BACKGROUND
Resistant hypertension (RH) remains a significant health condition, particularly in older adults [1]. Defined as uncontrolled blood pressure (BP) despite the use of three antihypertensive agents including a diuretic [2], RH is associated with markedly increased risks of cardiovascular events, renal impairment, and all-cause mortality [3].
The management of RH in elderly patients is especially challenging. Aging is accompanied by progressive changes in hepatic function, renal reserve, body composition, vascular stiffness, and drug-metabolizing capacity, all of which may alter pharmacokinetics (PKs) and pharmacodynamics (PDs) [45]. In parallel, older patients often exhibit disrupted circadian BP regulation, including attenuated nocturnal dipping and enhanced early-morning BP rise [6].
These age-related physiological and chronobiological changes suggest that a uniform strategy of “one size fits all” less convincing for elderly patients with complex clinical heterogeneity [7].
Aprocitentan, a novel oral dual endothelin A/B receptor antagonist, represents a promising therapeutic option for older adults with RH by targeting endothelin-mediated vasoconstriction [8], sodium–water retention, inflammation, and vascular remodeling. Its long effective half-life suggests once-daily administration, and available population PK analyses suggest broadly predictable exposure [9]. However, current evidence is largely derived from multinational non-Chinese-specific cohorts, and does not sufficiently address precision dosing in elderly Chinese patients [10].
This uncertainty is particularly relevant because age-related hepatic impairment and interindividual variability in uridine 5’-diphospho-glucuronosyltransferase 1A1 (UGT1A1)-mediated metabolism [11] may influence drug exposure, response, and safety [12].
Although PK and population-based analyses have provided important information, existing modeling approaches have not fully integrated drug exposure, BP circadian rhythm, subgroup-specific physiology, hepatic function, and metabolic variability within a unified quantitative framework [7913]. For antihypertensive therapy, treatment response depends not only on systemic concentration, but also on the timing of administration relative to endogenous BP rhythms [14]. Therefore, a physiologically based pharmacokinetic (PBPK)/PD-circadian rhythm model embedded within a model-informed precision dosing (MIPD) framework [15] is needed to suggest individualized dose selection and chronotherapy optimization.
This study aimed to develop a PBPK/PD-circadian rhythm model of aprocitentan in elderly Chinese patients with RH. The objectives were to characterize aprocitentan exposure under population-specific physiological conditions, evaluate the influence of hepatic function and UGT1A1 variability, integrate circadian BP dynamics into PD prediction, and simulate alternative dosing regimens and administration times. This framework is expected to provide a quantitative basis for optimizing aprocitentan therapy and advancing individualized chronotherapy in elderly Chinese patients with resistant hypertension.
METHODS
The overall modeling framework consisted of four sequential stages, as summarized in Fig. 1. First, a PBPK model was developed using physicochemical, physiological, and clinical information and validated against independent PK datasets.
The validated PBPK model was then used to predict aprocitentan concentration–time profiles and systemic exposure. Second, the predicted drug exposure was linked to the antihypertensive effect using a direct-effect sigmoid maximum drug effect (Emax) model. This concentration-dependent PD effect was subsequently integrated with the 24-hour circadian BP baseline to construct the final PK/PD-circadian rhythm model, which was evaluated against phase II and phase III BP data.
Third, the validated integrated model was extrapolated to elderly Chinese patients with RH (aged ≥ 65 years) by incorporating population-specific physiology, UGT1A1 activity, and the circadian BP baseline of the target population. Finally, the extrapolated model was used for sensitivity analysis and simulations of different populations, dose regimens, and administration times to generate the final 24-hour BP predictions.
Data sources and software
Physicochemical properties of aprocitentan were obtained from Food and Drug Administration drug review documents (Drugs@FDA) [16], DrugBank [17], PubChem [18], InvivoChem [19], and published clinical studies. Genetic polymorphism data were retrieved from the Clinical Pharmacogenomics Database (ClinPGx) [20]. Plasma concentration-time profiles and circadian BP curves were digitized using GetData Graph Digitizer (version 2.22).
PBPK model development and local one-at-a-time sensitivity analysis were performed using the Open Systems Pharmacology Suite (PK-Sim® and MoBi®, version 12.1). Statistical analyses were conducted in R version 4.6.0 under Windows 11 ×64. Morris global sensitivity analysis was performed using the ospsuite.globalsensitivity package (version 1.0.0).
