Timing and severity of COVID-19 during pregnancy and risk of preterm birth in the International Registry of Coronavirus Exposure in Pregnancy

Since its emergence, coronavirus disease 2019 (COVID-19) has proven uniquely harmful to certain populations [1, 2]. However, its effects on the pregnant population have been less easily discerned. Early studies suggested an elevated risk of preterm birth among pregnant people with COVID-19 [3‐7] but were limited by small samples from single hospitals, little variability of disease severity or timing of infection during pregnancy, exclusion of ongoing pregnancies from estimates, and lack of valid comparison groups. More recent studies have reported a small but consistent increase in preterm delivery among people with COVID-19 during pregnancy [8, 9] or at delivery [10‐12]. However, existing studies are limited in the extent to which they consider timing of infection; when gestational age at infection and start of follow-up are not aligned, “immortal time bias” may reduce, negate, or reverse any effect on prematurity [13]. In addition, although the daily rate of preterm delivery increases as week 37 approaches, the total risk of preterm delivery declines due to the shrinking window in which to deliver before term, making estimates of risk after infection at different gestational ages difficult to interpret. Finally, while associations between severe COVID-19 and preterm delivery may reflect biological effects of infection or the immunological response, they may also result from medically indicated delivery based on health concerns [5, 8, 11, 14]. No study has considered the timing of infection during pregnancy, the severity of disease, the indication of prematurity, and the methodological issues simultaneously.

Using data from a large, international pregnancy cohort, we investigated whether COVID-19 increased the risk of preterm birth. We used multiple analytic approaches to disentangle the role of severe disease and the timing of infection during pregnancy, allowing us to estimate gestational-age-specific risks and the respective roles of spontaneous and medically indicated preterm delivery after mild, moderate, and severe COVID-19 throughout pregnancy.

In a large, diverse pregnancy cohort, we found that severe COVID-19 late in pregnancy may double or triple the probability of preterm delivery but that increased risk due to milder disease, or earlier in pregnancy, is likely minimal. Much of the effect of severe COVID-19, and much or all of the smaller increase with mild and moderate disease, appears due to indicated preterm deliveries. Nonetheless, severe COVID-19-affected pregnancies also reported more spontaneous preterm labour or rupture of membranes, suggesting an effect on spontaneous preterm delivery.

Intrauterine bacterial infection is a major cause of spontaneous preterm delivery [20], mediated through the innate immune response [21], and concern about the effects of SARS-CoV-2 infection during pregnancy is justified by evidence from viral infections [22, 23]. Influenza infection increases risk of preterm birth [24‐27], and SARS and MERS outbreaks generated evidence of harms of coronavirus infection during pregnancy, with reports of preterm birth, intrauterine growth restriction, and mortality [28‐30].

A number of meta-analyses have pooled risk of preterm birth after COVID-19 from case reports, case series, and other early studies, though estimates vary widely [31‐38]. Apparent differences in risk result from selection of hospitalized patients, exclusion of ongoing pregnancies, and inclusion of pregnancies past 37 weeks’ gestation at symptom onset. Studies that include infections after 37 weeks underestimate the risk of preterm delivery, and those restricted to hospitalized patients could overestimate it or result in selection bias if factors affecting hospitalization and preterm risk are not accounted for. Our study improves upon previous estimates of risk by providing gestational-age specific risks among pregnant people with both severe and mild/moderate disease using data from ongoing (or lost to follow-up) as well as completed pregnancies.

Other studies, including some based on surveillance data, have also provided valuable comparisons with concurrent [8‐12, 39‐43] or historical [44‐46] COVID-19-negative pregnancies, or across the spectrum of disease severity [14, 47‐52]. Most [8, 10‐12, 14, 43, 44, 47‐52] but not all studies [40‐42] have provided evidence that any vs. no infection, or more vs. less serious disease, are associated with higher preterm risk, particularly with respect to indicated deliveries. However, the extent to which they have adjusted for confounders or avoided other sources of bias differs. For example, use of historical references is questionable given the pandemic’s broad effects [53]. Moreover, studies that define exposure to COVID-19 as “any time during pregnancy” will be biased by design, given the shorter opportunity for infection in preterm deliveries, and also obscure gestational-age-specific risks.

