External Validation of SELECT Trial-Derived Cardiovascular Risk Equations in a UK Population with Overweight or Obesity and Established Cardiovascular Disease Without Diabetes

BACKGROUND: Health economic models, many of which rely on risk equations, often play an important role in informing local healthcare decision making for the management of obesity. Risk equations for predicting cardiovascular outcomes have previously been derived from SELECT trial data. Validation of risk equations in real-world populations can provide confidence in the accuracy and reliability of health economic models. OBJECTIVE: The aim of this study was to validate the predicted outcomes of the novel, SELECT trial-derived cardiovascular risk equations in a real-world UK population with overweight or obesity and established CVD without diabetes using the Clinical Practice Research Datalink (CPRD) database. METHODS: and established CVD without diabetes. The risk equations were validated both in their original form, and after re-calibration to the CPRD analysis population. Discrimination and calibration were evaluated to assess the predictive performance of the risk equations. The SELECT risk equations, without recalibration, were also compared with published risk equations in terms of their discrimination and calibration. RESULTS: The discrimination C-indices at 4 years were 0.65 (95% confidence interval [CI] 0.64-0.65) for ACS and 0.71 (95% CI 0.71-0.71) for stroke. After re-calibration, at 4 years calibration, observed/expected ratio was 1.02 (95% CI 1.00-1.03) for ACS and 0.98 (95% CI 0.96-1.00) for stroke. Prediction accuracy was lower for those with a higher baseline risk of stroke, with over-prediction observed in these patients. For both ACS and stroke, the SELECT risk equations showed better discrimination than Framingham (ACS) (C-index 0.68 vs 0.54) and SMART-REACH (stroke) (C-index 0.74 vs 0.50). CONCLUSION: The recalibrated SELECT risk equations showed acceptable discrimination and good calibration when applied to a real-world population of patients with overweight or obesity and established CVD without diabetes from the CPRD database. These risk equations with calibration also demonstrated better predictive performance compared with published risk equations, supporting their use in health economic evaluations in this patient population.