nAb responses have been proposed as a SARS-CoV-2 correlate of protection across multiple vaccine platforms (20). We have previously analyzed nAb responses against the ancestral SARS-CoV-2 virus 28 days post-second dose primary series in our phase 3 study (13). As outlined in Sobieszczyk et al. (21), data for the follow-up period were obtained prior to and during a global SARS-CoV-2 Alpha wave and Omicron had not been identified at the time of data cut-off (July 30, 2021).

To provide context for this analysis of breakthrough infections, we assessed nAbs elicited against contemporary variants of concern (VoCs) using pooled sera obtained from vaccinees (n=210 participants; n=21 pools) 28 days following their second dose of primary series AZD1222. These samples were randomly selected from participants enrolled in the immunogenicity substudy of NCT04516746 who had consented to future use of biospecimens for exploratory analyses. Demographics of the participants who contributed to the pools (Supplemental Table 1) was representative of the wider NCT04516746 study population (13). nAb geometric mean titers (GMTs) were highest against the ancestral SARS-CoV-2 pseudovirus (202.1), with modest reductions observed against Alpha (133.1), Gamma (59.5), and Delta (103.1) pseudoviruses and larger decreases observed against Beta (32.7) and Omicron BA.1 (21.7) pseudoviruses (Supplemental Figure 1).

The analyses of SARS-CoV-2 infection in this manuscript are restricted to participants in the fully vaccinated analysis set (FVS) (i.e., those who were SARS-CoV-2 seronegative at baseline and remained on the study for ≥15 days after their second dose of primary series AZD1222 or placebo without SARS-CoV-2 infection) who experienced their first symptomatic RT-PCR-confirmed SARS-CoV-2 infection and provided samples during the 28-day illness period. The overall number of FVS participants who initiated illness visits and had confirmed SARS-CoV-2 positive symptomatic illness confirmed by adjudication was low (AZD1222: n=177/17,617; [1.00%]; placebo: n=203/8,528, [2.38%]) at the time of data cut-off. The demographics of participants initiating illness visits were broadly similar between AZD1222 and placebo arms (Supplemental Table 2). Although vaccinees were younger (median age: 42 versus 46 years, AZD1222 versus placebo), a greater proportion of vaccinees had a very high or high exposure risk to COVID-19 per Occupational Safety and Health Administration categories (27.1% versus 21.2%, AZD1222 versus placebo).

Participant illness e-Diary responses illustrated that vaccinees had an overall lower incidence of COVID-19 symptoms compared with placebo recipients (Figure 1A; Supplemental Table 3). Vaccinees had shorter mean symptom durations (by ≥1 day) than placebo recipients for chills (1.6 versus 2.8 days), cough (2.2 versus 3.2), fatigue (4.7 versus 6.7), muscle aches (2.9 versus 3.9), body aches (3.0 versus 4.0), new loss of smell (3.9 versus 5.3), congestion (5.3 versus 7.1), and runny nose (4.0 versus 5.5) (Figure 1B; Supplemental Table 3). A low incidence of COVID-19 disease in participants aged >65 years limited comparisons regarding symptom incidence and burden between age groups (Supplemental Table 3).

Analyses of SARS-CoV-2 viral loads in nasopharyngeal swabs and saliva samples revealed a trend towards lower GMT in vaccinees compared with placebo recipients at all timepoints throughout the illness period (Figures 2A, B). The median duration of viral shedding in saliva samples was shortened in vaccinees compared to placebo by 3 days (Figure 2C). Among cases, with sequence data at Illness Day 1 (ILL-D1), median overall viral loads in nasopharyngeal swabs (Figure 2D) and saliva samples (Figure 2E) were lower in vaccinees versus placebo recipients with consistent trends towards lowered viral loads observed for the ancestral SARS-CoV-2 virus and the Alpha variant across both sample types. Median vaccinee SARS-CoV-2 Epsilon viral loads were higher in nasopharyngeal swabs but were lower in saliva samples compared with those observed in placebo.

Median ILL-D1 anti-spike-binding titers in vaccinees were similar to peak titers seen 14 days after dose 2 of AZD1222 (13) and median titers in vaccinees were higher than those observed in placebo recipients at all timepoints (Figure 3A). Subgroup analyses illustrated that the kinetics and magnitude of the spike-binding antibody response differed by vaccinee age: median titers were lower at ILL-D1 and higher at ILL-D14 and ILL-D28 in vaccinees aged ≥65 versus those aged 18–64 years, although numbers of participants aged ≥65 years were small (AZD1222: n=10; placebo n=16).

