key: cord-0767670-z7zinuk1
authors: Hernandez-Diaz, Sonia; Smith, Louisa H; Dollinger, Camille; Rasmussen, Sonja A; Schisterman, Enrique F; Bellocco, Rino; Wyszynski, Diego F
title: International Registry of Coronavirus Exposure in Pregnancy (IRCEP) – Cohort Description and Methodological Considerations
date: 2022-03-08
journal: Am J Epidemiol
DOI: 10.1093/aje/kwac046
sha: 63c6b4a9a5cddc5964fa4756c1a050877d7054dc
doc_id: 767670
cord_uid: z7zinuk1

Limited data are available about the potential health effects of infection with the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) on pregnant women and their developing offspring. We developed the International Registry of Coronavirus Exposure in Pregnancy (IRCEP) to provide data on the risk of major adverse obstetric and neonatal outcomes among women with varying degrees of severity and timing of COVID-19 exposure during pregnancy. We describe here the cohort and share the lessons learned. The IRCEP enrolls women tested for SARS-CoV-2 or with a clinical diagnosis of COVID-19 during pregnancy and obtains information using an online data collection system. By March 2021, 17,532 participants from 77 countries had enrolled; 54% enrolled during pregnancy and 46% afterwards. Among women with symptomatic COVID-19 with a positive SARS-CoV-2 test (N=4,934), symptoms were mild in 41%, moderate in 52% and severe in 7%; 7.7% were hospitalized for COVID-19 and 1.7% were admitted to an intensive care unit. The biggest challenges were retention of participants enrolled during pregnancy, and the potential bias introduced when participants enroll after pregnancy outcomes are known. Multiple biases need to be considered and addressed when estimating and interpreting the effects of COVID-19 in pregnancy in these types of cohorts.

were admitted to an intensive care unit. The biggest challenges were retention of participants enrolled during pregnancy, and the potential bias introduced when participants enroll after pregnancy outcomes are known. Multiple biases need to be considered and addressed when estimating and interpreting the effects of COVID-19 in pregnancy in these types of cohorts.

Over one hundred million women have given birth worldwide since the Coronavirus Disease 2019 (COVID-19) pandemic started in China in late 2019 1,2 . To date there is no consistent evidence that pregnant women are more susceptible to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection [3] [4] [5] [6] ; and conflicting findings have been published on whether they have more severe COVID-19, if infected, than non-pregnant women of similar age 3, 5, 6 . In early reports, based on small samples, the risk of both symptomatic infections and death in pregnant women with COVID-19 was similar to that in non-pregnant women 7, 8 .

Subsequently, larger studies suggested that hospitalization rates, intensive care unit (ICU) admissions and mortality in pregnant women diagnosed with COVID-19 are higher than in nonpregnant women of similar age [8] [9] [10] [11] [12] . More recent studies concluded that pregnant women are not more likely to get seriously ill from SARS-CoV-2 infection. 13 These discrepancies may be at least partially explained by international and temporal differences in SARS-CoV-2 virulence, population vulnerability to its effects, or COVID-19 treatments.

Regardless, even if infection and severity risks were not above general population levels, consequences of severe COVID-19 are likely exacerbated when the patient is pregnant and has to deliver and take care of a newborn 14 . Moreover, severe COVID-19 may not only cause maternal morbidity and mortality, but also lead to iatrogenic preterm delivery due to concerns about COVID-19 transmission and progression, 3, 6, [15] [16] [17] [18] and potential coagulation disorders 19 that may cause placental-related complications such as miscarriages. 20 Vertical SARS-CoV-2 transmission is rare but can occur. 3, 5, 6, [21] [22] [23] A dearth of information, especially during the first months of the COVID-19 pandemic, led to increased levels of anxiety among pregnant women 24 , as well as unnecessary cesarean deliveries 3 and elective terminations. 25 Therefore, there was, and still is, an urgent need to gather and women worldwide and to assess the relative risk of major adverse obstetric and neonatal outcomes in pregnancies exposed to varying degrees of severity and timing of SARS-CoV-2 infection. We share here the lessons we learned from designing and conducting IRCEP.

