Favipiravir pharmacokinetics in Thai adults with mild COVID‐19: A sub‐study of interpatient variability and ethnic differences in exposure

Abstract

This sub‐study sought to characterize the pharmacokinetics (PK) of favipiravir (FPV) within Thai adults and quantitatively assess differences in exposure to those previously reported in other populations as a basis to understand putative differences in efficacy between studies conducted in different regions. It was nested within a prospective trial of adults with symptomatic COVID‐19 infection without pneumonia receiving 1800 mg FPV twice‐daily on day 1 and 800 mg twice‐daily thereafter. Individual PK profiles were fitted with a one‐compartment disposition model (first‐order absorption). Eight adults (seven female) with a median age of 39 years and BMI of 27.9 kg/m² were included. Seven adults achieved plasma concentrations above the EC₉₀ in vitro target (25 mg/L), with minimum–maximum concentrations decreasing with repeat dosing. The mean FPV apparent clearance observed in this study was 1.1 L/h (coefficient of variation [CV]: 60%), apparent volume of distribution 20.6 L (CV: 40%), absorption rate constant 6.1 h (CV: 100%), and 2.4 daily % change in apparent clearance (CV: 315%). Higher exposures were observed in these Thai adults compared with data from previous studies in Chinese, Japanese, and Turkish populations, respectively. Current FPV doses recommended in Thailand achieved target plasma concentrations with higher exposures than those described previously in other populations. The limited sample size prohibits firm conclusions from being drawn but the presented data warrants confirmation with a view to interrogate the appropriateness of doses used in randomized clinical trials that failed to demonstrate efficacy.

Population simulations of favipiravir plasma exposure using pharmacokinetic parameters and inter‐individual variabilities from the fitting to the Thai pharmacokinetic dataset at regimens (or a mix of regimens in the correct ratio) to match overlaid observed data from: (a) Thailand; (b) China; (c) Japan; and d) Turkey.

Abbreviations

AO

aldehyde oxidase

BMI

body mass index

C _avg

average plasma concentration

C _max

maximum plasma concentration

C _min

minimum plasma concentration

C _p

plasma concentration

C _trough

trough plasma concentration

CL/F

apparent clearance

FPV

favipiravir

FPV‐RTP

favipiravir ribofuranosyl‐5′‐triphosphate

HPLC‐UV

high performance liquid chromatography‐ultraviolet assay

ICU

intensive care unit

K _A

absorption rate constant

MKD

mg/kg/day

PC‐VPC

prediction corrected visual predictive check

PD

pharmacodynamics

PK

pharmacokinetics

RCT

randomized controlled trial

RdRP

RNA‐dependent RNA polymerase

XO

xanthine oxidase

V/F

apparent volume of distribution

1. INTRODUCTION

Favipiravir (6‐fluoro‐3‐hydroxy‐2‐pyrazinecarboxamide, T‐705, or FPV) is a nucleoside analog and pyrazine carboxamide derivative that inhibits COVID‐19 RNA‐dependent RNA polymerase (RdRP), blocking viral replication via lethal mutagenesis. ¹ , ² FPV has demonstrated broad‐spectrum antiviral activity, ³ , ⁴ , ⁵ a capacity to competitively inhibit viral replication in vitro, ⁶ , ⁷ and virological efficacy in COVID‐19 animal infection models. ³ , ⁸ Based on these observations, FPV has been studied in several COVID‐19 clinical trials of variable quality. For the most part, these doses were directly transposed from those of its primary indication (influenza) without a robust assessment of their plausibility.

An improved viral clearance and hindered viral progression has been described in some regions, ⁶ , ⁹ while conflicting results for clinical outcomes (e.g., intubation, in‐hospital mortality, and ICU administration) and/or virological efficacy have been reported in others when studied as a monotherapy or in combination. ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ A previous pharmacometrics assessment strongly suggested that maintenance doses of 1200 mg twice‐daily would be required to achieve intracellular target exposures based on the pharmacokinetics (PK) in non‐Thai populations. ¹⁵ Two recent phase II randomized controlled trials (RCT) conducted in the United Kingdom failed to demonstrate clinical efficacy of FPV, but both provided some evidence for an impact on viral replication consistent with PK exposure on the cusp of the efficacy threshold. ¹⁶ , ¹⁷ The FLARE trial utilized a 400 mg maintenance dose administered four times daily for 7 days, whereas the PIONEER trial utilized an 800 mg maintenance dose administered twice‐daily for 9 days (both had an 1800 mg loading dose administered twice on day 1).

