Safety of Administering Live Vaccines during Pregnancy: A Systematic Review and Meta-Analysis of Pregnancy Outcomes

Live-attenuated vaccines (LAV) are currently contraindicated during pregnancy, given uncertain safety records for the mother–infant pair. LAV might, however, play an important role to protect them against serious emerging diseases, such as Ebola and Lassa fever. For this systematic review we searched relevant databases to identify studies published up to November 2019. Controlled observational studies reporting pregnancy outcomes after maternal immunization with LAV were included. The ROBINS-I tool was used to assess risk of bias. Pooled odds ratios (OR) were obtained under a random-effects model. Of 2831 studies identified, fifteen fulfilled inclusion criteria. Smallpox, rubella, poliovirus, yellow fever and dengue vaccines were assessed in these studies. No association was found between vaccination and miscarriage (OR 0.98, 95% CI 0.87–1.10), stillbirth (OR 1.04, 95% CI 0.74–1.48), malformations (OR 1.09, 95% CI 0.98–1.21), prematurity (OR 0.99, 95% CI 0.90–1.08) or neonatal death (OR 1.06, 95% CI 0.68–1.65) overall. However, increased odds of malformations (OR 1.24; 95% CI 1.03–1.49) and miscarriage after first trimester immunization (OR 4.82; 95% CI 2.38–9.77) was found for smallpox vaccine. Thus, we did not find evidence of harm related to LAV other than smallpox with regards to pregnancy outcomes, but quality of evidence was very low. Overall risks appear to be small and have to be balanced against potential benefits for the mother-infant pair.


Introduction
Pregnancy induces immune modulation that renders women more susceptible to severe manifestations of infectious diseases, some of which can potentially be prevented by vaccination, e.g., influenza. In addition to protecting the mother, vaccination during pregnancy also serves as a vehicle to protect the unborn or newborn child via high-titer transplacental antibodies which can safeguard infants until active immunity through childhood vaccination has been established [1].
observational studies with no control group, and articles with unavailable full text were excluded from the review. No language or geographic restrictions were applied.
The search results from each database were exported to Mendeley 1.19.4 and deduplicated. Two reviewers (A.L. and D.B.) independently screened titles and abstracts, reviewed full texts of potentially relevant articles against inclusion criteria, and extracted data into a previously designed data collection form. Extracted data included study design, funding, country and setting, characteristics of the exposed and control groups, potential confounders, number of participants, vaccine(s) administered, trimester of pregnancy at vaccination, outcome definition and time points, number of events and sample size for all prespecified outcomes, crude and adjusted estimates of effect and variables included in adjusted analyses.
Included studies were independently evaluated by both reviewers using the ROBINS-I tool ("Risk Of Bias In Non-randomized Studies-of Interventions") [19]. The GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach was employed to assess overall quality of evidence for the prespecified outcomes. Any discrepancies were solved by discussion and consensus, consulting a third and fourth reviewer if necessary.

Data Analysis
A narrative synthesis was used to summarize the characteristics, results and internal validity of included studies, by vaccine. Meta-analyses were carried out using Review Manager software 5.3 (Cochrane Collaboration, Oxford, UK) for outcomes for which studies were deemed comparable. Pooled estimates of effect were obtained using the generic inverse-variance method under a random effects model, and presented as odds ratios (OR) and their 95% confidence intervals (CI). Fixed zero-cell correction was used for studies with no events in one or more groups. The degree of variability between studies was assessed using the I 2 statistic and heterogeneity was evaluated according to the Cochrane Handbook guidelines.
Given the variety of vaccine types that we encountered and their potentially different effect on pregnancy outcomes, we decided to do subgroup analyses by vaccine type, which had not been pre-specified in the protocol. Sensitivity analyses were conducted excluding studies at critical risk of bias. Given the greater susceptibility of the embryo to potential harmful effects of vaccines, we also decided to perform sensitivity analyses for the association of vaccination in the first trimester of pregnancy with miscarriage and congenital anomalies.
In addition to the studies included in the meta-analyses according to prespecified criteria, cohorts with no comparison group were summarized to obtain further safety data, especially regarding rare outcomes. The risk of congenital rubella syndrome (CRS) was calculated from studies of rubella vaccination, including controlled and uncontrolled cohorts. The upper limit for the risk of CRS was based on two-tailed 95% CI from the Poisson distribution.
There was no funding source for this study. The study protocol was registered with PROSPERO, number CRD42019138360.

