Updating the pooled cohort equations (PCEs) would improve the accuracy of atherosclerotic cardiovascular disease (ASCVD) risk estimates, especially for black adults, researchers found.
Predictions based on the 2013 risk calculator overestimated 10-year risk for atherosclerotic CVD (ASCVD) by an average of 20% across race and sex subgroups, compared with revised equations generated from newer data and statistical methods to assess for the same risks – nonfatal myocardial infarction, death from coronary heart disease, and fatal or nonfatal stroke.
The greatest discordance due to misestimation was noted among black adults, among whom 33% had 2013 equation risk estimates less than 70% or greater than 250% those of white adults with otherwise-identical risk factor values, reported Sanjay Basu, MD, PhD, of Stanford University in Stanford, California, and colleagues.
National CVD prevention guidelines recommend that clinicians inform decisions regarding aspirin, blood pressure, and statins treatment using PCEs’ 10-year atherosclerotic CVD risk estimates – derived in 2013 from five cohort studies. But the issue of risk overestimation arose immediately and has led to debate about their applicability.
“These findings have important clinical implications. Using the 10-year risk estimate of 7.5% of greater, approximately 11.8 million U.S. adults previously labeled high-risk by the 2013 PCEs would be relabeled lower-risk by the updated equations,” thus substantially reducing the number of U.S. adults recommended for statin therapy, the researchers noted.
At a 10-year risk less than 7.5%, the revised PCEs correctly reclassified 13 persons who did not have an atherosclerotic CVD event as low risk for each 1 person who did have such an event but was reclassified as low risk.
Defining a low 10-year risk more conservatively as less than 5%, the revised PCEs resulted in a reclassification ratio of 23:1, with 11.7 million U.S. adults no longer defined as high risk. Using a more liberal 10-year risk estimate of less than 10%, the reclassification ratio was 8:1.
The study used individual participant data from six longitudinal cohort studies, using the same eligibility criteria as the original 2013 PCEs, for a total of 26,689 white or black individuals ages 40 to 79 with no history of cardiovascular disease.
To increase generalization and limit overfitting of the model to the derivation cohort (a potential contributor to erroneous estimates for smaller subpopulations), the group used elastic net regularization for variable selection and model estimation. They also used modified logistic regression to replace the Cox proportional hazard models used in the 2013 PCEs.
Compared to the 2013 PCEs, applying the same derivation method to updated cohort data modestly improved discrimination but not calibration, while using both updated data and new derivation methods improved both calibration and discrimination.
In an accompanying editorial, Andrew Paul DeFilippis, MD, MSc and Patrick Trainor, MS, MA, both of University of Louisville, Kentucky, wrote that while “the novel approach has narrowed the black-white discordance in risk, the fundamental question remains: Does race change how the risk factors accounted for in the PCEs affect the propensity to have an ASCVD event?”
While such a difference was not evident in their prior work using MESA (Multi-Ethnic Study of Atherosclerosis), wrote DeFilippis and Trainor, given the increasing numbers of Hispanic and Asian Americans in the U.S., “we clearly need an accurate risk assessment tool for these growing American populations.”
The most notable limitation of the study was that results are from internal cross validation and prospective hold-out validation, editorialists wrote. Limitations noted by authors included the potential to incorrectly convert some people from high- to low-risk status due to use of cohorts with lower CVD rates; overestimation of the number considered high-risk due to inclusion of older cohorts; and as with the 2013 PCEs, the exclusion of heart failure or coronary revascularization from CVD outcomes to avoid biases due to ascertainment, definition, and physician recommendations.
Study funded primarily by National Institutes of Health.
Basu and the editorialists disclosed no conflicts of interest.
Robert Jasmer, MD Associate Clinical Professor of Medicine, University of California, San Francisco and Dorothy Caputo, MA, BSN, RN, Nurse Planner