Maternal Pertussis Vaccination, Infant Immunization, and Risk of Pertussis

The Links2HealthierBubs cohort is a population-based cohort of mother–infant dyads in 3 Australian jurisdictions: Northern Territory (NT), Queensland (QLD), and Western Australia (WA), representing one-third of all Australian births. The study protocol is published elsewhere.²⁰  In brief, probabilistic record linkage of maternal and infant administrative data sources was used to establish the population-based cohorts within each participating jurisdiction. All mothers and their infants with a gestational age ≥20 weeks and/or a birth weight ≥400 g who were born after the introduction of the jurisdictional maternal pertussis vaccination program were eligible for inclusion in the cohort (NT, WA: after April 1, 2015; QLD: after August 1, 2014).⁵ 

Data sources for measuring maternal dTpa vaccination, pertussis infection, and covariates were linked with the maternal–infant cohort, including perinatal, birth and death register, hospital inpatient, immunization, and notifiable disease data for each jurisdiction. With exception to immunization data, these jurisdictional databases each feed into national data collections and apply standardized data collection procedures.

Perinatal data collections are statutorily mandated databases summarizing information related to all births in each jurisdiction.²¹  We estimated the year and month of conception from the infant’s date of birth minus the best clinical estimate of gestational age in days. Socioeconomic status was assigned based on the mother’s residence at birth using the Index of Relative Socioeconomic Advantage and Disadvantage.²² 

During the study period, maternal pertussis vaccination was recommended during every pregnancy between 28 and 32 weeks of gestational age.⁵  Maternal vaccination information was obtained from jurisdictional immunization registers and databases summarizing provider-reported immunization data (2014–2016) in addition to perinatal data collections (≥2016). These data sources were used to determine dTpa vaccination status (yes/no) and the gestation at vaccination during pregnancy (<28 weeks, 28–31 weeks, ≥32 weeks). We also extracted information on influenza vaccination status during pregnancy (yes/no) and whether influenza vaccine was coadministered with dTpa or at a separate appointment. In NT and QLD, where infant immunization data were available, we additionally extracted information on pertussis-containing vaccines (diphtheria-tetanus-acellular pertussis [DTaP]) administered to the child up to 18 months of age.²³  Under the National Immunization Program, children in Australia receive their first dose of hexavalent pertussis–containing vaccine (diphtheria-tetanus-pertussis-hepatitis B-polio-Haemophilus influenzae type b vaccine [Infanrix hexa]) at 2 months of age, their second dose at 4 months of age, and their third dose at 6 months of age.²⁴  A booster dose of diphtheria-tetanus-pertussis vaccine (Infanrix or Tripacel) is recommended at 18 months of age.²⁴  We used the date of vaccination to assign the infant’s age in months at the time of vaccination with the first, second, and third doses of the primary series of DTaP vaccine.

In Australia, pertussis is a nationally notifiable disease, and information on medically notified or laboratory-diagnosed pertussis cases are collected and summarized in jurisdictional notifiable disease databases.²⁵  The criteria for notification includes a combination of clinical, epidemiologic, or laboratory evidence, with laboratory definitive evidence (ie, isolation of Bordetella pertussis in cell culture, detection by nucleic acid testing, or seroconversion) serving as a confirmed pertussis case²⁶ ; ≥94% of infant pertussis cases are notified following detection by polymerase chain reaction.²⁷  We identified notified pertussis infections and diagnosis date for infants through age 18 months. Pertussis cases included confirmed cases with definitive laboratory evidence or suggestive laboratory evidence coinciding with clinical evidence and probable cases with clinical and epidemiologic evidence.²⁶ 

Hospital inpatient records and death registrations were used to evaluate pertussis severity. Hospital inpatient records summarize public and private hospital admissions within the jurisdiction.²⁸  We used hospital admission date and date of pertussis diagnosis to assess whether the pertussis case was temporally associated with a hospitalization or ICU admission (admission date ≤10 days following diagnosis). Deaths, identified from the death register, coincided with the date of diagnosis (death date ≤10 days following diagnosis or diagnosed postmortem).

First Nations mothers and infants were identified using a previously validated algorithm combining information across multiple administrative data sources to avoid incomplete and inaccurate data.²⁹ 

We restricted the primary analysis to live, singleton births with complete covariate information. We excluded mother-infant pairs with a record of vaccination <14 days prior to delivery (classed as “indeterminate vaccination status”) (Supplemental Fig 3). We performed descriptive analyses using χ² tests for categorical variables and Wilcoxon rank sum tests for nonnormally distributed continuous variables to examine cohort characteristics by maternal vaccination status and pertussis infection.

