Fetal death and preterm birth associated with maternal influenza vaccination: systematic review

Transcription

1 DOI: / Systematic review Fetal death and preterm birth associated with maternal influenza vaccination: systematic review DB Fell, a,b RW Platt, a A Lanes, c K Wilson, d,e,f JS Kaufman, a O Basso, a D Buckeridge a a Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, QC, Canada b Better Outcomes Registry & Network (BORN) Ontario, Ottawa, ON, Canada c Department of Epidemiology and Community Medicine, University of Ottawa, Ottawa, ON, Canada d Department of Medicine, University of Ottawa, Ottawa, ON, Canada e Clinical Epidemiology Program, Ottawa Hospital Research Institute, Ottawa, ON, Canada f Institute for Clinical Evaluative Sciences, Ottawa, ON, Canada Correspondence: DB Fell, Better Outcomes Registry & Network (BORN) Ontario, Children s Hospital of Eastern Ontario Research Institute, 401 Smyth Road, Centre for Practice Changing Research, Ottawa, Ontario, Canada K1H 8L1. dfell@bornontario.ca Accepted 13 May Published Online 10 July Background Before 2012, few studies had addressed pregnancy outcomes following maternal influenza vaccination; however, the number of publications on this topic has increased recently. Objectives To review comparative studies evaluating fetal death or preterm birth associated with influenza vaccination during pregnancy. Search strategy We searched bibliographic databases from inception to April Selection criteria Experimental or observational studies assessing the relationship between influenza vaccination during pregnancy and fetal death or preterm birth. Data collection and analysis Two reviewers independently abstracted data from studies meeting the inclusion criteria. Main results We included one randomised clinical trial and 26 observational studies. Meta-analyses were not considered appropriate because of high clinical and statistical heterogeneity. Three studies of fetal death at any gestational age reported adjusted effect estimates in the range , and four of five studies of fetal death at <20 weeks reported adjusted estimates between 0.89 and 1.23, all with confidence intervals including 1.0. Adjusted effect estimates for four of five studies of fetal death at 20 weeks ranged from 0.44 to 0.77 (two with confidence intervals not crossing 1.0), whereas a fifth reported a non-significant effect in the opposite direction. Among 19 studies of preterm birth, there was no strong evidence suggesting any increased risk, and meta-regression did not explain the moderate between-study heterogeneity (I 2 = 57%). Authors conclusions Most studies reported no association between fetal death or preterm birth and influenza vaccination during pregnancy. Although several reported risk reductions, results may be biased by methodological shortcomings of observational studies of influenza vaccine effectiveness. Keywords Fetal death, influenza vaccination, monovalent pandemic H1N1 vaccine, preterm birth, systematic review, trivalent inactivated influenza vaccine. Please cite this paper as: Fell DB, Platt RW, Lanes A, Wilson K, Kaufman JS, Basso O, Buckeridge D. Fetal death and preterm birth associated with maternal influenza vaccination: systematic review. BJOG 2015;122: Introduction Complex physiological changes during pregnancy render pregnant women more vulnerable to severe illness when infected with a pathogen such as influenza virus. 1,2 The severity of influenza illness in pregnant women is of concern, as symptoms such as hyperthermia 3 and immune responses influencing placental function 4 have the potential to affect the developing fetus. Reports from previous pandemics 5,6 and from the recent H1N1 pandemic 7,8 suggest that infected pregnant women are at increased risk for experiencing fetal death, although maternal influenza illness in non-pandemic years has not been consistently implicated as a risk factor for adverse perinatal outcomes As pregnant women are considered to be a high-risk group for severe influenza illness, several countries now recommend vaccination of all pregnant women with inactivated influenza vaccine, typically comprising two influenza A strains and one influenza B strain (i.e. trivalent). 18 However, high-quality evidence on fetal safety associated with maternal influenza vaccination has been lacking because of the exclusion of pregnant women from placebo-controlled trials. 19 In addition, limited data on ª 2014 Royal College of Obstetricians and Gynaecologists 17

2 Fell et al. vaccine coverage, challenges in ascertaining pregnancy losses and requirements for large sample sizes (particularly for rare perinatal outcomes and given that maternal vaccination rates have typically been low 20,21 ) pose obstacles for observational studies. During the H1N1 influenza pandemic, despite limited pre-licensure safety data for the rapidly developed monovalent influenza A/H1N1 vaccines during pregnancy, pregnant women around the world were prioritised for H1N1 vaccination and were strongly advised to be vaccinated. 22,23 Enhanced surveillance of pregnant women during the pandemic, combined with higher than usual vaccination rates, provided a unique opportunity to study influenza vaccination during pregnancy using observational designs. Since April 2012, a number of comparative analytical studies of perinatal outcomes following pandemic H1N1 vaccination during pregnancy have been published reflecting obstetric populations around the world. This recently expanded evidence base also includes several new publications addressing trivalent inactivated influenza vaccines (TIVs) administered to pregnant women. Our objective was to systematically identify and synthesise findings from these studies to assess whether vaccination during pregnancy for the prevention of maternal influenza illness is associated with any increased risk of early fetal death (i.e. miscarriage), late fetal death (i.e. stillbirth) or preterm birth. Methods Prior to commencing this review, we registered a protocol specifying the exposures and outcomes of interest, study inclusion criteria and analytical methods (PROSPERO ID #CRD ). 