Optimal timing of influenza vaccine during pregnancy: A systematic review and meta‐analysis

Abstract Background Pregnant women have an elevated risk of illness and hospitalisation from influenza. Pregnant women are recommended to be prioritised for influenza vaccination during any stage of pregnancy. The risk of seasonal influenza varies substantially throughout the year in temperate climates; however, there is limited knowledge of how vaccination timing during pregnancy impacts the benefits received by the mother and foetus. Objectives To compare antenatal vaccination timing with regard to influenza vaccine immunogenicity during pregnancy and transplacental transfer to their newborns. Methods Studies were eligible for inclusion if immunogenicity to influenza vaccine was evaluated in women stratified by trimester of pregnancy. Haemagglutination inhibition (HI) titres, stratified by trimester of vaccination, had to be measured at either pre‐vaccination and within one month post‐vaccination, post‐vaccination and at delivery in the mother, or in cord/newborn blood. Authors searched PubMed, Scopus, Web of Science and EMBASE databases from inception until June 2016 and authors of identified studies were contacted for additional data. Extracted data were tabulated and summarised via random‐effect meta‐analyses and qualitative methods. Results Sixteen studies met the inclusion criteria. Meta‐analyses found that compared with women vaccinated in an earlier trimester, those vaccinated in a later trimester had a greater fold increase in HI titres (1.33‐ to 1.96‐fold) and higher HI titres in cord/newborn blood (1.21‐ to 1.64‐fold). Conclusions This review provides comparative analysis of the effect of vaccination timing on maternal immunogenicity and protection of the infant that is informative and relevant to current vaccine scheduling for pregnant women.


| INTRODUC TI ON
Pregnant women have a particularly high risk of illness and hospitalisation from influenza. During pregnancy, women experience physiological changes in their cardiopulmonary and immunological systems. 1,2 An increase in oxygen consumption, a decrease in lung capacity and the suppression of cell-mediated immunity to tolerate the growth of a genetically foreign foetus all increase pregnant women's susceptibility to infectious diseases and respiratory pathogens such as influenza. [3][4][5] The risks of hospitalisation and complications for respiratory illness during the influenza season are higher for pregnant women and increase by trimester. 6,7 Furthermore, pregnant women infected with influenza might be more likely to have adverse birth outcomes. 3,[8][9][10] Vaccination is the most effective preventative measure against influenza infection, 8,11 and influenza vaccines have been recommended for use in pregnant women for many decades. 12 The safety, effectiveness and immunogenicity of influenza virus vaccines during pregnancy have been studied extensively, and there is good evidence to support current vaccination recommendations. [13][14][15][16] The World Health Organization and the US Centers for Disease Control and Prevention prioritise pregnant women for vaccination, 17 and the Advisory Committee on Immunization Practices and the American College of Obstetricians and Gynecologists have recommended the inactivated seasonal influenza vaccine to women in any trimester since 2004. 17,18 Evidence of additional benefits of maternal influenza vaccination, such as the protection of young infants via placental transfer of protective antibodies to the foetus, provides further support for antenatal vaccination. 19,20 Moreover, the interruption of influenza virus transmission by vaccinating the mother, together with transplacental transfer of vaccine-associated antibody, also reduces the risk of infection for infants 3-4 months old (before direct vaccination is possible). 21 Despite the heightened risk of influenza illness in pregnant women and benefits of vaccination, vaccination coverage rates in this population remain suboptimal. In recent years, coverage rates in the United States and Australia have ranged from 20%-50%. [22][23][24][25] Surveys have attributed these low vaccine uptake rates in part to distrust in the healthcare system, unawareness of the risks of influenza infection during pregnancy, concerns about vaccine safety for the foetus and lack of encouragement from healthcare professionals. 8,[22][23][24] Recommendations for the timing of influenza vaccination during pregnancy have varied. Although immunisation is now recommended for women at any stage of pregnancy, 26 the timing of vaccination to optimise benefit to the mother and their infants is not well established. A structured analysis of the optimal timing of influenza vaccination during pregnancy would inform specific scheduling recommendations to pregnant women and maximise the benefit received by vaccination. knowledge of how vaccination timing during pregnancy impacts the benefits received by the mother and foetus.
Objectives: To compare antenatal vaccination timing with regard to influenza vaccine immunogenicity during pregnancy and transplacental transfer to their newborns.
Methods: Studies were eligible for inclusion if immunogenicity to influenza vaccine was evaluated in women stratified by trimester of pregnancy. Haemagglutination inhibition (HI) titres, stratified by trimester of vaccination, had to be measured at either pre-vaccination and within one month post-vaccination, post-vaccination and at delivery in the mother, or in cord/newborn blood. Authors searched PubMed, Scopus, Web of Science and EMBASE databases from inception until June 2016 and authors of identified studies were contacted for additional data. Extracted data were tabulated and summarised via random-effect meta-analyses and qualitative methods.
Results: Sixteen studies met the inclusion criteria. Meta-analyses found that compared with women vaccinated in an earlier trimester, those vaccinated in a later trimester had a greater fold increase in HI titres (1.33-to 1.96-fold) and higher HI titres in cord/newborn blood (1.21-to 1.64-fold).

