The combination of pre- and postcapillary forms of pulmonary hypertension (PH) is called ‘combined-postcapillary PH’. Nowadays, detection of this phenotype requires right heart catheterization.

Using dense phenotypic data (phenomapping), we aimed to detect noninvasively combined-postcapillary PH.

We included 82 patients (mean age 67.8±11.6 y, males 47%) from a referral center with mean pulmonary artery pressure (PAP)≥25mmHg and wedge pressure (WP)>15mmHg. Combined-postcapillary PH was defined by diastolic pressure gradient≥7mmHg and/or pulmonary vascular resistance (PVR)>3 WU. Phenotypic domains (including clinical, laboratory, imaging and hemodynamic variables) were imputed for missing values, filtered when correlated (coefficient>0.6), grouped using agglomerative hierarchical clustering and participants were separated using penalized model-based clustering. Phenogroups were compared for phenotypic domains and 2-years all-causes mortality was analyzed by Cox regression.

Mean PAP, WP, cardiac index and PVR were 39.6±9.8, 23.2±5.2, 2.96±0.77 and 3.21±1.77 respectively. Among 65 phenotypic domains, 31 entered the model from which heatmap was drawn (Fig 1) and 3 phenogroups were identified. Combined-postcapillary PH was found in 90% of phenogroup 3 remarkable by older age females (P<0.01) with more hypertension, atrial fibrillation, normal ejection fraction, smaller ventricular cavities, higher tricuspid regurgitation velocity and more severe functional status. Compared to phenogroup 1, mortality (Fig 2) was not increased in phenogroup 3 (Hazard Ratio 0.79 [0.2–3.1], P=0.73) but in phenogroup 2, characterized by younger males with altered ejection fraction and only 27% combined-PH forms (HR 3.74 [1.2–11.3], P=0.02).

From heterogeneous clinical presentation, phenomapping using statistical learning algorithms allowed to identify non-invasively a population at high risk of combined-postcapillary PH.