Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Sep 10.
Published in final edited form as: Pharmacoepidemiol Drug Saf. 2025 Jul;34(7):e70180. doi: 10.1002/pds.70180

How COVID-19 Treatment in Pregnancy Reflects Healthcare Utilization During a Pandemic: A Two-Stage Individual Participant Data Meta-Analysis Combining Case-Based Registries

Emeline Maisonneuve 1,2,3,4, Odette De Bruin 5,6, Guillaume Favre 1,7, Erin Oakley 8, Jenny Yeon Hee Kim 8, Fouzia Farooq 8, Nouf Al-Fadel 9, Abdulaali Almutairi 9, Maria del Mar Gil 10,11,12, Irene Fernandez Buhigas 10,12, Silvia Visentin 13, Erich Cosmi 13, Fernanda Surita 14, Renato T Souza 14, José G Cecatti 14, Maria Laura Costa 14, Jose Sanin-Blair 15, Jorge E Tolosa 15,16,17, Eran Hadar 18, Anna Goncé 19, Christophe Poncelet 20, Fabienne Forestier 20, Thibaud Quibel 21, Begoña Martinez de Tejada 22, Béatrice Eggel-Hort 1,6, Romina Capoccia Brugger 23, Daniel Surbek 24, Luigi Raio 24, Anda-Petronela Radan 24, Monya Todesco-Bernasconi 25, Cécile Monod 26, Leonard Schäffer 27, Anett Harnadi 27, Sayed Hamid Mousavi 28,29, Diogo Ayres-de-Campos 30, Léo Pomar 1,31, Joanna Sichitiu 1,32, Laurent J Salomon 32, Yves Ville 32, Andrea Papadia 33, Marie-Claude Rossier 7, Lavinia Schuler-Faccini 34, Natalya Goncalves Pereira 35, Adolfo Etchegaray 36, Albaro Jose Nieto-Calvache 37, Michael Geary 38, Javiera Fuenzalida 39, Claudia Grawe 40, Albert I Ko 41, Silke Johann 42, Marco De Santis 43, Cora Alexandra Voekt 44, Najeh Hcini 45, Karin Nielsen-Saines 46, Charles Garabedian 47, Loïc Sentilhes 48, Otto H May Feuerschuette 49, Grit Vetter 50, Manggala Pasca Wardhana 51, Irida Dajti 52, Kitty W M Bloemenkamp 6, Satu J Siiskonen 53, Emily R Smith 8, David Baud 1, Alice Panchaud 2,54, Miriam C J M Sturkenboom 5
PMCID: PMC12418691  NIHMSID: NIHMS2098232  PMID: 40665803

Abstract

Purpose:

To describe an international response to the COVID-19 pandemic by estimating the prevalence of medication use for COVID-19 treatment in pregnancy, stratified by hospitalization, trimester of pregnancy, and country.

Methods:

We conducted a two-stage individual participant data meta-analysis of proportions from primary data on medications used to treat COVID-19 during pregnancy. A common data model was developed to pool the data from single-country and international registries. Data from pregnant individuals with COVID-19 between February 2020 and October 2022 were included in study platforms across 9 data sources. Patient information was abstracted from medical records.

Results:

Among 24 937 pregnant individuals, the pooled prevalences of individuals receiving medications to treat COVID-19 were: 34.7% heparin, 9.8% antibiotics, 4.9% corticosteroids, 2.2% antivirals, 0.8% antimalarials, 0.3% convalescent plasma, 0.2% immunosuppressants, and 0.02% monoclonal antibodies. Prevalence of medication use was higher in hospitalized individuals than in non-hospitalized individuals: 58.4% versus 17.9% for heparin, 26.9% versus 5.7% for antibiotics, 17.5% versus 1.3% for corticosteroids, 10.3% versus 0.3% for antivirals, and 4.5% versus 0.1% for antimalarials. The prevalence of corticosteroid use was lower in the first trimester (0.1%) compared with the second (7.2%) and third (4.9%) trimesters of pregnancy. The prevalence of medications differed widely across countries.

Conclusion:

Medication to treat COVID-19 was more frequently used in pregnant individuals hospitalized for COVID-19. Corticosteroids were used less in the first trimester of pregnancy. The differences in use between countries could reflect differences in the clinical management and access to medications for this population at risk of severe disease.

Keywords: CONSIGN group, COVID-19, medication use, meta-analysis, pregnancy

Plain Language Summary:

Our aim was to provide insight into healthcare utilization during the COVID-19 pandemic by estimating the prevalence of medication use for COVID-19 treatment during pregnancy, based on gestational trimester, hospitalization status for COVID-19 episodes, and country. An international meta-analysis was conducted using primary data on medications used for COVID-19 treatment during pregnancy. A harmonized method was used to pool data from different international data study platforms. Pregnant individuals with COVID-19 in 2020–2022 were included across 27 countries. Patient information was abstracted from medical records. Among the 24 937 pregnant individuals with COVID-19 included in this meta-analysis, the combined prevalences of medications received to treat COVID-19 were as follows: 34.7% heparin, 9.8% antibiotics, 4.9% corticosteroids, 2.2% antivirals, 0.8% antimalarials, 0.3% convalescent plasma, 0.2% immunosuppressants, 0.02% anti-SARS-CoV-2 monoclonal antibodies, and 0.00% intravenous immunoglobulins. Prevalence was much higher in hospitalized individuals than in non-hospitalized individuals: 58.4% versus 17.9% for heparin, 26.9% versus 5.8% for antibiotics, 17.5% versus 1.3% for corticosteroids, 10.3% versus 0.3% for antivirals, and 4.5% versus 0.1% for antimalarials. The prevalence of using corticosteroids was lower in the first trimester of pregnancy (0.1%) when compared with the second (7.2%) and third (4.9%) trimesters. The prevalence of any medication varied widely across countries.

