Maternal smoking and congenital heart defects

Karin KaÈ lle nTornblad Institute, University of Lund, Lund, Sweden

Accepted in revised form 17 July 1999

Abstract. The Swedish Child Cardiology Registry(CCR) and the Swedish Medical Birth Registry(MBR) were used to investigate a possible associationbetween maternal smoking during pregnancy andcongenital heart defects. Among 1,413,811 infantsborn in 1983±1996 with known smoking exposure inearly pregnancy, 3384 infants with congenital heartdefects were selected (458 term infants with persistentductus arteriosus (PDA) identi®ed from MBR orCCR, and 2926 infants with other heart defects,identi®ed from CCR). After controlling for year ofbirth, maternal age, parity, and educational level, aweak, statistically signi®cant association between allheart defects and maternal smoking was found (oddsratio (OR): 1.09; 95% con®dence interval (CI): 1.01±

1.19). When infants with isolated PDA were removedfrom the case group the signi®cance disappeared(OR: 1.07; 95% CI: 0.98±1.17). For truncus abnor-malities (OR: 1.23; 95% CI: 1.02±1.49), atrial septaldefects (OR: 1.63; 95% CI: 1.04±2.57), and PDA(OR: 1.30; 95% CI: 1.05±1.62), a rather strong e�ectof maternal smoking was indicated. The increasedOR for PDA remained when gestational durationand intrauterine growth was also controlled for.Further research based on independent data sets isneeded to conclude whether the risk estimates formaternal smoking are true and truly di�er betweenthe groups. The classi®cation system used (with 24classes) is described in enough detail to permit arepetition of the study.

Key words: Classi®cation, Congenital heart defects, Malformation, Registry, Smoking

Abbreviations: CCR = Swedish Child Cardiology Registry; DORV = double outlet right ventricle;HLHS = hypoplastic left heart syndrome; ISC = International Society of Cardiology; MBR = SwedishMedical Birth Registry; PDA = persistent ductus arteriosus

Introduction

Congenital heart defects are among the most frequentmajor malformations but the etiology of most heartmalformations is still unknown. The overall terato-genic e�ect of tobacco smoke is uncertain, but forsome speci®c malformations a positive associationwith maternal smoking has been reported, and the®nding has been repeated by independent investiga-tors (oral clefts [1±5], limb reduction defects [6±8],and kidney malformations [9, 10]). Lilja [11] found apersuasive association between single umbilical arteryand maternal smoking.

Fedrick et al. [12] reported a highly signi®cant riskfor heart defects among infants of smoking mothers.Due to small numbers, the authors could not ascer-tain whether any speci®c types of heart defects wereparticularly associated with maternal smoking, butmentioned that it might be true for persistent ductusarteriosus (PDA) and tetralogy of Fallot. It is notclear if preterm infants with PDA were included inthe case group. Since the report by Fedrick et al.,several negative studies regarding maternal smokingand heart defects with none or only a crude (O3)

division into subgroups have been published [13±17].With the exception of the study by Malloy et al. [16],these studies were all comparatively small with lowpower to detect a moderate risk increase. Not manyinvestigators have broken down the group of heartmalformations to more homogeneous subgroups inorder to calculate adequate risk estimates for mater-nal smoking. Shiono et al. [18] computed odds ratios(ORs) for maternal smoking among several groups ofheart defects, most of them far too small to have anypotential power to detect a true association (e.g. therisk estimate for transposition of the great arterieswas based on eight cases). In 1991, Tikkanen andHeinonen [19] could neither ®nd any association be-tween maternal smoking and all heart defects, nordetect signi®cant di�erences in maternal smokingbetween the six groups of heart malformations theyinvestigated (ORs were not shown). They also in-cluded PDA (if still patent after the neonatal period)and again it is unclear whether preterm infants wereincluded. In 1992, Pradat [20] found no associationbetween heart defects and maternal smoking (infantswith isolated PDA and a birth weight below 2500 gwere excluded). Two years later, using an extended

European Journal of Epidemiology 15: 731±737, 1999.Ó 1999 Kluwer Academic Publishers. Printed in the Netherlands.

study period (1983±1990), some associations betweenmaternal smoking and certain speci®c heart anoma-lies appeared [21]. In 1996, Wasserman et al. [22]chose to especially study the putative associationbetween maternal smoking and conotruncal heartdefects. By combining the dose groups, the OR formaternal smoking could be estimated to 1.32 (95%con®dence interval (CI): 0.91±1.93).

