Early Childhood Health Outcomes Following In Utero Exposure to Influenza Vaccines: A Systematic Review

We conducted a systematic review of the literature related to IIV during pregnancy and childhood health outcomes, as guided by the minimum evidence-based set of items for reporting in systematic reviews outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.²⁹  The study protocol was registered with the National Institute for Health Research international prospective register of systematic reviews (CRD42014014384) before commencing the review.

We searched CINAHL Plus, Embase, Medline, Scopus, and Web of Science databases for peer-reviewed literature from inception to July 24, 2019, using a combination of medical subject headings and keywords related to prenatal influenza vaccination and early childhood health outcomes (Supplemental Tables 6–10). Our search included experimental and observational studies, including cross-sectional, case-control, and cohort studies. As recommended,^30,31  we consulted with a medical librarian to develop our search strategy.

Eligible studies included those investigating any mother–child pair population, in which exposure to influenza vaccines in utero (pandemic or seasonal) was reported, an unexposed mother–child pair control group is present, and ≥1 health outcome in children aged 6 months to 5 years was investigated. We made the following exclusions: articles not published in peer-reviewed journals, articles published in languages other than English, studies not conducted in humans, reviews, editorials, commentaries, letters, case studies, and case series. First, two independent reviewers (D.Y.P.F. and M.S.) screened and reviewed the titles and abstracts of records retrieved during the search for inclusion criteria. Second, the reviewers screened and reviewed the full-text articles for eligibility in accordance with study inclusion and exclusion criteria. Studies deemed to meet the inclusion criteria by the two reviewers were included in the final review. A third reviewer (A.K.R.) resolved any conflicts between the two reviewers during each screening and review stage.

We developed a standardized data collection form to extract information on study characteristics, including study design, geographic location, participant demographics, definition and ascertainment of exposure and outcomes (including type of influenza vaccine), effect sizes and confidence intervals (CIs), and confounding variables. Two reviewers (D.Y.P.F. and M.S.) independently extracted information from each of the included articles.

For observational studies, we used the Newcastle–Ottawa scale to assess risk of bias (Tables 1–3).³²  The Newcastle–Ottawa scale has a maximum score of 9, with a greater score indicating the lowest risk of bias. The scoring system is based on the following criteria: selection bias (maximum score: 4), comparability of study groups (maximum score: 2), and ascertainment of exposure for case-control studies or outcome for observational studies (maximum score: 3).³²  Observational studies were considered at low risk of bias if they scored ≥8 and moderate-high risk of bias if they scored <8.

For randomized controlled trials (RCTs), we used the Cochrane risk-of-bias tool to assess risk of bias.³¹  The Cochrane risk-of-bias tool addressed 5 major domains: selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and personnel and other potential threats to validity), detection bias (blinding of outcome assessment and other potential threats to validity), attrition bias (incomplete outcome data), and reporting bias (selective outcome reporting).³¹  Consistent with the Cochrane risk-of-bias tool scoring, articles were classified as low risk of bias, some concerns of bias, or high risk of bias.³¹  Two reviewers (D.Y.P.F. and M.S.) independently assessed the quality of case-control and cohort studies and RCTs using the Newcastle–Ottawa Scale⁴²  and Cochrane risk-of-bias tool,⁴³  respectively. Any conflicts between the two reviewers during the quality assessment process were resolved by a third reviewer (A.K.R.).

We developed a narrative description of the characteristics and results of included studies. Where possible, results were provided by trimester of vaccination and type of vaccine. To generate pooled effect estimates for each outcome, a random effects meta-analysis was planned a priori, dependent on whether a sufficient number of studies were retrieved using commonly defined end-points (n > 2).

This systematic review was supported in part by funding received from the National Health and Medical Research Council (GNT1141510), Curtin University Graduate Research School, and the Wesfarmers Centre of Vaccines & Infectious Diseases at the Telethon Kids Institute. The funders had no role in the design or implementation of the systematic review or the decision to publish findings from the review.

We identified 3647 records, of which 3551 were excluded after initial title and abstract screening. We reviewed 96 full-text articles, of which 9 were deemed eligible for inclusion (Fig 1). Reasons for exclusions included no measurement of influenza vaccination during pregnancy (n = 4), no reported outcome measure of early childhood health (n = 33), not a comparative observational study or RCT (n = 44), or not published in English (n = 6).

