Background

Pertussis is a vaccine-preventable bacterial respiratory infection leading to high morbidity, especially in young infants. Disease burden of pertussis remains significant despite high vaccination coverage in children in western countries [1]. In Germany, annual pertussis incidence ranged from 11 to 20 per 100,000 inhabitants during the years 2013–2018 [2]. Young infants < 6 months of age are at increased risk of pertussis related complications, such as otitis media, pneumonia, apnea, encephalopathy, as well as pulmonary hypertension which is caused by extreme lymphocytosis [3]. Severe and potentially lethal complications are most common in infants < 2 months of age [4]. In Germany, mean annual incidence of pertussis among infants aged ≤3 months was 80 per 100,000 during the past 5 years, while hospitalization rate in those young infants was > 75% [RKI, surveillance data, unpublished]. A recent German capture-recapture study suggests that incidences based on statutory surveillance are substantially underestimated by 39% [5]. Since the introduction of nationwide mandatory pertussis reporting in Germany in 2013, two pertussis-related deaths were notified in infants 2 and 4 months of age [6] [RKI, surveillance data, unpublished]. Due to young age, this vulnerable group cannot benefit from direct effects of vaccination. Studies have shown that the majority of pregnant women in western countries have insufficient concentrations of pertussis-specific antibodies to confer protection to the newborn via diaplacentally transferred maternal antibodies [7,8,9,10]. In contrast, vaccination during pregnancy results in high levels of antibodies in the mother and the newborn [11, 12]. Therefore, vaccination of pregnant women with an acellular pertussis vaccine has been introduced in a number of countries, including the United Kingdom, USA, Belgium, Switzerland, Spain and Australia [13,14,15,16,17,18,19].

So far, six systematic reviews have investigated the effectiveness and/or safety of pertussis vaccination during pregnancy [4, 20,21,22,23,24]. Importantly, however, a number of new studies on this topic were published only recently, and not all of these reviews addressed the entire spectrum of clinically relevant outcomes comprising safety as well as effectiveness for mother and child. Furthermore, some of the earlier reviews did not use the most advanced methodological tools recommended to address risk of bias and evidence quality, both being of key importance for decision-making regarding the implementation of vaccine programs during pregnancy.

We therefore performed a systematic review assessing the evidence for safety and effectiveness of pertussis vaccination during pregnancy.

Methods

Search strategy and selection criteria

The protocols of this systematic review were published in the Prospective Register for Systematic Reviews (PROSPERO; registration no, CRD42018087814 (for safety), CRD42018090357 (for effectiveness)). The review was performed according to the guidelines in the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement [25].

To be eligible, a study had to match the following PICO (population, intervention, comparator, outcome) criteria:

P – pregnant women and their newborns.

I – vaccination with an acellular pertussis component-containing vaccine during pregnancy.

C – placebo or no vaccination or vaccination with other, not pertussis component-containing vaccines, e.g. tetanus, tetanus-diphtheria, or influenza vaccination (only for effectiveness outcomes).

O – efficacy/effectiveness: (1) laboratory-confirmed pertussis in infant ≤3 months of age; (2) hospitalization due to (1); (3) death due to (1);

O – safety: (4) fever (≥38 °C) in pregnant woman; (5) pre-eclampsia/eclampsia; (6) chorioamnionitis; (7) preterm birth; (8) stillbirth; (9) low birth weight; (10) malformation; (11) neonatal intensive care unit (NICU) admission; (12) neonatal sepsis; (13) neonatal death.

Electronic databases searched were MEDLINE and EMBASE (date of initial search: 26 February 2018; last update: 10 January 2019). For details on the complete search strategy, see Additional file 1: Figure S1. Additionally, the Cochrane Data Base of Clinical Trials was searched, and a search in ClinicalTrials.gov was conducted for unpublished or ongoing trials. Electronic searches were complemented by manually screening reference lists of all identified studies and those of identified reviews. Search results (titles, abstracts, full texts) were independently assessed by three investigators (WH, TH, SVB). Differences were discussed until a consensus was reached.

Search was limited to studies published from 01 January 2010 onwards. We did not make restrictions with regard to setting, language or publication status (published/unpublished).

