INTRODUCTION
Outcome of traumatic brain injury (TBI) patients depends both on the primary and the secondary brain injury. Perfusion disturbances as well as blood-brain barrier permeability (BBBP) alterations constitute the pathophysiological mechanisms of secondary TBI. Currently established prognostic models in TBI include non-contrast computed tomography (NCCT) which is insensitive to early perfusion alterations associated with the secondary TBI. Perfusion CT (PCT) on the other hand offers insight into early perfusion abnormalities and allows the evaluation of BBBP as well. We hypothesized that PCT derived parameters of cerebral perfusion and BBBP differ between the outcome groups and that adding CT perfusion and permeability data to the established outcome predictors improves the performance of the prognostic model.
METHODS
A prospective cohort study of 50 adult patients with moderate and severe TBI was performed. PCT and routine control NCCT were performed 12-24 hours after admission and regions of interest were drawn on parametric maps in 6 major vascular territories. Glasgow Outcome Scale (GOS) was used 6 months later to categorize patients' functional outcomes to favorable (GOS>3) or unfavorable (GOS⡤3). Core prognostic model consisted of age, motor GCS score (GCSm), pupillary reactivity and NCCT Score. We added perfusion and permeability data as predictors and compared updated models to the core model using cross-validated AUCs.
RESULTS
Perfusion and BBBP parameters differed significantly between the outcome groups. The prognostic model containing both minimal cerebral blood volume (CBVmin) and the average fraction of cerebral extravascular-extracellular volume (VEmean) in addition to age, GCSm, pupillary reaction, and NCCT score, performed better than the core model (cv-AUC (95% CI) 0.75 (0.51, 0.84) vs. 0.6 (0.37, 0.74)).
CONCLUSIONS
PCT derived perfusion and permeability parameters differed significantly between the outcome groups. Adding CT perfusion and permeability data to the established outcome predictors may improve the performance of the prognostic model.
|