers of aa and similar results), and GSTT1 (present
vs absent) genotypes.
Results Without consideration of genotype, continuous maternal smoking during
pregnancy was associated with a mean reduction of 377 g (SE, 89 g) in birth
weight (odds ratio [OR], 2.1; 95% confidence interval [CI], 1.2-3.7). When CYP1A1 genotype was considered, the estimated reduction
in birth weight was 252 g (SE, 111 g) for the AA
genotype group (n = 75; OR, 1.3; 95% CI, 0.6-2.6), but was 520 g (SE, 124
g) for the Aa/aa genotype
group (n = 43 for Aa, n = 6 for aa; OR, 3.2; 95% CI, 1.6-6.4). When GSTT1
genotype was considered, the estimated reduction in birth weight was 285 g
(SE, 99 g) (OR, 1.7; 95% CI, 0.9-3.2) and 642 g (SE, 154 g) (OR, 3.5; 95%
CI, 1.5-8.3) for the present and absent genotype groups, respectively. When
both CYP1A1 and GSTT1 genotypes
were considered, the greatest reduction in birth weight was found among smoking
mothers with the CYP1A1 Aa/aa and GSTT1 absent genotypes (−1285 g; SE, 234 g; P<.001). Among never smokers, genotype did not independently confer
an adverse effect. A similar pattern emerged in analyses stratified by maternal
ethnicity and in analyses for gestation.
Conclusions In our study, maternal CYP1A1 and GSTT1 genotypes modified the association between maternal cigarette
smoking and infant birth weight, suggesting an interaction between metabolic
genes and cigarette smoking.
In the United States, 65% of all infant deaths occur among low-birth-weight
(LBW) infants (<2500 g); LBW infants account for 7.6% of all live-born
infants.1 The etiology of LBW is largely unknown,
but both environmental and genetic factors may play a role.2
Numerous studies have shown that maternal cigarette smoking during pregnancy
is associated with reduced birth weight or increased risk of LBW.3- 8
In 1997, 13.2% of US women reported smoking cigarettes during pregnancy.1 Maternal cigarette smoking is identified as the single
largest modifiable risk factor for intrauterine growth restriction in developed
countries.9,10 However, not all
women who smoke cigarettes during pregnancy have LBW infants. The reason for
this variability is largely unknown, but may be related to maternal genetic
Tobacco smoke contains approximately 4000 compounds11;
the most important carcinogens in tobacco smoke are polycyclic aromatic hydrocarbons
(PAHs), arylmines, and N-nitrosamines.12
The ability of an individual to convert toxic metabolites of cigarette smoke
to less harmful moieties is important for minimizing the adverse health effects
of these compounds. Using PAHs as an example, the metabolic processing of
PAHs in humans involves 2 phases. The phase 1 metabolism is an activation
process, in which the inhaled, hydrophobic PAHs are converted mainly through
aryl hydrocarbon hydroxylase activity into hydrophilic, reactive, electrophilic
intermediates that can bind covalently to macromolecules, especially DNA.13 These intermediates may be more toxic than the original
form. Aryl hydrocarbon hydroxylase, encoded by the CYP1A1 gene, is a well-studied phase 1 enzyme and is particularly relevant
to the metabolism of chemicals in cigarette smoke. The phase 2 metabolism
is a detoxification process, in which these metabolic intermediates are detoxified
by enzymes such as glutathione S-transferases (GSTs) or uridine diphosphate
(UDP)-glucuronosyltransferase through transformation into conjugated forms
that are sufficiently polar to be excreted from the body.14
GSTT1, encoded by the GSTT1 gene, is a major phase
2 enzyme. There is evidence that the adverse health effects of cigarette smoke
may depend on the combined effects of phase 1 and phase 2 metabolism.15- 17
Both CYP1A1 and GSTT1
genes are highly polymorphic in the population18- 20
and their polymorphisms have been associated with their encoded enzyme activities.21,22 The expression of different host
genotypes may explain varying susceptibility to the adverse health effects
of cigarette smoke.
We hypothesized that the association between maternal cigarette smoking
during pregnancy and reduced birth weight or increased risk of LBW is modified
by maternal genetic susceptibility. In this report, CYP1A1 and GSTT1 gene polymorphisms are used to
characterize genetic susceptibility and to assess the interaction between
metabolic genes and cigarette smoking. We chose to focus on these specific
gene polymorphisms not only because such an interaction is biologically plausible,
but also in light of previous research that found evidence of interaction
between these gene polymorphisms and benzene exposure on gestation duration.23 In addition, these gene variants are common in our
study population, permitting us to examine gene–cigarette smoke interactions.
