Abstract
Objective:
To evaluate whether the risk of Long COVID among individuals infected with SARS-CoV-2 during pregnancy differs from females who were not pregnant at time of virus acquisition.
Methods:
We conducted a multicenter observational cohort study at 79 National Institutes of Health (NIH) Researching COVID to Enhance Recovery (RECOVER) sites. Females aged 18–45 years with an index (first) SARS-CoV-2 infection on or after December 1, 2021 were included. The exposure was pregnancy (any gestational age) at time of index SARS-CoV-2 infection. The primary outcome was Long COVID 6 months after index infection defined as RECOVER-Adult Long COVID Research Index ≥ 11 based on a detailed symptom survey. To account for confounding and differential selection between participants who were pregnant and non-pregnant at infection, propensity score matching methods were used to balance the groups on variables potentially associated with both pregnancy status and Long COVID
Results:
Overall 2423 participants were included; 580 (23.9%) were pregnant at index SARS-CoV-2 infection. The median age at infection was 33 years (IQR: 28–38), and 2,131 (90.0%) of participants with known vaccination status were vaccinated. After propensity score matching, the adjusted Long COVID prevalence estimates 6 months after index infection were 10.2% (95% CI: 6.2%, 14.3%) among those pregnant at infection, and 10.6% (95% CI: 8.8%, 12.4%) among those not pregnant at infection. Pregnancy was not associated with a difference in adjusted risk of Long COVID (adjusted risk ratio [RR]: 0.96, [95% CI: 0.63, 1.48]).
Conclusion:
Acquisition of SARS-CoV-2 during pregnancy was not associated with a differential risk of Long COVID at 6 months when compared with similar-age females who acquired SARS-CoV-2 outside of pregnancy.
Precis
Acquisition of SARS-CoV-2 during pregnancy was not associated with a differential risk of Long COVID at 6 months when compared with similar-age females who acquired SARS-CoV-2 outside of pregnancy.
Introduction
Long COVID, or post-acute sequelae of SARS-CoV-2, is broadly recognized as a debilitating illness that affects a substantial proportion of the population. Manifestations vary widely and can be observed as signs and symptoms affecting multiple organ systems.1 Long COVID is hypothesized to be mediated, in part, by the immunologic response and resultant inflammation in response to the initial SARS-CoV-2 infection.2 During pregnancy, the immune system shifts to be more tolerant and less inflammatory with an increase in complement activity and monocytes, and a lower number of T cells and B cells, particularly in the third trimester.3,4 Thus, it is biologically plausible that inherent differences in the immune response when SARS-CoV-2 is acquired during pregnancy as compared with SARS-CoV-2 acquired in the non-pregnant state could result in a differential prevalence of Long COVID. Similarly, hormonal changes, such as those occurring in pregnancy,4 may also influence the development or symptomatology of Long COVID. For example, female sex has been associated with higher risk for developing Long COVID, with differential risk depending on patient age and pre- or post-menopausal status.5, 6
Some published studies have suggested that the incidence of Long COVID following SARS-CoV-2 infection during pregnancy is lower than when SARS-CoV-2 is acquired outside of pregnancy.7,8 Data from the NIH RECOVER Electronic Health Record (EHR) cohort including 5,397 pregnant and 83,915 non-pregnant similar-age female patients found that acquisition of SARS-CoV-2 during pregnancy was associated with a lower incidence of Long COVID at 6 months post-infection (26% vs 34%, aHR 0.85, 95% CI 0.81–0.91).7 Studies without a control group of non-pregnant females demonstrate varying prevalence of Long COVID after SARS-CoV-2 infection during pregnancy ranging from 9% to 40%.8,9 Thus, it remains uncertain whether there is a differential risk for Long COVID when SARS-CoV-2 is acquired during pregnancy. A study with a direct comparison to non-pregnant adult females is needed.
In this study, we evaluate whether the risk of Long COVID among individuals infected with SARS-CoV-2 during pregnancy differs from that of females who were not pregnant at the time they acquired SARS-CoV-2 infection. We hypothesize that the risk of developing Long COVID among individuals infected with SARS-CoV-2 during pregnancy is lower than that of similar-age females who were not pregnant when infected with SARS-CoV-2. We further hypothesize that the frequencies of Long COVID symptoms among individuals infected with SARS-CoV-2 during pregnancy differ from symptoms among similar-age females who were not pregnant when infected with SARS-CoV-2.
