A Digital Twin of the Angiotensin II Receptor Blocker Losartan: Physiologically Based Modeling of Blood Pressure Regulation

1. Introduction

Hypertension, defined as blood pressure of 130 mmHg or higher and/or a diastolic blood pressure of 80 mmHg or higher, is a major global health challenge and leading risk factor for cardiovascular disease, stroke, and kidney failure [1,2,3]. In 2019, over 1.2 billion people worldwide were affected, with prevalence exceeding 50% in some regions [2]. Despite its silent clinical course, hypertension contributes to over 8 million deaths annually [3]. Risk factors include age, genetics, obesity, inactivity, and unhealthy diet [1]. While treatment options exist, improved prevention and personalized therapies remain essential.

The renin-angiotensin-aldosterone system (RAAS) regulates blood pressure and fluid balance. Renin release from the kidney cleaves angiotensinogen to angiotensin I, which is converted by the angiotensin converting enzyme (ACE) into angiotensin II [4]. Angiotensin II raises blood pressure via vasoconstriction, stimulation of aldosterone and vasopressin release, enhanced sodium reabsorption, and increased sympathetic tone [5]. Dysregulation of the RAAS underlies hypertension, heart failure, and kidney disease.

Pharmacological treatments target RAAS at multiple points. Beta-blockers lower renin, direct renin inhibitors (e.g., aliskiren) block its activity, ACE inhibitors (e.g., ramipril) prevent angiotensin II formation, and angiotensin II receptor blockers (ARBs) (e.g., losartan) selectively antagonize AT1 receptors [5,6,7,8]. ARBs offer high specificity by blocking receptor-mediated effects of angiotensin II without interfering with its synthesis. Other antihypertensive drug classes include diuretics and calcium channel blockers [9,10].

Losartan, the first ARB approved in 1995, is widely prescribed for hypertension and related complications [11,12]. By competitively blocking AT1 receptors, it reduces vasoconstriction, aldosterone secretion, and water retention. Its major active metabolite, E3174, is generate through CYP2C9 mediated conversion and 10–40 times more potent [13], though losartan itself remains an effective antagonist. Both compounds contribute to clinical efficacy. Losartan is well tolerated, with dizziness and mild respiratory symptoms as the most common side effects [12]. It is available as monotherapy or combined with hydrochlorothiazide for greater blood pressure reduction [14]. Standard dosing ranges from 25–100 mg/day. Like other ARBs, it is contraindicated in pregnancy [11].

Losartan shows rapid oral absorption, with peak plasma levels after 1–2 h and bioavailability of about 33% due to first-pass metabolism [13,15]. E3174 peaks later (3–4 h) and has 4–8-fold higher systemic exposure [13,16]. Both compounds bind strongly to plasma proteins and are eliminated via hepatic metabolism and renal/fecal excretion, with terminal half-lives of 2 h (losartan) and 4–6 h (E3174) [17,18]. Only small fractions are excreted unchanged renally [19]. Minor metabolites such as L158 reflect additional metabolic pathways [18,20].

Pharmacodynamically, losartan lowers blood pressure by blocking AT1 receptors, thereby reducing aldosterone secretion [21,22,23]. AT1 blockade also increases renin and angiotensin levels due to loss of feedback control [24,25]. Enhanced stimulation of AT2 receptors may further contribute to vasodilation [26].

Liver cirrhosis reduces losartan clearance by about 50% and doubles bioavailability, leading to higher plasma levels, though E3174 exposure increases only modestly [11,27]. Renal impairment decreases clearance of losartan and E3174 but generally does not elevate E3174 levels due to compensatory mechanisms [19,28,29]. Neither compound is dialyzable, and dose adjustments are mainly required in hepatic dysfunction [30].

Drug response varies due to genetic polymorphisms in ABCB1 and CYP2C9 genes. Variants of ABCB1, a gene encoding for a drug efflux transporter also known as P-glycoprotein or MDR1, influence losartan absorption and blood pressure response, though findings are inconsistent [31,32,33]. CYP2C9 polymorphisms (*2, *3, *13) reduce metabolism to E3174, increasing losartan exposure and diminishing therapeutic effect [13,34,35,36,37]. The frequency of these alleles varies by population [38,39,40].

