Abstract by Ida Klitzing Storgaard
Model-based optimization of dosing of high-risk medication in older patients
Older adults are overrepresented in emergency department settings, yet are relatively underrepresented in clinical studies, in spite of, or perhaps due to, the many medication-related risk factors associated with aging. Malnutrition and decreased kidney function are two prevalent factors that can impact drug pharmacokinetics in a variety of ways and, if unaccounted for, can lead to inappropriate dosing. Drug pharmacokinetics, i.e. concentration-time data, can be analyzed by population pharmacokinetic modeling, where parameters such as drug clearance are simultaneously estimated for the population and the individual, and variability can be identified, quantified, and possibly explained by covariates.
The aim of this PhD thesis was to investigate and evaluate approaches to dose optimization of high-risk drugs in older medical patients. Through population pharmacokinetic modeling of data from a clinical trial, this thesis sought to describe and investigate sources of variability in the pharmacokinetics of cannabis-based medicine, and to assess the performance of ten equations for estimating kidney function (based on one, two, or four biomarkers) and three methods for measuring kidney function (with exogenous filtration markers) compared to the model-estimated clearance of the renally eliminated drug gentamicin.
In the clinical trial, 20 older patients with poor appetite were given two doses of 2-3 sprays Sativex®, an oromucosal mouth spray containing 2.7 mg Δ9-tetrahydrocannabinol (THC) and 2.5 mg cannabidiol per spray. The pharmacokinetics of THC and its active metabolite, 11-hydroxy-Δ9-tetrahydrocannabinol (THC-OH), were best described as a parent-metabolite model with one compartment for THC, two compartments for THC-OH, and absorption of THC through three transit compartments. There was large inter-individual and inter-occasion variability (coefficient of variation = 40.2–152%) on structural model parameters. This could not be explained by any of the available covariates, which included a variety of clinical and paraclinical characteristics and body composition parameters (e.g., muscle mass and fat percentage). The large variability may be due to underlying variability in bioavailability. The oromucosal route of administration should be used with caution in this patient population, as the results indicated a high degree of unreliability in predicting exposure levels.
The pharmacokinetics of gentamicin after intravenous administration to 52 older patients were best described by a two-compartment model with inter-individual variability on all structural parameters. Nine equations for estimated glomerular filtration rate (eGFR), one equation for estimated creatinine clearance (eCrCL) and measured glomerular filtration rate (mGFR) obtained by assessing the clearance of 99mTc-DTPA were tested as covariates on gentamicin clearance. Overall, it was found that expressing mGFR, eGFR, and eCrCL in absolute units (mL/min) instead of body surface area-indexed units (mL/min/1.73 m2) resulted in better prediction accuracy of gentamicin clearance. As expected, mGFR provided the best model fit based on drop in objective function value and reduction in inter-individual variability on clearance. Of the equations, eGFR calculated from a combination of creatinine and cystatin C levels performed best, along with eGFR calculated from a panel of four biomarkers (creatinine, cystatin C, β-trace protein, and β-2 microglobulin); the addition of two extra biomarkers did not improve performance. Equations based on two or more biomarkers performed significantly better than single-biomarker equations. The 1976 Cockcroft-Gault equation for eCrCL, which is still widely used for assessment of kidney function in population pharmacokinetic studies, performed significantly worse than the eGFR equations when all were expressed in absolute units.
A modified version of the gentamicin model was used to estimate gentamicin clearance in 51 older patients who simultaneously had mGFR determined by three methods: intravenous sampling of iohexol and 99mTc-DTPA, and dried blood spot sampling of iohexol. All three methods were found to be accurate predictors of gentamicin clearance when implemented as covariates in the model, with intravenous sampling of iohexol providing the best model fit. Based on both the model-based analysis and other performance metrics, dried blood spot sampling was found to be a reliable alternative to intravenous sampling of iohexol, presenting a less cumbersome option for determining mGFR in clinical settings, potentially making more frequent and accurate assessment of kidney function as mGFR feasible for optimized dosing of renally excreted drugs.
In conclusion, this thesis presents approaches to optimizing dosing of high-risk drugs in older patients, particularly renally excreted drugs. For this patient group, the findings shows that assessment of kidney function for the purpose of dose adjustment in both drug development and clinical practice could be improved by switching to eGFR equations based on a combination of creatinine and cystatin C levels, and that dried blood spot testing is a reliable alternative to intravenous sampling of iohexol when more accurate assessment is needed.