COVID-19 and pregnancy: An umbrella review of clinical presentation, vertical transmission, and maternal and perinatal outcomes

Background We conducted an overview of systematic reviews (SRs) summarizing the best evidence regarding the effect of COVID-19 on maternal and child health following Cochrane methods and PRISMA statement for reporting (PROSPERO-CRD42020208783). Methods We searched literature databases and COVID-19 research websites from January to October 2020. We selected relevant SRs reporting adequate search strategy, data synthesis, risk of bias assessment, and/or individual description of included studies describing COVID-19 and pregnancy outcomes. Pair of reviewers independently selected studies through COVIDENCE web-software, performed the data extraction, and assessed its quality through the AMSTAR-2 tool. Discrepancies were resolved by consensus. Each SR’s results were synthesized and for the most recent, relevant, comprehensive, and with the highest quality, by predefined criteria, we presented GRADE evidence tables. Results We included 66 SRs of observational studies out of 608 references retrieved and most (61/66) had "critically low" overall quality. We found a relatively low degree of primary study overlap across SRs. The most frequent COVID-19 clinical findings during pregnancy were fever (28–100%), mild respiratory symptoms (20–79%), raised C-reactive protein (28–96%), lymphopenia (34–80%), and pneumonia signs in diagnostic imaging (7–99%). The most frequent maternal outcomes were C-section (23–96%) and preterm delivery (14–64%). Most of their babies were asymptomatic (16–93%) or presented fever (0–50%), low birth weight (5–43%) or preterm delivery (2–69%). The odds ratio (OR) of receiving invasive ventilation for COVID-19 versus non-COVID-19 pregnant women was 1.88 (95% Confidence Interval [CI] 1.36–2.60) and the OR that their babies were admitted to neonatal intensive care unit was 3.13 (95%CI 2.05–4.78). The risk of congenital transmission or via breast milk was estimated to be low, but close contacts may carry risks. Conclusion This comprehensive overview supports that pregnant women with COVID-19 may be at increased risk of adverse pregnancy and birth outcomes and low risk of congenital transmission.

Introduction Women undergoing pregnancy, and those at the time of childbirth and puerperium constitute potentially vulnerable populations for COVID-19. Although our understanding of this disease is growing every day, many answers are still needed about the diagnostics and the clinical management methods in these groups, the impact of the disease in pregnant women and newborns, and the potential of mother-to-child transmission.
Although some living guidelines on COVID-19 target the pregnancy population, several clinical questions regarding pregnancy and childbirth remain unanswered [1]. The rate of COVID-19 in pregnant and recently pregnant women attending or admitted to hospital for any reason was around 10%. Pregnancy, in general, does not significantly increase the risk of being infected by SARS-CoV-2 [2].
The World Health Organization (WHO) stated that pregnant women or recently pregnant women who are older, overweight, and have pre-existing medical conditions such as hypertension and diabetes seem to have an increased risk of developing severe COVID-19 [3]. In general, there is a consensus that breastfeeding should be promoted due to its mutual benefits. However, it is not well known whether the virus can be transmitted through breastmilk [4].
Systematic reviews (SRs) constitute an organized effort to collect and comprehensively synthesize the best available evidence on a given topic. Through this panoramic review of SRs, we aimed to answer a series of clinical questions about COVID-19 and pregnancy by summarizing the body of evidence and highlighting the best reviews in completeness and methodological quality.

Objectives
To summarize the clinical presentation, vertical transmission, and maternal and perinatal outcomes in pregnant women with COVID-19 and their neonates.

Methods
We performed an overview of SRs or umbrella review (PROSPERO Registration number CRD42020208783) following Cochrane methods [5] and the Preferred Reporting Items for systematic Reviews and Meta-Analyses (PRISMA) statement [6] and a specific guideline for overviews [7] (S1 File) for reporting.
To be included, SRs had to meet at least four of these criteria, the first three of which were mandatory. The exposures of interest were defined as diagnosis of SARS-CoV-2 infection, SARS-CoV-2 risk factors, diagnostic tests, or treatments. Pregnant women without interventions or exposures under study, including active or inactive comparators, usual care, or placebo, were defined as comparison groups. Any pregnancy or neonatal outcomes, including clinical presentation, laboratory, and radiological findings, were included (S2 File).

