NATIONAL VETERINARY SCHOOL TOULOUSE Round table Principle of
NATIONAL VETERINARY SCHOOL TOULOUSE Round table: Principle of dosage selection for veterinary pharmaceutical products Bayesian approach in dosage selection EAVPT Torino September 2006 D. Concordet National Veterinary School Toulouse, France
Bayesian forecasting methods = Therapeutic drug monitoring
Why a bayesian forecasting method ? Consequence of PK Variability : Toxicity Efficacy the same dose gives different exposures Exposure
Why a bayesian forecasting method ? Consequence of PK Variability : We need to anticipate the "level" of exposure Toxicity Efficacy the same dose gives different exposures Exposure
How to predict exposure ? Exposure
How to predict exposure ? Exposure POPULATION PK Cannot be predicted with covariates Need further information Covariate : e. g. Age
The bayesian approach Same dose animals with the same age a priori information Probably a high exposure A blood sample at this time
The bayesian approach Same dose animals with the same age a priori information Probably a small exposure A blood sample at this time
The bayesian approach Same dose animals with the same age a priori information Exposure ? A blood sample at this time
Concentration Why population information is needed ? Exposure ? A blood sample at this time Time
The bayesian approach Same dose animals with the same age A blood sample at this time
The bayesian approach Frequency Same dose animals with the same age Exposure A blood sample at this time
Frequency The a posteriori distribution Distribution of exposure for animals that received the same dose have the same age have the same drug concentation at the sampling time Maximum a posteriori (MAP) = Bayesian estimate = most common exposure Exposure
The a priori information Frequency Same dose animals with the same age Exposure A blood sample at this time
The a priori information Frequency Same dose animals with the same age Exposure A blood sample at this time
The a priori information Frequency Same dose animals with the same age Exposure A blood sample at this time
Exposure How to predict exposure ? POP. PK Covariate : e. g. Age
Exposure How to predict exposure ? POP. PK + 1 concentration POP. PK Covariate : e. g. Age
Exposure How to predict exposure ? POP. PK + 2 concentrations POP. PK + 1 concentration POP. PK Covariate : e. g. Age
Problem of highly variable drugs ? Concentration 1 st Administration: fixed dose A blood sample at this time Time
Problem of highly variable drugs ? 2 nd Administration: same animal, same dose as 1 st Concentration Large inter-occasion variability A blood sample at this time Time
How does it work ? A population model jth concentration measured on the ith animal jth sample time of the ith animal
How does it work ? A set of concentrations on THE animal : (t 1, Z 1) , (t 2, Z 2) , … Maximize the a posteriori likelihood Minimize
To summarize Bayesian forecasting can be useful for: pets touchy drugs (narrow therapeutic index) It requires: results of a pop PK study some concentrations on the animal a recent computer Can’t work for large inter-occasion variability
- Slides: 24