Early signs of brain infarction at CT: observer reliability and outcome after thrombolytic treatment--systematic review.

PURPOSE

To review systematically all reported early computed tomographic (CT) signs in acute ischemic stroke to determine interobserver agreement and the relationship between early CT signs and patient outcome with or without thrombolysis.

MATERIALS AND METHODS

A systematic review of the literature was conducted by using Cochrane Stroke Group methodology to identify studies published between 1990 and 2003 that were performed to assess interobserver agreement about early signs of infarction on CT scans obtained within 6 hours after onset of stroke symptoms and determine the relation of early signs of infarction to clinical outcome, including any interactive effect of thrombolysis. Interobserver agreement was measured with the kappa statistic, sensitivity, and specificity. The relation of early signs to clinical outcome with or without thrombolysis was assessed with calculated odds ratios and 95% confidence intervals.

RESULTS

In 15 studies of interobserver agreement (median of 30 CT scans and six raters), the prevalence of all early infarction signs was 61% +/- 21 (standard deviation). Interobserver agreement (kappa statistics) ranged from 0.14 to 0.78 for any early infarction sign. The mean sensitivity and specificity for detection of early infarction signs with CT were 66% (range, 20%-87%) and 87% (range, 56%-100%), respectively. Experience improved detection, but knowledge of symptoms did not. In 15 studies of early infarction signs and outcome (including seven thrombolysis trials) in 3468 patients, any early infarction sign increased the risk of poor outcome (odds ratio, 3.11; 95% confidence interval: 2.77, 3.49). Two studies that sought interaction between early infarction signs and thrombolysis found no evidence that thrombolysis given in the presence of early infarction signs resulted in worse outcome than that due to early signs alone.

CONCLUSION

Further work is required to determine which signs are most reliably detected, whether scoring systems help to improve detection, and whether any early infarction sign should influence decisions concerning thrombolysis.