Association Between 3 Doses of mRNA COVID-19 Vaccine and Symptomatic Infection Caused by the SARS-CoV-2 Omicron and Delta Variants

Key Points

Question  What is the association between 3 doses of mRNA COVID-19 vaccine and symptomatic SARS-CoV-2 infection with the Omicron and Delta variants?

Findings  In this test-negative case-control analysis that included 70 155 tests from symptomatic adults, the likelihood of vaccination with 3 mRNA vaccine doses (vs unvaccinated) was significantly lower among both Omicron (odds ratio, 0.33) and Delta (odds ratio, 0.065) cases than SARS-CoV-2–negative controls; a similar pattern was observed with 3 vaccine doses vs 2 doses (Omicron odds ratio, 0.34; Delta odds ratio, 0.16).

Meaning  These findings suggest that vaccination with 3 doses of mRNA COVID-19 vaccine, compared with being unvaccinated and with receipt of 2 doses, was associated with protection against both the Omicron and Delta variants, although higher odds ratios for the association with Omicron infection suggest less protection for Omicron than for Delta.

Importance  Assessing COVID-19 vaccine performance against the rapidly spreading SARS-CoV-2 Omicron variant is critical to inform public health guidance.

Objective  To estimate the association between receipt of 3 doses of Pfizer-BioNTech BNT162b2 or Moderna mRNA-1273 vaccine and symptomatic SARS-CoV-2 infection, stratified by variant (Omicron and Delta).

Design, Setting, and Participants  A test-negative case-control analysis among adults 18 years or older with COVID-like illness tested December 10, 2021, through January 1, 2022, by a national pharmacy-based testing program (4666 COVID-19 testing sites across 49 US states).

Exposures  Three doses of mRNA COVID-19 vaccine (third dose ≥14 days before test and ≥6 months after second dose) vs unvaccinated and vs 2 doses 6 months or more before test (ie, eligible for a booster dose).

Main Outcomes and Measures  Association between symptomatic SARS-CoV-2 infection (stratified by Omicron or Delta variants defined using S-gene target failure) and vaccination (3 doses vs unvaccinated and 3 doses vs 2 doses). Associations were measured with multivariable multinomial regression. Among cases, a secondary outcome was median cycle threshold values (inversely proportional to the amount of target nucleic acid present) for 3 viral genes, stratified by variant and vaccination status.

Results  Overall, 23 391 cases (13 098 Omicron; 10 293 Delta) and 46 764 controls were included (mean age, 40.3 [SD, 15.6] years; 42 050 [60.1%] women). Prior receipt of 3 mRNA vaccine doses was reported for 18.6% (n = 2441) of Omicron cases, 6.6% (n = 679) of Delta cases, and 39.7% (n = 18 587) of controls; prior receipt of 2 mRNA vaccine doses was reported for 55.3% (n = 7245), 44.4% (n = 4570), and 41.6% (n = 19 456), respectively; and being unvaccinated was reported for 26.0% (n = 3412), 49.0% (n = 5044), and 18.6% (n = 8721), respectively. The adjusted odds ratio for 3 doses vs unvaccinated was 0.33 (95% CI, 0.31-0.35) for Omicron and 0.065 (95% CI, 0.059-0.071) for Delta; for 3 vaccine doses vs 2 doses the adjusted odds ratio was 0.34 (95% CI, 0.32-0.36) for Omicron and 0.16 (95% CI, 0.14-0.17) for Delta. Median cycle threshold values were significantly higher in cases with 3 doses vs 2 doses for both Omicron and Delta (Omicron N gene: 19.35 vs 18.52; Omicron ORF1ab gene: 19.25 vs 18.40; Delta N gene: 19.07 vs 17.52; Delta ORF1ab gene: 18.70 vs 17.28; Delta S gene: 23.62 vs 20.24).

