E. Hanze, M. Kågedal
AstraZeneca R&D Södertälje, Sweden
Objectives: In neuropathic pain, the primary endpoint for detecting the presence of dose-response has typically been the end of treatment (EOT) response in the Numeric Rating Scale (NRS) 1. The objective of this simulation study was to evaluate the precision in the dose-exposure-response relationship when using change from baseline at end of treatment as compared to when longitudinal data, i.e. all individual data from all time points are used. In addition the impact of introducing an effect delay in the longitudinal model was evaluated.
Methods: For the longitudinal data analysis a simulation model was developed based on placebo-response (NRS) data from a phaseIIa study in patients with Painful Diabetic Neuralgia. For both the longitudinal and EOT model an assumed dose-exposure-response relationship was applied with a sigmoid Emax model driving the response. The estimation models were simplified compared to the simulation model and was not fully identifiable. A no effect model, a linear model, an Emax and sigmoid Emax model was fitted to data and model selection was based on likelihood ratio test. Different study-designs with varying number of dose arms and number of patients were evaluated. For the longitudinal analysis, a direct response model was compared to a model with a time-delay in response reaching maximum effect half-way in the study. The precision in dose-exposure-response was evaluated by estimating the percentage of studies with an estimated delta NRS within 0.7-1.3 of given a true effect (simulated effect) of delta NRS equal to 1.
Results: In a simulated study with 4 dose arms and 54 patients in each dose arm the precision in dose-exposure-response was increased from 33% to 97% when using longitudinal data as compared to EOT. Introducing a time-delay in the efficacy reduced the precision from 97% to 83%.
Conclusions: Including all time point of efficacy assessments increased the precision in dose-exposure-response to a large extent. When performing a simulation based evaluation of a longitudinal analysis it is important to identify considerations that can lead to overestimating the precision.
References:
[1] Bornkamp B. et al. Innovative approaches for designing and analyzing adaptive dose-ranging trials. Journal of Biopharmaceutical Statistics. 17: 965-995, 2007.
Reference: PAGE 21 (2012) Abstr 2371 [www.page-meeting.org/?abstract=2371]
Poster: Study Design