ACCORD: Risk Calculator Outperforms AHA/ACC In CV Outcomes Prediction For Diabetes Patients
By Dave Quaile, /alert Contributor
December 7, 2020
A machine learning atherosclerotic cardiovascular disease risk calculator was more accurate in predicting atherosclerotic cardiovascular disease among adult patients with type 2 diabetes, according to the results of the ACCORD study.
The study from Wenjun Fan, MD, from the School of Medicine California Irvine and colleagues, was recently presented at AHA 2020 All-Virtual Scientific Sessions and simultaneously published in Circulation.
The study used a total of 6581 patients with T2DM from ACCORD trial with no prior CVD and calculated the 10-year risk using the ACC/AHA pooled cohort risk calculator (PCRC) to predict the composite outcome of MI, non-fatal stroke, and CV death.
The researchers used age, gender, race, systolic BP, antihypertensive, medication use, total cholesterol, high-density lipoprotein cholesterol, current smoking status and diabetes mellitus status as predictors.
The ASCVD risk calculator was developed using Random Forest learning machine algorithms as a basis and utilized follow-up data from the ACCORD study with the aforementioned predictors.
A total of 6581 T2DM participants without baseline ASCVD were included in our final sample with a median follow up of 9.1 years.
According to the researchers, the PCRC gave a modest performance, with an AUC of 0.604.
The ML model had much better performance by comparison, with an RF AUC of 0.866.
According to the abstract, age, SBP, total cholesterol and HDL-C were the most important features in predicting ASCVD among patients with T2DM.
While the results of the study are promising, the researchers also point out that there is still work to be done.
“Future studies need to validate these and other ML algorithms and to explore their applicability in guidelines,” they wrote.