FormalPara Key Points

The current scope of work related to drug/vaccine safety among pregnant individuals living in LMICs largely focuses on infectious diseases, and are frequently from Africa, and are often funded by high-income countries.

These publications largely failed to include maternal outcomes as an outcome of interest overall and longer-term maternal, birth, or neonatal/infant outcomes.

Current gaps in pregnancy PV in LMICs exist in terms of specificity of exposure timing (timing of initiation and duration of treatment) and the need to broaden focus to include non-communicable diseases.

1 Introduction

Pharmacovigilance (PV), or the ongoing safety monitoring that occurs after a medication or intervention has been licensed [1], plays a crucial role in pregnancy, as clinical trials often exclude pregnant people. Although PV encompasses a broad umbrella of activities, from animal testing to population-level surveillance, its main approaches include passive and active surveillance, targeted post-marketing studies, and record linkage studies utilizing patient data sources [1]. In this review, we utilize the term “PV projects” to maximize inclusion of PV-related work, such as cohort studies, programs, etc., that assess pregnancy outcomes in low- and middle-income countries (LMICs) to allow us to capture a larger breadth of PV in pregnancy projects, address a clear gap in knowledge regarding various exposure and adverse outcome associations that are not explicitly investigated in clinical trials. Currently, studies focused on therapeutic and vaccine efficacy continue to exclude pregnant people, resulting in them having delayed or no access to potentially beneficial interventions [2,3,4]. Furthermore, some adverse maternal, birth, and neonatal/infant outcomes are rare, requiring a large sample size, and cannot feasibly be studied in pre-approval trials. Thus, real-world evidence is needed to ensure pregnant people and their newborns are not exposed to harmful therapeutic agents, or denied access to treatments with a favorable benefit-risk profile.

Pregnancy PV projects in LMICs are of particular interest because some adverse maternal, birth, and neonatal/infant outcomes are more prevalent in LMICs compared to high-income countries (HICs) [5, 6]. Additionally, certain exposures, such as therapeutics for HIV, tuberculosis (TB), and malaria, are more prevalent in LMICs than in HICs. Investments in post-marketing surveillance systems may not be routine in many LMICs and population-level electronic health records, which can be quickly leveraged for pregnancy PV activities in HICs, are relatively uncommon. Additional data collection requirements may challenge overburdened health systems in LMICs.

The disproportionate lack of data from LMICs poses a threat to global health equity, with the majority of the world’s pregnant population living in LMICs. Thus, it is important to understand how pregnancy PV projects in LMICs have conducted their work to assess exposure-outcome relationships and learn enduring lessons for future related work.

In 2014, the World Health Organization (WHO) created the Global Alignment of Immunization Safety Assessment in pregnancy (GAIA) initiative in order to create a global standard for monitoring vaccine safety in pregnancy, especially focusing on LMICs [7]. This initiative has greatly supported standardization of case definitions and outcomes for vaccine safety research, and especially surveillance; however, it has not translated to more PV in pregnancy projects nor harmonization of related outcomes in LMICs. The challenges lie in ensuring that such standardized approaches align with realities of clinical maternal and neonatal care in LMICs beyond a focus on vaccine safety surveillance. Developing robust PV in pregnancy projects continues to be a challenge in LMICs due to many factors, including scarcity of medical personnel, under-resourced and often fragmented health infrastructure, and limited investment in pregnancy PV projects [3]. Yet, these projects are vital for ensuring optimal treatment of pregnant people and infants and it is important to understand how PV projects currently function and what can be learned for effective implementation in LMIC settings. To fill these knowledge gaps, we conducted a scoping review to assess how exposures and outcomes were measured/operationalized in recently published PV in pregnancy projects throughout LMICs.

2 Materials and Methods

2.1 Aim and Objectives

The aim of this scoping review was to examine measures and metrics for exposure-outcome relationships in relation to PV projects in pregnancy in LMICs. Specifically, we sought to evaluate how different published pregnancy PV projects have measured and operationalized their exposures and outcomes of interest. Exposures of interest were therapeutics or vaccinations utilized before or during pregnancy and outcomes of interest included maternal, birth, and neonatal/infant outcomes. This scoping review was conducted between October 2022 and September 2023. The conduct of our scoping review was informed by the PRISMA reporting guidelines for scoping reviews [8] (see supplementary materials for checklist).

The specific research questions that guided this scoping review were:

  1. 1.

    What published PV in pregnancy projects exist in LMICs?

  2. 2.

    How were exposure and outcomes measured/operationalized in pregnancy PV projects based in LMICs?

  3. 3.

    How may gaps in existing pregnancy PV projects guide future projects?

2.2 Literature Search

We conducted the first search of all relevant, indexed publications on December 28, 2022. We began our search by first meeting with a librarian, specializing in scoping reviews, at the University of Washington Health Sciences Library. This librarian aided us in creating a complete and relevant search string to be used in PubMed and Embase. After noticing a lack of LMIC results within these databases, we improved our search string by adding a comprehensive LMICs search term, based on World Bank categorization [9], to our existing search string, in addition to editing our search strings to be used in SciELO and Global Index Medicus. The updated search string was run on February 6, 2023. Search terms focused on the subjects, pregnant, birth, maternal, newborn, and neonatal outcomes, as well as pregnancy PV projects and an extensive list of LMICs (see supplementary materials for full search string). After primary review of our search results, from July 2023 to September 2023, we conducted a secondary review of the references from the primary publications. Additionally, using team knowledge, we compiled a list of key publications that were not identified in our primary or secondary search.

2.3 Eligibility Criteria

While inputting the search string into the databases, filters were placed to select publications in English pertaining to “Humans” and “Female” (see supplementary materials for full search string). Publications were selected from a 10-year range (from 2012 to 2022). Publications pertaining to animals were excluded. The LMICs strings were based on the Cochrane Effective Practice and Organization of Care (EPOC) LMICs filters 2020 v.4 with additional terms added to increase comprehensiveness [10].

These parameters were chosen for numerous reasons including comprehensiveness, practicality, and efficiency. Publications were selected from a 10-year range for the scoping review to be thorough, yet feasible for the research team. Animal studies were excluded to focus on the target population of the scoping review, pregnant people and their newborns. Last, we limited the search results in PubMed and Embase to publications in English, the common language spoken by the research team. Notably, SciELO publishes literature originating in Latin American, Caribbean, and Iberian countries and Global Index Medicus is a database created by the WHO that provides access to literature produced by LMICs. Thus, for these databases, search results were not limited to English. The research team wanted to make sure we were not excluding any important and relevant publications and so for these databases, if the English text was available for the publication, it was reviewed.

2.4 Publication Selection

After the finalization of the search string, two members of the research team (JS, MKV) conducted an initial selection of relevant publications based on title review, recording notes in a Google Sheet. Deduplication of publications was done via a validated deduplication tool through the Systematic Review Accelerator developed by Bond University [11]. The abstracts of these publications were then screened for relevance, and those meeting our inclusion criteria were compiled in a Google Sheet. Inclusion criteria at this stage included: (1) exposure delineated as a therapeutic and/or vaccine, (2) reported on maternal, birth and/or neonatal/infant outcomes, (3) assessment of relationship between exposure and outcome existed, (4) pregnant people were a main study population or publication contained significant data on pregnant people and neonates, and (5) publication focused on LMICs or global data. Although “pharmacovigilance” was a term included in our search string, authors did not need to explicitly name their project a PV project in order to be eligible for this review. Once a list of all potentially relevant publications based on abstract-only screening was compiled, full copies of the publications were obtained and entered in Zotero. We were successful in obtaining full copies of all publications at this stage. The list of publications was split between the two members (JS, MKV) conducting the publication review, and one member read the full publication to determine relevance. Additionally, case series and case reports were excluded, because we were interested in pregnancy PV projects as a whole, rather than particular cases of adverse events (AEs) or outcomes. When we included relevant systematic reviews or meta-analyses, we chose not to review each source included in the review as the goal of this scoping review was synthesis of existing publications, and not enumeration of specific findings as would be done in a systematic review or meta-analysis. A subset of publications was reviewed by both (JS, MKV) to ensure reliability in the review process. Any discrepancies were resolved via discussion and consensus with inclusion of a third reviewer (RCP). Publications felt to have unclear relevance based on the initial review were discussed during regular meetings with the core research team (JS, MKV, and RCP), and decisions to include or exclude were made by consensus. The same process was used for the secondary review of references from the primary publications. Two members of the research team (JS, MKV) conducted an initial selection of the reference publications based on publication title. The abstracts of these publications were screened for relevance and compiled in a separate page on the same Google Sheet. The inclusion and exclusion criteria was the same as before. After full copies of the publications were obtained, the publications were split for review between the two members.

3 Data Items and Synthesis of Results

Our Google Sheet was populated with each publication’s URL, title, project name, authors, publication year, main objective, overall study design, target population, inclusion and exclusion criteria, comparator groups, primary exposure(s), primary exposure measure(s), primary outcome(s), primary outcome measure(s), findings, and funding sources. The exposure variables reviewed were largely therapeutics for malaria, TB, and HIV, as well as vaccinations. The outcomes reviewed were maternal (e.g., maternal death, preeclampsia/eclampsia, and adverse therapeutic-/immunization-related events), birth (e.g., major congenital anomalies, stillbirth, and small for gestational age [SGA]), and neonatal/infant outcomes (e.g., neonatal mortality and infant mortality). Exposures were grouped by infectious diseases (with therapeutic use as exposure) and vaccinations based on topic frequency emerging among the publications. In order to standardize outcome categories, outcome variables were grouped by maternal, birth, and neonatal/infant outcomes based on recent WHO recommendations [12].