Physiologically-based pharmacokinetic/ pharmacodynamic (PBPK/PD) models integrate ADME processes (absorption, distribution, metabolism and elimination) with drug effects using differential equations [41]. They allow simulation of dosing regimens, organ dysfunction, and genetic variability on drug behavior. In this work, a PBPK/PD model of losartan was developed to investigate dose-dependent PK/PD effects, the influence of hepatic and renal impairment and the impact of ABCB1 and CYP2C9 variants. The aim is to clarify sources of variability in losartan response and support individualized therapy.

2. Materials and Methods

2.1. Systematic Literature Research

A systematic literature search was performed in PubMed and PKPDAI [42] on 27 August 2024 using the terms losartan AND pharmacokinetics. Eligible studies included human clinical trials with PK/PD data in healthy volunteers, hepatic/renal impairment, or CYP2C9 genotypes. Excluded were animal studies, pediatric/ single-patient reports, reviews, computational models, cocktail or combination designs, and PD-only studies. Where data were redundant, representative studies were selected. Additional in vitro reports were included to derive kinetic parameters.

Figure 1. PRISMA flow diagram of the literature search and study selection process.

Figure 1. PRISMA flow diagram of the literature search and study selection process.

2.2. Data Curation

Relevant study data were curated into PK-DB [43]. Extracted metadata included group and individual characteristics (age, sex, genotype, comorbidities), interventions (dose, route, regimen), and outcomes (concentration-time profiles of losartan and metabolites, RAAS biomarkers, blood pressure, heart rate). Digitization of graphical data was performed using WebPlotDigitizer [44]. Data were organized according to PK-DB standards into groups, individuals, interventions, and time courses, providing the heterogeneous dataset used for modeling and validation.

2.3. Computational Model

A physiologically-based pharmacokinetic/pharmacodynamic (PBPK/PD) model was built in the Systems Biology Markup Language (SBML) [45,46] using sbmlutils [47], visualized using cysbml [48], simulated with sbmlsim [49] based on libroadrunner [50,51], and shared under CC-BY 4.0 at Zenodo (v0.7.1) [52]. The model consists of submodels for intestine, liver, kidney, and RAAS, connected by systemic circulation. Hepatic impairment was implemented as progressive cirrhosis [53,54], aligned with Child-Pugh classes [55,56]. Renal impairment was modeled as reduced clearance based on glomerular filtration rate following KDIGO guidelines [57,58]. CYP2C9 genetic variability was incorporated using allele-specific activity scaling [59,60,61], and ABCB1 activity was adjusted according to published polymorphism data [62,63].

2.4. Parameter Optimization

Model parameters were estimated by minimizing weighted residuals between clinical data and simulations. The cost function was defined as

$$F\left(\overrightarrow{p}\right)=0.5\sum _{i,k}{\left(w_{i,k}·r_{i,k}\left(\overrightarrow{p}\right)\right)}^2,$$

with weights $w_{i,k}=n_k/{\sigma }_{i,k}$ based on study size and measurement error. One hundred optimization runs with varying initial conditions were performed. Fitting proceeded in two stages: first pharmacokinetic, then pharmacodynamic parameters. Independent datasets including multiple dosing and impairment scenarios were used for validation.

2.5. Pharmacokinetic and Pharmacodynamic Parameters

Pharmacokinetic parameters for losartan, E3174, and L158 were derived using non-compartmental methods. The elimination rate ($k_{el}$) was estimated from log-linear regression of the terminal phase. The area under the concentration-time curve (AUC) was computed by trapezoidal rule and extrapolation. Apparent clearance ($Cl/F$) and volume of distribution ($V_d/F$) were derived from $Cl/F=k_{el}·V_d$ and $V_d/F=D/\left(AU{C}_{\infty }·k_{el}\right)$, where D is dose. Pharmacodynamic outputs (renin, angiotensin, aldosterone, blood pressure) were summarized by maximal and minimal values.

3. Results

3.1. Losartan Database

An open database containing pharmacokinetic and pharmacodynamic data of losartan from 24 clinical studies was compiled, covering a range of dosing regimens, physiological conditions and different genotypes (Table 1). This dataset served as the basis for developing the losartan PBPK/PD model.