27 We prepared the manuscript according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting recommendations. 28 Search strategy and selection criteria We initially searched the following bibliographic databases between inception and May 2013: MEDLINE (via Ovid, including In-Process & Other Non-Indexed Citations), EM- BASE, CINAHL, ProQuest Dissertations & Theses Full Text, and DARE. Medical subject headings and keywords related to the population, exposure and outcomes were used to develop the search strategy in MEDLINE in consultation with a medical librarian. A second medical librarian reviewed the strategy using the PRESS standard, 29 and the syntax was then adapted for the other databases (see Supporting Information Appendix S1 for the full search strategy). We then updated the search in April 2014 using MEDLINE only (from 1946 to 7 April 2014), as all but one of the included studies (a PhD dissertation 30 ) identified in the initial search were indexed in MEDLINE. We considered studies for inclusion: (i) if they employed observational (i.e. cohort, case control, cross-sectional) or experimental [i.e. randomised clinical trial (RCT)] designs; (ii) if they compared influenza vaccination during pregnancy with no influenza vaccination during pregnancy; (iii) if the study participants comprised pregnant women; (iv) if the intervention consisted of either seasonal TIV or monovalent H1N1 influenza vaccine; and (v) if the study addressed any of the following outcomes: preterm birth (defined as birth prior to 37 completed weeks of gestation), early fetal death (defined as fetal death prior to 20 weeks of gestation) or late fetal death (defined as fetal death at a gestational age of 20 weeks or greater). Finally, we excluded any study not using a comparison group of non-influenza-vaccinated pregnant women (for example, pharmacovigilance studies, case series) and those not published in English, French or Spanish. Two epidemiologists (D.B.F. and A.L.) independently screened the titles and abstracts of all records to identify studies for full text review. Each then screened the full text of the remaining articles and excluded those that did not fully meet the eligibility criteria. Disagreements were resolved through discussion and consensus. Data extraction and assessment of methodological quality We developed a standard data collection form to extract study information, including: study design, population, inclusion and exclusion criteria, definition and ascertainment of exposure and outcomes, confounding variables considered in the analysis, and unadjusted and adjusted ratio measures of effect [i.e. relative risks (RRs), hazards ratios (HRs) or odds ratios (ORs)] with 95% confidence intervals (CIs). Each reviewer independently assessed study quality using the Newcastle Ottawa Scale 31 and the Downs Black quality assessment tool. 32 Assessments of study quality were primarily used to augment the interpretation of the results. Data synthesis and analysis We first summarised the descriptive characteristics of all studies, including study location, design, exposure and outcome definitions, and unadjusted and adjusted effect estimates. Outcome-specific Forest plots were generated to visually display individual study-adjusted effect estimates with their 95% CIs. To determine whether the meta-analysis of each outcome was appropriate, we considered clinical heterogeneity (e.g. clinical populations, outcome definitions, vaccine formulations), design heterogeneity (e.g. study design, analytical approach) 33 and statistical heterogeneity. The latter was measured by the I 2 statistic generated from a random effects model, 34 with the 95% CI around the I 2 statistic computed using the non-central v 2 18 ª 2014 Royal College of Obstetricians and Gynaecologists

3 Pregnancy outcomes and maternal influenza vaccination method. 35 Outcomes with high statistical heterogeneity (i.e. >75%) were deemed to be inappropriate for meta-analysis; however, we attempted to explain heterogeneity using random-effects meta-regression on study-level variables (i.e. vaccine type, study quality) when ten or more studies were available for a given outcome. 36,37 As a result of concerns about bias in observational studies of influenza vaccine effectiveness, 38 we only considered studies that reported an effect measure addressing confounding in the design (e.g. matching or randomisation) or analysis (e.g. multivariable regression) for possible meta-analysis. STATA SE 12.1 software (STATA-Corp LP, College Station, TX, USA) was used to generate Forest plots and for all analyses. Results Bibliographic database searches identified 4446 titles and abstracts, corresponding to 2199 unique records following de-duplication. Of these, 1544 were not relevant to our review, 472 were editorials (e.g. commentaries and annual influenza vaccination recommendations), 102 were review articles, 30 had no comparison group (e.g. pharmacovigilance study) or an inappropriate comparison group (e.g. comparison of one influenza vaccine directly with another 39 ) and five did not meet language criteria. We retrieved the full text for the remaining 45 articles, of which a further 18 were excluded for the reasons provided in Figure 1. Our full-text review included two conference abstracts for which we contacted the lead author to determine whether there were journal articles reporting the same study. One was a clinical trial of influenza vaccination during the third trimester in Mali (ClinicalTrials.gov identifier: NCT ), which was underway but incomplete, 40 and the second was a Canadian prospective cohort study that had not yet been published. 