Data processing and visualization were performed using dplyr (version 1.2.1), readxl (version 1.4.5), and ggplot2 (version 4.0.3). Circadian rhythm fitting was performed using the nlsLM() function from the minpack.lm package (version 1.2.4).
PBPK model development
Datasets
PK data (Supplementary Table 1) were obtained from seven independent phase I clinical studies, comprising 14 plasma concentration-time profiles [162122]. The datasets included healthy adult Caucasian subjects (n = 163), healthy elderly Caucasian subjects (n = 8), adults with severe renal impairment (n = 8), healthy Japanese subjects (n = 10), and adults with moderate hepatic impairment (n = 8).
The dataset covered single oral doses (5, 25, 50, and 100 mg) and multiple-dose regimens (5, 25, and 100 mg once daily for 10 days) with dense sampling. Ten profiles from three studies were used as the training dataset for parameter estimation, while the remaining four studies served as external validation. Additionally, 22 key PK parameters were extracted for quantitative goodness-of-fit (GOF) evaluation (Supplementary Table 2).
Model development
A base model was established using physiological parameters of a typical healthy adult Caucasian individual provided within the software. Drug-specific processes included 2 metabolic enzymes (UGT1A1 and uridine 5’-diphospho-glucuronosyltransferase 2B7) and 2 efflux transporters (ABCB1/MDR1 and ABCG2/BCRP) [11].
Based on in vitro and in vivo metabolic profiling, aprocitentan is primarily cleared via non-enzymatic hydrolysis and hepatic metabolism, as illustrated in Supplementary Fig. 1. Due to the absence of a non-enzymatic hydrolysis module in PK-Sim and incomplete characterization of enzyme-specific contributions [23], a functionally equivalent simplification strategy was adopted and was illustrated in Supplementary Fig.
2. Specifically, the relative contributions of UGT-mediated N-glucosidation and residual hepatic cytosolic/mitochondrial intrinsic clearance (CLint) processes were adjusted so that the resulting PBPK model reproduced the observed systemic clearance (CLsystemic) and plasma concentration-time profiles [24]. Excretion pathways were parameterized based on mass balance studies [25], with fixed fractions of parent drug and metabolites recovered in urine and feces used as constraints during optimization.
The drug-specific parameters used in the model were derived from various sources and optimized based on available experimental data. The final drug-specific input parameters used for the PBPK model are summarized in Table 1.
Model evaluation
Model calibration was performed using log-transformed GOF plots, comparing simulated and observed concentration-time profiles [26].
For quantitative validation, fold error (FE) was calculated to evaluate the model’s performance. The FE for each data point was computed using the following formula:
PK/PD-circadian rhythm model development
Datasets
PD data were obtained from a phase II dose-response study and the phase III PRECISION trial, comprising 6 cohorts (n = 760) with hypertension and multiracial enrollment, undergoing ambulatory blood pressure monitoring (ABPM) [102728].
The training dataset consisted of 4 dose groups from the phase II study (5, 10, 25, and 50 mg once daily for 8 weeks), while the validation dataset included two dose groups from the phase III study (12.5 and 25 mg once daily for 4 weeks) (Supplementary Table 3).
Model development
The PD model was designed to quantitatively describe the relationship between plasma concentration of aprocitentan and its antihypertensive effect, using a sigmoid Emax direct-effect model [29]:
where Edrug was expressed as a negative value to represent BP reduction, Emax the maximal effect, Cp the plasma concentration, concentration producing 50% of maximal effect (EC50) the concentration producing 50% of Emax, and γ the Hill coefficient.
To capture the intrinsic circadian variation in BP, a triple-harmonic Fourier function was employed to describe the rhythm [3031]:
where Ecircadian represents the baseline circadian effect, M is the mean BP level, and Ai and φi denote the amplitude and phase angle of the ith harmonic (24, 12, and 8 hours), respectively. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) parameters were estimated separately by bounded nonlinear least squares using minpack.lm::nlsLM() (version 1.2.4), which implements a Levenberg–Marquardt nonlinear least-squares algorithm with parameter bounds. Initial values were informed by the observed BP level and lower-order harmonic fits.
The maximum number of iterations was 200, with both function tolerance (ftol) and parameter tolerance (ptol) set to 1 × 10−9. Prespecified bounds were applied to the mesor, amplitudes, and phase parameters.
The integrated equation for absolute BP is given by [15]:
where E represents the final BP.
The final parameter estimates of the integrated PK/PD-circadian rhythm model are summarized in Supplementary Tables 4 and 5.
Model evaluation
The log-transformed GOF plots were generated by comparing simulated and observed BP–time profiles. The FE calculation was consistent with the approach used in the PBPK model.