Data extraction

Three independent reviewers (WH, TH, SVB) used standardized forms to extract study characteristics from eligible studies and to assess risk of bias. In case of disagreement, a final decision was made by consensus. The following data were extracted: study location, setting, study design, study period, participants, intervention, comparator, study size, outcomes, study sponsorship, conflict of interests, number (proportion) of vaccinated participants with outcome, number (proportion) of control participants with outcome, unadjusted estimates, adjusted estimates, and confounders.

Assessment of risk of bias and quality of evidence

For randomized controlled trials (RCTs), the Cochrane risk of bias tool was used to assess the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias [26]. RCTs were categorized as being at “high risk”, “low risk” or “unclear risk” of bias. For non-randomized studies, the ROBINS-I tool was used, comprising the following domains: bias due to confounding, bias in selection of participants into the study, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in measurement of outcomes, and bias in selection of the reported results [27]. Risk of bias was categorized as being “low risk”, “moderate risk”, “serious risk” or “critical risk”.

The methodology of the GRADE (Grading of Recommendations Assessment, Development and Evaluation) working group and the software “GRADE profiler” were used to assess the quality of evidence [28, 29].

Statistical analysis

Abstracted data were aggregated in tables. Risk ratios, odds ratios, risk differences and corresponding 95% confidence intervals (95% CI) were either calculated or extracted from the publications. A p-value <.05 was considered statistically significant. Vaccine effectiveness (VE) was either extracted from the publications or calculated as [1-(risk ratio or rate ratio comparing vaccine and control recipients)] × 100. Since heterogeneity between studies was judged to be high with regard to setting, study design, outcome definition and confounders considered in the analysis, no meta-analyses were performed.

Results

Search results

By systematic literature search, a total of 1273 publications were identified. Screening of titles and abstracts led to the exclusion of 1074 publications. Of the remaining 199 studies, 22 were found to match our inclusion criteria (see flowchart and list of excluded studies in the Additional file 1: Figure S1 and Table S1). The characteristics of the included studies are listed in Table 1.

Table 1 Characteristics of included studies
Full size table

Vaccine safety

Evidence base and risk of bias

Three RCTs [30,31,32] and 11 non-randomized studies [15, 33,34,35,36,37,38, 40,41,42, 50] from Belgium, United Kingdom, Canada, New Zealand, Vietnam and the USA reported maternal and/or infant safety outcomes (Table 2). Taking into account overlapping study populations of four studies based on the US Vaccine Safety Datalink project [34, 35, 38, 50], data from a total of 1.4 million pregnant women were included, of which 199,846 had received a pertussis-component-containing vaccine during pregnancy. In three RCTs [30,31,32] and one non-randomized study [42] the pertussis-containing vaccine used was Adacel®, whereas in four other studies it was Boostrix® [15, 37, 52, 53] and in the British study [36] it was Repevax®. In most studies from the US [33,34,35, 38, 40, 41, 50] the vaccine used was not specified.

Table 2 Safety outcomes
Full size table

In most studies, women who had received tetanus-diphtheria-acellular pertussis (Tdap) vaccines (Adacel® or Boostrix®) or Tdap-IPV-vaccines (Repevax®), were compared to women, who were either unvaccinated or had received placebo. In two RCTs, the comparison group was vaccinated with a tetanus-toxoid-containing vaccine [31, 32].

Risk of bias (RoB) was judged high for one [31] and low for two RCTs [30, 32] (Additional file 1: Table S2). Vaccine administrators were not blinded in the trials by Halperin et al. [32] and Munoz et al. [30], but we considered this to be unlikely to have influenced the outcomes “prematurity“, “pre-eclampsia/eclampsia“, “fever” and “malformations” in these studies.

