Mobile echoes on prosthetic valves are not reproducible. Results and clinical implications of a multicentre study.

AIMS To test the hypothesis that inter-observer variability accounts for the wide variation in reported prevalences of fibrin strands on prosthetic heart valves and to develop criteria for their identification and reporting. METHODS AND RESULTS A videotape with 30 sequences of prosthetic heart valves imaged by transoesophageal echocardiography and showing abnormalities such as strands, microbubbles, and spontaneous echocardiographic contrast, was assessed in 13 European and three American centres. There were three duplicated examples, unbeknown to the observers. Definitions and reported prevalence rates of the abnormalities were analysed, and inter- and intra-observer agreement estimated with the kappa statistic. Mobile echoes were identified in 40 to 80% of the sequences on the tape. The reported prevalence of mobile echoes correlated with the time spent reporting the tape. There was moderate inter-observer agreement for the identification of any mobile echoes (kappa = 0.38), but no agreement for their labelling (kappa = 0.22), in spite of similar definitions. Intra-observer reproducibility was good (agreement in 76% of the reduplicated sequences). CONCLUSIONS The true prevalence and potential significance of mobile echoes on prosthetic heart valves cannot be assessed unless inter-observer consensus on echocardiographic criteria for identifying such echoes is reached.