In a comparison with Framingham-based models, the Reynolds Risk Score proved to be the more accurate predictor of risk of cardiovascular disease (CVD) in women in an independent study using a large, multi-ethnic, prospective cohort. Results from the paper published online March 7 in Circulation have clinical implications, according to the authors of an accompanying editorial.

The American College of Cardiology and the American Heart Association designated both the Framingham Adult Treatment Panel III (ATP-III) coronary heart disease (CHD) risk prediction model and the Reynolds Risk Score as Class I recommendations in the 2010 guidelines for assessment of cardiovascular risk in asymptomatic adults (Circulation 2010;122:e584-e636). But the scoring systems have not been compared in an independent cohort that was not used to derive the score, and the models were developed and validated based on a primarily white population, Nancy R. Cook, ScD, of Brigham and Women’s Hospital and Harvard Medical School, both in Boston, and colleagues wrote.

Cook and colleagues designed their study to examine the clinical performance of two Framingham models and the Reynolds score using the Women’s Health Initiative Observational Study (WHI-OS). The validation cohort offered a racially and ethnically diverse patient population of more than 90,000 postmenopausal women in the U.S. For the analysis, they selected a sample of 1,722 cases of major CVD and a random sub-cohort of 1,994 women without prior CVD. The major CVD group included 752 MIs, 754 ischemic strokes and 216 other CVD deaths.

The researchers estimated the risk of the ATP-III score, the Reynolds Risk Score and the Framingham CVD model using weighting methods in a complex analytical approach. They found large differences in the predicted 10-year risk estimates between models; there was a 10 percent or higher risk in 6 percent, 10 percent and 41 percent of women using the ATP-III, Reynolds and Framingham CVD models, respectively. They demonstrated that the ATP-III and Framingham CVD models over-estimated the risk for CHD and major CVD, respectively, while the Reynolds model had the best agreement.

The researchers also recalibrated the models, which were designed to predict differing outcomes, to have the same CVD end point. The recalibrated analyses showed that the Framingham CVD score was not superior to the ATP-III score and neither discriminated as well as the Reynolds score.

“The number of women potentially eligible for statin therapy can vary greatly depending on the equation and end point used,” the authors wrote. They noted these are important criteria for developing clinical guidelines.

“Using data from the WHI-OS, among women with 10-year ATP-III risks of 5 to 10 percent, the Reynolds score would reclassify 15 percent to a lower risk category (less than 5 percent) and over 28 percent to a higher risk category (more than 10 percent), including 5 percent with estimated risk exceeding 20 percent.” They added that many of these women would be at risk of MI or stroke, and based on clinical trial data they may be candidates for statin therapy.

The authors concluded that the findings have clinical implications for CVD prevention, a view shared by editorial writers Erin D. Michos, MD, MPH, and Roger S. Blumenthal, MD. Michos and Blumenthal, both of the Johns Hopkins Ciccarone Center for Prevention of Heart Disease in Baltimore, pointed out the CVD is the leading cause of death in women in the U.S., with the majority being asymptomatic when they die of CVD.

“Therefore, it is of great importance to identify ‘at-risk’ women early, such that effective prevention strategies can be instituted,” they wrote. They added that the analyses by Cook and colleagues validated the existing models; showed that the Reynolds model had improved discrimination compared with the other models; that the improvement was consistent for white and black women alike; and that if applied in practice could have a significant impact.

“This finding has important clinical implications that would change LDL-cholesterol (LDL-C) targets for initiation of pharmacotherapy as well as alter LDL-C treatment goals for a substantial number of women,” Michos and Blumenthal wrote. “One hopes that this improved classification of risk and resulting modification of our treatment strategies will lead to improved cardiovascular outcomes; however, we do not have outcome data yet for this approach.”

They suggested that all three risk prediction models may be too complex for clinical use, and that many questions remain. “Clearly the time has come for improved risk stratification to better identify those who will most and least benefit from intensified primary prevention,” they proposed.