New strategies for identifying patients with high cardiovascular risk

Study shows machine-learning-based clustering using finite mixture models can help identify phenogroups of people with type 2 diabetes, distinct clinical characteristics and cardiovascular risk.

Individuals with type 2 diabetes are at increased risk for cardiovascular disease (CVD) morbidity and mortality. In a new study published in Diabetes Care, Segar et al. developed and validated different phenomapping strategies and their ability to classify cardiovascular risk in individuals with type 2 diabetes. Using these results, they aimed to identify the optimal phenomapping strategy and identify subgroups who might benefit from specific therapies – a risk-based approach that could be used to help prevent adverse cardiovascular events in patients with type 2 diabetes.

Participants with type 2 diabetes and free of baseline CVD in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial were included in the study (N=6466). Semi-supervised clustering was applied using four phenomapping strategies: Gaussian mixture models, latent class analysis, finite mixture models and principal component analysis. Variables used to classify clustering included demographics, medical and social history, laboratory values and diabetes complications. The interaction between the phenogroup and intensive glycaemic, combination lipid and intensive blood pressure therapy for the risk of composite of fatal myocardial infarction, non-fatal myocardial infarction or unstable angina, the primary outcome, was evaluated.

After a follow-up period of over 9.1 years, 789 (12.2%) participants had a primary outcome event. Clustering with finite mixture models was the optimal phenomapping strategy and identified three subgroups of patients with type 2 diabetes with distinct clinical phenotypes, cardiovascular disease risk and response to different therapies. Phenogroup 1 (n=663; 10.3%) had the highest burden of comorbidities and diabetes complications, phenogroup 2 (n=2388; 36.9%) had an intermediate comorbidity burden and lowest diabetes complications and phenogroup 3 (n=3415; 52.8%) was found to have the fewest comorbidities and intermediate burden of diabetes complications.

The best-performing clustering strategy was shown to be finite mixture models phenomapping with three phenogroups. Individuals with a lower burden of risk factors and favourable clinical profile had a beneficial treatment response to intensive glycaemic control, combination lipid therapy, intensive lifestyle intervention and early coronary revascularisation. The results of this study were published in Diabetes Care in March 2021.

Visit the following link for more information: https://link.springer.com/article/10.1007%2Fs00125-021-05426-2.

For more on phenomapping and type 2 diabetes, see our course Phenotypic variability and type 2 diabetes.

For more on CVD and type 2 diabetes, see our course Cardiovascular health.

The views expressed in this article are those of the author, Dr Eleanor D Kennedy.