4 Results

4.1 Selection of Sources of Evidence

In our primary review, our search string produced 53 publications across all databases. After deduplication, there were 50 publications remaining (Fig. 1, [13]). We identified 25 publications meeting our inclusion criteria based on title and abstract alone. We included these 25 publications for full text review. Of these, 14 were excluded for the following reasons: irrelevant (i.e., did not report maternal, birth, or neonatal/infant outcomes, n = 7), publication focus was non-LMICs (n = 5), or publication was a case series or case report (n = 2). The database sources for the 11 included publications were: PubMed (n = 8), Embase (n = 2), SciELO (n = 1), and none were identified from Global Index Medicus.

Fig. 1
figure 1

Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) publication selection procedure flow chart. LMIC low- and middle-income country

Full size image

From the secondary review of the references of the primary publications, we identified 632 references, of which 87 publications met our inclusion criteria based on title alone (Fig. 1 [13]). After abstract review, we excluded 66 publications and included 21 publications for full text review. These publications were excluded for being duplicates (n = 6), non-LMICs (n = 19), irrelevant (i.e., did not report maternal, birth, or neonatal/infant outcomes, n = 40), or for not being within the publication range (2012–2022, n = 1). After a full text review, we excluded 5 additional publications for the following reasons: retracted study ( n = 1), irrelevant (i.e., did not report maternal, birth, or neonatal/infant outcomes, n = 3), unclear exposure-outcome ascertainment (i.e., publication was a hypothetical trial, n = 1), leaving 16 secondary review publications.

After a full text review of all publications included in the primary and secondary review, we included 31 publications in our synthesis, with n = 11 included from the primary search string, n = 16 from secondary review, and n = 4 from team knowledge (Table 1).

Table 1 Publications included in scoping review (N = 31)
Full size table

4.2 Characteristics of Sources of Evidence

After compiling the data extraction for the 31 included publications, we further categorized the publications into two groups: publications relating to therapeutics for infectious diseases (n = 27, Table 2) and vaccines (n = 4, Table 3). The publications ranged across more than 28 countries, with several publications covering more than one country (n = 11); the majority of studies were from Africa (n = 28). Most publications ( n = 30) had pregnant people as the only study population. All of the publications covered infectious diseases/conditions, with most publications (n = 27) focusing on a therapeutic exposure (n = 17 for HIV, n = 7 for malaria, n = 9 for TB, n = 2 for multi-class therapeutics exposure) compared to publications about vaccinations (n = 1 for pertussis, tetanus, and diphtheria, n = 3 for influenza). As for outcomes, n = 15, n = 31, and n = 20 of the publications covered maternal, birth, and neonatal/infant outcomes, respectively.

Table 2 Summary of infectious disease-related publications
Full size table
Table 3 Summary of vaccine-related publications
Full size table

4.3 Synthesis of Results

The results are presented in two sections pertaining to infectious diseases-related (1) therapeutics and (2) vaccines. We discuss in detail common themes arising from the publications, including ascertainment of exposures and outcomes, timing of exposures during pregnancy, and comparator groups used in the exposure-outcome relationship assessment.

4.4 Infectious Diseases Therapeutics-Related Publications (Table 2)

For the 27 infectious diseases therapeutics-related publications, the major diseases analyzed in relation to pregnant people and their newborns were HIV (n = 17), malaria (n = 7), drug-resistant tuberculosis (DR-TB)/multidrug-resistant tuberculosis (MDR-TB) (n = 9), along with multi-class therapeutics exposure (n = 2; Table 2).

4.5 HIV

Among publications with HIV as the infectious disease of study, the primary exposure-outcome relationship of focus was exposure to maternal antiretroviral therapy (ART) (primarily dolutegravir-, efavirenz-, nevirapine-containing ART or zidovudine monotherapy) and adverse maternal, birth, and neonatal/infant outcomes.