Results
The search strategy identified 2831 articles following the removal of duplicates, of which 15 studies were included in the main body of this review: 12 cohorts (six retrospective, five prospective and one historically-controlled), two case-controls, and one secondary data analysis from clinical trials ( Figure 1). An overview of the key characteristics of included studies is provided Table 1. All retrievable articles were published in English and focused on populations from high and middle income countries. Sample size varies considerably, from a few dozens to over 9000 participants per group. Only one was industryfunded [20]. Included studies reported on outcomes after immunization with one of the following vaccines: smallpox (eight studies), rubella (three), OPV (two), YF (one), and dengue (one).  An overview of the key characteristics of included studies is provided Table 1. All retrievable articles were published in English and focused on populations from high and middle income countries (see Appendix A). Sample size varies considerably, from a few dozens to over 9000 participants per group. Only one was industry-funded [20]. Included studies reported on outcomes after immunization with one of the following vaccines: smallpox (eight studies), rubella (three), OPV (two), YF (one), and dengue (one).

Meta-Analysis of Pregnancy Outcomes
Meta-analysis was carried out for the outcomes of miscarriage, stillbirth, congenital anomalies, prematurity, and neonatal death (See Table 2). Data on congenital infection and LBW was considered inadequate for meta-analysis given the scarcity of studies reporting these outcomes, and the heterogeneity in the methods of outcome assessment for congenital infection. Miscarriage (pregnancy loss before 20 weeks EGA), stillbirth (fetal death after 20 weeks EGA) NR = Not reported, NA = Not applicable, DOB = Date of birth, LBW = Low birth weight, EGA = Estimated gestational age, SGA = Small for gestational age, SD = Standard deviation, IPV = Inactivated poliovirus, CDC = Centers for Disease Control and Prevention, LMP = Last menstrual period, BPA = British Pediatric Association, NBDPN = National Birth Defects Prevention Network. * Successful vaccination against smallpox implies development of a vesicle or pustule at the inoculation site. † N = 4238 in exposed group and 2214 in control group for assessment of LBW (longer follow-up to include all viable infants whose mothers were <3 months pregnant at vaccination). ‡ Exposure status (vaccination during pregnancy) was not determined at the individual level. § Mass campaign after dengue outbreak with use of organophosphate fogging.  (1) vaccines.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.

Miscarriage
Four studies on smallpox vaccine, two on rubella, and one each on OPV, YF, and dengue vaccine reported on miscarriage. Meta-analysis showed no association between immunization and spontaneous abortion overall and for each type of vaccine (pooled OR: 0.98; 95% CI 0.87-1.10). No statistical heterogeneity was detected across studies (I 2 = 0%) (Figure 2a). Four studies on smallpox vaccine, two on rubella, and one each on OPV, YF, and dengue vaccine reported on miscarriage. Meta-analysis showed no association between immunization and spontaneous abortion overall and for each type of vaccine (pooled OR: 0.98; 95% CI 0.87-1.10). No statistical heterogeneity was detected across studies (I2 = 0%) (Figure 2a). The analysis for exposure in the first trimester of pregnancy revealed a pooled OR of 2·66 (95% CI 0.73-9.64). However, the subgroup analysis showed a strong association between smallpox vaccination and miscarriage (OR 4.82, 95% CI 2.38-9.77, p < 0·0001) as shown in Figure 2b.
Quality of evidence was rated as very low, given a critical risk of selection bias (Table S2). When considering exposure to vaccination in the first trimester, downgrading was also granted for imprecision. Funnel plot revealed no evidence of publication bias ( Figure S1). The analysis for exposure in the first trimester of pregnancy revealed a pooled OR of 2·66 (95% CI 0.73-9.64). However, the subgroup analysis showed a strong association between smallpox vaccination and miscarriage (OR 4.82, 95% CI 2.38-9.77, p < 0·0001) as shown in Figure 2b.

