Efficacy, duration of protection, birth outcomes, and infant growth associated with influenza vaccination in pregnancy: a pooled analysis of three randomised controlled trials

Background: Maternal influenza immunisation can reduce morbidity and mortality associated with influenza infection in pregnant women and young infants. We aimed to determine the vaccine efficacy of maternal influenza immunisation against maternal and infant PCR-confirmed influenza, duration of protection, and the effect of gestational age at vaccination on vaccine efficacy, birth outcomes, and infant growth up to 6 months of age. Methods: We did a pooled analysis of three randomised controlled trials done in Nepal (2011–2014), Mali (2011–2014), and South Africa (2011–2013). Pregnant women, gestational age 17–34 weeks in Nepal, 28 weeks or more in Mali, and 20–36 weeks in South Africa, were enrolled. Women were randomly assigned 1:1 to a study group, in which they received trivalent inactivated influenza vaccine (IIV) in all three trials, or a control group, in which they received saline placebo in Nepal and South Africa or quadrivalent meningococcal conjugate vaccine in Mali. Enrolment at all sites was complete by April 24, 2013. Infants and women were assessed for respiratory illness, and samples from those that met the case definition were tested for influenza by PCR testing. Growth measurements, including length and weight, were obtained at birth at all sites, at 24 weeks in South Africa, and at 6 months in Nepal and Mali. The three trials are registered with ClinicalTrials.gov, numbers NCT01430689, NCT01034254, and NCT02465190. Findings: 10 002 women and 9800 liveborn infants were included. Pooled efficacy of maternal vaccination to prevent infant PCR-confirmed influenza up to 6 months of age was 35% (95% CI 19 to 47). The pooled estimate was 56% (28 to 73) within the first 2 months of life, 39% (11 to 58) between 2 and 4 months, and 19% (–9 to 40) between 4 and 6 months. In women, from enrolment during pregnancy to the end of follow-up at 6 months postpartum, the vaccine was 50% (95% CI 32–63) efficacious against PCR-confirmed influenza. Efficacy was 42% (12 to 61) during pregnancy and 60% (36 to 75) postpartum. In women vaccinated before 29 weeks gestational age, the estimated efficacy was 30% (–2 to 52), and in women vaccinated at or after 29 weeks, efficacy was 71% (50 to 83). Efficacy was similar in infants born to mothers vaccinated before or after 29 weeks gestation (34% [95% CI 12 to 51] vs 35% [11 to 52]). There was no overall association between maternal vaccination and low birthweight, stillbirth, preterm birth, and small for gestational age. At 6 months of age, the intervention and control groups were similar in terms of underweight (weight-for-age), stunted (length-for-age), and wasted (weight-for-length). Median centile change from birth to 6 months of age was similar between the intervention and the control groups for both weight and length. Interpretation: The assessment of efficacy for women vaccinated before 29 weeks gestational age might have been underpowered, because the point estimate suggests that there might be efficacy despite wide CIs. Estimates of efficacy against PCR-confirmed influenza and safety in terms of adverse birth outcomes should be incorporated into any further consideration of maternal influenza immunisation recommendations. Funding: Bill & Melinda Gates Foundation. Previous publications have described each of the three clinical trials.8, 9, 12 Funded by the Gates Foundation, the trials were initially designed as separate studies with overlapping features. Trial procedures were then coordinated for future pooled analyses from the planning phase onward, before completion of the trials, as previously outlined.5 Pregnant women were enrolled from April 25, 2011, to April 24, 2013, in Nepal, Sep 12, 2011, to April 18, 2013, in Mali, and March 3, 2011, to July 2, 2012, in South Africa. Infant follow-up visits ended on April 13, 2014, in Nepal, Jan 28, 2014, in Mali, and May 20, 2013, in South Africa. Infants were followed up for 6 months in Nepal and Mali and 24 weeks in South Africa. The studies in Nepal and Mali enrolled and vaccinated women year round, and the study in South Africa coincided enrolment with the influenza season. Multiple peaks of influenza activity were observed in Nepal and two peaks a year were observed in Mali (February and September–October). In South Africa, there were two