Lipid monitoring using non-invasive measurement technologies and machine learning: a systematic review

Cardiovascular diseases (CVD) are the leading cause of death among women, with risk increasing after menopause. Lipid levels are key biomarkers, yet conventional blood tests remain invasive and underutilized. Non-invasive technologies and machine learning (ML) may offer new approaches to lipid monitoring and risk assessment using wearable devices and biosensors.

This systematic review investigates the availability, accuracy, and clinical applicability of minimally and non-invasive lipid monitoring methods and ML-based cardiovascular risk estimation in adults.

A systematic search was conducted in MEDLINE, Embase, Cochrane Library, Web of Science, Scopus, and ClinicalTrials.gov (2010-2024). Studies in English were included; case reports and animal studies were excluded. Data extraction focused on devices, measurement approach, and predictive utility for cardiovascular outcomes. Methodological heterogeneity was addressed through narrative synthesis and thematic grouping (Thomas in Cochrane Handb Syst Rev Interv, 2024).

From 14,863 records, 37 studies were included. Near-infrared, saliva-based, and smartphone-enabled fingertip devices showed promising accuracy. ML models using wearable-derived physiological data demonstrated moderate success in predicting cardiovascular risk and lipid levels.

Minimally and non-invasive lipid monitoring and ML-based risk prediction may support accessible, personalized cardiovascular risk management. Despite encouraging findings, validation in large-scale, long-term studies is essential before clinical adoption.

Title registration number (on PROSPERO): CRD420251105896.