Methods
RECOVER-Adult is a multicenter cohort study of adults age 18 years or older, with and without prior infection with SARS-CoV-2. RECOVER-Pregnancy enrolled participants into the RECOVER-Adult study who had a pregnancy during the study enrollment period either with or without SARS-CoV-2 infection. The design of these studies has been described previously.9,10 Eligible participants for inclusion in the present analysis were adults age 18–45 whose sex was assigned female at birth, and who had a SARS-CoV-2 infection with index date on or after December 1, 2021 (the start of RECOVER study enrollment and also the first emergence of the Omicron variant in the U.S.). The index date was defined as the date of the first infection (suspected, probable, or confirmed SARS-CoV-2 infection as defined by World Health Organization criteria). Participants who were uninfected at enrollment but had a positive antibody result at enrollment (nucleocapsid for any participant, or spike protein for those who were unvaccinated) were considered infected, with an index date set to 90 days before the positive test. Additionally, participants in RECOVER who were uninfected at enrollment but had an on-study first infection were included in this analysis as infected with index date based on this infection.
This analysis included participants who enrolled in RECOVER no more than 4.5 months after their index date. At enrollment, participants completed a comprehensive set of surveys including demographic information, social determinants of health (SDoH), SARS-CoV-2 vaccination status, and health history. Participants were followed prospectively with study visits in 3-month intervals, and at each study visit completed a survey recording whether they had a comprehensive set of symptoms and selected symptom severities.10,11
The exposure of interest was pregnancy at the time of index SARS-CoV-2 infection, ascertained via participant self-report. Pregnancies at any gestational age, and those resulting in livebirths, stillbirths and miscarriages were included. Participants with unknown pregnancy status at index infection date were excluded.
The primary outcome was presence of Long COVID at the 6-month study visit after index infection, defined using a published symptom-based algorithm called the RECOVER-Adult Long COVID Research Index (LCRI).12 An LCRI value (ranging from 0 through 30) was computed based on each participant’s symptoms, and participants with an LCRI ≥ 11 were considered as meeting the definition for Long COVID. The remaining participants were considered Long COVID-indeterminate (not meeting classification criteria for Long COVID but not necessarily asymptomatic).
Cohort characteristics were summarized by pregnancy status at index, using counts and relative frequencies for binary and categorical variables, and means with standard deviation (SD) for continuous variables. We considered a broad range of relevant covariates, including demographic and enrollment factors, vaccination status at index (fully vaccinated defined as having received the primary COVID-19 vaccination series, partially vaccinated, or unvaccinated), comorbidities present prior to index, and a set of composite SDoH measures. Individual-level SDoH measures included economic instability, education and language access barriers, health care access and quality challenges, and lack of social and community support, as detailed in Appendix 2, available online at http://links.lww.com/xxx. Community-level measures included percent of ZIP code living below the federal poverty level and percent living in crowded homes as defined in the American Community Survey.13,14
To account for confounding and differential selection between participants who were pregnant and non-pregnant at infection, propensity score matching methods were used to balance the groups on variables potentially associated with both pregnancy status and Long COVID. The propensity score was defined as the probability of being pregnant at the time of SARS-CoV-2 infection, estimated using logistic regression. We then generated propensity score weights using a technique called constrained full matching, which used all participants to create weighted exposure groups that were balanced on baseline covariates to mitigate confounding.15 Specifically, we weighted the exposure groups to both have the same distribution of covariates as the broader cohort of females age 18–45 in RECOVER, resulting in Long COVID prevalence estimates that target an overall ‘average treatment effect’ (ATE). We assessed overlap of the distributions of propensity scores in each group using histograms, and plotted standardized mean differences of each covariate before and after propensity score weighting to assess covariate balance.
Prevalence of Long COVID at 6 months after index infection was estimated by pregnancy status at infection, both before and after propensity score weighting. We estimated the corresponding propensity score-weighted relative risk of Long COVID using binomial regression with log link, and risk difference using binomial regression with an identity link. Descriptive propensity score-weighted prevalence of each individual symptom among those with and without Long COVID were estimated by pregnancy status.
The 6-month Long COVID outcome was missing among participants who did not complete a 6-month symptom survey. We addressed this and other missingness in the covariates using multiple imputation via chained equations (M=25 imputed datasets), performing the propensity score analysis in each dataset and pooling the results. To minimize the influence of acute SARS-CoV-2 symptoms, for participants with a reported reinfection up to 30 days before or 7 days after their 6-month visit we also replaced their observed LCRI with missingness to be imputed.
Two additional secondary analyses considered Long COVID prevalence at 9 months and 12 months after index infection. These analyses addressed the potentially changing prevalence over time and also accounted for potential differences in measured symptoms at 6 months due to ongoing pregnancy or postpartum status among some participants in the pregnancy group. Because adjustment or selection based on post-exposure variables can distort estimates, we did not otherwise exclude or alter LCRI values for participants who were pregnant or in the early postpartum period (less than 12 weeks after delivery).
We performed several sensitivity analyses for the main comparison of 6-month Long COVID prevalence between exposure groups. We assessed the effect of changing the age range for eligibility from 18–45 to 18–40 years. We also considered alternative propensity score weights called ‘average treatment effect among the treated’ (ATT) weights, comparing Long COVID prevalences where the non-pregnant group was weighted to have the same distribution of covariates as the pregnant group.