Of the 11 non-randomized studies, we judged eight as having a serious RoB and three studies to show a critical RoB (Additional file 1: Table S2). The main reasons for these classifications were confounding, selection bias, and imprecise outcome assessment. Residual confounding could not be excluded in any of the studies. In addition, a likely healthy vaccinee bias was observed in most studies. Preexisting comorbidities (e.g. arterial hypertension [34, 38, 41], heart disease [38], diabetes [34, 41], pulmonary disease [34, 38]) and referral to high-risk obstetrics clinics [41] were more frequent in non-vaccinated women than in vaccinated women. In addition, health care utilization differed between vaccinated and non-vaccinated women (e.g. higher uptake of influenza vaccination [33, 37, 40, 54] and ultrasound examinations [40, 54] during pregnancy among Tdap vaccinated women). In several studies, Tdap-vaccinated women showed indications for better uptake or earlier start of prenatal care [33, 38, 40, 41]. Frequently, these results were statistically significant [33, 38, 41]. Healthy vaccinee bias might have shifted estimates towards more favorable outcomes in vaccinated women and their infants. Moreover, with respect to preterm birth, immortal time bias could also have influenced the results.

In two studies [38, 39], a large proportion (74 and 79% respectively) of eligible study participants was excluded from analysis, e.g., women with irregular health insurance status or with a history of multiple pregnancies, stillbirth or premature birth. This might have limited generalizability of study results. In addition, exclusion of pregnancies ending in stillbirth or abortion might have resulted in selection bias with regard to potentially associated outcomes, like congenital malformations.

Five studies were based on commercial health data bases using ICD codes [34, 35, 38,39,40]. Coding for commercial reasons, such as insurance claims, might be prone to favoring more severe diagnoses. This might be a relevant source of bias for some outcomes, such as chorioamnionitis, but not for others, like admission to NICU. In studies, which were based on medical records [33, 36], no standardized case definitions were used, such as those developed by the Brighton Collaboration [55]. Here, potential misclassification was likely to be non-differential.

Fever

Rates of fever after Tdap vaccination in pregnancy were assessed in four studies [15, 30, 31, 50]. The definition of fever varied considerably across studies (Table 2). Overall, fever following immunization was reported in 0.03 to 3% of pregnant women and occurred more frequently in Tdap-vaccinated women than in control women.

Stillbirth

One RCT [31] and four non-randomized studies [33, 36, 41, 42] assessed stillbirths. None of these studies reported an increased risk in Tdap-vaccinated women.

Neonatal deaths

Two non-randomized studies reported on neonatal death [36, 41]. A few cases were observed but there was no significant association with Tdap vaccination during pregnancy.

Preterm birth

Ten of the included studies reported on preterm birth, mostly defined as gestational age <37 weeks. In the three RCTs [30,31,32], only a few preterm births were observed without differences between vaccinated and control mothers. In seven non-randomized studies [33, 35, 37, 38, 40,41,42], risk of preterm birth was higher in unvaccinated than in Tdap-vaccinated women. In one of these studies [40], compared to no vaccination, Tdap-vaccination at 27–36 weeks gestation was associated with a decreased risk of preterm birth, whereas earlier vaccination (before 27 weeks) was associated with an increased risk (see Table 2).

Low birth weight

Low birth weight (LBW; < 2500 g) or very low birth weight (VLBW; < 1500 g) were assessed in three studies. Donegan et al. [36] reported on intrauterine growth retardation/LBW, Berensen et al. [33] assessed LBW and VLBW separately. Neither study reported an association between Tdap-vaccination during pregnancy and low birth weight. Griffin et al. [37] assessed the outcome “fetal growth restriction” and also found no association with Tdap-vaccination during pregnancy.

Congenital malformations

Definition and recording of congenital malformations varied considerably across the seven studies which reported on this outcome. Hoang et al. [31] did not detect any malformations within 30 days after birth, while Munoz et al. [31] observed one case of pyelectasia in an infant of a Tdap-vaccinated mother and two cases of cardiac malformation in infants whose mothers had not been vaccinated. In five non-randomized studies [15, 33, 34, 41, 42], the authors observed no association between Tdap-vaccination during pregnancy and malformations including infants of mothers who were vaccinated during the first trimester in the study by DeSilva et al. [34].

Neonatal septicaemia

Only one non-randomized study [40] reported on neonatal sepsis, but did not distinguish between early- and late-onset sepsis. Newborns of Tdap-vaccinated mothers were at lower risk of septicemia than newborns of unvaccinated mothers.

