Clinical risk factors of adverse outcomes among women with COVID-19 in the pregnancy and postpartum period: A sequential, prospective meta-analysis


                  Objective
                  This sequential, prospective meta-analysis (sPMA) sought to identify risk factors among pregnant and postpartum women with COVID-19 for adverse outcomes related to: disease severity, maternal morbidities, neonatal mortality and morbidity, adverse birth outcomes.
               
                  Data sources
                  We prospectively invited study investigators to join the sPMA via professional research networks beginning in March 2020.
               
                  Study eligibility criteria
                  Eligible studies included those recruiting at least 25 consecutive cases of COVID-19 in pregnancy within a defined catchment area.
               
                  Study appraisal and synthesis methods
                  We included individual patient data from 21 participating studies. Data quality was assessed, and harmonized variables for risk factors and outcomes were constructed. Duplicate cases were removed. Pooled estimates for the absolute and relative risk of adverse outcomes comparing those with and without each risk factor were generated using a two-stage meta-analysis.
               
                  Results
                  We collected data from 33 countries and territories, including 21,977 cases of SARS-CoV-2 infection in pregnancy or postpartum. We found that women with comorbidities (pre-existing diabetes, hypertension, cardiovascular disease) versus those without were at higher risk for COVID-19 severity and pregnancy health outcomes (fetal death, preterm birth, low birthweight). Participants with COVID-19 and HIV were 1.74 times (95% CI: 1.12, 2.71) more likely to be admitted to the ICU. Pregnant women who were underweight before pregnancy were at higher risk of ICU admission (RR 5.53, 95% CI: 2.27, 13.44), ventilation (RR 9.36, 95% CI: 3.87, 22.63), and pregnancy-related death (RR 14.10, 95% CI: 2.83, 70.36). Pre-pregnancy obesity was also a risk factor for severe COVID-19 outcomes including ICU admission (RR 1.81, 95% CI: 1.26,2.60), ventilation (RR 2.05, 95% CI: 1.20,3.51), any critical care (RR 1.89, 95% CI: 1.28,2.77), and pneumonia (RR 1.66, 95% CI: 1.18,2.33). Anemic pregnant women with COVID-19 also had increased risk of ICU admission (RR 1.63, 95% CI: 1.25, 2.11) and death (RR 2.36, 95% CI: 1.15, 4.81).
               
                  
                     Conclusion
                  
                  
                     We found that pregnant women with comorbidities including diabetes, hypertension, and cardiovascular disease were at increased risk for severe COVID-19-related outcomes, maternal morbidities, and adverse birth outcomes. We also identified several less commonly-known risk factors, including HIV infection, pre-pregnancy underweight, and anemia. Although pregnant women are already considered a high-risk population, special priority for prevention and treatment should be given to pregnant women with these additional risk factors.
                  
               

We pooled and re-analyzed data from global collaborators. We assessed high-priority risk factors 235 and two dozen, consistently defined maternal and newborn outcomes. Given the large sample, 236 including data from low-and middle-income countries, we generated estimates on rare outcomes 237 (maternal mortality, stillbirth) and risk factors (anemia, underweight, HIV) where data has been 238 lacking. 239 J o u r n a l P r e -p r o o f

