The VSD is a collaboration between the Centers for Disease Control and Prevention Immunization Safety Office and eight integrated healthcare systems. The aim of the VSD is to monitor the safety of vaccines routinely administered within the United States.¹⁶ We analyzed data from 7 participating VSD sites in 6 states (California, Colorado, Minnesota, Oregon, Washington, and Wisconsin) for the present study. The study was approved by the institutional review board at each participating site and the Centers for Disease Control and Prevention.

Our study population included pregnant women who delivered a live-born infant from January 1, 2004, through September 1, 2013, at a participating VSD site, identified using a validated algorithm¹⁷ and with linked birth records available. Women were required to have continuous insurance enrollment from 6 months before pregnancy through 6 weeks post-partum and have at least 1 medical encounter during pregnancy. Eligible women were also required to have, at minimum, a 1-week period during their first trimester overlap with a time period when influenza vaccine was available. For each influenza season, vaccine availability was defined as the time period between the first and last IIV administration within the cohort.

Infants surviving to 1 year of age were required to have at least 4 months of continuous insurance enrollment and at least 1 outpatient encounter. To help ensure identification of infants diagnosed with a severe major structural birth defect, those who died before 1 year of age or were hospitalized after birth for 30 days or longer were retained in the cohort even if they did not meet insurance enrollment or healthcare use criteria. We excluded women with multiple gestation pregnancies, those with prepregnancy or gestational diabetes, those with neoplasms, and those diagnosed with potentially teratogenic infections (eg, syphilis or toxoplasmosis) or exposures to teratogenic medications (eg, warfarin or phenytoin). In addition, women with exposures to live virus vaccine during their first trimester were excluded. Infants with birth weight of less than 350 g, gestational age of less than 20 weeks, or postnatally diagnosed chromosomal anomalies or congenital infections were also excluded. A full list of exclusions is available (Table I; available at www.jpeds.com).

Vaccinations were identified from standardized VSD files, which include data from electronic health records, medical and pharmacy claims, and linkage with state immunization information systems. Historical vaccines, administered at health fairs, schools, or the work place, were available if manually recorded at a healthcare visit. Using data on gestational age (in weeks) at birth and date of birth obtained from electronic health records or birth certificates, and based on clinician assessment at birth, we assigned a pregnancy start date to each pregnancy. Influenza vaccinations identified during pregnancy were categorized by gestational week of pregnancy. First trimester vaccination was defined as receipt of IIV before 14 weeks of gestation. Women receiving IIV in their second or third trimester or who did not receive IIV during pregnancy were classified as IIV unexposed during first trimester. The study period did include a period of monovalent H1N1 vaccine availability. Exposures to the monovalent H1N1 vaccine were captured in “other vaccines” received during pregnancy.

Major structural birth defects were identified from International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes occurring at outpatient, emergency, or inpatient visits in infant medical records. Our primary outcome was the presence of 1 or more prespecified major structural birth defects. As previously described, to improve specificity for our ICD-9–based outcomes, we applied outcome-specific algorithms.¹⁸ For example, congenital diaphragmatic hernia is defined as having 1 inpatient diagnosis by 3 months of age. Algorithms were developed and validated within a sample of the VSD population. These outcomes have been applied in a prior study on the safety of maternal pertussis vaccination.¹⁹ In secondary analyses, we examined structural birth defect outcomes by organ system (central nervous system, ophthalmologic or otologic, gastrointestinal, respiratory or orofacial, cardiac, genitourinary or renal, muscular or limb defects). We also examined severe cardiac defects, neural tube defects, microcephaly, and cleft lip or cleft palate as individual safety outcomes, applying outcome-specific algorithms for identification.

Descriptive statistics were used to compare baseline demographic variables, comorbidities, and healthcare use variables between first trimester exposed and unexposed women. Propensity scores, predicting the probability of influenza vaccination during first trimester, were developed using a regression generalized additive model with a smooth variable for maternal age and study week of last menstrual period. Other covariates, identified through diagnostic coding, pharmacy claims, or electronic health data and included in the propensity score were maternal race/ethnicity, neighborhood poverty level, Kotelchuck adequacy of prenatal care index,^20,21 smoking during pregnancy, hospitalization in the first 20 weeks of pregnancy, comorbidities (ie, hypertension and other cardiovascular disease, lung disease, nutritional deficiencies, neurologic conditions, renal disease, rheumatic disorders and thyroid disorders), exposure to a possibly teratogenic medication (ie, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, aliskiren, paroxetine, trimethoprim, and trimetrexate), and VSD site. Model fit for the propensity score was evaluated by the C-statistic and Hosmer-Lemeshow goodness-of-fit test.

