Craniosynostosis affects approximately 4/10,000 live births in the United States¹ and is characterized by premature fusion of one or more skull sutures. In normal development, the skull sutures begin to fuse around age 2 years and reach maturity in adulthood², allowing for the skull to expand to accommodate brain growth during childhood. With craniosynostosis, this fusion occurs earlier, which can result in dysmorphic skull and facial features, as well as restricted brain development.³ Craniosynostosis subtypes can be defined by the affected suture (sagittal, metopic, coronal, lambdoidal, and multiple sutures), with some variation in risk factors for each subtype. Risk factors consistently associated with craniosynostosis include infant sex (excess among males for all subtypes except coronal),^1,4–7 non-Hispanic White maternal race/ ethnicity,^1,5,8 plural gestation,^1,7,9,10 and older maternal or paternal age.^1,7,8,11 Valproic acid has been reported to be associated with increased risk of craniosynostosis,^12–14 but evidence for other modifiable risk factors, such as smoking^15–17 and fertility treatment,^14,18,19 is mixed.

Alcohol is a common exposure among women of reproductive age²⁰ and is known to be associated with other adverse pregnancy outcomes.²¹ Analysis of 1997–2005 data from the National Birth Defects Prevention Study (NBDPS) showed a null association between maternal periconceptional (one month before conception through the third gestational month) alcohol consumption and craniosynostosis among offspring, but an inverse association with later pregnancy alcohol consumption.²² Another small study also observed an inverse association between maternal alcohol consumption and sagittal craniosynostosis.²³ Both studies were based on maternal self-reported alcohol consumption and authors noted the possibility of exposure misclassification.

Given the relative paucity of literature on alcohol consumption and craniosynostosis, further exploration of this relationship is needed. Towards this end, we used the complete NBDPS dataset, incorporating an additional six years of participant data, to update and expand upon previous NBDPS analyses of maternal alcohol consumption during pregnancy and craniosynostosis. NBDPS is one of the largest population-based studies of birth defects. We capitalized on this robust sample to examine associations by craniosynostosis subtype and incorporate a formal analysis of bias due to potential alcohol exposure misclassification.

NBDPS was a multi-site, population-based case-control study designed to investigate risk factors for major structural birth defects. Methods have been described in detail elsewhere.²⁴ Briefly, NBDPS was implemented in 10 US states (Arkansas, California, Georgia, Iowa, Massachusetts, New Jersey, New York, North Carolina, Texas, and Utah) and enrolled people with pregnancies ending on or after October 1, 1997 with estimated dates of delivery (EDDs) through December 31, 2011. Mothers of infants with birth defects (cases) were identified via birth defects surveillance programs in each state; cases could be liveborn, stillborn at 20 weeks gestation or later, or prenatally diagnosed and terminated at any gestational age. Mothers of liveborn infants without birth defects (controls) were randomly selected from hospital records or birth certificates in the same geographic regions as cases. Overall, 65% of controls and 66% of craniosynostosis cases participated in the interview. Participants self-reported demographic and pregnancy exposure data via telephone interview at least six weeks and up to two years after their EDD. Mothers were eligible to participate in the interview if they could speak English or Spanish, were not incarcerated, had legal custodial rights to their child, and had not already participated in the NBDPS with a previous child. Mothers provided informed consent and institutional review boards at each site and the Centers for Disease Control and Prevention approved the study protocol.

We included mothers of 1,435 craniosynostosis cases and 11,216 controls in our main analysis. Case characteristics varied by affected suture, but in general, compared to controls, cases more frequently reported older age at delivery, non-Hispanic White race/ethnicity, family history of craniosynostosis, and fertility treatment (Table 1). We observed an excess of male infants among all craniosynostosis subtypes except coronal craniosynostosis, among which we observed an excess of female infants. Mothers of metopic and coronal cases more frequently reported obese BMIs, whereas mothers of lambdoidal cases reported disproportionately high rates of cigarette smoking during pregnancy.

Similar proportions of case and control mothers reported periconceptional alcohol consumption during pregnancy (Table 2). Overall, 36.9% (n=529) of case and 38.2% (n=4,280) of control mothers reported consumption, of whom about one-third of each reported at least one binge episode. In general, among those who reported alcohol consumption, patterns of frequency and quantity were similar between case and control mothers and most mothers reported consumption in only one (cases: 59.6%; controls: 56.0%) or two (cases: 29.5%; controls: 29.9%) gestational months. Further, most alcohol consumption was restricted to the periconceptional period, with 83.2% of case and 79.6% of control mothers reporting consumption only during the periconceptional period (Supplemental Table 2). Among those who reported consumption in the second or third trimester (n=92 cases, 924 controls), 75.0% of cases (n=69) and 70.7% of controls (n=653) reported consumption in the periconceptional period as well.

Estimated AORs for any periconceptional alcohol consumption and craniosynostosis in offspring, overall and by affected suture, ranged from 0.78–0.93, with all 95% CIs including the null (Table 3). Most AORs were slightly higher for associations between periconceptional binge drinking and craniosynostosis and its subtypes, but CIs were wider and included the null. Restricting to mothers who only reported alcohol consumption during the periconceptional period, estimates were slightly closer to the null but very similar to our main results (Supplemental Table 3). Our sub-analysis of later pregnancy consumption was based on 23 exposed cases, of whom only 1 mother reported binge drinking, prohibiting meaningful analysis by affected suture or quantity of consumption. Our estimate for the association between later pregnancy alcohol consumption and craniosynostosis was further from the null than our periconceptional estimate (AOR 0.67, 95% CI 0.43, 1.04) (data not shown). When stratified by time period, AORs for any periconceptional alcohol consumption and for periconceptional binge drinking were similar among mothers with 1997–2005 EDDs and 2006–2011 EDDs (Supplemental Table 4).

