Rhiannon Edge¹, Sam Matthews¹, John Pura2, Bahar Ahani3, Anastasia Aksyuk3, Lindsay Clegg4, Mark Esser3, Lee-Jah Chang3, Ian Hirsch1, Tonya Villafana3, John Perez3, Oleg Stepanov5, Katie Streicher3, Tom White1, Taylor Cohen3, Dean Follmann6, Peter Gilbert7, Seth Seegobin1
1Biometrics, Vaccines & Immune Therapies, BioPharmaceuticals R&D, AstraZeneca, 2Biometrics, Vaccines & Immune Therapies, BioPharmaceuticals R&D, AstraZeneca, 3Vaccines and Immune Therapies, BioPharmaceuticals R&D, AstraZeneca, 4Clinical Pharmacology and Quantitative Pharmacology, Clinical Pharmacology & Safety Sciences, R&D, AstraZeneca, 5Clinical Pharmacology and Quantitative Pharmacology, Clinical Pharmacology & Safety Sciences, R&D, AstraZeneca, 6Biostatistics Research Branch, National Institute of Allergy and Infectious Diseases, National Institutes of Health, 7Vaccine and Infectious Disease Division, Fred Hutchinson Cancer Center
Objectives The constantly shifting variant landscape of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has made development of monoclonal antibodies targeting the virus very challenging. New clinical studies evaluating the efficacy of monoclonal antibodies (mAbs) for pre-exposure prophylaxis (PrEP) of symptomatic COVID-19 cannot keep pace with the emergence of new variants. Thus, a robust method is needed to rapidly assess expected mAb efficacy against novel variants. We developed a model to relate observed efficacy to prevalence-adjusted neutralising antibody titres (PATs), to facilitate establishment of a threshold of protection for mAb PreP of COVID-19. Methods PATs were estimated by calculating predicted neutralising antibody titres from measured mAb serum concentrations and potency against different SARS-CoV-2 variants (in vitro IC50 values) [1], with the potency averaged across viral variants based on prevalence data at each day of follow-up. Daily PATs were derived from individual population pharmacokinetic predictions of serum mAb concentrations over time. Viral variant prevalence was based on publicly available SARS-CoV-2 variant surveillance data [2]. Individual participant PATs were related to efficacy with a time-varying Cox model. The model was built using data from the phase 3 PROVENT pre-exposure prophylaxis trial of tixagevimab–cilgavimab (NCT04625725). The model was then validated using independent data from the SUPERNOVA trial (NCT05648110), which assessed sipavibart efficacy against COVID-19 in participants with immunocompromising conditions during a different variant landscape. Finally, the model was refit to combined data from both studies, using a stratification term to account for potential differences in baseline hazards between studies. Results In PROVENT, there were 63 events of symptomatic COVID-19 amongst 3272 participants (1.8%) in the tixagevimab–cilgavimab group with available serum mAb concentration predictions from the previously-developed population pharmacokinetic model [3], and 55 events amongst 1730 participants (3.2%) in the placebo group [4]. In SUPERNOVA, the analysis included 122 events amongst 1649 participants (7.4%) in the sipavibart group with available serum mAb concentration predictions, and 178 events amongst 1631 participants (10.9%) in the comparator group [5]. The Cox model built upon PROVENT data estimated the variant-specific observed efficacies from SUPERNOVA for 3- and 6-months post any dose with Lin’s concordance (two-sided 95% bootstrapped confidence interval) of 0.86 (0.45, 0.89) and 0.75 (0.28, 0.92), respectively. These results indicate good agreement between model-predicted and observed efficacy across a range of SARS-CoV-2 variants with in vitro IC50 values ranging from 3.8 to 83 ng/mL. When the model was estimated based on pooled data from PROVENT and SUPERNOVA, predictions were largely consistent with those for the model fit to PROVENT data alone, supporting the generalizability of the findings. Conclusions While past analyses have established a relationship between neutralising antibody titres and protection from symptomatic COVID-19, this analysis is the first model built on individual patient-level data to account for multiple SARS-CoV-2 variants with shifting prevalence over time, and against which the analysed mAbs have varying potency. The resulting externally validated model could be used as a benchmark for inferring the expected efficacy of new or existing prophylactic mAbs, potentially facilitating rapid development and assessment of new mAbs, as well as real-time adjustment of the recommended dosing regimens for mAbs authorized PrEP of COVID-19 as new variants emerge.
[1] Clegg, L. E. et al. Serum AZD7442 (tixagevimab-cilgavimab) concentrations and in vitro IC(50) values predict SARS-CoV-2 neutralising antibody titres. Clin Transl Immunology 13, e1517 (2024). [2] Shu Y, McCauley J. GISAID: Global initiative on sharing all influenza data – from vision to reality. Euro Surveill. 2017;22:30494. [3] Clegg, L. E. et al. Accelerating therapeutics development during a pandemic: population pharmacokinetics of the long-acting antibody combination AZD7442 (tixagevimab/cilgavimab) in the prophylaxis and treatment of COVID-19. Antimicrob Agents Chemother 68, e0158723 (2024). [4] Levin, M. J. et al. Efficacy, Safety, and Pharmacokinetics of AZD7442 (Tixagevimab/Cilgavimab) for Prevention of Symptomatic COVID-19: 15-Month Final Analysis of the PROVENT and STORM CHASER Trials. Infect Dis Ther 13, 1253–1268 (2024). [5] Haidar, G. et al. Sipavibart for prevention of COVID-19 in immunocompromised persons (SUPERNOVA): a randomised, double-blind, phase 3 trial. Lancet Infect Dis, (2025).
Reference: PAGE 33 (2025) Abstr 11378 [www.page-meeting.org/?abstract=11378]
Poster: Drug/Disease Modelling - Other Topics