Certain countries permit using FPV to treat COVID‐19 (i.e., Turkey, Japan, India, Russia, Saudi and Arabia), ¹⁸ , ¹⁹ including the National Clinical Practice Guidelines of Thailand's Ministry of Public Health, for mild COVID‐19 infections. ²⁰ The recommended dosing regimen for adults <90 kg in Thailand is a loading dose of 1800 mg twice‐daily for the first day, and maintenance dose of 800 mg twice‐daily thereafter for a total duration of 5–14 days. ⁶ , ¹⁰ , ²¹ This regimen is adjusted to 2400/1000 mg twice‐daily for adults ≥90 kg. ²⁰ FPV is currently not recognized as a treatment for COVID‐19 by the World Health Organization, Centres for Disease Control and Prevention, Food and Drug Administration, and National Institutes of Health. ²² , ²³ , ²⁴ , ²⁵

FPV is a prodrug that undergoes intracellular ribosylation and phosphorylation to form an active metabolite (favipiravir ribofuranosyl‐5′‐triphosphate, or FPV‐RTP) which exerts antiviral activity upon incorporation into nascent viral RNA as a purine‐base analog by viral RdRp. ¹ , ³ , ⁶ , ⁹ The plasma half‐life of FPV has been reported to be 2.0 to 5.5 h. ³ , ⁵ FPV‐RTP is subsequently metabolized primarily by hepatic aldehyde oxidase (AO) and partially by xanthine oxidase (XO) into hydroxylated inactive metabolites (T‐705M1) that are excreted through urine. ³ , ⁵ , ²⁶ Studies report a wide 50% effective concentration (EC₅₀) range of 10 to 78 mg/L (62 to >500 μM), ⁷ , ¹¹ , ²⁷ , ²⁸ , ²⁹ , ³⁰ with some not identifying SARS‐CoV‐2 inhibition below 15.71 mg/L (100 μM) in vitro ¹² , ³¹ or only 50% virus inhibition at this concentration. ¹² The EC₉₀ was found to be 25.0 to 52.5 mg/L and cytotoxic concentration (CC₅₀) >400 to 500 μM in Vero E6 and Caco‐2 cells. ⁶ , ⁷ , ³⁰ As the predicted intracellular half‐life of the active metabolite is longer than the parent drug in plasma, twice‐daily administration at a higher dose may be expected to provide more robust intracellular concentrations across the dosing interval than administrations four times a day at a lower dose. ¹⁵

Numerous studies have explored FPV's clinical effects, but few assessed or analyzed its PK, which differ across ethnicities. ⁶ , ¹¹ , ²⁷ , ³² , ³³ , ³⁴ , ³⁵ A better understanding of FPV PK is required to rationalize its application for treatment of COVID‐19, influenza, and other viral infections. No robust description of FPV PK has been obtained in a Thai population. Characterizing the impact of regional and ethnic difference on PK is critical to ensure optimized FPV dosing against circulating viral variants in the region and higher doses may be needed in some populations than others. ²⁷ PK models may be useful to facilitate exposure–response relationship studies. ³⁶ , ³⁷

This PK study aimed to characterize and model FPV plasma PK parameters within a Thai adult population with mild COVID‐19 without pneumonia and make a preliminary quantitative assessment of differences between those reported in other populations.

