Simultaneously Predicting the Pharmacokinetics of CES1-Metabolized Drugs and Their Metabolites Using Physiologically Based Pharmacokinetic Model in Cirrhosis Subjects

Hepatic carboxylesterase 1 (CES1) metabolizes numerous prodrugs into active ingredients or direct-acting drugs into inactive metabolites. We aimed to develop a semi-physiologically based pharmacokinetic (semi-PBPK) model to simultaneously predict the pharmacokinetics of CES1 substrates and their active metabolites in liver cirrhosis (LC) patients. Six prodrugs (enalapril, benazepril, cilazapril, temocapril, perindopril and oseltamivir) and three direct-acting drugs (flumazenil, pethidine and remimazolam) were selected. Parameters such as organ blood flows, plasma-binding protein concentrations, functional liver volume, hepatic enzymatic activity, glomerular filtration rate (GFR) and gastrointestinal transit rate were integrated into the simulation. The pharmacokinetic profiles of these drugs and their active metabolites were simulated for 1000 virtual individuals. The developed semi-PBPK model, after validation in healthy individuals, was extrapolated to LC patients. Most of the observations fell within the 5th and 95th percentiles of simulations from 1000 virtual patients. The estimated AUC and Cmax were within 0.5–2-fold of the observed values. The sensitivity analysis showed that the decreased plasma exposure of active metabolites due to the decreased CES1 was partly attenuated by the decreased GFR. Conclusion: The developed PBPK model successfully predicted the pharmacokinetics of CES1 substrates and their metabolites in healthy individuals and LC patients, facilitating tailored dosing of CES1 substrates in LC patients.


Introduction
Liver cirrhosis (LC) is widely prevalent worldwide and results from a variety of causes including obesity, non-alcoholic fatty liver disease, high alcohol consumption, hepatitis B or C infection, autoimmune diseases, cholestatic diseases and iron or copper overload [1,2].The Child-Pugh score is often used to classify liver cirrhosis into Child-Pugh A (CP-A), Child-Pugh B (CP-B) and Child-Pugh C (CP-C) according to the severity of LC [3,4].In addition to the impairment of hepatic functions, LC also leads to remarkable alterations in a series of other physiological parameters such as functional liver volume, hepatic arterial blood flow, portal venous blood flow, glomerular filtration rate (GFR), α-acid glycoprotein, albumin content, drug-metabolizing enzymes and transporters.The alterations may directly affect the pharmacokinetics of drugs [5].For example, Duthaler et al. investigated the effects of LC on the pharmacokinetics of CYP450 cocktail probes with caffeine (CYP1A2), efavirenz (CYP2B6), flurbiprofen (CYP2C9), omeprazole (CYP2C19), metoprolol (CYP2D6) and midazolam (CYP3A).They found that liver cirrhosis increased the plasma exposure of tested probes, the extent of which depended on the type of probe Pharmaceutics 2024, 16, 234 2 of 29 and LC severity.The calculated ratios of the AUC in patients to that in controls (AUCR) of caffeine, efavirenz, flurbiprofen, omeprazole, metoprolol and midazolam in CP-C patients were 6.2, 0.8, 1.4, 10.5, 4.5 and 6.3, respectively.The calculated AUCR values of omeprazole in CP-A, CP-B and CP-C patients were separately 4.8, 6.5 and 10.5.The AUCR values of probes in LC patients were in line with those in the contents of hepatic CYP450s [6].LC also affects the renal excretion and intestinal absorption of drugs.Furosemide is primarily eliminated through the kidneys.It was reported [7] that clearance (CL) of furosemide significantly decreased from 154 mL/min in control subjects to 91 mL/min in CP-B or CP-C patients, which mainly resulted from decreases in renal clearance (CL K ).These results indicate that drug dosage adjustments are necessary for LC patients based on the severity of their condition.Thus, regulatory agencies recommend pharmacokinetic studies of drugs in LC patients [8].However, conducting pharmacokinetic studies in LC patients can be both costly and time-consuming.More importantly, it is difficult to recruit patients, especially patients with CP-C.Physiologically based pharmacokinetic (PBPK) modeling is considered an ideal technique for predicting the pharmacokinetics of drugs in patients with altered physiology.The alterations in physiological parameters, expression of hepatic drug-metabolizing enzymes and transporters under various degrees of severity of LC have been demonstrated.The possibilities for predicting the pharmacokinetics of drugs in LC patients using the PBPK model have been demonstrated [9].
Carboxylesterase1 (CES1) is one of the most abundant drug-metabolizing enzymes in human livers, constituting approximately 1% of the entire liver proteome.CES1 is responsible for 80-95% of total hydrolytic activity in the liver, which mediates the metabolism of a wide range of drugs, pesticides, environmental pollutants and endogenous compounds [10].CES1-mediated metabolism leads to the biotransformation of a pharmacologically active drug into its inactive metabolite, as exemplified by methylphenidate hydrolysis.CES1 also mediates the activation of some prodrugs.The typical examples are some angiotensin-converting enzyme inhibitors (such as enalapril, cilazapril and temocapril) and neuraminidase inhibitors (oseltamivir).CES1 also hydrolyzes cholesteryl ester in lipid metabolism in human macrophages and hepatocytes, inferring that CES1 could be a potential drug target for the treatment of metabolic diseases, such as diabetes and atherosclerosis [10][11][12][13].LC has been demonstrated to significantly downregulate expressions of the hepatic CES1 protein; the CES1 contents in CP-B patients and CP-C patients were decreased to 70% and 30% of those of healthy subjects, respectively, and the CES1 enzyme content in CP-A patients was comparable to that of healthy subjects [9].On this basis, LC can alter the plasma exposure of its substrate drugs such as enalapril and oseltamivir [14,15].Moreover, it is worth noting that metabolites of most CES1 substrates (such as enalapril and oseltamivir) are mainly eliminated via renal excretion.LC also injures renal functions, leading to decreases in renal clearance of the metabolites, indicating that alterations in the plasma exposure of metabolites by LC are attributed to the integrated effects of the decreases in hepatic CES1 activity and renal clearance.
This study aimed to develop a semi-PBPK model incorporating alterations in hepatic CES1 activity, liver/renal functions, gastrointestinal transit rate and relevant organ blood flow to simultaneously predict the pharmacokinetics of CES1 drugs and their metabolites in LC patients.Clinical pharmacokinetic studies of CES1 drugs were collected from data published on PubMed based on the following criteria.(1) The tested drug must be metabolized primarily by CES1.(2) Pharmacokinetic parameters (such as AUC or plasma drug concentrations) following intravenous (i.v.) and/or oral (p.o.) administration to liver cirrhosis populations must be available.(3) The clinical pharmacokinetic data might come from different reports.Based on these criteria, nine CES1 substrates were included in the simulations.The nine drugs are primarily metabolized by CES1 and include six prodrugs (enalapril, benazepril, cilazapril, perindopril, temocapril and oseltamivir) and three direct-acting drugs (flumazenil, pethidine and remimazolam).Flumazenil and remimazolam are mainly administered by intravenous injection.Pethidine is administrated via intravenous or oral routes.The remaining drugs are administered as oral immediate-release formulations.The predicted results were compared with clinical studies in patients with different statuses of LC.These results will assist in tailoring dosages of CES1 substrates in LC patients.

