Page 11 - 70_03

VOL. 70 (3), 663-679, 2004 PHARMACOKINETICS AND INDIVIDUALIZED DRUG THERAPY

EXAMPLES OF BAYESIAN DOSE PREDICTIONS

Today, there are computerized statistical techniques available
to help us to perform the correct interpretations of measured plasma
concentration values, although they are not used very much. The
most reliable one is based on the statistical theory called Baye-
sian techniques after the English statistician Bayes. As shown in
equation 1 the Bayesian analysis (Maximum likelihood estimates)
takes into consideration, apart from the measured plasma concen-
tration, what is known about the pharmacokinetic parameters in the
population under investigation and its statistical variability.

Bayes = Sum [(Cobs - Cpred)2/ S.D. ana2 + (Ppop – Ppred)2/ S.D.
pop2 + 1n (S.D. pred)2]

Cobs= Measured plasma concentration for a specific individual
Cpred= Predicted plasma concentration for a specific individual
Ppop= Pharmacokinetics population paremeter value
Ppred= Predicted pharmacokinetics parameter value for the individual
S.D. pop= Standard deviation of the population value
S.D ana= Standard deviation of the chemical analysis

By this technique, first introduced by Sheiner et al. (7), you can
calculate the pharmacokinetic parameters in the individual patient
with only one plasma concentration value available. Once you have
calculated the pharmacokinetic parameter of the individual patient
you can easily calculate the dose and dosing interval the patient
should use in order to get desired plasma concentrations. I will give
you some examples and the first one is shown in Fig. 5.

If this patient behaved as the normal average patient the recom-
mended dosing should give a plasma concentration-time profile as
the upper curve. A Bayesian estimate gave however the lower curve
with trough values lower then the 250 ng/mL needed for the drug to
have an effect in this individual. Obviously this patient needs a hig-
her dose in order to get an effect of the drug.

671