Of the 9 included studies, methodology and study outcomes were highly diverse (Table 4). All of the studies originated from high-income countries in North America (n = 5) or Europe (n = 4), and study periods ranged between 1980 and 2012. Estimates from the same population were reported in 2 articles.^33,37  One study was an RCT,⁴¹  5 studies were retrospective cohort studies, ^{33,35,36,38,39}  and 2 were prospective cohort studies^34,37 ; 1 study used a case-control design.⁴⁰  Record linkage was used in most studies to measure prenatal influenza vaccination and early childhood health outcomes (n = 7). Studies ranged in sample size from 306 to 275 500 mother–infant pairs. With the exception of 1 study in which children 6 to 12 months of age were specifically examined,⁴⁰  the follow-up period of included studies began at birth and ranged from 1 to 15 years of age.

Researchers in 6 studies investigated pandemic^{34,36,38,39,41}  influenza vaccines, and researchers in 3 investigated seasonal^33,37,40  influenza vaccines (Table 4). Of the 6 studies on 2009 monovalent pandemic influenza A virus subtype H1N1 (A/H1N1) vaccines, researchers of 4 studies exclusively investigated adjuvanted vaccines (MF59-adjuvant: n = 2^35,41 ; AS03-adjuvant: n = 2^34,38 ). Vaccination status during pregnancy was ascertained by self-report (n = 1),³⁵  random allocation (n = 1),⁴¹  written or electronic health records (n = 3),^33,37,40  and registry information (n = 4).^34,36,38,39 

Outcomes included infectious (n = 7), atopic (n = 2), autoimmune (n = 2), and neurodevelopmental (n = 3) conditions, neoplasms (n = 1), and all-cause morbidity (n = 2) and mortality (n = 2) (Table 5). Infections or infectious conditions included influenza, pneumonia, otitis media, sepsis, acute respiratory infections, gastrointestinal infections, viral infections, and all-cause infection-related primary care contact. Atopic conditions included asthma. Autoimmune conditions included celiac disease, ulcerative colitis, Crohn disease, juvenile arthritis, Sjögren syndrome, vasculitis, reactive arthropathy, idiopathic thrombocytopenic purpura, type-1 diabetes, Bell palsy, and Guillain-Barré syndrome. Neurodevelopmental conditions included epilepsy, autism spectrum disorder, intellectual disability, and sensory disorders. All-cause morbidity outcomes included 1-year all-cause hospitalization, 3-year all-cause hospitalization, 5-year all-cause hospitalization, urgent and inpatient health services used, and pediatric complex chronic conditions. International Classification of Diseases (ICD) clinical diagnosis codes were used in the majority of studies (n = 5) to identify outcomes.

Given the small number of studies in which similarly defined outcomes were addressed, meta-analyses were deemed not possible.

Among the 9 included studies, several maternal and child characteristics were included as potential confounders (Supplemental Table 11). Maternal characteristics included maternal age (n = 6), ethnicity (n = 2), place of birth (n = 3), place of residence (n = 3), socioeconomic status and income (n = 4), education (n = 3), BMI (n = 2), parity (n = 5), medical comorbidities (n = 5), pregnancy complications (n = 3), multiple gestations and/or births (n = 3), smoking during pregnancy (n = 4), season of conception (n = 2), and gestational age (n = 2). Child characteristics mostly included sex (n = 4). Other characteristics, such as marital status, calendar year of conception, use of antenatal care, medication use, the child’s ethnicity, and other birth outcomes, were less commonly controlled for as confounders (n = 1) (Supplemental Table 11). One study accounted for the potential influence of childhood influenza immunization by censoring at age at influenza vaccination.³³ 

In the 2 studies in which researchers examined influenza infection, infection status was ascertained either by a positive result by direct fluorescent antibody test (for laboratory-confirmed influenza)⁴⁰  or by using primary or secondary diagnostic codes for (1) influenza alone or (2) influenza and pneumonia.³⁶  Outcome measures were expressed as vaccine effectiveness (VE; calculated as [1 − odds ratio comparing the odds of infection in exposed versus unexposed] × 100%) or crude incidence rates for influenza infection among children exposed to IIV in utero versus unexposed. Benowitz et al⁴⁰  assessed laboratory-confirmed influenza and did not identify an association between IIV exposure in utero and VE in infants aged ≥6 months (unadjusted VE: −41.4%; 95% CI: −2257.4% to 91.5%) (Table 5). However, as noted by Benowitz et al,⁴⁰  because of the small sample size of infants aged ≥6 months, there was low statistical power to assess VE. In a study that followed infants ≤1 year of age, Fell et al³⁶  observed incidence rates of influenza that did not differ between children exposed to IIV in utero compared with unexposed children for each influenza time period examined, including the 2009 A/H1N1 pandemic season (incidence rate ratio: 0.61; 95% CI: 0.32 to 1.17) and the post-2009 A/H1N1 pandemic period (adjusted incidence rate ratio [aIRR]: 1.05; 95% CI: 0.81 to 1.37). Incidence rates for influenza and pneumonia were significantly higher among infants exposed to IIV in utero compared with unexposed infants in the post-2009 A/H1N1 pandemic period (aIRR: 1.17; 95% CI: 1.05 to 1.31). However, incidence rates for influenza and pneumonia did not differ significantly during the 2009 A/H1N1 pandemic season (aIRR: 1.04; 95% CI: 0.84 to 1.29) (Table 5).

