Abstract P236: Improving Cardiovascular Risk Prediction in Type II Diabetes Patients with Repeated Measurements of Clinical Biomarkers
Background: Patients with type 2 diabetes (DM2) are at increased risk of cardiovascular disease (CVD) and treated for risk factors such as hypercholesterolemia and hypertension. HbA1c, cholesterol, blood pressure are monitored at regular time intervals, eg quarterly or annually. It is unknown whether these repeated measurements improve CVD risk prediction in DM2 patients.
Methods: We simulated a cohort of 2,000 patients based on existing data, in which we modelled treatment of diabetes and risk factors according to Dutch guidelines. All subjects were treated with a sulfonylurea-derivative, and subjects with increased baseline CVD risk were treated with simvastatin and/or antihypertensive medication. Reductions in HbA1c, blood lipids, blood pressure and CVD risk were modelled according to literature (timepoint 1). If target values were not reached, patients were treated with insulin, atorvastatin, and a second antihypertensive agent (timepoint 2). In the simulated dataset CVD risk prediction at baseline was compared with 3 approaches to include repeated measurements: 1) use latest values of clinical biomarkers in existing risk score;2) remodel risk score using latest values;3) remodel risk score using baseline plus latest values. The approaches were compared by discrimination and calibration.
Results: Table 1 shows the performance of the approaches at both time points. Calibration was acceptable in all settings except approach 1 at time 2 (p = 0.03). C statistics showed small differences between approaches; it slightly diminished with approach 1, and improved when fitting a new model with both the repeated and baseline value. Other treatment scenarios yielded comparable results.
Conclusion: Differences in C statistics between the approaches are comparable in order of magnitude with differences between entire risk scores or added value of biomarkers in this clinical domain. Thus, although concrete recommendations cannot be made yet, including repeated measures in CVD prediction for DM2 patients may be promising.
Author Disclosures: L.M. Peelen: None. L. Nagtzaam: None. J.W.J. Beulens: None. S. Van Dieren: None. K.G.M. Moons: None. Y.T. Van der Schouw: None.
- © 2015 by American Heart Association, Inc.