Our bias-adjusted estimates for periconceptional alcohol consumption and craniosynostosis tended to be slightly higher than our conventional estimates (Table 4) but remained close to null. When we assumed higher sensitivity among case mothers and non-differential specificity, the bias-adjusted estimates moved lower. Under the most plausible scenario—lower sensitivity among case mothers and non-differential specificity—our bias-adjusted estimates shifted upwards to 1.03 (0.80, 1.33) for any periconceptional alcohol consumption and to 1.13 (0.87, 1.46) for any periconceptional binge consumption.

In our study, estimates for associations between maternal periconceptional alcohol consumption and craniosynostosis in offspring were either near or below the null, with many estimates imprecise and all 95% CIs including the null. These results are consistent with those previously published using NBDPS data^22,42 and with one non-NBDPS study of alcohol and craniosynostosis that we identified.²³

We encourage caution in interpreting the inverse associations we observed, as most participants who reported alcohol use in our study did not report any binge episodes. If craniosynostosis is associated with heavy alcohol consumption, but not lower levels of consumption, our study may have been underpowered to detect this effect. Our study had 80% power to detect a minimum effect size of 1.25 for binge consumption; it would require a study almost seven times larger to detect a statistically significant effect size of 1.09, the OR that we observed for binge consumption. Our ORs for participants who reported at least one binge drinking episode were generally higher than those for any alcohol consumption, but some estimates remained inverse, and all CIs included the null. NBDPS did not collect data on frequency of binge episodes, so we have limited ability to explore a dose-response effect with higher intensity consumption.

Unlike most other structural birth defects, which occur early in gestation, premature suture closure can occur any time, extending the etiologically relevant exposure window for craniosynostosis through the end of pregnancy. However, we were limited in our ability to analyze the association between later pregnancy alcohol exposure and craniosynostosis due to the small number of participants who reported alcohol consumption only in the second or third trimester. Most mothers who reported alcohol consumption during the second or third trimester also reported periconceptional consumption; to isolate the effect of timing, we excluded these participants from our sub-analysis. Our inverse AOR for second or third trimester alcohol consumption was similar to that reported earlier²² but should be interpreted with caution given the limited scope of this sub-analysis. We cannot rule out the possibility that later pregnancy alcohol consumption is associated with craniosynostosis.

Our study is limited by several potential sources of bias. We relied on self-reported exposure data. Alcohol is a known teratogen and participants were asked about consumption patterns after their pregnancy outcome was known. If case mothers were more likely to underreport alcohol consumption than control mothers, due to social stigma, this could bias estimates downward. Our bias-adjusted estimates based on this assumption (scenarios 1–3) were indeed higher than our conventional estimates for periconceptional alcohol consumption, but not high enough to suggest a meaningfully positive association.

Although the NBDPS interview was wide ranging in scope, allowing us to adjust for known confounders, it remains possible that our results are biased by unmeasured confounding. Misclassification of the covariates included in our model could also lead to incomplete control for confounding. We based our modeling strategy on a DAG, which we developed based on a review of the existing literature and expert opinion. However, we made certain assumptions which may or may not be correct, potentially affecting the validity of our statistical model. Nevertheless, our results presented in Table 1 generally confirmed associations in our sample that had been demonstrated in others informing our DAG and did not yield any additional variables strongly associated with craniosynostosis.

Finally, selection bias could play a role in our results. Almost one-third of mothers of eligible cases and controls did not participate in the NBDPS interview.²⁴ Non-participants are likely systematically different than participants, and to the extent that those differences are associated with craniosynostosis risk and/or alcohol use, our results could be biased in either direction. Furthermore, 2.6% of case and 2.9% of control participants did not respond to questions about periconceptional alcohol use and were excluded from our analysis. Among those non-respondents for whom data are available, we know they were younger at the time of delivery, more frequently reported non-Hispanic Black or Hispanic race/ethnicity, and more frequently reported unintended pregnancy than those included in the analysis (data not shown). These are characteristics generally associated with higher rates of alcohol use,⁴³ however given that exclusion due to missing data did not differ by case status, any bias is likely towards the null. In a previous NBDPS analysis of alcohol consumption and gastroschisis, results in which alcohol consumption status was imputed did not meaningfully differ from those in which participants with missing data were excluded.²⁸

Despite these limitations, our study has several notable strengths. The NBDPS is among the largest population-based studies of birth defects, enabling us to analyze associations with specific craniosynostosis subtypes. Our robust case classification process is another major strength, minimizing outcome misclassification, ruling out syndromic cases, and further supporting our ability to analyze subtypes. Our population-based, multi-center design also promotes geographic and demographic diversity in our sample, increasing external validity. Finally, our quantitative bias analysis is a distinct strength, offering an estimate of the magnitude and direction of bias due to the self-reported data on maternal alcohol exposure in our study.