Study cohort and design

This retrospective, population-based cohort study included all singleton, live-born children born in Western Australia (WA) between April 1, 2012 and July 1, 2016 and their mothers, as identified from birth registrations (Fig 1). Mother–child pairs were probabilistically linked with other population-based administrative health datasets by the WA Data Linkage Branch [14], including the Midwives Notification System (MNS) [15], the WA Antenatal Vaccination Database (WAAVD) [16], Hospital Morbidity Data Collection (HMDC) [17], Emergency Department Data Collection (EDDC) [18], and death registrations. The probabilistic linkage matches records from different sources using complex nonunique identifiers or field matching algorithms [19]. These algorithms compare common fields, such as given name, surname, date of birth, and other relevant fields (dependent on the contents and context of the dataset) [20] and provides a similarity weighting that is positively associated with the likelihood that 2 or more records belong to the same individual [19]. Clerical review is required to assess potential nonmatched records; this process has been shown to reduce the error rate of matching to less than 0.1% [21].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram of study participants included in the cohort.

WA, Western Australia.

https://doi.org/10.1371/journal.pmed.1003963.g001

The MNS is a legally mandated perinatal data collection of all children born at least 20 weeks’ gestation or birth weight of ≥400g (where gestational age is unknown). These data collection includes maternal demographic and health information, obstetric history, date of delivery, gestational age, and birth weight. Death registrations include the date and cause of all registered deaths in the state. This study received approval and a waiver or consent from the Department of Health WA Human Research Ethics Committee (RA#2016.56) and the Curtin University Human Research Ethics Committee (RA#20217–0808). This study is part of a larger linked cohort study and was guided by the study protocol described elsewhere [22]. This study is reported according to the REporting of studies Conducted using Observational Routinely-collected Data (RECORD) guideline (S1 RECORD Checklist) [23].

Maternal influenza vaccination

Data on influenza vaccination during pregnancy were obtained from the WAAVD, a statewide database, managed by the WA Department of Health, which includes information on the date of vaccination, vaccine brand, and batch number, and the estimated gestation at which vaccinations were administered as reported by their healthcare provider. In Australia, seasonal IIV during pregnancy has been recommended by the Australian Government since March 2000 and funded under the National Immunisation Program since January 2010 [4,24].

Children of mothers who had a record of receipt of seasonal IIV during pregnancy were considered “maternally vaccinated.” We estimated the gestational age at vaccination from the difference of completed weeks of gestation between the estimated date of conception and date of vaccination. Trimester at vaccination was categorized as first trimester (0 to ≤13 weeks gestation), second trimester (14 to ≤27 weeks gestation), and third trimester (≥28 weeks gestation). As antibody transfer may have been suboptimal when administered <2 weeks prior to birth [25], we excluded children of mothers that received an influenza vaccine <2 weeks prior to birth (i.e., “indeterminate” vaccination status).

Outcomes

Outcomes were identified from hospital and emergency department (ED) records [17,18]. The HMDC records all hospital inpatient admissions in the state’s public and private hospitals, including the date of admission and separation, up to 21 diagnosis codes (classified according to International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification [ICD-10-AM] codes) [17]. The EDDC includes all ED presentations in the state’s metropolitan public and private hospitals, including the date of presentation, 1 ICD-10-AM diagnosis code, and 1 symptom code. Outcomes evaluated in this study included (a) allergic or autoimmune diseases; (b) allergic diseases; and (c) autoimmune diseases, as identified by ICD-10-AM diagnostic codes in primary or additional diagnosis fields (S1 Table). Within allergic diseases, we separately evaluated asthma or anaphylaxis as health outcomes. To evaluate the possible influence of bias in the study results, we included presentations for all-cause injuries as a negative control outcome, since negative controls can be used to identify noncausal associations between prenatal exposure to seasonal IIV and the outcomes (S1 Table) [26].

Covariates

Maternal characteristics included age at the time of her child’s birth (≤19, 20 to 24, 25 to 29, 30 to 34, and ≥35 years), Aboriginal and/or Torres Strait Islander status (hereafter respectfully referred to as Aboriginal), socioeconomic status (quintiles between 1 (most disadvantaged) and 5 (least disadvantaged)) [27], body mass index, parity (primiparous, 1 prior birth, and ≥2 prior births), preexisting medical conditions (asthma, essential hypertension, and preexisting diabetes), pregnancy complications (gestational diabetes, gestational hypertension, and preeclampsia), smoking during pregnancy, and gestational age at the first prenatal care visit. Socioeconomic status was based on the Socioeconomic Index for Areas Index of Relative Socioeconomic Advantage and Disadvantage, an area-based index of relative access to resources for households within the same census collection district [27]. The referent group for socioeconomic status was the least disadvantaged quintile (i.e., quintile 5).

