Change in left atrioventricular coupling index predicts incident heart failure with preserved ejection fraction: the multi-ethnic study of atherosclerosis
D J Lim, M Hoballah, A Abdollahi, R L Mcclelland, W Post, T Pezel, S Shea, D Bluemke, B Ambale-Venkatesh, J A C LimaAbstract
Aims
Heart failure with preserved ejection fraction (HFpEF) is increasing, yet early risk markers remain limited. The left atrioventricular coupling index (LACI), a ratio of left atrial (LA) to left ventricular (LV) end-diastolic volumes, may capture atrioventricular remodeling preceding HF. We tested whether longitudinal change in LACI (∆LACI) from cardiac magnetic resonance (CMR) predicts incident HFpEF versus HF with reduced ejection fraction (HFrEF) in adults free of clinical cardiovascular disease (CVD).
Methods
We studied 2461 multi-ethnic participants free of CVD who underwent baseline and repeat CMR ~9.5±0.5 years apart. ∆LACI (%/year) was derived between scans. Participants were followed from second scan for 8.9±2.3 years to incident HF, classified as HFpEF or HFrEF (EF≥50% or <50%).
Results
During follow-up, 102 participants developed HF (HFpEF n=56; HFrEF n=48). Each 1-standard deviation (SD) increase in ∆LACI independently predicted HFpEF (hazard ratio (HR) 1.41, 95% confidence interval (CI) 1.21-1.61, p<0.001) but not HFrEF (HR 1.18, 95%CI 0.90-1.54, p=0.22). Participants with ∆LACI≥1.6%/year had >2-fold higher adjusted HFpEF risk (HR 2.23, 95%CI 1.29-3.85, p=0.004). Adding ∆LACI to risk factors improved discrimination (C-statistic 0.812→0.837, ∆C=0.025, p=0.01), model fit (likelihood ratio test (LRT) p<0.001) and preserved calibration (Grønnesby and Borgan test (G-B) p=0.11). For HFrEF, ∆LACI did not significantly improve discrimination (∆C=0.023, p=0.06), whereas ∆minimum indexed LA volume (∆LAVimin) provided the greatest incremental value (∆C=0.031, p=0.03, LRT p=0.009, G-B p=0.10).
Conclusion
In a multi-ethnic, community-based cohort, rising ∆LACI on CMR was an independent predictor of incident HFpEF, improving discrimination and model fit with preserved calibration, while offering limited value for HFrEF, for which ∆LAVimin performed best.Figure 1Figure 2