Pregnant women and infants are at increased risk of morbidity and mortality from influenza infections.^1–3 Infants are not eligible for influenza vaccination until 6 months of age;⁴ however, maternal influenza vaccination during pregnancy can confer passive immunogenic protection for infants through transfer of vaccine-induced maternal antibodies via the placenta and breastmilk.^5–8 Maternal influenza vaccination during pregnancy is associated with decreased risks of influenza-related hospitalization in pregnant women and in infants less than 6 months of age.^9,10 Inactivated influenza vaccine has been recommended for pregnant women in any trimester of pregnancy by the Advisory Committee on Immunization Practices (ACIP) and the American College of Obstetricians and Gynecologists since 2004.^11–14 Although influenza vaccination coverage among pregnant women in the United States has increased substantially since the 2009–2010 H1N1 pandemic,¹⁵ it was approximately 60% by 2020.^16–18

Concerns regarding fetal safety are an important barrier for maternal vaccination.^19,20 Existing evidence on the safety of seasonal and pandemic H1N1 inactivated influenza vaccines in relation to birth defects is reassuring.^21–28 However, ACIP considers the existing data on influenza vaccine administration specifically during the first trimester, the etiologically-relevant period for most structural birth defects, to be relatively limited.⁴ Most of the studies investigating any birth defect or any major birth defect have reported null associations with first trimester influenza vaccination^21–26 as have the few studies examining specific major birth defects or groups of major birth defects.^21,22,27,28 Because etiologies of specific birth defects can vary, safety signals can be missed when grouping all birth defects together.²⁹ Therefore, we conducted an analysis of early pregnancy influenza vaccination and selected individual and groups (e.g., neural tube defects [NTDs]) of major structural non-cardiac birth defects using data from the National Birth Defects Prevention Study (NBDPS). The association between early pregnancy influenza vaccination and cardiac birth defects is the focus of a separate analysis.

There were 8233 case and 4937 control children eligible for the analysis (Figure 1). Among mothers of these children, 11.1% reported having influenza vaccination during the month before pregnancy through delivery (B1P9). Among mothers with an EDD in 2010 or 2011, 18.4% reported having influenza vaccination during B1P9 (data not shown). During B1P3, 334 (4.1%) mothers of case children and 197 (4.0%) mothers of control children reported receiving influenza vaccination during B1P3. No mothers reported receiving nasal mist formulations of the vaccination (i.e. live attenuated influenza vaccine) during B1P3, which are contraindicated during pregnancy.¹⁴ Selected characteristics of control mothers are reported by vaccination status in Table 1. Control mothers who reported influenza vaccination more often delivered after 2009 with an EDD in April through September, had at least 13 years of education, were 35 years of age or older, were non-Hispanic White, more often reported drinking, and less often reported smoking compared with those who did not report influenza vaccination.

We calculated ORs for 19 non-cardiac major birth defects or groups of defects (Table 2). The CIs for only two birth defects, hypospadias and craniosynostosis, excluded the null before adjustment for covariates. However, the estimates were attenuated upon adjustment for the limited set of covariates [hypospadias adjusted (a)OR: 1.25 (95% CI 0.89, 1.76); craniosynostosis aOR: 1.23 (95% CI: 0.88, 1.74); Table 2]. Adjusted ORs for the limited set of covariates ranged from 0.53 for omphalocele with six exposed case children to 1.74 for duodenal atresia/stenosis with eight exposed case children. Most of the aORs (11 of 19) were ≤1 and all CIs included the null after adjustment for covariates. No aORs in the primary analysis exceeded 2.00. Results were similar when adjusting for the limited set of covariates using propensity score adjustment versus standard multivariable adjustment (Supplemental Table S1) and when adjusting for the limited versus expanded sets of covariates (Table 3).

Results from analyses including case children with isolated defects only, singletons only, and mothers whose first trimester overlapped with typical influenza vaccination months are presented in Table 4. Overall, results were similar to the primary analysis. Only two aORs exceeded 2.00 in sensitivity analyses, isolated duodenal atresia/stenosis, and isolated anophthalmia/microphthalmia, although results were imprecise. Adjusted ORs were less than 0.50 in sensitivity analyses for one birth defect, omphalocele. Furthermore, excluding case or control children with hypospadias and craniosynostosis that had a family history of the same birth defect did not change the adjusted associations materially (Table 5). Results were similar when excluding the 107 case and 57 control children with mothers who reported receiving influenza vaccination between 2 and 3 months before conception, including 16 mothers who also reported receiving influenza vaccination during B1P3 (Supplementary Table S2).

In our analysis of early pregnancy inactivated influenza vaccination and 19 selected major structural non-cardiac birth defects or birth defect groups in the NBDPS, all aORs had CIs that included the null. All aORs in the primary analysis were less than 2.00 and most were less than 1.00. We conducted several sensitivity analyses, and all analyses yielded similar results.

