Modelling the efficacy of antiviral strategies of SARS-CoV-2 in a context of emerging variants: from hospitalized patients to general community
Maxime Beaulieu
Université Paris Cité, Inserm, IAME
Objectives The severe acute respiratory syndrome coronavirus (SARS-CoV-2) is a highly versatile virus that has shown a large capacity to adapt and escape immune pressure. Although vaccines have fortunately remained effective against severe forms of the disease[1], the emergence of Omicron variants has made vaccines much less effective against infection and transmission[2]. Despite the fact that millions of people have undergone PCR tests with viral load testing during the Pandemic, we still ignore to what extent the antagonist effects of vaccination and variant emergence have modulated viral load kinetics and whether this could explain in part the changes in virus transmissibility with Omicron emergence. Further the virus continues to cause hospitalization, in particular in fragile individuals that are not vaccinated or do not response to vaccination, such as immunocompromised individuals[1]. In this population, the administration of monoclonal antibodies (mAbs) could be beneficial, but its efficacy may be limited by the fact that the treatment may intervene too late to reduce viral load[3,4], and that these treatments may not be effective against Omicron variants[5]. Based on these observations, the objectives of my PhD thesis were to address two questions: 1)What is the virological efficacy of mAbs in Covid-19 hospitalized patients? 2)Can we gain insights into the impacts of vaccination and Omicron emergence on infection and transmission by analyzing the millions of PCR tests that have been done during this period in the population? Methods 1.We relied on data from DisCoVeRy, a European, randomized, placebo-controlled, phase III clinical trial[6]. The study evaluates Evusheld (AZD-7442), a cocktail of two mAbs, on hospitalized patients for SARS-CoV-2 infection. Patients received either placebo or Evusheld (IV infusion of 600mg). Patients were included throughout the Alpha, Delta and Omicron variant waves. Blood-neutralizing antibody titres (measured as ED50) against Delta, Omicron BA.1 and BA.2, and viral load were sampled. We used a sigmoid Gompertz model[7] to describe the neutralization kinetics of the patients that we integrated into a mechanistic viral dynamics model[8]. The impact of Evusheld on neutralization evolution and viral load dynamics was evaluated through simulations of 5,000 in silico Evusheld-treated profiles using estimated parameters of the models. Then, for each virtual patient, we calculated the viral kinetics with and without treatment to determine the gain in viral load decline and time to reach undetectable viral load. Finally, based on simulated population, we estimated the level of neutralization activity needed to achieve 0.5 log difference in viral load levels at day 5 compared to placebo, which is associated to a reduction risk of severe disease in outpatients[9]. Modelling and simulation was performed with Monolix Software 2018R2[10]. 2.We evaluated by simulation the possibility to use mathematical modelling to detect a signal in the change of viral kinetic parameters using sparse but massive amount of PCR tests. For that purpose we used a Bayesian inference framework to estimate parameter of a viral dynamic model, focusing on the essential parameters driving viral kinetics, namely the incubation period, the duration of virus proliferation, the peak viral load, and the time to viral clearance. We considered a situation where a large number of individuals were diagnosed (going from hundreds to hundred of thousand individuals) with no or few repeated data, thereby mimicking data collected in the population, where most individuals are only sampled at diagnosis, which typically occurs shortly after symptom onset, with no or few additional samples afterwards. Results 1.The model identified that the neutralization activity was associated with viral kinetics. Reflecting the variant-dependent neutralization activity of Evusheld, the antiviral activity of Evusheld was larger in patients infected with pre-Omicron or Omicron BA.2 variants than in patients infected with Omicron BA.1 variant. The model predicted that Evusheld reduced the median time to viral clearance compared with placebo-treated patients by more than 5 days in patients infected by pre-Omicron (median: 5.9; 80% PI: 2.1–13.6) or Omicron BA.2 (median: 5.4; 80% PI: 2.0–12.4), respectively. The effect was more modest in patients infected by the Omicron BA.1 variant, reducing the median time to viral clearance by 2 days (median: 2.2; 80% PI: 0.4–8.9). The increase in neutralization activity provided by Evusheld in patients infected with pre-Omicron and Omicron BA.2 variants implies a viral reduction greater than 1 log, which means that Evusheld could have clinical benefit in hospitalized patients. 2.We showed that precise estimates of viral dynamic parameters could be obtained with as little as 200 individuals, corresponding to 260 data in total. However, our approach identified that a bias in all parameters emerges when only positive PCR tests are included, and when data from individuals having only negative PCR tests are ignored (eg, infected individuals that could be only diagnosed after viral clearance). This shows that such approach needs to account for all data available, including negative tests. We illustrate the benefit of this approach on a very large collection of PCR tests carried out between June 2021 and March 2022 in French city laboratories. In total, the dataset contains more than 6.6 million tests on more than 4.2 million individuals. Conclusion We developed a mathematical model for integrating the evolution of neutralization activity into a viral dynamics model for evaluating a mAb on hospitalized patients. We show that Evusheld reduced the time to SARS-CoV-2 viral clearance in hospitalized patients by 5–6 days in patients infected with pre-Omicron or Omicron BA.2 variants, and by 2 days in patients infected with Omicron BA.1 variant. This shows that, despite a lower antiviral activity against Omicron variants, mAbs could still be part of the arsenal against Covid-19. In addition, we have shown that viral modelling can also be used outside clinical trials, to better understand the impact of vaccination and variant emergence on viral load collected in the general community. This study paves the way for the real-time use of within-host dynamics modelling to inform on the evolution of viral and epidemiological characteristics in future epidemics. Acknowledgements I thank Jérémie Guedj (IAME, Inserm) for supervising my PhD thesis, Florence Débarre (iEES-Paris), François Blanquart (CIRB), and France Mentré (IAME, Inserm) for valuable discussions on modelling.