Stillbirth
Quality of evidence was rated as very low, given a critical risk of selection bias (Table S2). When considering exposure to vaccination in the first trimester, downgrading was also granted for imprecision. Funnel plot revealed no evidence of publication bias ( Figure S1).

Stillbirth
No association was found between vaccination and stillbirth (pooled OR: 1.04; 95% CI 0. 74-1.48) in the pooled analysis of nine studies, including six on smallpox, and one each on rubella, OPV, and dengue vaccine ( Figure 3). There was moderate statistical heterogeneity between studies (I 2 = 45%, p = 0.07) and quality of evidence was rated as very low due to serious risk of bias (Table S3). Funnel plot revealed no evidence of publication bias ( Figure S2). No association was found between vaccination and stillbirth (pooled OR: 1.04; 95% CI 0.74-1.48) in the pooled analysis of nine studies, including six on smallpox, and one each on rubella, OPV, and dengue vaccine (Figure 3). There was moderate statistical heterogeneity between studies (I 2 = 45%, p = 0.07) and quality of evidence was rated as very low due to serious risk of bias (Table S3). Funnel plot revealed no evidence of publication bias ( Figure S2).

Congenital Anomalies
Eight studies on smallpox vaccine, two on rubella, and two on OPV contributed data for the meta-analysis of congenital anomalies, revealing no evidence of an association with maternal immunization (pooled OR: 1.09; 95% CI 0.98-1.21); this also held true when considering only pregnancies exposed to vaccination in the first trimester (pooled OR: 1.22; 95% CI 0.87-1.72), as depicted in Figure 4a,b. Nonetheless, the subgroup analysis revealed an increase in the odds of congenital anomalies after smallpox vaccination (OR: 1.24; 95% CI 1.03-1.49) and a tendency towards an association with rubella vaccine, albeit with a very wide confidence interval (OR: 2.8; 95% CI 0.65-12.04).

Congenital Anomalies
Eight studies on smallpox vaccine, two on rubella, and two on OPV contributed data for the meta-analysis of congenital anomalies, revealing no evidence of an association with maternal immunization (pooled OR: 1.09; 95% CI 0.98-1.21); this also held true when considering only pregnancies exposed to vaccination in the first trimester (pooled OR: 1.22; 95% CI 0.87-1.72), as depicted in Figure 4a,b. Nonetheless, the subgroup analysis revealed an increase in the odds of congenital anomalies after smallpox vaccination (OR: 1.24; 95% CI 1.03-1.49) and a tendency towards an association with rubella vaccine, albeit with a very wide confidence interval (OR: 2.8; 95% CI 0.65-12.04).
No statistical heterogeneity was detected across studies for the overall effect of immunization on congenital defects (I 2 = 0%), but there was substantial heterogeneity (I 2 = 77.4%) between different vaccines when considering only first trimester immunization (p = 0·01 for subgroup differences, Figure 4B). Quality of evidence was rated as very low, due to serious risk of bias (Table S4) and suspicion of publication bias ( Figure S3).

Preterm Birth
Five studies were pooled in the meta-analysis for preterm birth, including two on smallpox, two on rubella vaccine, and one on OPV. No association was found between vaccination and prematurity (pooled OR: 0.99; 95% CI 0.90-1.08) overall and for each type of vaccine ( Figure S4). No statistical heterogeneity was detected across studies (I 2 = 0%), but there was moderate heterogeneity between subgroups (I 2 = 42·7%). Quality of evidence was rated as very low as a result of serious risk of bias (Table S5).  No statistical heterogeneity was detected across studies for the overall effect of immunization on congenital defects (I 2 = 0%), but there was substantial heterogeneity (I 2 77.4%) between different vaccines when considering only first trimester immunization (p = 0·01 for subgroup differences, Figure 4B). Quality of evidence was rated as very low, due to serious risk of bias (Table S4) and suspicion of publication bias ( Figure S3).

