Meanwhile, the adverse impact of periconceptional alcohol consumption on the growth and mental development of the fetus—collectively referred to as fetal alcohol syndrome—is well established (Jones et al., 1973; Hanson et al., 1976, 1978; Riley et al., 2003). Yet the association of alcohol consumption with major congenital anomalies, especially those not characteristic of fetal alcohol syndrome, remains a matter of debate. Some epidemiological evidence indicates that the relationship between alcohol use during pregnancy and congenital malformations varies by the amount of alcohol intake. One population-based case-control study showed that even a low level of alcohol consumption during pregnancy was significantly associated with an elevated risk of orofacial clefts in the offspring (Munger et al., 1996). Other studies have concluded instead that there was no significant association between low levels of maternal alcohol intake and increased risk of oral clefts (Werler et al., 1991; Shaw and Lammer, 1999; Meyer et al., 2003). Two of these studies (Werler et al., 1991; Shaw and Lammer, 1999), however, reported an increased probability of cleft lip in the offspring of women who engaged in binge drinking—that is, consumed five or more drinks per drinking occasion—during pregnancy. Likewise, there is considerable uncertainty regarding the role of alcohol intake during pregnancy in the development of conotruncal heart defects. At least three studies have shown that maternal alcohol consumption results in a moderately elevated risk of such defects (Shaw et al., 1992; Tikkanen and Heinonen, 1992; Carmichael et al., 2003), whereas two other studies found no association (Adams et al., 1989; Ferencz et al., 1997). The few epidemiologic studies that have evaluated the potential association between alcohol intake during early pregnancy and NTDs have concluded that such consumption did not increase the risk of defects in infants (Mills and Graugard, 1987; McDonald et al., 1992; Shaw et al., 1996a; Suarez et al., 2007).

Mothers of cases and controls were eligible to be interviewed if they were the biological mother and carried the pregnancy of the selected study subject, were not incarcerated, and if their primary language was English or Spanish. Overall, 1,355 mothers (80% of eligibles) in the case group and 700 mothers (77% of eligibles) in the control group were interviewed. Data were unavailable from 11% of mothers of eligible cases and 12% of mothers of controls who could not be located, and an additional 9% of mothers of cases and 11% of mothers of controls who refused to participate. Interviews were conducted primarily over the telephone in English or Spanish, no earlier than 6 weeks after the EDD. The median time between EDD and interview completion was 10 months for cases and 8 months for controls.

A standardized, computer-based questionnaire was used to elicit detailed information on smoking and alcohol consumption, as well as potential covariates, during the 4 month periconceptional time period, defined as the 2 months before through the 2 months after conception. To assess maternal smoking and alcohol consumption, women were asked to report the average number of cigarettes they smoked per day, how often they consumed alcohol, and the number of drinks they consumed per drinking day, during each month of the periconceptional period. To assess binge drinking, women were asked to estimate the number of days in each of the 4 months that they consumed five or more drinks on one occasion. Maternal smoking, defined as the average number of cigarettes smoked per day, and alcohol intake were analyzed separately for each of the 4 months of the periconceptional period. We examined two indicators of alcohol consumption: (1) the number of drinking days per week and (2) the number of drinks per drinking day. The reference group for the smoking analyses was comprised of women who indicated no smoking during the entire periconceptional period, whereas the reference group for the alcohol consumption analyses consisted of women who indicated no smoking or alcohol consumption during these 4 months.

At the start of the periconceptional time period, between 9 and 13% of the mothers in the case groups were classified as smokers, as compared to 11% of the mothers in the control group (Table 1). These percentages dropped steadily over the 4 month study period: smoking during the second month of pregnancy was reported by just 5% of the mothers of the control group and between 1 and 4% of the mothers in the case groups. The proportions that reported alcohol consumption dropped from 27–28% at 2 months prior to pregnancy to 6–7% during the second month of pregnancy. Binge drinking also declined: the percentage of mothers who reported consuming five drinks or more at one occasion fell from a range of 3–9% at the start of the periconceptional period to around 1% at the end.

Table 2 summarizes the unadjusted risk estimates associated with maternal smoking—classified as none, ≤5 cigarettes per day, and >5 cigarettes per day—during the first month of pregnancy. Among women who reported smoking five cigarettes or less per day, elevated risks were not observed for most defects, except for modest, but imprecise (as indicated by the wide confidence bands), increased risks for TOF (OR 1.6; 95% CI: 0.8, 3.2), multiple CP (OR 1.5; 95%CI: 0.4, 5.2), and multiple CLP (OR 1.4; 95% CI: 0.4, 4.8). By contrast, results for smoking five cigarettes or less per day and NTDs suggested an inverse risk, both when examined as a single group (OR 0.7; 95% CI: 0.3, 1.4) and when divided into the phenotypes anencephaly (OR 0.4; 95% CI: 0.1, 1.5) and spina bifida (OR 0.8; 95% CI: 0.3, 1.9). Similar results were observed among women who smoked more than five cigarettes per day for conotruncal heart defects as a group (OR 0.5; 95% CI: 0.2, 1.2) and for the phenotypes dTGA (OR 0.4; 95% CI: 0.1, 1.7) and TOF (OR 0.5; 95% CI: 0.2, 1.7). All of the 95% confidence bands associated with the ORs contained one and were wide, underlining the imprecision of the risk estimates. The results for the other 3 months of the periconceptional period were generally similar to those for the first month of the pregnancy (data not shown). The data were too sparse to simultaneously adjust for all the covariates in multivariable models; instead, single variable control was undertaken. Adjustment of the risk estimates for covariates did not yield results that were sufficiently different to change the interpretation associated with the crude risk estimates reported in Table 2 (data not shown).

The unadjusted analysis showed that women who consumed alcohol less than once a week during the first month of their pregnancy had a 1.5-fold greater risk (95% CI: 1.0, 2.2), relative to nondrinkers, of delivering infants with conotruncal heart defects (Table 3). This increase was attributed largely to the phenotypic subgroup dTGA, for which nearly a twofold elevated risk was observed (OR 1.9; 95% CI: 1.1, 3.2). Similarly, the risk of delivering infants with isolated CP was somewhat elevated (OR 1.4; 95% CI: 0.8, 2.5), while the risk of offspring with multiple CLP was 1.9-fold higher (95% CI: 0.8, 4.5). Drinking more frequently—that is, consuming alcohol at least once a week—was associated with elevated risk only for multiple CLP (OR 2.6; 95% CI: 1.1, 6.1). This elevated, albeit imprecise, risk of multiple CLP was also observed with respect to binge drinking, defined as consuming five or more drinks per drinking occasion, during the first month of pregnancy (OR 2.3; 95% CI: 0.7, 7.0). The data were too sparse for binge drinking to simultaneously adjust for all the covariates in multivariable models; instead, single variable control was undertaken. Controlling for the individual covariates did not produce risk estimates that were substantially different from the crude estimates reported in Table 3—all of the adjusted risk estimates changed by less than 12%.

After adjustment for covariates, the risk estimates associated with alcohol consumption during the periconceptional period were similar to the crude estimates reported above, with the exception of NTDs. Relative to non-drinkers, women who consumed alcohol less than once a week during the first month of pregnancy had a 1.6-fold increased risk of delivering infants with NTDs (95% CI: 0.9, 2.6), as compared to an unadjusted OR of 1.1, while women who consumed alcohol once a week or more had a 2.1-fold increased risk (95% CI: 1.1, 4.0), as compared to an unadjusted OR of 1.1. The results for the anencephaly and spina bifida phenotypes showed elevated risks that were similar to those reported for all NTD cases combined (data not shown).

The decreased risks of NTDs and conotruncal heart defects associated with maternal smoking during the periconceptional period could be indicative of true risks or could be artifactual explained by the modestly elevated risk (i.e., relative risks of less than 1.5) of spontaneous abortion (Risch et al., 1988; Armstrong et al., 1992) among mothers who smoked during pregnancy. Complete ascertainment of NTDs among spontaneous abortions is essentially not possible. Thus, we can not disentangle whether the reduced risk is real or artifact. However, our finding of a reduced risk of NTDs in offspring associated with maternal smoking during the periconceptional period is consistent with the results of Shaw et al. (1996a), who reported risk estimates of 0.9 and 0.7 for mothers who smoked 1–19 cigarettes per day and ≥20 cigarettes per day, respectively. Both of these sets of results are contrary to those in earlier studies that have either reported an increase in the risk of NTDs (Kelsey et al., 1978; Evans et al., 1979; Suarez et al., 2007) or have showed no association (McDonald et al., 1992; Wasserman et al., 1996). Our finding of a reduced risk of conotruncal heart defects associated with periconceptional smoking of more than five cigarettes per day is inconsistent with two previous studies that reported no association (Adams et al., 1989; Wasserman et al., 1996), and two others that reported ORs of 1.3 for mothers who smoked 20 cigarettes or less per day (Kelsey et al., 1978) and 1.9 for mothers who smoked just 1–9 cigarettes per day (Shaw et al., 1992). As these latter two studies examined conotruncal heart defects as a group and did not report results for the specific phenotypes, they do not offer a basis for comparison of our finding that maternal smoking was associated with a reduced risk of dTGA and TOF. Finally, our finding that maternal smoking during the periconceptional period does not play a significant role the etiology of orofacial clefts, with the possible exception of isolated CP, is supported by three studies (Evans et al., 1979; Werler et al., 1990; Lieff et al., 1999), yet contradicted by several others (Kelsey et al., 1978; Khoury et al., 1987, 1989; Shaw et al., 1996b; Chung et al., 2000; Honein et al., 2007). A number of our estimates of the risk of defects associated with maternal smoking are based on sparse data; therefore, the observed results might alternatively be attributable to random variation.