peaks in the 2011 season (June and September) and one peak in 2012 (August). Pregnant women were screened and enrolled from nine Village Development Committees in rural southern Nepal (vaccinated at 17–34 weeks gestational age). Women accessing prenatal care were screened and enrolled in Bamako, Mali (vaccinated at ≥28 weeks gestational age) and Soweto, South Africa (vaccinated at 20–36 weeks gestational age). The overall median gestational age at vaccination, 29 weeks, was used in the analysis to stratify early and late gestational age at vaccination. In Nepal, the dates of last menstrual period were prospectively collected to calculate gestational age, although the studies in Mali and South Africa both used estimates at vaccination. In Mali and South Africa, ultrasound, uterine height, and date of last menstrual period were used. Verbal informed consent was obtained in Nepal, and written informed consent was obtained in South Africa. All women provided informed consent in Mali. If a participant in Mali was illiterate, consent was obtained in the presence of a literate witness after listening to an audio recording of the consent form. The study protocols were reviewed and approved by institutional review boards of partner entities: Emory University; University of Maryland; the Ministry of Health, Mali; Johns Hopkins Bloomberg School of Public Health; the Institute of Medicine at Tribhuvan University, Kathmandu, Nepal; Nepal Health Research Council, Kathmandu, Nepal; and University of Witwatersrand, Johannesburg.5, 8, 9, 12 The Emory University institutional review board performed ongoing and periodic review of the pooled analyses. Women were randomly assigned 1:1 to a study group, in which they received trivalent IIV (VAXIGRIP; Sanofi-Pasteur, sourced from Mumbai, India in Nepal, and from Lyon, France in Mali and South Africa) in all three trials, or a control group, in which they received saline placebo in Nepal and South Africa or quadrivalent meningococcal conjugate vaccine (Menactra; Sanofi Pasteur, Lyon, France) in Mali. In Nepal, women were assigned in blocks of eight, stratified by gestational age at enrolment (17–25 weeks vs 26–34 weeks), using sealed envelopes with the participant number on the outside and an allocation code within. In Mali, women were assigned using a computer-generated list with blocks of six or 12. In South Africa, randomisation was also computer-generated in blocks of 30 by enrolment site. Women were immunised with a single 0·5 mL dose of either IIV or control product injected into the deltoid muscle. Infants and women at all sites were assessed weekly for illness through active surveillance, and samples were tested for influenza by PCR, as were samples from participants hospitalised with respiratory illness. At all three sites, the influenza strains were determined, including H1N1 and H3N2 influenza A strains or influenza B. Vaccine and non-vaccine matched strains were included in analyses. Across sites, the most predominant influenza strain was influenza H3N2 in infants (37·0 cases per 1000 person-years among controls), and influenza B in women (20·0 cases per 1000 person-years among controls). With the exception of strain-specific estimates, we included all influenza strains in pooled vaccine efficacy estimates. Infant growth measurements, including length and weight, were obtained at birth at all three sites, at 24 weeks in South Africa, and 6 months in Nepal and Mali. Measurements were excluded if infants were older than 72 h when measured at birth. Infants were also excluded if they were younger than 150 days or 210 days or older at their last visit. Outcomes analysed in this pooled analysis include overall vaccine efficacy of maternal influenza immunisation against maternal and infant PCR-confirmed influenza, duration of maternal and infant protection, the effect of gestational age at vaccination on vaccine efficacy, adverse birth outcomes (low birthweight, stillbirth, preterm birth, and small for gestational age), and infant growth up to 6 months of age (infant weight-for-age, weight-for-length, length-for-age, median centile change from birth to 6 months, and mean weight and length at birth at 6 months). A combined cohort size of 10 000 provided more than 99% power to see a 35% difference in PCR-confirmed infant influenza infection based on a baseline incidence of approximately 0·200 cases per infant-year (assuming 0·500 person-years per infant), more than 99% power to