Dose Adaptation Based on Pharmacometric Models

Detta är en avhandling från Uppsala : Acta Universitatis Upsaliensis

Sammanfattning: Many drugs exhibit major variability in both pharmacokinetic (PK) and pharmacodynamic (PD) parameters that prevents the use of the same dose for all patients. Variability can occur both between patients (IIV) as well as within patients over the course of time (IOV). In a drug with narrow therapeutic range and substantial IIV, dose selection may require individual adaptation. Adaptation can either be made before (a priori) or after (a posteriori) first drug administration. The former implies basing the dose on prior information known to be influential, such as kidney function indicators, weight or concomitant medication, whereas a posteriori dose adaptations are based on post-treatment observations. Often individualization cannot be based on the clinical outcome itself. In such cases, drug concentrations or biomarkers may be valuable for dose individualisation.In this thesis two therapeutic areas where dosing is critical have been investigated regarding the possibilities of a priori and a posteriori dose adaptation; anticancer treatment where myelosuppression is dose limiting, and tacrolimus used for immunosuppression in paediatric transplantation. In tacrolimus previously published models were found to be of little value for dose adaptation in the early critical days post-transplantation. New PK models were developed and used to suggest new dosing regimens tailored for the paediatric population, recognizing the changing pharmacokinetics in the early time post-transplantation.For several anticancer drugs covariates were identified that partly explained IIV in myelosuppression. IOV were found to be lower than IIV which implies that individual dose adaptations a posteriori can be valuable. Dose adaptation, using Bayesian principles in order to simultaneously minimise the risk of severe toxicity or subtherapeutic levels, was evaluated using simulations. Type and amount of data needed, as well as variability parameters influential on the outcome, were evaluated. Results show drug concentrations being of little value, if neutrophil counts are available.The models discussed in this thesis have been implemented in MS Excel macros for Bayesian forecasting, to allow widespread distribution to clinical settings without necessitating access to specific statistical software.