Recent research indicates that machine learning can effectively identify individuals with diabetes who are at high risk for diabetic cardiomyopathy (DbCM). This advancement, detailed in a study published in the European Journal of Heart Failure on September 6, offers potential for more targeted heart failure prevention strategies.
Led by Dr. Matthew W. Segar from the Texas Heart Institute in Houston, the study utilized a machine learning-based clustering approach to classify individuals based on echocardiographic data and cardiac biomarker parameters. The initial training model analyzed data from 1,199 diabetes patients participating in the Atherosclerosis Risk in Communities cohort, all of whom were free from cardiovascular disease at the start of the study. Validation was achieved through data from 802 participants in the Cardiovascular Health Study and a separate cohort comprising 5,071 electronic medical records.
The findings revealed that patients classified as phenogroup-3, representing the high-risk DbCM phenotype (comprising 324 patients), exhibited a significantly higher five-year incidence of heart failure compared to other phenogroups—12.1% for phenogroup-3 versus 4.6% for phenogroup-2 and 3.1% for phenogroup-1. Critical echocardiographic predictors for this high-risk group included elevated levels of N-terminal pro-B-type natriuretic peptide, increased left ventricular mass and left atrial size, along with deteriorated diastolic function. In the external validation cohorts, a deep neural network classifier identified 16% and 29% of participants with DbCM, respectively. Those in the high-risk phenotype group experienced a notably higher incidence of heart failure.
The authors assert that utilizing a machine learning-based approach to identify patients at risk for DbCM can facilitate a risk-based strategy in administering effective but potentially costly heart failure preventive therapies. This could significantly benefit the highest-risk subgroup of diabetes patients.
It’s important to note that several authors of the study disclosed affiliations with the pharmaceutical and biotechnology sectors.
Related topics:
Increase in Prescriptions for Weight-Management Drugs Among Type 1 Diabetes Patients
Rising Diabetes Rates in Indonesia: A Public Health Crisis
Decline in Type 2 Diabetes Incidence: A Mixed Picture for Australian Communities