detect a 20% difference in PCR-confirmed maternal influenza infection based on a baseline incidence of around 0·040 cases per person-year (assuming 0·500 person-years per mother), 80% power to detect a 40% difference in stillbirth based on a baseline incidence of around 0·015 cases per woman, and 80% power to detect a 18% difference in preterm birth based on a baseline incidence of 0·120 cases per live-birth. The combined analysis also had 80% power to detect a 15% difference in low birthweight with a combined cohort size of 9000 and a baseline incidence of around 0·150 cases per infant, and 80% power to detect a 13% difference in small for gestational age with a combined cohort size of 8500 and a baseline incidence of 0·300 cases per infant. For infant growth analyses, infant weight-for-age, weight-for-length, and length-for-age Z scores and percentiles were determined using the WHO growth standards.18 Analyses included cutoffs for underweight (weight-for-age Z score <–2), severely underweight (weight-for-age Z scores <–3), wasting (weight-for-length Z score <–2), severe wasting (weight-for-length Z scores <–3), stunting (length-for-age Z score <–2), and severe stunting (length-for-age Z scores <–3). Weight-for-age, weight-for-length, and length-for-age Z scores greater than 7 or less than negative 7 were excluded, resulting in excluding 50 infants from the at-birth analysis, and 17 infants from the 6-month analysis. Intergrowth newborn size standards were used to identify small for gestational age infants.19 Birthweight less than 2500 g was considered low birthweight. Data was pooled and analysed using a one-stage meta-analysis. Poisson regression models were used to calculate incidence rate ratios (IRR), and pooled Poisson models were based on random intercept models. Goodness-of-fit χ2 tests were used to assess the fit of the Poisson models. Log-binomial regressions were used to estimate risk ratios (RRs), and robust CIs were used. Pooled models were adjusted for the effects of site (each of the three trial locations), and interaction by site was evaluated for each pooled analysis. Interaction by site was significant for efficacy of maternal vaccination against PCR-confirmed influenza in infants less than 4 months of age (p=0·04), during the full study period for women (p=0·03), less than 4 months after vaccination for women (p=0·004), less than 6 months after vaccination for women (p=0·005), efficacy against H1N1 for women (p=0·01), and the low birthweight analysis (p=0·01). To further assess heterogeneity across sites by I2 testing, we did two-stage meta-analyses with site RRs. Vaccine efficacy was calculated as: (1 – IRR) × 100. Changes from birth to 6 months in median weight and length percentiles between intervention and control groups were compared using Wilcoxon Rank Sum Test. Mean weight and length at birth and 6 months were compared across study groups using two-sample t tests. Infants and women were considered always at risk for influenza infection and were censored at 180 days postpartum for Nepal and Mali and at 175 days postpartum in South Africa as per protocol. In age group analyses, 61 days was considered 2 months of age, and 122 days was considered 4 months of age. Analyses were as consistent as possible across sites, in terms of case definitions, person-time calculations, and data cleaning, for example. Therefore, numbers from this paper might differ slightly from site-specific publications. Women were excluded from during and after pregnancy analyses if the date of delivery was unknown, because we were not able to determine if the influenza episode occurred during or after pregnancy. Statistical analyses were done using Stata version 14.2. The three trials were registered with ClinicalTrials.gov, numbers {"type":"clinical-trial","attrs":{"text":"NCT01430689","term_id":"NCT01430689"}}NCT01430689, {"type":"clinical-trial","attrs":{"text":"NCT01034254","term_id":"NCT01034254"}}NCT01034254, and {"type":"clinical-trial","attrs":{"text":"NCT02465190","term_id":"NCT02465190"}}NCT02465190. The funder of the study had no role in study design or conduct of the three studies and no role in the data collection for the pooled analysis, but provided feedback on the pooled analysis study design, data analysis, data interpretation, and writing of the manuscript. The working group of authors had full access to all the data in the study and were responsible for the final decision to submit for publication.