Framingham-based scores greatly overstate cardiovascular disease risk

Four commonly used scores for predicting cardiovascular risk were shown to overestimate risk by between 37% to 154% in men and 8% to 67% in women, according to a new study. A fifth scoring system was seen overestimating risk in men by 8% while underestimating risk in women by 21%.

The study, published online Feb. 16 in Annals of Internal Medicine (doi:10.7326/M14-1281) adds weight to recent concerns that risk scores using algorithms based on Framingham algorithms, rooted in data from decades-old cohorts, need to be revisited. Of particular concern is excessive prescription of statin drugs based on overestimated risk of cardiovascular events. Dr. Andrew P. DeFilippis of the University of Louisville (Ky.) led the study, which included 4,227 participants in the Multi-Ethnic Study of Atherosclerosis (MESA) who were aged 50-74 without cardiovascular disease or diabetes at baseline and followed for 10.2 years. Dr. DeFilippis and colleagues evaluated subjects’ expected risk according to three older scoring tools based on the Framingham Original and Offspring Study cohorts for different cardiovascular endpoints (FRS-CHD, FRS-CVD, ATPIII-FRS-CHD) and the 2013 American Heart Association/American College of Cardiology tool (AHA/ACC). They also used the Reynolds Risk score, which includes some biomarker measures along with Framingham-derived measures. Of all of these, the Reynolds score predicted risk most accurately for the cohort as a whole, though it underestimated risk in women. Statin use, revascularization procedures, and potential underreporting of events did not explain the differences between expected and recorded events seen in the cohort, Dr. DeFilippis and colleagues found. Rather, the researchers suspected, “changing significance of risk factors between older cohorts that were used to develop these risk scores” and the modern study cohort was the likelier explanation. For example, the researchers noted, smoking has been a “yes or no” variable in the scoring tools, “while number of cigarettes smoked and content of cigarettes have both changed substantially over time.” Dr. DeFilippis and colleagues noted as a limitation of their study the fact that the cohort may have represented a more health conscious, and therefore healthier, group than the general population.

The study was funded by the National Heart, Lung, and Blood Institute. Dr. DeFilippis disclosed receiving honoraria from Roche and AstraZeneca, while his coauthor Dr. Michael Blaha disclosed fees from Pfizer.

Four commonly used scores for predicting cardiovascular risk were shown to overestimate risk by between 37% to 154% in men and 8% to 67% in women, according to a new study. A fifth scoring system was seen overestimating risk in men by 8% while underestimating risk in women by 21%.

The study, published online Feb. 16 in Annals of Internal Medicine (doi:10.7326/M14-1281) adds weight to recent concerns that risk scores using algorithms based on Framingham algorithms, rooted in data from decades-old cohorts, need to be revisited. Of particular concern is excessive prescription of statin drugs based on overestimated risk of cardiovascular events. Dr. Andrew P. DeFilippis of the University of Louisville (Ky.) led the study, which included 4,227 participants in the Multi-Ethnic Study of Atherosclerosis (MESA) who were aged 50-74 without cardiovascular disease or diabetes at baseline and followed for 10.2 years. Dr. DeFilippis and colleagues evaluated subjects’ expected risk according to three older scoring tools based on the Framingham Original and Offspring Study cohorts for different cardiovascular endpoints (FRS-CHD, FRS-CVD, ATPIII-FRS-CHD) and the 2013 American Heart Association/American College of Cardiology tool (AHA/ACC). They also used the Reynolds Risk score, which includes some biomarker measures along with Framingham-derived measures. Of all of these, the Reynolds score predicted risk most accurately for the cohort as a whole, though it underestimated risk in women. Statin use, revascularization procedures, and potential underreporting of events did not explain the differences between expected and recorded events seen in the cohort, Dr. DeFilippis and colleagues found. Rather, the researchers suspected, “changing significance of risk factors between older cohorts that were used to develop these risk scores” and the modern study cohort was the likelier explanation. For example, the researchers noted, smoking has been a “yes or no” variable in the scoring tools, “while number of cigarettes smoked and content of cigarettes have both changed substantially over time.” Dr. DeFilippis and colleagues noted as a limitation of their study the fact that the cohort may have represented a more health conscious, and therefore healthier, group than the general population.

The study was funded by the National Heart, Lung, and Blood Institute. Dr. DeFilippis disclosed receiving honoraria from Roche and AstraZeneca, while his coauthor Dr. Michael Blaha disclosed fees from Pfizer.

Four commonly used scores for predicting cardiovascular risk were shown to overestimate risk by between 37% to 154% in men and 8% to 67% in women, according to a new study. A fifth scoring system was seen overestimating risk in men by 8% while underestimating risk in women by 21%.

The study, published online Feb. 16 in Annals of Internal Medicine (doi:10.7326/M14-1281) adds weight to recent concerns that risk scores using algorithms based on Framingham algorithms, rooted in data from decades-old cohorts, need to be revisited. Of particular concern is excessive prescription of statin drugs based on overestimated risk of cardiovascular events. Dr. Andrew P. DeFilippis of the University of Louisville (Ky.) led the study, which included 4,227 participants in the Multi-Ethnic Study of Atherosclerosis (MESA) who were aged 50-74 without cardiovascular disease or diabetes at baseline and followed for 10.2 years. Dr. DeFilippis and colleagues evaluated subjects’ expected risk according to three older scoring tools based on the Framingham Original and Offspring Study cohorts for different cardiovascular endpoints (FRS-CHD, FRS-CVD, ATPIII-FRS-CHD) and the 2013 American Heart Association/American College of Cardiology tool (AHA/ACC). They also used the Reynolds Risk score, which includes some biomarker measures along with Framingham-derived measures. Of all of these, the Reynolds score predicted risk most accurately for the cohort as a whole, though it underestimated risk in women. Statin use, revascularization procedures, and potential underreporting of events did not explain the differences between expected and recorded events seen in the cohort, Dr. DeFilippis and colleagues found. Rather, the researchers suspected, “changing significance of risk factors between older cohorts that were used to develop these risk scores” and the modern study cohort was the likelier explanation. For example, the researchers noted, smoking has been a “yes or no” variable in the scoring tools, “while number of cigarettes smoked and content of cigarettes have both changed substantially over time.” Dr. DeFilippis and colleagues noted as a limitation of their study the fact that the cohort may have represented a more health conscious, and therefore healthier, group than the general population.

The study was funded by the National Heart, Lung, and Blood Institute. Dr. DeFilippis disclosed receiving honoraria from Roche and AstraZeneca, while his coauthor Dr. Michael Blaha disclosed fees from Pfizer.