A. Rostami-Hodjegan, GT Tucker
University Dept. of Medicine and Pharmacology, Royal Hallamshire Hospital, Sheffield, S10 2JF
The POP-PK literature is remarkably silent in providing guidance on estimating t1/2 and its variance when, as is usual, the analysis is parameterised for CL and V. Few POP-PK studies report half-lives and those that do rarely indicate their likely variance. An accurate knowledge of t1/2 is particularly pertinent in cases such as methadone population kinetics since the treatment of methadone toxicity involves the use of opiate antagonists that have much shorter half-lives than methadone itself. As a practical example, we calculated the population value of methadone half-life and its variance in 13 healthy subjects by four different approaches using the P-Pharm software. A bi-exponential disposition model was used since it was shown to be superior to models with mono-exponential disposition.
Values of t1/2 and its variance were estimated by following 4 methods:
1- The terminal elimination half-life (t1/2 z) was considered to be a secondary parameter and population values were calculated using the following equations:
Assuming no covariation between the components of t1/2 z, a measure of the variance of its population value was calculated using the following approximation [33] (P-Pharm does not provide a covariance matrix for parameters):
(where y is dependent variable, x1, ….,xn are independent variables and dy/dxi is the partial derivative of y with respect to xi).
2/3- Using individual values of half-life obtained from a maximum a posteriori probability (MAP) Bayesian fitting procedure two population estimates could be provided depending upon the assumption of a normal or a log-normal distribution for elimination half-life.
4- Finally, the population analysis was re-run with t1/2 z as a primary parameter with a log normal distribution.
The fact that population estimates of t1/2 z from the 4 methods were not similar (Fig. 1) suggests that assumption of a particular distribution (e.g. normal) for primary parameters (e.g. CL, V, etc.), leads to a distribution of secondary parameters (such as t1/2) that is
unknown and cannot reasonably be described by a log-normal distribution. Simulations, may help to define conditions where such an approximation is reasonable.
Reference: PAGE 6 (1997) Abstr 667 [www.page-meeting.org/?abstract=667]
Poster: poster