Correlation between carotid intraplaque hemorrhage and clinical symptoms: systematic review of observational studies.

BACKGROUND AND PURPOSE We sought to investigate the association between carotid intraplaque hemorrhage (IPH) and ipsilateral symptoms of cerebral ischemia. METHODS A search was performed for clinical observational studies comparing the incidence of IPH between symptomatic and asymptomatic patients. Odds ratios (ORs) for IPH as a factor in the pathogenesis of neurologic events were calculated and combined by a meta-analysis. Interstudy heterogeneity, estimated effects, and methodologic quality of the studies were assessed. RESULTS Thirty-one studies were included for analysis. The reported ORs varied widely. Overall, the incidence of IPH in the symptomatic groups was significantly higher than in the asymptomatic group. However, there was an apparent trend for heterogeneity (P<0.00001) between studies. The random-effects summary estimator of ORs was 2.25 (95% CI, 1.57 to 3.22; P<0.00001). To identify potential sources of heterogeneity, subgroup analyses were performed. The pooled ORs varied greatly by stratification. Major heterogeneity was found among studies with low quality, microscopic methods of examination, significant effects, small sizes, early publication, and unequal severity of carotid stenosis in both groups. Large, recent, macroscopic, or high-quality studies, as well as studies with equal degrees of stenosis, tended to yield insignificant associations. The methods in defining and evaluating hemorrhage were very heterogeneous. Characterizations of the age, size, number, and location of hemorrhages were poorly reported and highly variable. In addition, a lack of control of confounders and selection bias were frequently identified among studies. CONCLUSIONS Statistical inferences have suggested a plausible role in the production of cerebral ischemia; however, reliable interpretation was strongly undermined by poor methodologic quality, substantial heterogeneity, and suspicious publication bias. To preciously estimate the underlying correlation, a well-designed study with uniformity in definition and evaluation for IPH might be warranted.