By linking the Swedish Child Cardiology Registry(CCR) with the Swedish Medical Birth Registry(MBR) (the same sources as used by Pradat [20, 21],but with an extended study period) a comparativelylarge case group with prospectively collected infor-mation on smoking in early pregnancy could be ob-tained. The large case group made it possible todivide it into more homogenous subgroups.

Materials and methods

Infants with congenital heart defects were mainlyidenti®ed from the CCR [21, 23] which contains de-tailed information on infants who have been investi-gated at one of the four child cardiology departmentsin Sweden. All heart defects are coded with a slightlymodi®ed three-digit ISC (International Society ofCardiology) code [24]. Only malformations diag-nosed by echocardiography, cardiac catheterization,at operation, or at autopsy and detected before theage of one are included. PDA in infants younger than3 months is not recorded in the Child CardiologyRegister. Each case is identi®ed with the personalidenti®cation numbers of the mother and the infant.This makes it possible to link this registry with theMBR so that, e.g., information on smoking habitscan be obtained. For a special analysis of PDA, terminfants identi®ed from the MBR with a diagnosis ofisolated PDA but not recorded in the CCR wereadded.

The MBR contains medical information on nearlyall deliveries in Sweden (coverage about 99%) [25].All diagnoses are given as ICD codes. Nearly allpregnant women avail themselves of the free ante-natal service. Information from antenatal medicalrecords is available for 99.1% of the births registeredin the MBR. Among the women who used the ante-natal service, about 93% attended during the ®rst 14weeks of pregnancy. At the ®rst visit (usually duringweek 10±12), each woman is interviewed by a midwifeand, among other things, the smoking habit of thewoman is stated as none, <10 cigarettes per day, orP10 cigarettes per day. Standardized record formsare used at all antenatal clinics, all delivery units, andat all pediatric examinations of newborn infants.Copies of these forms are sent to the National Boardof Health where they are computerized. Registry in-formation on smoking habits is available from 1983.Since then, smoking information is available for 93%of all women giving birth in Sweden.

Expected birth weight was estimated using a pari-ty-speci®c birth weight for gestational age standard[26].

In order to collect information on maternal edu-cational level, the MBR was linked with the Registryof Education kept by Statistics Sweden. The latterregistry contains information on the educational level(nine years of compulsory school included) of eachwoman and the educational level on 1 January 1996was linked from it. Thus, the educational level used inthis study to estimate socio-economic status does notnecessarily refer to the actual educational length ofeach women at the time she is giving birth. This madeit possible to estimate socio-economic status also foryoung women. For women giving birth during 1996,no information on educational level could beobtained.

E�orts were made to make an embryologic classi-®cation of the malformations and the heart defectswere divided into 24 subgroups as speci®ed in theResults section. The details of the classi®cation areshown in Appendix A.

Infants with a known chromosome abnormality(n = 560) or with unknown smoking exposure inearly pregnancy were excluded from the study. Thereference group contains all births.

All odds ratios (OR) were calculated using Mantel-Haenszel's technique. Strati®cation was made foryear of birth, maternal age (®ve year classes), parity(previously born infants +1 (1, 2, 3, 4+)), and ed-ucational level (unknown, <10 years, 10±12 years,13±14 years, and P15 years). Other strati®cations aredescribed in the Results section. 95% CI were esti-mated using Miettinen's method [27]. When com-paring two strati®ed ORs, two-tailed z-tests werecarried out, using the same variance as used to esti-mate the 95% CI. Tests of homogeneity of the ORsacross strata were based on weighted sums of thesquared deviations of the stratum speci®c log-oddsratios from their weighted means.