1 |. Introduction

The coronavirus disease 2019 (COVID-19) pandemic raised unique concerns for pregnant individuals because SARS-CoV-2 infection increased their risk of death, severe maternal comorbidities, and neonatal morbidity [1]. For non-pregnant adults, international clinical trial groups, based on extensive multicenter collaboration and adaptive design in randomized controlled trials (RCT), provided an influx of good-quality evidence that enabled the immediate implementation of the findings into COVID-19 treatment guidelines [2–4]. Conversely, because pregnant women were not included in most of the RCT evaluating the efficacy and safety of medications to treat COVID-19, they may have been undertreated [5, 6]. Therefore, international pharmacoepidemiology studies were required to close the gaps in knowledge about the use of these treatments across all trimesters of pregnancy. A multinational study of European population-based cohorts reported the prevalence of medication use in 911 pregnant individuals hospitalized for COVID-19 but was not able to carry out a meta-analysis because of small numbers [7]. The literature provides a few meta-analyses on the use of medication in pregnant individuals. However, they were based primarily on case series and case reports with small sample sizes between 599 and 1742 pregnant individuals, limiting the pooling of homogenous, high-quality data [8–10]. The published studies are heterogeneous, covering either inpatients or outpatients or both settings, which adds to the complexity of analysis and interpretation [7, 11]. Moreover, their data were collected at the beginning of the pandemic, and the pharmacological treatments for COVID-19 differed according to national guidelines and have evolved since then [12].

Therefore, the aim of our international meta-analysis was to pool available primary data from around the world with similar protocols and settings, harmonize the variables created by the participating study sites using a common data model, and describe the use of medication to treat COVID-19 stratified by hospitalization status, trimester of pregnancy at COVID-19 infection, and country. The results and lessons learned from our collaborative strategy should be useful for improving health utilization in pregnant individuals in the next pandemics.

2 |. Methods

Through the initiative of the European Medicines Agency (EMA) and the International Coalition for Medicines Regulatory Agencies (ICMRA), the Covid-19 infection and medicines in pregnancy (CONSIGN) international group was launched by the EU PE&PV Research Network, utilizing data sharing through a common data model approach [13, 14].

2.1 |. Data Sources

Sites or networks with primary data collections on pregnancies testing positive for COVID-19 and containing information on medication used to treat COVID-19 were invited to participate. The search strategy and eligibility criteria are described in Figure 1 and online [14]. The included data sources were the United States Centers for Disease Control and Prevention (CDC) Surveillance for Emerging Threats to Pregnant People and Infants Network (SET-NET), COVI-PREG international registry, George Washington University Prospective Meta-Analysis (GWU PMA), and Saudi Food and Drug Authority (FDA). The 10 SET-NET jurisdictions and the 26 COVI-PREG countries were kept aggregated for statistical purposes and thus each contributed one data source. Saudi-FDA provided two data sources. GWU study sites (n = 5) were not aggregated due to missing variables differing between sites (Table S1). GWU PMA participants also included in COVI-PREG were excluded.

FIGURE 1 |.

FIGURE 1 |

Open in a new tab

Flowchart of the search and selection process. N corresponds to the number of data sources. CDC/SET-NET: United States Centers for Disease Control and Prevention/Surveillance for Emerging Threats to Pregnant People and Infants Network, CONSIGN: Covid-19 infectiOn aNd medicineS In preGNancy, GWU PMA: George Washington Prospective Meta-Analysis, Saudi FDA: Saudi Food and Drug Authority. A data source is defined by a dataset on pregnant individuals with COVID-19 collected by the same group of researchers. It can be a single study site (i.e., study sites from GWU PMA and Saudi FDA), a national multi-sites dataset (i.e., SET-NET) or an international multi-site collaboration (e.g., COVI-PREG). The names of the excluded data sources not shown on this flowchart are available in the Table 1 of deliverable 6 published on Zenodo: “Description and characterization of the data sources to be included in the CONSIGN-International meta-analysis” [14]. *Medications were not available for some GWU PMA sites at the time of the meta-analysis, as code was only recently developed.

2.2 |. Study Population

2.2.1 |. Inclusion Criteria

Pregnant individuals, identified at any point during pregnancy from February 2020 and that reached the theoretical gestational age of 42 weeks by October 31, 2022, in any health care facility for antenatal care or labor and delivery ward with a positive test for SARS-CoV-2 were eligible. For SET-NET, pregnant individuals with COVID-19 were identified through national reporting of positive SARS-CoV-2 testing and linkages with vital records, administrative datasets to ascertain pregnancy status [15].

2.2.2 |. Exclusion Criteria

Self-reported pregnancies were excluded in order to ensure that only accurate information on medications collected by health care workers was used.

