Electronic monitoring of variation in drug intakes can reduce bias and improve precision in pharmacokinetic/pharmacodynamic population studies

Abstract

Population pharmacokinetic (PK) and pharmacodynamic (PD) studies evaluate drug concentration profiles and pharmacological effects over time when standard drug dosage regimens are assigned. They constitute a scientific basis for the determination of the optimal dosage of a new drug. Population PK/PD analyses can be performed on relatively few measures per patient enabling the study of a sizable sample of patients who take the drug over a possibly long period of time. We expose the problem of bias in PK/PD estimators in the presence of partial compliance with assigned treatment as it occurs in practice. We propose to solve this by recording accurate data on a number of previous dose timings and using timing-explicit hierarchical non-linear models for analysis. In practice, we rely on electronic measures of an ambulatory patient's drug dosing histories. Especially for non-linear PD estimation, we found that not only bias can be reduced, but higher precision can also be retrieved from the same number of data points when irregular drug intake times occur in well-controlled studies. We apply methods proposed by Mentré et al. to investigate the information matrix for hierarchical non-linear models. This confirms that a substantial gain in precision can be expected due to irregular drug intakes. Intuitively, this is explained by the fact that regular takers experience a relatively small range of concentrations, which makes it hard to estimate any deviation from linearity in the effect model. We conclude that estimators of PK/PD parameters can benefit greatly from information that enters through greater variation in the drug exposure process.

Predicting Plasma Concentration of Risperidone Associated With Dosage Change

“…In addition, although we requested the subjects to record the time of last dose, some of these outliers may reflect a discrepancy between the reported and actual time of last dose and reflect individuals with poor adherence. Specifically, Vrijens and others [26][27][28] have demonstrated that the precision and bias of a population PK analysis improve substantially (a term that would reflect significant deviations in the predictions after accounting for between subject variability), and this improvement is explained with a more detailed and reliable measure of adherence to the medication regimen.…”

Electronic monitoring of patient adherence to oral antihypertensive medical treatment: a systematic review

“…This relatively new approach of analysing electronic monitoring data definitely contributes to a more detailed understanding of nonadherence, but it has not been in wider use so far. In combination with electronic monitoring of patient adherence, a variety of additional more or less complicated statistical methods have been proposed to more accurately analyse patient adherence [27,28,81,82].…”

Modelling the association between adherence and viral load in HIV-infected patients

“…Patients that regularly take their antiretroviral medication as prescribed experience a relatively small range of drug exposures relative to those patients that do not take their medications as prescribed, which makes it di cult to estimate deviation from normality in the viral load model. Estimates of the proposed model parameters could beneÿt greatly from information that enters through greater variation in the drug exposure process [15]. Such data could be collected from patients with poorer adherence, as observed in usual practice.…”