The IRCEP is an ongoing observational cohort that began enrollment in June 2020. Data on baseline characteristics, reproductive history, chronic medical conditions and medication use, exposure to SARS-CoV-2, COVID-19 manifestations, prenatal care, pregnancy outcomes, and neonatal outcomes are collected during pregnancy and postpartum. The IRCEP allows enrollment throughout pregnancy and during the first 180 days after the end of pregnancy.

The study population includes women 18 years of age or older from around the world with a current or recent pregnancy and tested for SARS-CoV-2 (regardless of the result) or with a clinical diagnosis of COVID-19 between the first day of the last menstrual period (LMP) and end of pregnancy. Inclusion criteria also require willingness to provide responses to a minimum set of demographic questions. The rationale for including cases without a positive SARS-CoV-2 test is that, in many countries, early in the pandemic, the reverse transcriptase-polymerase chain reaction (RT-PCR) test was inaccessible to a large proportion of the population. Therefore, the presence of clinical symptoms (e.g., typical pulmonary lesions on chest CT, loss of sense of smell or taste), confirmed by a health care professional, was sufficient to be considered a In addition to self-reported information, participants are asked to photograph and upload all available SAS-CoV-2 test results, delivery and neonatal medical records, and any other health care records they consider relevant, after redacting personal identifiers. These records could be used to validate maternally reported diagnoses and to allow for potential adjudication of outcomes by experts blinded to the maternal COVID-19 status for specific studies. Uploaded records are checked for redaction prior to storage and will be translated with professional software followed by human review as they became relevant for analyses (e.g., when evaluating malformations).

The identifying and anonymized data of individual study participants are linked via a unique subject identifier. The file containing the personal identifiers used during the informed consent process is securely stored in a separate server. Files for analysis do not include any personal identifiers. Once the participant completes the study, the key used to link analytic files with identifiers is deleted. obstetric and neonatal events in women from the same source population as the exposed, upon adjustment for potential confounders associated with either infection or testing. Women in the reference group that test positive later in pregnancy would be considered exposed at that point.

Sub-cohorts defined by COVID-19 severity are identified. The classification of severity is summarized in Web table 1. To evaluate the effect of disease severity on adverse pregnancy outcomes among women exposed to the virus, severe cases will be compared to those with mild COVID-19 presentations. The IRCEP will also be able to assess the potential effects of specific COVID-19 characteristics or treatments, predictors of COVID-19 severity, and frequency of infections in newborns.

The first day of the LMP is used to define the timing of pregnancy and the length of gestation.

Gestational age is determined by an algorithm using the best available information, including reported LMP, due date based on LMP, and due date based on ultrasound. Gestational timing is needed to define exposure since the etiologically relevant period for each outcome of interest varies. Web Table 2 .

The main outcomes of interest include SAB (spontaneous pregnancy loss prior to 20 weeks gestation), stillbirth (fetal death after 20 weeks gestation), major congenital anomalies (specific types will be explored where sample size permits), preterm delivery, maternal obstetric complications (e.g., preeclampsia, gestational hypertension, cesarean delivery, postpartum hemorrhage, postpartum depression), small for gestational age (≤10th centile on birth weight for the infant's sex and gestational age), head circumference at birth, admission to the neonatal intensive care unit, vertical transmission of SARS-CoV-2, and neonatal death. Questions on access to care during the COVID-19 pandemic, maternal mental health, and breastfeeding will allow evaluation of other aspects of wellbeing.

Primary analyses will be restricted to participants that enrolled before specific outcomes occur or can be known, to avoid both selection and recall biases 26 . For example, in the case of congenital anomalies, before any informative prenatal test. The exposure windows of interest will also

depend on the outcome; for example, to evaluate whether SARS-CoV-2 infection affects the risk for preterm delivery, those exposed in the month before a preterm delivery may be compared to the unexposed in that risk set defined by gestational age.

