Efficacy of primary series AZD1222 (ChAdOx1 nCoV-19) vaccination against SARS-CoV-2 variants of concern: Final analysis of a randomized, placebo-controlled, phase 1b/2 study in South African adults (COV005)

The study evaluates how efficacy of the AZD1222 (ChAdOx1 nCoV-19) COVID-19 vaccine varies against different SARS-CoV-2 variants of concern (VoC) in South African adults aged 18–65 years. Contextually, South Africa experienced sequential waves dominated by wild type (WT), Beta (B.1.351), and Delta variants. Prior analyses showed robust AZD1222 efficacy against ancestral virus and Alpha, but low efficacy against mild-to-moderate disease with Beta in South Africa. The research objective is to provide final overall and variant-specific vaccine efficacy (VE) and safety outcomes from the COV005 randomized, placebo-controlled trial, informing the durability and breadth of protection across VoCs and guiding vaccination policy.

Design and setting: COV005 is a multicenter, phase 1b/2, double-blind, randomized, placebo-controlled trial (NCT04444674; PACTR202006922165132) in healthy South African adults aged 18–65 years, enrolled June 24–November 9, 2020. Participants were randomized 1:1 to two doses of AZD1222 or saline placebo. Endpoints: Primary endpoint was efficacy against PCR-positive symptomatic COVID-19 >14 days after dose 2, and safety. Secondary endpoints included variant-specific VE and reactogenicity. Testing and case ascertainment: Nasopharyngeal swabs were tested by NAAT assays targeting N1 and N2 regions (Ct <40 for both required for positivity; inconclusive if only one target <40 and considered infected if repeat inconclusive). Human RNase P (Ct <40) confirmed specimen adequacy. Sequencing and variant assignment: Whole-genome sequencing performed at VIDA (Johannesburg) or KRISP (Durban) using cDNA synthesis (Superscript IV), amplicon protocols (QIAseq DIRECT SARS-CoV-2 Primer Panel or ARTIC V3), Illumina library prep and MiSeq sequencing. Genome assembly used Genome Detective and Coronavirus Typing Tool with quality filtering (bcftools), alignment (Nextalign), phylogeny (IQ-TREE), dating and annotation via a local Nextstrain pipeline for variant designation. A sensitivity analysis imputed variant for unclassified infections by calendar period: pre-October 2020 as WT, October 2020–April 2021 as Beta, and May–November 2021 as Delta. Follow-up and analysis population: Participants were followed from 14 days post–second dose until withdrawal, death, infection, unblinding, or censoring on November 15, 2021. Only participants who received both doses and had sufficient follow-up were included; participants were analyzed by treatment received. Statistical analysis: Incidence was calculated as cases divided by person-days at risk. VE was estimated as 1 − (incidence in vaccine group ÷ incidence in placebo group). Exact 95% confidence intervals used Clopper–Pearson methods. VE was assessed for any SARS-CoV-2 infection (including asymptomatic) and for mild/moderate COVID-19, overall and by variant (WT, Beta, Delta). Safety: Local/systemic adverse events were recorded for 28 days post each dose; serious adverse events collected throughout the study. Ethics approvals and informed consent were obtained per Declaration of Helsinki and GCP.

Participants: 2130 randomized (1,065 AZD1222; 1,065 placebo); analysis included 935 vaccine and 960 placebo recipients who received two doses and had sufficient follow-up. Median age 31 years; 82.7% aged 18–45; 42.6% smokers; 7.5% with hypertension, diabetes, or respiratory conditions. There were 104 participants living with HIV. Mean time from second dose to infection: 5.5 months (placebo) vs 4.5 months (vaccine) overall; for Delta infections, 9.3 vs 10.0 months, respectively. Efficacy against any SARS-CoV-2 infection (symptomatic and asymptomatic): - Overall: VE 38.3% (95% CI 15.1, 55.4). - WT: VE 90.6% (95% CI 35.4, 99.8). - Beta: VE 6.7% (95% CI −41.1, 38.5) — no meaningful protection. - Delta: VE 77.1% (95% CI 30.4, 94.4). Efficacy against mild/moderate COVID-19 (no severe cases observed before unblinding): - Overall: VE 45.2% (95% CI 19.7, 63.1). - WT: VE 88.5% (95% CI 17.2, 99.7). - Beta: VE 8.0% (95% CI −51.3, 44.2). - Delta: VE 75.7% (95% CI 25.6, 94.1). Safety: Reactogenicity consistent with prior reports; one vaccine recipient experienced Grade 3 fever deemed related to AZD1222, resolving within 48 hours. No severe COVID-19 cases were recorded prior to unblinding.

Final analysis demonstrates high AZD1222 efficacy against infection and mild/moderate disease caused by WT SARS-CoV-2, and substantial protection against Delta infections occurring 9–10 months post–primary series, suggesting durability of protection likely supported by anamnestic immune responses. The lower VE against Delta relative to WT may reflect waning antibodies, partial immune escape, and increased transmissibility of Delta. Importantly, no severe COVID-19 occurred prior to unblinding, which may reflect the younger, generally healthy trial population. Consistent with earlier interim results, efficacy against mild-to-moderate disease caused by the Beta variant was low, likely due to immune escape from vaccine-induced neutralizing antibodies. Nevertheless, external real-world evidence indicates AZD1222 maintains strong protection against severe outcomes (hospitalization/death) with Beta/Gamma, suggesting cellular immunity may preserve protection against severe disease despite reduced neutralization. The findings underscore that vaccine policy should not be based solely on efficacy against non-severe illness for a single variant; continuation of AZD1222 deployment during Beta likely contributed to preventing deaths during subsequent Delta waves. The study predates the emergence of Omicron; booster strategies and heterologous boosting have since been deployed, with real-world data supporting AZD1222 as a viable booster option after an AZD1222 primary series.

This final analysis of COV005 shows that a two-dose AZD1222 primary series confers high protection against WT and substantial protection against Delta infections and mild/moderate disease, with limited efficacy against mild/moderate Beta variant disease. The results support durability of protection against some variants months after vaccination and reinforce the importance of vaccination even when efficacy against non-severe disease is reduced for particular variants. Future research should assess VE against Omicron and its sublineages, evaluate durability beyond 10 months, and compare homologous versus heterologous booster strategies, including cellular immunity correlates of protection.

- Population limited to adults aged 18–65 years and largely healthy; those >65 years and individuals with uncontrolled chronic conditions were excluded, limiting generalizability and potentially explaining the absence of severe cases. - Some infections were unclassified by sequencing due to low viral loads; variant in sensitivity analyses was imputed based on calendar period, which may introduce misclassification. - Relatively small numbers of WT and Delta cases reduce precision of variant-specific VE estimates (wide CIs). - Follow-up and infection timing varied across waves; potential confounding by waning immunity over time. - Study conducted in South Africa; findings may not fully extrapolate to other settings with different variant circulation and demographics. - The analysis was censored and participants unblinded before Omicron emergence, precluding VE estimates for Omicron.