Adding LAA morphology and Lp(a) to a thrombus-risk model beats CHA2DS2-VA in atrial fibrillation, cohort of 418 (Int J Cardiol Heart Vasc 2026)

Original title: Integrating left atrial appendage morphology and lipoprotein(a) improves thrombus risk stratification in atrial fibrillation: a retrospective cohort study

Int J Cardiol Heart Vasc · 19 May 2026 · 7

Song X, Yang J, Li Y, Zhao Z, Xu D, Wang J, Wang X, Zhang J

Retrospective cohort of 418 atrial fibrillation (AF) patients (262 without thrombus, 156 with thrombus), building a six-predictor model (non-paroxysmal AF, CHA2DS2-VA score 2 or more, anticoagulation use, AF duration, left atrial appendage orifice ellipticity index [LAA OEI], and Lp(a)) for left atrial thrombus risk. The model showed excellent discrimination (AUC 0.861, 95% CI 0.829-0.892) and good calibration (Hosmer-Lemeshow p=0.462), significantly outperforming the CHA2DS2-VA score alone (AUC 0.643, delta AUC 0.2186, p<0.0001), with a net reclassification improvement of 0.844 and integrated discrimination improvement of 0.311 (both p<0.001). The authors conclude integrating LAA morphology and Lp(a) with clinical risk factors meaningfully improves thrombus risk stratification in AF compared with the CHA2DS2-VA score alone.

Original abstract

Background: Accurate assessment of thrombus risk in atrial fibrillation (AF) requires the integration of structural and metabolic factors, which are insufficiently addressed by current clinical risk scores. We developed a model that added LAA morphology and metabolic markers, and compared it with the CHA2DS2‑VA score.

Methods: This retrospective cohort study included 418 participants (262 AF patients without thrombus and 156 AF patients with thrombus). Data on medical history, contrast‑enhanced computed tomography with three‑dimensional reconstruction, and metabolic profiling were retrospectively collected.

Results: The final model comprised six independent predictors: non‑paroxysmal AF, CHA2DS2‑VA score ≥ 2, anticoagulation use, AF duration, LAA orifice ellipticity index (LAA OEI), and lipoprotein(a) (Lp(a)). The model demonstrated excellent discrimination (AUC = 0.861, 95% CI 0.829-0.892) and good calibration (Hosmer‑Lemeshow test p = 0.462). It significantly outperformed the CHA2DS2‑VA score (AUC = 0.643; ΔAUC = 0.2186; p < 0.0001), with a net reclassification improvement of 0.844 (p < 0.001) and an integrated discrimination improvement of 0.311 (p < 0.001).

Conclusions: Integrating LAA OEI and Lp(a) with clinical factors improves thrombus risk stratification in patients with AF compared with the CHA2DS2‑VA score alone.

Summary written by lp-a.org from the published abstract; figures as published. Page updated 17 August 2026. Methods.