Extrapolation
Elderly Chinese patients were defined as individuals aged ≥ 65 years throughout the study [32]. The validated PBPK model was extrapolated to this target population using the East Asian population setting in PK-Sim, together with age- and population-specific physiological characteristics, according to literature-reported baseline demographics of elderly Chinese patients [33]. No compound-specific parameters were re-estimated during this step.
In the PBPK/PD-circadian framework, the PK-driven exposure-response relationship was retained from the reference model, whereas the circadian baseline BP function was adapted to the target population. The mesor, amplitude, and phase parameters of the SBP and DBP baseline functions were adjusted using reported 24-hour BP profiles in elderly Chinese patients with hypertension [34].
Sensitivity analysis
Both local one-at-a-time (OAT) and global Morris sensitivity analyses [35] were conducted to evaluate the impact of key input parameters on major output metrics, including area under the curve (AUC), maximum plasma concentration (Cmax), and mean residence time (MRT) parameters [36].
Local sensitivity analysis was first performed by perturbing each parameter individually and assessing the resulting changes in model outputs. Based on these results, influential parameters were selected for global sensitivity analysis.
In the Morris method [3738], 500 Morris runs were generated to compute the elementary effect (EE) of each parameter (x):
The mean (μ*) and standard deviation (σ) of EE were subsequently calculated over n sampling iterations [39], thereby quantifying both the main effects and interaction effects of each parameter.
where n represents the number of random trajectories.
Parameters identified as influential in the local one-at-a-time sensitivity analysis were further evaluated using the Morris elementary-effects method implemented with ospsuite.globalsensitivity::runMorris() (version 1.0.0). Eleven parameters were included and assigned prespecified uniform or log-uniform distributions according to their physiological or physicochemical characteristics, and were shown in Supplementary Table 6.
Simulation and optimization strategy
UGT1A1 variability and hepatic impairment
To reflect metabolic variability in elderly Chinese patients, UGT1A1 activity was incorporated as a variable parameter in Monte Carlo simulations, as reduced-function UGT1A1 alleles are relatively common in East Asian populations [4041]. Six representative diplotypes were considered, including *1/*1, *1/*6, *1/*28, *6/*6, *6/*28, and *28/*28, corresponding to normal, intermediate, and poor metabolizer phenotypes [4243]. Accordingly, UGT1A1 activity was varied from 0.3 to 1.0 of the reference activity, representing poor to normal metabolic capacity.
Finally, 100 virtual elderly Chinese patients with RH were simulated [44]. In addition to the base population, a separate cohort with mild hepatic impairment was generated [12].
Dose and dosing-frequency simulation
Dose and dosing-frequency regimens were evaluated using steady-state exposure as the primary criterion [9]. The steady-state area under the curve (AUCss) and Cmax achieved by the standard clinical regimen were used as reference exposure metrics. Candidate regimens with different dose levels and dosing frequencies were simulated in both elderly Chinese patients and elderly Chinese patients with mild hepatic impairment.
The preferred regimen was defined as the strategy that maintained exposure close to the clinical reference [22] while reducing the risk of excessive systemic exposure, particularly in subjects with impaired hepatic function or reduced UGT1A1 activity.
Chronopharmacological optimization
After selecting the dose and dosing frequency, dosing-time optimization was performed using the PBPK/PD-circadian rhythm model. For each candidate administration time across 24 hours, the model predicted SBP and DBP profiles by integrating aprocitentan exposure, the PK-driven drug effect, and the elderly Chinese-specific circadian BP baseline [34]. An exhaustive grid search was conducted from 0 to 24 hours with a 0.1-hours step size.
The optimal dosing time was identified by minimizing the following objective function:
The first term represents the maximum predicted SBP over the 24-hour cycle and was used to prioritize control of the least well-controlled portion of the daily BP profile [45], and the second term represents the within-day blood pressure variability (BPV) [4647]. The 0.7/0.3 coefficients were prespecified study-specific modeling weights, assigning greater priority to maximum-SBP control while retaining BPV as a secondary optimization criterion. The candidate time yielding the lowest objective-function value was identified as the model-predicted global optimum, and additional local minima in the objective-function profile were retained for interpretation.
There was no iterative convergence criterion because all candidate clock times were explicitly evaluated.
RESULTS
PBPK model
As shown in Fig. 2, the fitted concentration–time profiles for the training dataset demonstrate acceptable agreement between model predictions and observed data across all dose groups and time points, particularly at steady-state peak and trough concentrations following multiple dosing.
