Development of a physiologically based pharmacokinetic (PBPK) population model for Chinese elderly subjects

Aims This study aims to develop and verify a physiologically based pharmacokinetic (PBPK) population model for the Chinese geriatric population in Simcyp. Methods Firstly, physiological information for the Chinese geriatric population was collected and later employed to develop the Chinese geriatric population model by recalibration of corresponding physiological parameters in the Chinese adult population model available in Simcyp (i.e., Chinese healthy volunteer model). Secondly, drug‐dependent parameters were collected for six drugs with different elimination pathways (i.e., metabolized by CYP1A2, CYP3A4 or renal excretion). The drug models were then developed and verified by clinical data from Chinese adults, Caucasian adults and Caucasian elderly subjects to ensure that drug‐dependent parameters are correctly inputted. Finally, the tested drug models in combination with the newly developed Chinese geriatric population model were applied to simulate drug concentration in Chinese elderly subjects. The predicted results were then compared with the observations to evaluate model prediction performance. Results Ninety‐eight per cent of predicted AUC, 95% of predicted C max, and 100% of predicted CL values were within two‐fold of the observed values, indicating all drug models were properly developed. The drug models, combined with the newly developed population model, were then used to predict pharmacokinetics in Chinese elderly subjects aged 60–93. The predicted AUC, C max, and CL values were all within two‐fold of the observed values. Conclusion The population model for the Chinese elderly subjects appears to adequately predict the concentration of the drug that was metabolized by CYP1A2, CYP3A4 or eliminated by renal clearance.


| INTRODUCTION
Elderly subjects aged over 65 are the largest population in the pharmaceutical market that takes an average of two to five kinds of medication per day. 1 However, it is also one of the least studied populations during drug development as they are generally not included in the clinical trial due to complex pathophysiology, variability in organ function and presence of co-medication. 2 Thus, the approved dosing regimen for young adults has usually been applied to geriatric patients, which may not be appropriate since the drug absorption, distribution, metabolism and elimination (ADME) in geriatric patients may be different. 3 In the absence of well-designed clinical trials focusing on geriatric patients, modelling and simulation technique may serve as an alternative way to understand pharmacokinetic (PK) and pharmacodynamic (PD) variability and to inform dosing regimen design. Physiologically based pharmacokinetic (PBPK) modelling can describe the drug ADME process by incorporating drug properties and physiological variables from a specific population. Over the past several years, it has been successfully applied to predict drug concentration in paediatrics, pregnant women and renal impairment patients. [4][5][6][7][8] To date, a few studies have been conducted on PBPK models to predict drug concentration in the geriatric population. In these models, alterations of certain physiological parameters with ageing were considered, such as alveolar ventilation, cardiac output, different organ (e.g., liver, brain, heart, kidney, spleen, left and right lung) weights and blood flows. [9][10][11][12][13] However, most models were developed in the Caucasian geriatric population whose physiological variables and PK characteristics are not necessarily the same in the Chinese geriatric population. In fact, such difference has been observed between Caucasian and Chinese young adults.
For instance, the differences in the frequency of cytochrome P450 (CYP)2C19 poor metabolizers (PMs), CYP2D6 PMs, and intermediate metabolizers (IMs) caused a reduction in clearance of phenacetin, omeprazole, desipramine, midazolam and alprazolam by 19-75% in Chinese subjects, compared to Caucasian subjects. 14 Thus, it may be risky to predict drug exposure in the Chinese geriatric population by adopting the Caucasian population model directly. On the other hand, elderly subjects usually have different PK characteristics from young people due to the degeneration of organ function, which also hampers the application of the Chinese adult population model to predict PK characteristics of the Chinese elderly subjects. For example, the clearance of theophylline after oral administration in people aged 62-93 was approximately 54% lower than in people aged 21-24. 15,16 Thus, it is necessary to develop a population model that is specific to the Chinese geriatric population.
Li et al. previously developed a Chinese population model by incorporating demographic information, microsomal protein per gram of liver (MPPGL), liver weight and CYP1A2 abundance that were specific to the Chinese population. 17 This model was used to predict theophylline concentration in Chinese geriatric subjects and achieved reasonable prediction with predicted maximum drug concentration (C max ) and area under the curve (AUC) within two-fold of the observed values. However, most of the demographic information came from Chinese young adults and thus limited the application of this model to much older Chinese subjects. Besides, this model did not account for the change of cardiac output and liver blood flow in the geriatric population. These may have little impact on accurate prediction of PK of drugs with low hepatic extraction ratio (such as theophylline in the study) but are important physiological parameters for predicting PK of medications with a medium or high hepatic extraction ratio.
The objectives of our study are (i) to develop a PBPK population model for Chinese elderly subjects aged 65 and above, notably including those older than 75, based on more comprehensive physiological data; and (ii) to verify the Chinese geriatric population model using six substrates specifically eliminated by different pathways.