2.3 |. Variables

2.3.1 |. Exposure

The medications mentioned in the World Health Organization (WHO) “Therapeutics and COVID-19” living guidelines, or in the National Institutes of Health (NIH) guidelines section “special considerations for pregnancy” were defined as medication used to treat COVID-19 (Table S2) [16–21]. Any pregnant individual with a report of administration of at least one of these medications to treat her after a COVID-19 episode was considered exposed. Medications given > 30 days after diagnosis were not included.

Medications were coded using the Anatomical Therapeutic Chemical (ATC) classification system [22]. Information on exact dose, duration of treatment, and calendar date of prescription were not collected. Further details on the timing of COVID-19-related medication are given in Table S3.

2.3.2 |. Covariates

Maternal socio-demographic and pregnancy characteristics, along with information on SARS-CoV-2 infection, were abstracted from medical records. The trimester of pregnancy at SARS-CoV-2 infection was determined based on calculated date of last menstrual period and/or estimated due date and date of first positive SARS-CoV-2 laboratory result [23].

Hospitalization for COVID-19 as primary diagnosis was a binary variable used for stratification, as criteria for disease severity differed across the study sites. Patients hospitalized for delivery or any obstetric reasons and screened for COVID-19 at hospital were considered not hospitalized for COVID-19.

2.4 |. Statistical Analysis

We conducted a two-stage individual participant data (IPD) meta-analysis. The protocol, the codebook with all common variables, and study analysis plan were established a priori and are available online [24].

2.4.1 |. Common Data Model

The Saudi FDA and COVI-PREG shared anonymized IPD and already transformed their variables into the common data model, allowing the CONSIGN team to perform the analysis. SET-NET and GWU PMA provided aggregated data aligned with study protocol and the common variable codebook.

2.4.2 |. Descriptive Analyses

Descriptive analyses of baseline maternal, obstetric, and COVID-related characteristics are presented by study site. The prevalence of medication use for COVID-19 treatment was defined as the proportion of pregnant individuals exposed to one medication, divided by the total number of included pregnant individuals [25]. Prevalence was calculated with 95% confidence intervals (95% CI) using normal approximation and stratified by hospitalization status, trimester of pregnancy at COVID-19 infection, and country.

2.4.3 |. Meta-Analysis

Data and pooled estimates are presented in forest plots to illustrate individual site estimates and 95% CI for each data source, with a diamond representing the pooled prevalence and 95% CI for each medication at ATC-second level. A random-effects model was used due to differences in source data [26]. Heterogeneity was assessed with the I2 statistic [27]. The statistical analyses were performed with R software, version 4.2.2 [28]. The pooled prevalence of medications was computed with the “meta” package and the “metaprop” function. The Logit transformation of proportions (parameter: sm = “PLOGIT”) was implemented to calculate an overall proportion. A generalized linear mixed model (GLMM), specifically, a random intercept logistic regression model, was utilized for the meta-analysis of proportions. Forest plots are displayed by hospitalization status and trimester at infection for the medications with a prevalence over 0.5%. The results of the respective statistical tests are indicated in the forest plots’ footnotes.

2.4.4 |. Missing Values

For each specific medication, we used the missing indicator methods; when the variable was unknown or missing, it was coded as not receiving treatment.

SET-NET did not collect specific information on the reason for hospitalization. SET-NET analysts developed an algorithm to extract information on whether there was a hospitalization for COVID-19 as the primary diagnosis. This was based on content in free text fields and diagnosis codes related to COVID-19 that required hospitalization. If it could not be determined that there was a hospitalization with COVID-19 as the primary diagnosis, then that variable was considered unknown. Individuals missing information on a stratification variable were excluded from that analysis for completeness.

3 |. Results

A total of 49 data sources containing information on pregnant individuals with COVID-19 and medication use were identified (Figure 1). Among these, 32 were data sources with IPD. Fourteen were excluded, and nine decided to withdraw from the meta-analysis. In total, nine study sites or networks were included in the meta-analysis of case-based registries: CDC/SET-NET with 10 jurisdictions; COVI-PREG with 26 countries; five study sites from GWU PMA in Brazil, Colombia, Italy, Spain, and the US; and two study sites from Saudi FDA.

The maternal, COVID-related, and obstetric characteristics of the 24 937 pregnant individuals from the nine data sources included in the meta-analysis are described in Table 1 and Table S4, respectively.

TABLE 1 |.

Maternal and COVID-related characteristics of the participants by study.

