European Journal of Clinical Pharmacology

, Volume 70, Issue 5, pp 549–555

The impact of frailty on pharmacokinetics in older people: using gentamicin population pharmacokinetic modeling to investigate changes in renal drug clearance by glomerular filtration


  • Claire Johnston
    • Sydney Medical SchoolUniversity of Sydney
    • Department of Clinical Pharmacology and Aged Care and the Kolling Institute of Medical ResearchRoyal North Shore Hospital
    • Sydney Medical SchoolUniversity of Sydney
    • Department of Clinical Pharmacology and Aged Care and the Kolling Institute of Medical ResearchRoyal North Shore Hospital
    • Laboratory of Ageing and Pharmacology, Level 12 Kolling Institute of Medical ResearchRoyal North Shore Hospital
  • Andrew J. McLachlan
    • Centre for Education and Research on AgingConcord RG Hospital
    • Faculty of PharmacyUniversity of Sydney
  • Slade T. Matthews
    • Sydney Medical SchoolUniversity of Sydney
  • Peter R. Carroll
    • Sydney Medical SchoolUniversity of Sydney
    • Department of Clinical Pharmacology and Aged Care and the Kolling Institute of Medical ResearchRoyal North Shore Hospital
  • Carl M. Kirkpatrick
    • Centre for Medicine Use and SafetyMonash University
Pharmacokinetics and Disposition

DOI: 10.1007/s00228-014-1652-7

Cite this article as:
Johnston, C., Hilmer, S.N., McLachlan, A.J. et al. Eur J Clin Pharmacol (2014) 70: 549. doi:10.1007/s00228-014-1652-7



Frailty, a multifactorial biological syndrome characterized by a cumulative dysregulation of physiological processes, is associated with changes in pharmacokinetics and pharmacodynamics. The aim of this study was to quantify the effect of frailty on glomerular filtration of drugs, using the probe drug gentamicin.


Gentamicin concentrations and clinical data including the Reported Edmonton Frail Scale score were pooled from two prospective observational inpatient studies, one on prophylactic gentamicin for urologic surgery and one on therapeutic gentamicin for the empiric treatment of sepsis. Population pharmacokinetic modeling was performed using non-linear mixed effects modeling (NONMEM program) to determine the impact of frailty on gentamicin clearance.


A one-compartment linear pharmacokinetic model best described the data and the addition of frailty to the model reduced the random variability in gentamicin clearance by 12 % after adjustment for renal function (estimated creatinine clearance using lean body weight) and lean body weight. Frail patients had an approximately 12 % lower (bootstrapping results: 14 % median) gentamicin clearance than non-frail patients (calculated as a fractional effect of frailty).


Frailty may independently predict reduced clearance of gentamicin in older patients. Frailty could be considered in the development of dosing guidelines for drugs that undergo significant excretion through glomerular filtration.


FrailtyAgeingPharmacokineticsPopulation pharmacokineticsRenal clearanceGlomerular filtrationGentamicin


As the population ages, it is increasingly important to consider the needs of older people in the development, approval and use of medicines [1]. In addition to chronologic age, it is important to consider the impact of biologic age or frailty on clinical pharmacology in older people. Frailty is a complex multifactorial biological syndrome that is characterized by a cumulative dysregulation of physiological processes and decline in homeostatic reserve [2]. Its prevalence increases with age and it is associated with adverse clinical outcomes, such as hospitalisation, morbidity and mortality [3], as well as different patterns of medication use including polypharmacy [4]. Frailty has a heterogeneous clinical presentation and a multitude of definitions [5]. The European Medicines Agency recognizes the need for an operational definition of frailty that can be used in research and to guide therapeutic decisions [1].