Abstract: 240
Objective: This sequential, prospective meta-analysis (sPMA) sought to identify risk factors 241 among pregnant and postpartum women with COVID-19 for adverse outcomes related to: disease 242 severity, maternal morbidities, neonatal mortality and morbidity, adverse birth outcomes. studies. Data quality was assessed, and harmonized variables for risk factors and outcomes were 252 constructed. Duplicate cases were removed. Pooled estimates for the absolute and relative risk of 253 adverse outcomes comparing those with and without each risk factor were generated using a two-254 stage meta-analysis. 255 256 Results: We collected data from 33 countries and territories, including 21,977 cases of SARS-257 CoV-2 infection in pregnancy or postpartum. We found that women with comorbidities (pre-258 existing diabetes, hypertension, cardiovascular disease) versus those without were at higher risk 259 for COVID-19 severity and pregnancy health outcomes (fetal death, preterm birth, low 260 birthweight). Participants with COVID-19 and HIV were 1.74 times (95% CI: 1.12, 2.71) more 261 likely to be admitted to the ICU. Pregnant women who were underweight before pregnancy were 262 at higher risk of ICU admission (RR 5.53, 95% CI: 2.27, 13.44), ventilation (RR 9. 36 States (N=7950) determined that pregnant women over 25 years of age, with pre-pregnancy 300 obesity, chronic lung disease, chronic hypertension, and pregestational diabetes mellitus had a 301 32% to 85% increased risk of moderate-to-severe COVID-19, compared to pregnant women free 302 of these conditions 9 . Pregnant women with three or more underlying health conditions had over 303 twice the risk of moderate-to-severe COVID-19 illness than those with no comorbidities 9 . 304 305 In the general population, nutritional status has been introduced as a potential risk factor for severe 306 COVID-19. A meta-analysis of seven studies (N=9,912) found that among people with COVID-307 19, those with anemia had 2.44 higher odds of severe illness than non-anemic people 10 . A 308 scientific review found sufficient intake of micronutrients, proteins, diet fiber, short-chain fatty 309 acids, and omega-3 polyunsaturated fatty acids may act as a protective factor against severe illness 310 in COVID-19 patients 11 . Further research is required for pregnant women, for whom nutritional 311 guidance would be particularly useful. 312

313
There is an urgent need to pool high-quality and internationally representative data assessing the 314 underlying risk factors and outcomes linked to COVID-19 in pregnancy. Currently, scarcity of 315 similarly collected and analyzed data hampers our ability to make strong recommendations for the 316 introduction and prioritization of new pharmaceutical interventions in pregnancy. The primary aim 317 of this sequential, prospective meta-analysis (sPMA) is to accrue harmonized global data to inform 318 policy and practice, grounded in the epidemiology of COVID-19 in the pregnancy, peripartum, 319 and postnatal periods. 320

321
Objectives 322 In this analysis, we sought to identify risk factors among pregnant and postpartum women with 323 SARS-CoV-2 infection for adverse outcomes related to: i) disease severity; ii) maternal 324 morbidities; iii) fetal and neonatal mortality and morbidity; iv) adverse birth outcomes. 325

Methods 327
We registered the protocol for this prospective meta-analysis via PROSPERO (ID: 188955) in 328 May 2020, and the full protocol has been published elsewhere 12 . The meta-analysis project was 329 determined to be exempt from IRB review. processed data to review data quality, identify outliers, and reconstruct variables to align with 346 harmonized definitions of outcomes as defined in our protocol. We shared results with 347 investigators for review and approval. For study sites unable to share IPD directly, the technical 348 team worked with investigators to implement a common set of Stata codes to complete the same 349 process of review, data quality checks, and harmonization. 350

351
In cases where studies collected data from overlapping catchment areas, we worked with 352 investigators to identify and remove potential duplicates from the analysis. Because of the 353 harmonization process and removal of overlapping data, there are some differences between our 354 study results compared to original published studies; these differences are summarized in Table  355 S1. 356 357 Assessment of risk of bias. We use an adapted Newcastle Ottawa Scale to review study quality and 358 risk of bias for each participating study; criteria for determination of high or low risk for each 359 study design element are presented in Table S2 13 . 360 361 Outcomes. We examined 24 outcomes related to: i) COVID-19 severity; ii) maternal morbidities; 362 iii) fetal and neonatal morbidity and mortality; iv) adverse birth outcomes. Specific definitions of 363 each outcome-as well as 4 alternative outcomes used in sensitivity analyses-are presented in 364 Nutrition-related risk factors included body mass index (BMI) and anemia. We relied on pre-379 pregnancy BMI to determine the category for each participant, and we examined two risk factors: 380 underweight (BMI <18.5 kg/m 2 ) and obesity (BMI >30 kg/m 2 ). Both risk factors are compared to 381 a reference group of participants who were normal weight or overweight pre-pregnancy (BMI 382 18.5-<30 kg/m 2 ). Anemia was diagnosed based on a hemoglobin measurement <11 g/dL at the 383 time of COVID-19 diagnosis. 384