Conclusions and Relevance  Among individuals seeking testing for COVID-like illness in the US in December 2021, receipt of 3 doses of mRNA COVID-19 vaccine (compared with unvaccinated and with receipt of 2 doses) was less likely among cases with symptomatic SARS-CoV-2 infection compared with test-negative controls. These findings suggest that receipt of 3 doses of mRNA vaccine, relative to being unvaccinated and to receipt of 2 doses, was associated with protection against both the Omicron and Delta variants, although the higher odds ratios for Omicron suggest less protection for Omicron than for Delta.

On November 24, 2021, health authorities in South Africa reported the emergence of a new SARS-CoV-2 variant, B.1.1.529 (Omicron).¹ Omicron has spread rapidly, and as of January 6, 2022, was identified in 149 countries across all 6 World Health Organization regions.² Omicron was first detected in the US on December 1, 2021, and by January 1, 2022, was estimated to be responsible for 95% of sequenced new cases.^3,4

Sequencing of early Omicron strains documented more than 30 mutations in the spike protein, including in the receptor binding domain.^5,6 These mutations, combined with observed exponential growth in case counts, even in settings with substantial rates of COVID-19 vaccination or previous SARS-CoV-2 infection, raised concerns about potential for increased transmissibility and immune escape.^2,7-10 There is an urgent need to understand the protection provided by current vaccination regimens against Omicron, including any additional protection derived from booster doses.

In this analysis, a subset of data from the national Increasing Community Access to Testing (ICATT) platform was used to estimate the association of receipt of 3 doses of a mRNA COVID-19 vaccine (vs unvaccinated and vs 2 doses) with symptomatic infection with the Omicron and Delta variants, using an internally validated genetic proxy for variant identification.

The human subjects advisor for the Centers for Disease Control and Prevention (CDC) National Center for Immunization and Respiratory Diseases determined that this analysis met the requirements for public health surveillance as outlined in 45 CFR §46.102(l)(2). Because data were collected during routine operational procedures, this secondary data analysis did not require informed consent and was conducted consistent with applicable federal law and CDC policy.

Data from the ICATT platform¹¹—a Department of Health and Human Services (HHS) partnership facilitating no-cost, drive-through SARS-CoV-2 testing at pharmacies across all 50 states, the District of Columbia, and Puerto Rico—were analyzed. Testing sites were selected by HHS to prioritize access in racially and ethnically diverse communities and areas with moderate-to-high social vulnerability. Data for this analysis were limited to tests occurring between December 10, 2021, and January 1, 2022, at testing sites that collected booster vaccination history and sent specimens to a single laboratory chain (Aegis Sciences Corp) for processing. The laboratory used the TaqPath COVID-19 Combo Kit (Thermo Fisher Scientific), which identifies SARS-CoV-2 infections by detecting 3 targets from the viral ORF1ab, S, and N gene regions. The laboratory reported overall test results and cycle threshold (Ct) values for each of these targets for SARS-CoV-2–positive specimens.

Individuals registered online for testing at drive-through sites where nasal swabs were collected. During registration, individuals self-reported symptom status (asymptomatic or symptomatic with ≥1 COVID-like symptom), race, ethnicity, sex, age, state of residence, history of prior SARS-CoV-2 infection, and underlying conditions. Fixed categories were used to capture data for symptoms, race, ethnicity, and underlying chronic conditions including the presence of an immunocompromising condition (defined in the questionnaire as “such as from immunocompromising medications, solid organ or blood stem cell transplant, HIV, or other immunocompromising conditions”). Race and ethnicity were collected as required data elements under HHS COVID-19 laboratory reporting requirements.¹² The testing program geocoded testing sites to identify their census tract Social Vulnerability Index (SVI) score.¹³

Patients also self-reported COVID-19 vaccination status, including the number of doses (up to 4), product, and month and year of receipt for each dose. For doses received in the same month or the month prior to testing, an additional question was asked to specify whether the dose was received 14 days or more before testing. Vaccination reporting was not mandatory, and information was not verified. Data were reported to HHS with an estimated 3-day lag and were deidentified to remove any personally identifying information.

