Objectives: To summarize the current state of cardiovascular disease (CVD) risk modeling literature with a focus on the U.S. patient population, and to describe evidence on which models best predict cardiovascular risk among patients with diabetes.
Data Sources: We searched MEDLINE for articles published January 1, 1999, to February 24, 2009, and reviewed all reference lists of included articles.
Review Methods: We included studies of asymptomatic adults in any geographic setting with any study design in which a CVD clinical risk prediction model was developed or validated. We excluded studies that 1) were not in English; 2) were without information pertinent to the key questions; 3) had fewer than 200 participants at enrollment; 4) were not original studies; and 5) lacked internal or external validation data. We captured study information such as cohort characteristics, risk model characteristics, model performance statistics, and quality review elements. We collected information about the study populations for stratification of results by variables, including sex and geographic area. We also searched online for available tools and documented their location and the model on which they purported to be based. We used the online tools to calculate risk for five test cases to identify variation in estimated risk.
Results: Of the 3,499 articles initially identified, 84 met inclusion criteria, providing data on 102 risk models. The majority of models (87 out of 102) were not externally validated. The most commonly externally validated risk models were the 1991 Framingham (FRS) model for CVD (26 evaluations), the 1998 FRS model for total coronary heart disease (CHD) (24 evaluations), the FRS Adult Treatment Panel III (ATP-III) model for hard CHD (16 evaluations), the Prospective Cardiovascular Münster (PROCAM) model for hard CHD (11 evaluations), and the Systematic Coronary Risk Evaluation (SCORE) model for CVD mortality (11 evaluations).
Conclusion: The FRS models performed well in U.S. populations, but there were absolute risk prediction problems when they were applied to populations substantially different from the source cohort. Sometimes this was due to particularly low or high baseline risk in the destination cohort, and at other times to systematic differences in risk attributable to specific factors. The 2001 ATP-III version demonstrated better risk prediction than older FRS models because it focuses on hard CHD outcomes, excludes patients with diabetes, and includes newer FRS data. Diabetes-specific process measurement variables are significantly related to cardiovascular outcome risk among patients with diabetes, and risk models that incorporate these factors outperform general risk prediction models when applied to these patients. Models excluding patients with diabetes outperformed general risk prediction models that included these patients in their development when applied to non-diabetic cohorts. Unfortunately, external validation of diabetes-specific risk models is lacking, particularly among U.S. cohorts.