Key result
LA reservoir strain alone matches the 11-variable CHARGE-AF score for predicting incident AF.
Why the study?
Left atrial function measures from two-dimensional echocardiograms are linked to AF development, but their standalone predictive value for long-term AF remained unclear.
Does standalone left atrial reservoir strain or a machine learning-based echocardiographic model accurately predict 5- and 10-year incident atrial fibrillation compared to the multivariable CHARGE-AF score in older adults?
Cohort (n=5,442)
Yes
Does standalone left atrial reservoir strain or a machine learning-based echocardiographic model accurately predict 5- and 10-year incident atrial fibrillation compared to the multivariable CHARGE-AF score in older adults?
A single echocardiographic measure—left atrial reservoir strain—predicts 5- and 10-year incident atrial fibrillation nearly as well as the 11-variable clinical CHARGE-AF score, and combining it with routine echocardiographic parameters via machine learning further improves long-term risk discrimination.
Background: Left atrial (LA) function measures from the two-dimensional echocardiograms are linked to atrial fibrillation (AF) development, but their standalone predictive value remains unclear. Objective: This study assessed the standalone predictive power of LA function measures for long-term AF prediction (5 and 10 years). Methods: We analyzed 5,442 older adults (median age 74) from the Atherosclerosis Risk in Communities (ARIC) study without prior AF. Participants were split into training and testing sets (7:3 ratio). We evaluated the standalone predictive performance of LA strain measures (reservoir, conduit, and contraction strain) with and without age for long-term AF. The performance was compared to that of the CHARGE-AF score and a machine learning-based predictive model that incorporates the most important routinely measured echocardiographic parameters with LA strain measures. Results: LA reservoir strain alone achieved a Harrell’s C-index of 0.664 and 0.646 for 5- and 10-year AF prediction, improving to 0.677 and 0.663 when combined with age, which outperformed the CHARGE-AF score (0.667 and 0.655) and other LA strain measures. A better predictive performance was achieved by the machine learning-based predictive model, including LA reservoir strain, age, race, and 9 echocardiographic parameter; the C-indexes were 0.732 (0.686-0.777) for 5-year and 0.725 (0.693-0.757) for 10-year prediction. Conclusion: The standalone predictive performance of LA reservoir strain for long-term AF risk is comparable to CHARGE-AF score (11 variables) and slightly lower than that of a machine learning-based model with 12 variables, highlighting the importance of LA reservoir in AF risk prediction.
No takes yet. Share an insight, caveat, or question.
Sun et al. (2025) conducted a cohort in Risk of incident atrial fibrillation (n=5,442). Left atrial reservoir strain (standalone and in a 12-variable LASSO machine learning echocardiographic model) vs. CHARGE-AF score (11 clinical variables) and other LA strain measures (conduit and contraction strain) was evaluated on 5-year and 10-year incident atrial fibrillation prediction (Harrell's C-index) (C-index 0.664 (standalone 5-yr); LASSO model C-index 0.732 (5-yr), 95% CI 0.686-0.777). Left atrial reservoir strain alone achieved C-indexes of 0.664 and 0.646 for 5- and 10-year AF prediction, comparable to the 11-variable CHARGE-AF score (0.667 and 0.655).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: