Impact of Dosing Regimens on Dropout Across Pregabalin Trials in the Treatment of Generalized Anxiety Disorder: Model Refinements and External Validation
Bojan Lalovic, Matt Hutmacher, Bill Frame, Raymond Miller
Pfizer Inc.
Background: We have previously analyzed the relationship of dizziness and somnolence as major determinants of study dropout in the treatment of generalized anxiety disorder (GAD) with pregabalin1. The relationship of dose to adverse-event was independently modeled using the two-part (incidence and severity) AE model2. The examination of dose-adverse-event-dropout relationship afforded a model-based strategy to mitigate incidence and severity of adverse events and reduce dropout, which was previously based only on subjective and empirical clinical judgment.
Objectives: To examine potential improvements to the initial modeling analysis by introducing several refinements to the adverse event and dropout sub-models and examine their predictive performance and impact of external validation.
Methods: Adverse-event incidence was modeled as a time-to event process, allowing incorporation of daily dosing (titrations) as a time-varying covariate. Conditional severity of adverse events was described as an ordered categorical variable with proportional odds accounting for both the time-course of effect and correlation between adjacent observations3. Parametric discrete-time, hazard models were fitted using dizziness severity as a time-varying covariate. The model-based predictions of dropout were evaluated against the nonparametric (Kaplan Meier) estimates (predictive check) and against data from an independent trial.
Results: Consideration of initial adverse event incidence within an individual as a time to event data represented a principled method in handling time-varying information across the both adverse event and dropout sub-models. Model predictions illustrate the benefit of a gradual titration on dropout as a result of lower initial dizziness incidence and severity. Predictive performance of the adverse-event dropout model was evaluated and the approach considered adequate, with the assessment based in part on an independent GAD trial providing an external validation.
Conclusions: Dropout represents an important clinical trial endpoint, which can be analyzed using time to event models that readily incorporate daily dosing, or other time varying information (covariates). Prospective simulations with the current model highlight the utility of this modeling approach in examining the impact of untested titrations schemes on dropout for future GAD trials.
References:
[1] Modeling Dropout from Longitudinal Adverse Event Data: Selecting Optimal Titration Regimens. Lalovic B, Hutmacher MM, Frame B, Ito K, Miller R. Poster Presentation, PAGE 2007.
[2] A two-part mixture model for longitudinal adverse event severity data. Kowalski KG, McFadyen L, Hutmacher MM, Frame B and Miller R. J Pharmacokinet Pharmacodyn. 2003 Oct ;30 (5):315-36.
[3] Exposure-response analysis for spontaneously reported dizziness in pregabalin-treated patient with generalized anxiety disorder. Ito K, Hutmacher MM, Liu J, Qui R, Frame B, Miller R. Clin Pharmacol Ther. 2008 Jul;84(1):127-35. Epub 2008 Feb 6.