Level 2: Mid-level evidence
A systematic review evaluated prediction of coronary or cardiovascular disease (CVD) risk with Framingham risk scores (Anderson or Wilson methods) in 27 studies with 71,727 participants (Heart. 2006;92:1752-1759). For the outcome of coronary heart disease (CHD), evaluated in 20 studies, the predicted-to-observed ratios ranged from 0.43 (underprediction) in people with a family history of CHD to 2.87 (overprediction) in low-risk women from Münster, Germany. Underprediction occurred in six studies with a 16%-48% observed 10-year risk and was found in persons with diabetes, a family history of premature CHD, and a higher-risk primary-care British population. Accurate prediction occurred in two studies with a 15%-16% observed 10-year risk.
Overprediction occurred in 12 studies with a 1%-12% observed 10-year risk.In seven studies that evaluated CVD, Framingham risk prediction was more accurate. Underprediction occurred in one study with a 14% observed 10-year risk and two studies with a 52%-53% observed 10-year risk. Overprediction occurred in one study with a 15% observed 10-year risk, and accurate prediction occurred in three studies with an 8%-14% observed 10-year risk.