Application of physiologically-based pharmacokinetic modeling to inform dosing decisions for geriatric patients

Abstract

Modeling approaches, including population pharmacokinetic (popPK) and physiologically-based pharmacokinetic (PBPK) modeling, have been applied to simulate the complex interplay between pharmacokinetics and aging-related pathophysiologic changes.¹ Compared with the limited popPK modeling, there are several published PBPK models for older adults (Table S1). Despite the increasing PBPK models, three research questions remain that are discussed below.

Aging is often accompanied by changes in the anatomy and physiology of multiple tissues and organs. Considerations for developing a mechanistic population model for White or Chinese geriatric patients are detailed in various publications (Table S1). Healthy geriatric population models have been implemented in or can be generated using built-in algorithms within PBPK modeling software (e.g., Simcyp, GastroPlus, PK-Sim). A workflow for developing a geriatric population model is presented (Figure 1).

In addition to the aging process, older adults are susceptible to chronic diseases that lead to functional changes in major organs governing drug disposition, including the liver and kidney.^{2, 3} These diseases are often accompanied by polypharmacy, resulting in complex disease-drug–drug interactions (D-DDIs), which further increase the risk of pharmacotherapeutic failure. While PBPK modeling has been widely used for DDI evaluation, few studies have explored prospective simulations and evaluations for geriatric patients with health impairments due to the compounded effects of aging and chronic disease on drug disposition, along with scarce clinical data for robust evaluation.

Virtual health-impaired geriatric populations are generally developed from virtual healthy geriatric populations via different methodologies. For example, a renal impairment (RI) geriatric model was developed using the healthy geriatric model within Simcyp, with adjustments made to the physiological parameters of the kidney corresponding to RI severity.⁴ The same approach was applied using PK-Sim. The geriatric model was first scaled from a younger adult model, after which drug systemic exposure was simulated for older adults with RI or hepatic impairment (HI) based on the prevalence of renal and hepatic dysfunction in this population (Table S1). This approach enabled consideration of organ dysfunction arising from both disease and aging. Another approach began with the Simcyp built-in RI or HI population model, which was adjusted for older age ranges within the models (Table S1). However, the lack of clinical data precluded comparing and evaluating these two approaches for the same drug, warranting further investigation.

In addition to considering the general changes resulting from aging and disease, enzymes or transporters involved in the disposition of a specific drug may necessitate further attention. For example, human organic anion transporter (hOAT) 3 mediates the renal tubular secretion of the anticoagulant rivaroxaban. During RI, circulating uremic solutes compromise hOAT3 activity. Therefore, modifications to hOAT3 activity were incorporated into the PBPK RI model to reflect the disproportionately greater deterioration in hOAT3-mediated secretion of rivaroxaban compared with glomerular filtration rate (GFR).⁴ Similarly, available clinical data for uridine 5′-diphosphate-glucuronosyltransferase (UGT) 2B7 and butyrylcholinesterase (BChE) substrates were leveraged to predict the impacts of renal dysfunction and aging on enzyme activities. Results were integrated into a PBPK model used to predict the disposition of the dual UGT2B7/BChE substrate mirabegron (Table S1). These studies provided valuable insights into the use of available pharmacokinetics of known drug substrates for prospective simulation of another drug sharing the same disposition pathway.

PopPK analysis provides another feasible approach using available clinical data for improving prospective PBPK simulations. A popPK analysis was conducted for Chinese geriatric patients with congestive heart failure (CHF) that involved the P-glycoprotein (P-gp) substrates and digoxin.⁵ Age was identified as a significant covariate of the clearance of both drugs. Subsequent PBPK sensitivity analysis highlighted the importance of adjusting aging-related variation in P-gp function in both the intestine and liver in addition to CHF-induced RI. Furthermore, PBPK predictions and popPK model-derived estimates were simultaneously made and compared for dose optimization. This approach was also applied to the antihistamine bilastine to identify key age-dependent variables impacting drug disposition among healthy older adults and support PBPK model-informed dose selection (Table S1). This integrated pharmacometric method, which leverages the strength of popPK analysis to identify covariates of variability and estimate pharmacokinetics via retrospective analysis of sparse data, applied PBPK modeling to prospectively simulate untested clinical scenarios.

PBPK platforms use different older adult databases for pharmacokinetic predictions. Simcyp and PK-Sim incorporate the database established by Thompson et al.⁶ and Schlender et al.,⁷ respectively. Stader et al.⁸ built a database using Matlab. The module within GastroPlus accounts for aging-related physiological changes to predict drug exposure; however, literature sources have not been published. These platforms include most physiological changes in older adults, including demographics, tissue weights, cardiac output, tissue blood flows, and GFR.

