Calculator uses AI to estimate 10-year risks for diabetes complications after bariatric surgery
LAS VEGAS— For people with type 2 diabetes and obesity, a novel calculator that employs machine learning technology can provide personalized, evidence-based risk information about future cardiovascular outcomes and mortality, with and without bariatric surgery, according to findings presented at ObesityWeek.
“Although weight loss surgery significantly improves type 2 diabetes and risk for the adverse events of diabetes and obesity on body organs, less than 1% of eligible patients undergo weight loss surgery,” Ali Aminian, MD, FACS, FASMBS, associate professor of surgery at the Cleveland Clinic, told Endocrine Today. “One reason for underutilization of weight loss surgery would be a lack of information on the benefits of procedures. Evidence-based risk prediction can be helpful to inform treatment decisions for patients and physicians who are considering metabolic surgery.”
The calculator — 10-year Individualized Diabetes Complications Risk Scores — was developed in two phases over the course of about 2 years, Aminian said. For the first phase, an observational study assessed 2,287 patients with type 2 diabetes and a BMI of at least 30 kg/m² who underwent bariatric surgery between 1998 and 2017 in the Cleveland Clinic Health System. Researchers also included 11,500 propensity-matched patients with similar characteristics who received usual medical care. The phase 1 results, published in JAMA in September and reported by Healio, showed that bariatric surgery for patients with type 2 diabetes and obesity was linked to lower incidence of major adverse CV events compared with nonsurgical methods of metabolic management.
For the second phase of the study, researchers used the same cohort to identify predictors for different health outcomes. Researchers built evidence-based models using random forest machine learning, integrated into a risk calculator, to estimate the likelihood of coronary heart disease, stroke, heart failure, diabetic kidney disease and mortality over 10 years in people with type 2 diabetes and obesity, with and without bariatric surgery. The prediction models were programmed to construct user-friendly, web-based and smartphone applications for clinical use.
The prediction tools demonstrated the following discrimination ability based on the area under the receiver operating characteristic curve (1 = perfect discrimination; 0.5 = chance) at 10-years in the surgical and nonsurgical groups, respectively: all-cause mortality (0.79 and 0.81), coronary artery event (0.66 and 0.67), heart failure (0.73 and 0.75) and nephropathy (0.73 and 0.76).
The calculator can be clinically useful for primary care physicians, obesity medicine specialists, endocrinologists, cardiologists, bariatric surgeons and people with type 2 diabetes and obesity to assess the risks and benefits of bariatric surgery, Aminian said.
“The calculator uses information that is readily available in clinical practice and will be useful in decision-making,” Aminian said.
The calculator will be accessible on the Cleveland Clinic Risk Calculator Library website and as a smartphone app (BariatricCalc). A second version of the app with additional calculators will be released during ObesityWeek 2019. – by Regina Schaffer
Aminian A, et al. Abstract #A101. Presented at: ObesityWeek 2019; Nov. 3-7, 2019; Las Vegas.
Disclosures: Medtronic funded this study but had no role in the design, conduct of the study or reporting of the results. Aminian reports no relevant financial disclosures.