Evaluating COVID-19 vaccine effectiveness during pre-Delta, Delta and Omicron dominant periods among pregnant people in the U.S.: Retrospective cohort analysis from a nationally sampled cohort in National COVID Collaborative Cohort (N3C)

Overall study setting and design

We conducted a retrospective cohort analysis of a prospectively collected EHR repository. The N3C consortium is a high-granularity EHR data repository of deidentified, patient-level data from medical centres across the USA. The overall N3C design, data ingestion and harmonisation, and sampling approach have been described previously.24 25 Additional details about N3C and ethical review are in online supplemental materials.

Our dataset included COVID-19 vaccines currently given Food and Drug Administration (FDA) authorisation, including two mRNA vaccines (Pfizer-BioNTech (BNT162b2) and Moderna (mRNA-1273)) and a viral vector vaccine (Johnson & Johnson/Janssen (JNJ-784336725)), as well as other vaccines (eg, AstraZeneca). We categorised initial full vaccination as completion of the recommended dosing regimen of any vaccine (ie, at least two doses for mRNA or other vaccines and one dose for the Janssen vaccine) and partial vaccination as one dose of mRNA vaccine. We categorised any additional doses after 90 days of initial dosing (eg, a third mRNA dose) as booster vaccinations and only accounted for the first booster vaccination in this analysis.26 The concepts and codes to define vaccinations can be found in online supplemental table 1.

Study period and analytical cohort selection

Our study observation period occurred from 10 December 2020 to 7 June 2022. We used N3C data version 84 (released 7 July 2022), noting that we truncated the observation period 1 month earlier to allow for adequate time for data reporting. We defined the beginning of the observation period, 10 December 2020, based on the date the FDA first approved COVID-19 vaccination.27 We chose the end date, 7 June 2022, to allow 90 days prior to when the bivalent booster became available (9 September 2022) so that this analysis amply precludes bivalent booster impact (separate analysis forthcoming).

We stratified the analysis by variants predominant during the pandemic: pre-Delta (10 December 2020, to 20 June 2021), Delta (21 June 2021 to 25 December 2021) and Omicron (26 December 2021 to 7 June 2022), with predominance periods defined based on Centers for Disease Control and Prevention (CDC) estimates for the USA in general.27

Cohort construction and bias control

Figure 1 illustrates our cohort construction. We included data from 43 total contributing sites, excluding sites in the bottom quartile of vaccination rates. Next, we generated subsets of this cohort to identify pregnant people (cohort 1) and vaccinated females (cohort 2).

Request permission

For the analysis of initial vaccinations with cohort 1, we included pregnant people aged 15–55 with COVID-19 vaccinations during or before their pregnancy on or after 10 December 2020 (ie, the earliest date that COVID-19 vaccination was approved by FDA). For unvaccinated pregnant people, we included those with pregnancy end dates after 1 March 2021. We used 1 March 2021, the first peak of the initial full vaccination date among vaccinated pregnant people, as the ‘proxy initial vaccination date’. Similarly, for the analysis of booster vaccinations with cohort 1, we included pregnant people with a booster vaccination during or before their pregnancy after 1 December 2021. For unvaccinated pregnant people, we included those with pregnancy end dates after 1 December 2021. We used 1 December 2021, the peak of booster vaccination date among vaccinated pregnant people, as the ‘proxy booster vaccination date’. We defined proxy vaccination dates to avoid ascertainment bias in person-time at risk being longer in the unvaccinated versus vaccinated groups.

For analysis of initial and booster vaccinations with cohort 2, we included non-pregnant vaccinated (ie, those initial vaccinations/booster vaccinations) females of reproductive age to match with the pregnant vaccinated (ie, those initial vaccinations/booster vaccinations) females.

Cohort 1 analysis (all pregnant people)

In the analysis for initial and booster vaccinations in cohort 1, we compared vaccinated versus unvaccinated pregnant people aged 15–55 to evaluate the vaccine effectiveness in pregnancy via incident and severe COVID-19 infections. We categorised vaccinated pregnant people based on vaccination timing into four groups: received any initial or booster vaccination during pregnancy (vaccinated during pregnancy) and received all initial or booster vaccinations before pregnancy (vaccinated before pregnancy).

To further investigate the effectiveness of booster vaccinations, we compared pregnant people with full initial vaccinations and those with booster vaccinations via incident and severe COVID-19 infections.

Cohort 2 (all vaccinated females)

In the analysis for initial and booster vaccinations in cohort 2, we compared pregnant people vaccinated during or before pregnancy (same as cohort 1) versus non-pregnant females aged 15–55 to examine the impact of pregnancy on COVID-19 vaccine effectiveness.

For both cohorts 1 and 2, we excluded pregnant people with COVID-19 infections after their initial/booster vaccination (or proxy initial/booster vaccination dates) but before pregnancy. In the analysis for booster vaccinations, the included people were a subsample of the analysis cohort for initial vaccination.

