Development and Multinational Validation of a Novel Algorithmic Strategy for High Lp(a) Screening

Area of Science:

• Cardiology
• Biomedical Informatics
• Machine Learning

Background:

• Elevated lipoprotein(a) [Lp(a)] is a significant risk factor for atherosclerotic cardiovascular disease (ASCVD) and major adverse cardiovascular events (MACE).
• Current Lp(a) testing rates are below 0.5%, hindering the evaluation and utilization of emerging targeted therapies.
• There is a critical need for improved screening strategies to identify individuals with elevated Lp(a).

Approach:

• Developed and validated a machine learning model, Algorithmic Risk Inspection for Screening Elevated Lp(a) (ARISE), using data from 4 multinational population-based cohorts (N=479,475).
• Utilized routinely collected clinical features such as lipid profiles, medication use, and ASCVD history as input for the ARISE model.
• Compared ARISE's performance against the pooled cohort equations (PCE) for predicting elevated Lp(a) and assessed its association with mortality and MACE.

Key Points:

• ARISE demonstrated superior performance in predicting elevated Lp(a) compared to PCE, significantly reducing the number-needed-to-test (NNT) by up to 67.3%.
• The model exhibited consistent performance across diverse external validation cohorts and subgroups.
• Higher ARISE probability scores were independently associated with increased risks of all-cause mortality, cardiovascular mortality, and MACE.

Conclusions:

• The ARISE model effectively optimizes targeted screening for elevated Lp(a) using readily available clinical data.
• Deployment of ARISE in electronic health records (EHR) and other clinical settings can enhance Lp(a) testing yield.
• Improved identification of individuals with elevated Lp(a) via ARISE will facilitate eligibility for novel targeted therapeutics in clinical trials and practice.