Timothy Knab (1), Ahmed Elmokadem (1), Emmanuel Chigutsa (2), Eric Jordie (1), Matthew Riggs (1), Patricia Brown-Augsburger (2), Christopher Wiethoff (2), Ajay Nirula (2), Jenny Y Chien (2), Lisa O’Brien (2).
(1) Metrum Research Group, Inc., 2 Tunxis Road, Suite 112, Tariffville, CT 06081, (2) Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46285
Objectives: Neutralizing monoclonal antibodies (mAb) to provide novel therapeutics for COVID-19 treatment, were urgently researched from the start of the pandemic. The selection of an optimal mAb candidate and therapeutic dose were expedited using open-access in silico models.
Methods: Candidate selection and effective therapeutic dose projection were supported by innovative adaptation of models obtained through open-science initiatives [1,2]. A physiologically-based pharmacokinetic (PBPK) model was modified by incorporating affinity-capture self-interaction nanoparticle spectroscopy data from the mAbs (AC-SINS). The PBPK model was then used to predict mAb clearance, tissue distribution, and estimated mAb exposures needed to maintain lung interstitial fluid (ISF) concentrations above IC90 of in vitro neutralization for up to 4 weeks in 90% of patients. AC-SINS score and in vitro neutralization of SARS-CoV-2 virus infection were determined as described (Jones et al., 2021 [3], Wu et al., 2015 [4]). Specific to this goal and to explore differentiation of the mAb candidates, the following characteristics were considered:
PK:
- Candidate mAb target distribution (ISF)
- mAb-specific properties affecting PK of individual candidates
- Interpatient variability based on intrinsic factors in the model (body weight, age, etc)
Viral Neutralization data:
- mAb-specific neutralization IC90 to establish target exposure for each candidate
A population PK model for bamlanivimab was fit to human data simulated from the PBPK model. Parameter estimates obtained using NONMEM® were scaled allometrically to predict serum concentration in primates and human. Inter-individual variability and parameter uncertainty (30-50%) were incorporated in the model predictions.
Serum concentrations of bamlanivimab were determined in cynomolgus and rhesus monkeys by ELISA, and bamlanivimab in vivo activity was determined in a rhesus prophylaxis model [3]. Serum concentrations of bamlanivimab in study PYAA were determined with a validated affinity capture, LC-MS/MS method.
Results: The PBPK model using molecule-specific AC-SINS predicted antibody clearance and organ-specific concentrations and projected 17-fold lower concentrations in lung ISF than corresponding plasma levels. Evaluation of early candidate antibodies primarily showed AC-SINS < 3, which alone was not sufficient to for final candidate selection. Therefore, the final candidate was selected based on projected therapeutic dose potency. In vitro neutralization data, with virtual patient simulations, were used support the selection of the most potent candidate where a clinical dose of 175 – 500 mg was expected to maintain target ISF mAb concentrations for over 28 days in 90% of patients.
The first-in-human trial with bamlanivimab (NCT04411628) proceeded with a 700 mg starting therapeutic dose, escalating to higher doses to evaluate the upper limit of safety and tolerability. Non-clinical and clinical a posteriori results confirmed the model predictions for PK, viral clearance, and ultimately the authorized dose.
Conclusions: The accelerated selection of bamlanivimab as the first neutralizing mAb drug candidate to enter clinical evaluation and the prediction of the maximum therapeutic dose of bamlanivimab for the treatment of COVID-19 in the absence of preclinical PK and animal PD data was supported by an in silico quantitative modeling and simulation framework.
References: [1] Jones et al. CPT:PSP 2019; 8(10): 738-747
[2] Shah and Betts. MAbs 2013; 5(2): 297-305
[3] Jones et al. STM 2021; eabf1906
[4] Wu et al. Protein Eng Des Sel 2015; 28(10): 403-414
Reference: PAGE 29 (2021) Abstr 9739 [www.page-meeting.org/?abstract=9739]
Poster: Drug/Disease Modelling - COVID-19