Many physiological changes that can occur with aging and frailty potentially impact the clinical pharmacology of medicines [6]. Older people have increased sarcopaenia (decline in skeletal muscle mass), which may be accompanied by increased adiposity, which is exaggerated in frailty [7]. Hepatic clearance of drugs decreases, to varying degrees, with age. Studies measuring the impact of different measures of frailty on different measures of hepatic clearance are inconsistent [4, 8]. Renal function declines with age, although the rate at which this occurs is contentious [6] and the independent impact of frailty has not been assessed [4]. As frailty progresses and homeostatic physiological systems become impaired, often with concurrent multimorbidity and polypharmacy, the differences between younger people, robust older people and frail older people become more marked, with implications for pharmacokinetics and drug dosing [9].

There is much debate and controversy over the operationalization and definition of frailty [5, 1012]. Frailty, however it is measured, is strongly associated with adverse outcomes such as falls, institutionalization, morbidity and mortality [4, 1320]. The two prevailing philosophies of frailty are based on physical function measures or more global measures of deficit accumulation. The ‘frailty phenotype’, based on the Cardiovascular Health Study [18], characterizes older people as frail, pre-frail or non-frail/robust based on five factors: weight loss, weakness, exhaustion, slow walking speed and low physical activity. It predicts adverse outcomes in many epidemiological studies. The ‘frailty phenotype’ is heavily influenced by acute illness, limiting its use in the acute hospital setting, and does not include cognitive, social, pharmacological and comorbidity contributions to the development and progression of the frailty syndrome. Frailty scales that are based on deficit accumulation such as the Frailty Index [21] and Edmonton Frailty Scale [22] do not have these limitations. The Reported Edmonton Frailty Scale (REFS, see Supplementary Material) [23], which was derived from the Frailty Index, has been validated against the Geriatrician’s Clinical Impression of Frailty, and is associated with drug use and clinical outcomes in studies of older hospitalized patients [24, 25].

Frailty has not been extensively investigated in relation to changes in pharmacokinetics. We are aware of only nine papers that have used some measure of frailty to look at changes in pharmacology [8, 2633]. Only three of these were published after publication of the first validated objective frailty scale (and only two investigate pharmacokinetics) [18]. All nine of these papers report changes, across various metabolic and elimination pathways, in frail older people compared to non-frail participants. The study by Hilmer et al. demonstrated that gentamicin clearance was significantly lower in frail patients compared to non-frail participants receiving prophylactic gentamicin [26]. Pharmacokinetic studies conducted during drug development often exclude older people, and rarely include frail older people, who may not be able to tolerate collection of multiple blood samples over prolonged periods and bring increased variability to the data [34]. Dosing recommendations are typically based on data from younger and/or more robust older people than the patients who receive the medicines in the setting of acute care, who are typically older, frail, receiving multiple medications, have multiple comorbid conditions and are extremely heterogeneous [1].

The use of a population approach to estimate the pharmacokinetic/pharmacodynamic profile of drugs is becoming common practice in a wide range of patient populations. It provides a description and quantification of the drug concentration-time profile in the body, the between-patient variability that can be explained by patient covariates, such as weight and renal function, the between-patient variability that cannot be described by covariates, as well as random, unexplained variability [35]. Population pharmacokinetic analysis facilitates pharmacokinetic studies in older people and could potentially identify and quantify the role of factors such as frailty in drug disposition and response. Modeling and simulation can also inform dose recommendations [36] and may be particularly useful in older people by taking into account these potential sources of variability [9].

The use of probe drugs to phenotype patients and to obtain a general picture of various metabolic and elimination pathways are well established [3739]. Gentamicin is a hydrophilic intravenous antibiotic that is primarily excreted unchanged in the urine by glomerular filtration [40]. This makes it a very suitable marker for renal drug clearance by glomerular filtration (not secretion). It has previously been suggested as a surrogate marker of drug-induced nephrotoxicity [41]. It is also a good example of a drug in which efficacy and toxicity depend on peak concentration (Cmax determines bacteriocidal action of gentamicin) and area under the concentration time curve (AUC, determines its nephrotoxicity) [40, 42, 43]. Dosing is challenging because, like other aminoglycosides, gentamicin has a narrow therapeutic index and its pharmacokinetics vary with age, body size and composition, renal function and pathophysiology [40, 43, 44].