The IRCEP collects data on a wide range of covariates including maternal demographic characteristics, comorbid medical conditions, habits, reproductive history, obstetric characteristics, use of medications, and measures of healthcare utilization. Women may enroll after the onset of COVID-19 infection but baseline questions collect information on characteristics existing before the infection to allow proper adjusted in future analysis 27 .

Participating women may withdraw from the IRCEP at any time at their own request, at which time all their information will be deleted from the database (except for select de-identified sociodemographic characteristics, which are retained to assess data representativeness in aggregate analyses relative to the initial population and correct for informative censoring if needed). The website inquires about the reason for withdrawal. When a participant does not respond to repeated online prompts requesting further information on the pregnancy, they will be considered lost to follow-up. Participants that do not complete all the modules by 180 days after the end of pregnancy will be censored from the cohort at the time of last contact.

The IRCEP protocol was approved by the Harvard Longwood Campus Institutional Review Board (IRB20-0622). Eligible participants provide consent electronically with an e-Form on the IRCEP website.

As of end of March 2021, 17,532 participants from 77 countries had enrolled in IRCEP. Of those, 54% enrolled during pregnancy and 46% after the end of pregnancy, i.e., postpartum or

after a pregnancy loss (Table 1 ). The frequency of participants by country of residence resembles the distribution of COVID-19 worldwide at that time, consistent with the IRCEP awareness campaigns online. Participants in the IRCEP are racially and socioeconomically diverse ( Table   2 ). The frequency of both negative and positive tests in asymptomatic women reflect screening intensity in those groups (e.g., more in those living in North America or Europe, white race, having asthma, or smoking tobacco). The median number of weeks post-LMP at enrollment was 26 (interquartile range [IQR] 17, 34) for those enrolled during pregnancy and 49 (IQR 44, 54) for those enrolled after the end of pregnancy (Figure 1 ).

At enrollment, 5,858 (34%) participants reported a positive test for SARS-CoV-2 (84% were symptomatic), 2.5% had a clinical diagnosis of COVID-19 without a positive test, and 10,215

(58%) had a negative test and no clinical diagnosis of COVID-19 (Table 3) . A small number of participants reported neither clinical diagnoses nor a SARS-CoV-2 test (n=20) and were excluded.

The timing of COVID-19 diagnosis was equally distributed throughout pregnancy (Web Figure   2 ). However, testing tended to cluster around the time of delivery, since it has become standard practice in many countries to screen for SARS-CoV-2 at the time of hospital admission for delivery. Increased screening of asymptomatic women pre-delivery resulted in more negative tests, and positive tests in women without symptoms, towards the end of pregnancy. 

than analgesics, the most common pharmacotherapies used to treat COVID-19 were azithromycin, oseltamivir, corticosteroids, and hydroxychloroquine/chloroquine.

Of the 9,471 participants enrolled during pregnancy, 5% have completed participation in the study and 6% are still pregnant and completing modules, as of March 31, 2021. Of the 8,061

participants enrolled postpartum or after fetal loss, 74% have completed participation (Figure 2 ).

Among those participants with complete follow-up, the frequencies of the most common obstetric outcomes were similar to what would be expected in the general population (i.e., 0.8% stillbirths, 6% preeclampsia, 1.9% twins or higher-level multiples, and 2.6% major congenital malformations) ( Table 4 .) The frequency of pregnancy losses is lower than the usual cumulative risk throughout pregnancy in the population since women that enroll late in pregnancy represent a survivor cohort and those that enroll retrospectively had shorter opportunities to have a test or an infection.

We have enrolled a large international cohort of pregnant women with and without COVID-19 in the IRCEP. Future studies will be able to assess a variety of specific research questions related to COVID-19 and pregnancy, such as the effect of COVID-19 severity on multiple outcomes. We learned many lessons designing and conducting the IRCEP, including the following:

The increasing number of pregnancies affected by COVID-19 and the wider availability of tests for SARS-CoV-2 facilitated enrollment over the study period. Nonetheless, we learned that social media awareness campaigns are key to enrolling participants in internet-based studies.

