Influenza infections can pose a significant threat to pregnant women and their newborns, including increased risk of maternal hospitalization and death and potential adverse birth outcomes such as premature birth, low birth weight, stillbirth, and birth defects [1,2]. Consequently, the World Health Organization, United States (US) Advisory Committee on Immunization Practices (ACIP), and the American College of Obstetricians and Gynecologists (ACOG) recommend influenza vaccination for pregnant women in any trimester [3–5]. Vaccination against influenza during pregnancy protects maternal and fetal health. It directly protects the mother from influenza infection, which in turn, protects the fetus from harmful consequences associated with maternal illness [2,6].

Even with global and national recommendations for influenza vaccination during pregnancy, vaccine coverage remains around 60 % [7–9]. Concern regarding the potential risk to fetal safety is one of the most recognized barriers to maternal vaccination [10,11]. Our analyses of 2006–2011 data from the US population-based, National Birth Defects Prevention Study (NBDPS) examining risk for over 30 major birth defects supported the safety of maternal influenza vaccination [12,13]. Influenza vaccine components may change from one year to another, requiring continued evaluation of the safety of seasonal vaccination. Of the 11 published studies identified using data since 2011 [2,14–23], four were summaries of post market surveillance reports of frequencies of several adverse infant events, including birth defects [18–20,23]. Of the remaining seven studies [2,14–17,21,22], none used population-based data to examine risk for individual major birth defects.

To maintain public confidence in vaccine safety and optimize maternal and fetal health outcomes, it is important to continue to evaluate the safety of maternal vaccinations using well-characterized, population-based datasets encompassing children with and without major birth defects. Building upon our prior publications [11,12], we used data for 2014–2019 deliveries enrolled in the US Birth Defects Study To Evaluate Pregnancy exposureS (BD-STEPS), a population-based, case-control study, to investigate potential associations between influenza vaccine exposure during the respective critical exposure period and selected individual birth defects.

The analytical sample for spina bifida included 1333 control children and 195 children with spina bifida (182 isolated case children and 13 with multiple defects); the sample for the other selected major birth defects included 1360 control children and 2033 children with other defects (1779 isolated case children, 253 with multiple defects, and 1 complex case). Characteristics of case and control participants are presented in Table 1, and characteristics of exposed and unexposed control participants are presented in Supplemental Table 2.

Influenza vaccination during B1P1 was reported by 4.5 % of control and 4.1 % (spina bifida) of case participants; reports during B1P3 were 14.1 % for control participants and ranged from 6.8 % (gastroschisis) to 17.5 % (diaphragmatic hernia) for case participants (Table 2). Of note, the analysis also evaluated the overlap between influenza vaccination periods (B1P1, B1P3) and the influenza season in the United States (September–March). For pregnancies with influenza vaccination exposure during the B1P3 period, the first trimester overlapped with the influenza season in 77.9 % of cases and 83.4 % of controls. A similar pattern was observed for pregnancies with influenza vaccination during the B1P1 period (83.3 % of cases and 83.5 % of controls).

In the primary analyses, the aOR for influenza vaccination was near the null for spina bifida (aOR: 0.9; 95 %CI: 0.4–2.0), and those for exposure during B1P3 and the other selected major birth defects ranged from 0.4 (95 %CI: 0.2–0.7) for gastroschisis to 1.3 (95 %CI: 0.9–1.9) for cleft palate alone (Table 2). Estimates greater than the null were observed for dextro-transposition of the great arteries, total anomalous pulmonary venous return, and cleft palate alone; null or inverse estimates were observed for anotia/microtia, coarctation of the aorta, hypoplastic left heart syndrome, tetralogy of Fallot, cleft lip ± cleft palate, esophageal atresia, transverse limb deficiency, diaphragmatic hernia, and gastroschisis. Only the 95 % CIs for cleft lip ± cleft palate (aOR: 0.6; 95 %CI: 0.4–0.9) and gastroschisis (aOR: 0.4; 95 %CI: 0.2–0.7) excluded the null.