GOF plots show that most of predicted-to-observed ratios fall within the 0.8–1.25 range (Supplementary Fig. 3). The prediction error for UGT-mediated metabolite fractions in plasma ranged within 0.8–1.25-fold, while urinary and fecal excretion predictions were within 0.5–2.0-fold, as illustrated in Supplementary Fig.
4.
PK/PD-circadian rhythm model
The circadian rhythm fitting results demonstrated that the triple-harmonic cosine function well described diurnal BP variation. For SBP, the residual sum of squares (RSS) was 16.70 and the root mean square error (RMSE) was 0.834 mmHg. For DBP, RSS was 14.26 and RMSE was 0.771 mmHg.
As illustrated in Supplementary Fig. 5, the predictive performance of the PD model was further evaluated using GOF plots based on all observed and predicted BP data from Supplementary Figs. 6, 7, 8, 9, 10, 11.
Overall, the predicted-to-observed ratios for both SBP and DBP were closely distributed around the unity line across different dose levels and sampling days. Most SBP predictions fell within the predefined 1.5-fold acceptance region, while DBP predictions showed even tighter agreement, with most points located within the 0.8–1.25-fold range.
Comparative analysis confirmed that the triple-harmonic Fourier function provided a better description of baseline circadian patterns than single- or dual-harmonic alternatives.
Sensitivity analysis results
Local sensitivity analysis illustrated in Supplementary Figs. 12, 13, 14, 15, indicated that aprocitentan PK outputs were mainly affected by ionization-, absorption-, and distribution-related parameters. Interstitial pH showed the strongest local influence on Cmax and AUC, followed by acid dissociation constant (pKa), plasma pH, intestinal luminal pH, lipophilicity, intestinal permeability, and plasma unbound fraction.
MRT and terminal half-life were particularly sensitive to lipophilicity, pKa, and plasma pH.
Morris global sensitivity analysis shown in Supplementary Figs. 16, 17, 18, 19 confirmed that the dominant drivers differed across PK endpoints. Plasma unbound fraction was the most influential parameter for cumulative and AUCss, with liver volume, interstitial pH, pKa, and lipophilicity also contributing substantially.
Lipophilicity was the primary determinant of early exposure, MRT, and terminal half-life, whereas pKa and plasma unbound fraction showed secondary but consistent effects. Several parameters showed relatively high σ values.
For PD outputs illustrated in Supplementary Fig. 20, Emax was the main determinant of the cumulative SBP-lowering effect, whereas EC50 showed an inverse influence. The peak SBP response was most sensitive to the Hill coefficient, indicating that the steepness of the exposure-response relationship strongly affected maximal BP reduction.
Overall, plasma unbound fraction, lipophilicity, pKa, pH-related parameters, liver volume, and key PD parameters showed the largest sensitivity measures across the evaluated PK and PD outputs.
Simulation and optimization results
Population simulation
Compared with healthy elderly Caucasian subjects (reference AUCss = 39,495.93 μmol*min/L), extrapolation simulations showed that the mean AUCss in elderly Chinese patients with RH increased to 47,619.47 μmol*min/L (around 20.6% increase), while in those with mild hepatic impairment it increased to 60,233.35 μmol*min/L (around 52.5% increase). Population simulation results are presented in Supplementary Fig. 21.
Dose and dosing-frequency
Simulated exposure metrics, including AUCss and Cmax, under different dosing regimens are shown in Fig. 3 for elderly Chinese patients with RH without and with mild hepatic impairment. The 0.8–1.25-fold range of the standard reference exposure was used as the predefined target window.
In elderly Chinese patients with RH without hepatic impairment, once daily (QD) 6.25 mg for 4 weeks produced AUCss and Cmax distributions that were consistently within the target exposure range. Although twice a day (BID) 6.25 mg showed acceptable median AUCss, its Cmax distribution was shifted upward relative to QD 6.25 mg. Similarly, once weekly (QW) 12.5 mg produced a Cmax close to the reference range.
Regimens involving escalation to 12.5 mg, including QD 6.25 mg followed by 12.5 mg and QD or BID 12.5 mg regimens, resulted in upward shifts in both AUCss and/or Cmax relative to the predefined reference range.
In elderly Chinese patients with mild hepatic impairment, systemic exposure increased further across most regimens. QD 6.25 mg remained the only daily regimen that maintained both AUCss and Cmax within or close to the predefined target window. In contrast, regimens containing 12.5 mg led to marked increases in systemic exposure, particularly for Cmax.