For the vast majority of HIV-centered publications (n = 17), a focal point of analysis was the time of ART initiation and the duration of treatment in relation to gestation, and birth, maternal, and neonatal/infant outcomes [14,15,16,17,18,19,20,21,22,23,24,25]. For example, some publications categorized the time of initiation into subcategories based on stages of pregnancy such as early (< 8 weeks), mid (9–20 weeks), and late pregnancy (21–36 weeks) [22] or first trimester (< 14 weeks), first half of second trimester (14–20 weeks), second half of second trimester (21–27 weeks), and third trimester (>28 weeks) [20]. Notably, Mehta et al specifically evaluated first trimester ART exposure and found that there was no association between first trimester exposure to efavirenz-containing ART regimens and congenital malformations yet found that first trimester exposure to nevirapine was associated with a greater risk of congenital malformations compared to births not exposed to ART during the first trimester [15]. The evaluation of first trimester exposure allowed Mehta et al to specifically comment on birth and infant outcomes in relation to specific exposure timing, as this cohort excluded pregnant people for whom the timing of ART was uncertain [15]. Similarly, by evaluating ART exposure based on varying lengths of treatment, Bengston et al was able to report that there was no evidence of elevated risk of low birth weight (LBW) infants for individuals receiving combination ART for any treatment length compared to individuals who never initiated ART [22].

However, most publications defined the categories of ART initiation as pre- (before) or post-conception (during pregnancy) (n = 8) [17,18,19,20,21, 23,24,25]. For publications defining ART initiation as pre- or post-conception, ART initiation preconception was defined as maternal ART that started before the calculated date of the last menstrual period (LMP), and post-conception ART initiation was defined as maternal ART that started after that date. To account for errors in estimation, one publication defined ART exposure prior to conception as 2 weeks before LMP, with broader exposure categories being based on the trimester in which ART treatment was initiated (e.g., 15 weeks post-LMP) [15].

Notably, many publications were able to comment on the significance of the timing of treatment initiation in relation to maternal, birth, and neonatal/infant outcomes due to exposure initiation stratification. To illustrate, Zash et al (2019) specifically sought to evaluate the effects of dolutegravir exposure at conception on the prevalence of neural tube defects and found that neural tube defects were more prevalent in association with dolutegravir-based treatment at conception than with non-dolutegravir ART at conception [18]. Ramokolo et al reported higher preterm delivery rates among pregnant people who initiated ART preconception compared to those who initiated ART post-conception [21]. Last, Chen et al evaluated ART exposure based on timing of ART initiation (before or after 32 weeks’ gestation), and thus were able to report that there were no significant differences in preterm delivery, SGA infants, or stillbirth rates based on timing of ART initiation [24].

To estimate gestational age in relation to the time of ART initiation, various factors were utilized across the publications. All publications utilized the estimated date of LMP and other factors depending on the data available in the publication. Some publications utilized LMP along with fundal height (n = 4) [20, 22,23,24], some used LMP and a dating ultrasound (if available) (n = 4) [15, 20, 24, 25], and one publication further assessed the accuracy of estimation by comparing the mean birthweight for each week of gestation age to a reference growth curve adjusted for the population [22]. For the majority of the reviewed HIV-related publications, whether they utilized data from a larger PV database or medical records, there was some degree of uncertainty in the exact timing of ART initiation relative to conception and/or trimester of exposure.

For HIV-centered publications that utilized comparator groups or had explicit information detailing their comparator groups, we found that all publications used contemporaneous/concurrent comparator groups (n = 10). Among these, the comparator groups included: pregnant people not living with HIV [15, 17,18,19,20], pregnant people living with HIV with various ART exposures [17,18,19, 21, 23, 24], pregnant people living with HIV with ART exposures at various gestational durations [22], and pregnant people living with HIV with no antenatal ART use [21]. Some publications collected data prior to WHO guidelines for universal ART use for pregnant people, thus overall ART exposure before and after this time period shifted [17,18,19, 23, 24].

4.6 Malaria

For publications with malaria as the infectious disease of study, the primary exposure-outcome relationship of interest was the association between antimalarial medication exposure during pregnancy and adverse maternal, birth, and neonatal/infant outcomes. These publications analyzed various antimalarial medications: artemisinin derivatives, such as artemisinin-based combination therapies (n = 6, [26,27,28,29,30,31]), quinoline derivatives, such as chloroquine, quinine, and amodiaquine (n = 3, [27, 28, 32]), and antifolates, such as sulfadoxine-pyrimethamine (n = 2, [28, 32]). The birth outcomes of interest were live birth, stillbirth, and miscarriages; birth maturity (preterm birth or full-term birth); birth weight; and congenital malformations. Among these publications there was an emphasis on examining the timing of exposure in relation to the outcome. Particularly, these publications sought to understand the effect of antimalarial medication exposure during the first trimester or preconception and adverse maternal, birth, and neonatal/infant outcomes [27, 29, 32], although two publications sought to understand the effect of antimalarial medication exposure during the second and/or third trimester compared to first trimester and preconception exposure (n = 2) [30, 31].