Researchers of two studies assessed infection-related primary care contact (fever, symptoms of infection of ≥1 organ system, or prescriptions for infectious symptoms) by examining medical records given by primary care providers³⁵  and infections (common cold, pharyngitis, otitis, pneumonia, and gastrointestinal infection) or fever by parent-reported daily diary cards reviewed by a research doctor.⁴¹  Neither van der Maas et al³⁵  (aIRR: 1.07; 95% CI: 0.91 to 1.28) nor Bischoff et al⁴¹  (aIRR: 0.99; 95% CI: 0.84 to 1.18) found a difference in incidence of infections between children of vaccinated and unvaccinated mothers in their studies (Table 5).

Studies in which upper respiratory tract infections, lower respiratory tract infections, and all-cause infections were examined found a statistically significant reduction in risk of upper respiratory tract infections (adjusted rate ratio [aRR]: 0.92; 95% CI: 0.85 to 0.99)³⁸  and all-cause infections (aRR: 1.71; 95% CI: 1.08 to 2.44)³⁸  but not lower respiratory tract infections (Table 5).^38,39  Hviid et al³⁸  observed an inverse association with upper respiratory tract infections and all-cause infections in children whose mothers were vaccinated with pandemic influenza vaccine in the second or third trimester and in the first trimester, respectively. However, after adjusting for multiplicity using a Bonferroni correction, these associations were no longer significant. Data from one of 2 studies^33,39  that investigated otitis media in infants ≤1 year of age reported a lower absolute adjusted VE (calculated as 1 – the ratio of the incidence of otitis media in exposed versus unexposed children) for children whose mothers received a pneumococcal conjugate vaccine while pregnant (37.6%; 95% CI: 23.1 to 49.4) than children whose mothers received a pneumococcal conjugate vaccine and trivalent IIV while pregnant (47.9%; 95% CI: 42.0 to 53.3), as compared with the children whose mothers received no vaccines during pregnancy.³³  The second study, by Walsh et al,³⁹  followed offspring ≤5 years of age and found no difference in incidence rates between children exposed to IIV in utero and unexposed children (aIRR: 1.03; 95% CI: 1.00 to 1.06). Walsh et al³⁹  and Hviid et al³⁸  each observed a statistically significant reduction in the risk of gastrointestinal infections (adjusted hazard ratio [aHR]: 0.94; 95% CI: 0.91 to 0.98 and aRR: 0.84; 95% CI: 0.74 to 0.94, respectively). Hviid et al³⁸  observed this association in children whose mothers were vaccinated in either the second or third trimester but not in the first trimester. Second or thirdtrimester prenatal influenza vaccination was also associated with a significant increase in risk of sepsis (aRR: 1.96; 95% CI: 1.26 to 3.05).³⁸  However, after taking multiple comparisons into account using Bonferroni-corrected CIs, associations for gastrointestinal infections and sepsis were not significant and could have been due to chance (Table 5).^38,39 

In the 2 studies in which asthma was examined, asthma status was identified from a regional asthma database³⁹  or by using primary or secondary ICD diagnostic codes.^38,39  Walsh et al³⁹  reported a small, but significant, increase in asthma among children exposed to IIV in utero compared with unexposed children (aHR: 1.05; 95% CI: 1.02 to 1.09) (Table 5). However, results were no longer statistically significant after adjusting for multiplicity using Bonferroni correction (aHR: 1.05; 95% CI: 1.00 to 1.11).³⁹  Hviid et al³⁸  reported no significant difference in asthma for first trimester vaccination (aRR: 1.50; 95% CI: 0.99 to 2.29) or second or third trimester vaccination (aRR: 1.02; 95% CI: 0.89 to 1.16).

