Identification of coronary plaque sub-types using virtual histology intravascular ultrasound is affected by inter-observer variability and differences in plaque definitions.

BACKGROUND Recent studies show that virtual histology intravascular ultrasound (VH-IVUS) can identify plaques at high risk of rupture, such as thin-capped fibroatheromata, raising the possibility of immediate targeted intervention. However, plaque classification entails border recognition and subjective assessment of plaque architecture, introducing inter-observer variability without confirmation by core-labs. Furthermore, the accuracy of local versus core-laboratory VH-IVUS plaque classification and effects of different plaque definitions have not been examined. METHODS AND RESULTS Local observers classified 100 VH-IVUS-defined coronary plaques to determine single center inter-observer variability; multi-center variability was determined by comparison with VH-IVUS core-laboratory analysis, and compared with gray-scale IVUS. Frequency of plaque types using different published plaque definitions also was determined. Single-center VH-IVUS inter-observer agreement was strong (kappa=0.86), but lower for thin-capped fibroatheromatas (k=0.59) because of observer judgments on presence and location of confluent necrotic core. Multi-center inter-observer agreement for plaque classification was lower again (k=0.71), particularly for thin-capped fibroatheromatas (k=0.56). Different plaque definitions further reduced VH-IVUS-defined thin-capped fibroatheromata numbers by 44%. The diagnostic accuracy of gray-scale IVUS to identify thin-capped fibroatheromata was poor for both observers (21 and 29% correct), with low inter-observer agreement (k=0.14). CONCLUSIONS VH-IVUS plaque classification, and particularly VH-IVUS-defined thin-capped fibroatheromata identification, varies significantly between local observers, and particularly in comparison with core-laboratory analysis. Differences in VH-IVUS plaque definitions introduce further variability between studies. These factors reduce the use of VH-IVUS plaque classification to guide intervention in a "live" clinical setting, and also affect comparison of diagnostic accuracy and natural history of plaques between studies.