We hypothesized that in older patients, frailty has an independent impact on renal drug clearance by glomerular filtration, measured using gentamicin clearance, due to the marked physiological differences between frail and robust older people. The aim of this study was to investigate the pharmacokinetics of gentamicin in a cohort of hospitalised older people after a single dose of gentamicin to quantify the impact of frailty on the clearance of gentamicin and identify influential covariates that describe the variability in gentamicin pharmacokinetics, as predictors of renal drug clearance by glomerular filtration in older people.


Patient population

Gentamicin dosing, concentration and clinical data were available from two prospective, observational studies of patients aged 65 years and over from Royal North Shore Hospital, a teaching hospital in Sydney, Australia. One study recruited 30 patients receiving prophylactic gentamicin prior to cystoscopy as reported previously [26]. This study found that there was a significant difference in the clearance of gentamicin (calculated using the TCIWorks software: between frail and non-frail older people. They also found that creatinine clearance calculated using ideal body weight [45] in the Cockcroft-Gault equation [46] best correlated with gentamicin clearance using a Bland-Altman analysis [47]. The second study recruited eight patients receiving empiric treatment with gentamicin for suspected sepsis. Both studies recruited patients receiving first-dose gentamicin only. Patients were excluded if they had severe cognitive, hearing or visual impairment (and no person responsible to provide consent on their behalf). Data collection was identical to the previously reported study of prophylactic gentamicin [26], including collection of number of medications, comorbidities [48] and renal disease risk factors [49].

Frailty was determined using REFS (see Supplementary Material) [23], which was administered by the researchers. This scale was chosen because it is validated to be administered in the acute setting by non-medically trained personnel with good inter-rater reliability (kappa = 0.84, n = 31). Furthermore, REFS is correlated with the current gold standard of frailty quantification, the Geriatrician’s Clinical Impression of Frailty. Participants with a REFS score greater than or equal to 8 out of a possible 18 were considered 'frail'. A cut-off REFS score of 8 was chosen based on the previous validation paper [23] and has been shown previously to predict drug use [25], pharmacokinetics and pharmacodynamics [24].

Blood samples were taken by venipuncture at 0.5–1, 2 and 4–6 h after gentamicin infusion. The sampling times were chosen to give maximum information about the pharmacokinetic parameters volume of distribution (0.5–1 h and some from 2 h) and clearance (some from 2 h and 4–6 h). These were based on previous studies of single dose gentamicin [50]. Blood samples were analysed by the nationally accredited hospital laboratory. For each patient, concentrations of serum creatinine were measured on the Roche Modular Autoanalyzer. Gentamicin concentrations were measured using a Siemens Dimension RxL analyzer with Siemens reagents and calibrators. At mean gentamicin concentrations of 2.5 mg L−1, 5.3 mg L−1 and 7.1 mg L−1 the coefficients of variation were 5.6 %, 2.8 % and 2.3 %, respectively. The lower limit of gentamicin quantitation was 0.5 mg L−1.

Data analysis

Population pharmacokinetic analysis was carried out in NONMEM [51] (version 6, Icon Development Solutions, Ellicott City, Maryland, USA, 2006) using a G77 FORTRAN compiler, with runs executed using Wings for NONMEM [52]). The first-order conditional estimation method with interaction was used to analyse data. One- and two-compartment linear pharmacokinetic models were evaluated as possible candidate models to describe gentamicin pharmacokinetics and assessed according to the lowest objective function (reduction in objective function ≥3.84 is equal to p < 0.05) obtained and visual inspection of goodness of fit plots. All models were parameterised in terms of volumes and clearances.