Child characteristics considered as covariates included the child’s Aboriginal status, year and season of birth, preterm birth, and small-for-gestational age birth status. All covariates were selected a priori for their association with maternal influenza vaccination [28–33]. Preterm birth was categorized as moderate to late preterm (32 to <37 weeks gestation), very preterm (28 to <32 weeks gestation), and extremely preterm (<28 weeks gestation) birth. Small-for-gestational age birth was defined as birth weight <10th percentile according to Australian national birth weight percentiles by sex and gestational age [34].

Participant involvement

No participants were involved in setting the research question or the exposure or outcome measures nor were they involved in the design and implementation of the study. There are no plans to directly involve study participants in the dissemination of the research findings.

Statistical analyses

We compared the demographic and health characteristics between vaccinated and unvaccinated mothers and their children using univariate logistic regression models. Based on the predicted probability of vaccination from multivariate logistic regression, we estimated the inverse probability of treatment (vaccination) weights (IPTWs) to control for baseline probability of vaccination in further analyses. The multivariate model included the maternal covariates: age, Aboriginal status, socioeconomic status, body mass index, parity, preexisting medical conditions, pregnancy complications, smoking during pregnancy, and gestational age at the time of first prenatal care visit. To assess the balance of maternal covariates between maternally vaccinated and unvaccinated children, we calculated the standardized mean differences of each covariate. IPTWs were applied to Cox proportional hazards regression models to estimate unadjusted and adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) for each of the study outcomes. Adjusted models additionally controlled for the child’s Aboriginal status and year and season of birth.

Children were followed up from the date of birth and were censored at the earliest of (a) the date the child reached 5 years of age; (b) the last date of available data provided by the WA Data Linkage Branch (i.e., July 1, 2017); (c) the date the child died; or (d) the date of the event. Subgroup analyses compared the risk of study outcomes by trimester of maternal vaccination. To assess the sensitivity of our definition of childhood asthma, we restricted the definition of asthma to the presence of a diagnosis code for asthma alone (i.e., J45 and J46) and did not include early symptoms of asthma (i.e., wheezing). To evaluate associations with more severe clinical outcomes, additional analyses defined outcomes based on hospital inpatient admission data only. To evaluate the importance of timing of maternal vaccination and duration of exposure to maternal antibodies, stratified analyses were planned a priori and performed (1) by trimester of vaccination and additional sensitivity analysis stratifying (2) by preterm birth status, to see if results differed by length of gestation. All analyses were performed in Stata version 15.1 (Stata, College Station, Texas, United States of America).

As per recommendations during the peer review process, additional analyses were conducted. To evaluate the potential competing risk of all-cause mortality, the mortality rate of maternally vaccinated and unvaccinated children was calculated. As the majority of studies in maternal influenza vaccination and early childhood health outcomes have focused on the first 6 months of life, we performed a sensitivity analysis restricting to children aged between 6 months and 5 years. As the study cohort includes siblings, this may imply statistical dependence of observations due to genetic factors and/or shared environment; therefore, we performed an additional sensitivity analysis restricting to 1 randomly selected child per mother. To further evaluate the potential influence of seasonality on our findings, we performed an additional sensitivity analysis that matched children by calendar year and month of birth, allowing the assessment of outcomes comparing children born during similar time periods. These models incorporated year and month of birth as stratum and removed year and season of birth from propensity score estimates.

Maternal influenza vaccination

Among the 124,760 children, 14,396 (11.5%) were maternally vaccinated: 2,785 (19.4%) exposed during the first trimester, 5,558 (38.6%) during the second trimester, and 6,053 (42.1%) during the third trimester. Vaccine exposure varied from 6.1% for births in 2012 to 12.0% in 2015. Most mothers received their influenza vaccine between March and July (n = 12,537; 87.1%) of any given year. Primiparous women, women with preexisting medical conditions and pregnancy complications, and women who were nonsmokers during pregnancy were more likely to have been vaccinated (Table 1). Vaccination was less common among women of the lowest socioeconomic status compared to women of the highest socioeconomic status (odds ratio [OR], 0.93; 95% CI, 0.88 to 0.98) and mothers of preterm infants compared to term infants (OR, 0.91; 95% CI, 0.85 to 0.98), and vaccination was more common among women who birthed during winter compared to summer (OR, 2.60; 95% CI, 2.46 to 2.74). After applying IPTW, standardized differences of maternal characteristics were balanced between maternally vaccinated and unvaccinated children (S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Odds of seasonal influenza vaccination by maternal and child characteristics for children born in WA between April 1, 2012 and July 1, 2016.

