Invited commentary: Clinical usefulness of the Framingham cardiovascular risk profile beyond its statistical performance.

Framingham risk functions (FRFs) have been developed for over 50 years. There have been numerous applications of them and within the last few decades they have been used in drug treatment guidelines. The Adult Treatment Panel III explicitly used a coronary heart disease FRF in their guidelines for cholesterol drug treatment. Evaluation of these functions has traditionally involved discrimination and calibration measures. One major goal of the FRFs is to see if they are valid in non-Framingham settings. In this issue of the Journal, Khalili et al. (Am J Epidemiol. 2012;176(3):177-186) apply a recent global cardiovascular disease FRF to the data from their Tehran Lipid and Glucose Study and demonstrate that the cardiovascular disease FRF performs extremely well: as good as the best risk function generated from the Tehrani data. The FRF is transportable to the Tehrani data without need for any calibration adjustment. The investigators then move beyond the traditional discrimination and calibration evaluations, look for utility, and apply the decision theory concepts of net benefit fraction. This application makes assumptions about treatment guidelines. There are both useful and negative aspects of this application, and caution is advised against a too enthusiastic acceptance of it to evaluate prediction rules for primary prevention of cardiovascular disease.

Duke Scholars

Author Michael J Pencina Biostatistics & Bioinformatics, Division of Biostatistics