41 Both were excluded for having insufficient information (Figure 1). Four other conference abstracts were excluded as duplicates, as all were superseded by journal articles that had already been included in the full-text review Finally, one PhD dissertation was identified and ultimately retained in our review after the author confirmed that the results had not been published elsewhere. 30 We included 27 studies, four of which were published prior to 2012, and the remainder between 2012 and April One study was an RCT, 43 three used a case control design 30,50,51 and 23 used cohort designs. Studies originated from 13 countries, with the majority performed in the USA (Table S3). Fourteen studies were of monovalent H1N1 vaccines, 8,44,45,52 62 ten examined TIVs, 30,42,43,46 50,63,64 and three 51,65,66 did not distinguish between TIVs (with or without H1N1) and monovalent H1N1 vaccines (Table S1). Among the 26 observational studies, scores for the Newcastle Ottawa Scale were in the range 5 9 (median 8.5; Table S3). The scores for the Downs Black quality assessment tool were in the range with a median value of 25. The major area of concern for the three studies with a low Downs Black score (<17) related to confounding (none of the three studies provided adjusted estimates of effect for the outcomes included in this review). 47,55,56 Association between influenza vaccination during pregnancy and fetal death Fetal death was reported in 15 publications (Table S4); however, synthesis of results across all 15 studies was not possible because of the high variability in the gestational age ranges underlying the outcome definition. To improve comparability, we used three categories based on similar gestational ages: (i) fetal death occurring at any time Figure 1. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram showing study selection process. ª 2014 Royal College of Obstetricians and Gynaecologists 19

4 Fell et al. during gestational follow-up; (ii) early fetal death (i.e. prior to 20 weeks); and (iii) late fetal death (i.e. at or after 20 weeks). Four cohort studies explored fetal death occurring at any time during gestational follow-up. 8,45,55,62 Crude and adjusted results for the four studies are shown in Table S4. Notwithstanding different ascertainment of fetal deaths (i.e. identified at 6 weeks by Pasternak et al. 45 and at 12 weeks by Haberg et al. 8 ), there was a high level of consistency in these two large studies, as both were performed in northern Europe, used similar statistical analyses and studied AS03-adjuvanted monovalent H1N1 vaccine from the same manufacturer. Adjusted HRs from these two studies were also similar (0.79; 95% CI, ; ; 95% CI, ) and did not suggest any increased risk of fetal loss following influenza vaccination. The third study reporting adjusted effect estimates also studied monovalent H1N1 vaccine in a European setting; however, most women (93%) received an unadjuvanted vaccine and the gestational age underlying their definition of fetal death was not provided. 62 Despite the similarity in the three studies with respect to the strain contained in the influenza vaccine and in the study design, the assessment of statistical heterogeneity was highly uncertain (I 2 = 0%; 95% CI, 0 68%) and we were unable to explore the reasons for heterogeneity owing to the small number of studies. We opted not to compute a pooled summary estimate; however, the adjusted effect estimates from the three studies are shown in Figure 2. Eight studies addressed early fetal death: three specified lower and upper gestational age bounds for the outcome, 44,45,50 three specified an upper bound only 47,65,66 and two did not provide a definition. 56,58 No adjusted effect estimates were provided in two of the studies, 47,65 and no events were recorded in another. 56 Of the remaining five studies with adjustment for confounding variables, four were cohort studies of H1N1-containing vaccines (either monovalent or within a TIV), and the fifth used a case control design to study TIVs. 50 Despite our classification scheme for early fetal death predicated on gestational age, a high level of heterogeneity in the outcome definitions in these five studies remained (Figure S1), suggesting that a pooled estimate across studies would be misleading. As the assessment of statistical heterogeneity was also uncertain (I 2 = 0%; 95% CI, 0 64%), we ultimately did not meta-analyse these studies. One of the studies reported a reduced risk of fetal death between 9 and 12 weeks of gestation (adjusted HR = 0.74; 95% CI, ); 44 however, the point estimates from the remaining studies generally fell close to the null value with CIs that crossed the line of no effect (Figure S1). We identified 11 studies of late fetal death: two investigated TIVs, 47,63 seven studied monovalent H1N1 vaccines 44,45,52 54,56,58 and two included both vaccine types. 65,66 Three of the studies did not report adjusted effect measures, 47,63,65 and three did not compute any measure of effect because of insufficient events, 56,58,66 leaving five studies of monovalent H1N1 vaccines for possible Figure 2. Individual study-adjusted effect estimates for fetal death at any time during gestational follow-up associated with influenza vaccination during pregnancy. Small, black diamond markers indicate individual study estimates, with corresponding 95% confidence intervals (CIs) represented by horizontal bars. *Adjusted estimate from original study. HR, hazards ratio. H1N1, monovalent H1N1 influenza vaccine. 20 ª 2014 Royal College of Obstetricians and Gynaecologists