Search strategy
From January to October 2020 an experienced librarian searched the Cochrane Library, MED-LINE, EMBASE, Latin American and Caribbean Health Sciences Literature (LILACS), Science Citation Index Expanded (SCI-EXPANDED), China Network Knowledge Information (CNKI), WHO Database of publications on SARS-CoV-2, EPPI-Centre map of the current evidence on COVID-19, guidelines published by national and international professional societies (e.g., ACOG, RCOG, FIGO), pre-print servers (ArXiv, BiorXiv, medRxiv, search.bioPreprint), and COVID-19 research websites (PregCOV-19LSR, Maternal and Child Health, Nutrition: John Hopkins Centre for Humanitarian health, the LOVE database) We also searched the reference lists of included SRs. No language or publication status restrictions were applied (The whole search strategy is presented in the S3 File).

Study selection data extraction and quality appraisal
Pairs of reviewers independently screened titles and abstracts. We retrieved all potentially relevant full-text study reports/publications, and two reviewers independently evaluated the fulltexts, recording the reasons for exclusion of the ineligible studies. Disagreements were resolved through discussion of the review team. This process was performed using the web-based software COVIDENCE [9].
Pairs of reviewers independently performed the data extraction through an online extraction form previously piloted in five studies. We recorded publication date, number of included studies, number of included participants, quality items, and the components of our research questions (population, exposition, comparisons, and outcomes). Discrepancies were resolved by consensus.
Pairs of reviewers independently assessed the quality of SRs through the AMSTAR-2 tool [10]. The instrument has 16 items. It is not intended to generate an overall score but provides a categorical rating based on critical domains: protocol register, adequacy of the literature search, justification for excluding individual studies, risk of bias from individual studies being included, appropriateness of meta-analytical methods, consideration of risk of bias when interpreting the results, assessment of publication bias. The overall quality or confidence in the results of the review can be rated as "high" (no or one non-critical weakness), "moderate" (more than one non-critical weakness), "low" (one critical flaw with or without non-critical weaknesses), and "critically low" (more than one critical flaw with or without non-critical weaknesses). Discrepancies were resolved by consensus. We did not assess the quality of the included primary studies in the SRs nor the quality of reporting of each SR.
The purpose of our study is to present and describe the current body of SRs evidence on COVID-19 in maternal and neonatal health. Therefore, we synthesized the results of all relevant SRs, regardless of topic overlap, considering that re-extracting and re-analyzing outcome data from non-overlapping studies was unfeasible and outside the scope of this overview. Additionally, for this scoping synthesis, we selected the best SR that answers a specific question according to pre-defined prioritizing criteria: most relevant, most comprehensive, most recent, and highest quality determined by AMSTAR-2 [10]. For these prioritized reviews, pairs of reviewers independently assessed the risk of bias of the priori SR using the tool Risk of Bias in Systematic Review (ROBIS) [11] and the GRADE approach evaluating the certainty of evidence of each outcome [12,13]. We did not assess the quality of the included primary studies in the SRs nor the quality of reporting of each SR.
We presented summaries of the findings in a format suitable for decision-makers, previously validated during the SUPPORT project [14], focusing on low-and middle-income countries (LMICs) for selected research questions.
Pre-specified subgroups were designated by sampling frame (universal, symptom-based, or risk-based testing), timing of suspicion/diagnosis (pregnancy or postnatal period), trimester of suspicion/diagnosis (first, second or third), country income-level (high or low-and middleincome country), and maternal risk status (low or high).
Among the primary studies included in these reviews, the main study designs were case reports, case series, and other observational studies with or without a comparison group.
The number of included SARS-CoV-2 positive pregnant women was highly heterogeneous across reviews. While the Li review [32] and the Mullins review [33] included only 19 of them, the Allotey review [2] included more than ten thousand. Fig 2 shows the percentual degree of overlap of included systematic reviews' primary studies, which in general was low (In S5 File we listed all primary studies with DOI included by each review and in S6 File the degree of overlap in absolute numbers).
Concerning the overall quality, based on AMSTAR-2 (see Table 1), most SRs were classified as "critically low" (n = 61), four as "low" [2,[34][35][36], and only one as "moderate" [37]. For the prioritized systematic reviews, we also used the ROBIS tool (one was classified as "low risk of bias" [2], one as "unclear risk of bias" [38] and the other two as "high risk of bias" [37, 39]) (S7 File, by each domain). The mean ± standard deviation of non-negative classifications was 9.21 ± 2.51. The most common weaknesses (>50% of SRs with unmet domain) were: not reporting the funding for the studies included in the review (n = 65), not providing a list of excluded studies and justifying the exclusions (n = 61), not accounting for risk of bias (RoB) in individual studies when interpreting or discussing the results of the review (n = 50), not providing a satisfactory discussion of any heterogeneity observed (n = 50), not providing a protocol (n = 47) and not using satisfactory techniques for assessing the RoB in individual studies included in the review (n = 39).
Below the main findings are described for the five available bodies of evidence identified in our overview: clinical presentation in pregnant women, pregnancy outcomes, clinical presentation in neonates, neonatal outcomes, and vertical transmission.
There was a high level of heterogeneity of the reported values for each outcome across the included SRs. Table 2 shows these value ranges (The S8 File shows the total number of newborns and pregnant women included and the numerical data of each outcome at the review level).
The main findings are described briefly below.