Results

The number of cases by smoking habits and site andadjusted ORs for maternal smoking are shown inTable 1. A weak association of borderline statisticalsigni®cance is indicated between all heart malforma-tions registered in the CCR. Overall, the estimatedORs for isolated or associated (with non-cardiacmalformation) heart defects do not substantiallydi�er. For truncus anomalies (statistically signi®-cant), atrial septal defects (statistically signi®cant)and complex heart malformations (borderline signif-icance), a moderate association with maternalsmoking is suggested. No signi®cant negative asso-ciation between maternal smoking and any of thestudied groups could be detected, but decreased ORfor maternal smoking is seen for ventricular septal

732

defects with atrial septal defects, endocardial cushiondefects, pulmonary or aortic valve anomalies, co-arctation of the aorta, and other ventricular anoma-lies, respectively. The indicated di�erence between theOR for maternal smoking for atrial septal defects andthe corresponding OR for ventricular septal defects isinteresting but not statistically signi®cant ( p = 0.08).The di�erence between the OR for maternal smokingfor truncus anomalies and the corresponding OR foraortic or pulmonary valve anomalies is statisticallysigni®cant ( p = 0.02). A rather strong, statisticallysigni®cant, association with maternal smoking isobserved for PDA in term infants. The ORs formaternal smoking among infants with PDA reported

from the CCR or the MBR, respectively, are of thesame magnitude.

When infants with PDA are included in the casegroup, a signi®cant association between heart con-ditions and maternal smoking is detected. There is nosign of any heterogeneity of the risk estimates overyear of birth, maternal age, parity, or educationallevel strata.

A homogeneity test was performed in order tostudy whether the risk estimates for maternalsmoking calculated for each of the 24 subgroupscould all be stochastic variations of the commonOR of 1.09. The estimated p value for homogeneitywas 0.28.

Table 1. Number of cases by site and smoking habits. ORs for smoking among mothers of cases vs. all births. Strati®ed foryear of birth, maternal age, parity, and educational level

Smokers Non-smokers OR (95% CI)

Double outlet right ventricle 26 84 0.90 (0.57±1.42)

Asplenia-polysplenia-situs inversuswith any heart malf. 7 22 0.96 (0.40±2.27)

Transposition of the great arteries 92 215 1.32 (1.02±1.71)

Tetralogy of Fallot 36 107 1.08 (0.73±1.61)Other truncus anomalies (including e.g. pulmonaryatresia, aortic atresia, common truncus, etc.) 34 80 1.19 (0.79±1.79)

All truncus anomalies 162 403 1.23 (1.02±1.49)Hypoplastic left heart syndrome, unspeci®ed 26 67 1.20 (0.75±1.92)Tricuspidal valve atresia with pulmonary atresia 9 14 1.69 (0.74±3.89)

Atrial septal defects 30 70 1.63 (1.04±2.57)Ventricular septal defects 77 234 1.02 (0.79±1.33)Septal defects with pulmonary, aortic, mitral,

or tricuspidal valve stenosis 29 80 1.22 (0.76±1.94)Single ventricle 7 10 1.94 (0.66±5.64)Atrial with ventricular septal defects 22 84 0.92 (0.56±1.52)

All septal defects 165 478 1.14 (0.94±1.37)Endocardial cushion defects 18 70 0.67 (0.38±1.15)Pulmonary or aortic valve anomalies 73 279 0.82 (0.63±1.07)

Tricuspidal valve anomalies 8 22 1.07 (0.49±2.36)Mitral valve anomalies 1 11Coarctation of the aorta 67 226 0.90 (0.68±1.20)

Other anomalies of arch and branches(aorta or pulmonary artery) 28 77 1.06 (0.67±1.68)

Anomalous pulmonary vein return 21 45 1.29 (0.75±2.22)

