Phenotyping of pulmonary arterial hypertension associated with congenital heart disease using latent class analysis: insights from a national prospective registry.

BACKGROUND: Pulmonary arterial hypertension associated with congenital heart disease (PAH-CHD) is characterized by substantial prognostic and therapeutic heterogeneity. Conventional clinical and anatomic-pathophysiological classifications fail to fully capture these intricacies. Unsupervised clustering approaches such as latent class analysis (LCA) have been used to address this limitation. RESEARCH QUESTION: Can LCA-derived phenogroups refine risk stratification and predict PAH-targeted treatment responses in adult patients with PAH-CHD? STUDY DESIGN AND METHODS: Adult PAH-CHD patients (N=889) from a multicenter prospective registry were stratified into post- and pre-tricuspid shunt subsets. LCA was performed in each subset using 28 candidate clinical, laboratory, and hemodynamic variables. Ten-year survival and the effect of PAH-targeted therapy across phenogroups were evaluated. RESULTS: LCA identified four phenogroups in each subset with distinct phenotyping patterns. In post-tricuspid shunts, Group 1 ("adaptation with open shunt") exhibited preserved right ventricular function and the best survival despite high afterload. Groups 3 ("maladaptation and failure") and 4 ("closed shunt and failure") faced 2 to 3-fold higher mortality than Group 1. Notably, Group 2 ("hyperkinetic state") uniquely benefited from combination therapy (interaction P=0.026). In pre-tricuspid shunts, Groups 1 ("high shunt and mild resistance") and 3 ("young female and low inflammation") demonstrated preserved survival despite varying afterloads. However, Group 2 ("old male and comorbidities") had significantly higher mortality and an attenuated treatment response (interaction P=0.037) despite sharing mild hemodynamics with Group 1. Group 4 ("high resistance and failure") faced severe hypoxemia and the worst prognosis. INTERPRETATION: LCA successfully stratified PAH-CHD patients into distinct phenogroups with varied clinical profiles, survival, and treatment responses. This data-driven phenotyping system provides a novel approach for risk stratification and may guide personalized clinical decision-making.