We used three designs, each addressing shortcomings of the others, to investigate whether COVID-19 affects risk of preterm birth. We addressed confounding with multivariable adjustment as well as with within-person comparisons. In addition, in our main time-to delivery analysis, we accounted for several timing-related issues: longer pregnancies are more likely to have been exposed to SARS-CoV-2, risk of preterm delivery depends on gestational age at time zero (infection or reference), and preterm delivery cannot be assessed among participants whose pregnancies are ongoing and under 37 weeks.

Nevertheless, our study has important limitations. Information on tests and gestational age at delivery was self-reported. However, mothers who suffered COVID-19 during pregnancy are likely to remember when it occurred, and all their estimated due date and date of delivery. In addition, we have limited clinical measures compared to studies based on medical records or direct observation, which may have resulted in misclassification of spontaneous vs. indicated preterm delivery or of severity of COVID-19 cases. Furthermore, indications for delivery in people with symptomatic COVID-19 likely differed geographically and over time, limiting our ability to assess effects on spontaneous delivery. We used objective and standard measures of COVID-19 severity (e.g., ICU, ventilation, ECMO) to maximize specificity, at the possible cost of sensitivity (e.g., hospitalizations may have been precautionary due to pregnancy). When comparing to mild/moderate disease, this misclassification would tend to bias toward the null; therefore, we may have underestimated associations with severe COVID-19. While we found no differences between risks after moderate and more mild disease, it is possible that this was due to imprecision in self-reported symptoms, and future research should investigate whether there are any signs or symptoms that confer additional risk. We did not include asymptomatic participants in our main analyses so were not able to assess whether they were at higher risk than uninfected pregnancies; however, previous studies have found risk to be elevated primarily after symptomatic disease [38]. In addition, some participants who tested negative may have later developed an asymptomatic infection that was not detected or reported on a survey, so may be misclassified. Finally, participants in this study were not vaccinated, so we were not able to investigate the effect of COVID-19 vaccination.

Furthermore, while we do not have outcomes on some participants with ongoing pregnancies, many others were lost to follow-up. We sent reminders via text message to participants to encourage complete follow-up, but we were not able to increase retention [15]. In our week-specific analyses we were able to use data until the last known week of continued pregnancy under the assumption that loss to follow-up was conditionally independent of preterm birth, given the covariates measured on the initial surveys. Predictors of selection would have to be strongly associated with preterm birth to explain away the increased risk we found with severe COVID-19 [54]. In addition, although we had participants worldwide, some countries were represented more than others, and within-country sampling was not random. Because of this design, our study is not representative of any particular well-defined population, and high-risk pregnancies that are more likely to be tested may be overrepresented. While it is more widely representative than previous studies, there are also undoubtedly people who aren’t represented, in particular, those without sufficient internet access. Although the intensity of the pandemic may differ geographically, biologic effects of COVID-19 may be less likely to vary across the population. However, if the effect is mediated through precautionary early C-sections to avoid transmission or maternal complications during labour, then populations unable or hesitant to conduct these would not see an increased risk of preterm delivery; instead, risk of other periconceptional morbidities may rise.

In conclusion, this study suggests that with respect to preterm birth, prevention of COVID-19 is especially important in the second half of pregnancy. However, while the risk of preterm birth appears to double or triple after severe COVID-19, the population attributable risk is likely modest, as severe COVID-19 is rare among young people, though may be more common during pregnancy [31]. Our study was not designed to estimate the risk of COVID-19 or of severe disease in pregnancy, but it is clear that protective measures should be taken to avoid SARS-CoV-2 infection and symptoms should be closely monitored to avoid progression. Much of the increased risk of prematurity due to severe disease is likely iatrogenic, due to urgent delivery in response to maternal or foetal decline. Vaccines lower the risk of infection and severe disease, and improved treatments for COVID-19 could lower risk of progression and thus prematurity by reducing indications for delivery.