Admission to NICU

Two non-randomized studies reported on NICU admission [33, 40]. Risk of NICU admission was lower in newborns of vaccinated mothers than in those of unvaccinated mothers in both studies.

Pre-eclampsia and eclampsia

The six non-randomized studies that reported on pre-eclampsia and eclampsia used heterogeneous definitions for this outcome (see Table 2). Layton et al. [40] observed a slightly decreased risk for pre-eclampsia and eclampsia in women who had been vaccinated after 26 weeks of gestation, compared to unvaccinated women (RR: 0.96; CI: 0.94–0.99). Similar findings were obtained by Griffin et al. [37] for severe pre-eclampsia (RR: 0.61; CI: 0.39–0.94). In the remaining four studies [15, 32, 36, 38], no association between Tdap-vaccination and pre-eclampsia or eclampsia was observed.

Chorioamnionitis

Chorioamnionitis was investigated in six non-randomized studies [33, 35, 37, 38, 40, 41]. All of them reported an increased risk of chorioamnionitis in women who had received Tdap-vaccination during pregnancy (Table 2). Of those, two studies investigated Tdap given at any week of gestation, while in the remaining four studies Tdap vaccination was performed in the third trimester (> = 27 to > = 32 weeks of gestation). In the four largest studies comprising 8178 to 123,780 vaccinated women, the outcome was defined using ICD-codes [35, 37, 38, 40]. One study [41] did not report the outcome definition. Only Berenson et al. [33] used a clinical case definition for identifying chorioamnionitis in electronic medical records. In the six studies, risk ratios ranged from 1.04 (95%CI: 0.98–1.11) to 1.53 (95%CI: 0.80–2.90). Size of risk estimates was unrelated to time point of Tdap vaccination during pregnancy. Estimates were statistically significant in three studies [35, 38, 40].

In order to minimize the influence of health seeking behavior, Layton et al. [40] conducted a subgroup analysis restricting the cohort to pregnant women who were vaccinated against influenza. In this subgroup, the association between Tdap-vaccination and chorioamnionitis was weaker (adjusted RR: 1.09; 95%CI: 1.03–1.15) than in the full cohort analysis (adjusted RR: 1.14; 95%CI: 1.10–1.18). When propensity score adjustment was used, the estimate was no longer statistically significant.

Vaccine effectiveness

Evidence base and risk of bias

Eight studies fulfilled the inclusion criteria for the assessment of vaccine effectiveness (VE), including four cohort studies [16, 43,44,45] and four case-control-studies [46,47,48,49] (Table 2). The populations of the studies by Amirthalingam et al. [43] and Dabrera et al. [47] were included in the study by Amirthalingam et al. [16] (personal communication, Gavin Dabrera, October 17th, 2018). Taking into account this overlap, a total of 855,546 mother-infant-pairs were included in the studies, including 682 pertussis cases in infants < 3 months (thereof 257 < 2 months) of age and 854,864 non-cases. Mothers of 84 cases (12%) and 205,919 non-cases (24%) were vaccinated.

Five [16, 43, 45, 47, 49] of the eight studies were judged as having a serious RoB, while three studies had a moderate RoB [44, 46, 48]. The main reasons for these classifications were selection bias, and imprecise outcome definitions. Using the screening method, Amirthalingam et al. [16, 43] could not adjust estimates for confounders other than age and time period. Residual confounding was judged likely to be present in the other studies as well, as adjustment for potential confounders was limited (see below). Therefore, and since there was evidence that vaccinated women had a more favorable health profile than those not vaccinated (e.g., uptake of influenza vaccination and ultrasound examination during pregnancy were more frequent in Tdap vaccinated women [45], smokers were more frequent in households of non-vaccinated women [46, 48]), a “healthy vaccinee bias” appeared likely, suggesting that VE based on these studies might be overestimated.

The study by Skoff et al. [49] was judged as having a serious RoB because two-thirds of the initially identified study population was excluded and evidence for selection bias was found (the level of education and geographical distribution of study participants and excluded population differed significantly). The most common reasons for exclusion were non-reachability and missing consent to participate. We judged the most recent cohort study from the USA by Becker-Dreps et al. [45] as having a serious RoB because no clear case definition based on laboratory criteria was reported and the proportion of lab-confirmed cases in the subgroups was unknown.