We used a mixed effects Cox model to compare pertussis infection rates among maternally vaccinated infants and infants with no record of maternal vaccination, with infant age (in months) as the underlying time variable. We included 2 random intercepts for (1) state/territory to account for clustering within region and (2) mothers to account for clustering among infants sharing the same mother. Infants were censored at either the (1) date of their first pertussis notification, (2) date of death, or (3) end of data availability. We used inverse probability of treatment (vaccination) weights to account for confounding. Treatment weights were derived from the predicted probability of pertussis vaccination during pregnancy, based on a fitted multivariable logistic regression model with maternal age, ethnicity, preexisting health conditions, pregnancy complications, parity, smoking status, initiation of prenatal care, year and season of conception, Socioeconomic Index for Areas quintile, and receipt of influenza vaccine during pregnancy as predictor variables. In additional analyses, we considered models adjusting for these variables rather than applying inverse probability of treatment weights. We evaluated the balance between maternally vaccinated and maternally unvaccinated groups by examining the absolute standardized mean differences in maternal characteristics before and after application of the weights (Supplemental Fig 4).

Vaccine effectiveness was estimated as one minus the inverse probability of treatment weighted hazard ratio (HR) from the Cox model × 100. We fit separate models to estimate VE against notified infection and hospitalization. Because of small numbers, we were unable to fit VE models against death or ICU admission. We fit additional models to evaluate VE by gestational age at vaccination (<28 weeks, 28–31 weeks, or ≥32 weeks versus unvaccinated) and the time between vaccination and birth (2–6 weeks, 7–11 weeks, or ≥12 weeks versus unvaccinated). To estimate VE of maternal pertussis vaccination before the possible influence of childhood immunization (and in the age group with highest incidence of severe pertussis), we conducted a sensitivity analysis, restricted to infants <2 months old.

For NT and QLD only, where childhood immunization data were available, we fit models to estimate infant VE against pertussis infection associated with receipt of the 3-dose series of DTaP immunization through 12 months of age to evaluate possible effect modification by maternal vaccination status (ie, “blunting” effects). We compared rates of pertussis infection by infant DTaP dose using Cox proportional hazard models with infant age as the time variable and DTaP dose number as the time-varying exposure. For each DTaP dose, infants were censored at either the (1) date of their first pertussis infection (outcome of interest), (2) date of death, (3) date of receipt of additional DTaP dose, (4) age 12 months, or (5) end of data availability. DTaP VE was estimated as 1 minus the HR adjusted for child covariates, including maternal socioeconomic status, the infant’s First Nations status, degree of prematurity, mother’s initiation of prenatal care (used as a proxy measure for family’s access to health services), and year of birth × 100. To evaluate potential effect modification, we included maternal vaccination status as a hospitalization term with DTaP dose. to compare rates of pertussis at each infant DTaP dose by maternal pertussis vaccination status, we constructed additional Cox proportional hazard models with maternal vaccination status as the exposure variable stratified by infant DTaP dose.

The cohort included 279 418 infants born to 252 444 mothers: 9996 mother–infant pairs in NT, 75 049 in WA, and 194 373 in QLD (Supplemental Fig 3). Of these, 51.7% (n = 144 429/279 418) had a record of pertussis vaccination during pregnancy: 5.0% (n = 14 028/279 418) before 28 weeks of gestation, 28.7% (n = 80 327/279 418) at 28–31 weeks of gestation, and 17.4% (n = 48 629/279 418) at ≥32 weeks of gestation (Supplemental Fig 5). For 1445 (0.5%) infants, the timing of maternal vaccination could not be determined. The characteristics of mother–infant pairs by pertussis vaccination during pregnancy are presented in Table 1.

A total of 331 notified pertussis cases were Identified in the cohort up to 18 months of age, equating to 118 cases per 100 000 infants; 119 cases were identified among infants of vaccinated mothers (82 per 100 000 infants) and 212 cases were identified among infants of unvaccinated mothers (157 per 100 000 infants). Among the 331 pertussis cases, 49 (14.8%) were diagnosed among infants <2 months old, and 124 (37.5%) were diagnosed among infants <6 months old (Supplemental Fig 6); 12.9% (n = 16/124) of cases among infants <6 months old coincided with a hospital admission, 4.8% (n = 6/124) coincided with an ICU admission, and 1.6% (n = 2/124) had a subsequent death recorded.

Maternal dTpa VE among infants <6 months of age was 65.1% (95% confidence interval [CI], 49.5–76.0) against notified pertussis infection and 60.2% (95% CI, −18.3 to 86.6) against hospitalized pertussis infection (Table 2; Fig 1; Supplemental Fig 7). There were insufficient data to estimate VE against ICU admission with pertussis and death. Results of analyses restricted to infants <2 months of age were similar to those for infants <6 months of age (Supplemental Table 4).

