P. Girard, L.B. Sheiner
INSERM & Departments of Laboratory Medicine and Pharmacy, University of California, San Francisco CA.
Drug taking behavior (compliance) varies between and within patients, as revealed by electronic monitoring devices placed on pill containers. In order to determine whether compliance depends on patient factors, and other covariates, we have developed an integrated model for dose taking times, that can be used either to test various assumptions, or for simulation purposes.
This model is based on the following assumptions: (i) each patient takes drug at times (T) distributed within an interval about a fixed number of individual nominal times (NT), the latter possibly different from the times prescribed by the physician, but equal in number; (ii) the time-errors made by the patient when taking his doses are normally distributed around T; (iii)the probability of taking or not taking a given dose within a given NT interval (interval boundaries bisect the inter-NT times) depends on the number of doses taken in the immediate previous interval, as well on time dependent (week-end, morning or evening doses, …) and time independent covariates (age, gender, …). We identify the full population model in two stages using maximum likelihood estimation for both fixed and random effect model parameters. All analyses are performed using the NONMEM software.
The methodological development of the model as well as an application to data collected over three months using the MEMS system, from HIV+ patients who had been prescribed AZT three times a day, will be presented. Applications of the model are numerous and promising: e.g., to explain poor compliance, drug holidays, or changing compliance using sociological or pharmacological covariates, or to enhance the precision of population PK/PD analyses
Reference: PAGE 5 () Abstr 560 [www.page-meeting.org/?abstract=560]
Poster: oral presentation