Many publications noted limitations in determining the timing of therapeutic exposure (due to incomplete records and related factors), particularly periconceptional exposure, which was dated retrospectively using estimated gestational age. Gestational age was most often estimated using LMP accompanied by other factors, depending on the publication [27, 29,30,31,32]. There were several combinations of estimation methods: 1) LMP, ultrasound, Dubowitz newborn assessment, fundal height formula validated for population [27], 2) LMP, Ballard score, fundal height, and ultrasound [31], or 3) LMP, fundal height, and date of quickening [30]. Last, one publication did not have such data available and used descriptive statistics to link records from outpatient and delivery/pregnancy complication registers [26].

For malaria-centered publications that utilized comparator groups or had explicit information detailing their comparator groups, we found that all publications used contemporaneous/concurrent comparator groups (n = 6). Among these 6 publications, the comparator groups included: pregnant people without malaria [27], pregnant people with an episode of malaria in first trimester [27], pregnant people with antimalarial therapeutic exposure during any period of pregnancy [28,29,30,31,32], and pregnant people (with or without malaria) with no therapeutic exposure [28,29,30,31,32].

4.7 Tuberculosis

For publications where the infectious disease of focus was TB, there were two main exposure groupings: 1) second-line DR-TB or MDR-TB therapeutics and 2) isoniazid-based therapeutics. The publications that focused on second-line therapeutics exposure investigated linezolid [33], a fluoroquinolone [34, 35] and bedaquiline, clofazimine, and levofloxacin specifically [35] to treat DR-TB or MDR-TB. The publications focused on isoniazid-based therapeutics and analyzed the safety of isoniazid preventive therapy (IPT) among pregnant people living with HIV [36,37,38].

Among these publications, common outcomes of interest were adverse birth and neonatal/infant outcomes such as preterm birth, LBW, stillbirth/miscarriage, neonatal mortality, and congenital anomalies [28, 33,34,35,36, 38,39,40]. All publications analyzed some maternal TB/IPT treatment outcomes and adverse maternal outcomes, but they differed in the specific outcomes that were analyzed. Some publications defined adverse maternal outcomes as overall maternal morbidity and mortality [33], whereas other publications defined adverse maternal outcomes as therapeutics-related adverse events such as liver impairment, kidney function impairment, gastrointestinal disorders, or psychiatric disorders [33] or loss of weight, dizziness, rash, nausea, and ototoxicity [34]. It is also notable that Gupta et al had a primary focus on treatment-related, maternal, AEs [36]. Tuberculosis treatment outcomes were typically framed by WHO TB guidelines that defined outcomes as: cured, completed, died, lost to follow up, or not evaluated [33,34,35, 39, 40].

While some publications focused on IPT among pregnant people living with HIV [36,37,38], the primary objective of most publications was to assess the safety of second-line TB therapeutics in relation to maternal, birth, and neonatal/infant outcomes [28, 33,34,35, 39, 40]. In some publications, particularly the publications centered on IPT in pregnant people living with HIV, a key point of analysis was the timing of exposure during pregnancy [36,37,38]. Specifically, some of these publications compared the effects of IPT initiated immediately during pregnancy for 28 weeks or at a deferred time after delivery (12 weeks after delivery) [36, 37], while another analyzed the effects of IPT at any time during the second or third trimester of pregnancy [38]. It should also be noted that, among the publications with a focus on second-line TB medication, one publication compared the timing of exposure between the first, second, or third trimester [40]. This publication noted that the outcome of live birth was significantly associated with trimester of initiation, with first trimester initiation being associated with the lowest rates of live birth [40]. Analyzing outcomes based on trimester exposure allowed this publication to conclude that TB treatment should be initiated in either the second or third trimester (preferably third), but not in the first trimester [40].

For TB-centered publications that utilized comparator groups or had explicit information detailing their comparator groups, we found that all publications used contemporaneous/concurrent comparator groups (n = 9). For publications with a focus on second-line TB therapeutics, these publications compared pregnant people with DR-TB using second-line TB treatment and those with DR-TB not using second-line TB treatment [34, 35]. However, Mokhele et al specifically compared pregnant people living with DR-TB and HIV with pregnant people living with DR-TB [34]. For publications that focused on IPT in pregnant people living with HIV, most compared a group that was immediately given IPT during pregnancy to a deferred group that received IPT after delivery (12 weeks after) [36, 37], while one matched pregnant people living with HIV exposed to IPT with a control comparison group of pregnant people living with HIV but with no IPT exposure [38].

4.8 Vaccine-Related Publications

Among the vaccine-related publications, the specific vaccines studied were for pertussis, tetanus, and diphtheria (n = 1) and influenza (n = 3; Table 3). Some of the publications (n = 2) assessed vaccine safety [41, 42], one publication focused on vaccine efficacy [43], and one publication focused on both vaccine safety and efficacy [44].