General Workflow
The workflow for developing a PBPK model (Figure 1) for LC patients.Initially, a semi-PBPK model (Figure 2) was developed for a virtual population of healthy individuals validated using clinical pharmacokinetic studies in healthy subjects.Then, the developed PBPK model was translated to LC patients by replacing the values of system-specific model parameters.Finally, pharmacokinetic predictions were conducted in 1000 virtual patients individuals and compared with clinical pharmacokinetic data from the literature.
intravenous or oral routes.The remaining drugs are administered as oral immediate-release formulations.The predicted results were compared with clinical studies in patients with different statuses of LC.These results will assist in tailoring dosages of CES1 substrates in LC patients.

General Workflow
The workflow for developing a PBPK model (Figure 1) for LC patients.Initially, a semi-PBPK model (Figure 2) was developed for a virtual population of healthy individuals validated using clinical pharmacokinetic studies in healthy subjects.Then, the developed PBPK model was translated to LC patients by replacing the values of system-specific model parameters.Finally, pharmacokinetic predictions were conducted in 1000 virtual patients individuals and compared with clinical pharmacokinetic data from the literature.

Model Development
A semi-PBPK model was developed to simultaneously predict the pharmacokinetics of CES1 substrate drugs and their metabolites in LC patients.The semi-PBPK model consists of the stomach, intestinal wall, intestinal lumen, portal vein, liver, kidney and systemic compartment, which are connected by the blood circulatory system.The elimination of most drugs mainly occurs in the liver and kidneys.Drugs are administrated via the intravenous route or oral route.It is generally accepted that absorption of most orally administered drugs may occur in the small intestine (duodenum, jejunum and ileum).Absorbed amounts of drugs in the stomach, caecum and colon are minor.The effective permeability coefficient (P eff ) is used to indicate the absorption capacity of a drug [16].In the simulation, it was assumed that elimination of the tested drugs only occurred in the liver and kidneys and absorption of drugs only occurred in the small intestine.