By age group, maternal dTpa VE was 70.4% (95% CI, 50.5–82.3) among infants <2 months old, 65.7% (95% CI, 41.8–79.8) among infants 3 to 4 months of age, 61.6% (95% CI, 37.5–76.4) among infants 5 to 6 months of age, and 43.3% (95% CI, 6.8–65.6) among infants 7 to 8 months of age (Fig 1). VE was not significant after 8 months of age (9–12 months: VE, 21.8% [95% CI, −31.8 to 53.6%]; 13–18 months: 34.1% [95% CI, 18.1–63.2]; 19–24 months: 44.0% [95% CI, −5.0 to 70.1]). The median age at notified pertussis infection among maternally vaccinated infants (median: 11 months; interquartile range, 8.75 months) was greater than the median age at pertussis infection among maternally unvaccinated infants (median: 7.5 months; interquartile range, 10 months; P < .001), although this was predominantly observed in QLD (Supplemental Fig 8). The VE for infants <6 months of age ranged from 11.6% (95% CI, –57.9 to 50.5) in WA to 76.0% (95% CI, 63.6–84.2) in QLD (Table 2). VE among infants <6 months of age was similar regardless of the gestational age at vaccination and the time between vaccination and birth.

A total of 171 840 infants were born in NT and QLD with immunization records available. Of these, 85 166 (49.6%) were exposed to maternal vaccine in utero and 86 674 (50.4%) had no record of maternal vaccination. The majority of infants who were either exposed to maternal vaccination or not received at least 1 infant dose of a pertussis-containing vaccine (93.4% and 84.8%, respectively); 87.3% of maternally vaccinated infants completed the 3-dose DTaP series and 76.6% of infants with no maternal vaccination completed the 3-dose series (Table 3).

Although the infant VE of 1 and 2 doses of DTaP was similar for maternally vaccinated and infants with no maternal vaccination (dose 1: 62.5% vs 71.2%, P value for interaction = .32; dose 2: 83.2% vs 83.6%, P value for interaction = .46), infant VE of the third dose of DTaP was lower (dose 3: 76.5% vs 92.9%, P value for interaction = .002) (Table 3). Despite this, the incidence of pertussis among those who received 3 DTaP doses was similar for maternally vaccinated infants and infants with no maternal vaccination (20.7 cases per 100 000 infants versus 23.1 cases per 100 000 infants; adjusted HR 0.70; 95% CI, 0.61–3.39). Survival curves indicate earlier and more frequent pertussis infection among infants with no record of maternal vaccination, regardless of DTaP immunization (Fig 2).

During the first 3 years of maternal pertussis immunization programs implemented in 3 Australian jurisdictions, we estimate that 52% of pregnant individuals received an acellular pertussis vaccine, and vaccination was associated with a 66% overall decrease in infant pertussis infection through 6 months of age. This protective effect of maternal pertussis vaccination was significant through to 8 months of age. Although we observed some evidence to support a reduced infant VE for 3 doses of DTaP immunization among maternally vaccinated infants, this did not appear to translate to greater risk of disease compared with infants with no record of maternal immunization. These results support ongoing pertussis vaccination during pregnancy to prevent pertussis-related morbidity among young infants.

Because of the size of our cohort, we were able to specifically evaluate VE against pertussis by the gestational timing of vaccination and the age of the infant. Although immunogenicity studies have observed higher antibody avidity in the cord blood of infants born to mothers vaccinated between 27 and 30 weeks of pregnancy (compared with later in pregnancy)^15,30  and recent studies have reported increased half-life of maternal antibodies following longer intervals between vaccination and delivery,³¹  we did not observe significant differences in VE by timing of vaccination. Similarly, a recent open-label randomized controlled trial in the United Kingdom comparing pertussis vaccination at 16 to 23 weeks, 24 to 27 weeks, and 28 to 31 weeks found no effect of gestational timing of vaccination on infant antibodies at birth.³²  Other surveillance-based studies of maternal pertussis vaccination at ≥17 weeks, 13 to 16 weeks, 8 to 12 weeks, 1 to 7 weeks before birth also suggested that timing has a limited effect on VE against infant disease.³³  Although our findings support the potential effectiveness of administering acellular pertussis vaccines during late second and third trimesters, it is worth noting that because of the recommendation in place during the study period (ie, vaccination at 28 weeks of pregnancy), few mothers were vaccinated before 25 weeks of completed gestation.

