Maternal smoking and the risk for clubfoot in infants^†

INTRODUCTION

Clubfoot is one of the most common major birth defects, with a prevalence of approximately 1 per 1,000 live births (Wynne-Davies, 1964; Ching et al., 1969; Moorthi et al., 2005). Infants born with clubfoot typically have forefoot adduction, hindfoot inversion, and hindfoot equinus (Ponseti, 1996). Their foot bones, ankle bones, foot muscles, and ligaments may all be abnormal, and their heel cords are often tight, preventing the development of a normal gait. These malformations require medical treatment and often surgery for correction.

The etiology of clubfoot is not well understood but numerous causes have been proposed. Intrauterine compression, vascular insufficiency, breech position, abnormal myogenesis, intrauterine infection, and neurologic abnormalities are several of the mechanistic theories suggested (Kite, 1964; Wynne-Davies, 1972; Hootnick et al., 1994; Martin et al., 1994; Mellerowicz et al., 1994). Genetic factors are also thought to play a role (Wynne-Davies, 1972; de Andrade et al., 1998; Lochmiller et al., 1998; Honein et al., 2000; Dietz, 2002). The concordance rate of clubfoot among monozygotic twins is 33%, in contrast to a concordance rate of about 3% among dizygotic twins (Wynne-Davies, 1972). Family studies show a higher recurrence of clubfoot among first-degree relatives than second-degree, or more distant, relatives (de Andrade et al., 1998; Lochmiller et al., 1998).

Racial and ethnic groups may have differing birth prevalences as well, with 0.39 cases of clubfoot per 1,000 livebirths among Chinese populations (Chung et al., 1969), 0.68 per 1,000 livebirths in Black populations (Moorthi et al., 2005), 0.76 per 1,000 livebirths in Hispanic populations (Moorthi et al., 2005), and 0.73–1.12 per 1,000 live births in White populations (Chung et al., 1969; Moorthi et al., 2005). In contrast, a recent study that examined the prevalence among racial and ethnic groups in Texas found very little difference among White, Black, and Hispanic (both foreign-born and US-born) populations (Moorthi et al., 2005).

Various risk factors have been examined in relation to clubfoot, although few consistent patterns have emerged. Clubfoot affects males about twice as often as females (Alberman, 1965; Cartlidge, 1984). Alderman et al. (1991) found no relationship between clubfoot and maternal age, parity, or season of birth. Similarly, Honein et al. (2000) found that maternal age and education were not significant risk factors. However, a case-control study by Skelly et al. (2002) found that maternal education level was higher among the cases than the controls.

Several studies have examined maternal smoking as a risk factor for clubfoot, and these findings have also been somewhat mixed. Van den Eeden et al. (1990) performed a case-control study of maternal smoking and various congenital malformations, including 171 cases of clubfoot. That study showed a weak, but positive, association between smoking and clubfoot (relative risk = 1.4; 95% CI: 1.0, 2.0). Alderman et al. (1991) conducted a case-control study investigating risk indicators for clubfoot among 175 cases and 1,470 controls. That analysis found that the association between maternal smoking and clubfoot was modified by infant sex, with an OR of 2.8 (95% CI: 1.6, 5.1) for male infants and 1.8 (95% CI: 0.8, 4.4) for female infants. In another case-control study of 239 cases of clubfoot and 365 controls, Skelly et al. (2002) found that case mothers were more likely to have smoked during pregnancy than control mothers. Moreover, smoking showed a dose-response relationship, with an OR of 1.5 for case mothers who smoked lightly, up to an OR of 3.9 for women who smoked 20 or more cigarettes per day. In a study of 346 infants with isolated clubfoot and 3,029 infants with no birth defects, Honein et al. (2000) found that having a family history of a foot anomaly was a strong effect modifier for clubfoot risk. The researchers found that while maternal smoking in the absence of a positive family history was weakly associated with clubfoot (OR 1.34; 95% CI: 1.04, 1.72), the combination of family history and smoking produced an OR of 20.30 (95% CI: 7.90, 52.17). Using national natality data files for 1997–1998, Honein et al. (2001) analyzed maternal smoking and clubfoot as reported on the birth certificate. That study included 3,894 cases of clubfoot, as reported on the birth certificate checkbox. Controlling for maternal age, education, and race/ethnicity, the authors found a prevalence ratio of 1.62 (95% CI: 1.49, 1.75) for clubfoot among women reporting any smoking during pregnancy. The authors in that study also noted that the association between smoking and clubfoot was stronger for male infants than for females, and that there was evidence of a positive dose-response relationship between the number of cigarettes smoked and the prevalence ratio.