5 Pregnancy outcomes and maternal influenza vaccination meta-analysis. Two of the five studies produced adjusted effect estimates indicating that influenza vaccination was beneficial, with CIs not including the null value, 45,52 whereas two showed a weaker, non-significant effect in the same direction. 44,54 The final study suggested an increased risk of fetal death; however, the CI was very imprecise, as the analysis was based on only three exposed cases and two non-exposed cases 53 (Figure 3). The extent of betweenstudy statistical variability was uncertain (I 2 = 0%; 95% CI, 0 64%), and clinical heterogeneity was felt to be rather high as a result of different vaccine components (i.e. adjuvanted versus unadjuvanted), outcome definitions and analytical methods; thus, no pooled estimate was computed. Association between influenza vaccination during pregnancy and preterm birth Of the 27 studies in this review, 24 assessed preterm birth, all using a definition of birth prior to 37 weeks of gestation, with some studies further limiting their definition to live births or singleton live births (Tables S1 and S5). Five studies did not report an adjusted effect estimate and were not considered for possible meta-analysis. 46,47,55,56,63 Among those remaining, ten explored monovalent H1N1 vaccines, 8,52 54,57 62 six addressed TIVs 30,42,43,48,49,63,64 and three studies included both (Figure 4). 51,65,66 Qualitatively, 18 of the 19 studies showed no evidence of harmful effects: the adjusted effect estimates ranged from 0.63 to 1.20, among which four studies reporting a decreased risk of preterm birth had CIs that did not include the null value. 54,60,61,64 One of the 19 studies reported an increased risk of preterm birth associated with influenza vaccination. 66 The risk increase was large (HR = 3.28; 95% CI, ); however, the CI was very wide and the authors noted that the average decrease in gestational length was 3 days, which they posited may not be of great clinical consequence. 66 The 1032 women in this study were sampled propsectively from a consultation service regarding drug or vaccine exposure during pregnancy, and thus their baseline risk for preterm birth may have been different from that of the general population. It is reassuring that even when including this study in the random-effects model used to assess statistical heterogeneity, the point estimate pooled across the 19 studies showed no evidence of harmful effects (Table S2). The between-study estimate of heterogeneity was moderate, with CI bounds ranging from low to high heterogeneity (I 2 = 57%; 95% CI, 18 73%). As we could not exclude high statistical heterogeneity 35 and had a sufficient number of studies, we explored the heterogeneity further through meta-regression. We examined vaccine type (TIV only or TIV/H1N1 combined, versus H1N1 only) and study quality (Downs Black score) as independent variables to determine whether either of these study-level factors could explain any of the between-study heterogeneity in effect estimates. Figure 3. Individual study-adjusted effect estimates for late fetal death associated with influenza vaccination during pregnancy. Small, black diamond markers indicate individual study estimates, with corresponding 95% confidence intervals (CIs) represented by horizontal bars. *Adjusted estimate from original study. HR, hazards ratio; OR, odds ratio; RR, risk ratio. H1N1, monovalent H1N1 influenza vaccine. ª 2014 Royal College of Obstetricians and Gynaecologists 21

6 Fell et al. Figure 4. Individual study-adjusted effect estimates for preterm birth at <37 weeks associated with influenza vaccination during pregnancy. Small, black diamond markers indicate individual study estimates, with corresponding 95% confidence intervals (CIs) represented by horizontal bars. *Adjusted estimate from original study. HR, hazards ratio; OR, odds ratio; POR, prevalence odds ratio; RR, risk ratio. H1N1, monovalent H1N1 influenza vaccine; TIV, trivalent inactivated influenza vaccine. Most of the heterogeneity remained unexplained following meta-regression: the I 2 statistic in the first model decreased to 50% (from 57% in the random-effects meta-analysis), indicating that the type of influenza vaccine explained only a very small amount of the heterogeneity between studies. In the second model, the I 2 statistic did not decrease (it increased by 1%) and the coefficient for the Downs Black score indicated no relative change in the overall effect with increasing study quality. Figure 4 presents the individual study-adjusted effect estimates, and results for the meta-regression are provided in Table S2. Finally, although funnel plot asymmetry suggests that small studies finding an increased risk of preterm birth may remain unpublished (Figure 5), as the region of the plot in which such studies appear to be missing contains areas of low and high statistical significance, other explanations for the asymmetry (e.g. different study populations or influenza seasons) are also possible. Discussion Main findings In view of the recommendations in many countries that all pregnant women should receive influenza vaccination, the issue of fetal safety is of high importance for healthcare providers, public health officials and the general public. 67 Figure 5. Funnel plot of all 19 studies of preterm birth at <37 weeks associated with influenza vaccination during pregnancy. Only studies that provided adjusted effect estimates for preterm birth and influenza vaccination are included in the funnel plot. In this systematic review, we summarised results from 27 studies of over pregnant women. Overall, we found no strong evidence indicating an increased risk of fetal death or preterm birth associated with influenza vaccination during pregnancy. On the contrary, studies generally reported either no association, or modestly decreased risks; however, the latter findings must be interpreted alongside 22 ª 2014 Royal College of Obstetricians and Gynaecologists