Between 1998 and 2000, we conducted a molecular epidemiological study
on environmental and genetic determinants of LBW (<2500 g) and preterm
birth (<37 weeks' gestation) among mothers who delivered at Boston Medical
Center, using a case-control design. Boston Medical Center serves a multi-ethnic
population of pregnant women, many of whom are from the inner city. The overall
rates of LBW and preterm birth are approximately 12% and 15% in this population
compared with the national average of 7.6% and 11.8%, respectively.1 More than 80% of study mothers had at least 1 prenatal
ultrasound examination; most examinations were performed prior to 20 weeks'
gestation. Cases were defined as women who delivered singleton, live, LBW
or preterm infants regardless of birth weight; controls were matched for age
and ethnicity and were defined as women who delivered singleton, live, term
infants with birth weight 2500 g or more. Three controls were identified for
every case. Multiple-gestation pregnancies (eg, twins, triplets) or newborns
with major birth defects were excluded. The study protocol was approved by
the Boston Medical Center Institutional Review Board and by the Massachusetts
Department of Public Health.
All eligible cases, including those women who delivered on weekends
and holidays, were approached postpartum by our research staff. The participation
rate was 90% and 85% among approached eligible cases and controls, respectively.
There was no significant difference between participants and nonparticipants
in infant birth weight, maternal ethnicity, or other sociodemographic characteristics.
After informed consent was obtained, a questionnaire interview was conducted
to obtain relevant information including demographic characteristics, cigarette
smoking, alcohol consumption, and medical and reproductive history. Maternal
and infant medical records were reviewed to obtain clinical data including
prenatal care, pregnancy complications, and birth outcomes (infant's sex,
gestational age, and birth weight). A maternal blood sample was obtained and
DNA was extracted according to standard protocol.24
The information on maternal smoking was based on maternal self-reporting
and was obtained for 4 time periods: 3 months before the index pregnancy and
the first, second, and third trimesters of the index pregnancy. In our study
sample, mothers' data were clustered in 3 groups in terms of cigarette smoking:
those who did not smoke throughout the index pregnancy; those who smoked during
early pregnancy but quit smoking during the first trimester; and those who
smoked continuously during the index pregnancy. Only 1 woman who did not smoke
in the 3 months before pregnancy or in the first trimester began smoking in
later pregnancy. None of the women who continued to smoke cigarettes in the
second trimester quit smoking in the third trimester. Therefore, in the analysis,
we defined "never smoker" as those women who did not smoke cigarettes during
any of the 4 time periods and used never smoker as the reference group. We
defined "ever smoker" as those who smoked any number of cigarettes during
any of the 4 time periods. We further divided ever smokers into 2 subgroups:
quitter, only smoked in the 3 months before pregnancy or during the first
trimester; and continuous smokers, smoked continuously from prepregnancy to
delivery. We are unable to adequately evaluate the timing of smoking in relation
to birth weight given the smoking pattern in our study sample. Maternal passive
smoke exposure was grouped into 2 categories based on maternal self-reporting:
unexposed or exposed to 1 or more smokers at home during the index pregnancy.
The detailed method for detection of the CYP1A1
MspI polymorphism can be found elsewhere.21
This method is able to detect all 3 possible genotypes for the polymorphism: AA (homozygous wild type), Aa
(heterozygous variant type), and aa (homozygous variant
type). In our preliminary analysis, we evaluated 4 possible genetic models:
dominant (AA = 0, Aa = 1, aa = 1), recessive (AA = 0, Aa = 0, aa = 1), additive (AA = 0, Aa = 1, aa = 2), and no restriction (no assumption made). We found that the
associations between maternal smoking and infant birth weight differed considerably
between the AA and Aa genotype
groups (mean, −234 g; SE, 99 g vs −577 g; SE, 126 g) but were
similar for the Aa and aa
genotype groups (−577 g; SE, 126 g vs −508 g; SE, 238 g). Thus,
our data did not suggest a recessive model. We combined the Aa and aa genotypes in the data analysis due
to the small number of enrolled mothers with the aa
The detailed method on detection of the GSTT1
deletion polymorphism can be found elsewhere.20
This method is only able to detect the present (at least 1 allele present, AA or Aa) or absent (complete
deletion of both alleles, aa) genotype.
Infant birth weight was evaluated as both a continuous and a binary
(<2500 g vs ≥2500 g) variable. Gestational age was assessed in 2 ways:
time since the first day of the last menstrual period and an algorithm based
on last menstrual period and the result of early ultrasound (<20 weeks'
gestation). This approach has been used in a large hospital-based preterm
study.25 Briefly, the last menstrual period
estimate was used only if confirmed by an ultrasound within 7 days or if no
ultrasound estimate was obtained; otherwise, the ultrasound estimate was used.