Enrollment in the IRCEP dropped substantially between campaigns, despite a large amount of pregnancy and COVID-19-related resources available on the Pregistry website. Web Figure 3 .

While completion of modules among retrospective enrollees has been over 70%, retention of participants registered during pregnancy is below 10%. The IRCEP tries to increase retention and data completeness by fostering an online community where participants not only share data but also receive relevant information. Optional automatic reminders are also sent via SMS to remind

participants to upload information. We did not offer economic incentives to participate in the study to minimize the risk of ineligible subjects enrolling for money. The study appeals to the altruism and solidarity of volunteers by conveying the importance of their loyalty to the study for the generation of evidence that will help other pregnant women. Unfortunately, these approaches were insufficient. Similarly, a small proportion of participants has submitted redacted photos of their medical records despite an easy system to upload documents. Consequently, our intention to validate and adjudicate self-reported outcomes may not be feasible.

Despite the attrition, this remains one of the largest cohorts of pregnant women with COVID-19.

However, the substantial losses to follow up might select a biased sample. Characteristics of participants lost to follow-up will be compared to the observed cohort and weights may be applied if censoring is not random. For example, among prospective enrollees, retention was higher among those negative for COVID-19, from North America or Europe, and of White race, higher education, and higher income (Web Table 3 ). Future studies should strive to attain higher retentions. For example, by allocating sufficient budget to a more personalized and proactive contact with enrollees or demanding stronger commitment to research from the beginning (e.g., making uploading test results an inclusion criteria).

While the multinational design is meant to facilitate generalizability of results across the globe, it also represents a challenge a) logistically, because of the need to translate the materials to many languages and to make them culturally appropriate across countries (e.g., race categories, health coverage modalities); and b) methodologically, because country of residence is a strong determinant of both COVID-19 infection and of testing, and of incidence and diagnosis of the outcomes. Therefore, the statistical analyses will need to take country of residence into consideration, for example with stratification or introducing a random effect component. In retrospect, a multinational study in fewer selected countries might have been preferable from a research perspective to guarantee sufficient numbers within strata while still providing a global perspective.

In addition, enrollment of women in pregnancy registries is voluntary and, therefore, participants are a non-random sample of all women with COVID-19. Consequently, the characteristics and experience of women who participate in a registry may differ from those of non-participants, and

these characteristics may modify the observed effects of SARS-CoV-2. Our primary awareness campaigns were designed to appeal to people of a wide variety of backgrounds, but our enrolled sample turned out to be more educated than the general population (65% with at least a college education). Although biological effects of viruses tend to be universal, the health consequences of COVID-19 among those volunteering to participate in the registry, who tend to be more educated and health conscious, is likely to underestimate the absolute impact on more vulnerable populations. Consequently, the participants who joined after delivery included more negative tests, because the prospective participants enrolled based on tests conducted before screening at delivery.

Therefore, study designs like ours are not appropriate to estimate the incidence of SARS-CoV-2 or COVID-19.

Non-random samples selected based on testing can provide valid estimates of COVID-19 effects on pregnancy outcomes if risk factors associated not only with the infection but also with testing are controlled. Asymptomatic participants with test results for SARS-CoV-2 (either positive or negative) represent populations with increased access to screening (e.g., more affluent) or highrisk groups (e.g., women with asthma). Therefore, the unbalanced characteristics observed between participants with COVID-19 and the reference group with negative tests may be risk factors for infection, or risk factors for testing. This selection introduced by the inclusion criteria (i.e., requiring a test) can be conceptualized as conditioning on a collider. 28 Although, this

potential selection bias is less likely to affect the assessment of COVID-19 severity within symptomatic cases, if milder cases with risk factors were still preferentially tested, we would underestimate the effect of severity. While we will adjust for factors associated with testing, residual confounding remains a concern. Web Figure 6.