However, not all studies reported a positive association between maternal smoking and clubfoot. In a study of 980 infants with clubfoot, Malloy et al. (1989) found a null OR of 1.02 for maternal smoking. McDonald et al. (1992) studied 396 nonsmokers and 218 smokers, analyzing risk according to the number of cigarettes smoked per day during the first 3 months of pregnancy. They found no association with clubfoot among any of the smoking strata, nor was there evidence of a dose-effect relationship (OR 1.06; 95% CI: 0.8, 1.4; OR 1.0; 95% CI: 0.8, 1.3; and OR 0.98; 95% CI: 0.8, 1.3, for infants of women who smoked 1–9, 10–19, and 20 or more cigarettes per day, respectively). Shiono et al. (1986) studied the relationship between maternal smoking and congenital malformations in a prospective cohort study of 86,946 livebirths and found an OR of 0.7 (95% CI: 0.6, 0.9) for clubfoot.

Part of the explanation for the inconsistent findings of a relationship between smoking and clubfoot may relate to the different case definitions used in the various studies. For example, some studies included infants with other malformations, while others considered only isolated clubfoot. Isolated and nonisolated cases may have distinctly different etiologies. There is also considerable variation in the diagnostic coding criteria used to define clubfoot among the different studies, with some studies using a relatively broad case definition and others using much more restrictive criteria. These differences are important to consider when comparing the different studies because they can result in widely varying results. The purpose of the present study is to further explore the link between maternal smoking during pregnancy and clubfoot in infants.

METHODS

This population-based case-control study used data from the North Carolina Birth Defects Monitoring Program (NCBDMP) and the North Carolina Composite Linked Birth File maintained by the North Carolina State Center for Health Statistics. The composite birth file consists of all North Carolina resident birth certificates linked to maternal and infant Medicaid paid claims and health department service data, such as the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) and maternity care coordination (Buescher et al., 1991). The NCBDMP is a statewide, population-based surveillance system, which uses active surveillance to identify infants born with birth defects. The program covers a surveillance population of approximately 120,000 births annually.

Cases of clubfoot were ascertained for the years 1999–2003 by the NCBDMP and consisted of all liveborn, singleton infants diagnosed with either isolated talipes equinovarus (TEV) or isolated clubfoot not otherwise specified (NOS) (British Pediatric Association (BPA) codes 754.500 and 754.730, respectively). Nonisolated cases, which were excluded from the analysis, were those with a major congenital anomaly in another organ system, such as neural tube defects (NTDs), or other major musculoskeletal defect not related to clubfoot, for example, limb reduction defects. Cases were not excluded if they had other minor malformations, such as postaxial polydactyly, overlapping toes, or similar conditions. Of the 704 cases initially identified, 17 (2.4%) were excluded because they were nonsingletons and 243 (34.5%) were excluded because they were nonisolated cases. One case infant was excluded because of missing information on maternal smoking, resulting in a total of 443 cases eligible for the study. The final case group included 173 (39.1%) infants diagnosed with TEV and 270 (60.9%) infants with clubfoot, NOS. The case definition criteria and percentage of initial cases meeting the inclusion criteria in this study are similar to those reported in previous research (Honein et al., 2000; Moorthi et al., 2005).

Controls were selected from a simple random sample of liveborn, singleton infants identified from North Carolina birth certificates during the years 1999–2003. Potential controls were verified against the NCBDMP clinical database to exclude infants with major birth defects. A total of 4,500 controls were initially ascertained. After excluding eight controls because of missing smoking data, there were a total of 4,492 included in the study.