Preterm Birth
Five studies were pooled in the meta-analysis for preterm birth, including two on smallpox, two on rubella vaccine, and one on OPV. No association was found between vaccination and prematurity

Neonatal Death
Three studies on smallpox vaccine, one on rubella, and one on OPV contributed to the meta-analysis of neonatal death, showing no association with vaccination (pooled OR: 1.06; 95% CI 0.68-1.65), as depicted in Figure S5. No heterogeneity between estimates of effect was observed (I 2 = 0%). Quality of evidence was rated as very low due to serious risk of bias (Table S6) and imprecision.

Sensitivity Analysis Excluding Studies at Critical Risk of Bias
Sensitivity analysis excluding four out of nine studies at critical risk of bias for the outcome of miscarriage revealed no association with vaccination (pooled OR: 0.97, 95% CI 0.86-1.09), consistent with the findings of the analysis including all studies. In the case of miscarriage after first trimester vaccination, all studies were at critical risk of bias and thus sensitivity analysis could not be conducted. For the rest of outcomes there were no studies at critical risk of bias.

Uncontrolled Cohorts
Additionally, twenty-three uncontrolled cohorts [8,[10][11][12][13]22,32, describing pregnancy outcomes after maternal vaccination were retrieved in the literature search. No cases of fetal vaccinia were found among 643 women vaccinated against smallpox [8,58,59], and rates of adverse pregnancy outcomes were within expected limits, except for a high frequency of stillbirth among those vaccinated in the first trimester [59] (Table S7).
Two controlled and twelve uncontrolled studies reported on the prevalence of CRS after maternal immunization [10- 13,22,32,[38][39][40][41][42][43]60,61], as shown in Table S8. No cases were detected among 3918 infants, including 2303 born to women susceptible to rubella before vaccination. The upper bound of the 95% CI for the risk of CRS is 0.09% for all infants, and 0.16% for those born to susceptible women. These cohorts included participants receiving the Cendehill, HPV-77 and RA 27/3 viral strains, which might differ in their fetotropic potential [10].
No signs of unexpected adverse pregnancy outcomes were found in four uncontrolled cohorts of YF vaccine [16,17,[62][63][64], except for an increase in minor dysmorphisms in a Brazilian cohort, which was probably explained by detection bias [62]. Only one case of probable intrauterine infection was detected among 422 infants tested for YF virus IgM (Table S9).