Pertussis in infants 0–3 months of age

Three cohort [16, 43, 44] and two case-control studies [47, 49] reported on the effectiveness of Tdap vaccination in pregnancy to prevent pertussis in infants 0–2 months of age (Table 3). In all studies except for the one by Skoff et al. [49], only laboratory confirmed cases were included. Skoff et al. [49] also included cases with an epidemiological link or with a clinical picture of pertussis in their analysis (6% of all cases). Confounder-adjusted VE estimates in these five studies ranged from 78 to 93%.

Table 3 Vaccine effectiveness outcomes
Full size table

Four studies - the aforementioned two UK-based cohort studies by Amirthalingam et al. [16, 43] and two case-control studies from Spain and Australia [46, 48] - reported vaccine effectiveness estimates for prevention of laboratory-confirmed pertussis in infants 0–3 months of age between 69 und 91% (Table 3). The age and time-period-adjusted point estimates of both studies by Amirthalingam et al. [16, 43] and the adjusted point estimate of the study by Bellido-Blasco et al. [46] were all 91%. Bellido-Blasco et al. [46] adjusted their analysis for breastfeeding, maternal level of education and presence of other children in the household. The VE estimate of Saul et al. [48] of 69% (95%CI: 13–89%) was adjusted for breastfeeding, household size and gestational age.

Pertussis-related hospitalization in infants 0–3 months of age

The confounder-adjusted VE estimates for the prevention of hospitalization due to pertussis in infants were 91 and 94% in two case-control studies from the US (California) [49] and Australia [48], respectively. Saul et al. [48] included infants ≤3 months of age, whereas Skoff et al. [49] focused on infants ≤2 months of age. The case definition used by Saul et al. [49] included only laboratory confirmed cases. Skoff et al. [49] included 6% clinical cases without laboratory confirmation (Table 3).

In the recent US American cohort study [45] which reported a VE of 66% for the prevention of hospitalization due to pertussis in infants ≤2 months of age, the percentage of laboratory confirmed cases was unclear.

Pertussis-related deaths in infants aged 0–3 months

One cohort study from the United Kingdom [16] reported an effectiveness of 95% for Tdap vaccination in pregnancy for the prevention of death due to laboratory confirmed pertussis in infants 0–3 months of age (Table 3). Due to the use of the screening method, only an unadjusted effect estimate could be reported, which was based on 11 cases.

Quality of evidence

Regarding safety outcomes, quality of evidence according to GRADE was judged as low to very low. Reasons for downgrading were serious to critical risk of bias and imprecision. The evidence related to three of four effectiveness outcomes (pertussis < 2 months of age, pertussis < 3 months of age, pertussis-related death) was also downgraded to low quality due to moderate to serious risk of bias and inconsistency. Evidence quality for the remaining effectiveness outcome (hospitalization) was assessed as moderate due to risk of bias (see the GRADE evidence profile in Additional file 1: Table S3 for details).

Discussion

In this comprehensive systematic review, we evaluated the safety and effectiveness of acellular pertussis vaccination during pregnancy. Using data from more than 1.4 million pregnancies, we found similar risks for all pre-specified safety outcomes in vaccinated and unvaccinated women and their infants except for two: slightly increased relative risks were detected for post-vaccination fever and chorioamnionitis at the time of delivery after Tdap vaccination in all studies reporting these outcomes. The risk increase was significant in one study reporting on maternal fever [50] and in three studies reporting on chorioamnionitis [35, 38, 40]. High effectiveness of Tdap-vaccination during pregnancy in preventing pertussis and related complications in the newborn and young infant was observed in all studies. Using GRADE methodology, the overall quality of evidence was rated as moderate to very low, depending on the outcome category.

Fever is a well-known adverse event after Tdap vaccination occurring at similar [30, 40, 56] or lower [40, 57] frequency in pregnant compared to non-pregnant women. The variation in the reported rates in the four included studies can be explained by the differences in the outcome assessments and fever definitions (self-reported versus measured versus medically attended fever). Based on the largest cohort study’s estimate [50], we calculated that about 6 additional cases of fever per 100,000 pregnant women would occur after Tdap vaccination, as compared to no vaccination (using the difference of absolute risks in vaccinated versus non-vaccinated women).