The exposure for vaccine safety-focused publications was during a specific time during the pregnancy. Some publications focused on vaccination during the second or third trimester [41, 42]. One publication assessed both vaccine safety and efficacy in Nepal and included two cohorts, one cohort in which individuals were vaccinated as soon as pregnancy was confirmed and a second cohort where individuals were identified as pregnant and then randomly allocated a week during pregnancy to receive the vaccine [44]. The primary outcomes of interest for this publication was incidence of maternal influenza-like illness, incidence of LBW, and incidence of laboratory-confirmed infant influenza [44]. The secondary outcomes for this publication were maternal laboratory-confirmed influenza, rate of infant influenza-like sickness, and SGA [44].

One publication assessed safety of two pertussis, tetanus, and diphtheria vaccines and had the following outcomes of interest: AEs following immunization, pregnancy outcomes (early abortion and complications like preterm labor and preeclampsia), and infant outcomes (congenital anomalies, APGAR scores, stillbirth) [41]. Another publication assessing influenza vaccine safety assessed whether the inactivated influenza vaccine may have non-specific effects that increase the risk of other infections in pregnant people. Thus, the outcomes of interest in this publication were maternal all-cause mortality, maternal mortality from presumed infectious causes, miscarriage/stillbirths, infant (aged up to 6 months) all-cause mortality, and infant (aged up to 6 months) mortality from presumed infectious causes (non-influenza related) [42].

One publication focused on assessing vaccine efficacy of maternal influenza vaccination but also assessed birth, maternal, and neonatal/infant outcomes [43]. This publication based in Nepal, Mali, and South Africa had different vaccination administration timing, ranging from 17 to 36 weeks gestation, depending on site location [43]. Omer et al examined maternal, birth, and neonatal/infant outcomes based on vaccination-site location and assessed vaccine efficacy at differing periods after vaccination [43]. For example, Omer et al concluded that vaccine efficacy against infant influenza was higher in the first 2 months of life yet did not demonstrate efficacy after 4 months of life [43]. This publication had several outcomes of interest, including: overall vaccine efficacy against maternal and infant PCR-confirmed influenza, duration of protection, the effect of gestational age at vaccination on efficacy, adverse birth outcomes (including LBW, stillbirth, preterm birth, and SGA), and infant growth up to 6 months [43].

For vaccine-centered publications that utilized comparator groups or had explicit information detailing their comparator groups, we found that all publications used contemporaneous/concurrent comparator groups (n = 3). Among these 3 publications, comparator groups included: pregnant individuals given a vaccine of interest during different times of the year [44], pregnant individuals given a vaccine other than the vaccine of interest [42, 43], and pregnant individuals given a saline placebo [42,43,44].

5 Discussion

5.1 Summary of Evidence

In this scoping review, we identified 31 publications spanning at least 24 LMICs for PV in pregnancy projects that focused on therapeutics or vaccines for infectious diseases, including HIV, malaria, TB, pertussis, and influenza. The study settings were frequently from Africa, study designs varied from cohort or cross-sectional studies to clinical trials, contemporaneous comparator groups were commonly used, and birth and neonatal/infant outcomes were most commonly reported, although marked heterogeneity existed in definitions and ascertainment of specific measures. While currently published pregnancy PV literature offers insights into the PV landscape in LMICs, further work is needed to standardize definitions and measurements in these projects. Given recent interest in pregnancy PV projects (e.g., since we ran our search string, there has been a number of highly pertinent publications [45,46,47,48]), we believe this scoping review will aid persons working in this field to better strategize for the next steps.

