Distribution pattern of left ventricular ejection fraction in patients with decompensated heart failure depends on sex results of a latent class analysis

Abstract

Abstract Funding Acknowledgements Type of funding sources: Private grant(s) and/or Sponsorship. Main funding source(s): unrestricted grant from Boehringer Ingelheim Background & Aims Since 2016, heart failure (HF) is classified using left ventricular ejection fraction (LVEF) thresholds of 40% and 50%. However, HF phenotypes may develop across the entire LVEF spectrum depending on individual patient characteristics including the risk and comorbidity profile. Using latent class analysis, we explored the sex-specific distribution of in-hospital LVEF in patients hospitalized for acute heart failure (AHF) at a tertiary care center in Germany. Methods Consecutive patients (≥18 years) hospitalized for AHF were recruited and phenotyped prospectively on a 7/24 basis. Exclusion criteria were high output heart failure, cardiogenic shock, and being listed for high urgency cardiac transplantation. LVEF was determined by transthoracic echocardiography using Simpson´s biplane or monoplane method. First, we estimated the distribution of LVEF in both sexes using histogram and kernel density estimation methods (bandwidth was selected by biased cross-validation). Then, Gaussian Mixture Models were fitted with increasing number of components. To identify the optimal number of subgroups we calculated the Bayesian Information Criterion (BIC). The minimum of the BIC criterion suggests the optimal number of subgroups for the final model. This analysis was performed on subsets including only male and only female patients. Results Out of 629 patients (39.8% female) admitted with AHF between 09/2014 and 12/2017, 93% patients received in-hospital echocardiography, and in 79.2% LVEF could be quantitatively assessed. The BIC suggested two subgroups each for male (Fig. A) and female patients (Fig. B). In the male two-subgroup model, mean ± SD LVEF values were 30 ± 9% and 59 ± 8%, thus covering 48% and 52% of the men, respectively (Fig. C). In the female two-subgroup model, respective LVEF values were 36 ± 13% and 65 ± 8%, thus covering 47% and 53% of patients (Fig. D). The "male" model suggested 45% as cut-point, whilst the "female" model suggested 51% as cut-point differentiating between lower and higher LVEF. Conclusions Using non-parametric and parametric statistical approaches, specific subgroups of patients hospitalized with AHF were identified among male and female patients hospitalized for AHF, which each time comprised subgroups with impaired vs. more preserved LVEF. Future analyses in larger AHF cohorts as well as in populations with chronic stable HF are warranted which take also into consideration sex differences in HF aetiology. Figure A) Minimum number of components (BIC) in men. B) Minimum BIC in women. C) LVEF distribution in men (2 components). D) LVEF distribution in women (2 components). The orange line indicates the respective cut-points between low and high LVEF. Abstract Figure.