Clinical presentation in pregnant women
Thirty-nine reviews [2, 15, 16, 18, 19, 21-23, 25, 28-37, 39-58] showed information regarding clinical manifestations during pregnancy. Fever and mild respiratory symptoms were the most frequently reported symptoms and raised C-reactive protein level, lymphopenia, raised white cell count and raised procalcitonin level were the most frequent laboratory findings. Signs of pneumonia on X-rays or computed tomography (CT) were also frequently reported ( Table 2,  Tables 1, 2 in S8 File). The Allotey review [2] was chosen as the best review for reporting clinical presentation in pregnant women diagnosed with COVID-19 according to predefined criteria described in the methods section (Table 3 and S9 File). It reported that fever (40%), cough (39%, involving 28 studies and 8317 pregnant women), and dyspnea (19%) were the most common symptoms. It also reported that raised C reactive protein levels (49%) and lymphopenia (35%) were the most common laboratory findings. Regarding findings on X-rays or CT, ground glass appearance had a prevalence of 69%, and any other abnormalities on CT had a prevalence of 65%. Finally,  The current evidence revealed a low possibility of vertical transmission of COVID-19, and antibodies against SARS-CoV-2 were detected among vertically exposed but negative infants. 6 Critically low A total of 1141 neonates were born of which, 281 (25%) were preterm (<37weeks). SARS-CoV-2 testing was positive for 39/1005 neonates (3.9%), 16/43 mother-baby dyads (37.2%) were preterm (<37weeks), 9 (21%) were low birth weight (<2500 g), and 27 (62.8%) were born by cesarean section ( Table 2). All 43 were tested for SARS-CoV-2 infections using nasopharyngeal or oropharyngeal specimen, and 19 neonates (44.2%) have positive RT-PCR for SARS-CoV-2.58 live-born SARS-CoV-2 cases, 4 (7%) were congenital in origin (2 confirmed, one probable, and one not sure), 41 were acquired in the postpartum period, and the remaining 13 neonates could not be classified due to non-available (Continued ) Of the ten reported cases, only three are likely to be vertically transmitted, while seven occurred in the post perinatal period and are likely to have been postnatally acquired. All neonates had a mild course, recovered, and were negative on re-testing.
When compared with non-pregnant women of reproductive age with COVID-19, the reported odds of admission to the ICU was 1.
Laboratory data (confirmed cases through Polymerase chain reaction [PCR], elevated SARS-CoV-2 IgM, and IgG antibodies) and imaging data (radiographic pneumonia) were also reported.

Vertical transmission
Forty-six reviews [4, 15, 17, 18, 20, 21, 23-27, 31, 33, 35-39, 41, 42, 44, 45, 47, 49-52, 54, 56-64, 66-68, 70, 71, 73-76] reported mother-to-child SARS-CoV-2 transmission ( Table 2,  Table 6 in S8 File). Most studies only reported the proportion of infants' positive cases without evaluating breast milk or congenital/perinatal transmission. In the reviews reporting breast milk or congenital/perinatal transmission the sample analyzed was generally small for these outcomes. The Centeno-Tablante review [38] was chosen as the most appropriate review to answer this question according to predefined criteria ( Table 3 and S10 File). This review included 37 papers with a total of 889 infants. Of the 72 infected mothers whose breast milk samples were laboratory-confirmed to contain the COVID-19 antigen, 14 infants were found to be infected with COVID-19. Eight of the twenty-three infants that were breastfed were infected, two of the eighteen infants that received a breast milk substitute were infected, two of four infants that received mixed feeding were infected, and two of the twenty-three infants that did not report  Table 3). This review included 24 studies, case series, and case reports, including a total of 155 neonates. Ninety neonates were tested for COVID-19, of which three were positive. The review also evaluated the presence of SARS-CoV-2 in amniotic fluid (1/32), umbilical cord blood (0/34), and placenta (1/3). While there was no vertical mother-to-child transmission, additional good-quality studies are needed to determine whether vertical transmission is possible.
The key messages from the prioritized systematic reviews (by the most current search date, the more significant number of included studies, and greater adequacy to address the outcomes) are presented in Table 3.