A retrospective test-negative case-control analysis was conducted on samples collected from December 10, 2021, to January 1, 2022, from adults 18 years or older with symptomatic COVID-like illness. The test-negative design is a commonly used observational method for evaluating the performance of vaccines in which participants are enrolled based on a clinical case definition, tested for the vaccine-preventable outcome of interest, and classified as cases or controls based on that testing; the odds of prior vaccination among cases and controls are compared as an estimate of the association between vaccination and the outcome.^14,15 A strength of the test-negative design is that all participants seek care (or testing) for a common clinical case definition, which can help reduce bias resulting from confounding by differential care-seeking behavior.^16,17

The unit of analysis for this study was tests; positive results were classified as cases, and negative results as controls. The short study period and restriction to symptomatic individuals limited the probability of individuals contributing more than 1 test.

Because protection from mRNA COVID-19 vaccines is substantially lower in immunocompromised individuals¹⁸ and the recommended vaccine dosing regimen is different from that of immunocompetent individuals,¹⁹ tests from individuals reporting an immunocompromising condition were excluded. Tests from persons reporting a positive COVID-19 test result within the previous 90 days were excluded to reduce potential misclassification. Tests with unknown vaccination status or incomplete vaccination data (ie, missing vaccination dates or products), from individuals reporting prior receipt of 1 or 4 mRNA vaccine doses or of non-mRNA COVID-19 vaccines, or from persons with improbable ages (defined as >100 years) were also excluded.

Additionally, among tests with positive results, Ct values were described by variant and by vaccination status (3 doses, 2 doses, and unvaccinated). Ct values reflect the number of cycles during polymerase chain reaction amplification needed to detect viral genetic material and are inversely proportional to the amount of target nucleic acid in the tested sample.²⁰ Ct values were examined to better understand the relative amounts of genetic material present in positive samples by variant and vaccination status.

The exposure of interest was self-report of any 3 doses of BNT162b2 (Pfizer-BioNTech) or mRNA-1273 (Moderna) vaccine (including mixed-product regimens) vs unvaccinated and vs any 2 doses of BNT162b2 or mRNA-1273 vaccine. For individuals reporting 2 vaccine doses, tests were excluded if the second dose was received less than 6 months prior to test date to ensure eligibility for a booster dose. For those reporting 3 doses, tests were excluded if the interval between second and third doses was less than 6 months, as per recommendations during the analysis period for booster doses among immunocompetent individuals,¹⁹ or if the most recent dose was received less than 14 days before testing. Another exposure examined was any 2 doses of BNT162b2 or mRNA-1273 vaccine, with the second dose received 14 days or more before testing vs unvaccinated; assessment of this exposure did not limit to those eligible for a booster dose (ie, did not limit to those with 6 or more months elapsed between the second dose and date of testing).

The primary outcome was symptomatic SARS-CoV-2 infection with the Omicron or Delta variant. For this analysis, an Omicron case was defined as presence of S-gene target failure (SGTF) in the test sample and a Delta case as absence of SGTF in the test sample. All samples that were determined by the processing laboratory to be SARS-CoV-2 positive had Ct values for at least 2 of the N, ORF1ab, and S genes. SARS-CoV-2–positive samples were considered to have SGTF if they had Ct values for the N and ORF1ab genes but not for the S gene; otherwise, samples were considered not to have SGTF.