Finally, in exploratory analyses of 6-month Long COVID prevalence, we performed separate propensity score matched analyses comparing the non-pregnant group with pregnant individuals whose infection occurred in particular time periods of gestation. We considered three subgroups infected during the second trimester, either first or second trimester, and third trimester. Those infected during the first trimester were not considered individually due to insufficient number of participants for analysis in that subgroup.
Statistical analyses were performed using R Software, using the ‘mice’ package for multiple imputation and the MatchThem package for propensity score estimation and weighting.16,17 All study data were stored in a Research Electronic Data Capture (REDCap) database housed in a FISMA moderate compliant environment. The current study was approved by the NYU Langone Health Institutional Review Board (IRB), which served as a single IRB for most sites, while others required local IRB approval. All participants provided written informed consent prior to enrollment. STROBE guidelines for cohort studies were followed.
Results
A total of 2,423 participants in RECOVER-Adult and RECOVER-Pregnancy met eligibility criteria for this study, of whom 580 (23.9%) were pregnant at index SARS-CoV-2 infection, and 1843 (76.1%) were not pregnant at index (Figure 1). There were 393 (16.2%) Hispanic or Latino participants, 311 (12.8%) non-Hispanic Black participants, and 216 (8.9%) non-Hispanic Asian participants. The median age at infection was 33 years (interquartile range [IQR]: 28–38), and 2,131 (90.0%) of participants with known vaccination status were fully vaccinated at least 2 weeks before infection (Table 1, Appendix 2, available online at http://links.lww.com/xxx). In the pregnancy group, 96.3% of reported pregnancy outcomes were live birth, with other adverse outcomes detailed in Appendix 3. available online at http://links.lww.com/xxx.
Figure 1.
Study population. Of the final study cohort, 25 pregnant and 72 nonpregnant participants intended to enroll into Researching COVID to Enhance Recovery (RECOVER) as uninfected participants, but were recategorized at enrollment as infected due to a positive SARS-CoV-2 antibody test. These participants had an index date defined as 90 days before the positive antibody test.
Table 1.
Baseline Characteristics
| Characteristic | During Pregnancy (n=580) | Not During Pregnancy (n=1843) | Overall (n=2423) |
|---|---|---|---|
| Age at index, Mean (SD) | 33 (5) | 33 (7) | 33 (6) |
| Race and Ethnicity | |||
| Hispanic | 71 (12.2%) | 322 (17.5%) | 393 (16.2%) |
| Non-Hispanic Asian | 36 (6.2%) | 180 (9.8%) | 216 (8.9%) |
| Non-Hispanic Black | 77 (13.3%) | 234 (12.7%) | 311 (12.8%) |
| Non-Hispanic White | 367 (63.3%) | 988 (53.6%) | 1355 (55.9%) |
| Multiracial / Other / Missing | 29 (5.0%) | 119 (6.5%) | 148 (6.1%) |
| Vaccination status two weeks before infection | |||
| Unvaccinated | 70 (12.7%) | 104 (5.7%) | 174 (7.3%) |
| Partially vaccinated / date of last dose unknown | 16 (2.9%) | 47 (2.6%) | 63 (2.7%) |
| Fully vaccinated | 466 (84.4%) | 1,665 (91.7%) | 2,131 (90.0%) |
| Missing | 28 | 27 | 55 |
| Comorbidities in year preceding infectiona | |||
| Obesity | 133 (23.1%) | 457 (25.0%) | 590 (24.5%) |
| Depression or anxiety | 247 (42.9%) | 923 (50.4%) | 1,170 (48.6%) |
| Asthma | 115 (19.9%) | 354 (19.3%) | 469 (19.5%) |
| Cardiovascular diseasesb | 42 (7.3%) | 133 (7.3%) | 175 (7.3%) |
| Diabetes | 26 (4.5%) | 61 (3.3%) | 87 (3.6%) |
| Rheumatologic, autoimmune or connective tissue disease | 47 (8.2%) | 209 (11.4%) | 256 (10.6%) |
| Chronic pain syndrome or fibromyalgia | 15 (2.6%) | 91 (5.0%) | 106 (4.4%) |
