Lindauer, A (1), Elshoff, J-P (2), Wilkins, J (1), Olsson Gisleskog, P (1)
(1) SGS Exprimo NV, Mechelen, Belgium, (2) UCB, Monheim, Germany
Objectives: To quantify the dose-exposure-response relationship of rotigotine in adult patients suffering from restless legs syndrome (RLS) and to assess the expected efficacy response to daily administration of rotigotine transdermal patches at different dose levels via simulations.
Methods: Pharmacokinetic (PK) and efficacy data from 3 large placebo-controlled clinical trials in adult RLS patients were used. The dataset included all verum patients that had at least 1 PK sample and all patients receiving placebo (N=709). Data from the remaining patients (N=582) were retained for external validation of the model. A simple PK model was developed to describe the average concentration at steady state (Cav,ss).
Subsequently, a two-part mixed-effects model for semi-continuous data [1] was employed, linking Cav,ss to International Restless Legs Syndrome Rating Scale (IRLS) score (range: 0; no symptoms – 40; most severe). IRLS scores >0 were treated as continuous data applying a longitudinal placebo sub-model and an Emax function for the concentration-response relationship; for IRLS scores equal to zero the likelihood of such observations was modeled with a logistic regression function. The F_FLAG functionality in NONMEM 7.2 was used for the simultaneous modeling of categorical and continuous data. The response at different dose levels was then simulated using Simulo 6.2 [2].
Results: An allometric relationship between body weight and apparent clearance with a fixed exponent of 0.75 significantly improved the fit of the PK model.
The two-part model successfully accounted for the disproportionately large number of IRLS scores of 0 in the dataset. A visual predictive check on the validation dataset demonstrated close correspondence between observed and simulated IRLS scores.
The likelihood of IRLS scores of 0, increased with time, up to a plateau of 13.5%, and with concentration. EC50 of 0.218 ng/mL was well-estimated (90%CI: 0.154-0.225) and corresponded to a dose of about 1.5 mg/24h in a typical adult patient.
Conclusion: A clear exposure-response relationship between rotigotine Cav,ss and the IRLS score could be established. The two-part mixed-effect model previously described by Olsen et al.[1] was successfully implemented in NONMEM and proved useful in modeling data with observations at the boundary of the measurement scale. The concentration-IRLS model provides a framework to simulate the expected response to rotigotine administration in other populations in order to aid in designing future clinical studies.
This study was sponsored by UCB.
References:
[1] Olsen M, Schafer J. A two-part random-effects model for semicontinuous longitudinal data. J Am Stat Assoc (2001) 96:730–45.
[2] http://www.simulo.eu
Reference: PAGE 24 () Abstr 3406 [www.page-meeting.org/?abstract=3406]
Poster: Drug/Disease modeling - CNS