CDC/SET-NET # USA N = 19 090 GWU PMA Cornell USA N = 291 GWU PMA Brazil N = 280 GWU PMA Colombia N = 406 GWU PMA Padua Italy N = 73 GWU PMA Madrid Spain N = 517 COVI-PREG International N = 3582 Saudi FDA Jeddah Saudi Arabia N = 56 Saudi FDA Riyadh Saudi Arabia N = 642
Maternal age (yrs)
 12–24 4577 (24.0) NA 54 (19.3) NA 7 (9.6) 38 (7.4) 359 (10.0) 4 (7.1) 74 (11.5)
 25–34 11 102 (58.2) NA 148 (52.9) NA 37 (50.7) 262 (50.7) 2149 (60.0) 34 (60.7) 373 (58.1)
 35–55 3411 (17.9) NA 78 (27.9) NA 28 (38.4) 217 (42.0) 1060 (29.6) 18 (32.1) 195 (30.4)
 Missing 0 291 (100) 0 406 (100) 1 (1.4) 0 14 (0.4) 0 0
Respiratory d.
 Yes 1115 (5.8) 29 (10.0) 21 (7.5) 20 (4.9) 0 16 (3.1) 104 (2.9) 2 (3.6) 23 (3.6)
 No 4381 (22.9) 262 (90.0) 259 (92.5) 306 (75.4) 73 (100) 501 (96.9) 3458 (96.5) 54 (96.4) 617 (96.1)
 Unknown/missing 13 594 (71.2) 0 0 80 (19.7) 0 0 15 (0.4) 0 2 (0.3)
Chronic renal d.
 Yes 84 (0.4) 0 1 (0.4) 1 (0.2) NA 2 (0.4) 25 (0.7) 1 (1.8) NA
 No 4980 (26.1) 291 (100.0) 279 (99.6) 325 (80.0) NA 515 (99.6) 3537 (98.7) 55 (98.2) NA
 Unknown/missing 14 026 (73.5) 0 0 80 (19.7) 73 (100) 0 20 (0.6) 0 642 (100)
Cancer
 Yes NA NA NA NA NA NA 17 (0.5) 1 (1.8) 3 (0.5)
 No NA NA NA NA NA NA 3545 (99.0) 55 (98.2) 639 (99.5)
 Unknown/missing 19 090 (100) 291 (100) 280 (100) 406 (100) 73 (100) 517 (100) 20 (0.6) 0 0
Cardiovascular d.
 Yes 315 (1.7) 5 (1.7) 2 (0.7) 0 1 (1.4) 7 (1.4) 56 (1.6) 1 (1.8) 3 (0.5)
 No 4917 (25.8) 286 (96.3) 278 (99.3) 326 (80.3) 72 (98.6) 510 (98.6) 3506 (97.9) 55 (98.2) 639 (99.5)
 Unknown/missing 13 858 (72.6) 0 0 80 (19.7) 0 0 20 (0.6) 0 0
Diabetes mellitus
 Yes 455 (2.4) 9 (3.1) 5 (1.8) 9 (2.2) 2 (2.7) 4 (0.8) 50 (1.4) 2 (3.6) 50 (7.8)
 No 17 144 (89.8) 282 (96.9) 275 (98.2) 317 (78.1) 28 (38.4) 513 (99.2) 3512 (98.0) 54 (96.4) 592 (92.2)
 Unknown/missing 1491 (7.8) 0 0 80 (19.7) 0 0 20 (0.6) 0 0
Chronic hypertension
 Yes 875 (4.6) 7 (2.4) 25 (8.9) 22 (5.4) 0 (0.0) 3 (0.6) 87 (2.4) 1 (1.8) 9 (1.4)
 No 16 750 (87.7) 284 (97.6) 255 (91.1) 304 (74.9) 73 (100) 514 (99.4) 3475 (97.0) 55 (98.2) 633 (98.6)
 Unknown/missing 2108 (11.1) 0 0 80 (19.7) 0 0 20 (0.6) 0 0
Auto-immune d.
 Yes 213 (1.1) 8 (2.7) 1 (0.4) 3 (0.7) 2 (2.7) 2 (0.4) 52 (1.5) 2 (3.6) NA
 No 4689 (24.6) 283 (97.3) 279 (99.6) 323 (79.6) 71 (97.3) 515 (99.6) 3079 (86.0) 54 (96.4) NA
 Unknown/missing 14 188 (74.4) 0 0 80 (19.7) 0 0 451 (12.6) 0 642 (100.0)
Obesity*
 Yes 5613 (29.4) 42 (14.4) 71 (25.4) 24 (5.9) NA 73 (14.1) 677 (18.9) 21 (37.5) 351 (54.7)
 No 13 415 (70.3) 102 (35.1) 127 (45.4) 80 (19.7) NA 438 (84.7) 2210 (61.7) 34 (60.7) 291 (45.3)
 Unknown/missing 62 (0.3) 147 (50.5) 82 (29.3) 302 (74.4) 73 (100) 6 (1.2) 695 (19.4) 1 (1.8) 0
Hematologic d. **
 Yes NA 1 (0.3) 1 (0.4) NA NA 4 (0.8) 26 (0.7) 0 NA
 No NA 290 (99.7) 279 (99.6) NA NA 513 (99.2) 3106 (86.7) 56 (100) NA
 Unknown/missing 19 090 (100) 0 0 406 (100) 73 (100) 0 450 (12.5) 0 642 (100)
Type of test
 RT-PCR 19 090 (100) 262 (90.0) 261 (93.2) 268 (66.0) NA 157 (30.4) 2968 (82.9) NA NA
 Antigen 0 0 0 10 (2.5) NA 305 (59.0) 496 (13.8) NA NA
 Serology 0 20 (6.9) 5 (1.8) 4 (1.0) NA 55 (10.6) 0 NA NA
 Unknown/missing 0 9 (3.1) 12 (4.3) 124 (30.5) 73 (100) 0 118 (3.3) 56 (100) 642 (100)
Hospitalization ***
 Yes 713 (3.7) 26 (8.9) 43 (15.4) 224 (55.2) 66 (90.4) 15 (2.9) 837 (23.4) 14 (25.0) 326 (50.8)
 No 11 431 (59.9) 15 (5.2) 195 (69.6) 104 (25.6) 7 (9.6) 500 (96.7) 2708 (75.6) 42 (75.0) 312 (48.6)
 Unknown/missing 6946 (36.4) 250 (85.9) 42 (15.0) 78 (19.2) 0 2 (0.4) 37 (1.0) 0 4 (0.6)
Time of infection
 First trimester 3975 (20.8) 2 (0.7) 23 (8.2) 21 (5.2) 2 (2.7) 128 (24.8) 483 (13.5) 3 (5.4) 36 (5.6)
 Second trimester 5909 (31.0) 30 (10.3) 65 (23.2) 49 (12.1) 18 (24.7) 208 (40.2) 1046 (29.2) 14 (25.0) 38 (5.9)
 Third trimester 9206 (48.2) 235 (80.8) 95 (33.9) 128 (31.5) 53 (72.6) 181 (35.0) 1837 (51.3) 39 (69.6) 568 (88.5)
 Unknown/missing 0 24 (8.2) 97 (34.6) 208 (51.2) 0 0 216 (6.0) 0 0
COVID-19 vaccine
 Yes 0 0 0 NA NA 192 (37) NA NA NA
 No 19 090 (100) 2891 (100) 280 (100) NA NA 322 (62) NA NA NA
 Unknown/missing 0 0 0 406 (100) 73 (100) 3 (1) 3582 (100) 56 (100) 642 (100)
World Bank class. $
 LMIC 0 0 280 (100) 406 (100) 0 0 633 (17.7) 0 0
 HIC 19 090 (100) 291 (100) 0 0 73 (100) 517 (100) 2794 (78.0) 56 (100) 642 (100)
 Unknown/missing 0 0 0 0 0 0 155 (4.3) 0 0
Open in a new tab