7 Pregnancy outcomes and maternal influenza vaccination known shortcomings of observational studies of influenza vaccine effectiveness. Strengths and limitations This review provides a comprehensive narrative summary of comparative analytical studies on a topic of high relevance to care providers. The reviewed studies, published mostly between 2012 and 2014, included several large, population-based observational studies investigating the relationship between influenza vaccination and fetal death, an outcome which would be too rare to investigate in an experimental study. 68 Nevertheless, our review has a number of limitations. We noted considerable clinical, design and statistical heterogeneity across the included studies, which ultimately precluded meta-analysis. In addition, with the exception of the preterm birth outcome, the small number of studies prevented the exploration of heterogeneity through meta-regression. Although studies of monovalent H1N1 vaccines represented about one-half of the studies in our review, they dominated our synthesis of studies providing adjusted measures of effect. In the meta-regression of preterm birth, there was no strong evidence suggesting that the type of vaccine explained between-study differences in effects. Although the monovalent H1N1 vaccines differed from usual TIV formulations, their safety profile was expected to be similar to seasonal TIVs, given the similar production processes. 69 However, pregnant women in many countries received an H1N1 vaccine containing an immune-boosting adjuvant (Table S1), which is not typically found in TIVs. We were unable to accurately characterise adjuvant exposure in many of the H1N1 studies in order to explore this as a source of clinical heterogeneity. The interpretation of any apparent protective effect of maternal influenza immunisation on the examined outcomes as causal is constrained by the limitations underlying observational studies. Finally, we could not fully dismiss the possibility of publication bias, in which small studies reporting an increased risk of preterm birth following influenza vaccination may remain unpublished. Interpretation Inoculation against influenza has been a public health strategy for many years, 18,70 and is considered to be the most effective means of preventing influenza and mitigating severe illness. 13,18,70 Theoretically, prevention of maternal influenza through vaccination could benefit the fetus by averting symptoms such as hyperthermia, 3 or consequences of illness influencing placental function. 4 However, this theoretical benefit is only plausible insofar as there is an independent risk of adverse pregnancy outcomes attributable to maternal influenza. Although epidemiological evidence concerning the effect of non-pandemic influenza on risks to the fetus has been inconsistent, 9 12 during the influenza pandemics of the 20th century, infected pregnant women experienced higher rates of pregnancy loss compared with non-infected pregnant women, 5 and increases in other adverse outcomes, such as preterm birth and birth defects, were also reported, 71 although not consistently. 72 The more recent literature pertaining to the H1N1 pandemic is dominated by case series reports; 73 however, two cohort studies found that maternal influenza infection was an independent risk factor for fetal death during the pandemic, 7,8 lending some support for the plausibility of studies reporting a reduced risk of fetal death, particularly as such studies all investigated monovalent H1N1 vaccines during the pandemic. Nevertheless, any apparent protective effect of influenza vaccination on the risk of fetal death observed during the pandemic, even if real, may not be evident during seasonal influenza epidemics with lower morbidity or with a poor match between the vaccine and the circulating virus. 74 Apart from the possibility that maternal influenza immunisation can truly prevent fetal death or preterm birth, another compelling explanation for apparent protective findings is bias. All but one of the studies in our review used non-randomised observational designs, raising concerns about the quality of the evidence, particularly in the light of treatment selection factors known to bias estimates of influenza vaccine effectiveness in observational studies in the elderly population. 38,75 Two studies in our review, 43,49 including the RCT, 43 attempted to assess the impact of bias by evaluating the association between influenza vaccination and study outcomes during time periods defined by viral activity. 38,75 A null association during the pre-influenza season and a reduction in preterm birth during the putative influenza season were observed by both studies. 43,49 This finding, together with the concordance between the observational study 49 and the RCT, 43 make it less likely that treatment selection bias is the sole explanation for the risk reduction; however, the magnitude of the reduction (72%) in preterm birth risk during the time period of high viral circulation in the observational study by Omer et al. 49 seems implausibly large, given that maternal influenza illness has not been shown to be strongly associated with preterm birth, and TIV efficacy is estimated to be only about 59%. 74 Most other studies in this review relied on conventional approaches to address confounding bias, such as multivariable adjustment and matching. Although these adjustment strategies yielded estimates that were generally similar to unadjusted estimates or closer to the null value compared with unadjusted estimates (unlike studies in the elderly population in which the bias often increases with adjustment for confounding variables 38 ), treatment selection bias in studies of the obstetric population remains a possibility. With the exception of one study, 66 it is reassuring that the individual ª 2014 Royal College of Obstetricians and Gynaecologists 23

