Critical Thresholds of Intracranial Pressure-Derived Continuous Cerebrovascular Reactivity Indices for Outcome Prediction in Noncraniectomized Patients with Traumatic Brain Injury.

Abstract

The aim of the study was to compare intracranial pressure (ICP)-derived cerebrovascular reactivity indices in their ability to predict six-month outcome, and to determine/compare critical thresholds related to outcome for each index in adult noncraniectomized traumatic brain injury (TBI). Using a retrospective cohort of nondecompressive craniectomy (non-DC) patients with TBI, we performed univariate and multi-variate binary logistic regression outcome analysis of: pressure reactivity index (PRx), pulse amplitude index (PAx), and a newly described index (RAC) calculated as the regression coefficient between ICP waveform amplitude and cerebral perfusion pressure (CPP). Finally, we performed sequential chi-square threshold analysis for each index as it related to six-month binary outcomes. Outcome was assessed via dichotomized Glasgow Outcome Scores (GOS): (A) favorable (GOS 4 or 5) versus unfavorable (GOS 3 or less), (B) alive versus dead. There were 358 non-DC patients with TBI included in all aspects of the analysis. In an analysis of the entire recording period for all patients using univariate binary logistic regression, the areas under the curves (AUCs) for favorable versus unfavorable outcome were: PRx (0.573, p < 0.0001), PAx (0.606, p < 0.0001), and RAC (0.655, p < 0.0001). Similarly, the AUCs for alive versus dead outcome were: PRx (0.651, p < 0.0001), PAx (0.705, p < 0.0001), and RAC (0.722, p < 0.0001). RAC displayed superior AUC statistics compared with PRx and PAx, using both univariate and multi-variate regression. RAC displayed more stable critical thresholds related to six-month outcomes. Thresholds for both favorable versus unfavorable and alive versus dead outcomes for PRx, PAx, and RAC across the entire recording period were: +0.35 and +0.35, 0 and +0.25, -0.10 and -0.05, respectively. In non-DC patients with TBI, RAC appears to be superior to PRx and PAx in six-month outcome prediction, using both univariate and multi-variate logistic regression. Further, RAC displayed more stable critical thresholds associated with binary outcomes at six months. Further analysis of RAC in TBI is required.