Other ventricular malformations 12 42 0.74 (0.39±1.43)Malformations of systemic arteries and veins 5 12 0.99 (0.36±2.70)Other heart malformations 24 52 1.34 (0.82±2.19)Complex heart malformations 103 263 1.24 (0.98±1.59)

Persistent ductus arteriosus (full term only)Registered in CCR 30 65 1.39 (0.88±2.19)Registered in MBR only 93 270 1.28 (1.00±1.64)

All PDA (full term) 123 335 1.30 (1.05±1.62)

All heart malformations registered in CCR

(PDA excluded)Isolated 638 1836 1.07 (0.97±1.18)Associated 118 334 1.06 (0.85±1.31)

Total 756 2170 1.07 (0.98±1.17)

All heart conditions (PDA included) 879 2505 1.09 (1.01±1.19)

Reference group 347,513 1,066,298

733

It is well known that infants born in the pretermperiod often have PDA. Figure 1 shows that the riskof PDA varies with gestational duration even in thefull term period. Infants born at 39 or 40 completedweeks are at the lowest risk, while infants born at 37±38 weeks of pregnancy or born after 42±43 completedweeks are at an increased risk of PDA compared toother term infants.

As shown in Figure 2, growth retardation seems tobe a risk factor for PDA in term infants. IncreasedORs for PDA is also seen among infants who weighmore than expected with respect to gestational du-ration, parity, and infant sex.

A series of analyses was performed in order tostudy whether the increased OR for maternal smok-ing among infants with PDA re¯ects a true risk in-crease, or is due to confounding by the well-knownassociation between gestational duration, growth re-tardation, and maternal smoking on one hand, andthe association between gestational duration, growthretardation, and PDA on the other (as shown inFigures 1 and 2). Table 2 shows the ORs for maternalsmoking among infants with PDA after adjustmentfor gestational duration and intrauterine growth, oneat a time and simultaneously. After maximum strat-i®cation, the OR for maternal smoking among in-fants with PDA remained signi®cant. No positivedose response e�ect was indicated, however.

Discussion

The group of heart malformations is large andprobably heterogeneous from an etiological pointof view. Thus, it was crucial to break up the groupof heart malformations into more homogenous

subgroups to investigate whether tobacco smokeseems to have a stronger e�ect on one group than onanother. However, when evaluating the results of thepresent study it is important to bear in mind thatwhen simultaneously studying a large number ofclasses, one or two risk estimates may reach signi®-cance due to mass-signi®cance.

The present study suggests the existence of a slightrisk increase for heart defects in infants of motherswho smoked during pregnancy. For some groups ofcongenital heart defects, especially truncus abnor-malities, atrial septal defects, and PDA, a ratherstrong e�ect of maternal smoking was indicated.

All the truncus anomalies result from abnormalaorto-pulmonary septation, a disturbance which isthought to be caused by abnormal neural crestmigration. If maternal smoking has an e�ect onneural crest migration, this could have implicationsfor many other malformations as neural crest cellsgive rise to diverse structures of the head and neck.Further research is needed in order to investigatewhether defects of other structures to which neuralcrest cells contribute are associated with maternalsmoking.

A heterogeneity of the e�ect of smoking on theincidence of atrial septal defects, ventricular septaldefects, or atrial with ventricular septal defects wasindicated but not statistically signi®cant. For allseptal defects, no signi®cant association with mater-nal smoking was found. The increased OR formaternal smoking among atrial septal defects mayre¯ect a true, speci®c risk increase, but the possibilityof the result being a random ®nding could not beneglected.

As only a small proportion of infants with PDAare registered in the CCR, cases with PDA were alsoobtained from the MBR in order to increase the

Figure 1. Gestational duration in full term infants as riskfactor for PDA. OR for PDA among infants born aftercompleted weeks of pregnancy as speci®ed compared to

infants born in any other week in the full-term period.Strati®ed for year of birth, maternal age, parity, educa-tional level, and smoking. 95% CI as vertical bars.

