AIMS: To determine the prognostic value of the Framingham equation and the United Kingdom Prospective Diabetes Study (UKPDS) risk engine in patients with newly diagnosed Type 2 diabetes. METHODS: A community-based cohort (n=428; aged 30-74 years) free of clinically evident CVD and newly diagnosed with Type 2 diabetes were studied over a median 4.2 (sd+/-0.62) years. Predicted (using baseline variables at diagnosis) and observed proportions of primary CVD and CHD events were compared using the Framingham equations and the UKPDS risk engine (only CHD events). The discrimination (c-statistic) and calibration (HLchi2) of the risk equations were calculated. The sensitivity and specificity of the Framingham equation at a 15%, 10-year CHD risk threshold (NICE guidelines) was compared with that of the ADA lipid threshold (LDLc>or=2.6 mmol/l or triglycerides>or=4.5 mmol/l). RESULTS: The Framingham equations underestimated the overall number of cardiovascular events by 33% and coronary events by 32% and showed modest discrimination and poor calibration for CVD [c=0.673; HLchi2=32.8 (P<0.001)] and CHD risk [c=0.657; HLchi2=19.8 (P=0.011)]. Although the overall underestimate was lower and non-significant with the UKPDS risk engine for CHD (13%), its performance in terms of discrimination and calibration were similar [c=0.670; HLchi2=17.1 (P=0.029)]. The 15%, 10-year CHD risk threshold with both the Framingham and UKPDS risk engines had similar sensitivity for primary CVD as the lipid level threshold [85.7 and 89.8% vs. 93.9% (P=0.21 and 0.34)] and both had greater specificity [33.0 and 30.3% vs. 12.1% (P<0.001 and P<0.001)]. CONCLUSIONS: In people with newly diagnosed Type 2 diabetes, both the Framingham equation and UKPDS risk engine are moderately effective at identifying those at high-risk (discrimination) and are poor at quantifying risk (calibration). Nonetheless, at a population level, a 15% 10-year CHD risk threshold using either risk calculator has similar sensitivity as an approach based on a single lipid risk factor level and may have benefits in terms of cost-effectiveness given the improved specificity.
AIMS: To determine the prognostic value of the Framingham equation and the United Kingdom Prospective Diabetes Study (UKPDS) risk engine in patients with newly diagnosed Type 2 diabetes. METHODS: A community-based cohort (n=428; aged 30-74 years) free of clinically evident CVD and newly diagnosed with Type 2 diabetes were studied over a median 4.2 (sd+/-0.62) years. Predicted (using baseline variables at diagnosis) and observed proportions of primary CVD and CHD events were compared using the Framingham equations and the UKPDS risk engine (only CHD events). The discrimination (c-statistic) and calibration (HLchi2) of the risk equations were calculated. The sensitivity and specificity of the Framingham equation at a 15%, 10-year CHD risk threshold (NICE guidelines) was compared with that of the ADA lipid threshold (LDLc>or=2.6 mmol/l or triglycerides>or=4.5 mmol/l). RESULTS: The Framingham equations underestimated the overall number of cardiovascular events by 33% and coronary events by 32% and showed modest discrimination and poor calibration for CVD [c=0.673; HLchi2=32.8 (P<0.001)] and CHD risk [c=0.657; HLchi2=19.8 (P=0.011)]. Although the overall underestimate was lower and non-significant with the UKPDS risk engine for CHD (13%), its performance in terms of discrimination and calibration were similar [c=0.670; HLchi2=17.1 (P=0.029)]. The 15%, 10-year CHD risk threshold with both the Framingham and UKPDS risk engines had similar sensitivity for primary CVD as the lipid level threshold [85.7 and 89.8% vs. 93.9% (P=0.21 and 0.34)] and both had greater specificity [33.0 and 30.3% vs. 12.1% (P<0.001 and P<0.001)]. CONCLUSIONS: In people with newly diagnosed Type 2 diabetes, both the Framingham equation and UKPDS risk engine are moderately effective at identifying those at high-risk (discrimination) and are poor at quantifying risk (calibration). Nonetheless, at a population level, a 15% 10-year CHD risk threshold using either risk calculator has similar sensitivity as an approach based on a single lipid risk factor level and may have benefits in terms of cost-effectiveness given the improved specificity.
Authors: X L Yang; W Y So; A P S Kong; P Clarke; C S Ho; C W K Lam; M H L Ng; R R Lyu; D D Yin; C C Chow; C S Cockram; P C Y Tong; J C N Chan Journal: Diabetologia Date: 2006-08-30 Impact factor: 10.122
Authors: Joseph Yeboah; Raimund Erbel; Joseph Chris Delaney; Robin Nance; Mengye Guo; Alain G Bertoni; Matthew Budoff; Susanne Moebus; Karl-Heinz Jöckel; Gregory L Burke; Nathan D Wong; Nils Lehmann; David M Herrington; Stefan Möhlenkamp; Philip Greenland Journal: Atherosclerosis Date: 2014-08-14 Impact factor: 5.162
Authors: Amber A W A van der Heijden; Monica M Ortegon; Louis W Niessen; Giel Nijpels; Jacqueline M Dekker Journal: Diabetes Care Date: 2009-11 Impact factor: 19.112