BACKGROUND: Cardiovascular disease (CVD) risk-prediction algorithms are key in determining one's eligibility for prevention strategies, but are often population-specific. Metabolic syndrome (MetS), a clustering of risk factors that increase the risk of CVD, does not currently have a risk-prediction algorithm available for prediction of CVD. The aim of this study was to compare the predictive capacities of an algorithm intended for 'healthy' individuals and one intended for 'diabetic' individuals. METHODS: Individual-specific data from 2700 subjects defined as MetS but free of diagnosed CVD from the Australian Diabetes, Obesity and Lifestyle study was used to estimate 5-year risk of CVD using the two algorithms, and compared using Wilcoxon-signed rank test. CVD end point data was used to assess the performance using discrimination and calibration techniques of the two algorithms. RESULTS: Five-year risk-prediction comparisons demonstrated that the UKPDS algorithm overpredicted risk in the younger age groups (25-54 years) and underpredicted risk in the older age groups (≥55 years) compared to the Framingham algorithm. A total of 133 CVD events occurred over a median follow up of 5.0 years. Model performance analyses demonstrated both the Framingham and UKPDS algorithms were poor at discrimination (area under receiver operator curve 0.513 and 0.524, respectively) and calibration (Hosmer-Lemeshow 467.1 and 297.0, respectively). CONCLUSIONS: Neither the Framingham or UKPDS algorithms are ideal for prediction of CVD risk in a MetS population. This study highlights the need for development of population-specific risk-prediction algorithms for this growing population group.
BACKGROUND:Cardiovascular disease (CVD) risk-prediction algorithms are key in determining one's eligibility for prevention strategies, but are often population-specific. Metabolic syndrome (MetS), a clustering of risk factors that increase the risk of CVD, does not currently have a risk-prediction algorithm available for prediction of CVD. The aim of this study was to compare the predictive capacities of an algorithm intended for 'healthy' individuals and one intended for 'diabetic' individuals. METHODS: Individual-specific data from 2700 subjects defined as MetS but free of diagnosed CVD from the Australian Diabetes, Obesity and Lifestyle study was used to estimate 5-year risk of CVD using the two algorithms, and compared using Wilcoxon-signed rank test. CVD end point data was used to assess the performance using discrimination and calibration techniques of the two algorithms. RESULTS: Five-year risk-prediction comparisons demonstrated that the UKPDS algorithm overpredicted risk in the younger age groups (25-54 years) and underpredicted risk in the older age groups (≥55 years) compared to the Framingham algorithm. A total of 133 CVD events occurred over a median follow up of 5.0 years. Model performance analyses demonstrated both the Framingham and UKPDS algorithms were poor at discrimination (area under receiver operator curve 0.513 and 0.524, respectively) and calibration (Hosmer-Lemeshow 467.1 and 297.0, respectively). CONCLUSIONS: Neither the Framingham or UKPDS algorithms are ideal for prediction of CVD risk in a MetS population. This study highlights the need for development of population-specific risk-prediction algorithms for this growing population group.
Authors: Bohdan B Khomtchouk; Diem-Trang Tran; Kasra A Vand; Matthew Might; Or Gozani; Themistocles L Assimes Journal: Brief Bioinform Date: 2020-12-01 Impact factor: 11.622
Authors: Sang Youl Rhee; So Young Park; Jin Kyung Hwang; Jung Il Son; Sang Ouk Chin; Young Seol Kim; Jeong-Taek Woo Journal: Diabetol Metab Syndr Date: 2014-09-12 Impact factor: 3.320
Authors: Anselm K Gitt; Roland E Schmieder; Eva Duetting; Peter Bramlage; Steffen Schneider; Diethelm Tschöpe Journal: Cardiovasc Diabetol Date: 2012-12-05 Impact factor: 9.951