A population pharmacokinetic/pharmacodynamic model for duloxetine in diabetic peripheral neuropathy, plus methods for handling missing data.
E. Yuen(1,2), I. Gueorguieva(1), L. Aarons(2)
(1)Eli Lilly & Co. Ltd, Global PK/PD, Windlesham Surrey; (2)Pharmacy Department, University of Manchester
Objectives: Duloxetine is a selective serotonin and norepinephrine reuptake inhibitor (SNRI) indicated for the treatment of diabetic peripheral neuropathic pain (DPNP). The population pharmacokinetics of duloxetine have been previously described [1]. In three phase 3 double-blind placebo-controlled trials, the efficacy of duloxetine was evaluated over a 12 week acute therapy phase. Patients receiving 60mg QD and 60mg BID duloxetine showed significant improvement in pain scores, starting from week one [2,3,4]. The aim of the current analyses was to develop a population PK/PD model in DPNP patients and to compare the different methods of imputing missing data for dropouts.
Methods: A total of 1106 patients (12549 PD observations) were randomised into placebo (N=327), 20mg QD (N=110), 60mg QD (N=335) and 60mg BID (N=334) groups. PK parameters from [1] were used to simulate duloxetine concentrations at steady state, and weekly average painscores (Likert scale, 0 to 10) calculated from 24-hour average scores were used as PD measures. Population Pk/PD analysis was carried out in NONMEM, describing the pain scores as well as a proportional odds model describing the probability of achieving clinically meaningful pain relief. Missing data for dropouts were imputed using last observation carried forward (LOCF), multiple imputation, and pattern mixture model methods. The PK/PD model was applied to these various enriched datasets and parameters obtained compared to those from non-missing data.
Results: The placebo response was described by an exponential decline model for each pain severity group, whilst drug effect was additive and described by a single Emax model. Across the 3 trials, completion rate was approximately 72%. More patients in the 60mg treatment groups discontinued due to adverse events, whilst more patients in the placebo group discontinued due to lack of efficacy. For the comparison of imputation methods, EC50 was the most sensitive estimate to the different methods. LOCF produced the smallest parameter estimates describing placebo slope. Both the multiple imputation and pattern mixture methods produced parameter estimates similar to the model with non-missing data.
Conclusions: The Likert scale pain scores were well described by the population PK/PD model. Overall, the different methods of handling dropouts produced comparable PD parameters to the model with non-missing data, mostly within 20% of the estimates using non-missing data.
References:
[1] Population pharmacokinetics of orally administered duloxetine in patients. Lobo E, Quinlan T, O'Brien et al. 2009, Clin Pharmacokinet, 48(3):189.
[2] A randomized controlled trial of duloxetine in diabetic peripheral neuropathic pain. Wernicke J, Pritchett Y, D'Souza D et al. 2006, Neurology, 67:1411.
[3] A double-blind, randomized multicenter trial comparing duloxetine with placebo in the management of diabetic peripheral neuropathic pain. Raskin J, Pritchett Y, Wang F et al. 2005, Pain Med, 6:346.
[4] Duloxetine versus placebo in patients with painful diabetic neuropathy. Goldstein D, Lu Y, Detke M et al. 2005, Pain, 116:109.