385
We considered two age groups as risk factors: younger maternal age (15-19 years) and older 386 maternal age (35-45 years). Both groups are compared to a reference group of women aged 20-34 387 years. Lastly, we considered being symptomatic for COVID-19, as compared to those with no 388 symptoms, as a risk factor for the outcomes of interest. 389

390
Generating study-specific estimates. We used a standard set of analysis codes to calculate study-391 specific estimates comparing those with and without each risk factor (proportions and relative risks 392 with 95% confidence intervals (CI)) for each participating study. Within each study, individual 393 participants were excluded from the analysis if they were missing data on the risk factor of interest. 394 Any study missing more than 25% of the data on an outcome of interest was excluded from that 395 specific analysis. 396 397 Data synthesis. We applied a 2-stage IPD meta-analytic framework to generate pooled absolute 398 risks and relative risks, with 95% CI for each risk factor-outcome pair when there were three or 399 more studies with available data. We presented unadjusted estimates because the goal of this study 400 was to present descriptive epidemiological data among a group of people (pregnant women with 401 COVID-19 and their infants), rather than to examine a causal relationship 18 , 19 . To estimate the 402 pooled absolute risk for each adverse outcome overall and within risk factor groups, we used a 403 logistic-normal random effects model 20 . In cases where the logistic-normal model did not 404 converge, we employed a random effects model with the Freeman-Tukey double arcsine 405 transformation, to ensure stable estimates and approximate asymptotic normality 21 . We used a 406 Dersimonian and Laird random-effects meta-analysis to generate relative risks for each risk factor-407 outcome pair and assessed heterogeneity across studies using the I 2 statistic. 408

