Predicting heart disease with markers is risky undertaking

W. Gregory Feero, MD, PhD
W. Gregory Feero, MD, PhD

Age, sex, BMI, smoking status, hypertension, diabetes, family history and cholesterol levels are the heavy hitters of risk assessment for coronary heart disease (CHD). All are firmly embedded in the minds of clinicians seeing patients age 40 and older. Whether or not we use formal scoring systems, like the Framingham risk score, a sort of mental calculus goes on involving these parameters when we consider prevention and treatment options for cardiovascular disease.

Combined, these risk factors explain a considerable amount of population risk, while in isolation each is only a weak predictor of risk. At the individual patient level, risk prediction is a tricky business. We all have the patient with the lifestyle of a saint that is a vasculopath at age 45 and the sinner that visits our clinic once a year at age 80.

Art 2

Recently, additional biomarkers like homocysteine levels and C-reactive protein (CRP) have been put forth to improve individual risk prediction. The debut of each new risk marker has been met with considerable controversy and years of scientific wrangling can follow in the attempt to come up with a clear picture of the value each adds to clinical care. The current debate around the value of CRP for CHD risk assessment could be considered a warm-up for things to come.

Genome-wide association studies have identified at least 22 new genetic markers for CHD risk. These newly identified CHD markers provide novel insights on pathways and mechanisms for cardiovascular disease pathogenesis and will prove invaluable over the long haul in developing new approaches to prevention and treatment (Think “the next statin”). Many of these associations are related to lipid profiles and other known (and already measurable) risk factors. Ultimately, some may prove useful in clinical care, since a person's DNA sequence isn't much affected by transient factors like diet or the presence of a cold.

More intriguingly, at least some of the genetic markers seem to be independent of all currently measurable clinical markers of cardiovascular disease risk. The hot (and commercially available) current markers are those near chromosome 9p21. The presence of risk variants at this loci increase individual risk for CHD by about 1.3-fold over the average individual, and protective variants exist in the population that diminish risk for those lucky winners of the genetic lottery. Immediately the question arises: Can these newly identified genetic risk markers for coronary heart disease be employed to improve disease risk prediction and health outcomes?

Answering this question for one marker is tough, and answering it for 22 (and maybe dozens to follow) will require a rethinking of how we address the question. The key point for the future is to keep your eye on the ball. We are considering using new risk markers in clinical medicine to improve health outcomes, not to refine current models for predicting risk. The process of sorting out the usefulness of these markers is occurring in the context of a health care system where cost-consciousness is mandatory, and cost savings would be a huge benefit.

It may turn out that adding markers to existing risk models makes little sense unless the additions dramatically improve risk discrimination. Consider a recent area under the curve (AUC) analysis of the potential for improved risk prediction for CHD. Adding 22 genetic cardiovascular risk markers to the traditional risk factors of age, sex, and BMI changes the AUC from 63% to 66%. While this small change is certainly significant, it is just as certainly not transformative for the individual patient. An additive, incremental approach to CHD risk assessment could conceivably drive up health care costs with little return on the investment.

Transformation will be best achieved through the discovery of additional genetic contributors to risk and a re-thinking the traditional way we use the information at hand to care for individual patients and populations. For example, under current practice it seems likely that those identified as having increased risk by virtue of these genetic markers at 9p21 might be subject to more intense management. Using the information this way may improve health outcomes and is a hypothesis that should be tested. A plausible alternative hypothesis is that health benefits and cost savings might be achieved if those shown to be at reduced risk had less aggressive management with expensive drugs and lab tests.

Most scientists and clinicians concur that it is too early to recommend the use of the new cohort of CHD risk markers in routine care. (See the Jan. 20 issue of Annals of Internal Medicine.) However, the story of CHD risk evaluation is rapidly evolving, and you can expect that future guidelines for care will incorporate at least some elements of genetic risk assessment. The big question is whether the new guidelines will transform or simply tweak cardiovascular health care.