The secondary analysis using standard multivariable adjustment and a limited set of covariates produced similar results (Table 2), supporting the propensity score-adjusted estimates produced in the primary analyses. Another secondary analysis using an expanded analytical sample that included participants with incomplete or imprecise influenza vaccination timing data also tended to yield similar results to the primary analyses (Table 3).

Additional secondary analyses that restricted case children to those classified as isolated (Supplemental Table 3), excluded case and control children with a family history of the same site defect (Supplemental Table 4), restricted case and control children to singleton pregnancies (Supplemental Table 5), and restricted participants to those with first trimester of pregnancy overlapping with influenza season (Supplemental Table 6) yielded results similar to those of the primary analysis. Results of our probabilistic bias analysis examining the potential effect of exposure misclassification (Supplemental Table 7) were also similar to those of the respective primary or secondary analysis.

Our study investigated potential associations between maternal influenza vaccination shortly before or during pregnancy and the risk of major birth defects in offspring. Vaccine uptake during B1P1 was reported to be less than 5.0 %, and uptake during B1P3 across defects ranged from 6.8 to 17.5 %. Primary adjusted analyses applying propensity scores showed no statistically significant positive associations between maternal influenza vaccination and any of the selected birth defects studied, with most associations being near or below the null. Several secondary analyses applying standard multivariable adjustment with a limited set of covariates, using an expanded analytical sample, or restricting to isolated case children, children with no family history of same site defect, singleton births, and participants whose first trimester of pregnancy overlapped with influenza season tended to yield similar results to the primary analyses. Our probabilistic bias analysis also tended to yield similar results to respective primary or secondary analyses, although we were limited to examining the potential effect of exposure misclassification using an estimate of specificity based on current-year vaccination status rather than a lower but potentially more accurate estimate based on previous-year vaccination status. Together, these analyses strengthen the evidence that influenza vaccination during critical exposure period does not increase the risk of selected major birth defects.

In our six-year BD-STEPS study period (2014–2019), vaccine uptake among participants (control, case) during B1P1 (4.5 %, 4.1 %) and B1P3 (14.1 %, 6.8–17.5 %) exceeded the respective proportions observed during our six-year NBDPS study period (2006–2011) for B1P1 (1.3 %, 1.0 %) and B1P3 (4.3 %, 2.4–8.5 %). We also observed differences (≥10 % absolute change) in the magnitude of the associations estimated in each study, with only the associations for cleft lip ± cleft palate and gastroschisis in BD-STEPS being statistically significant. Positive associations observed for vaccine exposure during B1P3 in NBDPS were attenuated in BD-STEPS for coarctation of the aorta (aOR = 1.3 to 0.9), total anomalous pulmonary venous return (aOR = 2.3 to 1.1), esophageal atresia (aOR = 1.2 to 0.6), and diaphragmatic hernia (aOR = 1.3 to 0.9). Inverse associations observed in NBDPS were strengthened in BD-STEPS for cleft lip ± cleft palate (aOR = 0.9 to 0.6) and gastroschisis (aOR = 0.8 to 0.4) but attenuated for hypoplastic left heart syndrome (aOR = 0.6 to 0.9), tetralogy of Fallot (aOR = 0.7 to 1.0), and transverse limb (aOR = 0.7 to 0.8). Associations in NBDPS for cleft palate strengthened from 1.0 to 1.3 in BD-STEPS. Lastly, associations for anotia/microtia (aOR = 0.7 versus 0.7) and dextro-transposition of the great arteries (aOR = 1.3 versus 1.2) were similar (<10 % absolute change) between the two studies. Our current findings support our earlier conclusions regarding the safety of influenza vaccine during early pregnancy.