Conclusions:
This review provides comparative analysis of the effect of vaccination timing on maternal immunogenicity and protection of the infant that is informative and relevant to current vaccine scheduling for pregnant women.

K E Y W O R D S
immunogenicity, influenza, pregnancy, timing, trimester, vaccination F I G U R E 1 Flow diagram detailing the study inclusion/exclusion process. Reasons for inclusion/exclusion: (i) Article was a review, recommendation, statement, did not study pregnant women, did not report on immunogenicity, or did not study influenza vaccine. (ii) Article was a cost-benefit analysis, studied vaccination uptake or attitudes, did not study pregnant women, did not report on immunogenicity, or did not study influenza vaccine. (iii) Article (n = 1) was a review of an included study, evaluated only adverse birth outcomes (n = 25), did not include any data on vaccination timing (n = 24), or only studied women in one trimester (n = 3). (iv) 19 studies contacted for stratified data: no response (n = 5), data not eligible due to time-points measured (n = 1), declined (n = 1), data published in another included study (n = 1). (v) One article identified through hand-searching of reference lists. (vi) References of studies included in each analysis Previous reviews of antenatal influenza vaccination have reported limited and mixed evidence on the association between influenza vaccination, influenza infection and adverse birth outcomes, and have not examined the relationship between vaccination timing and immunogenicity. 21,[27][28][29][30][31][32] This systematic review examined whether the timing of influenza vaccination during pregnancy affects the immunogenicity of the vaccine in the mother and transplacental transfer of antibody to the newborn.

| ME THODS
This review was conducted in accordance with the PRISMA 33 checklist (Table S1).

| Eligibility criteria
Our population of interest was women vaccinated during pregnancy with a seasonal or pandemic vaccine. Each study was required to include women vaccinated in different trimesters to enable comparison of outcomes between trimesters. All study designs were eligible for inclusion.
The primary outcome was geometric mean titre (GMT) measured by haemagglutination inhibition (HI) assays. Geometric mean titres had to be measured either at (a) both pre-vaccination and within one month post-vaccination, (b) both post-vaccination and delivery in the mother, or (c) delivery in cord blood or newborn blood (hereafter referred to as cord blood for simplicity). If reported in the included studies, seroprotection (HI titre of ≥1:40) and seroconversion (≥4fold increase in HI titre) rates were also discussed. 34 We did not assess vaccine safety or adverse outcomes in the mother or newborn. The records identified were assessed for eligibility in three phases by two independent reviewers (WC and RM) ( Figure 1): screening by title, abstract and full-text review. A third reviewer (NG) resolved any inconsistencies. Reference lists of papers that were identified through the database search were also searched for additional studies. If the study had not stratified GMTs by trimester of vaccination, authors were contacted to provide the required data. Grey literature was not searched.