Model Development
A semi-PBPK model was developed to simultaneously predict the pharmacokinetics of CES1 substrate drugs and their metabolites in LC patients.The semi-PBPK model consists of the stomach, intestinal wall, intestinal lumen, portal vein, liver, kidney and systemic compartment, which are connected by the blood circulatory system.The elimination of most drugs mainly occurs in the liver and kidneys.Drugs are administrated via the intravenous route or oral route.It is generally accepted that absorption of most orally administered drugs may occur in the small intestine (duodenum, jejunum and ileum).Absorbed amounts of drugs in the stomach, caecum and colon are minor.The effective permeability coefficient (Peff) is used to indicate the absorption capacity of a drug [16].In the simulation, it was assumed that elimination of the tested drugs only occurred in the liver and kidneys and absorption of drugs only occurred in the small intestine.
All available information on anatomical, physiological and ADME parameters of the tested drugs was collected for the initial model construction (Tables 1 and 2).Coding and solving of the PBPK model were conducted on WinNonlin 8.1 (Pharsight, St. Louis, MO, USA).The specific code and formulas for the model can be found in the supplementary material.After developing the initial model, parts of the plasma concentration curves of drugs from healthy subjects were used to estimate and optimize some parameters.Subsequently, the developed PBPK model was validated using plasma concentration-time curves from the rest of the clinical studies.All available information on anatomical, physiological and ADME parameters of the tested drugs was collected for the initial model construction (Tables 1 and 2).Coding and solving of the PBPK model were conducted on WinNonlin 8.1 (Pharsight, St. Louis, MO, USA).The specific code and formulas for the model can be found in the Supplementary Material.After developing the initial model, parts of the plasma concentration curves of drugs from healthy subjects were used to estimate and optimize some parameters.Subsequently, the developed PBPK model was validated using plasma concentration-time curves from the rest of the clinical studies.

PBPK Model Development in LC Patients
The anatomical and physiological parameters in healthy subjects were replaced with those (Table 1) in LC patients.The LC-induced alterations in parameters related to ADME were estimated according to their values in healthy (HT) subjects and the altered physiological parameters.
For CES1-mediated hepatic metabolism, CL int,CI,CES1 = CL int,HT,CES1 × f CES1 × f liver (1) where CL int,CI,CES1 and CL int,HT,CES1 represent the values of CES1-mediated intrinsic clearance in the liver of patients and healthy subjects, respectively.f CES1 and f liver represent the ratio of CES1 content in patients to that in healthy subjects and liver volume in patients to that in healthy subjects, respectively.For hepatic elimination of drugs mediated by other routes, CL int,CI,other = CL int,HT × f other × f liver (2) where CL int,cirr,other and CL int,heal,other represent the values of intrinsic clearance by other routes in the liver of patients and healthy subjects, respectively.f other is the ratio of other targets' content in patients to that in healthy subjects.e : Assumed values; f : Simulation by WinNonlin, cilazapril and cilazaprilat using 0.5 mg dose pharmacokinetic and remimazolam using 0.025 mg/kg dose pharmacokinetic in simulation; g : Calculated by WinNonlin, flumazenil using T.F.pharmacokinetic to calculate; h : CES1-mediated CL int ; i : CYP2B6-mediated CL int ; j : UGT intrinsic clearance of perindopril.
Among the tested drugs, pethidine binds mainly to α1-acid glycoprotein and the rest bind mainly to albumin [88,[93][94][95][96][97][98][99] (no data on binding protein for temocapril, so binding to albumin was assumed based on pka < 7.4, acidic).The free fraction of drugs in patient plasma was estimated using Equation (3) [21]: where f u,p,CI , f u,p,HT , P prot,CI and P prot,HT represent the unbound fraction of the drug in the plasma of patients and healthy subjects and the concentration of drug-bound proteins in the plasma of patients and healthy subjects, respectively.It was assumed that the free apparent volume of the distribution of the drug is unaltered; the apparent volume of distribution in cirrhosis patients (V sys,CI ) was derived from the apparent volume of distribution in healthy subjects, i,e., Liver cirrhosis also impairs renal function and is characterized by a decrease in the glomerular filtration rate (GFR).The renal intrinsic clearance (CL int,K,CI ) in patients may be estimated using equation [17]: where CL int,k,HT , GFR HT and GFR CI represent renal intrinsic clearance in healthy subjects and GFR in healthy subjects and patients, respectively.LC patients are often accompanied by impairment of the intestinal barrier [100].The Lactulose/Rhamnose ratio is used to assess intestinal permeability [26].The ratio of cirrhosis patients to healthy subjects was used to correct the absorption rate constant in LC patients: P eff,CI = P eff,HT × LR CI /LR HT (6) where P eff,CI and P eff,HT are P eff values in LC patients and healthy subjects, respectively.LR CI and LR HT are, respectively, the Lactulose/Rhamnose ratios in LC patients and healthy subjects.The four virtual populations (normal population, CP-A, CP-B and CP-C patients) were included in the simulations, each of which contained 1000 virtual individuals.For virtual population validation, each virtual individual was generated independently.CL int , CL int,K , f u,b , V system , P eff , k a , K L:P , K G:P , and K K:P were used to generate virtual individuals.A random individual could be generated by taking random values in the range of 80-120% of the above parameter values.The 5th and 95th percentiles and average values of the simulation derived from 1000 virtual subjects were obtained.Effects of cirrhosis on the plasma exposure of the tested drugs were indexed as AUCR or C max R AUCR = AUC CI AUC HT (7) Or AUCR = CL HT CL CI (8) where AUC CI , AUC HT , CL CI , CL HT , C max,CI and C max,HT are, respectively, the AUC, CL and C max of the tested drugs in cirrhosis patients and healthy subjects.