Six studies reported a small increased relative risk of chorioamnionitis after Tdap vaccination. Three studies showed a significantly increased risk. All three were performed in the USA and had very large numbers of participants. They used “presence of respective ICD-9 codes in electronic patient data” as definition of chorioamnionitis [35, 38, 40].

Chorioamnionitis is an inflammation of the fetal membranes, the amniotic cavity including its fluid and of the placenta, predominantly due to ascending bacterial infections [58, 59]. It may occur at any time during pregnancy or delivery and may be preceded [59] or followed [58] by premature rupture of membranes. Chorioamnionitis is defined by either clinical features [58, 60], microbiological findings, histopathological signs [59] or a combination of these. Clinically relevant chorioamnionitis is a frequent cause of preterm birth and may lead to neonatal sepsis [59]. In the USA, a clinical and/or histologically proven chorioamnionitis is diagnosed in 40–70% of preterm deliveries and in 1–13% of term deliveries [59]. From an immunological perspective, it appears plausible that vaccination can trigger an inflammatory process in pregnancy. During the course of pregnancy, the immune system of the expectant mother undergoes changes in its actiity, ranging from local inflammation that accompanies the tumor-like implantation of the fetus (first trimester) to the predominance of immune tolerance (second trimester) and ending up with inflammatory signals that lead to the induction of labor (third trimester) [61]. Tdap vaccination is only one of multiple activating stimuli to the maternal immune system during pregnancy. To our knowledge, so far no studies on pregnant animals have been published that examined the consequences of immune stimulation by vaccines for the outcome of pregnancy.

If chorioamnionitis were causally related to Tdap-vaccination in the studies that were included in this review, for instance through some as yet unknown immunologic mechanism, we would expect increased risks of preterm birth or sepsis in infants of Tdap-vaccinated women. However, in seven studies [33, 35, 37, 38, 40,41,42], including those reporting a significant association between Tdap vaccination and chorioamnionitis, rates of preterm birth were even lower in Tdap-vaccinated than in unvaccinated women. Furthermore, in three studies the risk of NICU admission [33, 40] and sepsis [18] was also lower in infants of Tdap-vaccinated mothers as compared to infants of unvaccinated mothers, including one of the studies that reported an increased risk of chorioamnionitis [33, 40].

Hypothesizing that ICD-codes derived from electronic databases might not correctly reflect the clinical, microbiologic or histopathologic diagnosis of chorioamnionitis, Kharbanda et al. [38] validated the diagnosis by subgroup analysis. They randomly selected a validation sample of 220 women with hospital discharge ICD-codes for chorioamnionitis from electronic health charts. “Probable chorioamnionitis” was defined as the presence of ICD-9-Code 658.41 in combination with at least two clinical signs (maternal or fetal tachycardia, uterine tenderness, purulent or foul smelling amniotic fluid). Based on this definition the authors calculated that the positive predictive value (PPV) of the ICD-code for “probable chorioamnionitis” was 50%. When applying this PPV to the whole study population the association between Tdap vaccination in pregnancy and chorioamnionitis remained statistically significant. However, for the subgroup of women vaccinated between 27 and 36 weeks gestation, the association was no longer statistically significant (p = 0.07).

Based on our analyses healthy vaccinee bias is a likely confounding factor in most studies, irrespective of study design. This could be an explanation for lower risks for potential sequelae of chorioamnionitis and for the increased frequency of diagnosing and coding chorioamnionitis as a consequence of better ante- and perinatal care including a more careful surveillance of vaccinated women (detection bias).