Note: The percentages in columns are presented in parentheses.

#

CDC/SET-NET included data from 10 US jurisdictions (Arkansas, Chicago, Houston, Michigan, Minnesota, Nebraska, New Jersey, New York state, except New York City, Puerto Rico, Washington)

*

Obesity was defined as body mass index ≥ 30 kg/m2

**

Hematologic disease refers to sickle cell disease and thalassemia

***

Hospitalization for COVID-19 as the main diagnosis and not for any other obstetric reason.

$

World Bank classification by Income level. Some studies are pooled data (such as SET-NET and COVI-PREG) and others are broken down by site (GWU has 5 sites that are not pooled; Saudi FDA has 2 sites that are not pooled).

Abbreviations: d.: disease; HIC, high income countries; LMIC, low and middle income countries; RT-PCR, reverse transcriptase—polymerase chain reaction; yrs, years.

Participants were predominantly detected with SARS-CoV-2 infection during the third trimester of pregnancy (51.1%). The distribution of hospitalization for COVID-19 ranged from 2.9% in the study in Madrid, Spain to 90% in the study in Padua, Italy.

All studies commenced patient inclusion in 2020, with most of them including infection in 2021. However, each study period ended at different time points between December 2020 and the most recent data collected in October 2022 (Table S1). Only the GWU Madrid study site had vaccination data, with 37% of participants receiving at least one dose of a COVID-19 vaccine. All other studies provided information prior to the availability of the vaccine or did not provide vaccine information.

Antibiotic use in hospitalized pregnant individuals was 26.9% (95% CI 23.2–30.9; 9 study sites; I2 = 70%), compared with 5.7% (95% CI 1.8–16.6, I2 = 99%) in non-hospitalized individuals, with high heterogeneity across study sites. The prevalence of use increased with the trimester of pregnancy, but the difference in prevalence across trimesters was not significant. The most frequently used antibiotic was azithromycin (2.8% [95% CI 0.8–8.9] in total and 12.2% [95% CI 4.9–27.6]) during hospitalization (Figure 2, Tables S5 and S6, Figure S1). After stratification by continent, there was still high heterogeneity in the prevalence of antibiotic use, with the highest prevalence in South America (18.1%; 95% CI 10.0–31.0) (Figure S2).

FIGURE 2 |.

FIGURE 2 |

Open in a new tab

Prevalence of antibiotic use among all pregnant individuals with COVID-19, stratified by hospitalization. The proportions are described as binomial proportions (number of medications use/number of participants) and range from 0.00 to 1. A dot indicates that the information regarding the studied medication was missing or excluded in some study sites. The vertical dotted (…) and dashed (−) lines represent the pooled proportions using the random effects and common effect models, respectively. The numbers of medications do not systematically add up to the total, because there were unknown or missing data regarding hospitalization for the COVID-19 event in some study sites. These cases were excluded from the stratification. After subgroup analysis using random effects model, there was a statistical difference for antibiotics use between hospitalized and non-hospitalized pregnant individuals (p = 0.004).