Influenza vaccine uptake among case and control participants is not directly comparable to previous studies, as we only had data for vaccination during the first trimester of pregnancy, not all trimesters of pregnancy. Additionally, among the seven studies cited that examined birth defect risk [2,14–17,21,22], only Louik et al. [16] reported findings for individual birth defects. Where data were available, our findings were comparable for the inverse association for cleft lip ± cleft palate and a positive, but nonsignificant, association for cleft palate alone. Conversely, our positive association for anomalous pulmonary venous return differed from the inverse association reported by Louik et al. [16]. Likewise, our inverse association for anotia/microtia and that near the null for diaphragmatic hernia differed from the positive associations reported by Louik et al. [16]. Our findings were not comparable to those for defect groups reported by Louik et al. [16], Kharbanda et al. [15], or Sarna et al. [21], due to the different individual phenotypes included within a group (e.g. congenital heart defects).

Our findings are limited to the major birth defects eligible for BD-STEPS. Additionally, our findings do not generalize beyond the US. Defects eligible for BD-STEPS are a reduced number compared to those included in the NBDPS and the study by Louik et al. [16] that examined risk for individual birth defects. An additional limitation was our exclusion of BD-STEPS eligible birth defects that had fewer than five exposed case children. Nonetheless, our use of population-based surveillance programs to identify children diagnosed with birth defects and rigorous, systematic review and classification of eligible BD-STEPS birth defects remained consistent with the approach used in NBDPS. BD-STEPS data also included fetal deaths and elective terminations and used population-based birth data to select control children, consistent with the approaches used in NBDPS. Our case ascertainment approach, however, was unable to systematically ascertain early fetal losses before 20 weeks gestation, a limitation shared with NBDPS.

Use of observational studies to monitor the safety of influenza vaccination during pregnancy poses challenges, such as the potential for influenza vaccine components to change from year to year, varied timing of administration, and low prevalence of outcomes, such as birth defects. As such, these studies require ongoing collection of maternal influenza vaccinations using large study samples. BD-STEPS collection of self-reports of influenza vaccination was conducted using telephone interviews. Although this approach was consistent with NBDPS methods, it may lead to exposure misclassification, particularly given the recall period spanned more than one influenza season. Based on the limited relevant literature identified that evaluated recall for influenza vaccination among reproductive-aged adults, [30] we were unable to determine the potential for differential misclassification for such recall. Compared to results from our primary analysis, our probabilistic bias analysis for non-differential misclassification showed similar aOR estimates with less precise 95 % CIs. Another limitation of our interview data collection was that receipt of influenza vaccine but not type of inactivated vaccine (e.g. egg-based, cell-culture-based, or antigen changes by season,) was requested. To minimize this limitation, we controlled for quarter and year of conception.

To control for confounding, we used propensity score adjustment for several covariates in our primary analyses and conducted a secondary analysis applying standard multivariable adjustment and a limited set of covariates, consistent with the analytical approaches we used with NBDPS data. Findings were similar between our primary analyses using propensity scores and secondary analyses using standard adjusted models, supporting our propensity score-adjusted estimates produced in the primary analyses. We also conducted several secondary analyses using more homogeneous samples (e.g. isolated case children only, singleton births only, etc.); findings from these analyses also tended to support those from our primary analyses. Even with these various analyses, we cannot rule out residual confounding.

Influenza infection during pregnancy puts both the mother and the child at increased risk of severe outcomes. As such, the World Health Organization, United States (US) Advisory Committee on Immunization Practices (ACIP), and the American College of Obstetricians and Gynecologists (ACOG) continue to recommend influenza vaccination for pregnant women during pregnancy [2–4]. Addressing vaccine hesitancy among pregnant women can optimize maternal and child health. Future studies would benefit from large, multinational samples and additional types of major birth defects to expand evaluation of the safety of influenza vaccination during pregnancy, although our findings reinforce the safety of influenza vaccination during pregnancy for the birth defects studied, supporting public health efforts to promote vaccination coverage among pregnant women.