User login
PHILADELPHIA – For patients with high-risk diabetes, a novel, machine learning–derived risk score based on 10 common clinical variables can identify those facing a heart failure risk of up to nearly 20% over the ensuing 5 years, an investigator said at the annual meeting of the Heart Failure Society of America.
The risk score, dubbed WATCH-DM, has greater accuracy in predicting incident heart failure than traditional risk-based models, and requires no specific cardiovascular biomarkers or imaging, according to Muthiah Vaduganathan, MD, MPH, a cardiologist at Brigham and Women’s Hospital and faculty at Harvard Medical School in Boston.
The tool may help inform risk-based monitoring and introduction of sodium-glucose transporter 2 (SGLT2) inhibitors, which have been shown in multiple clinical trials to prevent heart failure in at-risk patients with type 2 diabetes mellitus (T2DM), Dr. Vaduganathan said.
“Patients identified at high risk based on WATCH-DM should be strongly considered for initiation of SGLT2 inhibitors in clinical practice,” Dr. Vaduganathan said in an interview.
WATCH-DM is available online at cvriskscores.com. Work is underway to integrate the tool into electronic health record systems at Brigham and Women’s Hospital and at the University of Texas Southwestern Medical Center in Dallas. “I expect that to be launched in the next year,” he said.
The WATCH-DM score was developed based on data from the ACCORD (Action to Control Cardiovascular Risk in Diabetes) trial, including 8,756 T2DM patients with inadequate glycemic control at high cardiovascular risk and no heart failure at baseline.
Starting with 147 variables, the investigators used a decision-tree machine learning approach to identify predictors of heart failure.
“What machine learning does is automate the variable selection process, as a form of artificial intelligence,” Dr. Vaduganathan said.
The WATCH-DM risk score was based on the 10 best-performing predictors as selected by machine learning, including body mass index, age, systolic blood pressure, diastolic blood pressure, fasting plasma glucose, serum creatinine, high-density lipoprotein cholesterol, QRS duration, prior myocardial infarction, and prior coronary artery bypass grafting.
The 5-year risk of heart failure was just 1.1% for patients with WATCH-DM scores in the lowest quintile, increasing in a graded fashion to nearly 20% (17.4%) in the highest quintile, study results show.
Findings of the study were simultaneously published in the journal Diabetes Care.
Dr. Vaduganathan said he is supported by an award from Harvard Catalyst. He provided disclosures related to Amgen, AstraZeneca, Baxter Healthcare, Bayer AG, Boehringer Ingelheim (advisory boards), and with Novartis and the National Institutes of Health (participation on clinical endpoint committees).
SOURCE: HFSA 2019; Segar MW, Vaduganathan M et al. Diabetes Care. doi: 10.2337/dc19-0587.
PHILADELPHIA – For patients with high-risk diabetes, a novel, machine learning–derived risk score based on 10 common clinical variables can identify those facing a heart failure risk of up to nearly 20% over the ensuing 5 years, an investigator said at the annual meeting of the Heart Failure Society of America.
The risk score, dubbed WATCH-DM, has greater accuracy in predicting incident heart failure than traditional risk-based models, and requires no specific cardiovascular biomarkers or imaging, according to Muthiah Vaduganathan, MD, MPH, a cardiologist at Brigham and Women’s Hospital and faculty at Harvard Medical School in Boston.
The tool may help inform risk-based monitoring and introduction of sodium-glucose transporter 2 (SGLT2) inhibitors, which have been shown in multiple clinical trials to prevent heart failure in at-risk patients with type 2 diabetes mellitus (T2DM), Dr. Vaduganathan said.
“Patients identified at high risk based on WATCH-DM should be strongly considered for initiation of SGLT2 inhibitors in clinical practice,” Dr. Vaduganathan said in an interview.
WATCH-DM is available online at cvriskscores.com. Work is underway to integrate the tool into electronic health record systems at Brigham and Women’s Hospital and at the University of Texas Southwestern Medical Center in Dallas. “I expect that to be launched in the next year,” he said.