3.2. Computational Model

A PBPK/PD model of losartan was created that includes key factors that determine intra- and inter-individual variability (Figure 2). The model includes the organs involved in the pharmacokinetics of losartan (gastrointestinal tract, liver, and kidney), connected through the systemic circulation, as well as the main pharmacodynamic outputs of the RAAS (renin, angiontensin I, aldosterone, systolic and diastolic blood pressure). The main patient-specific factors that are known to affect the pharmacokinetics and pharmacodynamics of losartan are included in the model. Specifically, four key factors were included: dose dependency, liver impairment, renal impairment, and genotype variability.

3.3. Dose Dependency

Dose-dependent behavior of the losartan PBPK/PD model was assessed by simulating oral doses between 10 and 100 mg (Figure 3). The model captures the pharmacokinetics of losartan and its active metabolite E3174 in plasma, urine, and feces, as well as the associated pharmacodynamic responses (aldosterone, renin, angiotensin I, and blood pressure). Simulations reproduced key dose-related trends: increasing exposure with dose, a decreasing E3174/losartan ratio at higher doses, nonlinear increases in AUC and $C_{\max }$ (particularly for E3174), and dose-dependent prolongation of E3174 half-life. On the pharmacodynamic level, higher doses produced stronger suppression of aldosterone and blood pressure, accompanied by compensatory increases in renin and angiotensin I. These patterns are consistent with findings from published clinical studies (Figure 3 and Figure 4).

3.4. Hepatic Impairment

Figure 5 shows the impact of hepatic dysfunction on the pharmacokinetics and pharmacodynamics of losartan and E3174. As cirrhosis severity increases, losartan plasma and urine concentrations rise, while E3174 levels and the E3174/losartan ratio decrease. Fecal excretion remains largely unchanged, indicating that hepatic impairment predominantly affects metabolic clearance rather than biliary elimination.

Pharmacokinetic analysis reveals increasing AUC, C_max, and half-life for both losartan and E3174 with worsening cirrhosis, accompanied by a decline in the elimination rate constant (kel). While losartan accumulation is qualitatively consistent with clinical observations from patients with mild to moderate cirrhosis [8,11], the model underestimates absolute concentrations, particularly for E3174, which shows a delayed but modest increase in plasma levels at higher cirrhosis degrees. This discrepancy likely reflects unaccounted factors such as reduced hepatic blood flow, altered enzyme activity, or additional compensatory mechanisms [27].

Pharmacodynamically, impaired liver function attenuates the formation of E3174 and modifies RAAS regulation. Early renin and angiotensin I responses are reduced, while aldosterone and blood pressure increase initially, with prolonged elevation in more severe cirrhosis. These trends correspond to delayed normalization of hormonal and hemodynamic parameters. Despite sparse PD data and limited control groups, the model reproduces general concentration–time and dose–response patterns, supporting the clinical recommendation of dose adjustment in hepatic impairment. This is particularly relevant for patients with comorbidities such as heart failure, where sensitivity to RAAS inhibition is increased [35,36].

3.5. Renal Impairment

Figure 6 illustrates the effect of declining renal function on the pharmacokinetics and pharmacodynamics of losartan and its active metabolite E3174. As renal function decreases, losartan plasma concentrations remain largely unchanged, whereas E3174 levels increase, accompanied by reduced urinary excretion of both compounds. Fecal excretion of losartan shows a slight compensatory rise, and the E3174/losartan ratio increases with impaired renal clearance.

On the pharmacokinetic level, the model predicts stable AUC and C_max values for losartan, while E3174 shows rising AUC, C_max, and half-life as renal function declines, reflecting reduced clearance. These results align qualitatively with the physiology of renal elimination, though they overestimate metabolite accumulation compared to clinical findings [19,28], where E3174 exposure remained stable across renal function groups. This discrepancy may indicate unaccounted processes such as altered absorption, metabolism, or hepatic clearance of E3174 [11].

Pharmacodynamically, impaired renal function amplifies losartan’s antihypertensive effects. The model predicts stronger suppression of aldosterone and systolic blood pressure, along with compensatory rises in renin and angiotensin I. While direct PD data in renally impaired populations are limited, these predictions are physiologically plausible and extend clinical observations. Nevertheless, the degree of RAAS suppression may be slightly overestimated due to the absence of compensatory mechanisms such as altered hepatic clearance or long-term feedback regulation [36,37,77].

