乌鸦传媒

Most clinical assessment tools for predicting atherosclerotic cardiovascular disease failed a "crucial test of external validation."

Only the Reynolds Risk Score didn't uniformly overestimate risk in validation in the community-based Multiethnic Study of Atherosclerosis (MESA), of the University of Louisville, Ky., and colleagues reported online in the Annals of Internal Medicine.

But whereas that calculator had "superior calibration and equal discrimination compared with the other scores" (overestimating men's risk by only 9%), it underestimated risk by 21% in women.

The controversial 2013 American Heart Association/American College of Cardiology risk predictor overshot by 86% in men and by 67% in women.

Those at the recommended threshold for statin initiation -- with a predicted risk score of 7.5% to 10% with the AHA/ACC tool -- had an actual rate of heart attacks and strokes of only 3% in men and 5% in women.

Three older Framingham-based risk scores overestimated events by 37% to 154% in men and 8% to 67% in women.

Those findings from calculations on baseline factors appeared to hold for low- and high-risk groups alike after 10 years of follow-up.

"Accurate estimation of absolute risk underpins the effort by the AHA and ACC to update the prevention guidelines and is required to effectively balance therapeutic risk and benefit of an intervention for an individual patient or entire population," the group concluded.

"Physicians treating patients similar to those in MESA may consider interpreting the absolute risk generated by the new risk score with caution."

Those criticisms aren't new.

Right from introduction of the AHA/ACC guideline in 2013, its risk-assessment tool came under fire as inaccurate and potentially leading too many people to statin treatment.

, and , of Brigham and Woman's Hospital in Boston, quickly released an analysis of three large primary prevention cohorts showing that the tool .

The organizations stood by the calculator (which added consideration of stroke compared with prior iterations), arguing that those cohorts were not representative of the overall U.S. population.

But the guidelines themselves acknowledged that the tool overestimated risk in MESA and a second cohort against which it had been validated. None of the five risk assessment tools in DeFilippis' study had used MESA for development.

As in other analyses, MESA didn't show that medication use or interim revascularizations were good explanations for why risk is so overestimated by the calculators.

In a new editorial to go with the MESA findings, Ridker and Cook suggested that one probable explanation was that "the new [AHA/ACC] calculator was developed using data from cohorts that were decades old."

The acknowledged the guideline developers' emphasis that the calculator score was to help in the risk conversation, not to mandate pharmaceuticals, but offered some suggestions on what to do clinically in the face of consistent evidence of overestimation of risk:

"Clinicians might re-calibrate the algorithm so that it tracks more closely with contemporary evidence," they wrote, "simultaneously calculate multiple risk algorithms as currently done in some Mayo Clinic prevention programs, or elect to ignore the problem and accept that more persons will be treated with a class of drugs proven to reduce vascular event rates."

Comment