aTiming imputed for 19% of vaccinated group; No. of events and incidence rates in that group based on first imputed data set; adjusted estimates based on 10 imputed data sets. bAdjusted using stabilized inverse probability of treatment weights. cNo. of events = total children diagnosed with outcome. Point estimates are HRs from a Cox proportional hazards model. dSee Table 2. eNo. of events = total occurrences for each outcome. Point estimates are incidence rate ratios from a negative binomial model. fAll respiratory tract infections, gastrointestinal infections, and otitis media. gData suppressed (n < 5).
eFigure 1. Study Design
eFigure 2. Study Flow Diagram
eFigure 3. Sensitivity Analyses
eTable 1. Influenza Vaccination Exposure Assessment
eTable 2. Individual ICD-10-CA Diagnostic Codes to Identify Outcomes
eTable 3. Description of “Obstetric Conditions Affecting Pregnancy” Covariate
Customize your JAMA Network experience by selecting one or more topics from the list below.
Mehrabadi A, Dodds L, MacDonald NE, et al. Association of Maternal Influenza Vaccination During Pregnancy With Early Childhood Health Outcomes. JAMA. 2021;325(22):2285–2293. doi:10.1001/jama.2021.6778
Is seasonal influenza vaccination in pregnancy associated with adverse childhood health outcomes?
In this population-based retrospective cohort study that used inverse probability of treatment weighting and included 28 255 children with a mean duration of follow-up of 3.6 years, maternal influenza vaccination during pregnancy was not significantly associated with the risk of immune-related health outcomes (eg, asthma, infections) non–immune-related outcomes (eg, neoplasms, sensory impairment), or nonspecific health outcomes (eg, emergency visits, hospitalizations).
Seasonal influenza vaccination in pregnancy was not significantly associated with adverse early childhood health outcomes among offspring.
Seasonal influenza vaccination in pregnancy can reduce influenza illness among pregnant women and newborns. Evidence is limited on whether seasonal influenza vaccination in pregnancy is associated with adverse childhood health outcomes.
To assess the association between maternal influenza vaccination during pregnancy and early childhood health outcomes.
Design, Setting, and Participants
Retrospective cohort study, using a birth registry linked with health administrative data. All live births in Nova Scotia, Canada, between October 1, 2010, and March 31, 2014, were included, with follow-up until March 31, 2016. Adjusted hazard ratios (HRs) and incidence rate ratios (IRRs) with 95% confidence intervals were estimated while controlling for maternal medical history and other potential confounders using inverse probability of treatment weighting.
Seasonal influenza vaccination during pregnancy.
Main Outcomes and Measures
Childhood outcomes studied were immune-related (eg, asthma, infections), non–immune-related (eg, neoplasms, sensory impairment), and nonspecific (eg, urgent or inpatient health care utilization), measured from emergency department and hospitalization databases.
Among 28 255 children (49% female, 92% born at ≥37 weeks’ gestation), 10 227 (36.2%) were born to women who received seasonal influenza vaccination during pregnancy. During a mean follow-up of 3.6 years, there was no significant association between maternal influenza vaccination and childhood asthma (incidence rate, 3.0 vs 2.5 per 1000 person-years; difference, 0.53 per 1000 person-years [95% CI, −0.15 to 1.21]; adjusted HR, 1.22 [95% CI, 0.94 to 1.59]), neoplasms (0.32 vs 0.26 per 1000 person-years; difference, 0.06 per 1000 person-years [95% CI, −0.16 to 0.28]; adjusted HR, 1.26 [95% CI, 0.57 to 2.78]), or sensory impairment (0.80 vs 0.97 per 1000 person-years; difference, −0.17 per 1000 person-years [95% CI, −0.54 to 0.21]; adjusted HR, 0.82 [95% CI, 0.49 to 1.37]). Maternal influenza vaccination in pregnancy was not significantly associated with infections in early childhood (incidence rate, 184.6 vs 179.1 per 1000 person-years; difference, 5.44 per 1000 person-years [95% CI, 0.01 to 10.9]; adjusted IRR, 1.07 [95% CI, 0.99 to 1.15]) or with urgent and inpatient health services utilization (511.7 vs 477.8 per 1000 person-years; difference, 33.9 per 1000 person-years [95% CI, 24.9 to 42.9]; adjusted IRR, 1.05 [95% CI, 0.99 to 1.16]).