Gestational age was analyzed both as a continuous and a binary (<37 vs ≥37
weeks' gestation) variable. Since the results were similar for gestational
age based on last menstrual period vs the algorithm, we present the results
based on the latter approach. We used birth weight ratio (observed birth weight/mean
birth weight for gestational age) as a continuous measure of fetal growth
and defined intrauterine growth restriction as birth weight ratio less than
85%, an approach used in a previous study.26
We used multiple linear and logistic regression models to estimate the
individual and combined associations of maternal cigarette smoking and CYP1A1 and GSTT1 genotypes in
relation to infant birth weight, gestation, and fetal growth with adjustment
of major covariates. We first examined the association between maternal cigarette
smoking and infant birth weight without consideration of maternal genotypes.
Then we investigated whether the association between maternal cigarette smoking
and birth weight was modified by maternal genotypes by estimating the association
between maternal cigarette smoking and birth weight in maternal genotype groups
of each gene, respectively. Furthermore, we examined the combined association
of maternal cigarette smoking and maternal genotypes with birth weight in
8 subgroups. These subgroups were defined by maternal smoking status during
pregnancy (never vs continuous; the quitters were excluded from the analysis
due to small sample size) and by maternal genotype for CYP1A1 (AA, Aa/aa) and GSTT1 (present, absent).
Gene–cigarette smoke interaction was also tested by adding a product
term to the regression model. Similar analysis was applied to gestational
age and fetal growth. Finally, to address potential confounding by population
stratification, we performed the analysis stratified by maternal ethnicity.
All the analyses were adjusted for the following potential confounders:
maternal ethnicity (white, black, Hispanic, other), age (<20, 20-24, 25-29,
and ≥30 years), education (≤middle school, = high school, and >high
school), parity (0, 1, and ≥2), marital status (married, other), prepregnant
weight and height in both linear and quadratic terms, passive smoking (no,
yes), maternal self-reported alcohol use (nonusers, current users), and infant
sex. All P values were 2-sided and defined as P = .05 for statistical significance. We used statistical
software SAS (SAS Institute Inc, Cary, NC) for all analyses.
Our analysis included a total of 741 mothers: 567 never smokers and
174 ever smokers. A total of 207 cases were preterm or LBW infants (125 were
both preterm and LBW, 34 were LBW only, and 48 were preterm only) and 534
were non–LBW, full-term infants (control). As shown in Table 1, the never- and ever-smoking groups were similar in CYP1A1 and GSTT1 genotype frequencies,
age distribution, maternal prepregnancy weight and height, and infant sex.
However, the 2 groups differed in ethnicity, education, parity, marital status,
passive smoke exposure, and alcohol use. For the ever smokers, the mean birth
weight was 280 g lower (95% confidence interval [CI], −413 to −147)
and the odds ratio (OR) for LBW was higher (OR, 1.8; 95% CI, 1.3-2.7) compared
with the never smokers. The mean gestational age for ever smokers was 0.8
weeks shorter (95% CI, −1.3 to −0.2) and the OR of preterm birth
was higher (OR, 1.8; 95% CI, 1.3-2.7).
As shown in Table 2, without
consideration of genotype, continuous maternal smoking during pregnancy was
associated with an OR of 2.1 (95% CI, 1.2-3.7) for LBW and a mean reduction
of 377 g (SE, 89 g) in birth weight compared with the never smokers. When CYP1A1 genotype was considered, the association between
continuous maternal smoking and LBW differed remarkably by the genotype: the
OR for LBW was 1.3 (95% CI, 0.6-2.6) among mothers with the AA genotype (n = 75) but 3.2 (95% CI, 1.6-6.4) among mothers with the Aa/aa genotypes (n = 43 for Aa and n = 6 for aa). A similar
pattern emerged when GSTT1 genotype was considered:
the OR was 1.7 (95% CI, 0.9-3.2) and 3.5 (95% CI, 1.5-8.3) for the present
and absent genotypes, respectively. Consistently, when birth weight was analyzed
as a continuous variable, continuous maternal smoking was associated with
a mean reduction of 252 g (SE, 111 g) vs 520 g (SE, 124 g) in birth weight
for the CYP1A1 AA and Aa/aa genotypes, respectively; and a mean reduction of 285
g (SE, 99 g) vs 642 g (SE, 154 g) in birth weight for GSTT1 present and absent genotypes, respectively.
