Clinical relevance of the “Risk of Heart Disorders” Score from a cloud-based analysis of Holter ECG data (CardiolyseTM) for development and progression of heart failure

Introduction Markers derived from Holter electrocardiograms (ECGs) are known to be associated with clinical outcome in heart failure (HF), but their predictive value, particularly in comparison to established markers like NT-proBNP or HF risk scores such as the Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC) Risk Score, is not yet fully elucidated.

Aim: This study aimed to assess how the most predictive marker from a novel cloud-based Holter ECG analysis, can predict clinical outcome in individuals with HF.

Methods: Data from the MyoVasc study (NCT04064450), a prospective cohort on HF, were analyzed. Subjects underwent a standardized 5-hour examination, that included Holter ECG recording. The ECG data were analysed using a novel cloud-based analysis platform developed by Cardiolyse (Helsinki, Finland). To determine the most predictive parameter for worsening heart failure (WoHF) and all-cause death (ACD), elastic-net Cox regression models were utilized. Clinical determinants and the relationship to cardiac structure and function were assessed by linear regression models adjusted for sex, age, medication, and comorbidities. Finally, the best predictive parameter was evaluated using multivariable Cox regression models adjusted for sex, age, medication, and comorbidities.

Results: The analyzed sample included Holter ECG data from 953 subjects; symptomatic HF was present in 522 individuals. In total, 180 markers were analyzed by the cloud-based platform, with the composite parameter Risk of Heart Disorders (RoHD) proving to be most predictive. RoHD was associated with coronary artery disease (β -0.04 [-0.06; -0.01], p=0.011), diabetes mellitus (β 0.05 [0.01; 0.08], p=0.007), atrial fibrillation (β 0.65 [0.52; 0.79], p<0.0001) as well as intake of diuretics (β 0.09 [0.07; 0.11], p<0.0001). It was correlated to all 3 dimensions of cardiac function, thus left ventricular ejection fraction (β -2.71 [-3.29; -2.12], p<0.0001) global longitudinal strain (β 0.996 [0.679; 1.31], p<0.0001) and diastolic function measured by lateral left ventricular E/E’ (β 0.996 [0.679; 1.31], p=0.004). Additionally, RoHD was related to cardiac structure comprising left ventricular mass/height (β 2.45 [1.61; 3.29], p<0.0001), interventricular septum thickness in diastole (β 0.02 [0.00; 0.03], p=0.024) and left ventricle posterior wall thickness (β0.01 [0.00; 0.03], p=0.06), whereas elevation of RoHD indicated cardiac hypertrophy. The risk of WoHF was significantly elevated in individuals above the 95^th percentile of RoHD (derived from a subgroup of individuals without HF) after 4 years of follow-up (p<0.0001). RoHD predicted WoHF (hazard ratio per standard deviation (HR_SD) 1.33 [1.18; 1.50], p<0.0001) and ACD (HR_SD 1.34 [1.18; 1.52], p<0.0001) both in individuals with heart failure with preserved and reduced ejection fraction in fully adjusted models independent of the established markers NT-proBNP and MAGGIC risk score.

Conclusion: The composite ECG parameter “Risk of Heart Disorders” computed by cardiac analytics platform Cardiolyse (Helsinki, Finland) was related with the presence of cardiovascular risk factors, and cardiac structural and functional changes in individuals with HF. RoHD carried independent information from the established markers NT-proBNP and MAGGIC risk score for the prediction of clinical outcome in individuals with heart failure.