A generalized linear model with Poisson distribution and log link was used to estimate prevalence ratios (PRs) with 95% CIs. Poisson distribution with identity link with robust variance estimation was used to estimate prevalence differences (PDs) with 95% CIs. Our primary contrasts were between women with first trimester IIV exposures and those unexposed during their first trimester. In addition to unadjusted PRs and PDs, we obtained adjusted estimates by adding propensity scores to the models as quintiles. In secondary analyses, we report PRs and PDs for infants, comparing those with maternal first trimester IIV exposures with those with no IIV exposure during pregnancy. Based on our available sample size, in primary analyses, we had 80% power to detect a 0.23 per 100 live birth PD for selected major structural defects with a background rate of 1.6 per 100 live births. For severe cardiac defects, we had 80% power to detect an 8 per 10 000 live birth PD with a background rate of approximately 17 per 10 000 live births.¹⁸ All analyses were conducted using SAS version 9.3 (SAS Institute, Cary, North Carolina).

We initially identified 530 477 pregnant women with singleton live births meeting insurance enrollment criteria. Of these, 38 808 (7.3%) were excluded based on maternal factors (comorbidities, use of teratogenic medications, receipt of a live virus vaccine, or no recorded prenatal care in a VSD health system). An additional 38 284 (7.2%) were excluded because their first trimester of pregnancy did not overlap with influenza vaccine availability, as defined in the Methods. Of the 453 385 pregnant women remaining after maternal exclusions, 27 441 (6.1%) were excluded based on infant insurance enrollment, healthcare use, or clinical criteria (Figure 1; available at www.jpeds.com). Our final cohort consisted of 425 944 mother-infant pairs, of whom 52 856 mothers (12%) received IIV in their first trimester of pregnancy. Of the 373 088 women who did not receive IIV during their first trimester, 90 991 (24%) received IIV in their second or third trimester of pregnancy. Vaccinations occurred in 10 influenza seasons, with increases starting in the 2009-2010 season (Table II; available at www.jpeds.com).

In our final cohort, the majority of pregnant women were aged 25-34 years (62%), 39% were white non-Hispanic, 29% were Hispanic, and 16% were Asian (Table III). Most women received medical care in their first trimester and had an adequate/plus Kotelchuck prenatal care index.²¹ As compared with women unexposed to IIV in the first trimester, women exposed to IIV were significantly more likely to have been aged 25 years or older, white, non-Hispanic, have received medical care in their first trimester, or have had preexisting lung disease. Women exposed to IIV in the first trimester were also more likely to have received another inactivated vaccine during the first trimester (41% vs 23%). Other vaccines identified included the tetanus, diphtheria, acellular pertussis vaccine (77%), monovalent H1N1 vaccine (22%), and meningococcal vaccine (<1%) (Table I). Even after excluding women whose first trimester did not overlap with periods of influenza vaccine availability, the strongest predictor of receiving IIV in first trimester was the timing of the start of pregnancy or last menstrual period being in September (Figure 2; available at www.jpeds.com). Model fit for the propensity was adequate. The C statistic was 0.685 and the P value was less than .0001 (8 df) for the Hosmer-Lemeshow test.

Among 52 856 women who were exposed to IIV in the first trimester, 865 (1.6 per 100 live births) had an infant with a selected major structural defect versus 5730 (1.5 per 100 live births) in the first trimester unexposed group. The adjusted PR for having a selected major structural birth defect after first trimester IIV was 1.02 (95% CI 0.94-1.10). The most common subgroup of major structural birth defects observed was cardiac defects, occurring in 58 per 10 000 live births among first trimester IIV-exposed women versus 56 per 10 000 live births in unexposed women. The adjusted PR for cardiac defects was 1.0 (95% CI 0.89-1.10). A subset of 14 severe cardiac defects, likely to require urgent or emergent intervention at birth, occurred in 13 per 10 000 live births among first trimester IIV-exposed women and 12 per 10 000 live births among unexposed women, with an adjusted PR of 0.99 (95% CI 0.76-1.28) (Table IV).

Findings for all additional subgroups of major structural birth defects, stratified by organ system, were consistent, with no increased risks after first trimester maternal IIV exposure compared with those without first trimester maternal IIV exposure. There was also no increased risk observed for neural tube defects, microcephaly, or cleft lip and/or cleft palate after first trimester maternal IIV exposure (Table IV). Results were consistent in secondary analyses, where women with first trimester IIV exposures were compared with women who were unexposed to IIV throughout pregnancy (Table V).

We examined risks for more than 50 prespecified major structural birth defects among singleton, live births to 425 944 women, including 52 856 (12%) with IIV exposure in their first trimester. In this large observational study, we did not observe an increased risk for selected major structural birth defects after first trimester vaccination. Because IIV is currently recommended for all women who will be pregnant during periods of influenza circulation,²² these data should provide reassurance for women considering first trimester vaccination.