Antiviral use was higher in pregnant individuals hospitalized for COVID-19 (10.3%; 95% CI 4.6–21.4; 6 study sites; I2 = 90%) than those not hospitalized for this diagnosis (0.3%; 95% CI 0.0–3.6; I2 = 97%). Use of antivirals was low and did not differ across trimesters of pregnancy (Figure 3). Remdesivir was the most frequently used antiviral, exclusively in cases of hospitalization for COVID-19; the use of nirmatrelvir-ritonavir was not reported in any study site (Tables S7 and S8, Figure 3). The heterogeneity was high across data sources for antiviral use, even within the same country (Figure S4).

FIGURE 3 |.

FIGURE 3 |

Open in a new tab

Prevalence of antiviral use among all pregnant individuals with COVID-19. The proportions are described as binomial proportions (number of medications use/number of participants) and range from 0.00 to 1. A dot indicates that the information regarding the studied medication was missing or excluded in some study sites. The vertical dotted (…) and dashed (−) lines represent the pooled proportion using the random effects and common effect models, respectively.

The prevalence of antimalarials (hydroxychloroquine) use was low at 0.8% (95% CI 0.5–1.4; 6 study sites; I2 = 81%) (Figures S5 and S6).

Corticosteroids were more commonly used in pregnant individuals hospitalized for COVID-19 as the main diagnosis (17.5%; 95% CI 11.2–26.3; 7 study sites; I2 = 87%) versus non-hospitalized (1.3%; 95% CI 0.3–5.4; I2 = 98%) (Figure S7). Usage differed between trimesters, with the highest prevalence during the second trimester, followed by the third trimester, and remaining low during the first trimester (Figure 4). The most frequently used corticosteroid was dexamethasone (Tables S9 and S10). Corticosteroid use was similar across European study sites, but heterogeneity was high in American sites (Figure S8).

FIGURE 4 |.

FIGURE 4 |

Open in a new tab

Prevalence of corticosteroid use among all pregnant individuals with COVID-19, stratified by trimester of pregnancy. The proportions are described in fractions (number of medications use/number of participants) and range from 0.00 to 1. A dot indicates that the information regarding the studied medication was missing or excluded in some study sites. The vertical dotted (…) and dashed (−) lines represent the pooled proportion using the random effects and common effect models, respectively. The numbers of medications do not systematically add up to the total, because there were unknown or missing data regarding trimester of pregnancy for the COVID-19 event in some study sites. These cases were excluded from the stratification. Subgroup analysis using random effect model found statistical difference for corticosteroids use across trimesters of pregnancy (p = 0.006).

In this meta-analysis, the total prevalence of immunosuppressant use was very low at 0.18% of pregnant individuals (95% CI 0.02–1.36; 5 study sites; I2 = 90%). There were no pregnant patients treated with interferon. The prevalences of intravenous immunoglobulins and convalescent plasma uses were respectively 0.00% (95% CI 0.00%–4.4%, 4 study sites, I2 = 0%) and 0.26% (95% CI 0.02–2.7; 5 study sites, I2 = 97%). The prevalence of antiviral monoclonal antibodies (casirivimab-imdevimab and sotrovimab) was also very low at 0.02% (95% CI 0.01–0.10, 4 studies, I2 = 42%) (Figure S9).

The prevalence of antithrombotics (heparin) use for COVID-19 varied highly across sites but was higher in hospitalized individuals (58.4%; 95% CI 42.8–72.5; 6 study sites; I2 = 92%) than in non-hospitalized individuals (17.9%; 95% CI 9.1–32.1; I2 = 97%) (Figures S10 and S11).

4 |. Discussion

4.1 |. Main Findings

This was an international meta-analysis of pooled data from 24937 pregnant individuals with COVID-19 derived from nine data sources, including 27 countries and 4 continents, using a common data model approach designed to answer similar public health questions on medication used to treat COVID-19. The pooled proportions of medication use were: 34.7% of heparin, 9.8% antibiotics, 4.9% corticosteroids, 2.2% antivirals, and 0.8% antimalarials. The high heterogeneity observed across the strata highlights an important disparity in use of medications to treat COVID-19 between the countries and regions, but also between trimesters of pregnancy and by hospitalization status.

4.2 |. Comparison With the Existing Scientific Literature

The prevalence of antibiotics use was 9.8% in all pregnant individuals, which is much less than that of 36% and 46% found in meta-analyses including pregnant patients with COVID-19 occurring before February 2021 [8, 9]. Azithromycin was the most frequently prescribed antibiotic. It was initially considered to treat COVID-19 for its antimicrobial properties and immunomodulatory activity [29]. Regarding its safety in pregnancy, most studies found no increase in major congenital malformations after azithromycin exposure during the first trimester [30]. Since mid-2020, most national guidelines do not recommend prescribing antibiotics unless an additional bacterial pulmonary infection is suspected [31–36].

The prevalence of antivirals was 10.3% among hospitalized patients versus 0.3% in those not hospitalized for COVID-19. The meta-analysis published by Lassi et al. found that the prevalence of antiviral use during pregnancy was associated with severe COVID-19, which is consistent with our findings [6]. We observed that the most prescribed antiviral was remdesivir. The first published studies suggested some benefit of remdesivir in shortening the median recovery time, but no benefit in reducing mortality in the general population [37]. The recommendations on remdesivir use in pregnant individuals differ between national guidelines. It is considered for compassionate use during pregnancy in the US and UK [19, 38]. We found no patients treated with nirmatrelvir-ritonavir, which was not yet recommended by national guidelines of the participating countries at the time of data collection [12].