SGTF may serve as a proxy for the presence of the Omicron variant in samples tested with the TaqPath COVID-19 Combo Kit assay because of the presence of deletions in the S-gene region for Omicron that are not present in Delta^2,21; the deletions lead to S-gene–negative results in Omicron lineages BA.1 and B.1.1.529 but not the majority of Delta samples. While levels of non-Delta circulating variants other than Omicron remain low, samples with SGTF may be presumed to be Omicron.²² At the time of this analysis 99.9% of sequenced samples in the US prior to the emergence of Omicron were Delta.⁴ To validate the use of SGTF as a proxy for Omicron in the ICATT data, the frequency of SGTF was examined in a randomly selected subset of tests with positive results that were sequenced during the same period as the analysis. The sequenced subset was drawn from the complete database of test results from the laboratory, including tests not eligible for the main analysis. Sequencing data were not available for most cases included in the main analysis, which is why SGTF was used as a proxy. The sensitivity of SGTF for detecting Omicron (B.1.1.529 or BA.1 lineage) was 83.4% and the specificity was 99.2% (eTable 1 in the Supplement).

Secondary outcome measures included Ct values for the N and ORF1ab genes among Omicron and Delta cases and S gene among Delta cases.

The association between symptomatic infection with the Omicron or Delta variants and vaccination was estimated by comparing the odds of prior 3-dose vaccination vs unvaccinated and the odds of prior 3-dose vaccination vs 2-dose vaccination in cases vs controls using multivariable multinomial logistic regression. The odds ratio (OR) for 3 doses vs unvaccinated was used as an estimate of 3-dose vaccine effectiveness (effectiveness = [1 – OR] × 100%), with lower ORs suggesting more protection. The OR for 3 doses vs 2 doses was used as an estimate of relative vaccine effectiveness, reflecting additional protection from a booster dose relative to 2 doses. ORs were estimated for any combination of mRNA vaccine and separately for BNT162b2 and mRNA-1273.

Models included the number of days between the start of the analysis period and test date (as a continuous variable), age group, sex, race, ethnicity, testing site HHS region, testing site census tract SVI (dichotomized as 0 to <0.5 and ≥0.5-1), and number of underlying chronic conditions (0, 1, or ≥2) as covariates to adjust for potential confounding bias. Unknown race and ethnicity were coded as categories of their respective variables instead of null values to retain these records in regression models. Data with missing values for other model covariates (specifically, sex and SVI) were coded as null values and therefore dropped from adjusted regression models.

Two-sided 95% CIs were calculated for each reported OR, with 95% CIs that excluded 1 considered statistically significant. Two-sided P values for the association between vaccination and symptomatic infection with Omicron compared with Delta were corrected for false-discovery rate (FDR) using the Benjamini-Hochberg method (to account for type I error due to multiple comparisons), and results with Q values (ie, P values adjusted for the FDR) less than .001 were considered statistically significant.

To aid in the interpretation of associations of 3 vaccine doses vs unvaccinated and vs 2 vaccine doses, a secondary analysis was performed examining the association between 2 doses vs unvaccinated by time since receipt of second dose. The association between infection and 2 mRNA vaccine doses vs unvaccinated was examined separately for each product-variant combination by logistic regression incorporating the same covariates as above as well as month (0-11) since second dose using a 2-knot spline at months 3.5 and 7.5.

For comparison of Ct values among cases by variant and exposure status (3 doses vs unvaccinated, 3 doses vs 2 doses, and 2 doses vs unvaccinated), the 2-sided Mann-Whitney U test was used to identify significant differences in median Ct values. Correction for FDR was applied for each comparison of exposure status, and results with Q < .001 were considered statistically significant.

Statistical analyses were performed in RStudio and R, version 4.0.3 (R Foundation). Multinomial logistic regression was performed using the nnet R package, version 7.3-16.

A total of 70 155 tests from 4666 sites on samples collected between December 10 and January 1 across 49 states met inclusion criteria (Figure 1), including 23 391 cases (13 098 Omicron; 10 293 Delta) and 46 764 controls (Table 1) (mean age, 40.3 [SD, 15.6] years; 42 050 [60.1%] women). Included tests were most frequently performed on persons aged 25 to 34 years (30.4%), followed by those aged 35 to 44 years (19.3%), and on persons who reported being of White race (75.5%). More than one-third of tests (36.4%) were from people with reported underlying health conditions, with high blood pressure most common, followed by overweight. Compared with controls, cases were more frequently tests from persons aged 25 to 34 years (Omicron 35.1% and Delta 31.0% vs controls 28.9%), of Black/African American race (Omicron 24.4% and Delta 14.9% vs controls 11.9%), and of Hispanic/Latino ethnicity (Omicron 22.5% and Delta 17.7% vs controls 17.4%) (Table 1).