Discussion
According to the author´s knowledge, this is the first systematic review on the safety of the administration of LAV during pregnancy. Available evidence comes mainly from cohorts of women vaccinated against smallpox and, to a lesser extent, from observational studies regarding rubella, OPV, YF, and dengue vaccines. With the exception of smallpox vaccine, maternal immunization with these other vaccines shows no evidence of impact on pregnancy outcomes, but available evidence is of very poor quality.
Smallpox vaccine is the only LAV clearly implicated in adverse pregnancy outcomes, given the documented cases of fetal vaccinia, but, according to the available evidence, the risk of this severe complication appears to be low. In the present review, an increase in the odds of birth defects was detected and an association between vaccination in the first trimester and miscarriage was also apparent. However, the studies evaluating this outcome were at critical risk of bias because women were identified at prenatal clinics or hospitals, and early pregnancy losses were likely to be missed, which could have decreased the effect estimate. Furthermore, some studies included all vaccinated women, while others assessed only those considered as "successfully vaccinated", i.e., subjects who developed a cutaneous reaction. The proportion of women who underwent primary vaccination as opposed to revaccination during pregnancy varied widely between studies, and preexisting immunity could result in different responses to smallpox vaccine (see Table S7).
The largest studies included in this review were two population-based cohorts, comprising nearly 30,000 women, exploring the safety of OPV in the context of mass vaccination campaigns and showed no increased risk of adverse pregnancy outcomes [27,35]. In neither of these studies was exposure to vaccination determined at the individual level. However, mass campaign coverage was very high in both cases, increasing the confidence in their findings.
Despite the clear teratogenic action of wild-type rubella virus and the evidence of intrauterine infection in infants of women vaccinated during pregnancy, no cases of CRS have been documented in large cohorts of women inadvertently vaccinated during pregnancy or before conception in diverse settings and with different vaccine strains, which might have different fetotropic and teratogenic potential [10]. Maternal immunization with rubella vaccine does not appear to increase the risk of adverse pregnancy outcomes according to the available evidence from populations of women with varying degrees of previous immunity to rubella. However, the diversity of methods used to ascertain susceptibility and congenital infection, including hemagglutination-inhibition, IgM and IgG assays, IgG avidity tests, and viral isolation, could result in variable diagnostic accuracy.
Even though a total absence of risk cannot be proven, the likelihood of CRS is less than one per 1000 exposed pregnancies, according to available evidence [10- 13,22,32,[38][39][40][41][42]60,61]. Therefore, women inadvertently vaccinated during pregnancy should be reassured and therapeutic abortion would not be justified on the basis of a teratogenic potential of the vaccine.
The evidence regarding other LAV is even more scarce. In the case of YF vaccine, the only controlled study suggested an increased odds of miscarriage among vaccinated women, but the evidence for the association was weak. The mass vaccination campaign that provided the setting for this study was conducted after a dengue outbreak where organophosphate insecticides had also been used for vector control. Although reported exposure to organophosphates was similar between groups, measurement was imprecise, and residual confounding cannot be ruled out.
Other pregnancy outcomes after YF vaccination have been described only in uncontrolled studies. The largest of these cohorts followed 441 women inadvertently vaccinated in a mass campaign in Brazil [16,62]; the frequency of stillbirths (0.7%) and preterm birth (7.8%) were similar or lower than regional rates as reported by the authors. Major birth defects were found in 3.3% of 304 newborns. Minor dysmorphisms were detected in 62 infants, a frequency greater than expected, but this was attributed to evaluation bias. IgM antibodies were undetectable in blood samples from 341 infants, but antibody persistence was detected in one of 37 children assessed after two years of age, raising the possibility of intrauterine infection.
The only study analyzing the safety of maternal immunization with dengue vaccine suggested a possible increase in the frequency of stillbirths [20], but there were only two events in the exposed group, both in adolescent mothers. This association, however, should be furthered explored.
No controlled studies fulfilling the inclusion criteria were found for the live-attenuated influenza vaccine (LAIV). Similarly, no studies on varicella-zoster virus (VZV) vaccine fulfilled inclusion criteria, but data on 893 women accidentally vaccinated during or shortly before conception is available from the Merck/CDC Pregnancy Registry [65,66]. The prevalence of congenital anomalies (2.1%) was similar to that reported in the U.S. with no specific pattern of birth defects [67]. No cases of congenital varicella syndrome were detected among 810 infants.
The live-attenuated viral vectored vaccine against Ebola (rVSV-ZEBOV) is currently being administered to pregnant women as part of the ring vaccination strategy in Democratic Republic of Congo [68]. This will provide an opportunity to evaluate pregnancy outcomes to better inform decision-making in future outbreak scenarios.
Although the results presented in this manuscript are mostly reassuring, there are several limitations which preclude drawing final conclusions regarding the safety of maternal immunization with LAV. Besides the observational nature of the available evidence, most studies did not follow participants from the time of vaccination, leading to incomplete ascertainment of early pregnancy loses, which might include cases of congenital infection and malformations. Variable diagnostic criteria, ancillary methods and follow-up for ascertainment of birth defects can result in varying degrees of underreporting. The heterogeneity in study designs, settings, definitions and data collection methods leads to challenges in the comparison of different studies. On the other hand, the wide variety of countries and settings increases the generalizability to different populations. Furthermore, each LAV might have a different effect on pregnancy outcomes, which might be obscured by obtaining pooled effects. However, this was addressed by subgroup analyses which revealed an increased odds of miscarriage and congenital anomalies only after smallpox vaccination.
