Population PK Model for Pooled Data of Different Oral Diclofenac Formulations
Christian Bartels, Marianna Armogida, Bengt Hamren
Novartis Pharama AG
Objectives: Diclofenac is available in different oral formulations, which are known to have considerable interindividual and intraindividual variation in their plasma PK profiles. To provide a common framework for the description of existing oral formulations and their variability, a population PK model is developed based on data from multiple studies in healthy subjects.
Methods: The population PK analysis was performed using the non-linear mixed effects modeling software NONMEM(R) version VI 2.0. The data set was pooled from PK studies of six formulations, including two immediate release formulations with a total of 114 individuals, a mixed release formulation with 21 individuals, a slow release formulation with 12 individuals and an enteric coated form with 21 individuals. A total of 3399 plasma samples were used. The data includes single and multiple dose data. The absorption profiles for some of the formulations had characteristic delays and inter-occasion variability.
Results: A two compartment model is used. The absorption is modeled with two first order absorption compartments using lag times, coupled with two sequential first order processes. The two absorption compartments are characterized by a fast and slow absorption rate, respectively. Differences between formulations are described with covariates on absorption rates, relative bioavailabilities, residual errors, fraction that is rapidly or slowly absorbed, and lag times. Inter-individual variability is included on the clearance; inter-occasion variability is included on lag time and absorption rates.
Conclusions: Significant differences of the formulations, pronounced inter-individual and inter-occasion variability of some of the formulations pose challenges. The comparatively simple model provides a good description of the different formulations. The formulations are characterized by absorption rates ranging from 5 h-1 to 0.06 h-1; lag times range between 8 min and about 1 h. Possible improvements of the model are discussed.