Variability in echocardiography and MRI for detection of cancer therapy cardiotoxicity

OBJECTIVES: To compare variability of echocardiographic and cardiovascular magnetic resonance (CMR) measured left ventricular (LV) function parameters and their relationship to cancer therapeutics-related cardiac dysfunction (CTRCD). METHODS: We prospectively recruited 60 participants (age: 49.8±11.6 years), 30 women with human epidermal growth factor receptor 2-positive breast cancer (15 with CTRCD and 15 without CTRCD) and 30 healthy volunteers. Patients were treated with anthracyclines and trastuzumab. Participants underwent three serial CMR (1.5T) and echocardiography studies at ~3-month intervals. Cine-CMR for LV ejection fraction (LVEF), myocardial tagging for global longitudinal strain (GLS) and global circumferential strain (GCS), two-dimensional (2D) echocardiography for strain and LVEF and three-dimensional (3D) echocardiography for LVEF measurements were obtained. Temporal, interobserver and intraobserver variability were calculated as the coefficient of variation and as the SE of the measurement (SEM). Minimal detected difference (MDD) was defined as 2xSEM. RESULTS: Patients with CTRCD demonstrated larger mean temporal changes in all parameters compared with those without: 2D-LVEF: 4.6% versus 2.8%; 3D-LVEF: 5.2% vs 2.3%; CMR-LVEF: 6.6% versus 2.7%; 2D-GLS: 1.9% versus 0.7%, 2D-GCS: 2.5% versus 2.2%; CMR-GCS: 2.7% versus 1.6%; and CMR-GLS: 2.1% versus 1.4%, with overlap in 95% CI for 2D-LVEF, 2D-GCS, CMR-GLS and CMR-GCS. The respective mean temporal variability/MDD in healthy volunteers were 3.3%/6.5%, 1.8%/3.7%, 2.2%/4.4%, 0.8%/1.5%, 1.9%/3.7%, 1.8%/3.6% and 1.4%/2.8%. Although the mean temporal variability in healthy volunteers was lower than the mean temporal changes in CTRCD, at the individual level, 2D-GLS, 3D-LVEF and CMR-LVEF had the least overlap. 2D-GLS and CMR-LVEF had the lowest interobserver/intraobserver variabilities. CONCLUSION: Temporal changes in 3D-LVEF, 2D-GLS and CMR LVEF in patients with CTRCD had the least overlap with the variability in healthy volunteers; however, 2D-GLS appears to be the most suitable for clinical application in individual patients.