(P198) MANUAL VERSUS AUTOMATED LEFT VENTRICULAR LONGITUDINAL STRAIN ANALYSIS TO IDENTIFY PATIENTS WITH TRANSTHYRETIN CARDIAC AMYLOIDOSIS

Background: Transthyretin cardiac amyloidosis (ATTR-CM) is an underrecognized cause of heart failure that is often first suspected through echocardiographic findings. Relative apical sparing of left ventricular (LV) longitudinal strain has been proposed as a potential marker of ATTR-CM but its real-world reliability to predict ATTR-CM using either manual or automated strain analysis methods has not been well established. We aimed to compare the test characteristics of manual versus automated LV longitudinal strain analysis to predict ATTR-CM.

METHODS AND RESULTS: We performed a retrospective cohort study of consecutive patients with suspected ATTR-CM who underwent pyrophosphate scans at two tertiary care hospitals. LV strain analyses were performed using commercially available software (TomTec, Munich, Germany) by a Level III echocardiographer blinded to the result of the PYP scans. Following identification of the representative cardiac cycle in the apical views, LV longitudinal strain measurements were first obtained using the automated strain analysis feature and then refined via manual adjustments to optimize endocardial border tracking. Among 374 patients with suspected ATTR-CM who underwent echocardiogram and pyrophosphate scans within 1 year, of whom 54 patients were diagnosed with ATTR-CM. ATTR-CM diagnosis was associated with older age (82 vs 75 years, p< 0.0001), male sex (78% vs 60%, p< 0.01), and a history of heart failure (91% vs. 64%, p< 0.0001). Conventional echocardiographic features predictive of ATTR-CM included higher relative wall thickness (0.70 vs. 0.43, p< 0.0001) and higher indexed right atrial volume (47.8 vs 37.8 ml/m2, p< 0.0001). Patients with ATTR-CM demonstrated lower absolute LV global longitudinal strain (-15.3% vs. -18.8%, p< 0.0001) and higher relative apical longitudinal strain (RALS) ratio (1.2 vs. 0.7, p< 0.0001). Using the established RALS ratio >1.0 cut-off, manually derived LV longitudinal strain measurements yielded a 67% sensitivity and 94% specificity to identify ATTR-CM whereas automatically derived LV longitudinal strain yielded a 54% sensitivity and 87% specificity. On receiver operating characteristic curve analysis, manual strain had a higher area-under-the-curve than automated strain (0.85 vs. 0.75).

Conclusion: We present the test characteristics of the apical-sparing LV longitudinal strain pattern to predict ATTR-CM in a large real-world cohort of patients with suspected ATTR-CM, which demonstrate a relatively low sensitivity but relatively high specificity. Notably, manually-derived strain outperformed automatically-derived strain, highlighting the importance of manual adjustments to optimize LV endocardial border tracking during strain analysis.