Some studies that evaluated the effect of pregnancy on people with SARS-CoV-2 infection suggested a higher frequency of ICU admissions and hospitalizations among pregnant women with COVID-19, even after accounting for age and healthier overall status of pregnant women. 8, 12 However, results might be explained by preferential hospitalization and ICU admissions of pregnant women given the same disease severity and by inclusion of hospitalizations for pregnancy-related reasons in the outcome (Web figure 5 ). That is, even without COVID-19, pregnancy increases the likelihood of hospitalizations relative to nonpregnant women of similar age, even if only for obstetric reasons (e.g., delivery, preeclampsia).

We did not include non-pregnant women in our study and, therefore, the IRCEP cannot answer the question of whether pregnant women constitute a "vulnerable population" with respect to 

When assessing the effect of SARS-CoV-2 infection on pregnancy outcomes in observational studies, an association could be explained by a direct effect of the virus on the outcomes, an effect mediated through maternal symptoms (e.g., pneumonia or fever), or by confounding (e.g., women more likely to be infected might also be at higher risk of adverse pregnancy outcomes). It is also important to note that a factor may cause severe COVID-19 by increasing exposure to SARS-CoV-2, increasing susceptibility to being infected if exposed, or increasing the likelihood of progression to severe COVID-19 if infected. Studies that condition on the steps in the causal pathway (e.g., studying hospitalized participants) may be conditioning on colliders. 29 For example, when evaluating the effect of COVID-19 on preterm delivery and using a reference group recruited in the same center, within pregnancies hospitalized, those not admitted for COVID-19 would have other reasons for admissions (e.g., preeclampsia) that may be risk factors

for preterm birth. Web Figure 6 . Similarly, when testing is required to classify whether the person has COVID-19, if the outcome triggers testing, there would be a higher proportion of confirmed COVID-19 (and negative tests) among those with the outcome. 28 Future statistical analyses using the IRCEP data will consider these explanations for observed associations and carefully classify confounders, mediators, and colliders depending on the research question.

Of concern in cohort studies with primary data collection is the selection of non-lethal COVID-19 since most women will enroll after COVID-19 resolves. Although maternal mortality is expected to be low (<1%), this selection will result in optimistic descriptions of the nature of COVID-19 during pregnancy. Studies with population-based samples enrolled before infection (e.g., healthcare databases) will be able to provide the full picture.

Enrollment after pregnancy outcomes are known (during or after the end of pregnancy) may selfselect a group with adverse outcomes and more eager to share their experience in a study, thus overestimating risks; or it might underestimate the risk if the distressing event reduces the likelihood of participation. 30 Overall, when we compared participants that enrolled after vs before the end of pregnancy, the frequency of pregnancy outcomes was slightly higher for some (e.g., preterm delivery 10% vs 9%) and slightly lower for others (e.g., major malformations 2.6% vs 2.7%). Table 4 . If participation after an adverse outcome is diagnosed is more (or less) likely for patients with COVID-19, retrospective participation may lead to spurious associations.

Therefore, primary analyses will be restricted to women enrolled before the pregnancy outcome of interest is known. For example, analyses of malformations will be restricted to women that enroll before the results from informative prenatal screening tests are known.

However, events that occur right after COVID-19 may only enroll retrospectively. remained hospitalized through the end of pregnancy or had a natural or induced delivery and could only enroll after delivery. The above two scenarios challenged our plans to focus on prospective participants when evaluating the effects of infections near end of pregnancy. Studies that enroll the population before the infection (e.g., healthcare databases) will avoid the potential bias introduced by retrospective enrollment.