Median ILL-D1 nAb titers were comparable to peak titers observed 28 days after dose 2 of AZD1222 (13) and further increased throughout the illness visit period, with higher titers in vaccinees than in placebo recipients at all time points (Figure 3B). As with anti-spike-binding antibodies, subgroup analyses showed the impact of age on the kinetics and magnitude of nAb response – median titers in vaccinees aged ≥65 versus 18–64 years were lower at ILL-D1 and ILL-D14 but higher at ILL-D28. Median titers within the placebo arm peaked at ILL-D14 and were also higher in participants aged ≥65 versus 18–64 years.

Nucleocapsid antibodies were induced more slowly than spike-binding antibodies and were lower in overall magnitude, with higher median titers in placebo recipients versus vaccinees at ILL-D14 and ILL-D28 (Figure 3C). Subgroup analyses revealed median titers were higher in placebo participants aged ≥65 versus 18–64 years. GMTs were broadly similar in vaccinees aged ≥65 or 18–64 years, despite a lower median titer in vaccinees aged ≥65.

Median spike-binding titers differed with increased time since second dose of primary series vaccination (Figure 4). At ILL-D1, lower median spike-binding antibody titers were observed in vaccinees with ≥60 days since second dose primary series than those with <60 days. However, this did not prohibit participants with longer intervals between primary series and illness from mounting equivalent responses to participants with shorter intervals, as evidenced by ILL-D14 and ILL-D28 titers.

ILL-D1 nAb responses inversely correlated with SARS-CoV-2 virologic outcomes, with low-to-moderate correlations observed for nasopharyngeal viral loads (Pearson: AZD1222 –0.137; placebo –0.445) and viral shedding titers in saliva (Pearson: AZD1222 –0.469; placebo –0.398) (Figures 5A, B). ILL-D1 nAb responses in vaccinees displayed moderate correlations with duration of shedding, with weaker correlations observed in placebo recipients (Pearson: AZD1222 –0.441; placebo –0.113) (Figure 5C). As outlined in Aksyuk et al. (22), ILL-D1 spike-binding antibody responses also inversely correlated with virologic outcomes, with similar, albeit weaker, low-to-moderate correlations observed for viral shedding titers in saliva (Pearson: AZD1222 –0.436; placebo –0.251). As with ILL-D1 nAb responses, vaccinee spike-binding antibody responses displayed stronger correlations with duration of shedding than those in placebo recipients (Pearson: AZD1222 –0.323; placebo –0.067) (22).

Spike-specific CD4+ and CD8+ T-cell responses were assessed by intracellular cytokine staining (ICS) assay following stimulation of participant PBMCs with SARS-CoV-2 spike peptide pools. At ILL-D1, vaccinees had higher frequencies of spike-specific CD4+ (Figure 6A) and CD8+ (Figure 6B) T cells than placebo recipients (any response CD4+: P ≤ 0.01; any response CD8+: P ≤ 0.05). Furthermore, AZD1222 vaccination was associated with a higher proportion of responders than placebo recipients (AZD1222 versus placebo: CD4+ 14/15 [93%] versus 16/32 [50%]; CD8+ 10/15 [67%] versus 8/32 [25%]). Interestingly, AZD1222 vaccination did not impact participants de novo antiviral T-cell responses as SARS-CoV-2 nucleocapsid-specific CD4+ and CD8+ T-cell frequencies were not statistically different between both groups (Supplemental Figure 2).

Spike-specific T cells from responders were further assessed to determine the frequencies and proportions of individual cytokines produced. We found that vaccinees had a greater proportion of polyfunctional Th1 CD4+ T cells (i.e., T cells with the ability to secrete >2 cytokines from the panel) than placebo recipients (Figure 6C). Similar to CD4+ T cells, the proportion of polyfunctional spike-specific CD8+ T cells was higher among vaccinees compared to placebo recipients (Figure 6D).

At ILL-D14 spike-specific T-cell frequencies and the proportion of responders were increased in both groups (Supplemental Table 4), however, elevated responses were generally maintained in vaccinees (Figures 6E, F). Although the ancestral strain of SARS-CoV-2 accounted for nearly all breakthrough infections in this analysis, we wanted to determine whether the responders in (Figures 6A, B) were also capable of generating a T-cell response to an Omicron VoC lineage BA.1. Stimulation of vaccinee PBMCs with either ancestral SARS-CoV-2 or Omicron BA.1 spike peptide pools revealed equivalent CD4+ and CD8+ T-cell responses (Figures 6G, H) indicating that AZD1222 vaccination induces broad recognition of the SARS-CoV-2 spike protein as previously demonstrated (18). T cells from placebo participants were also able to recognize Omicron BA.1 spike peptides indicating natural infection can also induce a broad T-cell response.