2. METHODS

2.1. Study design and patients

This PK sub‐study of FPV (FUJIFILM Toyama Chemical Co., Ltd.; 200 mg per tablet) was nested within a prospective clinical trial of adults with symptomatic COVID‐19 infection without pneumonia. ³⁸ Patients aged ≥18 years, infected with COVID‐19 (PCR‐confirmed), and displaying mild‐to‐moderate symptoms were included in the parental study. Patients with pneumonia or in critical condition (>10 days after symptomatic onset), pregnant or lactating, on immunosuppressants or medication with potential antiviral activity (e.g., hydroxychloroquine, remdesivir), or with concurrent/concomitant infections were excluded from the parent study. ³⁸

2.2. Study procedure and pharmacokinetic assessment

Patients were randomized 2:1 into FPV and control arms in our parent study. ³⁸ The control arm received only symptomatic therapy (i.e., oxygen therapy, medication, intravenous fluid administration, etc.) and supportive care tailored to patients' needs. In addition to symptomatic and supportive care, the FPV arm received a loading dose of 1800 mg (nine tablets/dose) twice‐daily on the first day, and maintenance dose of 800 mg (four tablets/dose) twice‐daily for 5 to 14 days thereafter until clinical improvement or viral clearance (a negative result from the saliva RT‐PCR test). ³⁸ In this sub‐study, the patients in the FPV arm were invited. Blood samples were drawn pre‐dose, and 1‐h post‐FPV administration on days 1 to 5, 7, and 14. Samples were stored at −20°C until analysis.

2.3. Quantification of favipiravir

Plasma proteins were precipitated with ethanol (1:4) and the supernatant stored at −20°C until analysis. FPV and its metabolite T‐705M1 were quantified using a validated high performance liquid chromatography‐ultraviolet (HPLC‐UV) assay (calibration range of 0.1 to 50 mg/L), as reported previously (UHPLC Ultimate 3000, Thermo Fisher Scientific™, USA). ³⁹ The results of this validation are included in Table S1 and Figure S3.

2.4. Pharmacokinetic modeling and simulation

A two‐stage population PK analysis was applied to the data ⁴⁰ due to insufficient ability of the dataset to robustly inform estimates of interindividual variability in a full, nonlinear mixed effects, population PK analysis. This was accentuated by the small number of patients in the dataset (n = 8). The PK profile for each patient was individually fitted with the PK model according to their specific dosing record and the resulting individual parameter estimates reported in addition to summary statistics for parameters across the eight patients. Following exploratory analyses, and in accordance with previous PK models applied to FPV clinical data, ³² a one‐compartment PK disposition model with first‐order oral absorption input was fitted to the data parameterized by apparent clearance (CL/F, L/h) apparent volume of distribution (V/F, L), and absorption rate constant (K _A, h⁻¹). CL/F was estimated to change by a cumulative, multiplicative % factor each dosing day ³² per the following equation:

This model is able to describe the apparent induction (increase) of FPV clearance over time with repeated dosing seen in previous studies, ³² which was also observed in exploratory analyses of several individual patient profiles in this study and has been applied to the data in this study based on its previous use for FPV PK. ³² The model is also able to describe patient PK profiles where CL/F appears to decrease with repeated doses depending on whether the estimated “Factor” is positive or negative. While the mechanisms for this potential induction (or inhibition) of FPV clearance remain unknown, it has been hypothesized that induction of AO expression or activity by FPV, or that other factors related to illness (e.g., fluid redistribution, hypoalbuminemia, or non‐FPV‐related changes in AO activity associated with fever) may underlay the observed effects. ³² , ⁴¹

FPV population PK simulations were performed using the estimated means and variances of the PK model parameters (calculated from the n = 8 individual fittings) to randomly generate 1000 sets of individual model parameters—assuming log‐normal distributions for the parameters across a population. PK profile simulations from these parameters were generated and overlaid with observed data for regimens, or combinations of regimens, as follows: (a) 1800/800 mg (n = 1000) FPV twice‐daily loading/maintenance dosing regimen over 15 days, with the 90% prediction interval calculated from the range of the individual simulations overlaid with observed Thai plasma exposure data obtained under the same regimen (n = 8); (b) 1800/800 mg (n = 543) and 1600/600 mg (n = 457) FPV twice‐daily loading/maintenance dosing regimens over 15 days, with the simulated population prediction corrected visual predictive check (PC‐VPC) 90% prediction interval ⁴² overlaid with PC‐VPC observed Chinese plasma exposure data digitized from a prior publication by Wang et al. under the same regimens (n = 19 for 1800/800 mg, n = 16 for 1600/600 mg); (c) 1800/800 mg (n = 154) and 1600/600 (n = 846) FPV twice‐daily loading/maintenance dosing regimens over 15 days, with a simulated population PC‐VPC 90% prediction interval overlaid with PC‐VPC observed Japanese plasma exposure data digitized from a prior publication by Irie et al. under the same regimens (n = 6 for 1800/800 mg, n = 33 for 1600/600); and (d) 1600/600 mg (n = 1000) FPV twice‐daily loading/maintenance dosing regimen over 15 days, with the 90% prediction interval calculated from the range of individual simulations overlaid with observed Turkish (n = 21) mean plasma exposure profile obtained under the same regimen as reported by Gulhan et al.