Conclusions and Relevance
In this population-based cohort study with mean follow-up duration of 3.6 years, maternal influenza vaccination during pregnancy was not significantly associated with an increased risk of adverse early childhood health outcomes.
Seasonal influenza affects approximately 5% to 10% of adults and 20% to 30% of children around the world annually.1 Pregnant women are not more susceptible to acquiring influenza infection but, once infected, are at an increased risk of severe influenza illness and related complications.2-4 Maternal influenza infection, particularly during the 2009 influenza A(H1N1) virus ([H1N1]pdm09) pandemic, has also been associated with an increased risk of fetal death and preterm birth.5,6
Quiz Ref IDSeasonal influenza vaccination during pregnancy is an important public health intervention aimed at preventing maternal and neonatal influenza illness.7 Despite benefits of seasonal influenza vaccination in pregnancy, including newborn protection in the first months after birth,8 uptake remains low in many settings.9 Safety concerns are reported as a leading reason women do not receive influenza vaccination in pregnancy.10 Observational studies have not found an association between influenza vaccination during pregnancy and adverse fetal or perinatal outcomes11,12; however, few studies have investigated later childhood health outcomes,13 and this remains an evidence gap.14 The fetus may be sensitive to maternal immune responses during critical periods of development; vaccine exposure in utero may, therefore, be associated with later immune-related or other health outcomes in offspring.15 The purpose of this study was to assess the association between seasonal influenza vaccination in pregnancy and long-term immune-related or other childhood health outcomes.
Ethics approval was obtained from the IWK Health Centre research ethics board, which granted an exemption for informed consent. The study was further approved by the Joint Data Access and Health Data Nova Scotia access committees.
We conducted a population-based retrospective cohort study using the Nova Scotia Atlee Perinatal Database linked with hospitalization and physician billing records. The Atlee Database contains demographic information, prenatal risk factors, diagnoses, and procedures for every birth 500 g or greater and 20 weeks’ gestation or longer in Nova Scotia, Canada. Trained personnel abstract information from antenatal and medical charts, and a data quality program ensures data quality.16,17 The hospitalization database includes all emergency department and inpatient health care contacts for the province, recorded using the International Classification of Diseases, Tenth Revision, Canadian modification (ICD-10-CA) coding system. There are approximately 10 000 live births per year in Nova Scotia within a publicly funded universal health care system, ensuring data capture for all residents.
Quiz Ref IDThe study population consisted of all live births from October 1, 2010, to March 31, 2014, whose records linked with health administrative databases. Follow-up began on each child’s date of birth and continued until either death, migration out of the province, or March 31, 2016 (end of the study period), whichever occurred first. For analysis of time-to-event outcomes (ie, asthma, neoplasms, sensory impairment), the end of follow-up was the event date for those experiencing the outcome. A graphical depiction of the study design is presented in eFigure 1 in the Supplement.
Seasonal influenza vaccination in pregnancy was ascertained from the Atlee and physician billing databases (eTable 1 in the Supplement). The Atlee Database began capturing influenza vaccination during pregnancy in the 2010-2011 influenza season. This information is collected by clinicians during routine prenatal care, based on either patient self-report (eg, if the vaccine was received elsewhere between prenatal care visits) or on vaccination at the prenatal clinic. The physician billing database records fee codes for influenza vaccination administered during an office visit. We ascertained vaccination status predominantly through the physician billing data (81% of the vaccinated group). The influenza vaccines used during the study period were unadjuvanted trivalent inactivated influenza vaccines, consisting of 2 type A strains and 1 type B strain; no product-specific information was available in the study databases.
