Dynamics of competing SARS-CoV-2 variants during the Omicron epidemic in England

Since late 2020, SARS-CoV-2 variants of concern (VOCs) have emerged regularly, altering national and global epidemic dynamics. On 24 November 2021, lineage B.1.1.529 (Omicron) was designated and later classified as a VOC. Although Omicron has been associated with less severe disease, it carries numerous mutations, including in the receptor binding domain, enabling escape from a majority of pre-existing neutralising antibodies. Rapid growth in South Africa indicated higher transmissibility relative to earlier VOCs, attributed in part to immune evasion, reduced effectiveness of existing vaccines against infection, and increased risk of reinfection. This study uses unbiased community prevalence data from REACT-1 to quantify the dynamics, growth advantage, and replacement of Delta by Omicron (and subsequent BA.2 displacement of BA.1/BA.1.1) in England from September 2021 to March 2022.

The paper situates Omicron within prior VOC dynamics, citing evidence of immune evasion and reduced vaccine effectiveness against Omicron infection, as well as increased reinfection risk. Early assessments in South Africa and other contexts documented Omicron’s rapid spread, likely driven by immune escape. Prior REACT-1 publications focused on overall prevalence trends and variant transmissibility (e.g., Alpha, Delta). The authors note that routine testing data can be biased and emphasize the value of random community sampling for accurate prevalence estimation. References also discuss generation intervals, surveillance challenges, and analogies to influenza’s antigenic drift leading to recurrent waves.

Design and data sources: The study analyzes REACT-1 rounds 14–18 (9 September 2021–1 March 2022). Each round randomly samples England’s population at the lower-tier local authority level; participants self-collect throat and nose swabs for RT-PCR. Positivity is defined by N-gene Ct <37 or both N- and E-genes detected. Rim weighting adjusts for age, sex, deprivation, LTLAs and sampling method. Sequencing: Samples with N-gene Ct <34 and sufficient volume underwent next-generation sequencing using ARTIC protocols; libraries sequenced on Illumina platforms; analysis with a standardized bioinformatics pipeline; lineage assignment via Pangolin (database version 2022-02-28). Low-quality sequences and certain contextual duplicates were excluded. Statistical modeling: A mixed-effects Bayesian P-spline model estimated daily prevalence of Delta and Omicron, and later BA.2 vs non-BA.2 Omicron in rounds 17–18 (assuming >99% Omicron). Time was partitioned into ~6-day knots; prevalence on the logit scale was modeled as a sum of B-spline basis functions with a second-order random-walk prior on coefficients, sharing smoothness across lineages and allowing month-level random effects on first derivative changes. The total modeled prevalence was linked to daily weighted tests and positives via a binomial likelihood; the Omicron proportion was fit to daily lineage counts via binomial likelihood. Models were fit overall, by region, and by age groups using STAN’s No-U-Turn Sampler. Reproduction number and growth rates: Instantaneous growth rates for each lineage were derived from modeled prevalence. Rolling two-week average Rt was estimated assuming gamma-distributed generation times: for Omicron/non-Delta lineages (including BA.2 vs non-BA.2), rate=0.27, shape=0.89; for Delta, rate=0.48, shape=2.27. Constant growth-rate models (Bayesian logistic and multinomial logistic regressions) estimated daily growth in log-odds for Omicron vs Delta and BA.2/BA.1.1 vs BA.1. Phylogenetics and import/export: Maximum-likelihood phylogenies were inferred (IQ-TREE with HKY model); time-resolved trees generated with TreeTime using a relaxed clock (0.00085 substitutions/site/year). A discrete-trait model over England’s regions estimated migration patterns and pairwise genomic distances. GISAID comparisons identified close global matches within one week and ≤1 SNP to infer potential importations/exports after deduplication. Mobility and symptoms: Apple mobility indices (driving, walking, transit) were summarized as 7-day moving averages. Symptom proportions were compared between lineages using Wilson CIs and logistic regression, with sensitivity analyses adjusting for study round and N-gene Ct. Ethics: REACT-1 received ethical approval (IRAS ID: 283747); informed consent obtained.