| Data extraction
Summary characteristics of each study (study period, design, sample size, location and population, trimester of vaccination and vaccine administered) and outcome measures (GMTs, standard deviation, confidence intervals, seroprotection and seroconversion rates) were extracted and tabulated by one reviewer (WC) ( Table 1 and Table   S2). The methodological validity and internal bias of all included studies were assessed using critical appraisal tools ROBINS-I for non-randomised trials (including single-arm studies [35][36][37][38] ) and RoB 2.0 for randomised trials (Table S3).

| Analysis
Three analyses of study data were conducted. First, the effect of vaccination trimester on acute immune response was measured by Where not otherwise reported, standard errors for GMTs at each time-point were calculated from reported 95% confidence intervals (CI) using the t-distribution or from standard deviations. Standard errors of ratios were calculated using the z-distribution or, if the sample size did not change between time-points, the t-distribution. All analyses were conducted on the logarithmic scale and back-transformed to the original scale. Heterogeneity was examined using forest plots and the I 2 statistic. Due to the small number of studies included, funnel plots to check for publication bias were not produced. 39 Data were summarised using a random-effect (DerSimonian and Laird) model to accommodate between-study heterogeneity in true effects, and the estimate and 95% CI of the effect were presented on the logarithmic scale. 40 Timing of vaccination was stratified by trimester (1st trimester: <14 weeks, 2nd trimester: 14-27 weeks, 3rd trimester: ≥28 weeks).
Accordingly, three comparisons were made: 2nd-trimester vaccination compared with 1st-trimester vaccination, the 3rd trimester compared with the 2nd and the 3rd trimester compared with the   Cord blood d   3rd TRI  1st TRI  2nd TRI  3rd TRI  1st TRI  2nd TRI  3rd TRI   21  ----23  19   83  -59  83  -35  1st. For studies that included a two-dose vaccine group, these data were analysed in separate meta-analyses (trimester of vaccination defined by the timing of the first dose). 35 We also examined seroprotection and seroconversion rates reported in the included studies (Table S2). Where possible, we explored these outcomes stratified by trimester of vaccination;

Mother at delivery
however, a limited number of studies reported these stratified data.
We included some statistical results from the original papers (eg p-values from comparisons of seroprotection and seroconversion proportions).
For the acute immune response and transplacental antibody outcomes, we conducted three sensitivity analyses for the 3rd-versus 2nd-trimester comparison in which we restricted the studies included in the meta-analysis to those that also vaccinated women in the 1st trimester ( Figures S4, S5, S15). This sensitivity analysis allows us to explore whether the findings based on studies that in- women vaccinated in the 1st trimester. 35 Six other sensitivity analyses were undertaken whereby studies with concerns of bias (due to small sample size 42,43 or unbalanced loss to follow-up between study groups 38,41,44 ) were excluded (Table S3).
All analyses included all available cases with the unit of analysis being the participant. Data preparation and manipulation were undertaken using Microsoft Excel, and all meta-analyses and forest plots were produced in Stata 14.2. 45

| Literature search
After duplicates were removed, 1,285 articles were identified from the electronic databases search (Figure 1

| Acute immune response
Twelve studies measured both baseline HI titres immediately prior to vaccination and post-vaccination HI titres (Table S2).
However, this trend was not consistent across all three studies.
Restricting the meta-analysis to only studies that also vaccinated women in the 1st trimester 35 Figure S6; 11 studies, 903 participants).

| 3rd versus 1st trimester
The clarity of a dose-response relationship by trimester is strength-

| Antibody persistence
In addition to the post-vaccination blood draw, seven studies 35,36,38,41,44,46,51 also measured HI titres in the mother at delivery (Table S2). Thus, we can quantify the reduction in antibodies between the post-vaccination immune response and delivery (geometric mean titre fold decrease (GMFD)). All seven studies included women vaccinated in the 2nd and 3rd trimester, while only one 35 included women in all three trimesters.