QW 12.5 mg showed exposure values close to the reference range.
Based on the predefined exposure criterion, QD 6.25 mg was retained for subsequent administration-time and BP simulations.
Administration time
Using the circadian rhythm profile of elderly Chinese patients and 24-hour PD simulations, optimal dosing time was evaluated based on SBP. Two local minima were identified under once-daily dosing (approximately 2.0 and 12.4 hours).
As shown in Fig. 4, dosing at 12.4 hours (approximately 12:24) was associated with a lower objective-function value than surrounding alternative times and was retained as the model-predicted midday candidate for subsequent simulations.
Relative to the circadian baseline, the 12.4-hour simulation produced an additional 8.96 mmHg reduction in nighttime SBP and a 34.66% reduction in peak-to-trough fluctuation. Full optimization results are provided in Supplementary Fig. 22.
BP
To evaluate the performance of the model-predicted regimen, BP profiles were simulated by combining the circadian BP pattern with the PD effect of aprocitentan 6.25 mg once daily administered at the midday dosing time, and were shown in Fig. 5. BP levels were interpreted according to hypertension classification thresholds based on current Chinese guidelines.
In the general elderly Chinese RH population, aprocitentan 6.25 mg once daily reduced SBP and DBP by 13.04 and 8.97 mmHg, respectively. In elderly Chinese patients with RH and mild hepatic impairment (CP-A), the corresponding reductions were 13.63 and 9.37 mmHg. In both populations, the simulated 24-hour BP profile was maintained mainly within the Grade 1 hypertension range at steady state, without sustained elevation into the Grade 2 range.
DISCUSSION
The present study provides a quantitative and clinically oriented modeling strategy for optimizing aprocitentan therapy in elderly Chinese patients with RH. This design enabled simultaneous assessment of dose, dosing frequency, and administration time, thereby extending model-informed precision dosing [48] from exposure matching to 24-hour BP control. The approach is especially valuable for elderly Chinese patients, a population in which direct clinical evidence remains limited and treatment decisions may be affected by age-related physiological changes, UGT1A1 variability, and hepatic functional reserve.
By further evaluating mild hepatic impairment as a clinically important subgroup, the model provided a mechanistic rationale for avoiding unnecessary dose escalation and suggested 6.25 mg once daily and 12:00–13:00 as model-predicted alternative strategies.
Office BP does not fully reflect ambulatory BP-related risk. In the IDACO analysis, ambulatory BP provided prognostic information beyond office BP, while cohort studies showed that nighttime BP and specific nocturnal BP phenotypes were strongly associated with cardiovascular outcomes [495051]. However, evidence for routine bedtime antihypertensive dosing remains inconsistent: MAPEC-related and Hygia studies reported cardiovascular benefits with bedtime dosing [5253], whereas the TIME trial found no outcome difference between evening and morning administration [54].
The International Society of Hypertension therefore does not recommend bedtime dosing for all patients, but suggests ABPM-guided assessment of nocturnal hypertension, non-dipping status, and morning BP surge [55]. The recent OMAN trial in Chinese patients showed greater nocturnal BP reduction with bedtime dosing than morning dosing, but did not assess cardiovascular outcomes [56]. These findings suggest that administration time should be considered in relation to drug pharmacology and individual BP phenotype.
This may be particularly relevant in older patients, in whom nocturnal hypertension, nondipping patterns, morning BP surge, and increased BPV are common [5758].
The model-predicted midday window should be interpreted in the context of aprocitentan PKs. Aprocitentan reaches Cmax approximately 4–5 hours after dosing and has an effective half-life of approximately 41 hours [16]; therefore, at steady state, administration-time effects are unlikely to reflect major differences in total 24-hour exposure. Instead, they are more plausibly explained by a phase-alignment effect superimposed on sustained long-acting exposure, whereby relatively stable background pharmacological activity is maintained while the stronger post-dose exposure/effect phase shifts relative to the endogenous circadian BP rhythm.
This interpretation is consistent with circadian PK/PD modeling showing that optimal timing for mean BP reduction, BP-peak control, and nocturnal dipping may differ across the 0–24-hour cycle [45]. Accordingly, our objective function prioritized the maximum 24-hour SBP, representing the least well-controlled period, while retaining the BPV as a secondary measure [4757]. The 0.7/0.3 weighting was therefore a prespecified study-specific modeling choice rather than a clinically validated threshold.