We found a similar pattern for gestational age (Table 3). Without consideration of genotype, continuous maternal
smoking was associated with an OR of 1.8 (95% CI, 1.1-3.1) for preterm birth
and a 1.0 week (SE, 0.4-week) shortening in gestation. When CYP1A1 genotype was considered, the OR was 1.5 (95% CI, 0.8-2.8) and
2.2 (95% CI, 1.1-4.4) for the AA and Aa/aa genotypes, respectively. When GSTT1 genotype was considered, the OR was 1.4 (95% CI,
0.8-2.6) and 2.8 (95% CI, 1.2-6.7) for the present and absent genotypes, respectively.
When gestational age was analyzed as a continuous variable, there were reductions
in mean gestational age of 0.6 (SE, 0.5) and 1.5 (SE, 0.5) weeks for the CYP1A1 AA and Aa/aa genotypes, respectively; and reductions of 0.5 (SE, 0.4) and 2.1
(SE, 0.7) week for GSTT1 present and absent genotypes,
We also examined the association of continuous maternal smoking and
maternal genotype with birth weight ratio (Table 4). When CYP1A1 genotype was considered,
the OR of intrauterine growth restriction was 1.9 (95% CI, 0.9-4.1) and 4.1
(95% CI, 2.0-8.6) for continuous smokers with AA
and Aa/aa genotypes, respectively.
The pattern was similar when birth weight ratio was analyzed as a continuous
variable. However, this pattern was not found with stratification by GSTT1 genotype.
Table 5 presents the combined
association of maternal cigarette smoking and CYP1A1
and GSTT1 genotypes with infant birth weight, gestational
age, and birth weight ratio. There was a common pattern for the 3 outcomes.
Among nonsmoking mothers, genotype alone did not confer a significant adverse
effect. In the presence of maternal smoking, the greatest reduction in mean
birth weight (−1285 g; SE, 234 g), gestational age (−5.2 weeks;
SE, 1.0 week), and birth weight ratio (−0.120; SE, 0.048) was found
among the group with the CYP1A1 Aa/aa and GSTT1 absent genotypes. A test of interaction
between maternal smoking and maternal CYP1A1 and GSTT1 genotypes was statistically significant for birth
weight and for gestational age, but not for birth weight ratio.
We performed separate analyses for blacks and whites, the 2 largest
ethnic subgroups in our sample with adequate numbers of smokers. The percentages
of CYP1A1 AA, Aa, and aa genotypes were 58.0%, 36.3%, and 5.8%, respectively,
for blacks; 73.8%, 24.6%, and 1.6%, respectively, for whites (χ2 test, P = .005). The percentage of GSTT1 absent genotype was 22.5% for blacks and 18.9% for whites (χ2 test, P = .39). As shown in Table 6 without consideration of maternal genotype, the association
between maternal smoking and infant birth weight was comparable for blacks
and whites. However, when the genotype was considered, the estimated smoking
effects were different between the 2 groups and gene-smoking interactions
were only statistically significant in blacks.
There were 38 mothers with either gestational diabetes or diabetes mellitus
in our study. Neither adjustment for diabetic status in the regression model
nor exclusion of diabetic mothers from the analysis altered our results. Our
data did not show significant differences in pregnancy complications (preeclampsia,
eclampsia, chronic hypertension, diabetes, abruptio placentae, placenta previa,
incompetent cervix, oligohydramnios, polyhydramnios, meconium in amniotic
fluid) nor differences in method of delivery (vaginal vs cesarean) between
never and ever smokers. Further adjustment of pregnancy complications and
type of delivery in the regression analyses did not alter our results.
It has long been recognized that many human diseases arise from the
complex interplay of environmental exposures and host susceptibilities. Our
study represents the first step in investigating how genetic susceptibility
modulates risk of adverse reproductive outcomes from environmental exposures
such as cigarette smoke. Consistent with previous studies, we found that maternal
cigarette smoking was associated with reduced birth weight and an increased
risk of LBW,3- 8
shortened gestation and an increased risk of preterm birth,8,27- 29
and intrauterine growth restriction.3,9,10
Our data indicate that maternal cigarette smoking likely affects infant birth
weight via both reduced fetal growth and shortened gestation. More importantly,
our study shows consistent evidence that the adverse effects of maternal cigarette
smoking on infant birth weight and gestational age were modified by maternal CYP1A1 and GSTT1 genotypes. Our
data demonstrate that a subgroup of pregnant women with certain genotypes
appeared to be particularly susceptible to the adverse effect of cigarette
smoke, suggesting an interaction between metabolic genes and cigarette smoking.