Although outcomes proximal to birth are assessed according to internationally accepted standards following immunization during pregnancy,18,19 no guidance exists for monitoring later pediatric health outcomes. We prespecified immune-related outcomes (asthma, infections), as the fetal immune system may be sensitive to maternal immunization, with potential for beneficial or deleterious associations later in life through immunologic memory.15,20 We additionally included non–immune-related (neoplasms, sensory impairment) and nonspecific outcomes (all-cause emergency department visits and hospitalizations) that have been the focus of other maternal immunization studies.21,22
Exposure and covariate assessment were based on maternal records, while childhood outcomes were ascertained from children’s records (eFigure 1 in the Supplement). We measured outcomes using ICD-10-CA diagnostic codes from emergency department visits and hospitalizations (eTable 2 in the Supplement). Quiz Ref IDMeasurement of pediatric asthma was based on a modification of a previously validated algorithm (sensitivity, 89%; specificity, 72%) and consisted of at least 2 emergency department visits or 1 hospitalization for asthma, at age 6 months or older.23 Other outcomes were based on 1 or more diagnoses received during an emergency department visit or a hospitalization. Infectious outcomes included upper and lower respiratory tract infections, gastrointestinal infections, and otitis media, as well as a composite comprised of the total of these infections. Neoplasm was defined as any abnormal growth of body cells or tissues, and sensory impairment included hearing or vision loss. We prespecified a negative control outcome (all-cause injuries) to help detect any biases or spurious associations; this outcome shares similar potential sources of bias as other study outcomes but is not plausibly related to influenza vaccination during pregnancy.24
A broad range of potential confounders and their proxies were included in propensity scores to account for confounding. Included were maternal age at childbirth, parity, maternal smoking, multifetal gestation, preexisting maternal conditions (heart disease, pulmonary disease, asthma, renal disease, psychiatric illness), rural residence, neighborhood-level mean family income, gestational age at birth in weeks, infant sex, season and year of conception, type of clinician at delivery (proxy for the clinician that provided antenatal care), and district health authority.
Distributions of baseline variables, stratified by influenza vaccination during pregnancy, were summarized using frequency distributions and compared using standardized differences, for which an absolute standardized difference less than 0.1 was considered to indicate a well-balanced covariate.25 A propensity score for each child was developed using logistic regression to estimate the predicted probability of maternal influenza vaccination during pregnancy, conditional on all covariates listed in Table 1. The propensity scores were generated from 10 imputed data sets to account for less than 1% missing data; all covariates shown in Table 1 were included in imputation models using chained equations.26 We calculated inverse probability of treatment weights as the inverse of the propensity score for children exposed to maternal influenza vaccination during pregnancy (exposed group) and the inverse of 1 minus the propensity score for the unexposed children. We stabilized the weights to account for extreme values and reassessed the balance of covariates after weighting using standardized differences.25,27
Unadjusted incidence rates per 1000 person-years and crude rate differences with 95% confidence intervals, comparing vaccinated and unvaccinated groups, were computed for all outcomes. Cox proportional hazards models incorporating the inverse probability of treatment weights from each of the 10 imputed data sets were used to model time-to-event outcomes (asthma, neoplasms, sensory impairment), with the underlying time axis as the time in days since birth. Proportional hazards assumptions were met based on examining Schoenfeld residual plots and Wald tests for interactions between exposure status and time. To account for potential data overdispersion, negative binomial regression models, incorporating the inverse probability of treatment weights from each imputation data set, were used for outcomes analyzed as rates (infections, urgent and inpatient health services utilization, all-cause injuries). Each model included the total number of occurrences of each outcome as the dependent variable, with an offset of the log of person-years. Adjusted point estimates were combined across imputed data sets to produce a single estimate. We included a robust variance estimator in all models to account for correlation among siblings, and precision around point estimates was provided using 2-sided 95% CIs. Because of the potential for type I error due to multiple comparisons, study findings should be interpreted as exploratory.
