Phenomapping for the Identification of Hypertensive Patients with the Myocardial Substrate for Heart Failure with Preserved Ejection Fraction

J Cardiovasc Transl Res. 2017 Jun;10(3):275-284. doi: 10.1007/s12265-017-9739-z. Epub 2017 Mar 3.


We sought to evaluate whether unbiased machine learning of dense phenotypic data ("phenomapping") could identify distinct hypertension subgroups that are associated with the myocardial substrate (i.e., abnormal cardiac mechanics) for heart failure with preserved ejection fraction (HFpEF). In the HyperGEN study, a population- and family-based study of hypertension, we studied 1273 hypertensive patients utilizing clinical, laboratory, and conventional echocardiographic phenotyping of the study participants. We used machine learning analysis of 47 continuous phenotypic variables to identify mutually exclusive groups constituting a novel classification of hypertension. The phenomapping analysis classified study participants into 2 distinct groups that differed markedly in clinical characteristics, cardiac structure/function, and indices of cardiac mechanics (e.g., phenogroup #2 had a decreased absolute longitudinal strain [12.8 ± 4.1 vs. 14.6 ± 3.5%] even after adjustment for traditional comorbidities [p < 0.001]). The 2 hypertension phenogroups may represent distinct subtypes that may benefit from targeted therapies for the prevention of HFpEF.

Keywords: Cardiac mechanics; Heart failure with preserved ejection fraction; Hypertension; Machine learning; Speckle-tracking echocardiography.

Publication types

  • Multicenter Study

MeSH terms

  • Adult
  • Aged
  • Biomechanical Phenomena
  • Blood Pressure*
  • Cluster Analysis
  • Cross-Sectional Studies
  • Echocardiography, Doppler
  • Female
  • Heart Failure / diagnostic imaging
  • Heart Failure / etiology*
  • Heart Failure / physiopathology
  • Humans
  • Hypertension / classification
  • Hypertension / complications*
  • Hypertension / diagnostic imaging
  • Hypertension / physiopathology
  • Machine Learning*
  • Male
  • Middle Aged
  • Pattern Recognition, Automated
  • Phenotype
  • Predictive Value of Tests
  • Prognosis
  • Risk Assessment
  • Risk Factors
  • Stress, Mechanical
  • Stroke Volume*
  • United States