COMMENTARY

Who Needs a Statin? DNA Beats Current Risk Calculators

John Mandrola, MD

Disclosures

May 02, 2017

A middle-aged man whose brother recently had an MI wanted to know his cardiac risk and what steps he could take to avoid heart disease.

It's hard to believe that answering this central question of cardiology would be the same now as it was decades ago. We would calculate his cardiac risk using basic factors: smoking, age, blood pressure, and cholesterol levels.

Imprecise is a generous word to describe this method of predicting future cardiac events or tailoring preventive therapies. Consider, for instance, the most likely outcome for patients taking a statin drug for primary prevention is the same if they didn't take it—nothing.[1]

A recent paper in the Journal of the American Medical Association compared statin eligibility based on either the US Preventive Services Task Force  (USPSTF) and ACC/AHA recommendations.[2] On Twitter, Dr Eric Topol, the editor in chief of Medscape and a leading genomics researcher, wrote: "The debate about which statin guideline for primary prevention is archaic in an era with validated genetic risk scores (GRS)."

This spurred me to look at the evidence for genetic risk scores. I came away impressed and cautiously optimistic.

Background

The challenge of using genetic data to predict coronary artery disease (CAD) is that, unlike monogenetic diseases such as long-QT syndrome, CAD stems from many genes and environmental factors.

We have known for years—from genomewide association studies—that there are numerous genetic loci associated with CAD[3,4,5]. At these loci are single nucleotide polymorphisms (SNPs, a variation in one base pair) that may individually or together influence the risk of developing CAD.

The problem is that the individual predictive power of these SNPs is small. That's where risk scores become useful. These polygenetic scores cull many SNPs into a composite score.

This may sound technical, but it's not. Recent advances have reduced the cost of genotyping; identifying disease-associated SNPs is no longer beyond the realm of clinical use.

Dr Pradeep Natarajan, a cardiologist and genetics researcher at Harvard, told me that there is no standard test that you can order now but risk scores can be easily calculated from genomewide arrays that some direct-to-consumer (DTC) genetic testing companies provide. For example 23andMe used to report a gene risk score for CAD, but the FDA asked the company to stop (it can report on some conditions). Dr Topol and other researchers from Scripps are now in beta-testing for a no-cost mobile app (MyGeneRank) that takes a user's 23andMe data and delivers a genetic cardiac risk score.[6]

Compelling Evidence

Dr Gad Abraham (University of Melbourne, Australia) and colleagues asked whether a genomic risk score (GRS) comprising more than 49K SNPs could predict CAD.[7] They divided the score into quintiles from low to high risk, then validated its predictive ability in five prospective cohort studies—three FINRISK cohorts (n=12,676) and two Framingham Heart Study cohorts (n=3406). They also validated the score in a smaller Dutch cohort (ARGOS) of individuals with familial hypercholesterolemia.

There were four main findings of this paper:

  • The GRS associated with incident CAD in both the FINRISK and Framingham cohorts independently of established risk factors, including family history.

  • The GRS also associated with CAD in a high-risk group of patients with familial hypercholesterolemia.

The curious will note, as the study authors did, that these findings suggest genomic risk exerts its effect through molecular pathways largely independent of cholesterol, blood pressure, and smoking.

  • The GRS modestly improved on the 10-year prediction of current risk-factor–based scores.

  • Fourth, and most important, the GRS captured differing trajectories of cumulative CAD risk, For instance, applying Kaplan-Meier estimates of CAD events to the FINRISK cohorts revealed that men with the highest genetic risk reached a 10% level of cumulative risk 18 years earlier than men with the lowest genetic risk.

These are impressive findings, but there are limitations.

First, the differing trajectories of CAD incidence based on genetic risk were of lesser magnitude in women, which may be due to lower CAD event rates in women. Second, although the GRS modestly improved risk prediction of the population, some individuals who had an event were reclassified from higher risk per traditional risk calculation to lower risk based on the genetic score. Finally, the cohorts used in this study (and most studies of genetic risk scores) consisted largely of individuals of European descent. Utility of risk scores in people of varying ancestry requires validation.

Guiding Statin Therapy

Publishing in the Lancet, an international group of researchers used data from two community cohort studies and four randomized trials of statins (in both primary and secondary prevention) to validate a genetic risk score and determine those individuals who may derive greater benefit from statin therapy.[8]

The big finding from this study of more than 48K subjects was that patients with high genetic risk scores derived much greater risk reduction from statin therapy. The associations followed a gradient, with increasing relative risk reductions across low (13%), intermediate (29%), and high (48%) genetic-risk categories. Specifically, when the authors applied the gene risk score to the primary-prevention statin trials, they observed an approximate threefold decrease in the number needed to treat (NNT) to prevent one CAD event. For example, the NNT for high-genetic risk individuals in the ASCOT trial[9] was 20 (vs 72 for total coronary events in the original paper) and in JUPITER[10] it was 25.

Dr Natarajan and colleagues confirmed these gradients of risk reduction based on genetic risk scores in a meta-analysis of three primary-prevention statin trials.[11] In addition, they applied the gene score to participants in two observational cohort studies and noted that each standard-deviation increase in risk score associated with a 1.32-fold greater likelihood of having coronary artery calcification (CAC) and a 10% higher burden of carotid plaque.

A crucial factor to consider is that the search for genomic links to CAD continues.[12] Gene risk scores will get better as more SNPs are found.

Conclusions

I'm a clinician, not a geneticist, but the lack of excitement over the potential of genetic risk scores is surprising. Consider the statin debates. While nearly everyone agrees that statin benefit outweighs harm for secondary prevention, the argument in primary prevention turns on the smallness in absolute benefit (or high NNT). Gene scores may change that debate. If confirmed, a threefold reduction in the NNT for statin therapy has immense value—on both a patient and population level.

But the use of gene scores go beyond statin decisions. Knowledge is power. Heart disease may be a major killer, but it's highly preventable. The MI-GENES study randomized 200 patients to disclosure of CAD risk via a conventional risk score or traditional scoring plus a genetic risk score. The group with the added genetic info had lower LDL-C levels at 6 months than those given only a conventional risk score.[13]

The value in medical tests lie in their ability to change behavior or therapy. Obviousness dictates that (some) people at high risk might try to lower their risk. And the evidence—from a study of more than 55K individuals—is that among individuals at high genetic risk, adherence to a healthy lifestyle associated with a 50% lower risk of CAD relative to those with unfavorable lifestyles.[14]

Dr Natarajan reminded me that gene scores, unlike CAC scans, are not age-dependent and can identify higher risk people at younger ages. This is key because atherosclerosis is a chronic lifelong disease and the young stand to gain the most from preventive strategies.[1]

Of course, genetic risk scores need more study. I favor slow science. But the evidence looks promising. I'm afraid that in a few years we may look back on this data and think: what took us so long to see the signal in the genes?

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as:

processing....