Protein-Based Risk Score Shows Potential for Tailored Medicine in Cardiology
In the quest for personalized medicine, a new risk score based on proteins that circulate in blood plasma may more accurately predict cardiovascular events for patients with stable coronary heart disease than currently used risk factors.
Risk assessment for heart attack, stroke, heart failure, and cardiac death has improved greatly in recent years, and while tests based on C-reactive protein (CRP) levels and genetic screening have shown advancements, physicians are still unable to specifically stratify patients by phenotype.
Proteins are at “the end of the line,” Merry L. Lindsey, PhD (University of Mississippi Medical Center, Jackson), a proteomics researcher who was not involved in creating the risk score, told TCTMD. “Often changes that we see occurring at the gene level don’t get translated down to the protein product,” she said. “And so in terms of looking at a gene versus looking at a protein, at the end of the day, the protein is where the action is going to be.”
With this in mind, Peter Ganz, MD (University of California, San Francisco), and colleagues used previously unavailable protein sampling and identification technology (SomaLogic; Boulder, CO) to study 1,130 proteins from plasma samples from two distinct patient populations with stable coronary heart disease:
- Derivation cohort: 938 samples from outpatients who were seen at 12 San Francisco VA clinics and enrolled in the Heart and Soul study between 2000 and 2002, with follow-up through 2011
- Validation cohort: 971 samples from subjects in Norway who were enrolled in the HUNT3 study between 2006 and 2008, with follow-up through 2012
Of the more than 200 proteins determined to be associated with cardiac events, the researchers identified a total of nine that together had greater predictive accuracy than variables from the Framingham secondary event risk model including total cholesterol, systolic blood pressure, and smoking status. However, total discriminative accuracy for the nine-protein risk score was “only modest,” they reported.
In secondary analyses of paired samples, the nine-protein model was more accurate than the Framingham model in 139 patients who had cardiovascular events (P = 0.002) than in 375 who did not (P = 0.30).
Hence, future studies should focus on the protein score’s accuracy in a lower-risk population, Ganz and colleagues write.
Their research was published in the June 21, 2016 issue of JAMA.
This research is “state-of-the-art for our field,” Lindsey said, adding that the experimental design is “exceptional” given that both a derivation and a validation cohort were used. “Within their own study, they have built-in reproducibility,” and the paired strategy of using plasma samples from different time points “also gives strength because it helps to eliminate potential false positives and false negatives that we often see,” she explained.
One point of concern, however, is that more than 80% of the patients were male. “Cardiovascular disease is the number one cause of death in women in the US, and so being able to see how this translates across genders is something that is obviously a next step before trying to bring it out into a wide scale clinical assessment,” Lindsey said.
The issue of causality also hasn’t been established, she commented. “This just tells you that these are nine proteins that are associated with events—some of them are positive and some of them are negative,” Lindsey said. “It doesn’t necessarily tell you that of these nine proteins, which ones contribute to the disease process and which ones are more responders of the disease process, and so it’s possible within the body that some of these responder proteins are just natural protection mechanisms.”
Protein-Related Mechanisms Unclear
“We always talk about personalized medicine in cardiology, or in medicine in general, but we really until now haven’t had the tools to personalize medicine,” Ganz told TCTMD.
One of the reasons why genomic screening may not have achieved the success it was expected to in cardiology care is because “coronary heart disease is only partly genetic and much of it is environmental,” Ganz explained. “We believe that proteins actually capture both the genetic and environmental components. . . . Your genes won’t change during your lifetime regardless of what environmental factors you’re exposed to.”
He acknowledged the high likelihood that some of the nine proteins are “not the mediators of disease,” as Lindsey had suggested, but rather are markers of “end-organ damage.” However, one of the nine identified was troponin I—a biomarker many cardiologists are already using regularly to measure myocardial injury. The fact that this particular protein was independently reestablished in this study adds strength to the findings, Ganz said.
Still, Lindsey said that it will be important to go “back to the bench” and establish the mechanisms for each of the nine proteins’ influence on heart disease in order to gather as much information as possible. “That will help to strengthen whether or not all nine of them are really as informative as we think or whether we can even distill it down to fewer proteins than that,” she said.
Additionally, she called for future studies retroactively looking at samples of only patients who have cardiac events, specifically post-MI patients who go onto heart failure. “Heart failure is the next area of research we need to conquer,” Lindsey explained. “Really being able to know which patients are going to get heart failure and which are not is something that has a huge clinical [and] economic burden.”
Looking forward, Ganz said he has plans to study the risk score in a more diverse patient population, both with regard to gender and minority status. He is also excited about the possibility of using a protein score to predict whether patients will respond to various treatments.
SomaLogic is expected to commercially release a version of the protein score this fall, Ganz said, but whether it will include all nine or potentially only seven proteins remains to be seen. With regard to price, “it wouldn't necessarily be high,” he predicted, because clinicians will be using simpler microarray technology as opposed to the more expensive technology used in the study.
- CRP Can Identify Low CV Risk in Metabolically ‘Healthy’ Obese Patients
- High-Sensitivity CRP Predicts Long-term Mortality in Patients Receiving DES
- PROSPECT Substudy: Persistently High CRP Predicts Future MACE in ACS Patients After PCI
Ganz P, Heidecker B, Hveem H, et al. Development and validation of a protein-based risk score for cardiovascular outcomes among patients with stable coronary heart disease. JAMA. 2016;315:2532-2541.
- Ganz and Lindsey report no relevant conflicts of interest.