All but one of the included studies were considered at serious risk of bias due to confounding for all outcomes, since most of them did not measure potential confounding factors or did not control for them. Only the studies by Nishioka and Ryan provided an effect estimate adjusted for confounding [33,36]. However, confounding domains were not measured validly in the first case, and some important confounders were not controlled for in the second. Bar-Oz et al. controlled for confounding by matching, but no information was provided on the methods used to measure these potential confounders [22].
Four out of nine studies were considered at critical risk of selection bias for the outcome of spontaneous abortion [24,25,30,31], because participants were selected among women attending participating hospitals or clinics, thus missing those receiving care in different facilities and miscarriages occurring before onset of prenatal care.
The findings from this literature review cannot be directly extended to other LAV given the variable teratogenic potential of different vaccines. The use of modern technologies, such as viral vectored vaccines and novel adjuvants, might also modify the safety profile of live vaccines available in the near future. Furthermore, the power to detect a modest increase in the frequency of specific types of malformations is low, so the association of maternal vaccination with birth defects cannot be ruled out.
The development of new LAV that may target women of childbearing age, such as the viral vectored candidates for the prevention of Zika [69,70], and the use of the existing ones in mass vaccination campaigns or in response to outbreaks will need to be coupled with enhanced surveillance of pregnant women and their infants, as well as standardized case definitions to better characterize the impact of vaccination on the mother-offspring dyad [71]. Additionally, more accurate baseline information on maternal and fetal outcomes in low and middle income countries (LMIC) is desirable to make comparisons with appropriate local rates [71].
Furthermore, other factors need to be taken into account when considering maternal immunization, including implementation challenges and the need for integration with existing prenatal care services, as well as the potential interference of maternal antibodies with the development of infant humoral immune responses to vaccines in the first few months of life. This interference has been observed for several vaccines but its clinical significance is still uncertain and might differ depending on the specific antigen [72,73].
In summary, the risk of adverse pregnancy outcomes after maternal immunization with LAV appears to be small. This information can support decision-makers in the planning of vaccination campaigns that target women in the reproductive age, and may provide reassurance to healthcare workers taking care of women inadvertently immunized with LAV during pregnancy.
Our findings may also inform pressing decisions regarding vaccination of pregnant women during outbreaks. In this setting, benefits might exceed potential harms when the disease manifestations are severe for pregnant women and/or their offspring and the probability of exposure is high, as is the case with Ebola and Lassa Fever [74,75].
Inclusion of pregnant women in vaccine trials under conditions of enhanced safety monitoring and appropriate follow-up of mothers and infants has the potential to expand the benefits of immunization to these populations. The views of pregnant women themselves need to be taken into account in research design and policy making. Qualitative research on vaccine confidence [18,76] and willingness to participate in clinical trials, conducted in different settings and geographic locations, can help improve acceptance. We believe that pregnant women ought to be protected through research, not from research, in order to move forward in the promotion of health equity.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-393X/8/1/124/s1, Table S1: Detailed search strategy using MEDLINE and EMBASE databases, Table S2: Risk of bias assessment for miscarriage, according to the ROBINS-I tool, Table S3: Risk of bias assessment for stillbirth, according to the ROBINS-I tool, Table S4: Risk of bias assessment for congenital anomalies, according to ROBINS-I tool, Table S5: Risk of bias assessment for preterm birth, according to the ROBINS-I tool, Table S6: Risk of bias assessment for neonatal death, according to ROBINS-I tool, Table S7: Uncontrolled cohorts and pregnancy registries evaluating the pregnancy outcomes after maternal immunization with smallpox vaccine, Table S8: Uncontrolled cohorts and pregnancy registries evaluating the pregnancy outcomes after maternal immunization with rubella vaccine, Table S9: Uncontrolled cohorts and pregnancy registries evaluating pregnancy outcomes after maternal immunization with yellow fever vaccine, Figure S1. Funnel plot of the studies included in the meta-analysis of the effect of the association of maternal immunization with miscarriage, Figure S2. Funnel plot of the studies included in the meta-analysis of the effect of the association between maternal immunization and stillbirth, Figure S3. Funnel plot of the studies included in the meta-analysis of the effect of the association between maternal immunization and congenital anomalies, Figure S4: Meta-analysis of the effect of maternal immunization on preterm birth, Figure S5. Meta-analysis of the effect of maternal immunization on neonatal death.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A. Qualitative Synthesis of Included Studies
Findings of the included studies reporting on pregnancy outcomes are summarized below, grouped by vaccine.