Criterion of the Developed PBPK Model
The PBPK model was considered to be successful if the simulated AUC or C max fell within 0.5-to 2-fold of the observed data or the observed data were within the 5th and 95th percentiles of the simulation derived from 1000 virtual subjects [101].

Drug Data Set
Nine CES1 drugs, including six prodrugs (enalapril, benazepril, cilazapril, perindopril, temocapril and oseltamivir) and three direct-acting drugs (flumazenil, pethidine and remimazolam), were collected from data published on PubMed based on the following criteria.(1) The tested drug must be metabolized primarily by CES1.(2) Pharmacokinetic parameters (such as AUC or plasma drug concentrations) following intravenous (i.v.) and/or oral (p.o.) administration to liver cirrhosis populations must be available.(3) The clinical pharmacokinetic data might come from different reports.The collected pharmacokinetic parameters and drug information on clinical reports are listed in Tables 2 and 3, respectively.

Enalapril and Enalaprilat
Enalapril, an angiotensin-converting enzyme inhibitor (ACEI), is a prodrug, which is mainly metabolized to the active product enalaprilat via hepatic CES1 [12,102].Enalaprilat is eliminated primarily through the kidneys [103].In plasma, enalapril and enalaprilat are mainly bound to albumin, and their free fractions in plasma are 0.55 and 0.5 [33].Five clinical reports, including two reports involving liver cirrhosis, were selected in the simulations.

Benazepril and Benazeprilat
Benazepril, a prodrug, is metabolized by hepatic CES1 to the active product benazeprilat [12,102], which shows inhibition of angiotensin-converting enzyme (ACE).Benazeprilat is eliminated via renal excretion.Benazepril and benazeprilat are mainly bound to albumin, belonging to drugs with high plasma binding, and their free fractions in plasma are 0.03 and 0.05 [47], respectively.Six clinical reports, including one report involving liver cirrhosis, were selected in the simulations.

Cilazapril and Cilazaprilat
Cilazapril is also metabolized by hepatic CES1 into cilazaprilat [12,102].Cilazaprilat is mainly eliminated via the kidneys [52].Cilazapril and cilazaprilat are mainly bound to albumin, belonging to medium plasma-binding drugs, and their free fractions in plasma are 0.70 and 0.76 [50], respectively.Six clinical reports, including one report involving liver cirrhosis, were selected in the simulations.

Perindopril and Perindoprilat
The prodrug perindopril is mainly metabolized by hepatic CES1 to perindoprilat, which shows inhibition of ACE.The bioavailability of perindopril is 66% [64].Perindopril is primarily converted to perindoprilat in the liver, and other major metabolites of perindopril are perindopril glucuronide and perindopril lactam [142].Since it is not clear which isoenzyme of UGT metabolizes perindopril to perindopril glucuronide, the change rate of AUC 0-inf (0.62) for metoprolol in cirrhosis was used as a variation coefficient of intrinsic clearance for UGT [143].Perindoprilat is eliminated via renal excretion.Perindopril and perindoprilat are predominantly bound to albumin.Perindopril shows higher plasma binding (percent binding 60%) than perindoprilat (mean percent binding 15%) [72].
Four clinical reports, including two reports involving liver cirrhosis, were selected in the simulations.Cirrhosis in perindopril and perindoprilat only have pharmacokinetic parameters and no specific drug concentration-time profile, so only a comparison of parameters was made.

Temocapril and Temocaprilat
Temocapril is also a prodrug and metabolized by hepatic CES1 to temocaprilat.Temocaprilat is eliminated via both bile and the kidneys.The biliary clearance of temocaprilat was about two times the renal clearance [65].The CL int,K of temocaprilat was calculated to be 949.84mL/min [64].Thus, the CL bile,m of temocaprilat was estimated to be 1899.68mL/min, assuming that the ratio of CL bile,m to CL int,K was 2.0.Biliary excretion of temocaprilat is considered to be mediated by multidrug resistance-associated protein2 (MRP2) [144].One clinical report involving both liver cirrhosis patients and healthy subjects was selected in the simulations.

Oseltamivir and Oseltamivir Carboxylate
Oseltamivir, a prodrug, is metabolized via hepatic CES1 [12,102] to its active metabolite oseltamivir carboxylate (OC), which has an antiviral effect.About 80% of an orally administered dose of oseltamivir reaches the systemic circulation as the active metabolite.The absolute bioavailability of the active metabolite from orally administered oseltamivir is 75% [145].About 60 to 70% of an oral oseltamivir dose appears in urine as the active metabolite and less than 5% as oseltamivir.Oseltamivir carboxylate is primarily eliminated via renal excretion, accounting for 93% of intravenous doses [38].The CL int,K values of both oseltamivir and oseltamivir carboxylate exceed the GFR, indicating that renal elimination occurs via a combination of glomerular filtration and renal tubular secretion.Both oseltamivir and oseltamivir carboxylate are primarily bound to albumin; their bound fractions in plasma were approximately 42% and less than 3% [36].Four clinical reports, including one report involving liver cirrhosis, were selected in the simulations.