Our meta-analysis found an extremely low prevalence of antimalarial use at 0.8%. Since the results of the RECOVERY trial, published in June 2020, which found hydroxychloroquine ineffective against COVID-19, many national societies of obstetricians-gynecologists have successively discouraged prescribing this drug [17, 20, 27, 39–41]. Besides, a RCT assessing the efficacy and safety of hydroxychloroquine in pregnancy had a very low recruitment rate with only 75 pregnant individuals [42].

In our study, 17.5% of hospitalized pregnant individuals received corticosteroids. This finding aligns with the recommendations on corticosteroids from the RECOVERY trial (including six pregnant individuals), which demonstrated a significant reduction in mortality within 28 days for patients requiring mechanical ventilation or oxygen supplementation [2]. Corticosteroids for maternal COVID-19 were significantly more frequently prescribed in the second trimester of pregnancy than in other trimesters, with dexamethasone being the most used. This difference may be attributed to obstetricians’ willingness to combine corticosteroids for both maternal and fetal purposes during the second trimester of pregnancy. Meta-analyses published by Bailey et al. and Lassi et al. reported an overall prevalence of corticosteroid use of 6% and 14.7% in pregnant populations, respectively, which is consistent with our estimate [8, 9].

The prevalences of immunosuppressants, immunomodulatory treatments, and monoclonal antibodies were extremely low in all study sites. In the meta-analysis published by Lassi et al., 16.2% and 22.5% of pregnant individuals received immunotherapy and plasma therapy, respectively, which appears unusually high. This may be attributed to inclusion bias, as evidenced by one of their studies in which all patients were treated with interferon alpha-2b [9].

Our meta-analysis included only 14 women treated with tocilizumab, despite the availability of this medication since 2020. This was likely influenced by the WHO guidelines on tocilizumab, which have recommended it only in cases with severe or critical COVID-19 since July 2021 [17]. The available data do not raise serious safety signals, but they have significant limitations and are not sufficient to delineate the complete spectrum of potential adverse outcomes that may be associated with tocilizumab exposure during pregnancy [43].

Monoclonal anti-SARS-CoV-2 antibodies were only prescribed to five patients, but the window of time during which they were recommended was very short. Casirivimab-imdevimab was recommended by WHO only between September 2021 and January 2022, according to specific criteria [44]. After that period, it was no longer considered effective against the Omicron variant.

Of all pregnant individuals, 34.7% received anticoagulation related to COVID-19 diagnosis. The prevalence tended to be higher in the case of infection during the second trimester of pregnancy. An observational study including six European electronic healthcare databases of pregnant outpatients found a significantly higher use of antithrombotic medications after COVID-19 infection in the second and third trimesters [11]. Our results may be explained by the fact that national guidelines suggest stopping anticoagulation when a birth is expected within 12 h, which was more likely to occur in the third trimester of pregnancy [38, 45, 46]. Anticoagulation use concerned 58.4% of hospitalized cases, which is also in line with most national guidelines, which recommend prophylactic anticoagulation in pregnant individuals with severe COVID-19 or requiring hospitalization [21, 47, 48].

4.3 |. Understanding Healthcare Utilization in Pregnant Individuals

The current meta-analysis suggests that pregnant individuals may be undertreated in the first trimester compared with those infected later in pregnancy, when looking at the lower prevalence of corticosteroids in first trimester compared with second and third trimesters. However, we did not analyze the patients requiring oxygen therapy stratified by trimester of pregnancy and this lower proportion of corticosteroid use in the first trimester could be due to the greater severity of COVID-19 in the second half of gestation [49, 50]. For antibiotics, antivirals, antimalarials and antithrombotics, the prevalences were also lower in the first trimester of pregnancy, but the differences were not significant. This trend may also be explained by the indication bias with more severe COVID-19 cases in the second and third trimesters. Nevertheless, the only way to demonstrate a reluctance to treat pregnant individuals is to concurrently recruit non-pregnant participants to account for changing treatment guidelines in the respective countries. This methodology was applied in the other meta-analysis conducted by the CONSIGN consortium on secondary use of electronic health care records (Figure 1), using a control group of non-pregnant individuals [50]. In that meta-analysis, the US Sentinel surveillance system matched 40 518 pregnant with 40 518 non-pregnant US outpatients. In the 30 days after COVID-19 diagnosis, antibiotics were the most common medication dispensed among both cohorts (16% in pregnant versus 12.2% in non-pregnant individuals). Pregnant individuals were more likely to receive analgesics than non-pregnant individuals (8.9% versus 3.7%), but less likely to have a dispensing of psychoanaleptics (5.1% versus 9.4%) and corticosteroids (5.7% versus 8.4%). They also found that the use of COVID-19-specific medications including remdesivir and monoclonal antibodies was very limited in pregnancies with COVID-19 (< 1%) [51].

Our meta-analysis was performed using conventional statistical methods, without the integration of artificial intelligence (AI). The application of AI in global health during future pandemics will demand thoughtful consideration of ethical concerns, particularly around data use and algorithmic bias. These AI-driven systems should address worldwide differences in data access and ensure that AI technologies are tailored to specific local needs. Ultimately, AI’s potential to enhance healthcare efficiency and equity will rely on international, multidisciplinary collaboration and the development of inclusive AI systems that promote equitable outcomes for all, including pregnant individuals [52].