For vaccination history, the most common combination of 2 doses was BNT162b2/BNT162b2 (63.4% of 2-dose regimens), followed by mRNA-1273/mRNA-1273 (36.4%); the most common 3-dose combination was BNT162b2/BNT162b2/BNT162b2 (57.5% of 3-dose regimens) (Table 1). Among individuals receiving 2 doses only, the median time between the second dose and test date was 8 months. Among those receiving 3 doses, the median time between the second dose and test date was 8 months, between the second and third dose was 7 months, and between the third dose and test date was 1 month. Prior receipt of 3 vaccine doses was reported for 18.6% (2441/13 098) of Omicron cases, 6.6% (679/10 293) of Delta cases, and 39.7% (18 587/46 764) of controls (Table 1).

Among Omicron cases and controls, the adjusted OR for prior receipt of 3 mRNA vaccine doses vs unvaccinated was 0.33 (95% CI, 0.31-0.35); among Delta cases and controls, the adjusted OR was 0.065 (95% CI, 0.059-0.071; Q < .001 for comparison of ORs for Omicron and Delta) (Table 2). When models were stratified by mRNA product, the adjusted ORs for Omicron were 0.35 (95% CI, 0.32-0.38) for 3 doses of BNT162b2 vs unvaccinated and 0.28 (95% CI, 0.26-0.31) for 3 doses of mRNA-1273 vs unvaccinated. For Delta, the adjusted ORs were 0.077 (95% CI, 0.070-0.086) for 3 doses of BNT162b2 vs unvaccinated and 0.045 (95% CI, 0.038-0.053) for 3 doses of mRNA-1273 vs unvaccinated. Q values for all comparisons (Omicron vs Delta) of product-specific ORs were less than .001 (Table 2). An adjusted OR less than 1 indicates that relatively fewer test-positive cases had prior receipt of 3 doses (vs unvaccinated), with values closer to 0 representing a stronger magnitude of association.

Among Omicron cases and controls, the adjusted OR for prior receipt of 3 mRNA vaccine doses vs 2 doses was 0.34 (95% CI, 0.32-0.36); among Delta cases and controls, the adjusted OR was 0.16 (95% CI, 0.14-0.17; Q < .001) (Table 2). When models were stratified by mRNA product, the adjusted ORs for Omicron were 0.35 (95% CI, 0.32-0.37) for 3 doses of BNT162b2 vs 2 doses and 0.31 (95% CI, 0.28-0.34) for 3 doses of mRNA-1273 vs 2 doses. For Delta, the adjusted ORs were 0.17 (95% CI, 0.16-0.19) for 3 doses of BNT162b2 vs 2 doses and 0.13 (95% CI, 0.11-0.15) for 3 doses of mRNA-1273 vs 2 doses. Q values for all comparisons (Omicron vs Delta) of product-specific ORs were less than .001 (Table 2). An adjusted OR less than 1 indicates that relatively fewer test-positive cases had prior receipt of 3 doses (vs 2 doses), with values closer to 0 representing a stronger magnitude of association.

Among Omicron cases and controls, the adjusted OR for prior receipt of any 2 mRNA vaccine doses vs unvaccinated was lowest soon after the second dose and generally increased over time since vaccination (Figure 2). The upper bound of the 95% CI was consistently greater than 1 starting at 3 months after second dose for BNT162b2 and at 6 months after second dose for mRNA-1273. For Delta, the OR for any 2 doses vs unvaccinated was also lowest soon after receipt of second dose; however, the ORs remained less than 0.46 for BNT162b2 and less than 0.40 for mRNA-1273, with 95% CIs that did not include 1 (Figure 2).