Flumazenil
Flumazenil, a benzodiazepine receptor antagonist, is usually administered by intravenous injection [83].Flumazenil is inactivated by hepatic CES1 to flumazenil acid and probably by CYP450-catalyzed N-dealkylation to N-demethylated flumazenil [146].Flumazenil is predominantly bound to serum albumin, and its plasma protein binding is about 40% [85].Five clinical reports, including two reports involving liver cirrhosis, were selected in the simulations.

Pethidine
Pethidine (meperidine) is a synthetic opioid commonly used for analgesia in humans.Pethidine is metabolized in the body by two different pathways [88,102].The primary pathway is hepatic CES1 metabolism to pethidinic acid, an inactive metabolite.Another pathway is N-demethylation by CYP2B6 to normeperidine, a nonopioid active metabolite.The oral bioavailability of pethidine varies from 48-56% [147].Pethidine was predominantly bound to α1-acid glycoprotein.In the simulation for healthy subjects, the free fraction of pethidine in plasma was 0.418 [88].Ten clinical reports, including four reports involving liver cirrhosis, were selected in the simulations.

Remimazolam
Remimazolam, an ultrashort-acting sedative agent, is metabolized by hepatic CES1 to an inactive carboxy acid metabolite.The plasma protein binding of remimazolam is approximately 92%, predominantly serum albumin [77].In the clinic, remimazolam is normally administered intravenously.Two clinical reports, including one report involving liver cirrhosis, were selected in the simulations.

Prediction of
Extents of pharmacokinetic parameters under liver cirrhosis, AUCR and C max R were also predicted using the estimated pharmacokinetic parameters (Figures 5 and 6).AUC or C max values may come from different clinical reports or different doses, thus, the AUC or C max values were normalized by dose and their mean values were used for estimating the AUCR or C max R. The results showed that the vast majority of the ratios of predicted AUCR and C max R are close to observed values, with only a few individual values differing significantly, indicating a good prediction.All these show that the PBPK model successfully predicted the pharmacokinetics of drugs in cirrhosis.

Sensitivity Analysis of Model Parameters
The plasma concentration-time curve of enalapril and enalaprilat following oral administration (10 mg) was used as an example for pharmacokinetic sensitivity.Some parameters such as gastrointestinal motility rate (K t ), intestinal absorption (P eff ), hepatic arterial blood flow rates (Q LA ), portal vein blood flow rates (Q PV ), hepatic CES1 activity (CL int,L ), kidney blood flow rates (Q K ), GFR, f u,b and f u,b,m (free fraction of metabolites in blood) may affect the pharmacokinetics of drugs and were selected for sensitivity analysis.According to the variations in the corresponding parameters listed in Table 1, the variations of Q PV and Q K were set to be 1/2-, 1-and 2-fold; Q LA and CL int,L were 1/3-, 1-and 3-fold; variation in GFR was 0.5-, 1-and 1.5-fold; variation in f u,b was 0.7-, 1-and 1.3-fold for enalapril; and f u,b,m was 0.7-, 1-and 1.3-fold for enalaprilat.A report showed that K t values under diabetic status were lower by about 2-fold compared to healthy subjects [148].Table 1 also showed that K t values under liver cirrhosis were about 1.3-fold those of healthy subjects.Here, variations of K t were set to be 1/2-, 1-and 2-fold.Highly different P eff values of enalapril were reported [30,149,150].For example, Thoms et al.'s reported P eff value of enalapril was 0.00125 cm/min [149], while the P eff value of enalapril reported by Chaturvedi et al. was 0.0127 cm/min [150].Thus, variations in the P eff values of enalapril were set to be 1/3-, 1-and 3-fold.The results (Figure 7) show that these tested parameters affect the pharmacokinetics profile of drugs in varying degrees; their contributions to the AUC of enalapril were P eff > CL int,L > K t > f u,b > Q PV > GFR > Q K > Q LA and to that of enalaprilat were P eff > GFR > CL int,L > K t > f u,b,m > Q PV > Q K > Q LA .In addition to impairment of liver failure, LC patients were associated with increases in intestinal transit rates, intestinal permeability of drugs, Q LA and f u,b (due to decreases in plasma-binding protein levels) and decreases in GFR, Q K , CES1 activity and Q PV , although increases in Q L were reported in CP-C patients.The contributions of LC-induced alterations in K t , Q PV , CL int,L , P eff , GFR, Q K and f u,b to the plasma concentrations of enalapril and enalaprilat following an oral dose of enalapril maleate (10 mg) administered to CP-C patients and their integrated effects were also simulated.The results showed that decreases in the CL int,L and increases in the P eff of enalapril increased plasma concentrations of enalapril, while the increases in f u,b and K t and decreases in Q PV obviously decreased plasma concentrations of enalapril following an oral dose of enalapril maleate; the net effects were an increase in the plasma concentrations of enalapril.For enalaprilat, increases in P eff and decreases in GFR, Q K and Q PV significantly increased the plasma concentration profiles of enalaprilat, while decreases in CES1 activity and increases in the K t and f u,b,m of enalaprilat significantly decreased plasma concentrations following oral enalapril maleate administration.Their net effects were to decrease plasma concentrations of enalaprilat (Figure 7Q,R).