| 2nd versus 1st trimester
Horiya et al 35 were the only study to include women vaccinated in all three trimesters; therefore, no meta-analysis was possible. In this study, the fold reduction in GMT was greater for women vaccinated in the 1st trimester compared with women vaccinated in the 2nd tri-

| 3rd versus 2nd trimester
The fold reduction in GMT from immunisation to delivery was simi-

| Transplacental antibodies
Twelve studies measured HI titres in cord blood at delivery (Table   S2). Four studies 35

| Acute immune response
With  resulted in lower HI titres at delivery (P < 0.05), however did not find a difference in seroprotection rates in the mother at delivery be-

| Risk of bias
Largely influenced by the nature of the intervention and the outcome measures, we determined that there was low risk of internal methodological bias in all included studies (Table S3). However, the main limitations were low numbers of participants, missing data or loss to follow-up (Table 1) Figures S6, S10, S11, S13, S14, S17).

| Summary of findings
This systematic review explored the association between the timing of an influenza vaccination given during pregnancy and immunogenicity in the mother and newborn. The three main findings of this review were as follows. First, women vaccinated later during pregnancy had a greater immune response to vaccination. This effect size increased from the 1st to 3rd trimester. Second, maternal antibodies at delivery were reduced by a similar factor relative to post-immunisation titres, regardless of whether women were vaccinated in the 2nd or 3rd trimester. This observation suggests that antibodies wane faster in women vaccinated later in pregnancy; however, this hypothesis could not be explored further due to the low number of studies that vaccinated women in the 1st trimester. Regardless, antibody waning seems to occur over a fairly short time frame; some studies have estimated the antibody half-life to be as short as seven weeks. 54,55 Third, despite the observation that GMT at delivery was consistent across trimesters, there was strong evidence that vaccination in a later trimester increased the transfer of antibodies to the foetus.
To gain further insight into the impact of vaccination timing, we reported on the public health-relevant outcomes of seroprotection and seroconversion. An HI titre of 1:40 is recognised as an immunologic correlate corresponding to a 50% reduction in the risk of contracting influenza. 56 Even though some of the studies included in this review reported significantly higher seroprotection or seroconversion rates for women vaccinated in later trimesters at both the acute post-vaccination time-point and at delivery in the mother and cord blood, most women still achieved sufficient protection regardless of vaccination timing.
Other studies not meeting this review's inclusion criteria have found no difference in seroprotection rates by vaccination trimester. 12,[57][58][59] Some study designs included administering two vaccine doses, higher antigen doses or adjuvanted vaccines. 35,37,41,43,50 By trimester, the immune response was not greatly impacted by higher doses or adjuvanted vaccines; neither was a substantial increase in GMT conferred after a second dose. However, those who received two doses or a higher dose tended to have higher levels of antibodies at delivery and also transferred more antibodies to the foetus.