Our findings of elevated risk associated with periconceptional alcohol consumption are bolstered by experimental research that has suggested tenable mechanisms between alcohol and abnormal development (Pullarkat, 1991; Shean and Duester, 1993). Indeed, several studies have examined the link between maternal alcohol consumption and the risk for conotruncal heart defects in offspring, with conflicting findings. A population-based case-control study of California births between 1987 and 1988 noted that the risk of conotruncal heart defects in offspring was moderately elevated among women who consumed alcoholic beverages during the periconceptional period (Carmichael et al., 2003). These results agree with two earlier studies (Shaw et al., 1992; Tikkanen and Heinonen, 1992), but are contradicted by two others that found no association (Adams et al., 1989; Ferencz et al., 1997). Our study findings are consistent with the former research: we observed a risk of conotruncal heart defects—particularly dTGA—that was nearly twofold higher among infants whose mothers reported alcohol consumption during the periconceptional period.

The potential association between alcohol intake during early pregnancy and NTDs was first described in a clinical study 25 years ago (Freidman, 1982). Our results, however, are consistent with several subsequent epidemiologic studies (Mills and Graubard, 1987; McDonald et al., 1992; Shaw et al., 1996a), each of which found no significant increase in NTD risk associated with maternal alcohol use during early pregnancy. Similarly, our finding that maternal alcohol consumption during the periconceptional period was not associated with an elevated risk of most of the cleft phenotypes is generally consistent with previous studies (Werler et al., 1991; Shaw and Lammer, 1999). We did observe an increase in the risk of multiple CLP, unlike Munger et al. (1996), who found an effect only for isolated CLP. Our findings also suggested that binge drinking is associated with a higher risk of multiple CLP, but not other cleft phenotypes, which is in contrast to the findings of Werler et al. (1991) and Shaw and Lammer (1999). With the exception of isolated CLP (where no association was observed), however, our study lacked power to accurately estimate risks of clefts associated with binge drinking.

Any similarities and differences in results should be considered with caution given variation across studies in the definition of the periconceptional period, the methods for quantifying smoking and alcohol consumption, and the time period over which data were collected (and as a result, the interval between exposure and interview). For example, given the low prevalence of smoking in our study population, we were unable to replicate the broad smoking categories that were employed in previous studies (Shaw et al., 1992; Wasserman et al., 1996; Suarez et al., 2007; Honein et al., 2007), thus making direct comparison difficult. Similarly, we cannot easily compare our findings on alcohol consumption with previous studies, as we considered both the frequency (number of drinking days per week) and amount of alcohol consumption (number of drinks per drinking day), whereas several of the previous studies used a more limited classification of alcohol intake, for example, simply distinguishing between those who consumed alcohol and those who did not.

Another important consideration is the decline in the prevalence of smoking and alcohol consumption in the US during the past few decades: the share of the adult population in the US that smoked dropped by 38% between 1979 and 2004 (NCHS, 2006), while the share that consumed alcohol dropped by 13% between 1978 and 2003 (NIAAA, 2004). Among pregnant mothers, the decline in smoking prevalence was even steeper: 48% between 1989 (the first year for which data were reported) and 2004 (NCHS, 2006). According to data from two cross-sectional National Health and Nutrition Examination Surveys conducted in 1988–1994 and 1999–2002, the average number of cigarettes smoked per day fell by nearly 15% during this time period (O’Connor et al., 2006). The sales of high-tar and -nicotine delivery cigarettes have also dropped significantly since 1978, whereas sales of medium- and low-delivery cigarettes increased over the same time frame (NIAAA, 2004). Although the overall prevalence of smoking and alcohol consumption has declined, our results—that maternal alcohol consumption during the periconceptional period continues to increase risk of NTDs, dTGA, and multiple CLP, whereas maternal smoking decreases risk for NTDs and conotruncal heart defects—are consistent with earlier studies that were conducted when these exposures were more common.

A further issue to bear in mind is the potential impact of reporting bias. The behaviors that mothers of cases report could be influenced by increased public awareness of the potential teratogenic effects of smoking and alcohol consumption during pregnancy. Mothers may be inclined to underestimate their exposures, denying a possible causal role. On the other hand, they may overestimate their exposures, believing there is a causal role. Previous studies have suggested that for many chronic exposures, reporting bias is likely to be minimal in studies of congenital anomalies (Werler et al., 1989; Swan et al., 1992; Khoury et al., 1994). To help gauge whether such a bias exists in our study, we examined data from the National Birth Defects Prevention Study (NBDPS), an ongoing, multistate, case-control study of environmental and genetic risk factors for major birth defects. According to NBDPS data from 1999–2004, the percentage of control mothers in California who smoked ranged from 12% 2 months before pregnancy to 8% 2 months after conception—versus 11 and 5%, respectively, in our study. The prevalence of alcohol intake among the NBDPS control mothers ranged from 22% 2 months before pregnancy to 8% 2 months after conception—versus 27 and 6%, respectively, in our study. The comparison, therefore, reveals only modest differences in the prevalence of these exposures and offers no specific evidence that the mothers in our study may have over- or under-reported smoking and alcohol consumption.

Finally, we also cannot rule out residual confounding as a possible explanation for our findings. Although all of the analyses were ultimately adjusted for several covariates known to be associated with both exposures and the selected birth defects, it is possible that the omission of additional unmeasured confounders contributed to our results.