Chao Chen, Lia Liefaard
GlaxoSmithKline R&D
Objectives: The objective of the work described here was to quantify the potential for reducing the variability in exposure by personalising dose for a drug with moderate pharmacokinetic (PK) variability and narrow therapeutic window. Exposure (AUC) of the drug increases greater than proportionally to dose. At the same dose, the AUC is higher in women than in men.
Methods: A population PK model, including both between-subject and within-subject between-occasion variability, was developed using phase I data. Individual AUC in a large trial with a target AUC were then simulated. In the simulated trial, each subject received a test dose which was chosen to generate the target AUC for a typical person. A PK profile following this dose was obtained. Using this PK profile and the model, a personalised dose for the target AUC was derived with a pre-defined set of formulation strengths. A therapeutic AUC range was defined as +/- 20% of the target. The proportion of the people whose mean AUC was within this range was calculated for the test dose and for the personalised dose. Dose personalisation would be declared effective if the mean AUC of at least 30% more people were within this range.
Results: The model adequately described both the sex effect and the non-linear dose effect on exposure. The variability for apparent clearance was 23.8% between subjects or 14.4% between occasions. Because the AUC following the test dose was subject to between-occasion variability, this AUC could mis-inform the dose change. The simulations showed that the dose change brought the AUC, from outside the therapeutic range after the test dose, into the range for 82% of people. However, it also brought the AUC outside of this range for 20% of the people whose AUC after the test dose was already in the range. The proportion of people whose AUC was within the range increased from 62.3% with the test dose to 80.7% with personalised dose, representing a 18.4% gain by dose personalisation which was below the pre-defined success criteria of 30%.
Conclusions: Findings from this exercise provided the basis for ruling out single-profile based dose personalisation for this drug. The same simulation frame work can be applied to other situations with a defined set of conditions, such as the therapeutic exposure range, dose adjustment increments, number of profiles to inform dose change, and the success criteria.
Reference: PAGE 21 () Abstr 2342 [www.page-meeting.org/?abstract=2342]
Poster: Other Modelling Applications