Shotaro Yamamoto: No financial relationships to disclose
The results in this abstract have been previously presented in part at The 2024 annual meeting of the Japanese Society for Mathematical Biology, Hokkaido University Conference Hall, September 11, 2024.
Objectives: The COVID-19 pandemic has demonstrated the global threat of infectious diseases, highlighting the necessity of preparing for future outbreaks. Rapid development of antiviral drugs is a key aspect of this preparedness, requiring well-designed clinical trials. In this context, in silico randomized clinical trials (isRCTs) have emerged as promising tools to optimize trial design by integrating pharmacological and virological data. In a clinical trial of nelfinavir (NFV), a candidate of anti-SARS-CoV-2 drug, the required sample size was calculated using an isRCT combining in vitro antiviral activity data and a mathematical model of in vivo viral dynamics. However, the trial showed no significant difference in the primary endpoint, i.e., time to viral clearance, suggesting a possible discrepancy between in vitro and in vivo efficacy.
Methods: We analyzed clinical trial data to quantify the difference in nelfinavir’s antiviral efficacy between in vivo and in vitro. The analysis used salivary viral RNA load data from an open-label randomized controlled trial conducted at 11 sites in Japan from July 2020 to October 2021. We estimated potency reduction factor (prf), representing the ratio between in vivo and in vitro half maximal inhibitory concentration (IC50) using mathematical models while accounting for inter-participant variability. Then, we generated “virtual patients“ from the estimated distribution of model parameters and simulated clinical trials to evaluate the effect of prf on the endpoints. To explore better dosing regimens and conditions, the required sample sizes were calculated under varying scenarios.
Results: The estimated PRF ranged from 2 to 5, indicating that drug concentrations 2–5 times higher than those predicted by in vitro data were necessary to achieve 50% viral inhibition in vivo. As a result, the required sample size of each group exceeded 4,000 participants under the observed in vivo efficacy—approximately 70 times larger than 60 participants, designed in the NFV clinical trial. Even under an alternative high-dose, low-frequency regimen, the required sample size remained substantial, at approximately 15,000 participants per group. In contrast, assuming in vivo efficacy equivalent to in vitro estimates reduced the necessary sample size to about 60 per group.
Conclusions: These results underscore the importance of incorporating in vivo pharmacodynamic considerations into clinical trial design. In particular, quantifying the prf prior to trial initiation—potentially by analyzing viral kinetic data from compassionate use cases—could enhance the accuracy of efficacy predictions and improve the isRCT framework. Such approaches would strengthen the translational bridge between preclinical findings and clinical outcomes in antiviral drug development.
Citations: [1] Iwanami, S. et al. Detection of significant antiviral drug effects on COVID-19 with reasonable sample sizes in randomized controlled trials: A modeling study. PLoS Med 18, e1003660 (2021). https://doi.org/10.1371/journal.pmed.1003660 [2] Hosogaya, N. et al. Efficacy and safety of nelfinavir in asymptomatic and mild COVID-19 patients: a structured summary of a study protocol for a multicenter, randomized controlled trial. Trials 22, 309 (2021). https://doi.org/10.1186/s13063-021-05282-w [3] Esmaeili, S. et al. A unifying model to explain frequent SARS-CoV-2 rebound after nirmatrelvir treatment and limited prophylactic efficacy. Nat Commun 15, 5478 (2024). https://doi.org/10.1038/s41467-024-49458-9
Keywords: Modeling and simulation, potency reduction factor, drug repurposing