We assessed VE through to 18 months of age to evaluate the duration of protection against pertussis attributed to maternal vaccination as well as possible blunting of infants’ responses to primary pertussis vaccination. We observed significant protection against disease until at least 8 months of age; 2 months longer than reported in previous studies. We observed some evidence suggesting a lower VE for the third dose of infant DTaP vaccine for maternally vaccinated infants compared with infants with no maternal vaccine exposure. This finding aligns with results from previous randomized controlled trials^17,34,35  and nonrandomized studies,^16,36  that have consistently documented lower immunologic responses to infant pertussis vaccines among infants previously exposed to maternal pertussis immunization.

Few studies have directly measured the VE of infant immunization by maternal vaccination status. A recently published multicenter surveillance study of 376 infants 2 to 11 months old showed that VE of at least 1 dose of pertussis-containing vaccine against hospitalized pertussis infection was 74% among maternally vaccinated infants and 68% for infants with no history of maternal vaccination.³⁷  However, the authors acknowledge they were limited by small sample size and were unable to perform stratified analyses to formally investigate whether maternal vaccination modifies VE.³⁷  In our larger population-based study, we identified effect modification at third dose of DTaP, but lower incidence of pertussis through 18 months of age. Our interpretation of these findings, in combination with the published literature, is that maternal antibodies may “blunt” the response of infants to primary immunization, but maternal and/or infant antibodies are sufficient to protect maternally vaccinated infants from infection. However, further research is needed to confirm our findings.

Our study had several strengths and limitations. First, using data linkage systems in Australia, we were able to construct a large, population-based cohort of mother–infant pairs across multiple jurisdictions with comprehensive longitudinal information on maternal and infant health, helping to improve generalizability of findings and reduce selection bias. However, these data are observational and, as with any such study, results may be influenced by confounding. To restrict this influence, we applied several techniques, including adjustment for a range of available covariates and fitting weighted models by inverse probability of treatment (vaccination). Despite this, we cannot entirely exclude the possibility of uncontrolled confounding, especially for unobserved or latent variables. Another important limitation to our study was the method of outcome identification and the small number of infant pertussis cases, which precluded us from undertaking some analyses for infants <2 months, generating VE estimates against ICU admission and death, performing adjusted sensitivity analyses, and reduced the precision of our estimates by jurisdiction. Although cases were identified using a national, legally mandated data collection employing a standardized case definition,²⁶  we cannot exclude the possibility that some asymptomatic cases were not identified in the cohort, resulting in some outcome misclassification. Despite this, given most cases were polymerase chain reaction detected (with high specificity and sensitivity),³⁸  false positives in our cohort are unlikely and we do not expect that outcome misclassification would have occurred differentially across maternal vaccinated and unvaccinated infants. Finally, although our sources of data for immunization drew from medical records and registers that have high specificity,³⁹  it is possible that some vaccinated individuals were misclassified as unvaccinated. Because we have no reason to believe this misclassification would be differential by outcome, this could have biased our VE estimates toward the null, suggesting our estimates may underestimate the true VE.

From this large, population-based cohort, we estimate that immunization with acellular pertussis vaccine during pregnancy prevented 65% of pertussis infections through 6 months of age. Although the VE of the third DTaP dose may be lower for maternally vaccinated infants, we did not observe evidence supporting higher rates of pertussis infection associated with maternal vaccination through 18 months of age. These results indicate that maternal pertussis vaccination protects infants from pertussis infection during a period of greatest vulnerability to severe morbidity and mortality. Our findings support the infant health benefits of recommendations to administer a booster dose of pertussis vaccine near 28 weeks of gestational age.

The authors thank the Linkage and Client Services Teams at the Queensland and Western Australian Data Linkage Branches from their respective Departments of Health for their services in linking the data. The authors additionally acknowledge staff at the Midwives Notification System, the Hospital Morbidity Data Collection, the Western Australian Notifiable Infectious Diseases Database, the Western Australia Antenatal Vaccination Database, and the Western Australian Registry of Births, Deaths and Marriages. In Northern Territory, the authors thank Nicky O’Brien and others at SA-NT DataLink for their services in providing the linked datasets, as well as the data custodians and staff at the NT Centre for Disease Control for facilitating access to the NT Immunization Register, the NT Notifiable Conditions Register, and the NT Perinatal Trends Database. In Queensland, the authors thank the data custodians for the Vaccination Information and Vaccination Administration System and the Qld Perinatal Data Collection.

CI

confidence interval

DTaP

diphtheria-tetanus-acellular pertussis-containing vaccine

dTpa

diphtheria-tetanus-acellular pertussis vaccine

HR

hazard ratio

QLD

Queensland

NT

Northern Territory

VE

vaccine effectiveness

WA

Western Australia