Table 3.
Performance of CVD risk scores evaluated in individuals with diabetes
| Name of risk score | Validation population | Definition of diabetes | Outcome (n events) | Discrimination | Calibration | Sensitivity and specificity (%) | PPV/NPV (%) |
|---|---|---|---|---|---|---|---|
| UKPDS risk engine [47] | 428 men and women aged 30–74 years with newly diagnosed diabetes without pre-existing CVD, mean F/U 4.2 (0.6) years (GP recruitment in Poole, UK) [21] | Surveillance programme to identify diabetes cases; methods of diagnosis, N/A | CHD determined via hospital and GP notes, death certificates and post-mortem examination record (98) | CHD: aROC 0.670 (0.598–0.742) | Poor calibration for CHD: HL χ2 17.1 (p = 0.029) | 10 year CHD risk >15%: Sen 89.8 (95% CI 82.0–95.0) Spc 30.3 (95% CI 25.4–35.6) | N/A |
| 339 diabetic patients with 10 years of follow-up (Greece) [25] | N/A | CHD determined by coronary angiography (108) | aROC 0.61 | N/A | Sen 56; Spc 56 (no data on which risk threshold was used) | PPV 37; NPV 73 (no data on which risk threshold was used) | |
| 112 patients with type 2 diabetes with 10 years of follow-up (Spain) [22] | 1985 WHO criteria (FPG >7.7 mmol/l or, 2 h OGTT >11.0 mmol/l) | CHD event (19) | N/A | Men: overestimation by 108.8%; women: overestimation by 51.3% | 10 year CHD risk >20%: men: Sen 67; Spc 7; women: Sen 71; Spc 59 | 10 year CHD risk >20%: men: PPV 14; NPV 50; women: PPV 24; NPV 92 | |
| 798 diabetic patients identified via a computerised clinic database, follow-up 10 years [19] | N/A | CHD and CVD determined through computerised clinical records (358 CVD, including 269 CHD events) | CVD: c = 0.74; CHD: c = 0.65 | Poor calibration (p < 0.001); observed:predicted events ratio 1.20 and 1.60 for CVD and CHD (r = 0.89 and 0.86 respectively) | N/A | N/A | |
| 3,546 Chinese men and women with median age of 56 years without a history of CHD or HF, median F/U of 5.59 years (Hong Kong) [18] | Type 2 diabetes referred from GP and special clinics, and those discharged from hospitals | Hard CHD: myocardial infarction and CHD death (170) | Overall c statistic 0.610 (95% CI 0.581–0.639) | Poor calibration: HL χ2 = 230.1; p < 0.0001 | N/A | N/A | |
| Oxford risk engine (UKPDS risk engine version 3) [48] | 1,410 men and women aged 40–75 years on the placebo arm of the Collaborative Atorvastatin Diabetes Study (CARDS) (median F/U 3.9 years) [48] | 1985 WHO criteria (computerised registers of patients and by opportunistic assessment of individuals attending diabetes clinics) | CVD event (189) | N/A | Underestimation by 10.6%; 189 observed events vs 169 predicted events over 3.9 years | N/A | N/A |
| Diabetes Audit and Research in Tayside, Scotland (DARTS) [17] | Salford Diabetes Information System, F/U 5 years [17] | Treatment with diet or oral hypoglycaemic agents/age >35 years | CHD determined via hospital episode statistics (N/A) | c statistic = 0.69 (95% CI 0.58–0.78) | Included a graph but no statistical test | N/A | N/A |
| Swedish National Diabetes Register (NDR) [23] | 5,823 men and women aged 18–70 years with diabetes, no previous CVD, F/U 5.6 years (Swedish) [23] | Diagnosis of diabetes from the Swedish National Diabetes Register | Fatal and non-fatal CVD (N/A) | c statistic = 0.69 | Good calibration; observed:predicted CVD rate ratio 0.998 | N/A | N/A |
