Phenotyping of pulmonary arterial hypertension associated with congenital heart disease using latent class analysis: insights from a national prospective registry
Q L Li, Q Y Yu, G Z Zhang, H T Tian, Z Y Yu, C Z Zhang, Y C Chen, Y L Liu, W W Wu, X Z Zhu, C X Xiong, H H Han, Y W Wu, Q G Gu, J H HeAbstract
Background
Pulmonary arterial hypertension associated with congenital heart disease (PAH-CHD) is heterogenous on survival and treatment response. The anatomic-pathophysiological classification system—based on shunt type and location—appears to be better at predicting clinical outcome compared to the clinical system. However, significant heterogeneity still retained. The latent class analysis (LCA), a data-driven unsupervised clustering method, may identify distinct phenotypes within the PAH-CHD which help to refine prognostic prediction and management.
Methods
Adult patients (N=889) from an established multicenter prospective registry were stratified by post- and pre-tricuspid shunt. LCA was applied to each subset using 28 clinical and hemodynamic variables including age, sex, and inflammation markers. The optimal number of phenogroups was determined based on the Bayesian information criterion/Integrated Completed Likelihood, and entropy. Survival and effect of PAH-targeted therapy across the phenogroups were evaluated.
Results
Post-tricuspid shunts were associated with increased mortality, whereas clinical subgroups showed no significant differences in survival. Using LCA, four phenogroups were identified in each subset, showing distinct phenotypic patterns. In the post-tricuspid subset, Group 1 (N=212, 43.5%) was labeled as “adaptation and open shunt” and showed preserved RV function and the best survival despite a high afterload. Group 3 (“maladaptation and failure”; N=113, 23.1%) showed a similar proportion of clinical Eisenmenger syndrome with Group 1 but a 2.1-fold higher mortality risk. Group 2 (“hyperkinetic state”; N=97, 19.9%) and Group 4 (“closed shunt and failure”; N=66, 13.5%) both showed a 2-fold increased mortality risk compared to Group 1. Group 2 derived particular benefit from combination therapy (interaction P=0.026). Early referral for transplantation evaluation should be considered for Group 3 and 4 due to their high mortality. In the pre-tricuspid subset, Group 1 (N=70,17.6%) was labeled as“high shunt and mild resistance” with best survival. Group 2 (“old male and comorbidities”; N=72, 17.9%) showed hemodynamics similar to Group 1 but a significantly higher mortality risk. The effect to PAH-targeted therapy was attenuated in Group 2 (interaction P=0.037). Preserved survival was observed in Group 3 (N=173, 43.1%) with a “young female and low inflammation” phenotype despite a high afterload level. Group 4 was labeled as “high resistance and failure” with severe hypoxemia and the worst survival, and early referral for transplantation evaluation was warranted.
Conclusions
Using LCA, distinct phenogroups within PAH-CHD were identified with varying clinical profiles, long-term survival, and treatment responses. This phenotyping system may help to refine prognostic assessments and treatment strategies.HeatmapFor image description, please refer to the figure legend and surrounding text.Kaplan-Meier curvesFor image description, please refer to the figure legend and surrounding text.