These simulations and overlays illustrate the adequacy of the PK model and estimated parameters of the Thai dataset to describe various observed datasets. For the Thai observed data itself, the overlay plot serves as a VPC to the data of the model fittings and their mean parameter estimates. For the other datasets, the simulations based on the Thai PK model will indicate if expected exposures based on the Thai data PK model concur with what was observed in other populations. Where the observed data contains a mix of regimens and was obtained by digitization of a prior published PC‐VPC, the simulated population for overlay contains the same ratio of the different regimens to match the observed data, and population prediction correction ⁴² for calculation of the 90% prediction interval achieves dose‐normalization for plotting purposes. Simulations and overlays were also carried out using population PK parameter estimates from previously published PK models for FPV clinical exposure data from Chinese (Wang et al.) and Japanese (Irie et al.) populations (see Supplementary material for details). Model fittings and simulations were performed using R (v4.3.0), ⁴³ specifically the Pracma library for fittings ⁴⁴ and the lsqnonlin function for nonlinear regression (with a 1/[predicted²] weighted objective function).

3. RESULTS

Of the 62 patients in the FPV‐arm of our parent study, ³⁸ eight (seven females) were included in this PK sub‐study. The median (range) age of these patients was 39 (19–53) years and the BMI was 27.9 (18.0–33.6) kg/m². Three patients (37.5%) had underlying diseases, and none developed pneumonia or experienced disease progression. Patients No. 1 and 3–5 were treated with FPV for 14 days, and Patients No. 6–8 for 7 days. Patient No. 2 dropped out on the fourth day of treatment due to an erythematous cutaneous rash. This patient had relatively high FPV concentrations (99.7–201.2 mg/L) and low BMI (18.3 kg/m²). None of the other patients' clinical outcomes deteriorated during treatment. Individual plasma concentrations (C _p) versus time curves for FPV and T‐705M1 (after the first dose of FPV) are shown in Figure 1. Initial pre‐dose FPV concentrations for Patients No. 4 and 5 indicated they received FPV prior to our study.

FIGURE 1.

Individual plasma concentration of favipiravir (FPV) and T‐705M1 for: (A) Patient No. 1, a 33‐year‐old female with a body mass index (BMI) of 30.1 kg/m²; (B) Patient No. 2, a 38‐year‐old female with a BMI of 18.3 kg/m²; (C) Patient No. 3, a 45‐year‐old female with a BMI of 33.6 kg/m²; (D) Patient No. 4, a 29‐year‐old female with a BMI of 30 kg/m²; (E) Patient No. 5, a 48‐year‐old female with a BMI of 27.5 kg/m²; (F) Patient No. 6, a 53‐year‐old male with a BMI of 26.9 kg/m²; (G) Patient No. 7, a 40‐year‐old female with a BMI of 28.2 kg/m²; and (H) Patient No. 8, a 19‐year‐old female with a BMI of 18.3 kg/m². The dashed line represents an EC₉₀ of 25 mg/L.

Minimum plasma concentrations (C _min) of FPV were above the in vitro EC₉₀ target (25 mg/L) for all patients, except Patient No. 6 (who had the lowest C _min of 8.6 mg/L following their first dose). The C _min and maximum plasma concentration (C _max) of FPV decreased overtime and ranged from 12.1 to 132.2 mg/L and 14.5 to 201.2 mg/L, respectively. The FPV metabolite T‐705M1 remained relatively stable over time and ranged from 1.6 to 6.7 mg/L (C _min) and 2.0 to 24.8 mg/L (C _max). Coefficients of variation (CV) of 37% to 74% for FPV and 12%–62% for T‐705M1 indicated high interpatient variability consistent with previous reports.