The COVID-19 experience has taught us that we need to close the pregnancy-related data gap that is currently a barrier to gender equity in health innovation. Preparedness, networking, funding support, advocacy from trusted recommending organizations, confidence and education on how and when to include pregnant women in RCTs are paramount to ensure timely data collection and access to medicines and vaccines [53].

4.4 |. Strengths and Limitations

The strengths of our meta-analysis include the harmonized methods developed to standardize the perinatal variables and definitions of medications at ATC second and fifth levels. Through an international collaboration of researchers, perinatal epidemiologists, pharmacoepidemiologists, and healthcare workers, the common data model approach increased the accuracy of the data and its interpretation. Our international network of researchers enabled us to include a large number of pregnancies since the start of the pandemic. Additionally, our study populations were drawn from both high-income and middle- and low-income countries, enhancing the generalizability of our findings.

However, our meta-analysis presents several limitations. First, we considered “unknown” and “missing” medications as “no medication”, which may have underestimated the prevalence of use of the medications. Second, the study periods varied across the study sites, each with different predominant SARS-CoV-2 variants and recommended therapies, potentially confounding the prevalence estimates [12]. Third, heterogeneity was high, especially regarding the proportion of hospitalized individuals, likely driven by the different recruitment strategies utilized in the different case-based registries [54]. Fourth, we could not stratify on the COVID-19 severity according to WHO criteria. Fifth, most patients were not vaccinated because they were recruited in 2020 and early 2021, prior to vaccine availability. For pregnant individuals who contracted COVID-19 in 2021 and 2022, data on COVID-19 vaccination status was unavailable for several participating studies. This limitation may restrict the application of our findings to the current context, where most of the population has acquired immunity to COVID-19 through infection and/or vaccination. Finally, relatively limited data for 2021 and 2022 did not enable us to assess newer medications for COVID-19.

5 |. Conclusions

This meta-analysis, focusing on case-based registries of 24 937 pregnant individuals with COVID-19 from 2020 to 2022, found the following proportions of medication use: 34.7% for heparin, 9.8% for antibiotics, 4.9% for corticosteroids, 2.2% for antivirals, 0.8% for antimalarials, 0.3% for convalescent plasma, 0.18% for immunosuppressants, and 0.02% for monoclonal antibodies. These proportions were higher when pregnant individuals were hospitalized for COVID-19. The proportion of corticosteroid use was lower in the first trimester of pregnancy. The prevalence of medications used for treating COVID-19 varied highly across countries. The experience of pregnant individuals during the COVID-19 pandemic can provide valuable knowledge and preparedness to develop effective and inclusive strategies that protect their health in the future.

Supplementary Material

Supplementary material

Supporting Information

Additional supporting information can be found online in the Supporting Information section.

Summary.

  • In this meta-analysis, combining 24937 pregnant individuals from 27 countries, the pooled proportions of medications received to treat COVID-19 were as follows: 34.7% heparin (low molecular weight or unfractionated heparin), 9.8% antibiotics, 4.9% corticosteroids, 2.2% antivirals, 0.8% antimalarials, 0.3% convalescent plasma, 0.2% immunosuppressants, 0.02% monoclonal antibodies, and 0.00% intravenous immunoglobulins.

  • The prevalence of any medication use was significantly higher in pregnant individuals hospitalized for COVID-19 compared with those not hospitalized for COVID-19.

  • The prevalence of corticosteroid use to treat COVID-19 was significantly lower in the first trimester of pregnancy than in the second or third trimesters.

  • The prevalence of medication use to treat COVID-19 varied widely across countries, with the limitation that we used individual hospital registries and subnational studies to make these estimates.

  • In most countries in the world, there is no national surveillance system to monitor medication use in pregnancy. Therefore, multidisciplinary and international collaborations are needed, including the development of a common data model approach ready to use primary data in medical records from the start of a pandemic with a new pathogen.

Acknowledgements

Prof. Dr. Olaf Klungel (Utrecht University) for the EU PE&PV Research Network leadership; Dr. Hilde Engjom (Norwegian Institute of Public Health, Bergen, Norway) from the CONSIGN team.

Funding:

This study is part of the CONSIGN project funded by the European Medicines Agency (EMA) (SC04, FWC EMA/2018/28/PE, Lot 4). The participating sites had different funding to allow the primary data collection. COVI-PREG has been funded by the Swiss Federal Office for Public Health and the Lausanne University Hospital foundation. The George Washington University sequential prospective meta-analysis is funded by the Bill & Melinda Gates Foundation (INV-022057 to Emily Smith) (this doesn’t include the funding for participating countries).

Footnotes

Ethics Statement

In each country, primary data collection was approved by the respective Ethics Committee or carried out in compliance with applicable federal law or site policy, and the study was carried out in line with the General Data Protection Regulation (GDPR) and the Data Protection Impact Assessment (DPIA).

Conflicts of Interest

Begoña Martinez de Tejada received consultancy fees from Exeltis and Effik about medication for nausea and vomiting during pregnancy. All other authors declare no conflicts of interest.

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary material
NIHMS2098232-supplement-Supplementary_material.docx (3.6MB, docx)

RESOURCES