Exposures definitions

In the analysis with cohort 1 (all pregnant people), vaccination status is the key exposure category. In the analysis with cohort 2 (all vaccinated females), pregnancy is the key exposure category.

Outcomes definitions

Our first coprimary outcome was incident and breakthrough COVID-19 infections among vaccinated and unvaccinated people, respectively. Our incident or breakthrough COVID-19 positivity definition only included positive results based on: (1) PCR-positivity, (2) antigen-positivity or (3) ICD diagnosis condition, in that hierarchical order.24 Breakthrough infection was defined by incident COVID-19 positivity after 14 days of the last vaccine received for initial/booster vaccination.2 22

Our second coprimary outcome was severe COVID-19 infections. We defined severe COVID-19 infections as incident or breakthrough COVID-19 infections with hospitalisation records within 14 days before or 45 days after infections that do not overlap within 7 days before or after the recorded delivery or pregnancy end date (to avoid confounding of hospitalisation due to deliveries or other pregnancy outcomes).2 28 The concepts and codes to define COVID-19 infections and hospitalisations can be found in online supplemental table 1.

Person-time at-risk for vaccinated people accrued from 14 days after their last date of vaccination (initial/booster) until the earliest date of COVID-19 incident or severe infections, death, transfer to hospice or the date of their last record in N3C. Person-time at-risk for unvaccinated pregnant people accrued similarly from 14 days after the proxy initial or booster vaccination date. More details on person-time at risk are illustrated in online supplemental figure 1. In the comparison of pregnant people with booster vaccinations versus those with full initial vaccinations, we used the full initial vaccination date plus 90 days as the start date to calculate person-time-at-risk for pregnant people without booster vaccinations to avoid ascertainment bias. The 90-day time window was based on the definition of booster vaccinations, as mentioned in the previous section.

Other outcomes included (1) ICU admission within 45 days after COVID-19 infection; (2) severe COVID-19 disease with invasive ventilation or extracorporeal membrane oxygenation (ECMO) treatment within 45 days after COVID-19 infection and (3) 30-day mortality after COVID-19 infection.2 29 Note that ICU admission, ventilation and ECMO outcomes are hospital based and, therefore, subsumed in the second coprimary outcome of severe COVID-19 infection.

Statistical analysis

We present summary characteristics at the time of the first initial COVID-19 vaccination date and proxy vaccination date for vaccinated and unvaccinated people, respectively. We used descriptive statistics only for outcomes with low counts in various groups.

We applied Cox proportional hazard models to estimate adjusted HRs (aHR) with 95% CIs. We clustered SEs at data partner sites to account for heterogeneity across sites. We adjusted for individual-level sociodemographic characteristics, including age, race, ethnicity, specific clinical comorbidities, Charlson Comorbidity Index and whether people had any prior COVID-19 infections before initial/booster or proxy vaccinations.

We stratified the models by the variant predominance periods (ie, pre-Delta, Delta and Omicron) to allow for different HR estimations during our observation period, and we assessed model assumptions accordingly. Person-time at-risk was also calculated for each variant stratum (ie, one within each stratum). We further separated the analysis for initial and booster vaccinations.

Of note, given the timing of booster vaccinations occurring mainly during the Omicron period, we only analysed booster vaccinations during the Omicron period. Since we anticipated pregnant people to possibly undergo more frequent COVID-19 testing than non-pregnant people (eg, for the delivery hospitalisation), we conducted an a priori sensitivity analysis restricting non-pregnant people in cohort 2 to people with at least one recorded COVID-19 test. Due to the concern of different effectiveness of mRNA and non-mRNA vaccines,30 we additionally conducted sensitivity analyses only including people with mRNA vaccinations (ie, Pfizer and Moderna).

Details about covariates and modelling can be found in online supplemental materials. All analyses were conducted in the N3C Data Enclave using PySpark and SparkR/R.31

Patient and public involvement

This is a retrospective observational study, patients were not involved.

Cohort characteristics

A total of over 12 million people had records included in N3C from 10 December 2020 to 7 June 2022 (figure 1). We included 301 107 people in cohort 1 and 934 337 people in cohort 2.

Among cohort 1 (pregnant people), 34 907 (11.6%) had any initial vaccination during pregnancy, another 34 538 (11.5%) had all initial vaccinations before pregnancy and 231 662 (76.9%) did not have any COVID-19 vaccinations recorded (table 1). The median age was around 32.0 years for pregnant people with vaccinations and 30.2 (IQR 25.0–33.8) for those without vaccination. Among those who had any initial vaccination during pregnancy, 28 537 (81.8%) had full initial vaccinations, 8415 (24.1%) had booster vaccination and 25 540 (73.2%) had received a Pfizer vaccine for their first dose. Among those who had all initial vaccinations before pregnancy, 31 530 (91.6%) had full initial vaccination, 14 730 (42.6%) had booster vaccination and 23 547 (68.2%) had received a Pfizer vaccine for their first dose.