Individual fitted PK models for the eight patients' C _p versus time profiles after initial FPV administration are shown in Figure 2. Given the sampling design timepoints and rapid absorption, the precision of the K _A estimates was generally poor (Table 1). Some apparent discrepancies between nominal dosing, sampling times, and measured C _p were observed, particularly in Patients No. 5 and 8, leading to poor model fittings at C _max and C _trough on study days 3 and 6 and study days 2 and 3 for these patients, respectively. Regardless, PK parameter estimates for these patients were consistent with the other patients in this study (Table 1). In three individuals (Patients No. 2, 6, and 7), the daily % change in CL/F was estimated as negative to best describe their profiles—implying that in these individuals' CL/F appeared to decrease over the course of this study. The overall mean across the eight patients was, however, still a net positive daily % change in CL/F. This likely reflects both general interindividual variability in FPV disposition and other random variability in the study.

FIGURE 2.

Individual FPV PK model fitting profiles overlaid with observed plasma concentrations for: (A) Patient No. 1, a 33‐year‐old female with a body mass index (BMI) of 30.1 kg/m²; (B) Patient No. 2, a 38‐year‐old female with a BMI of 18.3 kg/m²; (C) Patient No. 3, a 45‐year‐old female with a BMI of 33.6 kg/m²; (D) Patient No. 4, a 29‐year‐old female with a BMI of 30 kg/m²; (E) Patient No. 5, a 48‐year‐old female with a BMI of 27.5 kg/m²; (F) Patient No. 6, a 53‐year‐old male with a BMI of 26.9 kg/m²; (G) Patient No. 7, a 40‐year‐old female with a BMI of 28.2 kg/m²; and (H) Patient No. 8, a 19‐year‐old female with a BMI of 18.3 kg/m². CL/F, apparent clearance; V/F, apparent volume of distribution; K _A, absorption rate constant; Daily Inc. CL/F, daily % change in apparent clearance.

TABLE 1.

Pharmacokinetic parameters of the eight Thai subjects used for Thai population simulations compared with Chinese, Japanese, and Turkish data.

Subject number	CL/F (L/h) [%RSE]	V/F (L) [%RSE]	K A (h−1) [%RSE]	Daily % change in CL/F [%RSE]
1	1.2 [14.7]	22.8 [18.5]	1.5 [1.0]	10.6 [41.3]
2	0.7 [17.5]	13.5 [16.3]	6.3 [1906.0]	−7.8 [324.0]
3	1.5 [117.7]	24.9 [48.6]	5.9 [6175.6]	6.3 [142.2]
4	0.8 [986.7]	14.7 [2154.2]	6.0 [1827.2]	8.1 [1901.4]
5	0.7 [1311.9]	14.1 [4893.9]	1.8 [14982.2]	9.5 [5274.3]
6	2.7 [23.1]	38.0 [34.5]	19.4 [0.5]	−3.8 [206.2]
7	1.0 [9.3]	17.8 [12.7]	7.1 [965.6]	−7.2 [50.3]
8	0.6 [25.4]	19.1 [35.1]	0.2 [51.7]	3.5 [286.8]
Range	0.6–2.7	13.5–38.0	0.2–19.4	−7.8‐10.6
Geometric mean	1.0	19.5	3.4	‐
Variance of log‐transformed parameter	0.3	0.1	1.8	‐
Arithmetic mean	1.1	20.6	6.0	2.4
Variance	0.5	66.2	36.0	57.3
SD	0.7	8.1	6.0	7.6
%CV	60.4	39.5	99.7	315.4

Abbreviations: CL/F, apparent clearance; CV, coefficient of variation; K _A, absorption rate constant; %RSE, percent relative standard error; V/F, apparent volume of distribution.