Figure 2. Growth retardation in full term infants as riskfactor for PDA. OR for PDA among infants weighing acertain percentage of the expected weight compared to in-

fants in all other percentage classes. The expected birthweight for each infant is calculated with respect to parity,gestational duration, and infant sex. Strati®ed for year of

birth, maternal age, parity, educational level, and smoking.95% CI as vertical bars. Only infants born at full term.

734

power of the investigation of a putative associationwith maternal smoking. The diagnoses of the casesobtained from the MBR are generally likely to be ofpoorer quality compared to those obtained from theCCR. However, the risk estimates for maternalsmoking were similar in the two case groups.

One complication in the investigation of the asso-ciation between PDA and maternal smoking is al-ready mentioned in the result section: the putativeconfounding by gestational age and intrauterinegrowth, both associated with PDA as well as withmaternal smoking. The results from the present studysuggest, however, that infants of smoking mothersare at an increased risk of PDA, irrespective of ges-tational duration and intrauterine growth. The con-striction of the ductus arteriosus at birth is inducedby an increase in local oxygen tension and normallyoccurs within 10 to 15 hours after birth in term in-fants. Several mechanisms behind the putative e�ectof maternal smoking on PDA could be speculated on.The PDA in infants of smoking mothers could forinstance result from a disturbance in the oxygenationof the infant at birth, be the result of a (compared togestational age and intrauterine growth) putativeimmaturity of infants of smoking mothers, a putativeabnormal hormonal balance among smoking women(during fetal life the ductus is thought to be keptpatent by circulating prostaglandins), or could be dueto some primary defect of the ductus arteriosus. If thesuggested association between maternal smoking andPDA is due to circumstances around birth, then thesmoking information used in this study (the end of®rst trimester) does not refer to the relevant period,and a bias toward unity may result.

Potential confounding factors are untested.Maternal diabetes, epilepsy, rubella infections, andsevere alcohol abuse are all generally accepted ascardiac teratogens. Among these risk factors, alcoholabuse is the only one which is known to be associatedwith maternal smoking and thus must be regarded asa putative confounder when investigating the associ-

ation between maternal smoking and heart defects.Reliable information on the percentage of pregnantwomen with an alcohol problem in Sweden is lacking,but the percentage is probably too low to seriouslyconfound the results of the present study.

The smoking information used in the present studyis self-reported and could obviously di�er from theactual smoking status of the pregnant women.Maternal exposure to passive smoking is anotherpossible source of misclassi®cation of the exposure.However, as the smoking information used in thepresent study is unbiased by the outcome of thepregnancy, a putative misclassi®cation would bias therisk estimates towards unity and could not explainthe positive results of the present study.

As mentioned in the Introduction, most studiesinvestigating the association between maternalsmoking and all heart malformations were too smallto have a potential power to detect a risk increase ofthe magnitude suggested in the present study.Compared to the OR for maternal smoking for allheart defects reported by the only positive previousstudy (Fedrick et al. [12]), the corresponding ORobtained in the present study is substantially lower.The authors of the previous study did not provideany risk estimate, but by using the raw numbersgiven, an OR of 1.53 (1.14±2.04) is obtained. Theirstudy was based on 204 cases of heart malforma-tions. As mentioned in the Introduction it is unclearwhether infants with PDA only born in the pretermperiod were excluded. If infants with PDA born inthe preterm period were included in the case groupin the present study, the overall OR for maternalsmoking among all heart defects increased to 1.16(1.08±1.24), an estimate still far below the resultsobtained from the study of Fedrick et al. It shall benoted that the case group in that study includedmainly stillbirths and neonatal deaths while thereference group comprised all births. The compara-tively high rate of smokers among mothers of in-fants in the case group could thus re¯ect the joined

Table 2. Maternal smoking as risk factor for PDA. ORs with 95% CI after strati®cations as speci®ed. (Only full terminfants.)