Among Omicron cases, median Ct values were significantly higher in samples from persons reporting 3 mRNA vaccine doses vs unvaccinated for the ORF1ab gene (19.25 vs 18.58; Q < .001) but not for the N gene (19.35 vs 18.71; Q = .002) (eTable 2 in the Supplement). Among Delta cases, median Ct values of the N, ORF1ab, and S genes were significantly higher in samples from persons receiving 3 doses vs unvaccinated (N gene: 19.07 vs 18.28; ORF1ab gene: 18.70 vs 17.84; S gene: 23.62 vs 19.58; Q < .001 for all comparisons) (Figure 3; eTable 2 in the Supplement). In comparing median Ct values in samples from persons reporting 3 mRNA vaccine doses vs 2 doses, all were significantly higher (Omicron N gene, 19.35 vs 18.52; Omicron ORF1ab gene, 19.25 vs 18.40; Delta N gene, 19.07 vs 17.52; Delta ORF1ab gene, 18.70 vs 17.28; Delta S gene, 23.62 vs 20.24; Q < .001 for all comparisons) (Figure 3; eTable 2 in the Supplement).

In this analysis of SARS-CoV-2 tests performed at sites across the US during a 23-day period when incidence of Omicron was rapidly increasing, vaccination with a third dose of an mRNA COVID-19 vaccine was significantly less common among individuals infected with either the Omicron or Delta variants compared with uninfected individuals. Because of the timing of booster recommendations in the US, most booster recipients had recent vaccination (median of 1 month since booster).

The magnitude of the association between vaccination and infection depended on the referent group and variant. For 3 doses vs unvaccinated, the ORs corresponded to an estimated effectiveness (1 – OR) of 67.3% (95% CI, 65.0%-69.4%) for Omicron and 93.5% (95% CI, 92.9%-94.1%) for Delta. For 3 doses vs 2 doses, the ORs corresponded to an estimated relative effectiveness of 66.3% (95% CI, 64.3%-68.1%) for Omicron and 84.5% (95% CI, 83.1%-85.7%) for Delta. For Omicron, the similarity between ORs for 3 doses using the unvaccinated referent group and the 2-dose referent group is consistent with the attenuation of the OR for 2 doses vs unvaccinated with time since second dose, which reflected no significant association by 6 months after second dose for both products. For Delta, the association between infection and 2 doses vs unvaccinated also attenuated over time since second dose, which is consistent with previous reports^23-26; however, the ORs were statistically significant even up to 11 months after the second dose.

Although these findings provide evidence supporting that 3-dose schedules are protective and that booster doses are more protective than primary series alone, the significantly higher OR for Omicron suggests that booster doses are less protective against Omicron than against Delta. These results are consistent with in-vitro neutralization assays that suggested the potential for immune evasion with Omicron.^27-30 They also highlight that, in the setting of Omicron, higher booster coverage rates may be needed to achieve the same public health benefit as during Delta predominance. Additionally, nonpharmaceutical interventions may provide an important adjunct to slow the spread of Omicron.

Among both Omicron and Delta variant cases, Ct values were generally higher (reflecting less genetic material detected) among those with 3 vaccine doses compared with unvaccinated or to 2 doses; with 1 exception (N gene for Omicron, 3 doses vs unvaccinated), all comparisons were statistically significant. Ct values are not a direct measure of viral load or infectiousness and can vary for a range of reasons including timing of sample collection relative to infection onset, specimen transport times, and laboratory assays and conditions. However, they have previously been used as a crude indicator of transmission potential, with higher values representing decreased likelihood of a case being infectious.^31-37 In this analysis, the Ct values were based on 2 targets from a single assay, performed at a single laboratory chain, increasing their comparability. The significantly higher Ct values found in individuals reporting receipt of 3 doses vs unvaccinated or 2 doses, for both the Omicron and Delta variants, may suggest decreased infectiousness in those receiving an mRNA booster dose. However, caution should be taken in interpreting these differences between groups as epidemiologically meaningful because all differences were less than 1 unit on the log scale.

