Fifteen studies assessing inter-observer agreement were included in the review. These studies assessed a total of 1,300 scans, with 707 readers.
Fifteen studies comparing CT signs and outcome were included in the review (3,481 patients). Five studies were secondary analyses of 7 published RCTs and 10 (n=947) were observational studies.
Assessment of inter-observer agreement.
Eight of the 15 studies gave explicit definitions of early signs of infarction. In 9 studies the readers were blinded to clinical factors.
Studies included a median of 30 scans and 6 observers. The prevalence of all early infarction signs (15 studies) was 61% +/- 21 (standard deviation). Inter-observer agreement (kappa statistic) ranged from 0.14 to 0.78 (6 studies). The mean sensitivity and specificity values for the detection of early infarction signs with CT were 66% (range: 20 to 87) and 87% (range: 56 to 100), respectively. The data were insufficient to allow a reliable determination of the effects of observer experience, knowledge of symptoms, or prevalence of early signs on inter-observer agreement; data from individual included studies indicated that experience improved detection, but knowledge of symptoms did not.
Comparison of CT signs and patient outcome.
The presence of any early infarction sign at CT increased the risk of poor outcome (death or dependency) (OR 3.11, 95% CI: 2.77, 3.49). The presence of each of five individual infarction signs also increased the risk of poor outcome in each case (full data reported in the article). Statistically significant between-study heterogeneity was present in all but one of the five groups of studies used to generate pooled ORs. Two studies that examined the relationship between early infarction signs and the effectiveness of thrombolysis found no evidence that thrombolysis resulted in worse outcomes.