| 3,068 men and women aged 18–70 years with diabetes, no previous CVD, F/U 4 years (Swedish) [23] | Diagnosis of diabetes from the Swedish National Diabetes Register | Fatal and non-fatal CVD (261) | c statistic = 0.69 | Good calibration; observed:predicted CVD rate ratio 0.96 | N/A | N/A | |
| Hong Kong Diabetes Registry for CHD [18] | 3,546 Chinese men and women with median age of 56 years without CHD or HF, median follow-up of 5.59 years (Hong Kong) [18] | Type 2 diabetes referred from GP and special clinics, and those discharged from hospitals | CHD: myocardial infarction or ischaemic heart disease (170) | Overall aROC 0.704 (95% CI 0.675–0.733) Adjusted aROC 0.737 | Good calibration: HL χ2 = 14.05; p > 0.05 | 5 year CHD risk >5.2%: Sen 67.6; Spc 68.5 | 5 year CHD risk >5.2%: PPV 8.6; NPV 98.0 |
| Framingham 1991 [50] | 3,898 men and women aged 25–65 years with diabetes without history of MI, angina or stroke (UKPDS) [24] | Newly physician-diagnosed diabetes, and FPG >6 mmol/l on two further occasions | Fatal CVD and fatal CHD (CVD: 7.4%, CHD: 6.3%) | aROC 0.76 | 32% underestimation | N/A | N/A |
| 112 patients with type 2 diabetes with 10 years of follow-up (Spain) [22] | 1985 WHO criteria (FPG >7.7 mmol/l or 2 h OGTT >11.0 mmol/l) | CHD event (19) | N/A | Men: overestimation by 64.8%; women: overestimation by 31.6% | 10 year CHD risk >20%: men: Sen 75; Spc 17; women: Sen 71; Spc 67 | 10 year CHD risk >20%: men: PPV 17; NPV 75; women: PPV 28; NPV 93 | |
| 428 men and women aged 30–74 years with newly diagnosed diabetes without pre-existing CVD, mean F/U of 4.2 (0.6) years (GP recruitment in Poole, UK) [21] | Surveillance programme to identify diabetes cases, methods of diagnosis, N/A | CVD and CHD determined via hospital and GP notes, death certificates and post-mortem examination record (98) | CVD: aROC 0.673 (0.612–0.734); CHD: aROC 0.657 (0.581–0.732) | Poor calibration for CVD and CHD CVD HL χ2 = 32.8 (p < 0.001); CHD HL χ2 = 19.8 (p = 0.011) | 10 year CHD risk >15%: Sen 85.7 (95% CI 77.8–99.5); Spc 33.0 (95% CI 30.7–34.7) | N/A | |
| 938 men and women with type 1 and type 2 diabetes, F/U 4 years (Cardiff, UK) [20] | Type 1 and type 2 diabetes identified through the Cardiff Diabetes Database; methods of diagnosis, N/A | CHD via primary or subsidiary diagnosis in hospital record linkage (172) | Men: aROC 0.64; women: aROC 0.66 | N/A | 10 year CHD risk >30%: men: Sen 34%, women: Sen25%; 10 year CHD risk >20%: men: Sen 75%; women: Sen 58% | 10 year CHD risk >30%: men: PPV 26%, NPV 83%; women: PPV 33%, NPV 86%; 10 year CHD risk >20%: men: PPV 24%, NPV 88%; women: PPV 24%, NPV 88% | |
| Framingham 1998 [52] | 112 patients with type 2 diabetes with 10 years of follow-up (Spain) [22] | 1985 WHO criteria (FPG >7.7 mmol/l or 2 h OGTT >11.0 mmol/l) | CHD event (19) | N/A | Men: overestimation by 97.8%; women: overestimation by 38.2% | 10 year CHD risk >20%: men: Sen 75; Spc 11; women: Sen 71; Spc 69 | 10 year CHD risk >20%: men: PPV 16; NPV 67; women: PPV 29; NPV 93 |
| 339 diabetic patients, follow-up of 10 years (Greece) [25] | N/A | CHD determined using coronary angiography (108) | aROC 0.65 | N/A | Sen 55; Spc 65 (no data on which risk threshold was used) | PPV 43; NPV 75 (no data on which risk threshold was used) | |
| PROCAM score [53] | 798 diabetic patients identified via computerised clinic database with 10 years of follow-up (UK) [19] | N/A | CHD and CVD determined through computerised clinical records (358 CVD, including 269 CHD events) | CVD: c = 0.67; CHD: c = 0.76 | Poor calibration (p < 0.001); observed:predicted events ratio 2.79 and 2.05 for CVD and CHD (r = 0.79 and 0.81 respectively) | N/A | N/A |
