Predictive Primates: Enabling Rapid Go/No-Go Decisions on COVID-19 Vaccine Candidates with a Predictive, Translational MBMA and Clinical Immunogenicity Data
Seth Robey* (1), Nele Plock* (2), Anna Largajolli (2), Bhargava Kandala (1), Akshita Chawla (1), Kenny Watson (2), Raj Thatavarti (2), Sheri A Dubey (1), S. Y. Amy Cheung (2), Rik de Greef (2), Jeffrey R. Sachs (1)
(1) Merck & Co., Inc., Kenilworth, NJ, USA; (2) Certara, Princeton, NJ, USA; *These two authors will co-present
Objectives: The Covid-19 pandemic is motivating extraordinarily rapid development of safe and efficacious vaccines against SARS-CoV-2, increasing the need for innovative quantitative decision tools to optimize speed and efficiency. This need can be met through leveraging nonclinical immunogenicity and protection (efficacy) data (NIPD), early clinical immunogenicity data (ECID), and published clinical efficacy data (CED). We present a quantitative framework (QF) to predict likelihood of clinical success by integrating, across various vaccine platforms: NIPD, study attributes (design, dose level, regimen, duration etc.), assays, endpoints, and ECID. This framework can be used to support Go/No-Go (GNG) and study design decisions, enabling enhanced prioritization and, potentially, accelerated development of vaccine candidates.
Methods: A literature review based on pre-specified inclusion/exclusion criteria was conducted using PubMed-LitCovid, BioRxiv and MedRxiv databases. Summary level data of 22 Rhesus Macaque (RM) studies across various platforms with serum neutralizing titers (SN) and viral loads (VL) were curated in the database, along with data from several clinical vaccine candidate studies. For the RM model, peak post-challenge viral load (VL) in bronchoalveolar lavage (BAL), lung, and nasal tissues was used as an efficacy endpoint to indicate severity of infection, interpreting it as a putative translational surrogate for clinical disease. Model Based Meta-Analysis (MBMA) was performed to assess the relationships between pre-challenge SN and peak sub-genomic RNA (sgrna) VL post-challenge with SARS-CoV-2. Various study attributes (e.g. design, or assay differences) were evaluated as potential covariates during model development. The VL predictions from the model were scaled to incidence rate of symptomatic disease (IR) in humans, enabling prediction of clinical efficacy based on Phase 1/2 ECID. To standardize assay differences across clinical studies, ECID was normalized to reported convalescent titers1,2. A generalized least square approach allowed accounting for the species differences (between RM and human) in assay interpretation. To qualify the model, Phase 1/2 immunogenicity data of COVID-19 vaccine candidates not used in model construction were used to predict vaccine efficacy and compared to their respective observed Phase 3 efficacy data.
Results: Clinical efficacies predicted from the translated MBMA model and Phase I/II ECID agreed with the efficacies reported for the respective novel COVID-19 vaccine candidates. The prediction came from a sigmoidal decrease in peak viral load (increasing protection) with increase in SN identified by the MBMA analysis. Results also indicated, as anticipated, that higher levels of SN titers are needed to reduce VL in nasal swabs than in BAL and lung tissue, and that there was no substantial difference between models for BAL and lung tissue data.
Conclusions: This translational MBMA model identified SN as quantitatively predictive of efficacy against COVID-19. This result is consistent across species after adjustment for assay and species differences. It can, thus, be leveraged to support GNG, dose-level, and regimen decisions for new COVID-19 vaccine candidates.
 K.A. Earle, D.M. Ambrosino, A. Fiore-Gartland et al. Evidence for antibody as a protective correlate for COVID-19 vaccines. medRxiv preprint: https://doi.org/10.1101/2021.03.17.20200246; posted March 20, 2021
 D.S. Khoury, D. Cromer, A. Reynaldi et al. What level of neutralising antibody protects from COVID-19? medRxiv preprint: https://doi.org/10.1101/2021.03.09.21252641; posted March 11, 2021