We performed post hoc subgroup analyses by trimester of vaccination; information on timing of vaccination was obtained from the physician billing data for 81% of the exposed group (8272/10 227 vaccinated) and trimester of vaccination was imputed for the remaining 19% (summary of the imputation approach is provided in the eMethods in the Supplement). Several sensitivity analyses were conducted to assess the robustness of the findings. We added maternal asthma as an additional covariate in weighted outcome models and, separately, gestational age in days to assess whether results were sensitive to using a continuous measure; we applied 3 types of trimming to the inverse probability of treatment weights (excluding those lying outside of the 0.01 to 99.9th percentiles, 0.05 to 99.5th percentiles, and 1st to 99th percentiles); and, in separate models, we adjusted for maternal obstetric conditions affecting pregnancy (described in eTable 3 in the Supplement) and maternal influenza illness during pregnancy. Analyses were conducted using SAS version 9.4 (SAS Institute Inc).
Among 28 356 eligible live births, 95 (0.32%) were excluded because of an infant death occurring on the date of birth and 6 (0.02%) because of missing parity or rural residence information, leaving 28 255 children (eFigure 2 in the Supplement). Almost all children (93.4%) were followed up to the end of the study period, ranging from a minimum of 2 years to a maximum of 5.5 years (mean, 3.6 years [SD, 1.1 years]); 69 children (0.2%) were censored because of death. A total of 10 227 children (36.2%) were exposed to influenza vaccination in utero; these children were more likely than unexposed children to be born to mothers who were 35 years or older (17.8% vs 16.6%), nulliparous (48.3% vs 44.4%), had a preexisting medical condition (eg, pulmonary disease, 8.3% vs 6.5%; asthma, 8.0% vs 6.2%), and who had a family physician attending their delivery (52.9% vs 40.3%). Children born to vaccinated mothers were less likely to be born to current or former smokers (16.2% vs 21.6%) and less likely to be born within the lowest neighborhood mean family income quintile (17.6% vs 21.3%) compared with the unexposed group (Table 1). After propensity score weighting, baseline characteristics of the study population were well-balanced across exposure groups and all standardized differences were less than 0.1.
There was no significant association between influenza vaccination during pregnancy and pediatric asthma (incidence rate, 3.0 vs 2.5 per 1000 person-years; difference, 0.53 per 1000 person-years [95% CI, −0.15 to 1.21]; adjusted hazard ratio [HR], 1.22 [95% CI, 0.94 to 1.59]), neoplasms (0.32 vs 0.26 per 1000 person-years; difference, 0.06 per 1000 person-years [95% CI, −0.16 to 0.28]; adjusted HR, 1.26 [95% CI, 0.57 to 2.78]), or sensory impairment (0.80 vs 0.97 per 1000 person-years; difference, −0.17 per 1000 person-years [95% CI, −0.54 to 0.21]; adjusted HR, 0.82 [95% CI, 0.49 to 1.37]) (Table 2).
After adjustment for confounding, exposure to influenza vaccination in utero was not significantly associated with overall infections (incidence rate, 184.6 vs 179.1 per 1000 person-years; difference, 5.44 per 1000 person-years [95% CI, 0.01 to 10.9]; adjusted incidence rate ratio [IRR], 1.07 [95% CI, 0.99 to 1.15]), urgent and inpatient health services utilization (511.7 vs 477.8 per 1000 person-years; difference, 33.9 per 1000 person-years [95% CI, 24.9 to 42.9]; adjusted IRR, 1.05 [95% CI, 0.99 to 1.16]), or with the negative control outcome (70.6 vs 63.4 per 1000 person-years; difference, 7.16 per 1000 person-years [95% CI, 3.84 to 10.5]; adjusted IRR, 1.05 [95% CI, 0.97 to 1.13]).
There was no statistically significant difference in crude incidence rates of upper and lower respiratory tract infections and otitis media between vaccine-exposed and unexposed children (Table 2). Crude rates of gastrointestinal infections were 30.2 vs 28.0 per 1000 person-years among children born to vaccinated vs unvaccinated mothers (crude rate difference, 2.22 per 1000 person-years [95% CI, 0.04 to 4.40]; however, there was no statistically significant association (adjusted IRR, 1.04 [95% CI, 0.94 to 1.15]).
Overall, results in the post hoc analyses by trimester of vaccination were not qualitatively different from the primary analyses (Figure). Although the magnitude and direction of the point estimate were similar to the primary analysis, third-trimester influenza vaccination was significantly associated with lower respiratory tract infection in this post hoc subgroup analysis (adjusted IRR, 1.19 [95% CI, 1.01 to 1.40]). The results were robust in sensitivity analyses (eFigure 3 in the Supplement).