Estimates on rates of autism spectrum disorder were provided in 2 studies, of which 1 group evaluated pandemic influenza vaccine³⁹  and the other evaluated seasonal influenza vaccine.³⁷  For both studies, autism spectrum disorder was identified using ICD diagnostic codes. Zerbo et al³⁷  reported a slightly elevated risk of autism spectrum disorder after prenatal influenza vaccination in the first trimester (aHR: 1.20; 95% CI: 1.04 to 1.39) but not in the second or third trimester (Table 5). However, this association was no longer significant after adjusting for multiplicity using a Bonferroni correction (P = .1). Hviid et al³⁸  found no significant association between IIV exposure in utero and autism spectrum disorder. Additionally, Hviid et al³⁸  observed an increase of risk in Sjögren syndrome, only after prenatal influenza vaccination in the second or third trimester (aRR: 1.59; 95% CI: 1.04 to 2.44) but not in the first trimester. After adjusting for multiplicity using a Bonferroni correction, this association for Sjögren syndrome was no longer significant (aRR: 1.59; 95% CI: 0.82 to 3.11).

Across all studies, there were no other significant associations between prenatal influenza vaccination and celiac disease,³⁸  Crohn disease,³⁸  ulcerative colitis,³⁸  juvenile arthritis,³⁸  vasculitis,³⁸  reactive arthropathy,³⁸  idiopathic thrombocytopenic purpura,³⁸  idiopathic urticaria,³⁸  type-1 diabetes,³⁸  Bell palsy,³⁸  epilepsy,³⁸  Guillain-Barré syndrome,³⁸  intellectual disability,³⁸  neoplasms,³⁹  and sensory disorders (Table 5).³⁹ 

All-cause childhood morbidity outcomes included all-cause hospitalizations and urgent care services; researchers in 1 study examined a nonspecific outcome, including a complex of pediatric chronic conditions.³⁹  Walsh et al³⁹  found no significant associations between IIV exposure in utero and urgent and inpatient health services used³⁹  or the nonspecific complex of chronic conditions (Table 5).³⁹  Hviid et al³⁸  observed a statistically significant reduction in the risk of 1-year all-cause hospitalization (aHR: 0.94; 95% CI: 0.89 to 0.99), 3-year all-cause hospitalization (aHR: 0.95; 95% CI: 0.90 to 0.99), and 5-year all-cause hospitalization (aHR: 0.95; 95% CI: 0.91 to 0.99) in children whose mothers were vaccinated in the second or third trimester.

Pediatric mortality was defined as death occurring from birth through 5 years of age³⁹  or from 7 days to 4.6 years of age.³⁴  No significant associations were identified in the 2 studies that examined mortality as an outcome. Although Walsh et al³⁹  reported a reduced association between pandemic influenza vaccine given during pregnancy and pediatric mortality (aHR: 0.83; 95% CI: 0.64 to 1.08), this result was not statistically significant (Table 5). Similarly, Ludvigsson et al³⁴  reported no significant association between pandemic influenza vaccine during pregnancy and pediatric mortality overall (aHR: 0.97; 95% CI: 0.69 to 1.36) or by trimester of vaccination (first trimester: aHR: 0.86; 95% CI: 0.51 to 1.47; second trimester: aHR: 1.10; 95% CI: 0.69 to 1.76; third trimester: aHR: 0.93; 95% CI: 0.54 to 1.60). Secondary analyses, using unvaccinated siblings as controls, also revealed no significant associations (overall: aHR: 0.78; 95% CI: 0.52 to 1.19; first trimester: aHR: 0.47; 95% CI: 0.22 to 1.01; second trimester: aHR: 1.44; 95% CI: 0.74 to 2.78; third trimester: aHR; 0.65; 95% CI: 0.30 to 1.39).³⁴ 

For observational studies, risk of bias scores ranged from 6 to 8 on the Newcastle–Ottawa scale, of which 4 studies were deemed to be at low risk of bias,^34,36,38,39  and 4 studies were deemed to be at moderate-high risk of bias (Tables 1–3).^33,35,37,40  Common causes of potential bias included inadequate representativeness of the exposed cohort (n = 3)^33,35,37  and inadequately described follow-up of exposed and nonexposed cohorts (n = 6). The 1 RCT included was deemed to be at high risk of bias because of the lack of blinding of outcome assessment (detection bias) and inadequate ascertainment of outcomes (reporting bias).⁴¹ 

