MICA: Pharmacokinetic/Pharmacodynamic (PKPD) Model Development to Inform SARS-CoV-2 Antiviral Development

The COVID-19 pandemic has exposed two major weaknesses in our preparedness for respiratory viral threats. Firstly, there is a critical lack of available antiviral drugs which can be deployed at the first signs of symptoms or as post-exposure prophylaxis (given as a short course to people who have been in contact with an infected individual). Secondly, a basic principle of treating viral infections is that a combination of drugs with different modes of action is usually required, and for respiratory viruses, antiviral combinations are only effective if started in the first day or two following symptom onset. As with other respiratory viruses such as influenza, SARS-CoV-1 and MERS-CoV, SARS-CoV-2 viral replication rapidly slows following symptom onset with the later severe stage of disease mediated more by the body's response to the infection rather than active viral replication. Most clinical trials to-date have used single antiviral agents rather than combinations, and have studied hospitalised patients (i.e. late stage of the disease) when antivirals are unlikely to work. Most prioritised studies have been Phase III ttrials of agents that have not first been proven to reduce viral load in Phase II. Unsurprisingly, none of the repurposed monotherapies studied in this way have yet shown any benefit, and in the case of (hydroxy)chloroquine, have been proven to cause harm.

There is an urgent need to rationally develop combination antivirals which reduce viral load, disease severity and risk of onward transmission. For vaccines, rational development meant small Phase II studies to assess antibody response, with successful vaccines taken forward to Phase III. The analogy for antivirals is small Phase II studies to find antiviral combinations that reduce viral load before progressing successful ones to Phase III. Repurposing trials such as RECOVERY and PRINCIPLE which took antiviral monotherapies with limited in vitro activity straight to Phase III have now comprehensively proven to be an inefficient way to find effective antiviral combinations. A more rational approach based on sound principles of antiviral drug development is now required.

This work will focus on mathematical modelling of SARS-CoV-2 viral dynamics in order to optimally design and analyse the results for Phase II antiviral trials. Looking at the difference in viral load in patients receiving antivirals compared to placebo is complicated by the fact that in the normal course of the disease, viral load changes by the hour: after initial infection viral load in the nose and throat rises to a peak around the time of symptom onset, and then falls away again such that by Day 7 up to a third of people no longer have detectable virus. Viral load trajectories also differ in patients of different age, disease severity, and potentially when infected with different variants of the virus. Therefore a mathematical model of the expected time course is needed to tease out drug effects from these other variables.

Using data we have collected during a recent individual patient-level meta analysis, we will firstly compare the performance of various recently published viral dynamic models on how they predict viral load with time. Using data from two ongoing Phase II trials, FLARE and FANTAZE, the models will be refined to account for new variants (both are double blind randomised trials with daily viral loads and whole genome viral sequencing) and to develop models of the repurposed drug combinations being tested (favipiravir, lopinavir/ritonavir and nitazoxanide). We will also work with Pfizer to apply these models to novel agents in their antiviral pipeline, and apply the models to real world data from three London hospitals to assess whether certain patient groups with prolonged viral shedding may benefit from antiviral treatment.

The final output will be a modelling framework for the design and analysis of combination antiviral Phase II trials.

Antivirals generally need to reduce viral load to positively influence clinical endpoints such as hospitalisation or mortality. To prioritise antivirals for Phase III trials, Phase II trials seeking significant viral load decrease versus placebo are required. Interpreting viral load is challenging since it changes with time since infection (rises then falls) and by sampling site. This proposal seeks to develop a pharmacometric nonlinear mixed effects model of SARS-CoV-2 viral dynamics.

The model will be used to design and analyse efficient antiviral Phase II trials, and prioritise antivirals and antiviral combinations for Phase III. The target cell limited model, along with its common extensions (eclipse phase and innate and adaptive immune components) and simplifications (quasi-steady-state assumption between infected cell and free virus numbers), will be compared using graphical and numerical model diagnostics. A preferred model will be chosen and applied to viral load data in ongoing trials and to simulate outcomes to optimally design future trials. The model will be applied to real-world
clinical data to seek subgroups of hospitalised patients who may benefit from antivirals. It will also be used to perform a model-based appraisal of whole genome sequence-derived biomarkers such as subgenomic RNA, to be assessed as a possible Phase II endpoint.

This work will be carried out by a collaboration of academic, clinical and pharmaceutical industry investigators who will develop and share SARS-CoV-2 antiviral modelling best practice.