https://doi.org/10.1371/journal.pmed.1003963.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Risk of allergic or autoimmune diseases associated with prenatal exposure to seasonal inactivated influenza vaccine among children <5 years of age, by trimester of prenatal vaccination.

https://doi.org/10.1371/journal.pmed.1003963.t002

Allergic or autoimmune diseases

During the study period, we identified 8,568 (6.9%) children with a diagnosis of an allergic and/or autoimmune disease. There was no significant difference in the risk of allergic or autoimmune diseases among maternally vaccinated children compared to maternally unvaccinated children (aHR, 1.02; 95% CI, 0.95 to 1.09) (Table 2). We identified no associations after stratifying by preterm birth status (S2 and S3 Tables) or when restricting to hospital inpatient admissions (S4 Table), children aged between 6 months and 5 years (S5 Table), or 1 random child per mother (S6 Table).

Allergic diseases

There were 8,417 children with a diagnosis of an allergic disease. The most common outcomes included asthma (n = 3,538; 42.0%), urticaria (n = 2,805; 33.3%), and anaphylaxis (n = 987; 11.7%). Maternal influenza vaccination was not associated with presentation to ED or hospitalization for an allergic disease among children aged <5 years (aHR, 1.02; 95% CI, 0.95 to 1.10) (Table 2). We observed similar results when stratifying by preterm birth status (S2 and S3 Tables) and when restricting to hospital admissions only (S4 Table), children aged between 6 months and 5 years (S5 Table), or 1 random child per mother (S6 Table).

Asthma

A total of 3,772 children had an episode of care for asthma: 1,020 (27.1%) diagnoses of asthma and 2,752 (73.3%) diagnoses of wheezing were identified. We observed no difference in the risk of asthma among maternally vaccinated children compared to maternally unvaccinated children (aHR, 1.00; 95% CI, 0.89 to 1.12) (Table 2).

In the sensitivity analysis restricting the definition of asthma to the presence of an asthma diagnosis code alone, we observed a negative association of asthma following vaccination during the third trimester (aHR, 0.70; 95% CI, 0.50 to 0.97) (Table 2). This was mostly attributed to a negative association observed among term children following vaccination during the third trimester (aHR, 0.68; 95% CI, 0.48 to 0.96) (S2 Table). When considering year and month of birth as a stratum, this negative association between asthma and vaccination during the third trimester was no longer significant (S7 Table). No other associations were observed following vaccination during earlier trimesters (Table 2) or after stratifying by preterm birth status (S2 and S3 Tables) or when restricting to hospital admissions only (S4 Table), children aged between 6 months and 5 years (S5 Table), or 1 random child per mother (S6 Table).

Anaphylaxis

In total, there were 1,157 children diagnosed with an episode of anaphylaxis. We observed a negative association of anaphylaxis associated with exposure to seasonal IIV administered during the third trimester (aHR, 0.67; 95% CI, 0.47 to 0.95). This was attributed to a negative association observed among term children following vaccination during the third trimester (aHR, 0.62; 95% CI, 0.43 to 0.91) (S2 Table). The negative association between anaphylaxis and vaccination during the third trimester remained even after treating year and month of birth as a stratum (aHR, 0.69; 95% CI, 0.49 to 0.97) (S7 Table). No other associations were observed following vaccination during earlier trimesters (Table 2) or after stratifying by preterm birth status (S2 and S3 Tables) or when restricting to hospital admissions only (S4 Table), children aged between 6 months and 5 years (S5 Table), or 1 random child per mother (S6 Table).

Autoimmune diseases

There were 174 children with a diagnosis of an autoimmune disease, including 16 (9.2%) who were maternally vaccinated. The most common outcomes included diabetes mellitus (n = 54; 31.0%), coeliac disease (n = 40; 23.0%), idiopathic thrombocytopenic purpura (n = 33; 19.0%), and juvenile arthritis (n = 21; 12.1%). There was no significant difference in the risk of autoimmune diseases between maternally vaccinated and unvaccinated children (aHR, 0.93; 95% CI, 0.55 to 1.59) (Table 2) nor any differences detected by trimester of vaccination (Table 2), after stratifying by preterm birth status (S2 and S3 Tables), or when restricting to hospital admissions only (S4 Table), children aged between 6 months and 5 years (S5 Table), or 1 random child per mother (S6 Table).

Negative control

A total of 21,730 children aged <5 years presented to ED or were hospitalized with an injury. We observed no difference in the risk of all-cause injuries between maternally vaccinated and unvaccinated children (aHR, 1.04; 95% CI, 0.99 to 1.11) (S8 Table).