Appendix A.1. Smallpox Vaccine
Three prospective and four retrospective cohorts, as well as one case control study fulfilled the inclusion criteria. All were conducted between the 1940s and 1960s, except for one more recent cohort published in 2008. Bellows et al. (1949) conducted a prospective cohort of 720 women who presented to antepartum clinics before the fourth month of pregnancy and who were vaccinated against smallpox during a mass campaign in New York, and 173 unvaccinated controls [23]. Among those less than 5 months pregnant at the moment of the vaccination campaign, 15 out of 428 (3.5%) exposed women and 4 out of 103 (3.9%) control women experienced miscarriage (OR: 0.90; 95% CI 0.29-2.77). Stillbirths occurred in 23 of 705 (3.3%) pregnancies that reached the fifth month in the exposed group, compared to 4 out of 169 (2.4%) in the control group (OR 1.39; 95% CI 0.47-4.08). Birth defects were documented in 3.3% and 2.4% of births (OR: 1.39, 95% CI 0.47-4.08), and neonatal deaths in 1.6% and 0.6% in the vaccinated and control groups respectively (OR: 2.69; 95% CI 0. . In another study carried out in New York, Greenberg et al. (1949) retrospectively studied 4172 infants born to mothers vaccinated in the first trimester of pregnancy and 2186 controls identified in the same hospitals and time period [26], and found no evidence of increased risk of major malformations among exposed newborns (0.7% versus 0.6%, OR: 1.25; 95% CI 0. 65-2.40).
Similarly, no evidence of increased risk of adverse pregnancy outcomes was found by Liebeschuetz (1964) in another retrospective cohort that included 157 vaccinated and 1657 unvaccinated or unsuccessfully vaccinated (no reaction at injection site) women attending a maternity hospital during a 6-month period after a mass vaccination campaign in London [30]. The majority of exposed women received the vaccine in the first trimester of pregnancy. The proportion of miscarriage (5.1% versus 4.5%, OR: 1.15; 95% CI 0.54-2.43) and stillbirth (2% versus 1.3%, OR: 1.53; 95% CI 0. 45-5.19) was similar between exposed and unexposed women. Birth defects were diagnosed in 3 out of 149 births in the vaccine group and 22 of 1583 births in the control group (OR 1.46; 95% CI 0. 43-4.93). No cases of fetal vaccinia were diagnosed among all live or stillborn infants.
Bourke and Whitty (1964) studied 112 women exposed to smallpox vaccine during pregnancy and 448 unexposed mothers attending antenatal clinics in four hospitals in Dublin [24]. The proportion of spontaneous abortion was the same in both groups (OR 1.0; 95% CI 0.11-9.04). Stillbirth was more common among exposed women (3.6% versus 1.3%) but the confidence interval of the odds ratio was compatible with no evidence for an association (OR 2.7; 95% CI 0.77-9.29). Similarly, congenital anomalies were detected more frequently in the vaccine group (3 of 113 births) in comparison with the control group (3 of 449 births) but the estimate of effect had a wide confidence interval including unity (OR 4.05; 95% CI 0.81-20.36). Two of the three mothers of malformed infants in the exposed group received the vaccine late in pregnancy (23 and 27 weeks) making a causal link unlikely. The frequency of neonatal deaths was 0.9% in both groups.
The pregnancy outcomes of 1522 women successfully vaccinated against smallpox at any stage of pregnancy during a mass campaign in Iran were prospectively described by Naderi (1975) and compared to 2014 unvaccinated women attending the same clinics during the following year [31]. All exposed women had received the smallpox vaccine in the past. Eighteen pregnancies in the exposed group and twenty-one in the control group ended in miscarriage (OR: 1.14; 95% CI 0.60-2.14). There were 23 stillbirths among vaccinated women and forty-eight among controls (OR: 0.63; 95% 0.38-1.04). The proportion of infants with malformations was similar between groups (1.4% versus 1.2%, OR: 1.17, 95% CI 0.66-2.08) but it was higher in the vaccine group when only those exposed in the first trimester were taken into account (3.5% versus 1.2%, OR: 2.90; 95% CI 1.24-6.79). With regards to specific defects, an increased risk for club foot was detected among infants exposed in the first trimester in comparison to the unexposed group (1.5% versus 0.25%, OR: 6.06; 95% CI 1. 44-25.57). No cases of fetal vaccinia were detected in this cohort.
In a more recent study,  analyzed a cohort of infants born to active-duty military women using data from the U.S. Department of Defense Databases [36]. Birth defects were diagnosed in 40 of 882 infants exposed to smallpox vaccine during gestation and in 867 of 23,685 infants born to unvaccinated mothers (OR: 1.25; 95% CI 0.77-1.40). Estimates of effect were similar when considering only those exposed in the first trimester of pregnancy (OR: 1.24; 95% CI 0.85-1.81). These authors did not find an increased risk of club foot in infants of vaccinated mothers. The prevalence of preterm birth was similar between groups (7.5% versus 7.0%, OR: 1.07; 95% CI 0.83-1.38).
In the case control study by Saxen et al. (1968), 835 stillbirths and 642 newborns with congenital anomalies notified to the National Board of Health in Finland were compared to 1477 infants born next after the cases in the same district [37]. Birth defects were not significantly associated to vaccination during pregnancy or shortly before conception (OR: 1.18, 95% CI 0.88-1.57) or to vaccination during the first trimester (OR: 1.05; 95% CI 0.64-1.74). The odds of exposure to smallpox vaccine were lower among stillbirths in comparison to controls (OR: 0.71; 95% CI 0.52-0.96).