The following are some common themes arising from the publications included in this review, which also illuminates gaps to address in future work: (1) we identified that all publications regarded infectious diseases and largely took place in Africa, (2) most included neonatal/infant outcomes, and (3) almost all covered a single therapeutic exposure. First, the heavy focus on infectious diseases may reflect their large burden and sizable treatment campaigns in the African region [49] but also the funding priorities of sponsors, largely coming from HICs [50]. For instance, our review yielded no publications on vaccinations largely only applicable in LMICs, such as yellow fever, meningococcal, or malaria vaccinations. Related to this, the burden of non-communicable diseases is increasing in LMICs [51] and is predicted to become the leading cause of morbidity and mortality in LMICs in coming decades. Thus, expansion of PV projects to include non-communicable disease medications is urgently needed. Second, all publications included birth outcomes and a majority included neonatal outcomes, usually assessed at birth or within the first 30 days after birth. Some (48 % of publications) included maternal outcomes, such as maternal death and preeclampsia/eclampsia, though maternal safety measures, such as AEs, or chronic exposures pre-pregnancy were often overlooked (noting only one publication used the WHO Vigibase system meant to standardize adverse drug events reporting for LMICs) [52]. Perhaps the most overlooked set of outcomes was longer-term child outcomes; 65% of publications included short-term infant outcomes but no studies tracked child growth outcomes out to two years past birth, let alone tackled neurodevelopmental outcomes out to even longer periods of time (though admittedly, such data exist outside of the PV paradigm) [53,54,55,56]. Last, nearly all the publications focused on one therapeutic only, i.e., a “vertical” approach to the work. If these projects were better leveraged to cover additional medications, including for vaccines, the broadened scope of the work would have greater returns for the investments and be uniform across health services, regardless of diseases or therapeutics. For instance, no studies investigated exposures of known teratogens, such as valproic acid or certain antiepileptics, newly emerging HIV pre-exposure prophylaxis (PrEP) drugs in their settings, especially for the various HIV-focused projects, which could have quickly pivoted to include this (though PrEP rollout has been relatively recent) [57], or COVID-19 vaccinations (similarly recent), as highlighted in the WHO COVID-19 pregnancy cohort study and similar initiatives [58,59,60]. This is a missed opportunity across many of the projects covered in this review and would be more scientifically robust while also maximizing benefits. Funder priorities or investments in single-topic projects drive such decisions, rather than building broader PV projects in existing health systems, which could more nimbly pivot towards additional analyses when needed. Undoubtedly, work to expand beyond single disease or therapeutic focus and to include longevity in the projects that span multiple years, requires long-term investments to strengthen health systems and research.

For many of the publications, the timing of treatment initiation or vaccination, specifically by trimester, was important, sometimes even deemed “critical,” in determining maternal and neonatal safety. However, a frequent barrier highlighted throughout the infectious diseases therapeutics-related publications is the underreporting of and inaccessibility to health data, including accurate gestational age data within LMICs. To illustrate, many of the analyzed HIV-related publications were concerned with the timing of ART initiation, especially prior to conception. However, it is difficult to calculate the exact time of exposure in terms of gestational days or weeks. While many publications used a variety of factors to assess periconceptual exposure, such as LMP or fundal height, these data were not uniform across publications. This is a substantial concern that hinders pregnancy PV efforts, since without accurate gestational age data, it is difficult to assess whether adverse outcomes are a result of disease exposure or therapeutic exposure (e.g., both malaria itself or anti-malaria drugs in pregnancy that lead to pregnancy loss). Furthermore, there are fundamental issues in some overlap, or even risk of confounding by indication, where the underlying condition being treated may contribute, alongside the treatment exposure itself, to the specific adverse pregnancy outcome; thus, PV projects need to effectively disentangle the two. It is also likely that each exposure-outcome relationship ascertainment is uniquely tied to gestational week-level development in a pregnancy. Therefore, the ultimate goal in PV projects in pregnancy would be to have granularity in exposure timing down to the gestational week or days, especially as a growing number of pregnant people may be exposed to medications in early pregnancy due to the increasing availability of various chronic or short-term medications. An additional challenge in LMICs would be capturing exposures to prescription medications accessed directly from private pharmacies, over-the-counter medications, herbal therapeutics, or intermittent medications. Incomplete pregnancy PV impedes the accessibility to safe, evidence-based, and successful treatment for pregnant people and their infants, as well as supports the continued use of potentially harmful treatments.

In our review, we found a wide breadth of outcomes being measured and reported, including for maternal or neonatal mortality, congenital malformations detected by birth surface exams, stillbirth, preterm birth, LBW and SGA. Among these outcomes, we noticed that definitions of congenital malformations varied the most (especially for neural tube defects), while stillbirth, preterm birth, and LBW were the most standardized. We found that for publications that were explicit in their definition of stillbirth, the gestational age cutoff was between 22 to 28 weeks. This is most likely the result of WHO guidelines that define stillbirth as no signs of life in the fetus after 22 gestational weeks but recommend that this cutoff be extended to 28 weeks if resources for care of very premature neonates are lacking [61]. Furthermore, Loveday et al noted that in South Africa, the legal definition of stillbirth is an infant born dead after 27 complete weeks [35]. Additionally, across the majority of publications, preterm birth was defined as birth prior to 37 weeks of gestation and LBW was defined as birth weight below 2500 g.