| Limitations
Of the original publications, only six reported the effect of vaccination timing on vaccine immunogenicity in detail with stratification by trimester. Thus, much of the data acquired for this review were not originally collected to address our objectives. Observational studies are also prone to inherent biases, and many of the included stud- All studies in this review included women vaccinated in the second and third trimesters, but only seven studies included women vaccinated in the first trimester (likely influenced by the short time between pregnancy diagnosis and the end of first trimester). This limited the comparisons of vaccine immunogenicity against women vaccinated in the first trimester. The robustness of our results is likely affected by the low number of studies available for inclusion in these meta-analyses ( Figure 1). However, the I 2 values were less than 44% in all our meta-analyses, indicating low between-study heterogeneity. Finally, we interpreted our results by considering overall trends rather than statistical significance and did not account for multiple comparisons.
Immunity prior to vaccination can both affect the immune response generated from the vaccine and mask the true immunogenicity of the vaccine. Seroprotection rates may over-estimate immunogenicity, while seroconversion rates may result in under-estimation if there is high baseline population protection. We used fold increases and fold decreases in GMT in an attempt to control for the pre-vaccination immune state-a valid method when combined into a meta-analysis. 60 Furthermore, exposure to wild-type influenza virus between vaccination and delivery may have impacted HI titres measured at delivery.
Finally, the widely accepted standard correlate of protection (ie based on challenge studies in healthy adults, that an HI titre >1:40 corresponds to a 50% reduction in the risk of contracting influenza) is not grounded in strong evidence. 61,62 While a higher HI titre is predictive of some protection, there is evidence to suggest that neuraminidase inhibition (NAI) titre is more predictive of protection and reduced infection. 63 NAI titres were not measured in any of the included studies.
There remains a need to standardise serological assays and better define correlates of protection, which may vary according to individual characteristics, populations, age groups and vaccine types.

| IMPLI C ATI ON S AND CON CLUS I ON
Vaccinating a woman in early pregnancy will provide protection against influenza for a greater proportion of pregnancy, but may increase the probability that this immunity will not last until delivery.
Lower antibody levels at delivery may reduce transplacental antibodies and the benefit to the newborn, and loss of immunity in the mother may increase the chance of her becoming a viral source to the newborn. A similar issue arises with antenatal pertussis vaccination, where there is recent evidence that immunogenicity is higher in the second trimester compared with the third trimester, which has resulted in some countries bringing forward their recommendations for the optimal timing of pertussis immunisation in pregnancy. 64 Furthermore, for women vaccinated early during pregnancy, but late in the influenza season, clinical protection may be reduced if the circulating strain of the following influenza season does not match the strain included in the previous season's vaccine, regardless of whether antibody levels remain high. Given that women immunised earlier in pregnancy show evidence of immune waning by the point of delivery, our findings support current recommendations for women immunised early in their pregnancy to receive a second dose if they are still pregnant in the following influenza season.
Due to the increased risk of adverse birth outcomes caused by maternal influenza infection, and the subsequent increased risk of lifelong chronic diseases associated with these birth outcomes, the economic burden of maternal infection is substantial. 65 There is recent evidence that the risk of foetal death and other adverse birth outcomes is highest for women who become infected with seasonal influenza during their first trimester. 66 That research compounds the complexity of optimising the scheduling of influenza vaccination for pregnant women and, in conjunction with our findings, highlights that the infant is at risk both in utero and after birth. The protection of the infant from adverse birth outcomes and infection in early life has not been measurable outcomes in this review. Nevertheless, this is a key aspect of antenatal vaccination policy and there is an extensive amount of documented interest in these benefits. 51,52,[67][68][69] While studies of vaccine effectiveness and efficacy were not included in this systematic review, the increased immune response by gestational age suggests that post-vaccination influenza infection might be less likely for women vaccinated in later trimesters.
However, preventing clinical disease depends more on the timing of vaccination relative to the seasonal influenza epidemic rather than to gestational age. For example, vaccination should not be delayed for a woman in her first trimester if the influenza season has begun and the vaccine is available.
The findings of this systematic review are informative and relevant to current vaccine scheduling for pregnant women. We found that women vaccinated later in pregnancy had a stronger immune response and transferred more antibodies to the foetus.
A limited number of published studies have considered the impact of vaccine timing on maternal and infant protection. We need to understand the full implications of vaccination timing for protection of mother, foetus and newborn. This knowledge is key to enabling future health policies to optimise protection for both mother and infant, developing future vaccine scheduling recommendations, and informing professional advice. In addition, a better understanding of the benefits of influenza vaccination during pregnancy may help increase vaccination rates among pregnant women.

ACK N OWLED G EM ENTS
The authors would like to thank Dr Juan-Pablo Villanueva-Cabezas and Dr Patricia Campbell for their assistance in revising the manuscript.