PBPK model development to study the effect of age, race and organ dysfunction
PBPK modeling and simulations were performed using Simcyp V19 simulator (Certara UK Ltd., Sheffield, UK) for clinical trial scenarios as shown in Supplementary Figure 1 . Well established or fit-for-purpose PBPK models were verified against available clinical data before being utilized to simulate scenarios where no clinical data are available. PBPK models of acalabrutinib, baricitinib, ruxolitinib, ritonavir, darunavir and dapagliflozin were previously verified extensively and they have been submitted previously as part of new drug application submission to the FDA. The rest of the PBPK models (azithromycin, atazanavir, lopinavir, chloroquine, hydroxychloroquine) were reported in peer reviewed journals with reasonable level of verification. These verified PBPK compound files were obtained from either Simcyp Simulator compound library file or repository (https://members.simcyp.com/account/libraryFiles/) or were built using reported compound input parameters. The PBPK model input parameters are shown in Supplementary Table 1 and key drug parameters required for simulating lung concentration are shown inSupplementary Table 2 . Trial designs for all simulations were set to 10 trials of 10 patients with the relevant populations constrained by age range and the proportion of females which reflected the reported study design of clinical studies where available. Dosing amounts and schedules were chosen as per the respective compound dosing recommendations and administered as an oral single-dose or multiple-dose across all simulations. The predictive performance PBPK model of probe compounds was measured by the ratio of the mean predicted AUC to the mean observed AUC where possible (Supplementary Table 3 ). For organ dysfunction simulations, AUCdiseased to AUChealthy ratios for both observed and predicted scenarios were calculated as predictive performance measures. The predictability was considered acceptable if this metric fell between the upper and lower limits (0.5 to 2-fold of observed data). Differences in exposure from race in Caucasian, Japanese and Chinese populations were predicted using Simcyp’s population library repository using the “Healthy-Volunteer,” “Japanese,” and “Chinese Healthy Volunteer” population models, respectively. Simulations for hepatic impairment were performed using Simulator’s “Cirrhosis CP-A”, “Cirrhosis CP-B”, and “Cirrhosis CP-C” populations, while for renal impairment, populations “RenalGFR_30-60”, and “RenalGFR_less_30” were used. The effect of age, race and organ impairment simulations were not simulated for baloxavir, remdesivir and large molecules, as adequate PBPK modelling was not possible due to limited data availability in the literature. Based on recent findings from IQ consortium [16] the impact of renal impairment can be well predicted within 2-fold of observed PK (AUC) and for hepatic impairment >70% of the cases were predicted within 2-fold, supporting the utility of PBPK modeling for the study of organ dysfunction.