Although there are few published data on genetic susceptibility to cigarette
smoke in relation to birth weight or gestation, this susceptibility is biologically
plausible. Both CYP1A1 and GSTT1 genes are highly polymorphic in our study population. These gene polymorphisms
have been associated with their encoded enzyme activity; CYP1A1 MspI variant genotypes may increase enzyme activity,30 while the deletion type of GSTT1 leads to an absence of enzyme activity.31
There is evidence that increased CYP1A1 enzyme activity
associated with MspI variant genotype or absence of GSTT1 enzyme activity associated with deletion genotype can be detrimental
to pregnancy outcomes in the presence of cigarette smoke exposure.
Major classes of carcinogens present in cigarette smoke are converted
into DNA-reactive metabolites by cytochrome P450–related enzymes and
some cytochrome P450 variants have been associated with increased risk of
various cancers.12 On the other hand, the GSTT1 enzyme is important in protecting against certain
genotoxic damages, such as sister chromatid exchanges32,33
and the formation of hemoglobin adducts due to ethylene oxide present in tobacco
smoke.34 Everson et al35
tested human placental specimens for DNA adducts and found that DNA adducts
were almost exclusively present in those specimens from mothers who were smokers.
Positive dose-response relationships were shown between levels of the smoking-related
adducts and biochemical doses of maternal tobacco smoke exposure during pregnancy.
Alexandrov et al36 further demonstrated that
the levels of benzo(a)pyrene diol-epoxide–DNA adducts and bulky DNA
adducts were significantly and positively correlated with CYP1A1 enzyme activity. A similar finding was demonstrated in another
independent study.37 Furthermore, Perera et
al38 found that newborns with elevated levels
of PAH-DNA adducts had significantly decreased birth weight (P = .05), birth length (P = .02), and head
circumference (P<.001) compared with newborns
with lower adducts (n = 135). Consistently, our study found that smoking mothers
who had CYP1A1 MspI variant genotypes or GSTT1 deletion genotype had lower birth weight and birth weight ratio
and shorter gestational age compared with the reference groups. Furthermore,
our study found that smoking mothers who had both CYP1A1 MspI variant genotype and GSTT1 deletion
genotype had the greatest reduction in birth weight, gestation, and birth
A number of methodological limitations should be considered when interpreting
our results. Maternal smoking was based on self-report and thus may be subject
to reporting bias. Nevertheless, studies have shown fair agreement between
self-reported smoking amount and serum or urinary level of cotinine (a biochemical
marker of cigarette smoke).5,39,40
Our results are consistent with the vast body of literature that demonstrates
a detrimental effect of cigarette smoke on the fetus. Maternal genotypes were
objective measurements and neither the mothers nor the research staffs were
aware of maternal genotypes at the time of interview and medical record review.
Second, smoking mothers differed from never smokers in terms of ethnicity,
education, parity, marital status, passive smoke exposure, and alcohol use.
In the regression analyses, we adjusted for these variables. However, we cannot
exclude the possibility of confounding effects by uncontrolled or inadequately
controlled risk factors. For example, no attempt was made to assess nutritional
status. There could be comorbidity between alcohol or illicit drug and tobacco
use. Exclusion of alcohol or illicit drug users from the analysis did not
significantly alter the results.
Third, cigarette smoke is a complex mixture of chemicals3
and other metabolic genes may be involved. This study only examined CYP1A1 and GSTT1 genotypes. The
relative role of metabolic genes vs other genes in determining genetic susceptibility
to adverse reproductive outcomes of cigarette smoking is yet to be understood.
Furthermore, there is a possibility of unrecognized linkage disequilibrium
between the candidate marker and another gene that is the real susceptibility
Population stratification is a potential issue in genetically heterogeneous
populations like that of the United States. This is an inherent weakness of
a case-control study design. A family-based association study, such as transmission/disequilibrium
testing, is more desirable to address this issue. Assessing the confounding
of interactions is an evolving area of epidemiology. Factors that may not
be confounders in a regular analysis may still change the estimate of the
effect of the gene polymorphism–smoking interaction. In addition, we
only examined maternal genotypes, and the role of fetal genotypes in modifying
the adverse effect of cigarette smoke and maternal-fetal gene interaction
remains to be determined.
The rapid advances in the Human Genome Project, bioinformatics, and
biotechnology have provided unprecedented opportunities as well as challenges
in understanding the genetic basis for individual differences in susceptibility
to environmental exposures.41 As discussed
in a recent commentary,42 much work remains
to be done and many methodological challenges remain to be addressed in this
research area. A coherent gene-environment approach, with attention to genetically
susceptible populations who are disproportionately exposed to environmental
reproductive hazards, may provide further insights into the etiology of intrauterine
growth restriction and preterm birth and may help identify high-risk subpopulations
for clinical or public health interventions.