Maternal smoking status during pregnancy was ascertained from the birth certificate, and was grouped into the following categories for the purpose of this study: no smoking, 1–5, 6–10, and >10 cigarettes per day. Subjects reporting no smoking versus any smoking were also compared. In order to assess the reliability of smoking data from the North Carolina birth certificate, maternal smoking as reported on the North Carolina Pregnancy Risk Assessment Monitoring System (NCPRAMS) survey for the years 1999–2003 was compared to smoking as indicated on the corresponding matched birth certificate for each NCPRAMS survey respondent. NCPRAMS, which is an ongoing survey of a representative, weighted random sample of North Carolina women who have recently given birth, asks women about a variety of behaviors and experiences before, during, and after delivery, including smoking. The level of agreement in maternal smoking (any smoking vs. none) as reported from the two data sources was examined using the kappa coefficient (Fleiss, 1981), and was based on the weighted NCPRAMS sample results. The kappa coefficient measures the level of agreement in a measurement or observation obtained from two independent data sources, while correcting for the probability of agreement by pure chance alone.

Descriptive analyses of cases and controls were conducted on the following variables: maternal age, maternal race/ethnicity, marital status, education, number of previous live births (parity), maternal Medicaid and WIC status, timing of prenatal care initiation, infant's sex, gestational age, birthweight, and infant death. Frequencies of all variables were calculated by case-control status, and bivariate analyses were carried out on selected variables to identify potential confounders. Variables that were statistically significantly associated with case-control status in bivariate analyses were entered into a logistic regression model, and backward elimination was used to develop a final model containing a set of predictors for clubfoot. Additionally, two variables, infant sex and maternal race/ethnicity, were examined as possible effect modifiers using cross-product terms in a logistic regression model. These variables were chosen as potential effect modifiers based on previous studies that found that family history and sex of the infant modified the association between smoking and club foot (Alderman et al., 1991; Honein et al., 2000, 2001). Because data on family history were not available for the subjects in the present study, race/ethnicity was examined as a possible effect modifier in its place.

The association between maternal smoking and clubfoot was examined using multiple logistic regression. Prevalence ORs and 95% CIs were determined, controlling for potential confounders identified in the bivariate analyses. Analyses were conducted using SPSS, version 13.0 for Windows and with SAS version 9.1 (SAS Institute, 2003). The study protocol was reviewed and approved by the University of North Carolina-Chapel Hill School of Public Health Institutional Review Board prior to analysis.

RESULTS

A comparison of the correlation in maternal smoking as reported on the birth certificate and the NCPRAMS survey indicated good agreement between the two sources (kappa statistic = .77) (Table 1). Overall, about 95% of the birth certificates were in agreement with the NCPRAMS survey results. Examining the discordant cells in Table 1 also reveals little evidence of systematic error in reporting, although the birth certificate tended to underreport smoking somewhat, as compared to NCPRAMS; 2.3% of birth certificates indicated maternal smoking when the NCPRAMS survey did not, and 2.8% of birth certificates indicated no smoking when NCPRAMS respondents reported that they did smoke. Note that the percentages shown in Table 1 are based on the weighted data, and do not correspond to the actual number of respondents shown in each cell.

There was no evidence of significant interaction with infant sex or maternal race/ethnicity, although there was a stronger effect for females compared to males (OR 1.83; 95% CI: 1.23, 2.73; and OR 1.34; 95% CI: 0.96, 1.88, for females and males, respectively). Compared to controls, mothers of case infants were more likely to be younger, non-Hispanic White, nulliparous, enrolled in WIC or Medicaid, and their infants were more likely to be male and to be born moderately preterm (33–36 weeks gestation) (Table 2). Approximately 18% of the case mothers reported any smoking during pregnancy, compared to about 13% of the controls (unadjusted OR 1.49; 95% CI: 1.15, 1.92). There was no evidence of a dose-response relationship between the number of cigarettes smoked and clubfoot risk; in fact, the OR was highest among women in the lowest smoking group (1–5 cigarettes per day). Controlling for maternal age, race/ethnicity, timing of prenatal care initiation, and sex of the infant, the adjusted OR associated with any maternal smoking and clubfoot was 1.40 (95% CI: 1.07, 1.83) (Table 3). The adjusted OR among women who smoked 1–5 cigarettes per day compared to nonsmokers was 1.79 (95% CI: 1.17, 2.74). Smoking 6–10 or >10 cigarettes per day was weakly associated with clubfoot risk (adjusted OR 1.26 and 1.27, respectively), and the lower bound of the 95% CI included the null value for each.