Strati®ed for Any smoking Smoking <10 Smoking P10

OR 95% CI OR 95% CI OR 95% CI

Year of birth, maternal age, parity,and educational level 1.30 (1.05±1.62) 1.37 (1.07±1.76) 1.19 (0.85±1.67)

Year of birth, maternal age,parity, educational level, andgestational duration 1.28 (1.02±1.59) 1.33 (1.03±1.71) 1.19 (0.85±1.67)

Year of birth, maternal age, parity,educational level, andintrauterine growth 1.27 (1.02±1.59) 1.28 (0.99±1.67) 1.18 (0.84±1.66)

Year of birth, maternal age,

parity, educational level, intrauterinegrowth, and gestational duration 1.30 (1.02±1.65) 1.34 (1.01±1.78) 1.25 (0.86±1.80)

735

e�ects of maternal smoking on infant death andcongenital heart malformations.

Using the same source as used in the present study,but covering only the period 1983±1990, Pradat [21]calculated the OR for maternal smoking among allheart defects to 1.07 (NS). For septal defects, a sta-tistically signi®cant association with maternal smok-ing was found (OR: 1.32; 95% CI: 1.02±1.71) whereasthe OR for truncus anomalies was unity. Thus, theresults of Pradat may seem contradictory to thoseobtained in the present study where a signi®cant as-sociation with maternal smoking was seen amongtruncus anomalies and atrial septal defects, but notwith ventricular septal defects or all septal defects.One explanation to the discrepancy between thestudies is that with fewer cases, Pradat [21] was forcedto make a cruder grouping of the heart defects thanthat made in the present study. The group of truncusmalformations included conotruncus septationanomalies, aorto or pulmonar valve anomalies, andDouble Outlet Right Ventricle (DORV). In thepresent study, the results from these groups werecalculated separately, and a heterogeneity of the riskestimates for maternal smoking on the incidence ofthese heart defects was indicated. The pathogenesis ofDORV is not well understood, but it is most likelycaused by some error in the re-alignment of the heartduring the 5th and 6th weeks, putting each ventriclein relation to the out¯ow tract. The reason why theassociation between septal defects and maternalsmoking found by Pradat [21] could not be repeatedis not obvious, but could be due to di�erent principleswhen among multiple heart defects ascertain theprimary heart diagnoses.

However, when a homogeneity test was performed,it could not reject the hypothesis that the individualrisk estimates for the 24 groups shown in Table 1were all stochastic variations of the common OR of1.09. The obvious risk of the ®ndings being due tomass-signi®cance is already mentioned, and furtherresearch based on independent data sets is needed toconclude whether the risk estimate for maternalsmoking is true and truly di�ers between the groups.In the present study, e�orts have been made to des-cribe the inclusion criteria, and especially important,the criteria for reduction of multiple heart defects tounderlying malformations, detailed enough to eval-uate whether the results of the present study could berepeated. If the OR for maternal smoking among thegroups are truly di�erent, this fact would be inter-esting not only when evaluating the teratogenic e�ectof tobacco smoke, but also for studies on thepathogenesis of di�erent heart defects. If the riskestimates for maternal smoking persist, but no spe-ci®city could be proven, this does not need to speakagainst maternal smoking as being a true teratogenfor heart defects. Any putative primary defectresulting from maternal smoking may result inmalformations of several regions of the heart.

Appendix A. Classi®cation system

E�orts have been made to make a classi®cation in

which the postulated primary defect ± not necessarily theclinically most signi®cant one ± is considered. Thus, theclassi®cation used considerably di�ers from the most com-

mon hierarchical model [16, 21]. The present classi®cationhas similarities with that used by Pradat [14, 15] but ismore detailed. Below is a list of the 24 groups with

comments.1 Double outlet right ventricle.2 Asplenia-polysplenia-situs inversus with any heart

malformation.

3±5 Defects resulting from a disturbance in the aorto-pulmonary septation:

3 Transposition of the great arteries.

4 Tetralogy of Fallot.5 Other truncus anomalies (e.g. Common truncus).

6 Unspeci®ed hypoplastic left heart syndrome (HLHS).

HLHS secondary to aortic atresia is included in the`other truncus anomalies' group.