Despite these limitations, overall description of the observed data by the model in most individual subjects (Figure 2) and precision of parameter estimates other than K _A are generally adequate, with mean PK parameters (Table 1) consisting of: CL/F of 1.1 L/h (CV across eight patients: 60%), V/F of 20.6 L (CV: 40%), K _A of 6.1 h⁻¹ (CV: 100%), and 2.4% daily increase in CL/F (CV: 315%).

Population simulations of plasma exposure based on this model and parameter estimates carried out to match observed (PC‐VPC) datasets from Thailand (this study), China, ³² Japan, ³⁴ and Turkey ³³ are illustrated in Figure 3A–D, respectively. Higher plasma C _max, C _min, and on average (C _avg) would be expected based on PK parameters estimated from the Thai data than were observed for given regimens in Chinese, Japanese, and Turkish studies. Similarly, simulations based on the population PK models of Wang et al. (Chinese Population dataset) and Irie et al. (Japanese Population dataset) demonstrated that observed Thai exposures are higher than expected from either of these models, while the Chinese, Japanese, and Turkish datasets were all broadly aligned with expected exposures based on the Wang et al. and Irie et al. PK models (Figures S1 and S2).

FIGURE 3.

Population simulations of FPV plasma exposure (90% prediction interval for n = 1000 simulated individual profiles) using PK parameters and inter‐individual variabilities from the fitting to the Thai PK dataset at regimens (or a mix of regimens in the correct ratio) to match overlaid observed data from: (A) Thailand (1800/800 mg twice‐daily loading/maintenance dosing regimen, n = 8); (B) China (PC‐VPC data of 1800/800 mg and 1600/600 mg twice‐daily loading/maintenance regimens from Wang et al. (2020b), ³² n = 19 and n = 16, respectively); (C) Japan (PC‐VPC data of 1800/800 mg and 1600/600 mg twice‐daily loading/maintenance regimens from Irie et al. (2021), ³⁴ n = 6 and n = 33, respectively); and (D) Turkey (1600/600 mg twice‐daily loading/maintenance regimen, data from Gulhan et al. (2022), ³³ n = 21, mean profile).

4. DISCUSSION

This study evaluated a PK model fitted to FPV plasma PK data in Thai adults to make a preliminary quantitative assessment of differences between FPV plasma PK parameters reported in other populations. The recommended 1800/800 mg twice‐daily FPV dosing regimen in Thailand provided FPV C _p above a SARS‐CoV‐2 target EC₉₀ of 25 mg/L. Since FPV is a broad‐spectrum antiviral drug, the data presented herein may also be valuable for future dose optimization in Thai patients in the context of other viruses.

FPV activity may be more efficacious in‐vivo than in‐vitro due to its nonlinear PK. ⁶ , ³⁴ Circulating concentrations of active FPV can initially increase through dose‐ and time‐dependent auto‐inhibition of hepatic AO, hypothesized to promote cellular uptake of FPV and its conversion to FPV‐RTP. Therefore, the resulting net decrease in circulating FPV may not translate to decreased FPV‐RTP exposure in tissues or target cells. ³ , ⁶ , ¹² , ²⁷ , ³⁴ , ⁴⁵ , ⁴⁶ Intracellular PK of the active FPV‐RTP metabolite have not been assessed in any population, but are expected to be different from the parent drug in plasma. ¹⁵ For FPV, and other drugs in this class, it may not be necessary to maintain EC₉₀ targets of the parent drug in plasma across the entire dosing interval, as the predicted intracellular half‐life of the active metabolite is longer than the parent drug in plasma. ¹⁵