This study has several limitations. First, vaccination status and symptoms were based on patient self-reported data, potentially leading to misclassification. Second, because the testing data do not include identifiers, tests rather than persons were used as the unit of analysis, and individuals may have been included more than once. However, the analysis was restricted to symptomatic individuals to reduce inclusion of individuals serially testing for reasons other than symptomatic disease, and the short study period (23 days) reduces the probability of individuals contributing multiple test results.

Third, individuals remaining unvaccinated or unboosted may differ from individuals with 3 doses in ways that cannot be adjusted for with the variables in this data set. Fourth, some factors that could potentially be associated with both vaccination and risk of infection and thus confound the observed associations (eg, masking and social distancing) were not measured. Fifth, US adults were recommended to receive booster doses at different dates depending on age, underlying conditions, and occupation; therefore, some subgroups may have had greater access to boosters before broader recommendations were made.

Sixth, sequencing data were limited for tests included in this analysis so SGTF was used as a proxy for Omicron infection; however, sensitivity of 83.4% and specificity of 99.2% in internal validation suggest that samples classified as Omicron by SGTF were almost always Omicron. Although some Omicron samples may have been misclassified as Delta, this would not affect the association between Omicron and vaccination status and would bias the association of Delta and vaccination toward that of Omicron such that the reported differences between Omicron and Delta are conservative.

Seventh, the analysis of the association between symptomatic infection and 3 vs 2 doses did not directly account for waning of the primary series, although all tests were performed at least 6 months after receipt of a primary series, at which point protection from 2 doses was quite reduced, particularly for the Omicron variant. Eighth, associations between infection and vaccination are not constant and will likely continue to change with time since last dose.

Among individuals seeking testing for COVID-like illness in the US in December 2021, receipt of 3 doses of mRNA COVID-19 vaccine (compared with unvaccinated and with receipt of 2 doses) was less likely among cases with symptomatic SARS-CoV-2 infection compared with test-negative controls. These findings suggest that receipt of 3 doses of mRNA vaccine, relative to being unvaccinated and to receipt of 2 doses, was associated with protection against both the Omicron and Delta variants, although the higher odds ratios for Omicron suggest less protection for Omicron than for Delta.

Corresponding Author: Emma K. Accorsi, PhD, COVID-19 Response, US Centers for Disease Control and Prevention, 1600 Clifton Rd Mailstop H24-6, Atlanta, GA 30329 (vgi0@cdc.gov).

Accepted for Publication: January 13, 2022.

Published Online: January 21, 2022. doi:10.1001/jama.2022.0470

Author Contributions: Drs Accorsi and Britton had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Accorsi and Britton contributed equally as co–first authors; Drs Verani and Schrag contributed equally as co–senior authors.

Concept and design: Accorsi, Britton, Fleming-Dutra, Shang, Miller, Schrag, Verani.

Acquisition, analysis, or interpretation of data: Accorsi, Britton, Fleming-Dutra, Smith, Shang, Derado, Schrag, Verani.

Drafting of the manuscript: Accorsi, Britton, Schrag, Verani.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Accorsi, Britton, Shang, Derado.

Obtained funding: Miller.

Administrative, technical, or material support: Britton, Smith, Miller, Verani.

Supervision: Miller, Schrag, Verani.

Conflict of Interest Disclosures: None reported.

Funding/Support: Funding for the Increasing Community Access to Testing platform was provided by the US Department of Health and Human Services. Funding for this analysis was provided by the Centers for Disease Control and Prevention (CDC).

Role of the Funder/Sponsor: The CDC was involved in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and approval of the manuscript; and decision to submit the manuscript for publication. CDC controlled publication decisions.

Disclaimer: The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the CDC.