| Immunocompromised conditions | 6 (1.0%) | 94 (5.1%) | 100 (4.1%) |
| ME/CFS, POTS or neurologic condition | 22 (3.8%) | 122 (6.7%) | 144 (6.0%) |
| Tobacco use in year preceding infection | 62 (10.7%) | 272 (14.8%) | 334 (13.8%) |
| Study participation via self-referral, community outreach, or Long COVID clinic referral | 206 (35.5%) | 985 (53.4%) | 1,191 (49.2%) |
| Social Risk Factor Index:c Economic Instability | |||
| 0 | 356 (61.4%) | 1,040 (56.4%) | 1,396 (57.6%) |
| 1 | 127 (21.9%) | 478 (25.9%) | 605 (25.0%) |
| 2 | 58 (10.0%) | 227 (12.3%) | 285 (11.8%) |
| 3+ | 39 (6.7%) | 98 (5.3%) | 137 (5.7%) |
| Social Risk Factor Index: Education and Language Barriers | |||
| 0 | 480 (82.8%) | 1,562 (84.8%) | 2,042 (84.3%) |
| 1+ | 100 (17.2%) | 281 (15.2%) | 381 (15.7%) |
| Social Risk Factor Index: Health Care Access and Quality Challenges | |||
| 0 | 380 (65.5%) | 1,114 (60.4%) | 1,494 (61.7%) |
| 1 | 158 (27.2%) | 544 (29.5%) | 702 (29.0%) |
| 2+ | 42 (7.2%) | 185 (10.0%) | 227 (9.4%) |
| Social Risk Factor Index: Lack of Social and Community Support | |||
| 0 | 292 (50.3%) | 486 (26.4%) | 778 (32.1%) |
| 1 | 172 (29.7%) | 566 (30.7%) | 738 (30.5%) |
| 2 | 72 (12.4%) | 446 (24.2%) | 518 (21.4%) |
| 3+ | 44 (7.6%) | 345 (18.7%) | 389 (16.1%) |
| Neighborhood-level proportion below Federal Poverty Level | |||
| Lowest Quintile | 123 (23.1%) | 343 (21.1%) | 466 (21.6%) |
| Second Quintile | 140 (26.3%) | 372 (22.9%) | 512 (23.7%) |
| Third Quintile | 90 (16.9%) | 291 (17.9%) | 381 (17.7%) |
| Fourth Quintile | 75 (14.1%) | 226 (13.9%) | 301 (14.0%) |
| Highest Quintile | 105 (19.7%) | 392 (24.1%) | 497 (23.0%) |
| Missing | 47 | 219 | 266 |
| Neighborhood-level proportion of ‘crowded’ households | |||
| Lowest Quartile | 236 (44.3%) | 538 (33.1%) | 774 (35.9%) |
| Second Quartile | 109 (20.5%) | 286 (17.6%) | 395 (18.3%) |
| Third Quartile | 114 (21.4%) | 400 (24.6%) | 514 (23.8%) |
| Highest Quartile | 74 (13.9%) | 400 (24.6%) | 474 (22.0%) |
| Missing | 47 | 219 | 266 |
Missingness of individual comorbidities not shown, but was no more than 0.8% in either group. Overall, 2,062 (85.1%) of participants had complete data for all of the covariates. Categorical variable percentages are calculated excluding missing values.
Cardiovascular disease includes heart failure, myocardial infarction, and chronic hypertension (with or without treatment).
Social Risk Factor Index values are sums of individual binary items detailed in Appendix 2, available online at http://links.lww.com/xxx
Overall, 1,840 (75.9%) of participants had an LCRI available at 6 months, among whom 33 (7.5%) of 440 participants who were pregnant at infection had an LCRI ≥ 11 and were classified as likely Long COVID, and 155 (11.1%) of 1400 participants who were not pregnant at infection were classified as likely Long COVID (Appendix 4, available online at http://links.lww.com/xxx).
From this analysis, the corresponding propensity score weighted estimates of prevalence of Long COVID at 6 months after index infection were 10.2% (95% Confidence Interval [CI]: 6.2%, 14.3%) among those pregnant at infection, and 10.6% (95% CI: 8.8%, 12.4%) among those not pregnant at infection. The propensity score distributions by pregnancy group had reasonable overlap, and the resulting propensity score weighted groups were well balanced on the covariates (Appendix 5, available online at http://links.lww.com/xxx). In the primary propensity score weighted analysis reported in Table 2, there was not a statistically significant difference in risk of Long COVID between those pregnant and those not pregnant at the time of SARS-CoV-2 infection (adjusted risk ratio [RR]: 0.96, [95% CI: 0.63, 1.48]).
Table 2.
