Contrast extravasation on computed tomography angiography predicts clinical outcome in primary intracerebral hemorrhage: a prospective study of 139 cases.

BACKGROUND AND PURPOSE

Several retrospective studies suggested that contrast extravasation on CT angiography predicts hematoma expansion, poor outcome, and mortality in primary intracerebral hemorrhage. We aimed to determine the predictive value of contrast extravasation on multidetector CT angiography for clinical outcome in a prospective study.

METHODS

In 160 consecutive patients with spontaneous intracerebral hemorrhage admitted within 6 hours of symptom onset, noncontrast CT and multidetector CT angiography were performed on admission. A follow-up noncontrast CT was done at 24 hours. Multidetector CT angiography images were analyzed to identify the presence of contrast extravasation. Clinical outcome was assessed by modified Rankin Scale on discharge and at 90 days.

RESULTS

A total of 139 patients with primary intracerebral hemorrhage were included in the final analysis. Contrast extravasation occurred in 30 (21.6%) patients. The presence of contrast extravasation was associated with increased hematoma expansion (P<0.0001), in-hospital mortality (P=0.008), prolonged hospital stay (P=0.006), poor outcome on discharge (P=0.025), increased 3-month mortality (P=0.009), and poor clinical outcome (P<0.0001). In multivariate analysis, contrast extravasation was a promising independent predictor (OR, 10.5; 95% CI, 3.2-34.7; P<0.0001) for 90-day poor clinical outcome followed by the presence of intraventricular hemorrhage (OR, 3.4; 95% CI, 1.5-7.7; P=0.003) and initial hematoma volume (OR, 1.0; 95% CI, 1.0-1.1; P=0.013).

CONCLUSIONS

The presence of contrast extravasation on multidetector CT angiography in patients with hyperacute-stage intracerebral hemorrhage is an independent and strong factor associated with poor outcome. Any patient with intracerebral hemorrhage with such sign on multidetector CT angiography should be monitored intensely and treated accordingly.