3.6. ABCB1 Genotypes

Figure 7 illustrates the influence of ABCB1 (P-glycoprotein) transporter activity on the pharmacokinetics and pharmacodynamics of losartan and E3174. Simulations across varying transporter activity levels, including common diplotypes (GG/CC, GT/CT, TT/TT), show that reduced ABCB1 activity leads to increased plasma and urinary concentrations of both compounds, while fecal excretion of losartan declines due to decreased intestinal and biliary efflux.

Pharmacokinetic analysis indicates modest increases in AUC and C_max for losartan and E3174 with lower ABCB1 activity, whereas elimination rate constants (kel) and half-life remain largely unchanged. The plasma and urine E3174/losartan ratio decreases slightly under low transporter activity. These trends are qualitatively consistent with clinical observations [32], although absolute concentrations are slightly underestimated by the model.

Pharmacodynamically, reduced ABCB1 activity slightly enhances RAAS-related effects, including modest increases in renin and angiotensin I and marginal reductions in aldosterone and blood pressure. While these changes are small and unlikely to necessitate dose adjustments based on ABCB1 diplotype alone, the model demonstrates the capacity to predict subtle genotype-dependent differences in drug exposure and pharmacodynamic response. Future refinement could include better representation of renal P-gp expression and its contribution to systemic clearance, as well as additional clinical data on pharmacodynamic outcomes to improve predictive accuracy [32].

3.7. CYP2C9 Genotypes

Figure 8 illustrates the impact of CYP2C9 genetic variability on the pharmacokinetics and pharmacodynamics of losartan and its active metabolite E3174. Simulations across a continuous range of CYP2C9 activity, as well as genotype-specific simulations for *1/*1, *1/*2, *1/*3, *1/*13, *2/*2, *2/*3, and *3/*3, demonstrate that reduced enzyme activity leads to impaired metabolic conversion, resulting in higher plasma, urine, and fecal levels of losartan and decreased E3174 exposure.

Pharmacodynamically, lower CYP2C9 activity diminishes RAAS inhibition, with smaller reductions in aldosterone and systolic blood pressure, and attenuated compensatory increases in renin and angiotensin I. AUC and C_max for losartan rise with declining enzyme activity, while E3174 exposure decreases; elimination rate constants and half-life remain relatively stable. These model predictions are consistent with clinical observations, which report 1.6- to 3-fold increases in losartan AUC and significantly reduced E3174 levels in carriers of reduced-function alleles [35,36,37], although absolute concentrations tend to be slightly underestimated.

The use of a continuous CYP2C9 activity scale enables the simulation of nonlinear relationships between enzyme activity and metabolite formation across a spectrum of genotypes. While CYP2C9 genotype is a major determinant of inter-individual variability, other factors such as age, comorbidities, and concomitant medications may further influence pharmacokinetics and pharmacodynamics. From a clinical perspective, these findings support the potential of genotype-guided dosing for patients with reduced-function alleles, although high variability within genotype groups suggests that genetic testing should be applied selectively, focusing on individuals with inadequate response or elevated risk of adverse effects.

Figure 9. continued: Pharmacokinetics and pharmacodynamics of losartan and metabolites across varying CYP2C9 activity levels. (F) Simulated versus observed losartan pharmacokinetics. Data from [35,40,66]. Simulated versus observed SBP and DBP for different CYP2C9 genotypes [36].

Figure 9. continued: Pharmacokinetics and pharmacodynamics of losartan and metabolites across varying CYP2C9 activity levels. (F) Simulated versus observed losartan pharmacokinetics. Data from [35,40,66]. Simulated versus observed SBP and DBP for different CYP2C9 genotypes [36].

4. Discussion

In this study, we developed and evaluated a physiologically based pharmacokinetic/ pharmacodynamic (PBPK/PD) model of losartan, informed by a curated database of 24 clinical studies covering diverse dosing regimens, populations, and physiological conditions. The model integrates intestinal absorption, hepatic metabolism via CYP2C9, and renal excretion, while linking systemic exposure of the active metabolite E3174 to downstream pharmacodynamic effects through a simplified RAAS submodel. This framework enables simulation of losartan and metabolite concentrations in plasma, urine, and feces, as well as prediction of pharmacodynamic endpoints such as renin, angiotensin I, aldosterone, and systolic blood pressure under baseline and perturbed conditions. By reproducing key clinical observations, the model demonstrates its potential to provide mechanistic insights into inter-individual variability and support personalized dosing strategies.