8 Fell et al. study results for preterm birth are generally in agreement with findings from the only published RCT on this topic. 43 Even so, one should be aware of the methodological limitations underlying the observational studies when interpreting these findings. Conclusions Although observational studies of influenza vaccination during pregnancy have important limitations, our systematic review results do not indicate any increased risk of fetal death or preterm birth associated with maternal influenza vaccination. The general agreement between observational studies of preterm birth and the single RCT on this topic is also reassuring. The results from another ongoing RCT may become available in the near future and add to this evidence base if they report on perinatal outcomes such as preterm birth in addition to infectious and immunological endpoints. In the meantime, the certainty that observational studies will continue to play a central role underscores the importance of developing robust datasets and methodologies that can overcome major biases and reduce the high clinical and design heterogeneity between studies. This would particularly benefit future research on the relationship between early fetal death and influenza vaccination, an area in which high-quality evidence is lacking. Disclosure of interests The authors have no conflicts of interest to declare. Contribution to authorship D.B.F. developed the study protocol, developed the search strategy, conducted the literature search, screened articles, extracted data from the included studies, analysed the data, interpreted the results, and drafted the manuscript. R.W.P. developed the study protocol, provided statistical advice, interpreted the results, and edited the manuscript. A.L. screened articles, extracted data from the included studies, and edited the manuscript. K.W., J.S.K., O.B. and D.B. interpreted the results and edited the manuscript. All authors critically reviewed the article for intellectual content and approved the final version. Details of ethics approval This study was approved by the McGill University Faculty of Medicine Institutional Review Board (#A03-M44-13B). Funding This study did not receive any external funding. Acknowledgements We acknowledge Genevieve Gore, MLIS, Medical Librarian at McGill University, for her contribution to developing the search strategy, and Margaret Sampson, MLIS, PhD, AHI, Medical Librarian at the Children s Hospital of Eastern Ontario, for peer-reviewing the search strategy and providing assistance with updating the search. D.B.F. is supported by a Canadian Institutes of Health Research (CIHR) Doctoral Award. R.W.P. is supported by a Chercheur-national (National Scholar) award from the Fonds de Recherche du Quebec Sante (FRQ-S) as well as core support to the Research Institute of the McGill University Health Centre from FRQ-S. A.L. is supported by an Ontario Graduate Scholarship. Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. Individual study-adjusted effect estimates for early fetal death associated with influenza vaccination during pregnancy. Table S1. Exposure and outcome information from the 27 included studies. Table S2. Results for random-effects meta-regression analyses for preterm birth outcome. Table S3. General characteristics of the 27 included studies. Table S4. Results of studies on fetal death. Table S5. Results of studies on preterm birth. Appendix S1. Primary search strategy. & References 1 Pazos M, Sperling RS, Moran TM, Kraus TA. The influence of pregnancy on systemic immunity. Immunol Res 2012;54: Jamieson DJ, Theiler RN, Rasmussen SA. Emerging infections and pregnancy. Emerg Infect Dis 2006;12: Steinhoff MC, MacDonald NE. Influenza pandemics pregnancy, pathogenesis, and perinatal outcomes. J Am Med Assoc 2012;308: Steinhoff MC, Omer SB. A review of fetal and infant protection associated with antenatal influenza immunization. Am J Obstet Gynecol 2012;207:S Harris J. Influenza occurring in pregnant women. J Am Med Assoc 1919;72: Freeman DW, Barno A. Deaths from Asian influenza associated with pregnancy. Am J Obstet Gynecol 1959;78: Pierce M, Kurinczuk JJ, Spark P, Brocklehurst P, Knight M. Perinatal outcomes after maternal 2009/H1N1 infection: national cohort study. BMJ 2011;342:d Haberg SE, Trogstad L, Gunnes N, Wilcox AJ, Gjessing HK, Samuelsen SO, et al. Risk of fetal death after pandemic influenza virus infection or vaccination. N Engl J Med 2013;368: Hartert TV, Neuzil KM, Shintani AK, Mitchel EF, Snowden MS, Wood LB, et al. Maternal morbidity and perinatal outcomes among pregnant women with respiratory hospitalizations during influenza season. Am J Obstet Gynecol 2003;189: McNeil SA, Dodds LA, Fell DB, Allen VM, Halperin BA, Steinhoff MC, et al. Effect of respiratory hospitalization during pregnancy on infant outcomes. Am J Obstet Gynecol 2011;204:S ª 2014 Royal College of Obstetricians and Gynaecologists

