BETHESDA, MD—Five years ago, NIH started promoting a paradigm of medicine—one that was predictive, personalized, preemptive and always with the participation of the patient. That paradigm began with the ability to predict who was at risk for certain diseases, including cardiovascular disease.
While science has identified any number of risk factors for CV disease, can it accurately predict who will actually develop the disease? Philip Greenland, MD, director of the Clinical and Translational Sciences Institute at Northwestern University in Chicago, and a member of several National Heart, Lung, and Blood Institute (NHLBI) boards, emphasizes that knowledge of risk factors is not the same as predictability and suggests that the answer to that question remains elusive.
“The question many of us have been working on is whether we can actually accomplish this,” Greenland said at a recent talk on the subject on the campuses of NIH. “Any of us who have been a patient would like medicine to look like this. But how far are we along that pathway in terms of cardiovascular medicine?”
Lack of Specificity
The last 20 years has seen the literature on CV risk factors grow exponentially, and numerous guidelines for heart health have been released, including ones from NHLBI, the American Heart Association and the World Heart Foundation.
“All of them come down to this: We have an evidence-base based on numbers of one sort or another, and if you want to manage your risk, you have to know your numbers,” Greenland said.
There has even been legislation mandating CV risk screening. In 2009, the state of Texas enacted a bill stating that, if a health plan provides coverage for any kind of medical-screening procedures, it must also provide minimum coverage for detection of CV disease. “Texas has decided that the evidence is strong enough, and we should be offering these tests to patients,” Greenland said. “But can we as scientists justify that recommendation?”
In reality, much of the literature deals with risk factors, which is not the same as prediction, Greenland said. The major risk factors, such as hypertension, diabetes, smoking, alcohol use, obesity and an unhealthy diet, are quite good at determining who will develop CV disease, but they lack specificity.
For example, diabetes will increase a person’s chances of developing CV disease yet many people with diabetes will never have any coronary problems whatsoever. While diabetes increases risk, it cannot be accurately used to predict who will and will not develop CV disease.
“In studies, we found that risk factor exposures are actually very common. The risk factor models we have are not all that good at predicting who will get a coronary event,” Greenland said.
Who Won’t Get a Heart Attack
On the other hand, science has a lot of good information on who is unlikely to get a coronary event. In the Chicago Heart Association Study, which examined more than 360,000 men over 40 years, and in similar studies, researchers were able to look at a number of factors and their effect on heart health over time. Those factors included age, history of diabetes, years of education, smoking, blood pressure, (BP), BP mediation, ethnicity, heart, weight, serum cholesterol and resting electrocardiogram (ECG).
The hypothesis was that, if a patient had a systolic BP of less than 120 mm, a diastolic BP less than 80 mm, cholesterol less than 200 mg/dl, was nonsmoking and did not have diabetes, then they had little risk of having a CV event over the next 15 to 20 years. That hypothesis proved true, with patients meeting those criteria having a 92% lower rate of CV events than the rest of the cohort.
Those patients with none of the major risk factors were found to live an average of 10 years longer than those who had multiple risk factors. However, researchers found that this was not a common profile among Americans—comprising less than 10% of the cohort in the Chicago study.
“If you want to predict who’s not going to have CV disease, the answer to that is pretty well established,” Greenland said.
Last year, the American Heart Association adopted this profile as the optimal CV health profile, widely promoting it and developing a website to help people track their risk factors.
Evaluating New Predictors
“But how do we predict the high-risk people,” Greenland asked.
The last few years have seen studies performed on a number of possible predictors. One of these was C reactive protein (CRP), a protein found in the blood, the levels of which rise due to inflammation. It was thought that the protein could be a viable biomarker for CV disease. However, studies found that CRP had a minimal impact on prediction and may actually result in poorer predictions, Greenland said.
Researchers had similar outcomes when looking at genetic predictors for CV disease. Adding the genetic profiles linked to CV disease into risk calculations had no impact on prediction, Greenland said. “They may be able to help us with drug targets, but in terms of prediction they’re not very successful. Most of the single-nucleotide polymorphisms (SNPs) that have been identified have odds ratios of 1.1 or 1.2, which is pretty unimpressive in terms of odds ratios.”
Another possible predictor is coronary calcium—specks of calcium that form along the arterial walls—which can be identified using computed tomography. Coronary calcium came to the public attention in 2005, when Time magazine ran a cover article on the possibility of using the scan as a predictor of CV disease.
“I got interested in the topic mainly because of the hype,” Greenland said. “I was convinced that the hype was unjustified.”
In a study published in The New England Journal of Medicine in 2008, Greenland looked at coronary calcium (CAC) as a predictor in a group of 6,724 people aged 45 to 84. He and his colleagues tracked the group over four years of follow-up, documenting the relatively few CV events that occurred.
Greenland found that people with a very high CAC score compared with those with a low score had a hazard ratio of about 10—much higher than CRP or the various genetic predictors. “It’s still not quite what we’re looking for [in terms of predictability], but it’s certainly a significant improvement,” Greenland said. “Is it ready for prime time? Probably not, because there is hazard associated with radiation exposure.”
The next step in evaluating the CAC test would be a clinical trial looking at actual measurement of patient outcomes.
“That’s something we’re currently working on,” Greenland said. “If you’re faced with someone who’s of intermediate risk, this imaging test may help you differentiate. The selective use of the test may help you. But that’s a clinician’s answer, not a health policy answer.”