Potentially influential covariates were screened and added according to biological plausibility, firstly, then to statistical significance and finally to the reduction in between subject variability, in a step-wise fashion. Covariates available for screening for inclusion into the population pharmacokinetic model are listed in Table 1. Lean body weight was calculated with the semi-mechanistic equation [53], which has been evaluated against dual energy X-ray absorptiometry in robust and frail older people previously [7, 54]. The round-up method for the lower limit of serum creatinine of 0.06 mmol−1 (which allows for the decreased production of creatinine in some patients and has been shown to be the most suitable cut-off point [55]) and normalised values for body weight and CRCL were also evaluated. The final model was assessed using visual predictive checks showing 10th, 50th and 90th percentiles; nonparametric bootstrapping (n = 1,000 simulated patients) and calculation of 95 % confidence intervals; and examination of observed vs predicted values.
Table 1

Participant characteristics




Number of patients


74 % Male

79 % Prophylactic

50 % Frail

Age (years)



Weight (kg)



Ideal Body Weight (kg)a



Lean Body Weight (kg)b



Number of Medications



Number of Comorbiditiesc



Number of Renal Disease Risk Factorsd



Estimated Creatinine Clearance (mL/min)e



eGFR (mL/min/1.73m2)f



Samples per Patient



Gentamicin Dose (mg)



Gentamicin Dose (mg/kg)



aCalculated using the Devine et al., 1974 equation [45]

bCalculated using the Janmahasatian et al., 2005 equation [51]

cCalculated using Charleson Comorbidity Index criteria [48]

dCalculated using the criteria from Levy et al. [49]

eCalculated using Cockcroft-Gault [46] equation with lean body weight [51] as the weight variable

fEstimated glomerular filtration rate, calculated using the MDRD equation [56]


The pharmacokinetics of gentamicin, as a marker of renal drug clearance by glomerular filtration, were investigated in 38 older patients (30 receiving prophylactic gentamicin and eight therapeutic gentamicin treatments) with a total of 89 gentamicin concentration observations. Participants had a mean age of 80 years, 50 % were classified frail and 74 % were male (Table 1).

A one-compartment linear pharmacokinetic model best described the disposition of gentamicin in this cohort of patients. The population pharmacokinetic model included a separate residual error model for comparing the patients receiving prophylactic gentamicin with those patients receiving therapeutic gentamicin, due to the potential increased error in the execution of the therapeutic part of the study (i.e. dosing errors, sampling errors). The random component of the between-subject variability in gentamicin volume of distribution was decreased from 14 to 10.4 % with the addition of lean body weight (normalized to the median). Between-subject variability in gentamicin clearance was reduced from 32.1 % to 23 % with the addition of CRCL using lean body weight (normalized to the median and using the round-up method for serum creatinine concentration) to clearance. The effect of frailty (as a fractional effect) further reduced the variability to 20.5 % when added to the clearance parameter (Table 2). The final fractional effect of frailty was 0.88, i.e. frail patients have a 12 % decrease in gentamicin clearance compared to non-frail patients, even after controlling for other factors known to impact gentamicin clearance. The average volume of distribution was the same for frail and non-frail patients and for both those being treated for sepsis and those receiving prophylactic gentamicin: 0.20 L/kg for total weight and 0.27 L/kg for lean body weight.
Table 2

Parameter estimates for the base structural model and the final covariate model with bootstrapped results for 1,000 simulated patients


Base Model

Covariate Model

1,000 Bootstrap Replicates Median (95 % CI)

Objective Function Value




CL (mL/min)



101.7 (91.7–111.7)

V (L)



14.6 (13.9–15.3)



86.2 (75.5–99.0)

Random Parameters (CV%)




19.8 (7.3–27.2)




9.6 (2.2–16.8)

CL Gentamicin clearance, V Volume of distribution, BSV Between subject variability, FFRAL Fractional effect of frailty


This study, using population pharmacokinetic modeling, found that in older adults, a validated measure of frailty may independently predict reduced clearance of gentamicin, a drug excreted by glomerular filtration.