Estimates Comparing Long COVID Prevalence Among Participants Pregnant at Infection vs. Not Pregnant at Infection (Reference) Using Propensity Score Weighting
| Adjusted Long COVID Prevalence Estimate (95% CI) |
||||
|---|---|---|---|---|
| Adjusted Risk Ratio (95% CI) |
Adjusted Risk Difference (95% CI) |
Pregnant at Infection | Not Pregnant at Infection | |
| Primary Outcome | ||||
| 6 Month Long COVID Prevalence | 0.96 (0.63, 1.48) | −0.4% (−4.8%, 4.0%) | 10.2% (6.2%, 14.3%) | 10.6% (8.8%, 12.4%) |
| Secondary Outcomes | ||||
| 9 Month Long COVID Prevalence | 0.87 (0.55, 1.39) | −1.3% (−5.6%, 3.0%) | 9.2% (5.4%, 13.1%) | 10.5% (8.8%, 12.2%) |
| 12 Month Long COVID Prevalence | 0.95 (0.64, 1.43) | −0.5% (−5.5%, 4.4%) | 12.3% (7.8%, 16.7%) | 12.8% (10.8%, 14.8%) |
| Primary Outcome Sensitivity Analyses | ||||
| Restriction to Age 18–40 years | 0.93 (0.59, 1.46) | −0.7% (−5.1%, 3.7%) | 9.7% (5.6%, 13.7%) | 10.4% (8.4%, 12.3%) |
| Alternative propensity score weights based on ATTa | 0.83 (0.50, 1.38) | −1.6% (−6.1%, 2.9%) | 7.8% (5.3%, 10.4%) | 9.5% (5.8%, 13.1%) |
ATT: Average Treatment Effect on the Treated. Weights applied to non-pregnant group only, to match covariate distribution of pregnant group.
Figure 2 reports symptom prevalences among those with Long COVID by pregnancy group, after propensity score weighting. Among the symptoms comprising the Long COVID definition, though there is substantial uncertainty in the estimated differences, the symptoms more common in the non-pregnant group and with the largest differences between groups were brain fog (45.2% vs 67.7%, difference: −22.5% [95% CI: −45.9%, 0.9%]), post-exertional malaise (76.7% vs 91.3%, difference: −14.6% [95% CI: −34.4%, 5.1%]), and chest pain (14.7% vs. 27%, difference: −12.3% [95% CI: −30.2%, 5.5%]). Symptoms more common in the pregnant group with the largest differences were chronic cough (38.3% vs. 28%, difference: 10.3% [95% CI: −12.3%, 33%]), shortness of breath (39% vs. 30.7%, difference: 8.3% [95% CI: −15.3%, 31.9%]), and thirst (35.8% vs. 29.4%, difference: 6.4% [95% CI: −16.3%, 29%]). The median LCRI value among those with Long COVID was similar between those pregnant at infection (13.6, [IQR: 12–16]) compared with those not pregnant at infection (14.4 [IQR: 12–18]).
Figure 2.
Symptom frequencies (%) by pregnancy at infection and long COVID status, after propensity score weighting. The Long COVID Research Index (LCRI) is calculated by summing the score contributions for each symptom an individual has, out of the top 11 listed. The result is a score ranging 0–30. LCRI ≥ 11 is used in this analysis to classify individuals as having long COVID. P-E, post-exertional.
Among participants pregnant at index who had known pregnancy status 6 months after infection, 44.5% were still pregnant or less than 12 weeks postpartum. At 9 months after infection, this proportion dropped to 14.3%, and by 12 months after infection was only 1.2% (Appendix 3, available online at http://links.lww.com/xxx). At 9 months and 12 months respectively, 1755 (72.4%) and 1644 (67.8%) of participants had the Long COVID outcome available, with comparable retention in both groups (Appendix 4, available online at http://links.lww.com/xxx). The adjusted risk ratios for Long COVID prevalence between pregnant and not pregnant at index remained similar in magnitude and non-significant both at 9 months (RR: 0.87; [95% CI: 0.55, 1.39]) and at 12 months (RR: 0.95; [95% CI: 0.64, 1.43]). Compared to the prevalence of Long COVID at 6 months, the estimated prevalences in each group at 9 months (9.2% and 10.5%) were slightly lower and at 12 months (12.3% and 12.8%) were slightly higher.
Additional sensitivity analyses for Long COVID at 6 months after infection showed consistent results with the primary analysis (Table 2). Restriction to individuals age 18–40 years, which excluded 27 (4.7%) individuals pregnant at infection and 308 (17%) individuals not pregnant at infection, yielded similar results (RR: 0.93; [95% CI: 0.59, 1.46]), as did using the alternative ATT propensity score weight definition making the characteristics of the non-pregnant group similar to the pregnant group (RR: 0.83, [95% CI: 0.50, 1.38]). The corresponding estimates of prevalence using ATT weights were lower than the primary analysis, because they represented prevalence in a population with covariates similar to the pregnant group, which was younger and generally had fewer comorbidities and higher socioeconomic status than the combined cohort of similar age females represented by the ATE weights used in the primary analysis.
Finally, in exploratory analyses of the 6 month outcome stratifying pregnancy group by trimester of infection, results were consistent with the primary analysis comparing pregnant individuals with a second trimester infection with the non-pregnant group (RR: 0.92, [95% CI: 0.47, 1.81]; adjusted risk difference −0.7%, [95% CI −6.6%, 5.2%]), and comparing pregnant individuals with a first or second trimester infection with the non-pregnant group (RR: 0.99; [95% CI: 0.62, 1.56]; adjusted risk difference −0.1%, [95% CI −5.0%, 4.8%]). The estimated Long COVID prevalence was somewhat lower for the group with a third trimester infection relative to the non-pregnant group (RR: 0.60; [95% CI: 0.33, 1.10]; adjusted risk difference −4.4%, 95% CI −8.6%,−0.2%), however there remains substantial uncertainty in the estimates due to reduced sample size within these pregnancy subgroups (Appendix 6, available online at http://links.lww.com/xxx).