7 Tricuspidal valve atresia with pulmonary atresia. A

condition parallel to mitral atresia with aortic atresialeading to HLHS.

8 Atrial septal defects.9 Ventricular septal defects.

10 Septal defects with pulmonary, aortic, mitral, or tri-cuspidal valve stenosis. As the signi®cance of the de-fects could not be obtained from registry data, severe

valve stenoses leading to septal defects could not beseparated from those minor ones which were found atclose cardiac examinations due to conditions caused

by other heart defects (e.g. major septal defects). Thecombination septal defect±valve stenosis is kept as a(probably heterogenous) entity.

11 Single ventricle.12 Atrial with ventricular septal defects. The combination

is kept as an entity as no obvious hierarchical orderbetween the individual defects exists. The combination

could be di�cult to separate from endocardial cushiondefects.

13 Endocardial cushion defects.

14 Pulmonary or aortic valve anomalies.15 Tricuspidal valve anomalies.16 Mitral valve anomalies. HLHS secondary to mitral

atresia (without aortic atresia) included.17 Coarctation of the aorta.18 Other anomalies of aortic arch and branches (aorta or

pulmonary artery).

19 Anomalous pulmonary vein return.20 Other ventricular malformations.21 Malformations of systemic arteries and veins. Does

not include single umbilical artery.22 Other isolated heart malformations.23 Complex heart malformations. See below.

24 Persistent ductus arteriosus (PDA). Only in term in-fants (born after 37 or more completed weeks ofpregnancy).

Each infant was included in one group only. For caseswith more than one diagnosis, the following principles were

followed when deciding the main diagnosis:± Double outlet right ventricle was considered as the

main diagnosis even when combined with other heart

defects.

736

± Cardiac defects associated with asplenia, polysplenia, orsitus inversus were coded as an entity irrespective of

type(s) of heart defect(s) (Ivemark's syndrome).± Atrial septal defects or ventricular septal defects were

counted only if occurring alone, together, or combinedwith valvular stenoses.

± Valvular stenoses were counted if occurring alone orcombined with atrial or ventricular septal defects. If notalone, regurgitation of the pulmonary, aortic, mitral, or

tricuspidal valve was not counted.± Aorta hypoplasia was not counted if occurring together

with coarctation of the aorta or hypoplastic left heart

syndrome.± Fibroelastosis or cardiomyopathy was not counted if

combined with anything else than valvular regurgitation,

valvular stenosis, atrial septal defect, or ventricularseptal defect.

± PDA was counted only if no other heart defects werepresent (`isolated').

All combinations not mentioned above were classi®ed as

complex heart malformations.

References

1. Ericson A, KaÈ lle n B, Westerholm P. Cigarette smoking

as an etiologic factor in cleft lip and palate. Am JObstet Gynecol 1979; 135: 348±351.

2. Khoury MJ, Weinstein A, Panny S, Holtzman NA,

Lindsay PK, Farrel K, Eisenberg M. Maternal cigarettesmoking and oral clefts: A population-based study. AmJ Public Health 1987; 77: 623±625.

3. Khoury MJ, Gomez-Farias M, Mulinare J. Does ma-ternal cigarette smoking during pregnancy cause cleftlip and palate in o�spring? Am J Dis Child 1989; 143:333±337.

4. Keels MA. The role of cigarette smoking during preg-nancy in the etiology of cleft lip with or without cleftpalate. [Dissertation]. Univ. of North Carolina at

Chapel Hill: 1991.5. KaÈ lle n K. Maternal smoking and orofacial clefts. Cleft

Palate J 1997; 34: 11±16.

6. Aro T. Incidence, secular trends and risk indicators ofreduction limb defects. Health Services Research bythe Helsinki: National Board of Health in Finland,

1984.7. Czeizel A, Kodaj I, Lenz W. Smoking during preg-

nancy and congenital limb de®ciency. Br Med J 1994;308: 1473±1476.

