Risk score may help predict future glycemic control after new diagnosis of diabetes

The most important predictors of the trajectory of glucose control were body mass index and levels of HbA1c and triglycerides, according to the study of registry data from the Netherlands.


A new risk score that incorporates individual patient data may help predict future glycemic control in primary care patients with a new diagnosis of diabetes.

Researchers in the Netherlands developed the score by using data from two large registries of patients with diabetes. Patients were included if they had received a new diagnosis of type 2 diabetes between Jan. 1, 2006, and Dec. 31, 2013, for the development cohort and Jan. 1, 2009, and Dec. 31, 2014, for the validation cohort. The development cohort contained 10,258 patients, while the validation cohort included 3,337 patients. Patients in both cohorts were followed from diagnosis of type 2 diabetes until the end of the study period or until censoring due to no further availability of HbA1c measurements. Machine learning models were developed to predict glycemic trajectories over five years that were identified with latent growth mixture modeling. Data used for prediction were patient characteristics that could be obtained easily in daily clinical practice. The study results were published online Nov. 2 by Diabetes, Obesity and Metabolism.

Mean patient age was 62.9 years in the development cohort and 63.7 years in the validation cohort. Slightly over half of the patients in each cohort were men (51.6% and 52.3%, respectively). The development cohort was followed for four years, and the validation cohort was followed for five years. The researchers identified three glycemic trajectories in the derivation cohort: stable, adequate glucose control, seen in 76.5% of patients; improved glycemic control, seen in 21.3% of patients; and deteriorated glycemic control, seen in 2.2% of patients. Trajectories were similar in the validation cohort. The most important predictors of trajectory were body mass index and levels of HbA1c and triglycerides. The receiver-operating characteristic area under the curve for the predictive model was 0.96 in the derivation cohort.

The authors noted that the patterns of the trajectories they identified may have been affected by glucose-lowering drugs and insulin prescriptions and that most of the patient population was white. However, they concluded that body mass index, HbA1c, and triglycerides can be used to accurately predict glycemic response in patients who have received a new diagnosis of type 2 diabetes. “The model can be used in practice as a quick, easy and accurate tool to determine patients' care needs and provide tailored diabetes treatment,” they wrote.