Sensitivity Analysis of Model Parameters
The plasma concentration-time curve of enalapril and enalaprilat following oral administration (10 mg) was used as an example for pharmacokinetic sensitivity.Some parameters such as gastrointestinal motility rate (Kt), intestinal absorption (Peff), hepatic arterial blood flow rates (QLA), portal vein blood flow rates (QPV), hepatic CES1 activity (CLint,L), kidney blood flow rates (QK), GFR, fu,b and fu,b,m (free fraction of metabolites in blood) may affect the pharmacokinetics of drugs and were selected for sensitivity analysis.According to the variations in the corresponding parameters listed in Table 1, the variations of QPV and QK were set to be 1/2-, 1-and 2-fold; QLA and CLint,L were 1/3-, 1-and 3fold; variation in GFR was 0.5-, 1-and 1.5-fold; variation in fu,b was 0.7-, 1-and 1.3-fold for enalapril; and fu,b,m was 0.7-, 1-and 1.3-fold for enalaprilat.A report showed that Kt values under diabetic status were lower by about 2-fold compared to healthy subjects [148].Table 1 also showed that Kt values under liver cirrhosis were about 1.3-fold those of healthy subjects.Here, variations of Kt were set to be 1/2-, 1-and 2-fold.Highly different Peff values of enalapril were reported [30,149,150].For example, Thoms et al.'s reported Peff value of