Appendix A.2. Rubella Vaccine
Two prospective and one retrospective cohorts evaluated pregnancy outcomes after exposure to rubella vaccine, with or without measles component. The rubella strain received by participants was RA 27/3 in two of the studies, and is not reported in the other one.
A retrospective cohort conducted by Ebbin et al. (1973) in Los Angeles County included 60 women inadvertently vaccinated during pregnancy or less than 3 months before conception and 47 controls matched for age, race, parity, sex of the infant and private/non-private hospital status [25]. No association was found between vaccination and miscarriage, with three cases documented in each group (OR: 0.77; 95% CI 0.15-4.01). Fourteen pregnancies in the exposed group were interrupted and rubella virus was isolated in two cases from products of conception. One of these isolates was characterized as wild-type and the other as attenuated rubella virus, although molecular methods were not yet available for confirmation. No virus was isolated in cultures of pharyngeal and rectal swabs taken from 34 exposed infants.
Bar-Oz et al. (2004) prospectively followed 94 women who called a telephonic counselling service for pregnant and lactating women in Toronto after being vaccinated against rubella within three months before or after estimated conception, and 95 controls counselled at similar gestational ages for non-teratogenic exposures [22]. No increased prevalence of miscarriage was found among vaccinated mothers as compared to controls (6.4% versus 8.4%, OR: 0.74; 95% CI 0.25-2.3) but seven pregnancies were interrupted due to fear of teratogenicity of rubella vaccine. No stillbirths were documented and congenital anomalies were diagnosed in three newborns in each group, showing no association with vaccination (3.7% versus 3.4%, OR: 1.08; 95% CI 0. 21-5.5). There were no cases of congenital rubella syndrome among 81 exposed newborns.
A prospective cohort conducted by Namaei et al. (2008) included 106 women vaccinated within three months before or after conception during a mass campaign in Iran, and 40 unexposed controls [32]. Women with a history of previous rubella infection or vaccination were excluded from the study. One pregnancy in the exposed group (0.9%) and none in the control group ended in stillbirth. The prevalence of prematurity was 7.5% among exposed newborns and 7.1% among the unexposed. Birth defects were detected in 2 of 108 infants in the vaccine group and in none of the controls. No evidence of congenital infection was found on the basis of serological essays performed in all livebirths, and none of the children exhibited signs of CRS.