409
We excluded studies with zero total events from that particular analysis. In case of zero events 410 within a risk factor subgroup, we applied a continuity correction of 0.5 when calculating pooled 411 absolute risks. For pooled relative risks, we applied a continuity correction of the inverse number 412 of events in the opposite group within the same study for the risk factor-outcome pair. All meta-413 Study selection. We included data from 21 studies conducted across 33 countries and territories 417  earning at least 4 out of 5 or 4 out of 6 stars across all outcome categories where that study was 454 included in the analysis. The most common cause for high risk of bias rating was related to 455 representativeness of the study population; 5 of 21 studies did not collect data on the reason for 456 screening for individual patients. Another 8 studies primarily used methods to identify cases that 457 were deemed to be at higher risk of bias (such as testing for clinical concern based on symptoms 458 or travel). In total, 13 of 21 studies had elevated risk of bias in this area. 459 Overall incidence. Overall event incidence for each site is shown in Figure 2. There is considerable 462 heterogeneity between studies for most assessed outcomes. This is likely due to a combination of 463 factors including varying sampling frames across studies, true differences in the incidence of 464 outcomes in the general population, and underlying differences in the standard of care provided 465 by health systems in each setting. Comorbidities. We found that pregnant women with COVID-19 who also had chronic illnesses, 470 including diabetes, hypertension, and cardiovascular disease, were at higher risk for most 471 outcomes related to COVID-19 severity, as well as pregnancy-related death (  Table S8), compared to those without these chronic health conditions. 477 478 Pregnant women with COVID-19 and one of these chronic conditions were at higher risk for 479 maternal morbidity, including placental abruption, preeclampsia, preeclampsia or eclampsia, 480 hypertensive disorders of pregnancy, preterm labor, and any cesarean delivery. Those with 481 hypertension or cardiovascular disease were also at increased risk of having an intrapartum 482 cesarean delivery. Babies born to mothers with both COVID-19 and one of these chronic 483 conditions were at higher risk for mortality (stillbirth, perinatal death, and neonatal death), as well 484 as NICU admission. These infants were more likely to be born preterm, low birthweight, and small-485 for-gestational age. 486 487 Although less data was available on HIV coinfection with COVID-19 during pregnancy, we found 488 coinfection increased the risk of severe COVID-19 disease ( Although data was limited, we found an increased risk of COVID-19 severity among pregnant 522 women with anemia at the time of COVID-19 diagnosis compared to those without anemia (Table  523 3 Primiparity. Overall, we found limited differences in risks of adverse outcomes among CI: 0.46, 0.77, 8 studies, 4,249 pregnancies) and were more likely to experience preeclampsia or 553 eclampsia, any hypertensive disorders of pregnancy, or intrapartum cesarean delivery, compared 554 to multiparous women (Table S15). 555 556 Symptomatic SARS-CoV-2 Infection. We found increased risks for adverse outcomes related to 557 COVID-19 severity among pregnant women with symptomatic infection compared to those with 558 asymptomatic SARS-CoV-2 infection, including ICU admission, any critical care, and pneumonia 559 (Table S16). However, most other outcomes related to maternal morbidity, fetal and neonatal 560 mortality and morbidity, and adverse birth outcomes were similar across symptomatic and 561 asymptomatic groups, with a few exceptions. Pregnant women with symptomatic COVID-19 were 562 more likely to have an intrapartum cesarean delivery (RR 1.25, 95% CI: 1.05, 1.48) compared to 563 those with asymptomatic infection (Table S17). 564 565 We also found increased risk of preterm and moderate preterm birth among symptomatic pregnant 566 women (RR 1.30, 95% CI: 1.06, 1.60, and RR 1.65, 95% CI: 1.00, 2.73, respectively). However, 567 when we restricted to only pregnant women with infection onset prior to 37 weeks' gestation for 568 preterm birth and prior to 34 weeks' gestation for moderate preterm birth, we found asymptomatic 569 pregnant women had an increased risk of preterm and moderate preterm birth (RR 0.71, 95% CI: 570 0.52, 0.97, and RR 0.57, 95% CI: 0.41, 0.81), compared to symptomatic pregnant women. 571 572

Comment
As in the general population, we found that pregnant women with comorbidities including 575 diabetes, hypertension, cardiovascular disease and obesity were at increased risk for severe 576 COVID-19-related outcomes, as well as maternal morbidities, and adverse birth outcomes, 577 compared to pregnant women without these comorbidities. Given pooled global data, we also 578 identified several less commonly-known risk factors for pregnant women with COVID-19, 579 including HIV coinfection, being underweight at the start of pregnancy, and anemia at the time of However, in a sensitivity analysis restricted only to participants infected prior to 37 weeks 637 gestational age, we found that asymptomatic pregnant women are more likely than symptomatic 638 pregnant women to have a preterm birth. These seemingly conflicting results may be related to 639 features of study sampling; for example, this difference may be due to the large percentage of 640 asymptomatic participants who are identified during screening at labor and delivery. Across the 641 were identified at or after 37 weeks gestational age, compared to 26% of babies born to 643 symptomatic participants. 644 645 Strengths and limitations. IPD meta-analyses are considered the gold-standard method for 646 generating aggregate estimates. Here, we standardized data quality assessment and harmonized 647 definitions of risk factors and outcomes. This is especially valuable for outcomes such as stillbirth, 648 preterm birth, and perinatal mortality, which have varying definitions globally. We included data 649 from 33 countries and territories, including many low-and middle-income countries, whereas the 650 bulk of the published literature on COVID-19 in pregnancy comes from middle-or high-income 651 countries. Therefore, by pooling global data we were able to investigate risk factors such as HIV 652 status, undernutrition, and anemia, which are more common in low-income countries, but for 653 which individual studies may not have adequate power to draw meaningful conclusions. We were 654 also able to identify risks linked to rare outcomes such as pregnancy-related death and stillbirth. it is not possible to draw inferences about the absolute risk of adverse outcomes. The heterogeneity 665 in baseline rates of adverse outcomes globally further complicates interpretation of the absolute 666 risks. However, the relative risks comparing those with and without the risk factors of interest 667 generally appear consistent between sites and heterogeneity is relatively low for pooled estimates. 668 Additionally, although this analysis pooled a large, global sample of pregnant and postpartum 669 women with COVID-19, half of the overall sample for critical care outcomes (ICU admission, 670 ventilation, any critical care, pneumonia, and mortality) was derived from the Mexican National 671 Registry, which collected no information on maternal morbidity, birth or neonatal outcomes. This 672 analysis also did not examine risk factors related to social determinants of health, which may 673 exacerbate the biological risk factors identified in this analysis. 674