Our findings add to the current literature supporting the absence of an association between early maternal vaccination and major structural birth defects in offspring. In 2014, McMillan et al¹⁴ systematically reviewed 12 studies of maternal influenza vaccination, including approximately 4000 women with first trimester exposures. Across these studies, odds ratios for first trimester IIV and birth defects (where specific birth defect outcomes varied across studies) ranged from 0.67 to 2.18; no study reviewed reported an association that was statistically significant.¹⁴ In 2015, Polyzos et al²³ conducted a meta-analysis of largely overlapping studies of 4733 women exposed to IIV in the first trimester and reported an odds ratio of 0.98 (95% CI 0.95-1.20) for major birth defects in offspring. More recently, studies from US and Swedish cohorts, including nearly 15 000 first trimester IIV exposures, along with 1 US-based case-control study, have demonstrated safety of first trimester influenza vaccination with respect to birth defects in offspring.^24-26

One strength of our study was our large population of women with first trimester IIV exposure. As such, we were able to examine subgroups of major structural birth defects, along with a few individual birth defects. Our findings were consistent across all subgroups when stratified by organ system or specific birth defect. Our adjusted PRs generally ranged from 90 to 1.05, with tight CIs around these point estimates. In addition, our birth defect outcomes, ICD-9–based algorithms, were developed and validated for use in the VSD population, consistent with recommendations for maternal vaccine safety surveillance.¹⁵ As part of this process, we determined that ICD-9 codes were valid for identifying defects by system but were prone to error for distinguishing similar defects, such as differentiating cleft palate with cleft lip from isolated cleft palate. Thus, a majority of outcomes analyzed were for groups of defects and not isolated defects. As previously described, for most groups of defects, background rates in the VSD were similar to those from population-based surveillance systems.¹⁸

Additional strengths of our study were that we were able to systematically identify and exclude pregnant women and infants who were a priori at increased risk for major structural birth defects owing to comorbidities (eg, diabetes), or drug exposures (eg, oral retinoids or valproate), diagnosed chromosomal anomalies, or congenital infections. Also, we were able to identify and adjust for other important potential confounders through the use of a propensity score. Of note, even after excluding mothers whose first trimester did not overlap with influenza vaccine availability (eg, last menstrual period is May 1,and thus first trimester ends August 13) the strongest predictor of maternal first trimester vaccination was timing of pregnancy start. Thus, as in our approach, calendar month of conception or pregnancy start would be important to account for in observational studies of maternal vaccine or drug exposures, especially if there is also seasonality for the outcomes. For some birth defects, seasonality has been described, although results have varied by study.^27-29

Several limitations should be noted. First, our selected major structural birth defect outcomes were based on ICD-9 codes and were not confirmed by clinical review or adjudication of cases. To improve specificity, we applied outcome-specific algorithms, developed and validated for use in the VSD. These algorithm outcomes would be classified as Brighton level 3, or meeting case definitions with the lowest level of diagnostic certainty.¹⁵ Achieving a higher level of diagnostic certainty would require additional review of clinical, laboratory, or imaging data, which is not routinely available in our automated VSD files. Manual review of charts for more than 5000 infants with major structural birth defects identified in this study was cost prohibitive. Also, because our birth defect definitions were consistent across IIV-exposed and -unexposed groups, potential misclassification of outcomes would likely be nondifferential. Additionally, because our list of outcomes was prespecified to only include those with valid and specific ICD-9 codes available, our outcomes may differ from those applied in other studies of maternal vaccine safety. Furthermore, our approach was not designed to detect syndromes or associations of defects. Other limitations were that we were only able to identify and account for risks for birth defects if they were documented in electronic healthcare data. For example, we were able to identify and exclude women with potential exposures to teratogenic prescription medications, but we could not determine whether women were taking folic acid or whether there was maternal alcohol or illicit drug use during the first trimester. Similarly, the potential that influenza vaccines administered through health fairs or the workplace would not be captured in our standardized files could lead to some misclassification of exposure status.³⁰ Our cohort was limited to women with continuous insurance coverage; nearly all had a medical visit during first trimester. As such, infants in our study were likely at lower risk than the general US population for having a major structural birth defect. Finally, owing to limitations in our data sources, we were unable to include birth defects identified in stillbirths, spontaneous abortions, or resulting in therapeutic abortions; therefore, our results are specific to pregnancies ending in a singleton live birth.

In this large observational study, we did not observe increased risks for first trimester maternal IIV exposure and major structural birth defects in offspring. This study supports the safety of current recommendations for IIV to be administered to women who may be pregnant, in any trimester, during influenza season. ∎