Discussion
Hepatic CES1 mediates the inactivation of direct-acting drugs or the activation of some prodrugs, most of whose active metabolites are mainly eliminated via the kidneys.In addition to hepatic dysfunction, LC is also associated with alterations in organ blood flow, decreases in plasma protein levels, increases in intestinal permeability of drugs and impairment of renal functions, commonly affecting the pharmacokinetics of CES1 substrate drugs and their metabolites.Both the whole-PBPK model and the semi-PBPK model have been widely applied to predict the pharmacokinetics of drugs, but compared with the whole-PBPK model, semi-PBPK model needs fewer parameters without losing key dynamic information [151], which may avoid overparameterization in the whole-PBPK model.Moreover, the semi-PBPK model may avoid some of the parameter estimation difficulties of whole-PBPK models [152].The main contributions of the study were the successful development of a semi-PBPK model involving intestinal absorption, hepatic metabolism and renal excretion to simultaneously predict the pharmacokinetic profiles of nine CES1 substrates (six prodrugs and three direct-acting drugs) in both healthy subjects and LC patients.Most clinical observations were within the 5th and 95th percentiles of simulations derived from 1000 virtual subjects.Most of the estimated AUC and C max values were also within 0.5-2.0-fold of observations.
The extent of LC-induced alterations in the plasma exposure of CES1 substrates and their metabolites was also assessed using AUCR and C max R. It was found that although most of the clinically observed plasma concentrations for the tested agents were within the 5th and 95th percentiles of simulations, poorly predicted AUCR or C max R values were found in benazepril, temocaprilat, perindopril and perindoprilat.The predicted AUCR values of flumazenil and pethidine were lower than the clinical observations.Benazepril and temocaprilat belong to highly bound compounds, and their f u,b values were 0.03 and 0.025, respectively.In general, it is difficult to obtain an accurate plasma-binding measurement for highly bound compounds [153].In addition to CES1, UGTs also mediate perindopril metabolism [142].The isoenzyme of UGT involved in the metabolism of perindopril has not been identified.In the simulation, it was assumed that LC-induced alterations in the CL int, UGT of perindopril were similar to that of metoprolol [143].LC patients with different etiologies show different amounts of hepatic CES1.In addition to CES1, other enzymes also mediate the metabolism of flumazenil [146].Pethidine is cometabolized by CES1 and CYP2B6 [88,102].Several reports have demonstrated extensive interindividual variability in the expression of CYP2B6 [154] and CES1 [102].All of these factors may be reasons leading to the differences between the predicted and the observed AUCR values, which need further investigation.
In general, LC-induced impairments of hepatic CES1 activity increase the plasma exposure of CES1 substrates, but sensitivity analysis revealed that the increases in the plasma concentrations of CES1 substrates in LC patients were only partially attributed to the impairment of hepatic CES1.Increases in the intestinal permeability of drugs were also observed in LC patients, contributing to increased plasma exposures of CES1 substrates.In contrast, LC-induced increases in intestinal transit rate and decreases in plasma-binding proteins and Q PV obviously decreased the plasma exposure of CES1 substrates, which partly attenuated the increases in the plasma exposures of CES1 substrates caused by liver cirrhosis.Metabolites of the tested CES1 substrates are eliminated via the kidneys.The decreases in the plasma exposure of metabolites induced by the impairment of hepatic CES1 activity were also partly attenuated by LC-induced alterations in GFR and Q K .Even under some conditions, levels of the metabolites are increased rather than decreased due to impaired renal function.For example, the AUC values of perindoprilat in CP-A and CP-B patients were obviously higher than those in healthy individuals; the observed AUCR values were 2.89 and 1.2, respectively, which were near to predictions (1.98 in CP-A patients and 2.04 in CP-B patients).These findings may partly explain clinical findings that although liver cirrhosis obviously increases the plasma levels of enalapril and perindopril, the magnitude of serum ACE-lowering effects by the two drugs was fairly comparable between LC patients and healthy subjects [14,120,121].
Plasma levels of the direct-acting drugs flumazenil, pethidine and remimazolam following their administration to LC patients were also successfully simulated.The observed AUCR values of remimazolam in LC patients could not be calculated due to a lack of observed pharmacokinetic parameters in LC patients, contrasting our expectation that the AUCR values in CP-B patients and CP-C patients would be 0.76 and 0.61, which may be explained by the fact that the increased plasma concentration by the impairment of hepatic CES1 may be attenuated by increases in hepatic arterial blood flow and increases in f u,b (Figure S2).The above simulations showed that the LC-induced impairments of hepatic CES1 activity may increase plasma levels of CES1 substrates (parent drug) and decrease plasma levels of their metabolites, if dosage adjustments are dependent on characteristics.For example, although LC obviously increases the plasma levels of enalapril and perindopril, the levels of enalaprilat and perindoprilat and the extent of decreases in serum ACE activity were obviously unaltered [14,120,121], indicating that no dosage adjustment of enalapril and perindopril in LC patients is required.The simulated levels of pethidine in the plasma of LC patients were higher than those in healthy subjects, which explained why the results in LC patients were consistent with the clinical observation that LC enhanced the CNS toxicity of pethidine [155], indicated that reduced dosages of pethidine in patients with hepatic insufficiency are needed [156].
However, this study also has some shortcomings.The predictions for healthy subjects were based on "ideal" healthy subjects (body weight assumed to be 70 kg) without considering gender, body weight, race and genetic variance of CES1.Genetic variation in CES1 also affects the pharmacokinetics of CES1 substrates [102].During the simulation in LC patients, LC patients were considered "ideal" CP-A, CP-B or CP-C patients without considering LC etiology, gender and race.It was reported that the amount of CES1 protein in patients with hepatitis C cirrhosis was approximately 1.47-fold that of patients with alcoholic cirrhosis [157].Similarly, it was reported that flumazenil might improve memory in patients with alcoholic cirrhosis but not in patients with nonalcoholic cirrhosis [158].Moreover, the mean absolute CES1 protein expression in female livers was reported to be 17.3% higher than that in male livers [159].
LC patients are often accompanied by impairment of the intestinal barrier and renal function.LC may impair the intestinal barrier and renal function via various mechanisms [100,160].The most common causes of LC are chronic liver diseases related to alcohol consumption, hepatitis virus infection, obesity and/or usage of drugs.Alcohol and drugs may directly impair the intestinal barrier.LC also leads to microbial alterations, which affect the intestinal epithelial barrier function directly or indirectly.For example, increased endotoxin levels directly downregulate the expression of intestinal tight junctions.Portal hypertension is a severe consequence of cirrhosis, which may lead to ascites, variceal hemorrhage and an impaired intestinal barrier [100].LC may impair renal functions via activating the renin-angiotensin system, the sympathetic nervous system or nonosmotic hypersecretion of arginine vasopressin.Moreover, the translocation of bacteria and bacterial products from the intestinal lumen to the mesenteric lymph nodes stimulates inflammatory responses, increasing the production of proinflammatory cytokines.Moreover, the increased circulating levels of endotoxin or bacterial DNA also increase serum levels of cytokines, in turn, impairing renal function [160].

Conclusions
The developed PBPK model may successfully be applied simultaneously to predict the pharmacokinetics of CES1 substrate drugs and their active metabolites in healthy subjects and LC patients.The impact of physiological alteration under different degrees of LC on the pharmacokinetic behaviors of drugs may be accurately simulated.The simulated results will help in deciding whether the dosage of CES1 substrates should be adjusted for LC patients.

Figure 1 .
Figure 1.Workflow for developing a semi-PBPK model.It involved establishing a PBPK model in normal subjects and validating it with a virtual population.Afterward, the parameters were changed according to the effects of cirrhosis and a model of PBPK in cirrhosis patients was created.Simulations were performed in virtual populations and compared with clinical pharmacokinetic data.