Discussion
In this multicenter cohort, the risk of Long COVID at 6, 9, or 12 months after index infection did not differ between similar-age females who acquired SARS-CoV-2 during pregnancy and those who acquired SARS-CoV-2 outside of pregnancy. In adjusted analyses, approximately 10% of the cohort met the definition of Long COVID. Symptoms of Long COVID differed between those who were and were not pregnant at the time of index infection with chronic cough, shortness of breath, and thirst being more common among those who acquired SARS-CoV-2 during pregnancy. Conversely, brain fog, post-exertional malaise, and chest pain were reported more frequently among those not pregnant at the time of SARS-CoV-2 acquisition.
Our results differ from those of the NIH RECOVER-EHR Cohort, which consists of deidentified data from large hospital systems participating in existing PCORnet and N3C consortia, with minimal overlap with the participants including in this observational cohort. The RECOVER-EHR cohort demonstrated initially7, and then validated in another larger data set18, that the acquisition of SARS-CoV-2 during pregnancy was associated with a lower risk of developing Long COVID. There are several possible explanations for this difference in results: (1) the EHR results rely on diagnostic codes rather than prospective symptom surveys and the use of Long COVID codes may have been more frequent for those who were not pregnant at the time of infection as some Long COVID symptoms may have been attributed to pregnancy or the postpartum state, (2) the EHR cohort encompasses more of the pandemic period and there may have been differential risk with the earlier variants of SARS-CoV-2 that were not observed in our cohort that was limited to the Omicron era, (3) the EHR relies on clinical documentation of SARS-CoV-2 infection which may be differentially prevalent in individuals who are pregnant compared with those who are not at time of infection, which may lead to greater ascertainment of more mild cases which are less likely to result in Long COVID, or (4) this prospective study may have insufficient sample size to detect the small risk differences that could be detected in the large EHR cohort studies.
While the observed rates of Long COVID were lower in the group who acquired SARS-CoV-2 during pregnancy (7.5% vs 11.1%), this difference was mitigated with propensity score matching to balance the groups at baseline. Pregnant people are typically younger, and have fewer comorbidities than those who do not pursue pregnancies. In other published literature, comorbidities are known to be associated with an increased risk of Long COVID.19–21 In an exploratory analysis, infection in the third trimester of pregnancy was associated with a slightly lower prevalence of Long COVID when compared with non-pregnant females; however, this should be interpreted with caution given the small size of this subgroup, and differences in statistical significance when evaluating the risk ratio and risk difference. Nonetheless, this is an area for future investigation. Some of the symptoms of Long COVID, such as shortness of breath, may overlap with symptoms seen in those without Long COVID during pregnancy or postpartum. However, sensitivity analyses at 9 and 12 months after index infection demonstrated similar results, at which time the majority of participants were no longer pregnant or within the first 12 weeks postpartum. For patients who have persistent symptoms beyond pregnancy and the postpartum period or symptoms beyond those expected as part of a normal pregnancy course, a diagnosis of Long COVID should be considered with appropriate referral to a specialist for evaluation and management.
Strengths of this study include a standardized survey that addresses an extensive list of symptoms of Long COVID at regular intervals after index infection.10,11 Both the pregnant and the non-pregnant group completed the same survey questions allowing for direct comparison. The population was geographically and ethnically diverse, as the cohort was recruited from 79 sites located in all regions of the U.S. We restricted the analysis to participants enrolled at the time of acute infection or within 4.5 months of index infection to more closely mimic a prospective cohort study in order to reduce selection bias introduced by enrollment of participants who were known to already have Long COVID at the time of enrollment. Finally, we were able to adjust for a number of baseline characteristics including sociodemographic characteristics that have been associated with Long COVID.8,19–21
Limitations include that participants were only enrolled during the time of the Omicron variant of SARS-CoV-2 or later. Thus, the results may not be generalizable to those who acquired SARS-CoV-2 early in the pandemic. The LCRI is a score that is used for research purposes and may not accurately reflect all patients who would be diagnosed with Long COVID in the clinical setting. We had a fixed sample size for this analysis, and a type II error is possible; however, the point estimates for the adjusted prevalence of Long COVID were notably similar between groups (10.2 vs 10.6%). The majority of the cohort had a pregnancy that resulted in a live birth; thus, we are not able to address whether pregnancy outcome affects the prevalence of Long COVID. Participants in this cohort were infected with SARS-CoV-2 during the time period of Omicron dominance, which precluded us from examining the effects of other variants. While we did not calculate a LCRI for participants with a re-infection in close proximity to a study visit (30 days before to 7 days after), other re-infections were not accounted for in our analysis and may have affected Long COVID symptoms. It is possible that we enrolled a higher risk group of individuals with pregnancies given that the majority of enrollment of the pregnancy cohort occurred at tertiary academic centers.