- Rapid replacement of Delta by Omicron: Omicron prevalence in REACT-1 was 0.11% (0.07%, 0.16%) by 7 Dec 2021 and surpassed Delta during December. Delta prevalence fell below 0.1% by 3 Jan 2022. - Peak Omicron prevalence: Maximum prevalence reached 6.96% (5.34%, 10.61%) on 30 Dec 2021 (with national peak timing varying across regions). - Growth advantage dynamics: The average daily growth rate in the log-odds of Omicron vs Delta was 0.21 (0.20, 0.23) from 23 Nov 2021 to 14 Mar 2022, declining from 0.37 (0.20, 0.49) on 3 Dec to 0.11 (0.03, 0.17) on 5 Jan, consistent with an Omicron generation time approximately 28% shorter than Delta and early spread in younger, more socially active groups. - Delta Rt decline: Delta’s Rt halved from 1.00 (0.91, 1.10) on 9 Dec to 0.35 (0.26, 0.46) by 30 Dec 2021, consistent with susceptible depletion due to Omicron’s spread and behavior change (mobility declines). - Regional heterogeneity: Omicron prevalence rose rapidly in all regions. Maximum recorded prevalence was highest in the North East at 7.37% (6.42%, 9.97%) and lowest in the East of England at 3.95% (3.83%, 5.00%). London peaked earlier at 6.45% (6.15%, 10.27%) on 29 Dec 2021. - Age patterns: Prevalence peaked at 10.74% (8.52%, 14.74%) on 28 Jan 2022 in 5–17-year-olds, higher and later than other groups. For ages 18–34, the maximum was 7.65% (6.08%, 12.63%) on 1 Jan 2022. Ages ≥55 had lower peak prevalence at 3.67% (3.25%, 4.88%) on 7 Jan, but Rt in this group was 1.14 (0.97, 1.33) on 1 Mar. - BA.2 vs BA.1/BA.1.1 dynamics: On 30 Dec 2021, Omicron composition was BA.1 84.6% (82.9%, 86.2%), BA.1.1 15.2% (13.6%, 16.6%), BA.2 0.2% (0.1%, 0.3%). By 1 Mar 2022, BA.1 declined to 9.6% (8.1%, 11.3%), BA.1.1 to 6.1% (5.7%, 7.1%), and BA.2 rose to 68.7% (64.6%, 72.3%). - Transmission advantage of BA.2: On 1 Mar, Rt was 1.17 (1.05, 1.28) for BA.2 vs 0.77 (0.69, 0.87) for non-BA.2 Omicron; multiplicative advantage for BA.2 ranged from 1.46 (1.40, 1.52) on 31 Jan to 1.54 (1.46, 1.60) in early February; consistent advantages were observed across regions and age groups with some regional heterogeneity. - Symptoms: A higher proportion of BA.2-infected individuals reported key COVID-19 symptoms (loss/change of smell/taste, fever, new persistent cough) at 53.5% (51.1%, 59.4%) vs 45.4% (43.3%, 47.6%) for BA.1, suggesting symptom-based surveillance may more effectively capture BA.2 infections. - Import/export signals: Phylogeographic and sequence similarity analyses suggested substantial international connectivity, with numerous inferred importations and exportations, especially involving London and neighboring regions.

Using unbiased, randomly sampled REACT-1 data, the study delineates the rapid replacement of Delta by Omicron in England, quantifies time-varying growth advantages, and shows how BA.2’s higher transmissibility prolonged the Omicron wave. The findings clarify that prevalence-based metrics offer a more direct measure of population exposure risk than routine test counts. Observed declines in Delta’s Rt likely reflect both susceptible depletion due to Omicron’s immune-evasive spread and behavioral changes. The faster generation time and early spread in younger cohorts explain the initially high growth advantage that attenuated over time. Regional and age-specific analyses show broadly synchronous dynamics with notable differences in timing and magnitude, while BA.2’s consistent advantage across settings explains rising Rt in late February despite widespread recent infections. Limited vaccine effectiveness against Omicron infection is evident from rising prevalence in older age groups, underscoring the need for boosters and updated vaccines. Overall, the study demonstrates that ongoing immune evasion and sub-lineage competition can drive recurrent waves even in highly immune populations.

The study provides a detailed, prevalence-based reconstruction of England’s Omicron wave, demonstrating rapid Delta displacement, a peak Omicron prevalence near 7% in late December 2021, and subsequent dominance of BA.2 with a clear transmission advantage that extended the wave. By modeling variant-specific prevalence and Rt, the work highlights time-varying growth advantages, regional and age-group heterogeneity, and differences in symptom reporting between sub-lineages. Given the immune-evasive evolution of SARS-CoV-2, intermittent waves of similar magnitude are plausible, implying a ‘new normal’ of periodic resurgences. Continued genomic and community surveillance, booster vaccination, and potential vaccine updates are recommended to mitigate health impacts. Future research should refine estimates of generation time by lineage, assess lineage-specific viral kinetics and Ct value distributions to reduce bias, and further integrate mobility/contact data with prevalence to disentangle behavior from immunological drivers.

- Sampling in discrete rounds produced gaps (e.g., late December 2021), widening credible intervals during key growth periods. - Sequencing was restricted to samples with N-gene Ct <34 and sufficient volume; lineage-specific differences in Ct or viral kinetics could bias lineage proportions. - Estimating prevalence (PCR positivity) vs incidence can bias Rt and growth estimates due to varying duration of positivity and delays; Rt trends may reflect such temporal effects. - Regional and age-stratified Rt estimates assume within-stratum mixing, which may not hold and should be interpreted cautiously. - Some sequences were excluded due to low coverage or quality, potentially affecting phylogenetic inferences; import/export inference relies on close temporal and genetic matches and may miss earlier events.