The WATCH-DM score was developed based on data from the ACCORD (Action to Control Cardiovascular Risk in Diabetes) trial, including 8,756 T2DM patients with inadequate glycemic control at high cardiovascular risk and no heart failure at baseline.
Starting with 147 variables, the investigators used a decision-tree machine learning approach to identify predictors of heart failure.
“What machine learning does is automate the variable selection process, as a form of artificial intelligence,” Dr. Vaduganathan said.
The WATCH-DM risk score was based on the 10 best-performing predictors as selected by machine learning, including body mass index, age, systolic blood pressure, diastolic blood pressure, fasting plasma glucose, serum creatinine, high-density lipoprotein cholesterol, QRS duration, prior myocardial infarction, and prior coronary artery bypass grafting.
The 5-year risk of heart failure was just 1.1% for patients with WATCH-DM scores in the lowest quintile, increasing in a graded fashion to nearly 20% (17.4%) in the highest quintile, study results show.
Findings of the study were simultaneously published in the journal Diabetes Care.
Dr. Vaduganathan said he is supported by an award from Harvard Catalyst. He provided disclosures related to Amgen, AstraZeneca, Baxter Healthcare, Bayer AG, Boehringer Ingelheim (advisory boards), and with Novartis and the National Institutes of Health (participation on clinical endpoint committees).
SOURCE: HFSA 2019; Segar MW, Vaduganathan M et al. Diabetes Care. doi: 10.2337/dc19-0587.
PHILADELPHIA – For patients with high-risk diabetes, a novel, machine learning–derived risk score based on 10 common clinical variables can identify those facing a heart failure risk of up to nearly 20% over the ensuing 5 years, an investigator said at the annual meeting of the Heart Failure Society of America.
The risk score, dubbed WATCH-DM, has greater accuracy in predicting incident heart failure than traditional risk-based models, and requires no specific cardiovascular biomarkers or imaging, according to Muthiah Vaduganathan, MD, MPH, a cardiologist at Brigham and Women’s Hospital and faculty at Harvard Medical School in Boston.
The tool may help inform risk-based monitoring and introduction of sodium-glucose transporter 2 (SGLT2) inhibitors, which have been shown in multiple clinical trials to prevent heart failure in at-risk patients with type 2 diabetes mellitus (T2DM), Dr. Vaduganathan said.
“Patients identified at high risk based on WATCH-DM should be strongly considered for initiation of SGLT2 inhibitors in clinical practice,” Dr. Vaduganathan said in an interview.
WATCH-DM is available online at cvriskscores.com. Work is underway to integrate the tool into electronic health record systems at Brigham and Women’s Hospital and at the University of Texas Southwestern Medical Center in Dallas. “I expect that to be launched in the next year,” he said.
The WATCH-DM score was developed based on data from the ACCORD (Action to Control Cardiovascular Risk in Diabetes) trial, including 8,756 T2DM patients with inadequate glycemic control at high cardiovascular risk and no heart failure at baseline.
Starting with 147 variables, the investigators used a decision-tree machine learning approach to identify predictors of heart failure.
“What machine learning does is automate the variable selection process, as a form of artificial intelligence,” Dr. Vaduganathan said.
The WATCH-DM risk score was based on the 10 best-performing predictors as selected by machine learning, including body mass index, age, systolic blood pressure, diastolic blood pressure, fasting plasma glucose, serum creatinine, high-density lipoprotein cholesterol, QRS duration, prior myocardial infarction, and prior coronary artery bypass grafting.
The 5-year risk of heart failure was just 1.1% for patients with WATCH-DM scores in the lowest quintile, increasing in a graded fashion to nearly 20% (17.4%) in the highest quintile, study results show.
Findings of the study were simultaneously published in the journal Diabetes Care.
Dr. Vaduganathan said he is supported by an award from Harvard Catalyst. He provided disclosures related to Amgen, AstraZeneca, Baxter Healthcare, Bayer AG, Boehringer Ingelheim (advisory boards), and with Novartis and the National Institutes of Health (participation on clinical endpoint committees).
SOURCE: HFSA 2019; Segar MW, Vaduganathan M et al. Diabetes Care. doi: 10.2337/dc19-0587.
REPORTING FROM HFSA 2019