Phenotypic Trajectories From Acute to Stable Phase in Heart Failure With Preserved Ejection Fraction: Insights From the PURSUIT‐HFpEF Study

Figure 1. Study flowchart. A total of 1231 patients were enrolled in the PURSUIT HFpEF registry. Nineteen patients died during hospitalization and there were 112 patients who did not have the prognostic information. We had 324 variables in this registry and excluded 175 variables such as data on admission, prognostic information, and information of treatments after discharge. In addition, we excluded variables with significant missingness (≥20%) and high degree of correlation (correlation coefficient >0.60). A total of 1100 patients and 63 variables were finally included in the machine learning‐based unsupervised cluster analysis (latent class analysis).

Figure 2. Discriminative power ranking. Discriminative power of the 26 variables selected by latent class analysis is shown as a bar graph. We computed the discriminative power of each variable which was defined as the logarithm of the ratio between the probability that the variable is relevant for clustering and the variable is irrelevant for clustering.

Figure 3. Clinical features in 5 phenogroups. Relative differences in the clinical features among the 5 phenogroups are illustrated in the radar chart. A, Overlay of all 5 phenotypes (1–5) for overall comparison. B through F, Emphasized profiles of Phenotypes 1 through 5, respectively, highlighting their distinct characteristics. Continuous variables are displayed as means of Z‐standardized values and categorical variables are displayed as proportion. AF indicates atrial fibrillation; BMI, body mass index; GGT, gamma‐glutamyl transpeptidase; CRP, C‐reactive protein; CTR, cardio thoracic ratio; GNRI, geriatric nutritional risk index; HF, heart failure; LAVI, left atrial volume index; LVDd, left ventricular diastolic dimension; LVMI, left ventricular mass index; PVS, plasma volume status; sBP, systolic blood pressure; and TRPG, transtricuspid pressure gradient.

Figure 4. Clinical outcomes. Survival analysis was conducted with Kaplan–Meier methods for a composite of all‐cause death and heart failure readmission (A), all‐cause death (B), and heart failure readmission (C). HF indicates heart failure.

Figure 5. Phenotypic trajectory from hospital admission (acute phase) to discharge (stable phase). This figure comprises 4 pie charts, each representing the percentage distribution of stable‐phase phenotypes (at hospital discharge) across acute‐phase phenotypes (at hospital admission). BNP indicates brain natriuretic peptide; TRPG, transtricuspid pressure gradient; TAPSE, tricuspid annular plane systolic excursion; and CRP, C‐reactive protein.