Analysis of Differences in Assessment of Left Ventricular Function on Echocardiography and Nuclear Perfusion Imaging

Two widely used methods for left ventricular (LV) ejection fraction (EF) determination, echocardiography (echo) and gated single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI), often have wide limits of agreement. Factors influencing discrepancies between core laboratory echo and MPI LVEF determinations were examined in a large series of heart failure (HF) subjects and normal controls. 879 HF and 101 control subjects had core lab analyses of echo and MPI (mean time between procedures 7-8 days). LVEF differences were analyzed using one-way analysis of variance and Bland-Altman plots. Relationships between LVEF differences and patient characteristics and outcome endpoints (mortality and arrhythmias) were explored with logistic regression, Cox proportional hazards models, and Kaplan-Meier survival analyses. There was a systematic difference between the 2 modalities; echo LVEF was higher with more severe LV dysfunction, MPI LVEF higher when systolic function was normal. LVEF results were within ±5% in only 37% of HF and 23% of control subjects. Considering discordance around the LVEF threshold 35%, there was disagreement between the 2 methods in 305 HF subjects (35%). Male gender (odds ratio (OR) = 0.200), atrial fibrillation (OR = 2.314), higher body mass index (OR = 1.051) and lower LV end-diastolic volume (OR = 0.985) were the strongest predictors of methodologic discordance. Cardiac event rates were highest if both LVEF values were ≤35% and lowest when both LVEF values were >35%. In conclusion, substantial disagreements between LVEF results by echo and MPI are common. HF patients with LVEF ≤35% by both techniques have the highest 2-year event risk.