8. KaÈ lle n K. Limb reduction malformations and maternalsmoking during pregnancy. Am J Public Health 1997;87: 29±32.

9. Li DK, Mueller BA, Hickok DE, Daling JR, FantelAG, Checkoway H, Weiss NS. Maternal smokingduring pregnancy and the risk of congenital urinary

tract anomalies. Am J Public Health 1996; 86:249±253.

10. KaÈ lle n K. Maternal smoking and urinary organ mal-formations. Int J Epidemiol 1997; 26: 571±574.

11. Lilja GMC. Single umbilical artery and maternalsmoking. Br Med J 1991; 302: 569±570.

12. Fedrick J, Alberman ED, Goldstein H. Possibleteratogenic e�ect of cigarette smoking. Nature 1971;

231: 529±530.13. Kelsey JL, Dwyer T, Holford TR. Maternal smoking

and congenital malformations: An epidemiologicalstudy. J Epidemiol Commun Health 1978; 32: 102±107.

14. Evans DR, Newcombe RG, Campbell H. Maternalsmoking habits and congenital malformations: A pop-ulation study. Br Med J 1979; 2: 171±173.

15. Christianson RE. The relationship between maternalsmoking and the incidence of congenital anomalies.Am J Epidemiol 1980; 112: 684±695.

16. MalloyMH,Kleinman JC, Bakewell JM, SchrammWF,Land GH: Maternal smoking during pregnancy: Noassociation with congenital malformations in Missouri

1980±1983. Am J Public Health 1989; 79: 1243±1246.17. Van den Eeden SK, Karagas MR, Daling JR, Vaughan

TL. A case±control study of maternal smoking andcongenital malformations. Paedtr Perinatal Epidemiol

1990; 4: 147±155.18. Shiono PH, Klebano� MA, Berendes HW. Congenital

malformations and maternal smoking during preg-

nancy. Teratology 1986; 34: 65±71.19. Tikkanen J, Heinonen OP. Maternal exposure to

chemical and physical factors during pregnancy and

cardiovascular malformations in the o�spring. Tera-tology 1991; 43: 591±600.

20. Pradat P. A case±control study of major congenitalheart defects in Sweden ± 1981±1986. Eur J Epidemiol

1992; 8: 789±796.21. Pradat P. Epidemiological characteristics of major

congenital heart defects [Dissertation]. Lund, Sweden:

University of Lund, 1994.22. Wasserman CR, Shaw GM, O Malley CD, Tolarova

MM, Lammer EJ. Parental cigarette smoking and risk

for congenital anomalies of the heart, neural tube, orlimb. Teratology 1996; 53: 261±267.

23. Carlgren LE, Ericson A, KaÈ lle n B. Monitoring of

congenital cardiac defects. Pediatr Cardiol 1987; 8:247±256.

24. International Society of Cardiology. Classi®cation ofheart disease in childhood. The Netherlands: VRB

O�setdrukkerij, Groningen, 1970.25. Cnattingius S, Ericson A, Gunnarskog J, KaÈ lle n B. A

quality study of a medical birth registry. Scand J Soc

Med 1990; 18: 143±148.26. KaÈ lle n B. A birth weight for gestational age standard

based on data in the Swedish Medical Birth Registry,

1985±1989. Eur J Epidemiol 1995; 11: 601±606.27. Miettinen OS. Simple interval estimation of risk ratio.

Am J Epidem. 1974; 100: 515±516.28. Fyler DC, Buckley LP, Hellenbrand WE, Cohn HE,

Kirklin JW, Nadas AS, Cartier JM, Breibart MH.Report of the New England regional infant cardiacprogram. Pediatrics 1980; 65[suppl]: 375±461.

Address for correspondence: Karin KaÈ lle n, Tornblad In-

stitute, Biskopsgatan 7, S-223 62 Lund, SwedenPhone: 46 (46) 222 7538; Fax: 46 (46) 222 4226E-mail: karin.kallen@anatom.lu.se

737