Within this framework, the local minima at approximately 02:00 and 12.4 hours may reflect different alignments with the morning and later-day BP rises reported by ambulatory monitoring studies [59]. Given the 4–5-hour time to maximum concentration, dosing around 02:00 would shift the stronger post-dose phase toward the morning BP rise, whereas dosing at 12:00–13:00 would align it more closely with the later daytime or early-evening rise, while the long half-life maintains background coverage across the rest of the cycle. Aprocitentan-specific PRECISION analyses also demonstrated sustained daytime and nighttime ambulatory BP lowering [60].
We therefore regard 12:00–13:00 as a simulation-derived candidate window for patients with conventional sleep–wake patterns. Any potential adherence advantage of lunch-time dosing was not evaluated, and the alternative 02:00 minimum should not be extrapolated to night-shift workers without individualized assessment based on sleep–wake timing and ABPM [10].
The hepatic-impairment findings require cautious interpretation. In the present simulations, AUCss was 52.5% higher in the elderly Chinese cohort with mild hepatic impairment than in healthy elderly Caucasian subjects. This difference should not be interpreted as the isolated effect of mild hepatic impairment, because the comparator also differs in ethnicity and age-related physiology.
Current prescribing information [16] does not require dose adjustment in mild hepatic impairment, but recommends baseline assessment of aminotransferases and bilirubin and clinically indicated monitoring, while use is not recommended in moderate-to-severe hepatic impairment. Hepatic safety also remains a recognized consideration for endothelin receptor antagonists: earlier experience with bosentan and sitaxsentan highlighted hepatotoxicity concerns, mechanistic studies have implicated hepatobiliary transporter inhibition and disturbed bile-acid handling [6162]. But macitentan showed a more favorable hepatic safety profile in SERAPHIN [6364].
Therefore, the absence of mandatory dose adjustment in mild hepatic impairment should not be interpreted as absence of hepatic relevance. Within this combined elderly Chinese mild-hepatic-impairment population, regimens containing 12.5 mg produced greater upward shifts in exposure, whereas 6.25 mg once daily remained closer to the predefined reference range. These findings suggest 6.25 mg once daily as a conservative, model-derived candidate for prospective evaluation, but do not demonstrate that the approved 12.5 mg dose is unsafe or that clinical dose reduction is required.
Prospective validation should combine sparse aprocitentan plasma sampling with clock-time-resolved 24-hour ABPM in elderly Chinese patients with RH, together with relevant demographic, hepatic, renal, metabolic, and sleep–wake covariates. Observed PK and time-resolved BP data could then be compared with model predictions to evaluate performance in the intended population [65]. Following external validation, a prospective timing study could compare morning, midday, and bedtime administration.
Sparse PK and ABPM data may also suggest future Bayesian model updating [48]. However, because no validated therapeutic concentration range for aprocitentan exists, routine therapeutic drug monitoring cannot currently be recommended.
This study has several limitations. First, individual-level PK/PD observations were unavailable for elderly Chinese patients with RH, limiting characterization of interindividual covariate relationships and precluding external validation in the intended target population. Second, longitudinal ET-1 concentrations, receptor-occupancy measurements, and other mechanistic biomarkers were unavailable [22].
Consequently, the present model cannot distinguish direct concentration-dependent effects from more complex delayed or indirect PD mechanisms. Third, exact dosing clock times were not consistently available in the source clinical datasets, and literature-derived circadian BP profiles may not fully represent the heterogeneity associated with sleep–wake patterns, dipping phenotypes, comorbidities, and concomitant treatments in older patients. Finally, the proposed 6.25 mg once-daily regimen and 12:00–13:00 administration window are model-predicted candidates.
The present simulations cannot establish their comparative efficacy, safety, tolerability, adherence, or cardiovascular benefit in real-world patients with RH. Prospective individual-level PK and 24-hour ABPM validation, followed by comparative clinical studies, is therefore required before clinical implementation.
CONCLUSIONS
This study developed an integrated PBPK/PD-circadian rhythm framework for evaluating aprocitentan dosing in elderly Chinese patients with RH. The model characterized population-specific exposure changes associated with aging, UGT1A1 variability, mild hepatic impairment, and circadian BP variation. Within the assumptions of the model, 6.25 mg once daily provided a favorable simulated exposure profile, and 12:00–13:00 emerged as a candidate administration window for patients with a conventional sleep-wake rhythm.
These findings should be regarded as model-based, hypothesis-generating predictions. Prospective studies incorporating individual PK measurements, 24-hour ABPM, safety, adherence, and clinical outcomes are required to confirm the proposed dose and administration-time strategies.