Prediction of coma and anisocoria based on computerized tomography findings in patients with supratentorial intracerebral hemorrhage

Abstract

Objectives: Coma and anisocoria are the two common signs of a crucial state of neurological dysfunction. The ability to forecast the occurrence of these conditions would help clinicians make clinical risk assessments and decisions.

Patients and methods: From October 2006 to September 2008, 118 patients with supratentorial intracerebral hemorrhage (SICH) were enrolled in this retrospective investigation. Patients were distributed into 3 groups according to occurrence of the signs of coma and/or anisocoria in the observation unit during a 30-day period. Group 1 included 52 patients who had normal or impaired consciousness, group 2 included 27 patients who had coma with no anisocoria and group 3 consisted of 39 patients who had coma with anisocoria. The clinical characteristics and parameters on computerized tomography (CT) findings were compared using univariate analysis to determine the factors that were related to the level of consciousness. Logistic regression models established the predictive equations for coma and anisocoria.

Results: Univariate analysis revealed that hematoma volume, the score of intraventricular hemorrhage (IVH score) and the amplitude of midline shift were the factors related to coma and anisocoria. Mean hematoma volume was 24.0 ± 13.0 ml, 53.6 ± 12.6 ml and 80.5 ± 24.6 ml, the mean amplitudes of midline shift were 1.3 ± 2.0 mm, 5.9 ± 4.9 mm and 10.1 ± 5.5 mm, and the mean IVH score was 0.8 ± 1.3, 3.3 ± 3.3 and 5.9 ± 3.4 in groups 1, 2 and 3, respectively. Multivariate analysis showed that hematoma volume and IVH score were independent prognostic factors for coma and anisocoria. The predictive equations for coma and anisocoria were LogitP = 0.279X(HV) + 0.521X(IVH)-18.164 and LogitP = 0.125X(HV)+0.326X(IVH)-6.864, respectively.

Conclusions: Hematoma volume and IVH score were the independent prognostic factors for coma and anisocoria. Logistic regression models established the fitted predictive equations, which could help clinicians make clinical risk assessments and decisions.