Protein-based CHD risk score developed

By Eleanor McDermid

Researchers have screened over 1000 plasma proteins to develop a predictive score in patients with stable coronary heart disease (CHD).

As reported in JAMA, the researchers screened plasma samples from 938 participants of the Heart and Soul study using aptamers, which are made of modified DNA and bind specific proteins, so they essentially function as synthetic antibodies.

From this, Peter Ganz (University of California-San Francisco, USA) and team identified nine proteins that were prognostic for myocardial infarction, stroke, heart failure and all-cause death over the 4 years after a sample was taken.

The nine proteins were angiopoietin-2, matrix metalloproteinase-12, chemokine ligand 18, complement 7, α1-antichymotrypsin complex, angiopoietin-related protein 4, troponin I, growth differentiation factor 11/8 and α2-antiplasmin.

When the researchers adjusted for the variables in the Framingham risk prediction model (refit in the Heart and Soul cohort), the nine-protein model remained predictive of cardiovascular events, "suggesting that the 9 proteins contained prognostic information that was at least partly independent of traditional risk factors."

However, the nine-protein model had only modest discriminative accuracy, with a C-statistic of 0.74 (where 1.0 is perfect discrimination) in the derivation cohort and 0.70 in a validation cohort of 971 participants of the HUNT3 study.

Discriminative accuracy was greater than for the refit Framingham model alone, but combining the two models improved the C-statistic only slightly, to 0.75 and 0.71 in the derivation and validation cohorts, respectively.

However, editorialist Marc Sabatine (Brigham and Women's Hospital, Boston, Massachusetts, USA) suggests that "this modest accuracy should not necessarily dampen enthusiasm for their results."

Although a diagnostic test demands high accuracy, the usefulness of a predictive test is based more on its ability to categorise patients according to their risk level, he says.

The 4-year event rate predicted by the protein score was within 2 percentage points of the observed rate in the validation cohort. And when participants were divided into deciles according to protein risk score, the expected and observed event rates for each decile were within 5 percentage points of each other. Expected rates rose from 7.8% in the first to 59.2% in the 10th decile.

"These data would be important to physicians and patients alike", says Sabatine, but he suggests that the score be tested on samples from patients who underwent active treatment in a clinical trial.

"Although more accurate risk prediction is always welcome, clinicians more readily embrace measuring a prognostic biomarker or calculating a risk score if the results could alter therapeutic decision making."

Licensed from medwireNews with permission from Springer Healthcare Ltd. ©Springer Healthcare Ltd. All rights reserved. Neither of these parties endorse or recommend any commercial products, services, or equipment.