Modeling and simulation of the modified Rankin Scale and National Institutes of Health Stroke Scale neurological endpoints in intracerebral hemorrhage

Abstract

Intracerebral hemorrhage (ICH) is a form of stroke characterized by uncontrolled bleeding into the parenchyma of the brain. There is no approved therapy for ICH and it is associated with very poor neurological outcomes with around half of subjects dying within 1 month and most subjects showing complete or partial disability. A key challenge is to identify subjects who could benefit from intervention using characteristics such as baseline hemorrhage volume and the increase in hemorrhage volume in the first few hours, which have been correlated with final outcomes in ICH. Combined longitudinal models were developed to describe stroke scales using categorical data (Modified Rankin Scale, mRS), continuous bounded data (National Institutes of Health Stroke Scale, NIHSS), and time to death. Covariate effects for baseline hematoma volume and maximum increase in hematoma volume were incorporated to assess the improvement in outcome when hematoma volume increase would be reduced by a potential treatment. The combined model provided an adequate description of stroke scales, with patients split into a Non-survival and a High-survival sub-population, and dropout due to death was well described by a constant hazard survival model. Models were compared indicating that the combined mRS/NIHSS model provided the most information, followed by the NIHSS-only model, and the mRS-only model, and finally the traditional statistical analysis on dichotomized response at 90 days. Simulations showed that substantial reductions in hematoma volume increase were required to increase the probability of a favorable outcome.