One possible explanation for the association between Tdap vaccination in pregnancy and chorioamnionitis might be confounding with epidural anesthesia. In a secondary analysis of data from a randomized trial, Abramovici et al. [62] reported a statistical association between use of epidural anesthesia and chorioamnionitis defined as the presence of fever and a physician’s diagnosis warranting antibiotics. In this study, placental histopathologic examination revealed acute inflammation in 70% of those cases of clinical chorioamnionitis in which placental pathology was available for review (64%). Placental culture was not performed. In the cohort study by Maertens et al. [15], 70% of vaccinated versus 57% of unvaccinated women received epidural anesthesia, and in the subgroup analysis of Kharbanda et al. 95% of the 220 women with ICD-9 codes for chorioamnionitis had received an epidural anesthesia and 91% antibiotic treatment (all 220 women had received Tdap vaccination) [38]. However, this information was not available for the whole study population. Epidural anesthesia is often associated with prolonged labor and maternal fever [63, 64], often leading to prophylactic antibiotic use [64]. Transient or non-specific maternal fever might thus get coded as chorioamnionitis [62], leading to an overcoding of this diagnosis in women who had received epidural anesthesia.

As vaccinated women in the 3 studies with a significant association between Tdap-vaccination and chorioamnionitis obtained better prenatal care (earlier and more frequent ante-natal clinic visits [38] and more frequently ultrasound examinations [40]), they may also have requested epidural anesthesia more frequently than non-vaccinated women. Unfortunately, rates of epidural anesthesia were not analyzed in any of the studies investigating chorioamnionitis after Tdap vaccination.

In our systematic review, vaccine effectiveness data of 855,546 mother-infant-pairs from Australia, Spain, UK, and the USA were analyzed. In the USA, Tdap vaccination in pregnancy has been recommended since 2011 [65], in Australia and Valencia (Spain) since 2015 [46, 48]. In the UK, pertussis vaccination in pregnancy was introduced in 2012 as an emergency measure during a nationwide pertussis outbreak with 14 infant deaths [13]. In 2014, it was decided to continue with the program, since pertussis incidence remained high in the overall population and the available evidence showed good safety and effectiveness of the intervention [16, 66]. Compared to safety studies, VE studies in our review had a lower risk of bias and higher quality of evidence. In all studies considering laboratory confirmed pertussis as the outcome, VE was high: it ranged from 69 to 91% for prevention of pertussis, from 91 to 94% for prevention of hospitalization and was 95% for prevention of death in infants 0–3 months of age. The effect was diluted by additional inclusion of clinically suspected cases without laboratory pertussis confirmation in the study by Becker-Dreps [45], resulting in lower vaccine effectiveness estimates.

Our up-to-date systematic review has several strengths. We focus entirely on clinical outcomes (rather than immunological [serological] markers) and include a critical evaluation of a recently detected possible safety signal, i.e. chorioamnionitis. Using ROBINS-I, we applied the most advanced ROB tool to assess internal validity of the included observational studies, allowing a very detailed judgement. However, our review also has limitations that are mainly due to the limitations of the included studies. The majority of studies investigating safety outcomes had a considerable risk of bias, which impairs the ability of drawing firm conclusions on the risk of adverse events. Moreover, the three RCTs were designed and powered for the assessment of the immune response in pregnancy and, thus, were hampered by participant numbers that were too small for the assessment of rare safety outcomes. Regarding studies that investigated VE outcomes, those with the highest numbers of participants used the screening method to calculate VE. Since this method uses population estimates rather than individual data, controlling for confounders was not possible.

Conclusions

In this systematic review we summarize the currently available evidence on safety and effectiveness of pertussis vaccination in pregnancy. Vaccine effectiveness for prevention of infant pertussis, hospitalization and death is high. Two safety issues were observed in the included studies: fever and chorioamnionitis. Six additional cases of fever per 100,000 vaccinated women are to be expected, which is a small number and makes fever an adverse event of minor importance. Increased ICD-coding of chorioamnionitis, even though statistically associated with Tdap vaccination during pregnancy in some of the studies, does not seem to be clinically relevant. However, when implementing Tdap vaccination in pregnancy, surveillance of all safety endpoints, including chorioamnionitis and its sequelae, is needed in view of a likely residual healthy vaccinee bias in currently available studies and in view of the overall low quality of the evidence. Given the high vaccine effectiveness, pertussis vaccination during pregnancy has an overall positive benefit-risk ratio, particularly if the incidence of pertussis in infancy is high.