FPV loading doses previously explored in adults ranged from 400 to 6000 mg twice‐ or thrice‐daily. ⁴⁷ Irie et al. characterized FPV PK in Japanese populations (predominantly male, median age: 68 years) with severe COVID‐19 primarily under a 1600/600 mg twice‐daily, 5‐day dosing regimen. They found that trough concentrations (C _trough) were lower than the target EC₅₀ of 9.7 mg/L, let alone the IC₉₀—potentially limiting therapeutic effects. Gulhan et al. studied the same dosing regimen but in a Turkish population (predominantly male, median age: 58 years) with mild to moderate COVID‐19. They noted that C _trough decreased rapidly from days 2–4 (21.26 mg/L to 1.61 mg/L) after treatment. Wang et al. simulated FPV profiles under 1600/600 and 1800/800 mg twice‐daily, 9‐day dosing regimens in critically‐ill patients (predominantly male, median age: 63.5 years). Higher proportions of patients had C _trough above 20 mg/L for longer durations in the 1800/800 mg twice‐daily group compared to the 1600/600 mg twice‐daily group. Early administration of dosing regimens higher than 1600/600 mg twice‐daily were required to reach target FPV and FPV‐RTP concentrations to achieve associated antiviral effects. ³² , ³⁴ PK parameters estimated from the Thai dataset in this study resembled those in previous literature, ³² , ³³ , ³⁴ but with the mean CL/F of 1.1 L/h approximately 3–9 times lower compared with PK model parameter estimates from Chinese (3.0 L/h, Wang et al. ³² ), Japanese (5.1 L/h, Irie et al. ³⁴ ), and Turkish (8.6 to 10.7 L/h, Gulhan et al. ³³ ) populations, respectively. These CL/F values directly reflect the observed differences in plasma exposures between these populations. FPV's low volume of distribution of 10 to 20 L ³ , ⁴⁵ , ⁴⁸ was previously influenced by vascular and extravascular fluid volume as well as protein binding. ⁴⁸ C _max is typically reached 2 h after an oral dose, ³ , ⁵ , ⁴⁵ with mean and median C _p at least twice that of its EC₉₀. ⁴⁹ Plasma exposures in this Thai dataset exceed the steady‐state C _trough exposure of 4.4 mg/L achieved at an efficacious intraperitoneal dose of 1000 mg/kg preclinically in hamsters. ⁵⁰ Prolonged use of FPV has been associated with decreased parent FPV steady‐state concentration in plasma, particularly by day 4 of administration. ³ , ¹⁵ Many covariates affect this elimination process, for instance, creatinine clearance, older age, and loading doses ≤45 mg/kg/day (MKD). ⁷ , ¹⁰ , ⁵¹

The patients in this study, in contrast to those discussed by Irie et al., Wang et al., and Gulhan et al., were not only younger (median age: 39 years) but predominantly female. Gulhan et al. noted FPV C _trough and C _max were higher in females than males, and clearance higher in males than females, both 2 and 4 days after treatment. This difference in demographics may have further contributed to differences in PK elimination and effect profiles. A multivariate analysis by Rattanaumpawan et al. ⁵¹ found that old age and FPV loading doses ≤45 MKD were associated with poor prognostic factors for clinical improvement by day 7 of FPV treatment. Suboptimal FPV dosing for SARS‐CoV‐2 may contribute to the therapeutic failure observed in patients who did not reach adequate FPV exposure. Lower C _trough values on day 2 of FPV treatment were reported by Gulhan et al. in deceased patients than those who had recovered. ³³ Ensuring that patients rapidly reach target FPV concentrations after treatment initiation could help achieve therapeutic efficacy (faster symptomatic resolution) before peak viral loads are reached, as observed in mild influenza cases (McKimm‐Breschkin et al. in Wang et al. ⁷ ). Meanwhile to promote therapeutic efficacy in those with no recorded clinical improvement or that experience clinical deterioration within days 1 to 3 of treatment, an increased FPV dose may be required on day 4 of treatment. ³³

Consistent with PK studies in other populations, a high interpatient variability was observed across the eight patient profiles in this study. Higher C _max, C _min, and C _p were seen in the Thai data and fitted model compared to simulations of this regimen from Chinese and Japanese PK models, and dose normalized compared with observed data in a Turkish study. Multiple extrinsic and intrinsic factors affect the high interpatient variability and complex PK profile of FPV. ²⁷ In addition to the physiological conditions of patients themselves, ³⁴ AO variability is one known metabolic factor ⁵² —further influenced by drug–drug interactions, weight, and ethnicity. ³² Lower FPV concentrations in American than Japanese populations ³ and decreased PK exposure in African American populations than Chinese populations ¹⁵ suggest that pharmacogenetic factors may also affect clearance for different patient groups. ²⁷ The higher plasma thresholds observed in the Thai‐fitted model, compared to Chinese, Japanese, and Turkish data could be due to posology, pharmacogenetics, clinical condition, or other unknown factors. This may further explain FPV's clinical efficacy in reducing the time to sustained clinical improvement in a Thai population (median of 2 days in FPV arm vs 14 days in control arm), as observed in our parent study. ³⁸