9 Pregnancy outcomes and maternal influenza vaccination 11 Irving WL, James DK, Stephenson T, Laing P, Jameson C, Oxford JS, et al. Influenza virus infection in the second and third trimesters of pregnancy: a clinical and seroepidemiological study. BJOG 2007;107: Banhidy F, Acs N, Puho EH, Czeizel AE. Maternal acute respiratory infectious diseases during pregnancy and birth outcomes. Eur J Epidemiol 2008;23: National Advisory Committee on Immunization (NACI). Statement on seasonal influenza vaccine for Can Commun Dis Rep 2011;37: Gruslin A, Steben M, Halperin S, Money DM, Yudin MH. Immunization in pregnancy. JObstetGynaecolCan2009;31: ACOG Committee Opinion Number 305, November Influenza vaccination and treatment during pregnancy. Obstet Gynecol 2004;104: Mak T, Mangtani P, Leese J. Influenza vaccination in pregnancy: current evidence and selected national policies. Lancet Infect Dis 2008;8: Luteijn JM, Dolk H, Marnoch GJ. Differences in pandemic influenza vaccination policies for pregnant women in Europe. BMC Public Health 2011;11: Artenstein AW. Influenza. In: Artenstein AW, editor. Vaccines: A Biography. New York, NY: Springer; pp Fiore AE, Uyeki TM, Broder K, Finelli L, Euler GL, Singleton JA, et al. Prevention and control of influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP), Morb Mortal Wkly Rep 2010;59: Schatz M, Chambers CD, Jones KL, Louik C, Mitchell AA. Safety of influenza immunizations and treatment during pregnancy: the Vaccines and Medications in Pregnancy Surveillance System. Am J Obstet Gynecol 2011;204:S MacDonald NE, Riley LE, Steinhoff MC. Influenza immunization in pregnancy. Obstet Gynecol 2009;114: Centers for Disease Control. Use of influenza A (H1N1) 2009 monovalent vaccine: recommendations of the Advisory Committee on Immunization Practices (ACIP), Morb Mortal Wkly Rep 2009;58: World Health Organization. Pregnancy and pandemic influenza A (H1N1) 2009: Information for programme managers and clinicians [ pregnancy.pdf]. Accessed 11 January Centers for Disease Control. Seasonal influenza and 2009 H1N1 influenza vaccination coverage among pregnant women 10 states, influenza season. Morb Mortal Wkly Rep 2010;59: Liu N, Sprague AE, Yasseen AS III, Fell DB, Wen SW, Smith GN, et al. Vaccination patterns in pregnant women during the 2009 H1N1 influenza pandemic: a population-based study in Ontario, Canada. Can J Public Health 2012;103: Gilmour H, Hofmann N. H1N1 vaccination. Health Rep 2010;21: Stewart L, Moher D, Shekelle P. Why prospective registration of systematic reviews makes sense. Syst Rev 2012;1:7. 28 Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ 2009;339:b McGowan J, Sampson M, Lefebvre C. Evidence based checklist for the peer review of electronic search strategies (PRESS EBC). Evid Based Libr Inf Pract 2010;5: [ Accessed 11 January Ahrens K. Consequences of controlling viral diseases during pregnancy: antiherpetic medication and influenza vaccination. PhD Dissertation, Boston University, 2012: Wells GA, Shea B, O Connell D, Peterson J, Welch V, Losos MTP. The Newcastle Ottawa Scale (NOS) for Assessing the Quality of Nonrandomized Studies in Meta-Analyses. Ottawa, ON: Ottawa Hospital Research Institute. [ epidemiology/oxford.htm]. Accessed 11 January Downs SH, Black N. The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and non-randomised studies of health care interventions. J Epidemiol Community Health 1998;52: Deeks JJ, Higgins JPT, Altman DG, on behalf of the Cochrane Statistical Methods Group. Analysing data and undertaking meta-analyses. In: Higgins JPT, Green S, editors. Cochrane Handbook for Systematic Reviews of Interventions, Version (updated March 2011). The Cochrane Collaboration; pp [ Accessed 11 January Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ 2003;327: Ioannidis JPA, Patsopoulos NA, Evangelou E. Uncertainty in heterogeneity estimates in meta-analyses. BMJ 2007;335: Thompson SG, Higgins JPT. How should meta-regression analyses be undertaken and interpreted? Stat Med 2002;21: Ioannidis JP, Patsopoulos NA, Rothstein HR. Reasons or excuses for avoiding meta-analysis in forest plots. BMJ 2008;336: Nelson JC, Jackson ML, Weiss NS, Jackson L. New strategies are needed to improve the accuracy of influenza vaccine effectiveness estimates among seniors. J Clin Epidemiol 2009;62: Conlin AMS, Bukowinski AT, Sevick CJ, DeScisciolo C, Crum-Cianflone NF. Safety of the pandemic H1N1 influenza vaccine among pregnant U.S. military women and their newborns. Obstet Gynecol 2013;121: Doumbia M, Teguete I, Orenstein EW, Orenstein LA, Dembele S, Traore A, et al. Abstract 236: Rates and causes of serious adverse events in the first 25 weeks of a clinical trial of maternal influenza vaccine during the third trimester in Bamako, Mali. 61st Annual Meeting of the American Society of Tropical Medicine and Hygiene, Atlanta, GA, [ Meeting_Archives&Template=/CM/ContentDisplay.cfm&ContentID= 4683]. Accessed 14 April Guo Y, Allen V, Bujold E, Coleman B, Drews S, Gouin K, et al. Abstract P087: efficacy and safety of pandemic influenza vaccine in pregnancy. 9th Canadian Immunization Conference, Quebec City, QC, Can J Infect Dis Med Microbiol 2010;21: Dodds L, Macdonald N, Scott J, Spencer A, Allen VM, McNeil S. The association between influenza vaccine in pregnancy and adverse neonatal outcomes. J Obstet Gynaecol Can 2012;34: Steinhoff MC, Omer SB, Roy E, El Arifeen S, Raqib R, Dodd C, et al. Neonatal outcomes after influenza immunization during pregnancy: a randomized controlled trial. Can Med Assoc J 2012;184: Sammon CJ, Snowball J, McGrogan A, de Vries CS. Evaluating the hazard of foetal death following H1N1 influenza vaccination; a population based cohort study in the UK GPRD. PLoS One 2012;7: e Pasternak B, Svanstr om H, Mølgaard-Nielsen D, Krause TG, Emborg H-D, Melbye M, et al. Vaccination against pandemic A/H1N influenza in pregnancy and risk of fetal death: cohort study in Denmark. BMJ 2012;344:e Black SB, Shinefield HR, France EK, Fireman BH, Platt ST, Shay D. Effectiveness of influenza vaccine during pregnancy in preventing hospitalizations and outpatient visits for respiratory illness in pregnant women and their infants. Am J Perinatol 2004;21: ª 2014 Royal College of Obstetricians and Gynaecologists 25