After adjustment for other risk factors, male sex remained positively associated with clubfoot risk (adjusted OR 1.88; 95% CI: 1.53, 2.32), and Black, non-Hispanic mothers were 33% less likely to have had an infant with clubfoot than White, non-Hispanic mothers (OR 0.66; 95% CI: 0.51, 0.87). Mothers who identified their race/ethnicity as “other” likewise were about one-third less likely to have an infant with clubfoot than White, non-Hispanic mothers, but the CI was wide and included the null value (OR 0.66; 95% CI: 0.35, 1.23). The difference between Hispanic mothers and White, non-Hispanic mothers was small and not statistically significant. Late initiation of prenatal care was inversely associated with the risk of the infant having clubfoot (OR 0.71; 95% CI: 0.52, 0.97). There was also a trend toward increasing clubfoot risk with decreasing maternal age, with women in the youngest age group (<20) having the highest risk (OR 1.58; 95% CI: 1.12, 2.22) relative to women of ages 30 or more.

In order to determine whether the risk associated with maternal smoking varied by type of clubfoot diagnosis, separate logistic regression models were run for infants with TEV (n = 173) and infants with clubfoot NOS (n = 270). Because of the sparse data in the higher exposure levels, smoking was treated as a dichotomous variable in these analyses. Among infants with TEV, the adjusted OR was 1.29 (95% CI: 0.84, 1.97). The adjusted OR among infants with clubfoot NOS was slightly higher (1.47, 95% CI: 1.06, 2.05) but not appreciably different from that of the group with TEV.

DISCUSSION

The results of this study support the hypothesis that maternal smoking is associated with isolated clubfoot in infants. Controlling for potential confounding factors, smoking during pregnancy was associated with a modestly increased risk for clubfoot among offspring. These results are similar to previous research on this topic, thus lending increased credence to studies of the link between maternal smoking and clubfoot. Unlike the findings of Honein et al. (2001) and Skelly et al. (2002), the present report found no evidence of a dose-response relationship between the number of cigarettes smoked and clubfoot risk. This may be due, in part, to the fact that there were very few subjects who fell into the higher exposure categories. For example, in the Honein et al. (2001) study, the prevalence ratio was substantially elevated only among women reporting >20 cigarettes per day; in the present study fewer than 1% of the subjects reported smoking more than 20 cigarettes per day, so it was not possible to assess the risk at this highest level of exposure. It is also possible that heavy smoking is more likely to cause other major birth defects in infants with clubfoot. Because this study excluded nonisolated clubfoot cases, this could have diminished evidence of a dose-response effect. The Honein et al. (2001) study was not limited to isolated cases.

Also, in contrast to previous reports by Alderman et al. (1991) and Honein et al. (2000, 2001), the present study found the association between smoking and clubfoot to be somewhat stronger among female infants than among males. This pattern held when the cases were stratified by the type of diagnosis (TEV or clubfoot NOS). Future studies are necessary to explore these sex differences in greater detail.

There are at least two possible biologic pathways through which maternal smoking can affect the developing fetus: (1) fetal hypoxia leading to vascular disruption, and (2) arrested foot development as a result of chemicals present in cigarettes. Fetal hypoxia may be a result of functional and structural changes in the placental tissue due to maternal smoking (Rama Sastry, 1991; Demir et al., 1994). It has been observed that intervillous blood flow decreases 15–20% after a pregnant woman inhales cigarette smoke, and this could cause vascular disruption, thus leading to clubfoot (Demir et al., 1994). Additionally, if a fetus is hypoxic, oxygen uptake by the hind limbs may decrease in order to supply oxygen to the brain and other vital organs (Boyle et al., 1992). Another plausible biologic mechanism between cigarette smoke and clubfoot is that of a disruption of foot development. Typically developing fetuses' feet go through positional changes, and at one point assume a varus position that later takes on a normal foot position (Victoria-Diaz and Victoria-Diaz, 1984). It is possible that nicotine and carbon monoxide could retard or arrest typical foot development because of their antimitotic or antimetabolic effects (Werler et al., 1985; Boyle et al., 1992; Johnston and Bronksky, 1995).

