Phenotypic clustering identifies heterogeneous cardiovascular risk among patients with elevated lipoprotein(a)

Lipoprotein(a) [Lp(a)] is an established cardiovascular risk enhancer, yet fixed concentration thresholds may not fully capture the heterogeneity of cardiovascular risk among individuals with elevated levels. We retrospectively analyzed 17,653 patients with Lp(a) measurements from three tertiary hospitals (2017-2024). After exclusions, 2355 patients with Lp(a) ≥ 50 mg/dL underwent k-means clustering based on 23 demographic, comorbidity, and laboratory variables, which identified two phenotypic groups. Cluster validation using elbow, silhouette, and NbClust consensus methods supported a two-cluster solution. Cluster 1 consisted of older, male-predominant patients with a higher cardiometabolic burden and lower renal function, whereas Cluster 2 included younger, female-predominant patients with fewer comorbidities and relatively treatment-naïve dyslipidemia. Despite similar Lp(a) levels, over a median follow-up of 2.3 years (interquartile range [IQR] 1.0-3.0), major adverse cardiovascular events (MACE) occurred more frequently in Cluster 1 than Cluster 2 (8.9% vs. 2.0%, log-rank p < 0.01). In multivariable Cox models, Cluster 1 was associated with higher MACE risk compared with the Lp(a) < 30 mg/dL reference group (HR 1.39, 95% CI 1.13-1.72), whereas Cluster 2 showed no significant risk difference (HR 1.08 95% CI 0.69-1.68). These findings suggest that phenotypic clustering of high-Lp(a) patients delineates subgroups with distinct cardiovascular risk profiles. Incorporating phenotype-guided risk assessment may refine cardiovascular risk stratification beyond fixed Lp(a) thresholds.