Quiz Ref IDThe results of this population-based study suggest that maternal influenza vaccination during pregnancy was not associated with immune-related adverse health outcomes (eg, asthma, infections), non–immune-related health outcomes (eg, neoplasms, sensory impairment), or nonspecific health services utilization (eg, emergency department visits and hospitalizations) in young children. These results remained largely consistent in trimester-specific analyses and in sensitivity analyses that considered possible residual confounding (eg, from maternal asthma or influenza diagnoses), changes in the gestational age definition, and implementation of trimming methods to account for possible influential inverse probability of treatment weights. A small positive association with the negative control outcome in unadjusted analyses was attenuated after propensity score weighting, implying that the models accounted for confounding. Previous studies on this topic have largely focused on exposure to influenza A(H1N1)pdm09 vaccination in pregnancy13; this study, therefore, makes an important contribution to long-term evaluation of seasonal influenza vaccination during pregnancy.
Quiz Ref IDEfficacy of trivalent inactivated influenza vaccination in pregnancy for preventing influenza infection in mothers and their infants younger than 6 months has been demonstrated in several randomized clinical trials.28-31 Among studies that have evaluated childhood health outcomes after age 6 months, 3 investigated seasonal influenza vaccine and a limited number of childhood outcomes, while 7 investigated influenza A(H1N1)pdm09 vaccine and a wider range of outcomes.13,32 Overall, these studies did not identify any concerning associations between influenza vaccination in pregnancy and adverse childhood health outcomes.13 For seasonal influenza vaccines, no significant association was reported with autism spectrum disorder (median follow-up of 8.3 years; adjusted HR, 1.10 [95% CI, 1.00 to 1.21]).33 No significant association was reported between seasonal influenza vaccination in pregnancy and lab-confirmed influenza hospitalization among children aged 6 to 12 months.34 Acute otitis media and medically attended acute respiratory infections in the first year of life were inversely associated with seasonal influenza vaccination during pregnancy.35
For influenza A(H1N1)pdm09 vaccine, no significant associations were reported between vaccination in pregnancy and many infection-related outcomes in the first year of life, such as influenza and pneumonia,36 infection-related primary care visits,37 or any infections.38 Previously reported inverse associations between influenza A(H1N1)pdm09 vaccination and rates of gastrointestinal infections up to 5 years of age (adjusted rate ratio, 0.84 [95% CI, 0.74 to 0.94]21; adjusted HR, 0.94 [95% CI, 0.91 to 0.98]22) were not observed in this study of seasonal influenza vaccine. Although the reason for the inverse association with gastrointestinal infections in previous studies is unclear, both studies speculated there may have been residual confounding by rotavirus vaccination that could not be adjusted for. This was likely not a problem in the current study because of lack of provincial health care coverage for the rotavirus vaccine during the study period.
A Danish cohort study, with a mean of 4.6 years of follow-up, found no significant association between influenza A(H1N1)pdm09 vaccine exposure during the first trimester (adjusted rate ratio, 1.50 [95% CI, 0.99 to 2.29]) or second or third trimesters (adjusted rate ratio, 1.02 [95% CI, 0.89 to 1.16]) and childhood asthma.21 A Canadian cohort study with a median length of follow-up of 5 years reported a weak association between any influenza A(H1N1)pdm09 influenza vaccine exposure and childhood asthma (adjusted HR, 1.05 [95% CI, 1.02 to 1.09]); however, the magnitude was small and may have been a result of residual confounding, as there was also a weak association with the negative control outcome in that study.22 The Canadian and Danish studies reported no significant association between influenza vaccination in pregnancy and a range of other immune-related child health outcomes (eg, upper or lower respiratory tract infections, otitis media, meningitis, sepsis, viral infections, any infections), and non–immune-related child health outcomes (eg, neoplasms, sensory disorders, epilepsy, intellectual disability, Bell palsy).21,22 In addition, a Swedish cohort study found no significant association between influenza A(H1N1)pdm09 vaccination and autism spectrum disorder (mean follow-up, 6.7 years; adjusted HR, 0.95 [95% CI, 0.81 to 1.12]).32
This study has several strengths. First, the study used a population-based design that included nearly all births in the province, which minimized selection bias and could therefore potentially generalize to other settings. Second, assessment of influenza vaccine exposure was based primarily on physician claims data (81%) but influenza vaccination reported in the perinatal database was also included, which likely improved exposure assessment as compared with previous research.9 Third, numerous sensitivity analyses yielded results similar to those from the primary analyses, supporting the robustness of the findings.