675
We identified risk factors for adverse maternal morbidities, fetal, and neonatal outcomes among 676 pregnant women with COVID-19, and these are generally consistent with risk factors for adverse 677 pregnancy outcomes including pre-existing diabetes or hypertension [36][37][38]  in the general non-pregnant population. Nonetheless, this study provides high-quality evidence 684 that pregnant women with these risk factors are also at risk for adverse outcomes from COVID-19 685 illness.
Although pregnant women are already considered a high-risk population by the WHO and should 689 be given equitable access to safe and effective preventives and therapeutics, special priority should 690 be given to pregnant women with additional risk factors, including chronic and infectious 691 comorbidities, nutritional status, and maternal age. This data strongly supports the need for access 692 to vaccines and treatments for SARS-CoV-2 infection for pregnant women, prioritizing those with 693 risk factors for severe illness and adverse birth outcomes. 694 695 J o u r n a l P r e -p r o o f 3 The Cancovid-Preg study follows a cohort of pregnant women with SARS-CoV-2 infection and their infants; because the study was ongoing at the time of data submission, risk factor data availability and sample size is slightly different for maternal COVID-19 severity outcomes and neonatal/birth outcomes. We present the data as two subsets of the same cohort for this ongoing study. In the "Maternal Subset", we present data on pregnant women with COVID-19, including outcomes on ICU admission, ventilation, and critical care (n=2,045). In the "Infant Subset", we present data on live births to pregnant women with COVID-19, including outcomes on preterm birth (n=2,626). 4 Data from Cancovid-Preg represents all provinces, with missing data randomly distributed across provinces except for the risk factor "pre-existing hypertension", which is unavailable for the full cohort from Ontario. 5 Note that for the UKOSS study, 100% of patients are hospitalized. However, the reason for hospitalization may not be COVID-19 and some participants presented at the hospital for an unrelated reason and were found to have an incidental COVID-19 infection. from all participating studies with at least 1 adverse event for the given outcome using a DerSimonian-Laird random effects model meta-analysis. For any study with zero events in one arm (Risk Group or Reference Group), we used a continuity correction of the inverse of the number of events in the oppposite group within the same study.   The PRISMA flow diagram outlines the identification and recruitment of studies and final inclusion of individual patient data for this study.

Figure 2. Incidence of outcomes by study
This figure presents the incidence and 95% confidence intervals of selected adverse outcomes across the 21 participating studies, including: A) ICU admission, B) ventilation, C) pregnancy-related death, D) preeclampsia, E) cesarean delivery, F) stillbirth, G) neonatal death, H) low birthweight, and I) preterm birth. Studies are grouped by World Bank income group levels: lower-middle income countries are shown in red; upper-middle income countries are shown in green; those from high income countries are shown in blue. Two studies (shown in purple) are multi-country studies that contain countries from multiple income groups. The complete list of countries for each of these multi-country studies is presented in Table 1.