Approved Research

Development and Internal Validation of a model to IdeNtify individuals with Elevated lipoprotein(a) [Lp(a)] levels. [DIVINE LP(a)].

Lay summary

Lipoprotein (a) or "Lp(a)" is a lipoprotein particle synthesised by the liver and its structure is similar to the well-known low-density lipoprotein (LDL), but with an additional attached protein, Apo(a). Lp(a) levels are genetically determined by ~90% and controlled by the LPA locus located in chromosome 6.

Over the last two decades, a plethora of evidence from experimental, observational and genetic studies have established the causal association of high Lp(a) with ASCVD and calcified aortic calve disease. Diet and exercises seem to have minor to no effect on Lp(a) levels, but it may decrease the overall ASCVD risk through other mechanisms. Novel drugs targeting Apo(a) production result in Lp(a) reductions by >90% and are expected to achieve a similar cardiovascular benefit to cholesterol-lowering medications.

Lp(a) levels can be used to identify individuals at high ASCVD risk, but they are often not measured in primary care settings where the patients reside. Available guidelines advocate for different strategies identifying people likely to have high Lp(a) levels. Our research aims to find such individuals using a branch of Artificial Intelligence called Machine Learning (ML). ML consists in a set of techniques where machines attempt to replicate a specific human behaviour by making predictions based on large amount of data and using complex mathematical models. Our research anticipates at the development of ML techniques which could predict Lp(a) levels and outperform current clinical guidelines. Our project is expected to be completed within 24 months from submission.

If our best ML model is better at identifying individuals with elevated Lp(a) levels than the current recommendations, it would be an ideal tool in primary care clinical practice. Primary care physicians would be able to identify individuals more likely to have an elevated Lp(a) and confirm it by measuring their Lp(a), low NNS. Therefore, the early diagnosis of this condition and its appropriate management could help prevent major adverse cardiovascular events. We hope that our research will inform how the successful use of both electronic health records and machine learning techniques can significantly improve public health.