Our results generally align with previous studies of first trimester maternal influenza vaccination and birth defects. Studies of the association between first trimester inactivated influenza vaccination, either seasonal or pandemic H1N1, and any major birth defect suggest no increased risk.^21,22,24,26 The Slone Epidemiology Center Birth Defects Study, also a multi-site case–control study, investigated selected birth defects in relation to first trimester influenza vaccination.^27,28 The aORs were generally not elevated for seasonal influenza vaccination (2011–2012, 2012–2013, and 2013–2014 seasons) for 42 specific birth defects or categories of defects considered.²⁷ In an analysis of first trimester pandemic H1N1 influenza vaccination (2009–2010 and 2010–2011 seasons) in the Slone Epidemiology Center Birth Defects Study, most aORs were close to null for the 41 specific birth defects or categories of defects considered.²⁸ Three defects had aORs >2.00 following first trimester pandemic H1N1 influenza vaccination; however, each of these associations were based on two exposed cases.²⁸ We were able to investigate one of these three defects, anophthalmia/microphthalmia, following any influenza vaccination during B1P3 and observed CIs to include the null [aOR: 1.60 (95% CI: 0.62, 4.14)]. Using automated health care data, the Vaccine Safety Datalink, a collaboration between the Centers for Disease Control and Prevention and eight health care systems, reported no association between first trimester inactivated influenza vaccination and risk of any of the major birth defects or categories that were considered, including the defect categories that we evaluated, i.e., NTDs and cleft lip and/or cleft palate.²² A study using registry data from Sweden reported no increased risk of NTDs, oral clefts, or limb deficiency following first trimester pandemic H1N1 influenza vaccination.²¹ Similarly, we observed no increased risk of NTDs aOR: 0.87 (95% CI: 0.41, 1.83), oral clefts aOR: 0.94 (95% CI: 0.70, 1.31), or limb deficiency aOR: 0.98 (95% CI: 0.60, 1.60).

Our study has several strengths including the population-based data source. Furthermore, standardized interview protocols allowed for the collection of information on many potential confounders. NBDPS utilized detailed protocols to classify cases, allowing us to study specific birth defects and ensuring the accuracy of case classification.³⁰ Also, we were able to separately analyze isolated defects. Although there is an increased risk of type I error when studying multiple specific birth defects, we observed no harmful aOR for inactivated influenza vaccination and the selected birth defects. Although the design of the current study is similar to the Slone Epidemiology Center Birth Defects Study, our analysis had more case children with certain birth defects available.

Our study has limitations to consider. Only 18% reported influenza vaccination during pregnancy among those with an EDD in the last 2 years of the study (2010–2011), whereas seasonal influenza vaccination coverage during pregnancy in the United States was between 32% and 40% during the 2009–2010 and 2012–2013 influenza seasons, respectively, with H1N1 vaccine coverage around 45% during the 2009–2010 influenza season.^38,39 This discrepancy suggests under-reporting of influenza vaccination in our study. Previous validation studies among similar age groups in the general population have demonstrated sensitivities of 90–97%, and specificities of 68–92%, for self-reporting influenza vaccination status, including for the previous influenza season, compared with information from vaccine registries and/or medical records.^40–42 However, we were not able to confirm self-reported vaccination status, and we suspect underreporting of influenza vaccination in the current study to be greater than reported in previous validation studies due to 1) the lag between exposure and interview, especially among women who had to recall vaccination status from an influenza season more than one season before the interview, and 2) the interview indirectly querying about influenza vaccination status (i.e., “… did you take any medications, remedies, or treatments that we haven’t already talked about? For example, flu or allergy shots…”). Non-differential exposure misclassification would tend to bias results downward, potentially masking positive associations (i.e., harmful effects). Differential misclassification could bias results in either direction, again potentially masking positive associations. Also, exposure data were available up to three months before pregnancy only, and we could not exclude women who were less likely to receive influenza vaccination during pregnancy because they had received the vaccine earlier in the season and prior to 3 months before pregnancy. This may have resulted in an underestimate of influenza vaccination prevalence given that the denominator may have included women vaccinated before this reporting period. Furthermore, it may have resulted in violations of the positivity assumption, a requirement for causal inference. However, excluding children whose mothers reported influenza vaccination between 2 and 3 months before conception did not materially change the results. The prevalence of reported influenza vaccination in B1P3 across all study years was only 4%, which limited the number of birth defects we were able to assess and led to imprecise estimates for many of the defects. Additionally, the type of vaccination (seasonal or pandemic) was often unknown, and together with limited statistical power prevented us from studying type and season of influenza vaccination. Selection bias from lack of inclusion of early pregnancy loss could be a concern; however, existing studies generally do not support an association between inactivated influenza vaccination and early pregnancy loss.¹ Missing covariate values were more common among control mothers who were unexposed versus exposed. Nevertheless, the potential for selection bias in the complete case analysis was minimized by the relatively low levels of missing data (i.e., usually 5% or less). Although the vaccines in this study were administered more than a decade ago, we expect the findings to be applicable for contemporary influenza vaccine exposures as the type of influenza vaccines used in pregnant women during the study years (i.e., inactivated influenza vaccine) is still recommended.¹⁴ Influenza virus sub-groups included in influenza vaccines vary by year, but are not expected to affect the safety profile.

Although we cannot rule out small increased risks for some of the birth defects analyzed due to imprecise estimates, we observed no aORs of 2.00 or higher for early pregnancy inactivated influenza vaccination and any of the 19 selected major structural non-cardiac birth defects or birth defect groups that we studied. Furthermore, we cannot entirely rule out increased risks for the birth defects analyzed due to the potential for exposure misclassification. Our study, together with other studies using automated health care data, population-based registry data, and population-based case–control study design, strengthens the evidence that inactivated influenza vaccination in early pregnancy, including the month before conception, is not strongly associated with the risk for major structural birth defects.