The IRCEP collects information directly from participants. Women often know more about their habits, occupations, and compliance with medication use than their health care providers;

however, clinicians might provide more complete and accurate information regarding diagnoses. 31 The accuracy of recall in the IRCEP is facilitated by using structured questionnaires, detailed questions that allow only plausible responses, and calendars to help establish gestational timing and enhance recall of dates. 32 To reduce misclassification of infection, primary analyses can be restricted to COVID-19 confirmed with laboratory testing, which became more available over time. Web Figure 7 . Misclassification of COVID-19 severity (e.g., need for respiratory assistance or ICU) is unlikely. Outcome misclassification could be non-differential or differential between COVID-19 cases and the reference group. Concern that COVID-19 might pose a risk could lead to more prenatal diagnostic measures such as ultrasound and to more careful examination of infants for defects postnatally, potentially leading to differential accuracy in detection and classification of defects among exposed and unexposed. This potential surveillance bias can be minimized by focusing on major outcomes that are less vulnerable to differential misclassification (e.g., prematurity, SABs, malformations).

Many publications on COVID-19 during pregnancy that shared our limitations failed to address important sources of biases. We summarized the lessons we learned so that future studies can do better. 

Pregnant mothers and babies born during COVID-19 pandemic threatened by strained health systems and disruptions in services

19-pandemic-threatened-strained-health. Accessed August, 2020. 2. WHO. Pneumonia of unknown cause -China

Clinical characteristics and intrauterine vertical transmission potential of COVID-19 infection in nine pregnant women: a retrospective review of medical records

Pregnancy and Perinatal Outcomes of Women With Coronavirus Disease (COVID-19) Pneumonia: A Preliminary Analysis

COVID-19 infection among asymptomatic and symptomatic pregnant women: Two weeks of confirmed presentations to an affiliated pair of New York City hospitals

Clinical Characteristics of Pregnant Women with Covid-19 in Wuhan, China. The New England journal of medicine

Mortality of a pregnant patient diagnosed with COVID-19: A case report with clinical, radiological, and histopathological findings

Characteristics of Women of Reproductive Age with Laboratory-Confirmed SARS-CoV-2 Infection by Pregnancy Status -United States

Public Health Agency of Sweden's Brief Report: Pregnant and postpartum women with severe acute respiratory syndrome coronavirus 2 infection in intensive care in Sweden

Characteristics and Maternal and Birth Outcomes of Hospitalized Pregnant Women with Laboratory-Confirmed COVID-19 -COVID-NET, 13 States

SARS-CoV-2 Infection Among Hospitalized Pregnant Women: Reasons for Admission and Pregnancy Characteristics -Eight U.S. Health Care Centers

Update: Characteristics of Symptomatic Women of Reproductive Age with Laboratory-Confirmed SARS-CoV-2 Infection by Pregnancy Status -United States

Characteristics and outcomes of pregnant women admitted to hospital with confirmed SARS-CoV-2 infection in UK: national population based cohort study

Preparing for influenza after 2009 H1N1: special considerations for pregnant women and newborns

What are the risks of COVID-19 infection in pregnant women?

Clinical manifestations and outcome of SARS-CoV-2 infection during pregnancy

Clinical analysis of 10 neonates born to mothers with 2019-nCoV pneumonia

Coronavirus disease 2019 and first-trimester spontaneous abortion: a case-control study of 225 pregnant patients

Coagulation abnormalities and thrombosis in patients with COVID-19

COVID-19 and acute coagulopathy in pregnancy

Perinatal Transmission of 2019 Coronavirus Disease-Associated Severe Acute Respiratory Syndrome Coronavirus 2: Should We Worry? Clinical infectious diseases : an official publication of the Infectious Diseases Society of America

Possible Vertical Transmission of SARS-CoV-2 From an Infected Mother to Her Newborn

Antibodies in Infants Born to Mothers With COVID-19 Pneumonia

Does having a high-risk pregnancy influence anxiety level during the COVID-19 pandemic? European journal of obstetrics, gynecology, and reproductive biology

Is termination of early pregnancy indicated in women with COVID-19? European journal of obstetrics, gynecology, and reproductive biology

Assessment of recording bias in pregnancy studies using health care databases: An application to neurologic conditions. Paediatric and perinatal epidemiology

Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology

Association Between Congenital Cytomegalovirus and the Prevalence at Birth of Microcephaly in the United States

The birth weight "paradox" uncovered?

Effects of gestational age at enrollment in pregnancy exposure registries