Among cohort 2 (vaccinated females), 864 892 (92.6%) were non-pregnant vaccinated females (table 1). The characteristics of the vaccinated pregnant groups were the same as above for cohort 1. Among vaccinated non-pregnant females, the median age was 37.1 (IQR 28.1–47.2) years, 765 929 (88.6%) had full initial vaccinations, 460 967 (53.3%) had booster vaccination and 576 921 (66.7%) had received a Pfizer vaccine for their first dose.

Cohort 1 findings: pregnant people with initial vaccinations had a lower risk of incident COVID-19 infections compared with unvaccinated pregnant people during pre-Delta and Delta periods, and they had a lower risk of severe infections consistently throughout the pandemic stages

In cohort 1, 4323 (12.4%) and 4577 (13.3%) of pregnant people with initial vaccinations during and before their pregnancy had incident COVID-19 infections across the entire study period, and 352 (1.0%) and 185 (0.5%) of them had severe COVID-19 infections, respectively. Among unvaccinated pregnant people, 39 313 (17.0%) had incident COVID-19 infections and 3868 (1.7%) developed severe COVID-19 infections (table 1).

The adjusted models for incident COVID-19 infections resulted in aHRs of 0.36 (95% CI 0.25 to 0.51), 0.77 (95% CI 0.62 to 0.96), 0.88 (95% CI 0.73 to 1.07) comparing pregnant people with any initial vaccination during their pregnancy with unvaccinated pregnant people during the pre-Delta, Delta and Omicron periods, respectively (table 2). For severe COVID-19 infections, the aHRs were 0.65 (95% CI 0.47 to 0.90) and 0.79 (95% CI 0.51 to 1.21) for the same group during the Delta and Omicron periods, respectively. The model results for both coprimary outcomes for pregnant people with initial vaccinations before their pregnancy and unvaccinated pregnant people were similar.

Compared with unvaccinated pregnant people, the aHR of incident COVID-19 infections for those with booster vaccinations during their pregnancy was 0.52 (95% CI 0.47 to 0.58) during the Omicron period, and the aHR of severe COVID-19 infections was 0.35 (95% CI 0.22 to 0.54). The result of the comparison between unvaccinated pregnant people and pregnant people with booster vaccinations before their pregnancy was similar. In addition, compared with pregnant people with full initial vaccinations during their pregnancy, the aHR of incident COVID-19 infections for pregnant people with booster during their pregnancy was 0.64 (95% CI 0.58 to 0.71), and the aHR of severe COVID-19 infections was 0.35 (95% CI 0.24 to 0.50).

Cohort 2 findings: pregnant people had a lower risk of COVID-19 infections but a higher risk of severe COVID-19 infections compared with vaccinated non-pregnant people during the Delta and Omicron periods

In cohort 2, among vaccinated non-pregnant females, 159 106 (18.4%) had breakthrough COVID-19 infections during the entire study period, of whom 3717 (0.43%) developed severe COVID-19 infections (table 1).

For breakthrough COVID-19 infections, the aHRs were 0.92 (95% CI 0.61 to 1.40), 0.79 (95% CI 0.69 to 0.91), and 0.61 (95% CI 0.55 to 0.69) comparing pregnant people with initial vaccinations during their pregnancy with vaccinated non-pregnant people for the pre-Delta, Delta and Omicron periods, respectively (table 3). For the same comparison, the aHRs for severe COVID-19 infections were 3.25 (95% CI 1.76 to 5.99) and 2.71 (95% CI 1.31 to 5.60) during the Delta and Omicron periods, respectively. The adjusted aHRs of breakthrough and severe COVID-19 infections for pregnant people with initial vaccinations before their pregnancy versus vaccinated non-pregnant people were similar.

Compared with non-pregnant people with booster vaccinations, the aHRs for breakthrough infections during the Omicron period were 0.79 (95% CI 0.69 to 0.89) and 0.73 (95% CI 0.64 to 0.84) for pregnant people with booster vaccinations during and before their pregnancy, respectively. Furthermore, the aHRs for severe COVID-19 infections was 3.43 (95% CI 2.59 to 4.55) comparing non-pregnant people with booster vaccinations and pregnant people with booster vaccinations during their pregnancy.

More descriptive results on person-time at risk distribution, the timing of vaccinations and incident/breakthrough infections can be found in online supplemental figures 2–7. Full primary model results can be found in online supplemental tables 2 and 3. The sensitivity analyses limiting the cohort to only people with at least one recorded COVID-19 test or those vaccinated receiving only mRNA vaccines showed similar results as our main analysis (online supplemental tables 4 and 5). Descriptive numbers of included people stratified by pre-Delta, Delta and Omicron predominant periods can be found in online supplemental table 6.