This study has several limitations. First, both physician billing information and the Atlee Database were used for exposure assessment, and the latter relied partly on patient self-report, which may be less reliable. However, of the 10 227 children exposed to influenza vaccination in utero, vaccination was recorded in the physician billing database 81% of the time and use of the Atlee Database to supplement exposure ascertainment had the advantage of capturing data on influenza vaccination in settings other than a physician’s office, such as workplaces or pharmacies. Second, outcome misclassification cannot be ruled out; the majority of study outcomes have not been validated, although an adapted validated algorithm was used for childhood asthma. Third, outcome measurement was limited to diagnoses made in emergency departments and during hospitalizations, although this approach reflected more serious illnesses and could potentially improve specificity compared with use of diagnoses from nonurgent settings.22,23 Fourth, residual confounding from unmeasured or unknown variables cannot be ruled out, although the lack of significant association with the negative control outcome provides additional reassurance for the validity of conclusions from the main analyses. Fifth, the point estimates were slightly increased for some outcomes (eg, adjusted HR of 1.22 for asthma and 1.26 for neoplasms), and the 95% CIs around these estimates indicate that some increase in risk cannot be ruled out. Given the observational nature of the study, the small events numbers for neoplasms, and the imprecise 95% CIs around some point estimates (especially for neoplasms and trimester-specific analyses), these results should be interpreted cautiously.
In this population-based cohort study with mean follow-up duration of 3.6 years, maternal influenza vaccination during pregnancy was not significantly associated with an increased risk of adverse early childhood health outcomes.
Corresponding Author: Deshayne B. Fell, PhD, Children’s Hospital of Eastern Ontario (CHEO) Research Institute, 401 Smyth Rd, Centre for Practice Changing Research, Room L-1154, Ottawa, ON K1H 8L1, Canada (firstname.lastname@example.org).
Accepted for Publication: April 15, 2021.
Author Contributions: Dr Mehrabadi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Mehrabadi, Dodds, MacDonald, Benchimol, Ortiz, Sprague, Walsh, Wilson, Fell.
Acquisition, analysis, or interpretation of data: Mehrabadi, Dodds, MacDonald, Top, Benchimol, Kwong, Ortiz, Fell.
Drafting of the manuscript: Mehrabadi, Fell.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Mehrabadi, Dodds, Fell.
Obtained funding: Dodds, Ortiz, Fell.
Administrative, technical, or material support: Dodds, Benchimol, Kwong, Sprague.
Supervision: Wilson, Fell.
Conflict of Interest Disclosures: Dr Top reported receiving consultancy fees from Pfizer and receiving grants from GlaxoSmithKline. Dr Benchimol reported receiving legal consulting fees from Hoffmann La-Roche Limited for matters unrelated to the influenza and other vaccine products. Dr Ortiz reported receiving consultancy fees from Pfizer, Foundation for Influenza, and Seqirus and receiving research support paid to his research unit from the National Institutes of Health, Pfizer, and GlaxoSmithKline. Dr Wilson is CEO of CANImmunize Inc, which hosts a national digital immunization record. No other authors reported disclosures.
Funding/Support: This study was funded by Canadian Institutes of Health Research operating grant AO1-151541.
Role of the Funder/Sponsor: The Canadian Institutes of Health Research had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The data (or portions of the data) used in this report were made available by Health Data Nova Scotia of Dalhousie University. Although this research is based on data obtained from the Nova Scotia Department of Health and Wellness, the observations and opinions expressed are those of the authors and do not represent those of either Health Data Nova Scotia or the Department of Health and Wellness.