Although there has been increasing interest in the potential impacts of IIV exposure in utero on later child health, we identified relatively few studies in which health outcomes through the age of 5 years have been evaluated. In this systematic review, we summarized results from 9 studies including information on over 750 000 children, 163 924 of whom were exposed to IIV in utero. Researchers in 2 studies suggested lower risk of upper respiratory tract infection, all-cause hospitalization, and gastrointestinal infection associated with pandemic IIV exposure in utero. While the data from some studies suggested a potential increase in the risk of asthma, sepsis, and Sjögren syndrome after exposure to pandemic IIV in utero and an increased risk of autism spectrum disorder after exposure to seasonal IIV in utero, after adjusting for multiple comparisons using Bonferroni-corrected CIs, these associations were no longer statistically significant.

Narrative synthesis of these few studies indicates there is limited evidence evaluating health outcomes through the age of 5 years after exposure to IIV in utero, particularly for seasonal IIVs. Existing studies assessed a range of outcomes and the majority examined exposure to pandemic influenza vaccines. Only 3 studies on seasonal influenza vaccines in children were identified, and these were prone to bias; additional high-quality studies are warranted. Furthermore, given seasonal IIV is recommended in all trimesters of pregnancy and fetal development varies by gestational age, outcomes should be assessed according to gestational age at vaccination. Only 3 studies assessed outcomes by trimester of vaccination, of which 1 study combined second and third trimesters. The number of vaccinated individuals, particularly in the first trimester, was small and consequently limited the statistical power to detect differences between the children in the exposed and unexposed group. Differences in study quality, exposure ascertainment, vaccine type (ie, pandemic or seasonal, adjuvanted or nonadjuvanted, monovalent or trivalent), and study outcomes and outcome ascertainment, made it difficult to compare results across studies and precluded statistical pooling of results. These differences may have also influenced the reliability of findings. For example, diagnoses recorded in medical records and not using a validated standardized clinical assessment could lead to potential outcome misclassification.

Importantly, while the potential confounding influence of maternal age, parity, and pre-existing medical conditions was accounted for in most studies, the study by van Santen et al³³  was the only one that factored for receipt of childhood immunizations as a potential confounding variable by censoring children after receiving their own influenza vaccine.³³  No study factored for receipt of other recommended childhood vaccines. Given that receipt of vaccines during pregnancy may be predictive of childhood vaccination as well as other health behaviors,⁴⁴  it is possible that residual bias in the observational studies identified may have influenced findings, especially the observed protective associations. Future studies should aim to account for childhood immunization status as a potential confounding variable.

Finally, there was little geographic variation in the geographic location of studies identified. All included studies were conducted in high-income countries in North America or Europe, and thus these findings may not generalize to children in low and middle-income countries, whose population demographics and risk profiles differ markedly. Research on the safety of influenza vaccination during pregnancy from low and middle-income countries would be useful for guiding future vaccine policies and programs in these settings.

Maternal influenza vaccination is a growing public health tool for improving the health of mothers and their infants. Despite long-standing recommendations, maternal influenza immunization policies were not widely adopted until the 2009 influenza A/H1N1 pandemic, and, as identified as part of this review, the long-term health effects of in utero exposure of influenza vaccines in children >6 months of age are underinvestigated. We identified relatively few studies in which early childhood health outcomes were evaluated in children aged >6 months of age in maternally vaccinated and unvaccinated children, and the same outcomes were rarely measured in the few existing studies. This made formal meta-analyses impractical. Although our review indicates that exposure to IIV in utero is not associated with adverse health outcomes in childhood, additional epidemiological studies of early childhood health outcomes after maternal influenza vaccination are still needed to address this gap. Future well-controlled research in which seasonal IIV administered in different trimesters and in different settings is considered is required to provide a stronger evidence-base for the long-term safety of maternal influenza vaccination. Notwithstanding the need for additional research in this area, the results of the studies to-date have not revealed any adverse effect of maternal influenza vaccination on childhood health outcomes during the first 5 years. These findings are reassuring and can help support health care providers and pregnant women in making decisions about influenza vaccination during pregnancy.

The authors would like to acknowledge contributions made by Ms Diana Blackwood, a research librarian of the Faculty of Health Sciences, who provided guidance with the literature search strategy.

A/H1N1

influenza A virus subtype H1N1

aHR

adjusted hazard ratio

aIRR

adjusted incidence rate ratio

aRR

adjusted rate ratio

CI

confidence interval

IIV

inactivated influenza vaccine

ICD

International Classification of Diseases

RCT

randomized controlled trial

VE

vaccine effectiveness