Collectively, this work highlights challenges facing PV in pregnancy work in LMICs. First, that our use of the term “PV” in our search string likely limited inclusion of relevant publications and raises the fundamental concern that consensus is lacking around what types of activities constitute PV, which becomes ever more relevant in LMIC settings since post-marketing or government-supported surveillance is not frequently conducted. For instance, medication testing with animal models often offers us the earliest evidence of possible teratogenicity. Spontaneous reporting, although passive, also offers some early signals of possible safety concerns. Thus, some conceptual work and consensus building is required on how best to incorporate various study designs or models for PV work when related nomenclature varies markedly by fields. Second, exposure-outcome relationships in pregnancy require granularity on the frequency, type, and timing of exposure throughout pregnancy and postpartum, and the feasibility of accurate exposure timing in relation to the pregnancy are often limited in LMIC settings. This may be in part because of incomplete records on LMP or exposures, limited access to ultrasonography for dating gestational duration in early pregnancy, delayed antenatal care-seeking, and suboptimal linkages between medication exposures and pregnancy outcomes in LMICs. Third, a focus on building comprehensive, consistent medical record systems in LMICs is paramount in order to inform maternal and neonatal care and safety and to quickly pivot towards a new or emerging threat. While future projects may utilize both active and passive PV surveillance, active forms of PV surveillance are needed to facilitate signal detection and assessment in pregnancy in particular by improving underlying quality of clinical data collected [62]. New pregnancy PV projects in LMICs may expand on existing structures, often bolstered by local or national governments, to become more robust and adapt to the needs of LMICs with changes in health patterns or disease burden. Such an example is the recent Western Cape Pregnancy Exposure Registry [63], which leverages electronic health records collected at the district level. Our findings may help guide future policy to address the gaps identified in existing PV structures, such as in quality and consistency of data collected or study designs utilized, as the limited existing evidence surrounding medication and vaccination use during pregnancy often shapes conservative guidelines that may negatively impact the health of pregnant people. Despite unique challenges facing PV work in LMICs, our review demonstrates that significant work is already underway and lessons from these published studies can help strengthen ongoing or planned PV in pregnancy projects in LMICs.

5.2 Limitations

Although the first of its kind, there are limitations in our work. First, to make our scoping review reproducible, we limited our search of the pregnancy PV projects to published studies. If we expanded our search to non-published projects, for example through gray literature or co-author networks, we would have been able to include a larger number of projects. However, many of the details needed for abstraction could not have been collected in a standard or reproducible fashion. Second, although we had developed search strings to be comprehensive and include terms to pull LMICs pregnancy PV publications, LMICs publications were scarce in PubMed. Furthermore, although we decided to use databases such as Global Index Medicus and SciELO to help mitigate these concerns, we found very few publications from these databases to be relevant to our scoping review focus. This may have skewed the selection of publications that we were able to review and our review may not fully encompass the extent of PV in pregnancy in LMICs. Additionally, the majority of the relevant publications identified were based in sub-Saharan Africa. While this may reflect the burden of infectious diseases and funding priorities of HICs, it is important to note that this may be indicative of a geographical bias and affect the generalizability of our findings. Third, although we altered our search strings to capture the most relevant publications, we still found it necessary to conduct a secondary review to develop a more comprehensive collection of publications. We did not originally plan to incorporate this secondary review; however, it yielded many relevant publications. We also note that our secondary review identified more relevant publications than our primary review, which may indicate that our search string is suboptimal. This could largely be due to our inclusion of the term “PV”, which as we have already discussed, may not be a term universally used. Broader terms, such as “surveillance,” “monitoring,” or “evaluation,” or more specific terms, such as “safety surveillance,” could capture a greater breadth of publications. Indeed, preliminary alterations of our search string to include such terms yielded higher hit results, but we did not identify any additional applicable publications on title review alone. Notwithstanding these limitations, this scoping review is the first of its kind to document published PV in pregnancy work in LMICs and help guide existing and planned related efforts.

6 Conclusion

Our scoping review of pregnancy-centered PV in LMICs publications examined the focus and methodology of current pregnancy PV projects and highlighted the need for more comprehensive projects that thoroughly promote the health of pregnant people and their infants. The selection and defining of exposure variables appeared to be heterogeneous among current projects, particularly in relation to the timing of initiation and duration of treatment, indicating a need for a more harmonized approach. Definition and measurement of outcomes were more consistent throughout the current pregnancy PV literature, likely reflecting the feasibility of measurement of certain adverse maternal, birth, and neonatal/infant outcomes. In order to create the most effective, comprehensive, and responsive PV in pregnancy projects, significant challenges in LMICs will have to be overcome, from capacitating exposure timing granularity to building electronic health records systematically to broadening the focus from infectious diseases to include non-communicable diseases. This scoping review highlights the need for a more thorough body of pregnancy PV research in LMICs and to address the current gaps in LMICs pregnancy PV projects, to ensure access to safe, evidence-based, and effective health care for pregnant people and their infants.