D. Breilh, C. Pobel, L. Lagagnier, P. Maire, M.C. Saux
Laboratoire de Pharmacocinétique et de Pharmacie Clinique C.H.U. de Bordeaux- Pharmacie Haut-Levêque - Avenue de Magellan 33604 Pessac Cedex
The aim of this study is to constitute two Bayesian estimators. For adaptative control of Vancomycin from the same population with two different methodologies and to value the performances between these two methods using a reference population distributed with USCPACK-PC software which has served to the validation. Plasma concentrations are measured by FPIA (TDx Abbott). In totaly, 50 subjects were inclued in this study : 32 for constitution of the population and 18 for its validation. The open bicompartimental model is used and two statistic approaches of vectors of population parameters are obtained : one parametric method with the NONMEM software population and one no parametric method with NPEM software population. To value the performances between these two estimators, a retrospective comparison between observed and predictive plasma concentrations has been realised with calculated and compared bias, precision and correlation coefficient. The results are as follows :
- there is no bias in these two approaches
- the correlation between observed and predictive values is good (r > 0.9)
- precisions are similar with those of literature; RMSE@ 3 to 4 µg/ml without significant difference in the two approaches
In conclusion, the Bayesian estimators used to follow adaptive control of Vancomycin are the same and answered to the fixed goal of therapeutical study plan.
Reference: PAGE 3 () Abstr 865 [www.page-meeting.org/?abstract=865]
Poster: poster