Correlations between ILL-D1 T-cell and humoral responses and virologic outcomes were assessed. ILL-D1 nAb titers in vaccinees directly correlated with CD4+ (Spearman rank: 0.65; p=0.02) but not CD8+ (Spearman rank: 0.05; p=0.85) T-cell responses (Supplemental Figure 3). Among participants with matched PBMCs and virology samples, CD4+ T-cell responses inversely correlated with nasopharyngeal viral loads with strong correlations observed in vaccinees (Pearson: AZD1222 –0.881; placebo 0.796) (Supplemental Figure 4). However, trends were inconsistent between groups for durations of shedding (Pearson: AZD1222 –0.421; placebo –0.344) and shedding titers in saliva (Pearson: AZD1222 –0.681; placebo 0.126). CD8+ T-cell responses also inversely correlated with virologic outcomes, with strong correlations observed with quantitative viral loads (Pearson: AZD1222 –0.932; placebo –0.805), and the duration of shedding (Pearson: AZD1222 –0.926; placebo –0.899) in both groups (Figure 7). Similar to CD4+ T cells, moderate correlations were observed between ILL-D1 CD8+ T-cell responses and viral shedding titers in saliva (Pearson: AZD1222 –0.513; placebo –0.169) for vaccinees but not placebo recipients. However, the numbers of data points were limited in these analyses and findings should be interpreted accordingly.

As previously reported (13, 21), NCT04516746 was designed as a double-blind, placebo-controlled, phase 3 study of the safety and efficacy of 28-day primary series AZD1222 for the prevention of symptomatic COVID-19 in participants ≥18 years of age whose conditions were medically stable and who were at increased risk for SARS-CoV-2 infection. Participants were recruited from 88 sites in the United States, Chile, and Peru. The study was conducted in accordance with the principles of the Declaration of Helsinki and the International Council for Harmonization Good Clinical Practice guidelines. All participants provided written informed consent before enrollment.

Study participants were randomly assigned in a 2:1 ratio to receive two injections of AZD1222 (5×10¹⁰ viral particles), or saline placebo administered 4 weeks apart on days 1 and 29 (−3 to +7 days). Randomization was stratified according to age (≥18–64 years and ≥65 years), with a target of ≥25% participants being ≥65 years of age.

Study participants who experienced any duration of fever, shortness of breath, difficulty breathing, ≥2 days of chills, cough, fatigue, muscle aches, body aches, headache, new loss of taste, new loss of smell, sore throat, congestion, runny nose, nausea, vomiting, or diarrhea were requested to contact their study site to initiate illness visits (13, 21).

All participants with qualifying symptoms underwent an initial illness visit for confirmatory SARS-CoV-2 RT-PCR testing and provided nasopharyngeal swabs, self-collected saliva samples (US-sites only), sera, and PBMCs (select-sites only) for analysis. Participants were also trained to operate an illness e-Diary to document their symptoms (Supplemental Methods). Only participants with confirmed SARS-CoV-2 infection continued the full 28-day illness visit course, which comprised an “at home” period with self-collection of saliva samples on days 3, 5, 8, and 11, and additional site-visits with collection of nasopharyngeal swabs, saliva samples, sera, and PBMCs (day 14 illness visit only) on days 14, 21, and 28. Some saliva samples were collected at home based upon investigator preference. If a participant had multiple sets of illness visits, the first set of illness visits with positive RT-PCR test result was used for the summary.

Participants were trained by site staff on how to record their symptoms in an illness e-Diary during the day 1 illness visit. Participants who tested positive for SARS-CoV-2 continued recording their symptoms until symptom resolution or until the day 28 illness visit. Site staff monitored the health status of participants via Illness e-Diary responses and called participants as needed based on their responses.

Not all participants contributed data at every illness visit. Of the 380 participants (AZD1222, n=177; placebo, n=203) with symptomatic infection, nasopharyngeal swabs, and saliva samples from 201 (AZD1222, n=97; placebo, n=104) and 222 (AZD1222, n=89; placebo, n=133) participants, respectively, were available for virologic assessments, sera from 323 (AZD1222, n=151; placebo, n=172) participants were analyzed for anti-SARS-CoV-2 serologic responses, and PBMCs from 47 (AZD1222, n=15; placebo, n=32) participants were analyzed by ICS to evaluate SARS-CoV-2 spike-specific CD4+ and CD8+ T-cell responses.