This study had some limitations. One limitation is the use of birth certificate data for the ascertainment of maternal smoking. Birth certificate data are accurate for many variables, but there has been question of the validity of some data such as smoking. Piper et al. (1993) examined the consistency of maternal smoking as reported on the birth certificate compared to medical records in Tennessee, and found the sensitivity of the birth certificate to be 73.5%. In a similar study in North Carolina, Buescher et al. (1993) reported that the percent agreement in maternal smoking between the birth certificate and medical records was 84.4. In a multistate study of birth certificate reporting of maternal smoking, Dietz et al. (1998) calculated that the sensitivity of the birth certificates ranged from about 71–82%. DiGiuseppe et al. (2002) conducted a multihospital study of the reliability of several birth certificate variables, including smoking, as compared to medical records. That study found a sensitivity of 72.2%, and a 91.9% concordance overall between the two data sources. In the present study, 1,031 of 1,284 NCPRAMS respondents who reported smoking during pregnancy on the self-administered maternal questionnaire also reported smoking on the birth certificate (sensitivity 80.3%), and the overall agreement between the two data sources was approximately 95%. Although the two data sources are not entirely comparable in the way that smoking is measured, the results indicated reasonably good agreement, and were consistent with previous studies. However, most studies suggest that the birth certificate does tend to underreport smoking to some degree, which would lead to misclassification of mothers who are smokers as nonsmokers. Such underreporting would likely be nondifferential, and would tend to bias the results toward the null.

Another limitation of this study is that it was not possible to look at family history as a risk factor in the analysis. Family history is one of the few consistently identified risk factors for clubfoot (Wynne-Davies, 1964, 1972; de Andrade et al., 1998; Lochmiller et al., 1998; Honein et al., 2000). Unfortunately, the data sources that were used for this study do not include this information.

This study has several strengths. First, it was a population-based study and was therefore less likely to be affected by selection bias. The cases were ascertained by the NCBDMP and the controls were selected using simple random sampling of natality data, both during the study years of 1999–2003. In addition, the sample size was large enough to allow us to identify a modest level of risk with reasonable precision. Although the study did not have a sufficient sample size to permit a detailed analysis of TEV and clubfoot NOS separately, the adjusted OR for smoking in relation to each of these diagnoses was similar, and in fact, was slightly higher for clubfoot NOS. This finding suggests that combining the BPA codes for TEV and clubfoot NOS is a reasonable approach to examining the relationship between smoking and, possibly, other exposures in relation to clubfoot.

Another strength of this study is that the case and control parameters used are comparable to other studies that have produced similar results. The cases were eligible to be included if they were diagnosed with TEV or clubfoot not otherwise specified, but were ineligible if any other major malformation or chromosomal abnormality was found. Studies that have used less refined case definitions have had more mixed results, making inferences about their findings problematic. Clubfoot in conjunction with other major birth defects is thought to have etiologies of a different nature than clubfoot alone or clubfoot with minor anomalies, and cases of this kind are commonly excluded for this reason (Alderman et al., 1991; Honein et al., 2000; Skelly et al., 2002; Moorthi et al., 2005).

This study shows that birth defects data, when linked with vital statistics and health services information, can be a useful, convenient, and informative approach to exploring risk factors for birth defects. The findings of this study are consistent with the hypothesis that maternal smoking during pregnancy confers an increased risk of an infant being born with clubfoot. The fact that even low levels of exposure (1–5 cigarettes per day) were associated with an increased risk is an important finding from a prevention perspective, as it reinforces the public health education message that merely cutting back on smoking during pregnancy is not sufficient for preventing adverse effects on the fetus. More research is necessary to elucidate the biologic mechanisms by which maternal smoking affects a developing fetus, and the combined role that genetics and environment play in this process. This study, in conjunction with the results of previous investigations, underscores the need to educate women of childbearing age of the array of adverse outcomes that may affect the infant whose mother smokes during pregnancy. These include birth defects as well as low birth weight and sudden infant death syndrome.