Some physiological changes in older adults are usually not considered. For example, due to conflicting reports on gastric emptying time in older adults, absorption parameters from younger adults are often applied to older adults. Regarding distribution, the changes in adipose tissue mass with age differ between databases. Stader's database suggested that adipose tissue weight increases with age until 78 years, while Schlender's database reported a peak at 70 years in females and 65 years in males.^{7, 8} Although Schlender's database described changes in adipose tissue weight with age, the relevant estimation equation was not provided, possibly because fat distribution is related to the age of menopause in females.

Lack of information on changes in transporters in older adults was observed in PBPK models used to predict the pharmacokinetics of relevant drugs in this population. For example, the aforementioned PBPK model for bilastine incorporated intestinal transporters to describe secretion and absorption (Table S1). When predicting systemic exposure in older adults, transporters were assumed to be independent of age due to data limitations. Similarly, relative transporter abundance was assumed to be unchanged with age in older adults when predicting the renal excretion of ganciclovir. This assumption may lead to prediction inaccuracy because ganciclovir is primarily excreted via hOAT1, and no experiment has been conducted with older adults. Besides transporter function, Alikhani et al.⁹ suggested the entire age spectrum equation. This estimated GFR equation was developed using GFR data obtained from individuals aged 2–97 years and represents an alternative method for predicting GFR in older adults. This model is based on biological age instead of chronological age, which can capture organ health quality.

Another limitation of the established databases and models for older adults is the racial and ethnic distributions of the study populations. Thompson's database primarily represents Japanese and White males, whereas Schlender's and Stader's databases focus on older European adults and the White population, respectively.^6-8 Additionally, Cui et al.⁵ developed PBPK models for Chinese older adults. This demographic is now included in the Simcyp virtual population. Clearly, gaps remain in PBPK modeling for older adults from other racial and ethnic groups.

Consumers, including older adults, are turning increasingly to botanical and other natural products for myriad purported beneficial effects. As aforementioned, older adults are prone to polypharmacy, rendering them at high risk for natural product–drug interactions. Like drugs, natural products can inhibit (e.g., grapefruit juice) and induce (e.g., St. John's wort) drug metabolizing enzymes and transporters, leading to an increase or decrease in object drug exposure. Relative to drugs, robust PBPK models for natural products remain lacking, which is due largely to their inherent complex chemical composition and the scant human pharmacokinetic knowledge of key phytoconstituents. With the increasing sensitivity of bioanalytical instrumentation and continual updates to the various PBPK modeling platforms, progress has been made in recent years. For example, models have been developed and verified to describe the disposition of select phytoconstituents contained in cannabis, goldenseal, and kratom in healthy adults.¹⁰ These models were then used to predict the interaction risk with various object drugs. As for drugs, refinement of these models with new clinical pharmacokinetic data, along with considering aging- and disease-related changes, should enable prediction of phytoconstituent disposition and drug interaction risk in this population.

A framework for developing a PBPK model for health-impaired geriatric patients that incorporates our focused insights is presented (Figure 1). Although physiological changes due to aging and disease may be partially elucidated through in vitro experiments, clinical data, whether sparse or abundant in the database, are crucial and necessary for refining PBPK models for this vulnerable cohort. This multipronged strategy should enhance both the quality and precision of prospective PBPK simulations for health-impaired geriatric populations. Such simulations can help inform dosing decisions for geriatric patients, as well as the design of future clinical trials.

More studies involving older adults are needed to resolve conflicting reports of aging-related physiological changes. Addressing the racial and ethnic diversity gaps in databases also requires attention. Relying solely on chronological age to predict physiological changes in older adults may lead to inaccurate results. Future studies should consider assessing biological aging by evaluating genetics, epigenetics, composite biomarkers, proteomics, and metabolomics. As for all patients, linking pharmacodynamic end points with PBPK models can further inform dosing decisions for geriatric patients. However, knowledge of pharmacodynamic changes related to age-associated factors is limited, requiring further research. Finally, generative artificial intelligence and deep machine learning could be used to extend current efforts in predicting the pharmacokinetics and pharmacodynamics of drugs in geriatric patients.

In summary, the judicious development and application of robust PBPK models improve the prediction accuracy of drug pharmacokinetics in geriatric patients and guide model-informed dosing regimen design. By bridging knowledge gaps and offering viable examples, this perspective stimulates further research in geriatric PBPK modeling, thereby promoting the advancement of personalized medicine for this understudied patient population.

Z.Z. was supported by the 2023 ASCPT Darrell Abernethy Early Stage Investigator Award, NIH/NIGMS (Grant R16 GM146679). M.F.P. was supported by NIH/NCCIH (Grant U54 AT008909). E.C.Y.C. was supported by the Joseph Lim Boon Tiong Urology Cancer Research (Grant A-0002678-01-00).

M.F.P. is a member of the Scientific Advisory Board for Simcyp, Certara UK Limited. All other authors declared no competing interests for this work.