Purpose: Inter-institutional agreement of imaging measurements is important in shared clinical management and in trials. 2D strain is sensitive to changes in LV function, automated and may be less variable than ejection fraction (EF), which is used widely but has inherent variability. We sought whether strain would have better concordance between different centers and that feedback from a calibration exercise would reduce the variability among institutions. Methods: 108 global longitudinal strain (GLS) measurements, calculated from 1944 segmental strains, were performed blindly by 21 experienced readers from 12 different institutes (5 Europe, 5 Asia, 1 North America and 1 Australia) in 6 cases. Intraclass correlation coefficients (ICCs) were used to determine concordance. All individual measurements were reviewed and some key points were identified to optimize strain measurement. After feedback, strain was remeasured and improvement of agreement sought using coefficient variance (CV) and mean difference (MD) from the reference standard. Results: GLS was -17.9±3.3%, while EF was 60±7%. The ICC in GLS (0.994 [95%CI 0.983, 0.999]) was better than that of 2DEF (0.928 [0.809, 0.988], p<0.001) at baseline. Two main sources of discordance in GLS measurements were the width and location of regions of interest, especially at mitral annulus and apex. After feedback, re-measurement showed the CV (p=0.02) and MD (p=0.03) of segmental strain improved, but comparison of 2nd vs 1st GLS showed no changes of ICC (p=0.85), CV (p=0.80) or MD (p=0.92). Conclusions: Feedback significantly decreased segmental strain variability, but did not improve the concordance in GLS, which had better precision than EF at baseline.
Read full abstract