10 Fell et al. 47 Deinard AS, Ogburn P. A/NJ/8/76 influenza vaccination program: effects on maternal health and pregnancy outcome. Am J Obstet Gynecol 1981;140: Munoz FM, Greisinger AJ, Wehmanen OA, Mouzoon ME, Hoyle JC, Smith FA, et al. Safety of influenza vaccination during pregnancy. Am J Obstet Gynecol 2005;192: Omer SB, Goodman D, Steinhoff MC, Rochat R, Klugman KP, Stoll BJ, et al. Maternal influenza immunization and reduced likelihood of prematurity and small for gestational age births: a retrospective cohort study. PLoS Med 2011;8:e Irving SA, Kieke BA, Donahue JG, Mascola MA, Baggs J, Destefano F, et al. Trivalent inactivated influenza vaccine and spontaneous abortion. Obstet Gynecol 2013;121: Louik C, Ahrens K, Kerr S, Pyo J, Chambers C, Jones KL, et al. Risks and safety of pandemic H1N1 influenza vaccine in pregnancy: exposure prevalence, preterm delivery, and specific birth defects. Vaccine 2013;31: Fell DB, Sprague AE, Liu N, Yasseen AS, Wen SW, Smith G, et al. H1N1 influenza vaccination during pregnancy and fetal and neonatal outcomes. Am J Public Health 2012;102:e Heikkinen T, Young J, van Beek E, Franke H, Verstraeten T, Weil JG, et al. Safety of MF59-adjuvanted A/H1N1 influenza vaccine in pregnancy: a comparative cohort study. Am J Obstet Gynecol 2012;207:177.e K allen B, Olausson P. Vaccination against H1N1 influenza with Pandemrix( â ) during pregnancy and delivery outcome: a Swedish register study. BJOG 2012;1190: Launay O, Krivine A, Charlier C, Truster V, Tsatsaris V, Lepercq J, et al. Low rate of pandemic A/H1N influenza infection and lack of severe complication of vaccination in pregnant women: a prospective cohort study. PLoS One 2012;7:e Lin T-H, Lin S-Y, Lin C-H, Lin R-I, Lin H-C, Chiu T-H, et al. AdimFlu-S ( â ) influenza A (H1N1) vaccine during pregnancy: the Taiwanese Pharmacovigilance Survey. Vaccine 2012;30: Ludvigsson JF, Zugna D, Cnattingius S, Richiardi L, Ekbom A, Ortqvist A, et al. Influenza H1N1 vaccination and adverse pregnancy outcome. Eur J Epidemiol 2013;28: Oppermann M, Fritzsche J, Weber-Schoendorfer C, Keller-Stanislawski B, Allignol A, Meister R, et al. A(H1N1)v2009: a controlled observational prospective cohort study on vaccine safety in pregnancy. Vaccine 2012;30: Pasternak B, Svanstr om H, Mølgaard-Nielsen D, Krause TG, Emborg H-D, Melbye M, et al. Risk of adverse fetal outcomes following administration of a pandemic influenza A(H1N1) vaccine during pregnancy. J Am Med Assoc 2012;308: Richards JL, Hansen C, Bredfeldt C, Bednarczyk RA, Steinhoff MC, Adjaye-Gbewonyo D, et al. Neonatal outcomes after antenatal influenza immunization during the 2009 H1N1 influenza pandemic: impact on preterm birth, birth weight, and small for gestational age birth. Clin Infect Dis 2013;56: Rubinstein F, Micone P, Bonotti A, Wainer V, Schwarcz A, Augustovski F, et al. Influenza A/H1N1 MF59 adjuvanted vaccine in pregnant women and adverse perinatal outcomes: multicentre study. BMJ 2013;346:f Beau AB, Hurault-Delarue C, Vidal S, Guitard C, Vayssiere C, Petiot D, et al. Pandemic A/H1N1 influenza vaccination during pregnancy: a comparative study using the EFEMERIS database. Vaccine 2014;32: Sheffield JS, Greer LG, Rogers VL, Roberts SW, Lytle H, McIntire DD, et al. Effect of influenza vaccination in the first trimester of pregnancy. Obstet Gynecol 2012;120: Legge A, Dodds L, Macdonald NE, Scott J, McNeil S. Rates and determinants of seasonal influenza vaccination in pregnancy and association with neonatal outcomes. Can Med Assoc J 2014;186: E Cantu J, Biggio J, Jauk V, Wetta L, Andrews W, Tita A. Selective uptake of influenza vaccine and pregnancy outcomes. J Matern Fetal Neonatal Med 2013;7058: Chambers CD, Johnson D, Xu R, Luo Y, Louik C, Mitchell AA, et al. Risks and safety of pandemic H1N1 influenza vaccine in pregnancy: birth defects, spontaneous abortion, preterm delivery, and small for gestational age infants. Vaccine 2013;31: Jacobson RM, Targonski PV, Poland GA. Why is evidence-based medicine so harsh on vaccines? An exploration of the method and its natural biases. Vaccine 2007;25: Neuzil KM. Influenza vaccination in : achieving 100% participation. J Am Med Assoc 2013;310: Moro PL, Broder K, Zheteyeva Y, Revzina N, Tepper N, Kissin D, et al. Adverse events following administration to pregnant women of influenza A (H1N1) 2009 monovalent vaccine reported to the Vaccine Adverse Event Reporting System. Am J Obstet Gynecol 2011;205: Steinhoff MC. Epidemiology and prevention of influenza. In: Nelson K, Masters Williams C, editors. Infectious Disease Epidemiology. Theory and Practice, 2nd edn. Sudbury, MA: Jones and Bartlett Publishers; pp Hardy JM, Azarowicz EN, Mannini A, Medearis DN, Cooke RE. The effect of Asian influenza on the outcome of pregnancy, Baltimore, Am J Public Health 1961;51: Widelock D, Csizmas L, Klein S. Influenza, pregnancy, and fetal outcome. Public Health Rep 1963;78: Mosby LG, Rasmussen SA, Jamieson DJ pandemic influenza A (H1N1) in pregnancy: a systematic review of the literature. Am J Obstet Gynecol 2011;205: Osterholm MT, Kelley NS, Sommer A, Belongia E. Efficacy and effectiveness of influenza vaccines: a systematic review and meta-analysis. Lancet Infect Dis 2012;12: Simonsen L, Viboud C, Taylor RJ, Miller MA, Jackson L. Influenza vaccination and mortality benefits: new insights, new opportunities. Vaccine 2009;27: ª 2014 Royal College of Obstetricians and Gynaecologists