| Data collection
To generate a robust Chinese geriatric population model, comprehensive physiological data in elderly subjects of Chinese ancestry (referred to as Chinese elderly subjects) were collected. Data on age distribution, height, weight, cardiac output and serum creatinine were collected from a national survey (National physical fitness and health database, 2006-2011), involving approximate 7000 Chinese elderly subjects, the majority of whom are Han ethnicity. 19 The survey adopted a multi-stage and stratified random sampling method to ensure the representativeness of samples. Liver weight and kidney F I G U R E 1 Overall workflow of the Chinese geriatric physiologically based pharmacokinetic (PBPK) population model development and verification weight were taken from a national autopsy report, which summarized the healthy organ weights of 8273 subjects. 20 The physicochemical parameters of the model drugs were adopted from the Simcyp compound library or public domains (e.g., Scifinder, Drug Bank, Pubchem, FDA-approved drug labels).
The clinical data were collected from literatures written in either English or Chinese. They were included for model verification and refinement when they met the following criteria: (1) PK parameters of the clinical study are available or C max , AUC or clearance (CL) can be obtained or calculated from the concentration-time curve; (2) the number of subjects in the clinical study is greater than or equal to 4; and (3) the clinical study was conducted in subjects of either Chinese or European ancestry. Also, the clinical research was excluded for our model verification when: (1) the population in the study was not stratified by young adults and elderly subjects; and (2) the clinical trials were performed using intranasal administration route, as our primary aim was the verification of the Chinese geriatric population following intravenous (IV) and oral administration.

| Development of Chinese geriatric population model
The Chinese geriatric population model was developed based on an inbuilt Chinese adult population model in Simcyp with demographic information, cardiac output and serum creatinine recalibrated using the physiological data collected in Chinese elderly subjects aged above 65. 14 As illustrated in Supporting Information Figure S1, the age distribution was defined based on observed data from the national survey.
Height and weight were modelled against age using a polynomial  when compared to males (i.e., 269.9 ± 61.7 g). No public data on liver or kidney were available for Chinese elderly subjects older than 75.

| Development of the Chinese geriatric population model
Physiological data were then used to develop the Chinese geriatric population model in Simcyp by recalibrating the equations for describing the change of these physiological characteristics with age in the Chinese adult population model. Table 2 shows the recalibrated equations. Chinese elderly subjects. In general, the 5% and 95% percentiles of simulated height, weight, cardiac output, serum creatinine as well as the mean value of liver and kidney weight were comparable to the observed data at the age over 65, suggesting that the new Chinese geriatric population model was able to describe the physiological characteristics in the Chinese geriatric population (see Table 3).

| Development and verification of drug models
Drug-dependent parameters for simvastatin, theophylline, gentamicin, vancomycin and ceftazidime were adopted from the Simcyp library or developed based on in vitro and clinical data. The drug models were verified using clinical data from Caucasian adults, Chinese adults and Caucasian elderly subjects (Supporting Information Table S2). Figure 2 illustrates that more than 95% predicted AUC, C max , and CL values were within two-fold of the observed values, indicating the drug models for all six drugs were verified.

