Epidemiology
Unbiased clustering of 661 Portuguese adults reveals a distinct high-Lp(a) profile: older, male, hypertensive and diabetic (Front Cardiovasc Med 2025)
Original title: Unveiling patient profiles associated with elevated Lp(a) through an unbiased clustering analysis
In this Portuguese cohort of 661 adults referred for Lp(a) measurement, an unsupervised clustering analysis using demographic and clinical variables, with Lp(a) itself deliberately excluded from the algorithm, identified two distinct clusters that nonetheless differed sharply in Lp(a) level. Cluster 1 (n = 336) had significantly higher median Lp(a) than Cluster 2 (n = 325, p = 0.004), with 46.4% exceeding the 75 nmol/L (30 mg/dL) risk threshold versus far fewer in Cluster 2 (p < 0.001). Cluster 1 was characterised by older age (median 57 vs. 45 years), lower BMI (27.17 vs. 29.40), predominantly male composition (73.8% vs. 26.5%), and higher prevalence of hypertension (56.5% vs. 31.1%), diabetes (38.7% vs. 17.2%), and dyslipidaemia (88.7% vs. 55.4%). The findings show that a recognisable clinical phenotype, independent of the Lp(a) value itself, can flag patients most likely to have elevated Lp(a) and warrant proactive screening.
Original abstract
Introduction: Lipoprotein(a) [Lp(a)] has been recognized as key factor in cardiovascular research. This study aimed to identify key patient profiles based on the characteristics of a Portuguese cohort of adults who were referred for Lp(a) measurement.
Method: An unsupervised clustering analysis was performed on 661 Portuguese adults to identify patient profiles associated with lipoprotein a [Lp(a)] based on a range of demographic and clinical indicators. Lp(a) levels were deliberately excluded from the algorithm, to ensure an unbiased cluster formation.
Results: The analysis revealed two distinct clusters based on Lp(a) levels. Cluster 1 (n = 336) exhibited significantly higher median Lp(a) levels than Cluster 2 (n = 325; p = 0.004), with 46.4% of individuals exceeding the 75 nmol/L (30 mg/dl) risk threshold (p < 0.001). This group was characterized by older age (median 57 vs. 45 years), lower body mass index (27.17 vs. 29.40), and a majority male composition (73.8% vs. 26.5%). Additionally, Cluster 1 displayed a higher prevalence of hypertension (56.5% vs. 31.1%), diabetes mellitus (38.7% vs. 17.2%), and dyslipidemia (88.7% vs. 55.4%). These data suggest that the Cluster 1 profile has a potential increased risk for cardiovascular complications and underscore the importance of considering specific patient profiles for Lp(a) screening and cardiovascular risk assessment.
Conclusion: Despite the study limitations, including single-institution data and potential selection bias, this study highlights the utility of cluster analysis in identifying clinically meaningful patient profiles and suggests that proactive screening and management of Lp(a) levels, particularly in patients with characteristics resembling those of Cluster 1, may be beneficial.
Summary written by lp-a.org from the published abstract; figures as published. Page updated 18 August 2026. Methods.