| Joint British Societies Risk Chart (JBSRC) [54] | 798 diabetic patients identified via computerised clinic database with 10 years of follow-up (UK) [19] | N/A | CHD and CVD determined through computerised clinical records (358 CVD, including 269 CHD events) | CVD: c = 0.80; CHD: c = 0.77 | Poor calibration (p < 0.001) | N/A | At CVD risk >30%: PPV 85%; at CHD risk >30%: PPV 66% |
| CardioRisk Manager calculator (CRM) [55] | 798 diabetic patients identified via a computerised clinic database with 10 years of follow-up (UK) [19] | N/A | CHD and CVD determined through computerised clinical records (358 CVD, including 269 CHD events) | CVD: c = 0.76; CHD: c = 0.73 | Poor calibration (p < 0.001) observed:predicted events ratio 2.30 and 1.74 for CVD and CHD (r = 0.98 and 0.97 respectively) | N/A | N/A |
| Myocardial Infarction Population Registry of Girona (REGICOR) [57] | 112 patients with type 2 diabetes with 10 years of follow-up (Spain) [22] | 1985 WHO criteria (FPG >7.7 mmol/l or 2 h OGTT >11.0 mmol/l) | CHD event (19) | N/A | Men: underestimation by 17.6%; women: underestimation by 34.2% | 10 year CHD risk >20%: men: Sen 25; Spc 67; women: Sen 0; Spc 100 | 10 year CHD risk >20%: men: PPV 14; NPV 80; women: PPV –; NPV 85 |
| SCORE [16] | 3,898 men and women aged 25–65 years with diabetes, without history of MI, angina or stroke within 1 year (UKPDS) [24] | Newly physician-diagnosed diabetes and FPG >6 mmol/l on two further occasions | Fatal CVD and fatal CHD (CVD: 7.4%, CHD: 6.3%) | aROC 0.77 | 18% overestimation | N/A | N/A |
| DECODE score [58] | 3,898 men and women aged 25–65 years with diabetes, without history of MI, angina or stroke within 1 year (UKPDS) [24] | Newly physician-diagnosed diabetes and FPG >6 mmol/l on two further occasions | Fatal CVD (CVD: 7.4%, CHD: 6.3%) | aROC 0.67 | 11% underestimation | N/A | N/A |
| UKPDS for stroke [49] | 1,370 patients aged >30 years of the Wisconsin Epidemiologic Study of Diabetic Retinopathy cohort, follow-up 8.3 years (USA) [49] | N/A | Fatal stroke (197) | N/A | Underestimation of fatal stroke by 12.5% | N/A | N/A |
| 3,541 Chinese diabetic patients without history of stroke (Hong Kong) [26] | Physician-diagnosed diabetes (GP and special clinics, and those discharged from hospital) | Stroke: hospital discharge diagnosis confirmed by physicians (182) | Unadjusted aROC 0.588 (95% CI 0.549–0.626) | N/A | N/A | N/A | |
| Hong Kong Diabetes Registry for stroke [26] | 3,541 Chinese diabetic patients without history of stroke, median follow-up of 5.37 years (Hong Kong) [26] | Physician-diagnosed diabetes referred from GP and special clinics, and those discharged from hospital | Stroke: hospital discharge diagnosis confirmed by physicians (182) | Adjusted aROC 0.770 for haemorrhagic stroke; 0.785 for ischaemic stroke | N/A | 5 year stroke risk >6.1%: Sen 65.7; Spc 74.9 | 5 year stroke risk >6.1%: PPV 11.6 |
| Framingham stroke risk [59] | 178 men and women with diabetes, with or without metabolic syndrome (Spain) [60] | FPG and OGTT based on WHO criteria | Stroke (9) | N/A | Predicted 10 year risk varied from 17.3 to 19.1%; observed 5 year cumulative incidence varied from 3.5 to 5.8 | N/A | N/A |
F/U, follow-up; HF, heart failure; HL χ2, Hosmer–Lemeshow χ2; MI, myocardial infarction; N/A, not available; NPV, negative predictive value; PPV, positive predictive value; Sen, sensitivity; Spc, specificity