Patterns of left ventricular remodeling post-myocardial infarction, determinants, and outcome.

ABSTRACT

AIM: Left ventricular remodeling (LVR) after myocardial infarction (MI) can lead to heart failure, arrhythmia, and death. We aim to describe adverse LVR patterns at 6 months post-MI and their relationships with subsequent outcomes and to determine baseline. METHODS AND RESULTS: A multicenter cohort of 410 patients (median age 57 years, 87% male) with reperfused MI and at least 3 akinetic LV segments on admission was analyzed. All patients had transthoracic echocardiography performed 4 days and 6 months post-MI, and 214 also had cardiac magnetic resonance imaging performed on day 4. To predict LVR, machine learning methods were employed in order to handle many variables, some of which may have complex interactions. Six months post-MI, echocardiographic increases in LV end-diastolic volume (LVEDV), LV end-systolic volume (LVESV), and LV ejection fraction (LVEF) were 14.1% [interquartile range 0.0, 32.0], 5.0% [- 14.0, 25.8], and 8.7% [0.0, 19.4], respectively. At 6 months, ≥ 15% or 20% increases in LVEDV were observed in 49% and 42% of patients, respectively, and 37% had an LVEF < 50%. The rate of death or new-onset HF at the end of 5-year follow-up was 8.8%. Baseline variables associated with adverse LVR were determined best by random forest analysis and included stroke volume, stroke work, necrosis size, LVEDV, LVEF, and LV afterload, the latter assessed by Ea or Ea/Ees. In contrast, baseline clinical and biological characteristics were poorly predictive of LVR. After adjustment for predictive baseline variables, LV dilation > 20% and 6-month LVEF < 50% were significantly associated with the risk of death and/or heart failure hazard ratio (HR) 2.12 (95% confidence interval (CI) 1.05-4.43; p = 0.04) and HR 2.68 (95% CI 1.20-6.00; p = 0.016) respectively. CONCLUSION: Despite early reperfusion and cardioprotective therapy, adverse LVR remains frequent after acute MI and is associated with a risk of death and HF. A machine learning approach identified and prioritized early variables that are associated with adverse LVR and which were mainly hemodynamic, combining LV volumes, estimates of systolic function, and afterload.