There remains a discrepancy that the prior published and digitized PC‐VPC observed datasets will have been prediction‐corrected (and hence normalized) using the population model predictions from the specific model fittings to datasets in those publications. The overlay with a simulation here from a different model will not strictly be a like‐for‐like comparison. However, the major effect of dose normalization for the mix of regimens present is accounted for and the broad comparison of expected versus observed exposures remains useful. The low patient numbers and sparse sampling are another major limitation of this study. However, the data strongly suggests regional/ethnic differences in FPV's PK. Other RNA viruses have previously been shown to be susceptible to FPV, including: positive‐strand RNA viruses (i.e., Noroviruses, Flaviviruses), negative‐strand RNA viruses (i.e., Arenaviruses, Bunyaviruses, Ebola virus), and other subtypes or strains of influenza. ² , ²¹ , ³⁰ , ⁵¹ A much better understanding of the PK‐pharmacodynamic (PD) relationship is required for FPV to avoid discarding a potentially valuable intervention due to a disregard for optimal dose selection.

To conclude, clinical trials for FPV to date have been contradictory but a signal for virological effect has been demonstrated in robustly conducted phase II RCTs in the United Kingdom. ¹⁶ , ¹⁷ We hypothesize that dosing of FPV was suboptimal for SARS‐CoV‐2 in non‐Thai studies at the doses directly transposed from influenza. However, in the eight Thai adults described here, the 1800/800 mg twice‐daily regimen enabled all but one patient to reach C _p above an EC₉₀ of 25 mg/L. Dose‐normalized C _p in Thai adults and observed exposures were higher than those described in Chinese, Japanese, and Turkish populations. Irrespective of the current data, the importance of a comprehensive pharmacometrics assessment, while challenging during a major healthcare emergency, cannot be overstated so that appropriate and population‐specific doses of antiviral drugs are studied in clinical trials and potentially useful drugs are not discarded due to suboptimal dose selection.

AUTHOR CONTRIBUTIONS

IP, WM, SN, HP, AO, RS, and KC conceptualized and devised the methodology of the study; WM, RS, and LC conducted the study‘s clinical investigation; IP, SN, TRC, PM, KC and HP carried out the formal analysis of its findings; KC and KCO drafted the manuscript; and all authors assisted with its revision.

CONFLICT OF INTEREST STATEMENT

O.A. is the Director of Tandem Nano Ltd and coinventors of patents relating to drug delivery. O.A. has been coinvestigator on funding received by the University of Liverpool from ViiV Healthcare and Gilead Sciences unrelated to COVID‐19 in the past 3 years. O.A. has received personal fees from Gilead and Assembly Biosciences in the past 3 years unrelated to COVID‐19. O.A. was a member of the Trial Management Group for the AGILE phase I/II platform trial until January 2023 and AGILE received funding from Ridgeback and GSK in the past 3 years for which O.A. was not a coinvestigator. All other authors: None to declare.

ETHICS STATMENT

This sub‐study was approved by the local institutional ethics committee (EC, certificate number: Si 434/2020) and registered in the Thai Clinical Trials Registry (No. TCTR20200514001, https://thaiclinicaltrials.org/). It was conducted in accordance with the International Council on Harmonisation's Good Clinical Practice, Belmont Report, and Declaration of Helsinki. All enrolled patients provided written informed consent both to the parent study and this sub‐study's PK assessments.

Supporting information

Data S1:

Prasanchaimontri I‐o, Manosuthi W, Pertinez H, et al. Favipiravir pharmacokinetics in Thai adults with mild COVID‐19: A sub‐study of interpatient variability and ethnic differences in exposure. Pharmacol Res Perspect. 2024;12:e1233. doi: 10.1002/prp2.1233

DATA AVAILABILITY STATEMENT

The data that supports the findings of this study are available in the supplementary material of this article.