Discussion
This systematic review of SRs integrated the most consolidated evidence synthesis regarding the effects of COVID-19 on maternal and neonatal health.
Most SRs (92.4%) were classified as "critically low" in overall confidence, using the AMSTAR-2 tool, likely due to the urgent demand of information for this hot topic. For the prioritized systematic reviews, we also used the ROBIS tool (one was classified as "low risk of bias" [2], one as "unclear risk of bias" [38] and the other two as "high risk of bias" [37,39]) and the GRADE approach for each of their outcomes. The certainty of evidence was rated as "low" to "very low" due to study design, risk of bias, inconsistency and/or imprecision.
The COVID-19 related symptoms manifest from one third to two thirds less often in pregnant women than in non-pregnant women of reproductive age [2]. While testing for SARS-CoV-2 in non-pregnant women is usually based on symptoms or contact history, testing in pregnant women is generally done for reasons that might not be related to COVID-19. Affected women were at higher risk of requiring admission to an ICU or invasive ventilation. Pregnant women with COVID-19 are also at an increased risk of receiving cesarean sections, delivering preterm and their babies being admitted to a NICU. Higher age, higher body mass index, and pre-existing comorbidities might be associated with severe disease [2]. Stillbirth and neonatal death rates are low in women with suspected or confirmed COVID-19. However, this evidence is based on a few large comparative studies. The substantial heterogeneity identified could be related to using different sampling techniques, the different sampling techniques, and the differential baseline risk of participants [2].
Most infected neonates were asymptomatic. The most frequent symptoms were fever (0-50%) or mild respiratory symptoms [39]. Low birth weight and preterm birth were the most frequently reported neonatal outcomes. The neonates born to mothers diagnosed with COVID-19 presented an odds three times higher admission to a NICU compared with those born to mothers without the disease [2]. The risk of congenital transmission [37] or transmission via breast milk [38] is estimated to be low to very low, but there is a higher risk of transmission due to close contact by droplet or airborne transmission.
The high rate of asymptomatic presentation in pregnant women with COVID-19 [2] may be explained by the screening strategy for COVID-19 and the low thresholds for testing during pregnancy. Even detecting more pregnant women with mild disease, higher admissions to the ICU, or invasive ventilation were observed when they are compared with non-pregnant women of reproductive age with COVID-19 [79]. A Swedish study also suggested an more admissions to an ICU and higher requirement of invasive ventilation in pregnant women than non-pregnant women [80].
Similar to the general population, pre-existing comorbidities seemed to be risk factors for severity of COVID-19 in pregnancy [81]. Adverse outcomes related to COVID-19 were not found to be higher in women at the third trimester nor in multiparous ones-but the existing sample sizes are not large (less than 300 women). Chronic hypertension and pre-existing diabetes were associated with maternal death in pregnant women with COVID-19, and both are recognized risk factors in the general population. The low numbers of studies does not allow to stablish the cause of death for these women. A slight increase in rates of preterm birth in pregnant women with COVID-19 was observed when compared to those without the disease. These preterm births could be medically indicative, since rates of spontaneous preterm births in affected women were similar to those before the pandemic. More than 60% of pregnant women underwent cesarean section in the non-comparative studies. This is three times the global rate of cesarean sections worldwide [82], and deserves future research.
Surely, the precision will improve as more data is published. The overall rates of stillbirths and neonatal mortality are not observably higher than the background rates. The indicators for admissions to the NICU, of about 25% of neonates from affected mothers were not reported. Countries' regulations on isolation of exposed infants to the virus may have influenced these rates.
Sixty-seven percent of newborns delivered by mothers with COVID-19 antibodies had SARS-CoV-2 IgG, but not IgM antibodies [83]. This finding against vertical transmission is consistent with our own findings of low to very low risk of this mechanism of congenital transmission.
To our knowledge, there is only one overview of SRs published that reports maternal and perinatal outcomes related to COVID-19 and pregnancy [84], including 52 SRs. This overview searched studies until September 2020 and did not include 14 SRs that were found in our overview, probably explaining the lower level of overlap observed in our study. The authors did not assess the quality of reporting of each SR, but they assessed the risk of bias of each included SR using the ROBIS tool [11]. The high risk of bias identified in this overview is consistent with the "critically low" confidence presented in our study by applying the AMSTAR-2 tool [10].
Initial studies involved women from China, but later in 2020, reports came also from regional or national data from European countries and Latin America. The study design has also changed from small case series and case reports to extensive observational data, with recent studies also being comparative. Variations in the criteria for doing testing (symptomsbased, close contact) and sampling methodologies explain differences in the prevalence of COVID-19. Moreover, the findings only apply to women attending the hospital for any reason. The true prevalence of COVID-19 in pregnancy is likely to be lower when all pregnant women are included.