The curated database was essential for calibration and validation of the model, but it also highlights important limitations. While losartan pharmacokinetics are well characterized, data on RAAS dynamics remain limited, particularly with respect to circadian variability and long-term feedback regulation. Excretion data—especially for fecal elimination and the secondary metabolite L158—were sparse, and pharmacodynamic datasets showed high variability across studies. These gaps restrict the predictive accuracy of the model under complex physiological conditions but do not preclude its use in capturing system-level dose–exposure–response relationships.

Our dose dependency simulations reproduced the nonlinear pharmacokinetics of losartan and E3174, demonstrating a decreasing metabolite-to-parent ratio at higher doses due to saturable CYP2C9 metabolism. Consistent with clinical findings, increasing doses resulted in stronger and more sustained suppression of aldosterone and systolic blood pressure, accompanied by compensatory increases in renin and angiotensin I. This supports the model’s ability to capture both nonlinear PK behavior and dose-dependent RAAS responses.

In hepatic impairment, the model predicted elevated losartan plasma concentrations, reduced clearance, and attenuated metabolite formation, reflecting impaired CYP2C9 metabolism and hepatic blood flow. While the model underestimated the reported accumulation of E3174 in cirrhosis, it nonetheless confirmed the need for dose reduction in patients with liver dysfunction, particularly in comorbid conditions such as heart failure where sensitivity to RAAS inhibition is heightened.

Renal impairment simulations predicted reduced clearance of both losartan and E3174, leading to prolonged systemic exposure. While these results are qualitatively consistent with PK data, the model overestimated E3174 accumulation compared to reported clinical findings. Importantly, in the absence of PD data from impaired populations, the model allowed prediction of enhanced RAAS suppression and blood pressure reduction—although these effects may be slightly exaggerated due to the omission of compensatory mechanisms.

Genetic variability was also explored. For CYP2C9, the model captured the pronounced impact of reduced-function alleles on E3174 formation and downstream pharmacodynamic effects, reproducing the diminished blood pressure reduction observed clinically. The use of a continuous enzyme activity scale enabled simulation across a spectrum of genotypic variability. In contrast, ABCB1 (P-gp) variability had only minor effects on systemic exposure and blood pressure, consistent with limited clinical data, although the model suggests possible contributions via altered intestinal efflux and renal secretion. These findings emphasize the potential of PBPK/PD modeling to disentangle the contributions of genetic variability to losartan response.

Looking forward, several opportunities for refinement exist. Incorporating circadian rhythm, sodium and water balance, and autonomic feedback would enhance physiological fidelity of the RAAS submodel. Expanding datasets for fecal excretion, L158 kinetics, and pharmacodynamic outcomes would strengthen calibration. Future model extensions could also integrate additional sources of variability, such as age, comorbidities, or transporter–enzyme interactions, to support population-level simulations. Ultimately, the losartan PBPK/PD framework provides a mechanistically grounded platform for exploring inter-individual variability and optimizing dosing strategies, with potential applications in precision medicine and clinical decision support.

5. Conclusions

In this study, a comprehensive physiologically based pharmacokinetic/ pharmacodynamic (PBPK/PD) model of losartan was developed and validated using an extensive clinical database. The model successfully reproduced key pharmacokinetic and pharmacodynamic behaviors, including dose-dependent RAAS inhibition and nonlinear metabolite formation resulting from saturable CYP2C9 metabolism. Simulations captured the attenuated pharmacodynamic response under hepatic impairment and predicted enhanced RAAS suppression in renal dysfunction. Furthermore, the model quantified the pronounced impact of CYP2C9 genetic variability on metabolite exposure and blood pressure effects, while ABCB1 activity contributed only minor modulation of systemic exposure.

Despite certain limitations in clinical data and model scope, the framework provides valuable mechanistic insight into inter-individual variability in losartan response. It highlights the potential of systems pharmacology modeling to support dose optimization across physiological and genetic conditions. Together, the losartan database and PBPK/PD model establish a solid foundation for future integration into digital twin platforms and personalized therapy design.