| Verification of the Chinese geriatric population model
The verified drug models were combined with the newly developed population model to predict drug concentration in Chinese elderly subjects. As illustrated in Figure 3, PBPK models could well describe the PK profiles of midazolam, theophylline, ceftazidime, vancomycin and gentamicin following the administration of drugs. For simvastatin, due to lack of concentration data reported in the literature for Chinese elderly subjects, no comparison of PK profile was possible.
Besides, the predicted AUC, C max and CL values were within the two-fold error range of observed values for all six drugs, among which 70% prediction on AUC, 78% prediction on C max and 70% prediction on CL were within a 1.25-fold error range of observed values (see Table 4).

| DISCUSSION
The geriatric population exhibits significant physiological changes such as a rise in gastrointestinal pH, delayed gastric emptying, decreased body muscle, increased body fats, lower GFR, reduced number of nephrons, and decreased hepatic blood flow. Alteration of these physiological changes may affect drug ADME. [37][38][39][40][41][42][43][44][45][46] In the previous study, it was shown that the PBPK model can be used to predict drug PK in Caucasian elderly subjects by incorporating these changes.
However, the study of the PBPK model to predict drug PK in Chinese elderly subjects is very limited, probably due to the lack of a robust Chinese geriatric population model. In this study, we collected physiological data, such as demographic information, cardiac output, liver weight, kidney weight and serum creatinine from Chinese elderly subjects and developed a Chinese geriatric population model in Simcyp.
Our study suggests that the newly established Chinese geriatric popu-

| Comparison of physiological parameters between Chinese and Caucasian geriatric populations
In the Caucasian geriatric population, body height declines by 2% per age decade after the age of 60, and the difference between men and women is constant. 13 Body weight increases in Caucasian subjects in their 50s and 60s and decreases after that by about 10% in each age decade. 13 Similar trends are observed in Chinese geriatric population.
In Caucasian males and females, cardiac output decreases by 5-10% every age decade after the age of 60. 13 A similar trend is observed in the Chinese geriatric population, with a 5% reduction in cardiac output from 65-75 to 75 years old and above.
Compared to adults aged 18-64, the liver weight of Caucasian elderly subjects (>65 years old) generally has 10-15% and 20% reduction for women and men, respectively. For kidney weight, the reduction starts with a loss of 5% at 70, 15% at 80, and later 25% at the age of 100 in both sexes. 13 In the Chinese population, the liver weight of 60-75 years old is 9% lower than that of men aged 18-59, and 2% for women. The kidney weight of men and women both reduce by 3% at 60-75 years old when compared to those aged 18-59 years old.
There is a progressive rise in the serum creatinine level in both males and females. 47 The serum creatinine level in elderly subjects aged 65-75 was increased by about 5% when compared to the Caucasian population aged 20-64. In elderly subjects over 75 years old, the serum creatinine level increased by 5-15%. Similarly, this increase was also observed in the Chinese geriatric population (see Table 1).

| Development of the Chinese geriatric population model
A novel Chinese geriatric population model was developed based on the Chinese adult population model in Simcyp. We recalibrated the height, weight, cardiac output and serum creatinine, which may potentially impact drug ADME, by using the data from the Chinese geriatric population aged 65 and above. Liver or kidney weight is usually estimated based on BSA. 48 As can be seen in Table S4 and Figure

| Limitation of the current study
In the current study, we established a Chinese geriatric population model in Simcyp. Different from the previous population model, which did not account for the change of cardiac output in the older population and thus is only capable of predicting drugs with low hepatic extraction, our population model can predict the concentration of drugs with low, medium and high hepatic extraction ratio. As there were only three drugs included in the analysis, further study will be needed to confirm the finding.
Secondly, we have confirmed that the model to estimate liver and kidney weight in healthy Chinese volunteers can also be used for older Chinese adults of 65-75 years old. However, due to a lack of data, model verification was not done in Chinese elderly subjects over 75 years old. Thus, the model should be extrapolated with caution for Chinese elderly subjects over 75 years old.
Lastly, as our new population model did not incorporate the number of nephrons and abundance of renal transporter, it may not be able to predict drugs with significant active secretion.

DATA AVAILABILITY STATEMENT
The data used to support the findings of this study are available from the corresponding author upon request. ORCID