In conclusion, we did not find a statistically significant association between the acquisition of SARS-CoV-2 during pregnancy and development of Long COVID when compared with non-pregnant similar-age females. Our findings are in contrast to previously published studies, which further emphasizes the importance of ongoing investigation as to the role of pregnancy (if any) in risk modification for Long COVID. Future work could examine whether different variants of SARS-CoV-2, reinfections, or trimester at the time of infection differentially affect the development of Long COVID and symptoms of Long COVID. Regardless, approximately 10% of similar-age females developed Long COVID, which has significant ramifications at the population level.
Supplementary Material
Funding Source:
This study is part of the NIH Researching COVID to Enhance Recovery (RECOVER) Initiative, which seeks to understand, treat, and prevent the post–acute sequelae of SARS-CoV-2 infection (PASC). For more information on RECOVER, visit https://recovercovid.org/. We would like to thank the National Community Engagement Group (NCEG), all patient, caregiver and community representatives, and all the participants enrolled in the RECOVER initiative.
This research was funded by the National Institutes of Health (NIH) Agreements OTA OT2HL161847, OT2HL161841 and OT2HL156812 as part of the Researching COVID to Enhance Recovery (RECOVER) Research Initiative. Additional support was provided by grant R01 HL162373 from the NHLBI, NIH (Drs Reeder and Foulkes). No funding source had any role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Financial Disclosure:
Torri D. Metz is the site PI for a Pfizer study of Paxlovid in pregnancy, is a site PI for a Moderna study of RSV vaccination in pregnancy, and was the site PI for a Pfizer study of RSV vaccination in pregnancy. She has received UptoDate royalties for two topics on trial of labor after cesarean. Valerie Flaherman reports payment received from the Research Triangle Institute. Carmen Beamon reports payment from Wellcare of North Carolina for participation on the physician advisory board. Jeanette Brown reports payment from Breas Medical for educational seminars and from Baxter for an advisory board not related to this work. Kelly S. Gibson disclosed that her institution received funding from Materna. Rachel Hess received payment from Astellas Pharmaceuticals for Data Safety Monitoring Board service. Sally Hodder has served as an advisor for Gilead Sciences, Viiv Healthcare, and Merck and her institution has received funding from Merck and Gilead for research study conduct. Brenna L. Hughes disclosed receiving payments from UptoDate, Moderna, and John’s Hopkins. Grace A. McComsey served as an advisor for Merck, Gilead, and GlaxoSmithkline. Patrick S. Ramsey disclosed receiving payments from UptoDate and payment to his institution from the Texas Collaborative for Health Mothers and Babies (Texas PQC). Daniel W Skupski reports receiving payments from Organon, Inc. and Cooper Surgical. Alan Tita reports payment from Pfizer. M. Camile Hoffman disclosed her institution received payment for her expert testimony for one medicolegal trial from Wheeler, Trigg, and Associates (a defense attorneys firm). Her institution also received payment for a disease state presentation on postpartum depression and zuranolone from SAGE/Biogen. Leora Horwitz reports payment from the National Academy of Medicine. Andrea Foulkes reports payment from RoundTable.
Footnotes
The other authors did not report any potential conflicts of interest.
Each author has confirmed compliance with the journal’s requirements for authorship.