Strengths and limitations
This overview has several strengths. First, we followed sound methodology to conduct the present overview of systematic reviews. Second, we included systematic reviews without language restrictions. Third, we adhered to rigorous quality appraisal for the conduction of systematic reviews (AMSTAR-2 tool), which was independently assessed by pairs of reviewers and discrepancies solved by consensus. We summarized and critically appraised an important amount of evidence that is relevant to health decision-making (See examples of policy briefs in S9 and S10 Files) and highlighted evidence gaps that could guide future research. Finally, we conducted a sensitive and comprehensive search strategy to reduce the risk of missing relevant studies. We synthesized the results of all relevant SRs, highlighting their overlap through a matrix of primary studies by SR, and we also selected the best SR that answered a specific question according to pre-defined criteria of relevance, comprehensiveness, data update, and quality. We presented these SRs through a summary of the findings tables with the certainty of evidence according to the GRADE approach. Our review integrated the evidence generated by different independent groups, which could improve the robustness of our findings.
Our study is not exempt from limitations. A main limitation is that the last search was run in October 2020. This is due to the time needed to perform a thorough SR. The general confidence of the included SRs was "critically low", and the certainty of evidence was "low" to "very low". Additionally, there is a scarcity of data comparing pregnant women with non-pregnant women or comparing pregnant women with and without COVID-19 [2]. We did not evaluate the risk of bias of the primary studies, nor did we undertake a pooled analysis by the outcome, as was originally stated in our protocol. Nevertheless, there was great heterogeneity of methods, study designs, and estimations that would preclude a meta-analysis.

Implications for clinical practice and research
Although pregnant women with COVID-19 could be less symptomatic than the general population, the overall pattern is similar. However, the admission rates to ICUs and invasive ventilation in pregnant women with COVID-19 could be higher than non-pregnant women. Mothers with pre-existing comorbidities will need to be considered as a high-risk group for COVID-19, along with those who are obese and of advanced maternal age.
Physicians need to weight the need for antenatal visits against unnecessary exposure and use of virtual meetings whenever possible. Infected pregnant women with before-term gestations might need care in special facilities for these cases since the neonates born to mothers diagnosed with COVID-19 are at three times the risk of admission to NICUs compared with those born to mothers without the disease.
Further data are needed to assess robustly if pregnancy-related maternal and neonatal complications are increased in women with COVID-19 than those without the disease. Similarly, the association between other risk factors, such as ethnicity and pregnancy-specific risk factors such as preeclampsia and gestational diabetes on both COVID-19 related and pregnancyrelated outcomes needs evaluation. Preeclampsia was reported to be associated with severe COVID-19 in small studies, but requires a further assessment as the clinical presentation of severe preeclampsia could mimic worsening COVID-19 [85]. Robust registers of pregnancy data by trimester of exposure are essential to determine the effects of COVID-19 on early maternal and neonatal outcomes.
Systematic reviews of RCTs are the highest quality evidence to inform guidelines, and poorquality systematic reviews will still directly impact on clinical care. Despite urgency for evidence, systematic reviews still need to adhere to the highest standards of reporting and conduction, more so in the presence of pre-prints, reports, and media statements. Primary studies will need to explicitly state if duplicate data have been included to avoid double counting participants in evidence synthesis. Individual participant data meta-analysis and network metaanalysis of the emerging cohorts are critical to evaluate clinical manifestations and outcomes by underlying risk factors and also to determine the differential effects of interventions to reduce the complication rates.