Figure 1 .
Figure 1.Workflow for developing a semi-PBPK model.It involved establishing a PBPK model in normal subjects and validating it with a virtual population.Afterward, the parameters were changed according to the effects of cirrhosis and a model of PBPK in cirrhosis patients was created.Simulations were performed in virtual populations and compared with clinical pharmacokinetic data.

Figure 2 .
Figure 2. Schematic structure of the semi-PBPK model.Kti represents the gastric emptying rate and intestinal transit rate.GWi represents the gut wall of the duodenum, jejunum and ileum.kai represents the rate of drug absorption into the gut wall.QGWi represents the blood flow rate in the gut wall.QLA, QL and QPV represent the hepatic artery blood flow rate, hepatic blood flow rate and portal vein blood flow rate, respectively.CLint, CLbile and CLint,K represent the intrinsic hepatic clearance, biliary intrinsic clearance and renal intrinsic clearance, respectively.

Figure 2 .
Figure 2. Schematic structure of the semi-PBPK model.K ti represents the gastric emptying rate and intestinal transit rate.GWi represents the gut wall of the duodenum, jejunum and ileum.k ai represents the rate of drug absorption into the gut wall.Q GWi represents the blood flow rate in the gut wall.Q LA , Q L and Q PV represent the hepatic artery blood flow rate, hepatic blood flow rate and portal vein blood flow rate, respectively.CL int , CL bile and CL int,K represent the intrinsic hepatic clearance, biliary intrinsic clearance and renal intrinsic clearance, respectively.

Figure 7 .
Figure 7. Sensitivity analysis of enalapril and enalaprilat following oral 10 mg enalapril maleate.Enalapril: (A) K t ; (B) CL int,L ; (C) GFR; (D) Q LA ; (E) Q PV ; (F) Q K ; (G) f u,b ; (H) P eff ; Enalaprilat: (I) K t ; (J) CL int,L ; (K) GFR; (L) Q LA ; (M) Q PV ; (N) Q K ; (O) f u,b,m ; (P) P eff .f u,b varies by 0.7-fold and 1.3-fold; f u,b,m varies by 0.7-fold and 1.3-fold; GFR varies by 0.5-fold and 1.5-fold; K t , Q PV and Q K are varied by 1/2-fold and 2-fold; and the rest are varied by 1/3-fold and 3-fold.Individual contributions of LC-induced alterations in K t , CES1 activity, GFR, f u,b , P eff , Q K and Q PV to plasma concentrations of enalapril (Q) and enalaprilat (R) following oral 10 mg enalapril maleate administration to LC patients and their integrated effects.

Table 1 .
Physiological parameters used in the physiologically based pharmacokinetic model in adults with and without cirrhosis.

Table 2 .
Simultaneously predicting the pharmacokinetics of CES1-metabolized drugs and their metabolites using the physiologically based pharmacokinetic model.
a : Data from www.drugbank.com,accessed on 4 February 2024; b : Bile intrinsic clearance of temocaprilat; c : Recalculated from CL L,b ; d : Calculations using Rodgers-Rowland method;

Table 3 .
Clinical information about CES1 substrates in the simulations.

Table 4 .
Observed and predicted values of AUC 0-t and C max of enalapril and enalaprilat following oral enalapril maleate administration to healthy (HT) subjects and liver cirrhosis patients.

Table 5 .
Observed and predicted values of AUC 0-t and C max of benazepril and benazepril following benazepril hydrochloric administration to healthy (HT) subjects and cirrhosis.
NR: Not reported.

Table 6 .
Observed and predicted values of AUC 0-t and C max of cilazapril following oral cilazapril to healthy (HT) subjects and LC patients.
NR: Not reported.

Table 7 .
Observed and predicted values of AUC 0-t and C max of perindopril following oral perindopril tert-butylamine administration to healthy (HT) subjects and LC patients.
NR: Not reported.

Table 8 .
Observed and predicted values of AUC 0-t and C max of temocapril and temocaprilat following oral temocapril hydrochloride administration to healthy (HT) subjects and LC patients.
NR: Not reported.

Table 9 .
Observed and predicted values of AUC 0-t and C max of oseltamivir and oseltamivir carboxylate (OC) following oral oseltamivir phosphate administration to healthy (HT) subjects and cirrhosis.

Table 10 .
Observed and predicted values of AUC 0-t (µg × h/mL) or CL (L/min) and C max (ng/mL) of flumazenil to healthy (HT) subjects and LC patients.

Table 11 .
Observed and predicted values of AUC 0-t (µg × h/mL) or CL (L/min) and C max (ng/mL) of pethidine following oral and intravenous pethidine HCl administration to healthy (HT) subjects and LC patients.

Table 12 .
Observed and predicted values of AUC 0-t of remimazolam following intravenous remimazolam besylate administration to healthy (HT) subjects and LC patients.
NR: Not reported.