References:
- 1.National Academies of Sciences, Engineering, and Medicine. 2024. A Long COVID definition: A chronic, systemic disease state with profound consequences. Washington, DC: The National Academies Press. 10.17226/27768. https://nap.nationalacademies.org/catalog/27768/a-long-covid-definition-a-chronic-systemic-disease-state-with. Accessed February 2, 2025. [DOI] [Google Scholar]
- 2.Bohmwald K, Diethelm-Varela B, Rodriguez-Guilarte L, et al. Pathophysiological, immunological, and inflammatory features of long COVID. Front Immunol 2024; 15:1341600. doi: 10.3389/fimmu.2024.1341600 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Peterson LS, Stelzer IA, Tsai et al. Multiomic immune clockworks of pregnancy. Semin Immunopathol 2020; 42(4): 397–412. doi: 10.1007/s00281-019-00772-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Abu-Raya B, Michalski C, Sadarangani M, Lavoie PM. Maternal immunological adaptation during normal pregnancy. Front Immunol 2020; https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2020.575197/full, Accessed July 11, 2025. doi: 10.3389/fimmu.2020.575197 [DOI] [Google Scholar]
- 5.Fernández-de-Las-Peñas C, Martín-Guerrero JD, Pellicer-Valero ÓJ, et al. Female Sex Is a Risk Factor Associated with Long-Term Post-COVID Related-Symptoms but Not with COVID-19 Symptoms: The LONG-COVID-EXP-CM Multicenter Study. J Clin Med. 2022. Jan 14;11(2):413. doi: 10.3390/jcm11020413. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Shah DP, Thaweethai T, Karlson EW, et al. Sex Differences in Long COVID. JAMA Netw Open 2025; 8(1):e2455430. doi: 10.1001/jamanetworkopen.2024.55430. doi: 10.1001/jamanetworkopen.2024.55430. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Bruno AM, Zhang C, Xu Z, et al. Association between acquiring SARS-CoV-2 during pregnancy and post-acute sequelae of SARS-CoV-2 infection: RECOVER electronic health record cohort analysis. EClinicalMedicine. 2024. May 24;73:102654. doi: 10.1016/j.eclinm.2024.102654. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Metz TD, Reeder HT, Clifton RG, et al. Post-Acute Sequelae of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) After Infection During Pregnancy. Obstet Gynecol 2024; 144(3): 411–20. doi: 10.1097/AOG.0000000000005670 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Oliveira AMDSS, Carvalho MA, Nacul L, et al. Post-Viral Fatigue Following SARS-CoV-2 Infection during Pregnancy: A Longitudinal Comparative Study. Int J Environ Res Public Health. 2022. Nov 26;19(23):15735. doi: 10.3390/ijerph192315735. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Metz TD, Clifton RG, Gallagher R, et al. ; RECOVER Initiative. Researching COVID to enhance recovery (RECOVER) pregnancy study: Rationale, objectives and design. PLoS One. 2023. Dec 21;18(12):e0285351. doi: 10.1371/journal.pone.0285351. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Horwitz LI, Thaweethai T, Brosnahan SB, et al. Researching COVID to Enhance Recovery (RECOVER) adult study protocol: Rationale, objectives, and design. PLoS One. 2023. Jun 23;18(6):e0286297. doi: 10.1371/journal.pone.0286297. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Geng LN, Erlandson KM, Hornig M, et al. 2024 Update of the RECOVER-Adult Long COVID Research Index. JAMA. 2024. Dec 18. doi: 10.1001/jama.2024.24184. Epub ahead of print. [DOI] [Google Scholar]
- 13.Krieger N, Chen JT, Waterman PD. Using the methods of the Public Health Disparities Geocoding Project to monitor COVID-19 inequities and guide action for social justice. Harvard T. H. Chan School of Public Health. March 7, 2024, https://www.hsph.harvard.edu/thegeocodingproject/covid-19-resources/ [Google Scholar]
- 14.United States Census Bureau: 2014–2018 American Community Survey (ACS) 5-year estimates. 2020. Deposited 2020. https://www.census.gov/programs-surveys/acs/technical-documentation/table-and-geography-changes/2018/5-year.html. Accessed July 21, 2025.
- 15.Stuart EA, Green KM. Using full matching to estimate causal effects in nonexperimental studies: examining the relationship between adolescent marijuana use and adult outcomes. Dev Psychol. 2008. Mar;44(2):395–406. doi: 10.1037/0012-1649.44.2.395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.van Buuren S, Groothuis-Oudshoorn K (2011). “mice: Multivariate Imputation by Chained Equations in R.” Journal of Statistical Software, 45(3), 1–67. doi: 10.18637/jss.v045.i03. [DOI] [Google Scholar]
- 17.Pishgar F, Greifer N, Leyrat C, Stuart E (2021). “MatchThem:: Matching and Weighting after Multiple Imputation.” The R Journal. doi: 10.32614/RJ-2021-073, https://journal.r-project.org/archive/2021/RJ-2021-073/. [DOI] [Google Scholar]
- 18.Zang C, Guth D, Bruno AM, et al. Long COVID after SARS-CoV-2 during Pregnancy in the United States. Nat Commun. 2025. Apr 1;16(1):3005. doi: 10.1038/s41467-025-57849-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Wang C, Ramasamy A, Verduzco-Gutierrez M, Brode WM, Melamed E. Acute and post–acute sequelae of SARS-CoV-2 infection: a review of risk factors and social determinants. Virol J. 2023. Jun 16;20(1):124. doi: 10.1186/s12985-023-02061-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Jacobs ET, Catalfamo CJ, Colombo PM, et al. Pre-existing conditions associated with post–acute sequelae of COVID-19. J Autoimmun 2023. Feb;135:102991. doi: 10.1016/j.jaut.2022.102991. Epub 2023 Jan 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Stephens MD, Gazmararian JA, Khakharia A. Prevalence and risk factors of post–acute sequelae of COVID-19 among United States Veterans. Ann Epidemiol 2024. Jan;89:1–7. doi: 10.1016/j.annepidem.2023.11.006. Epub 2023 Nov 15. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.