PAGE. Abstracts of the Annual Meeting of the Population Approach Group in Europe.
PAGE 25 (2016) Abstr 5905 [www.page-meeting.org/?abstract=5905]
Click to open
Oral: Clinical Applications
Rik Schoemaker(1), Satyaprakash Nayak(2), William S. Denney(2), E. Niclas Jonsson(3), Mats O. Karlsson(3), Lutz O. Harnisch (2), on behalf of the VISTA-ICH Collaborators
(1) Occams, (2) Pfizer Inc., (3) Pharmetheus AB
Objective: To develop a modeling and simulation framework for longitudinally observed mRS and NIHSS scores in patients with intracerebral hemorrhage (ICH) based on the placebo subject-level data from the VISTA-ICH  database.
Methods: mRS and NIHSS were modeled using categorical and continuous data models, respectively, where an underlying Emax time-profile allowed for an individual improvement or deterioration over time. The two endpoint scales were combined into a single model to estimate the correlation between model components and improve the model power. Covariate effects were implemented for the models by estimating the effect of ICH hematoma volume on mRS and NIHSS at baseline, and by estimating the effect of change in ICH hematoma volume (a future treatment target) on the fraction of patients improving. Covariate effects were simulated to quantify impact on outcomes for both scales at day 90 and to quantify the effect of reducing the hematoma volume increase on those outcomes. Trial simulations allowed power estimates for various longitudinal models, different sampling schedules, and different effect sizes using the Parametric Power Estimation approach  and contrasting it to a traditional logistic regression approach.
Results: Both the mRS-alone model and the combined mRS/NIHSS model provided an adequate description of the evolution of mRS scores over time. The combined model established two populations, improvers or deterioraters in the VISTA-ICH placebo data. The parameters for the base model component, Emax and fraction improvers were found to be highly correlated across these two scales. Simulations showed that for VISTA placebo data, small hematoma baseline volumes were associated with a favorable outcome, but substantial reductions in hematoma volume increase would be required to increase the probability of a favorable outcome.
Conclusion: The model predicted that a large reduction in hematoma volume increase would likely have a substantial impact on mRS outcome scores, providing us with an estimate for a desired target effect size. Since mRS and NIHSS profiles are highly correlated NIHSS used in conjunction with mRS will improve the prediction precision of the clinical outcome and therefore lead to much smaller trial sizes. The effect of a combined longitudinal endpoint could be quantified through power calculations to guide the design of a future trial.