As previously described (13, 21), SARS-CoV-2 viral load was assessed in nasopharyngeal swabs and saliva samples using the TaqPath™ SARS-CoV-2 RT-PCR Assay (ThermoFisher Scientific, Waltham, MA, USA). Nasopharyngeal swabs were analyzed by Labcorp, Indianapolis, IN, USA, while saliva samples were analyzed by Infinity Biologix, Rutgers, NJ, USA. SARS-CoV-2 genomic assessments are detailed in full in Supplemental Methods. Assessments of viral titer and sequencing were only performed on participants from whom sample was available after the completion of a central RT-PCR assay, thus limiting the availability for results in all participants.

Serum anti-SARS-CoV-2 spike-binding and nucleocapsid IgG antibody titers were tested at PPD^® in a validated multiplex electrochemiluminescence serology assay using the MSD V-PLEX^® SARS-CoV-2 Panel 2 (IgG) as outlined in (79). nAbs were assessed in a validated lentivirus-based SARS-CoV-2 phenosense pseudovirus assay (Monogram Biosciences, South San Francisco, CA, USA) as described previously (13). nAbs titers are reported as the reciprocal of the serum dilution conferring ID₅₀ of pseudovirus infection.

Formulae to enable conversion from arbitrary units per milliliter (AU)/mL to the WHO international standard (National Institute for Biological Standards and Control [NIBSC] 20/136) binding units (BAU/mL) and ID₅₀ to the WHO international standard (NIBSC 20/136) International units (IU/mL) are detailed in Supplemental Methods.

An ICS assay was used to evaluate T-cell responses, as previously described (18). Please refer to Supplemental Methods for full details on T cell stimulation with SARS-CoV-2 spike peptides, ICS antibody staining, and the flow cytometry gating strategy used in these analyses.

The analyses presented in this manuscript are restricted to baseline-seronegative participants with PCR-confirmed SARS-CoV-2 infection ≥15 days after dose 2 of AZD1222 or placebo per the protocol definition of breakthrough infection (13). Definitions of the study populations (i.e., the FVS and immunogenicity analysis set [IAS]) used to compile the participant demography (Supplemental Table 2) and participants with illness visits e-Diary data (Figure 1; Supplemental Table 3) tables are included in Supplemental Material. For ethical reasons, study participants could be unblinded and receive non-study COVID-19 vaccinations once available through emergency-use authorizations. The censoring implications of allowing participants to receive non-study COVID-19 vaccinations are detailed in Supplemental Methods GMTs were calculated and summarized at each illness visit for viral load assessments in nasopharyngeal swabs and saliva samples, and for anti-SARS-CoV-2 spike-binding, neutralizing and nucleocapsid antibodies in AZD1222 and placebo groups. SARS-CoV-2 spike-binding, neutralizing, and nucleocapsid antibodies were also assessed by participant age and time since second dose of primary series. GMT endpoints were analyzed on the natural log scale by separate ANOVA models including treatment and age as categorical covariates. On the log scale, the models were used to estimate a mean response for the vaccine and control groups and the difference (vaccine – control), with corresponding 95% confidence limits. Descriptive statistics for GMTs included the number of participants, geometric mean, 95% CI, minimum, and maximum. A GMT was calculated as the antilogarithm of Σ(log base 2-transformed titer/n), i.e., as the antilogarithm transformation of the mean of the log-transformed titer, where n is the number of participants with titer information. The 95% CI was calculated as the anti-logarithm transformation of the upper and lower limits for a two-sided CI for the mean of the log-transformed titers.

SAS 9.4 procedure SGPANEL was used to create the scatter plots for the correlative analyses. The REG statement generated the fitted regression line along with confidence limit intervals (CLI) and confidence limit for the mean (CLM) options to create the prediction limits and confidence limits respectively.

Informal comparisons between groups were done by post-hoc two-tailed Mann-Whitney tests and categorized as: not significant (NS), p>0.05; *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

The protocol and amendments for this trial (ClinicalTrials.gov number, NCT04516746) were approved by the ethics committee or institutional review board at each center, and the trial was conducted in compliance with the principles of the Declaration of Helsinki and the International Council for Harmonization Good Clinical Practice guidelines. Prior to enrolment all participants provided informed consent.