The reduction of gentamicin clearance with frailty is biologically plausible since frailty is associated with sarcopaenia (25), and gentamicin is a highly hydrophilic drug that is distributed into muscle mass or lean body weight [40]. Lean body weight, when used on the volume parameter and in the creatinine clearance equation, was shown to improve the model compared to using either total or ideal body weight. This lean body weight equation has shown consistency and adequate accuracy in predicting lean body weight compared to a gold standard in frail older people [7, 54]. This reflects the importance of using appropriate size metrics when calculating doses of hydrophilic drugs like gentamicin and in the calculation of GFR, unlike in the MDRD equation [56]. However, frailty was observed to have an additional independent relationship with clearance.

The volume of distribution of gentamicin (when expressed as a fraction of body weight) showed no difference between participants treated for sepsis or prophylaxis, or between frail and non-frail patients. We had expected to observe a difference between septic and prophylactic patients as septic patients have previously been shown to have increased volume due to leakage of fluid from the vessels [57].

This research highlights the usefulness of population modeling and simulation in pharmacokinetic studies in older people. The previous study by Hilmer et al. demonstrated that amongst older patients receiving prophylactic gentamicin, clearance was significantly lower in frail than non-frail patients [26]. The present study quantified the impact that frailty has on gentamicin clearance, independent of renal function and weight. Population pharmacokinetic modeling allows for a sparse sampling of patients, which is important for pharmacokinetic studies in acutely unwell older inpatients. It identifies and quantifies the effect of covariates such as age and weight in reducing the variability that is seen in the population. This is invaluable in geriatric pharmacology research as there is an innate increase in variability and heterogeneity seen in the aging population and attempts to identify and quantify this will allow for more accurate dosing and dose adjustment of those aged over 65 years with multiple medications, multiple comorbid conditions and changes in physiology seen with age and frailty [9].

Strengths of the study include detailed, objective clinical data collection; inclusion of patients with a wide range of weights, renal functions and ages (no upper age limit); both prophylactic and therapeutic indications for gentamicin; prospective collection of frailty status using a validated tool; and the gentamicin assay and serum creatinine were conducted at a single, accredited hospital laboratory.

Limitations of our study include the relatively small sample size (n = 38), with only eight patients receiving therapeutic gentamicin for sepsis. This highlights the difficult nature of recruiting older, acutely unwell patients for a pharmacokinetic study [58]. Longer sampling times may also be useful for future studies to more accurately capture the terminal phase of elimination, which may mean a two-compartment model is more suitable for describing the data, as it has been in some previous gentamicin studies [44]. Gentamicin is renally cleared by glomerular filtration and these results cannot be applied to drugs that undergo active renal tubular secretion or to hepatically cleared drugs.

This study has shown that frail patients, assessed using an objective validated scale, have decreased gentamicin clearance compared to non-frail patients after body size/composition and renal function have been taken into account. Frailty is a multifactorial biological syndrome that is independently associated with reduced clearance of gentamicin.

These findings provide some insight into the effects of frailty on drugs cleared renally by glomerular filtration and highlights the need for further studies into this area. The study demonstrates the feasibility and importance of using population pharmacokinetic modeling and simulation to understand and apply changes in pharmacokinetics in frail older people, potentially improving the safety and efficacy of medicines for this vulnerable population.


We would like to thank the Geoff and Elaine Penney Aging Research Fund for financial support. We would also like to acknowledge Dr. Doug Chesher and PaLMS for the analysis of the drug concentrations; Dr. Kashyap Patel for assistance with the model evaluation; and Ms. Kim Tran, Dr. Patrick Rubie and Dr. Jason Wright for the collection of the prophylactic data.

Conflict of interest

The authors declare no conflicts of interest.

Supplementary material

228_2014_1652_MOESM1_ESM.docx (253 kb)
Supplementary material(DOCX 253 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2014