Comparing strengths and limitations of hsCRP and CAC in refining CVD risk prediction

We hope that our recent paper in The Lancet comparing the predictive power of high-sensitivity C-reactive protein and coronary calcium challenged many physicians’ established beliefs about risk prediction in primary prevention.

For decades, the cardiology community has relied principally on the traditional Framingham risk factors for risk prediction. More recently, some clinical researchers have championed high-sensitivity C-reactive protein (hsCRP) as an additional tool for discerning vascular inflammation and CV risk.

Michael J. Blaha

Khurram Nasir

Roger S. Blumenthal

We posit the following question: What is the future of risk prediction — continued reliance on single measures of risk factors reflecting a single point in time, or incorporation of direct measurements of the disease (atherosclerosis) we propose to treat?

CAC in the MESA JUPITER population

In our paper, we identified 950 participants from the Multi-Ethnic Study of Atherosclerosis (MESA) who fit all criteria for the JUPITER study, and followed them for nearly 6 years. By definition, all of these patients had “normal” LDL concentrations (<130 mg/dL) and high hsCRP (≥2 mg/L). The JUPITER study showed that patients such as these received a 46% reduction in major CV events with 20 mg of rosuvastatin (Crestor, AstraZeneca). We asked the following question: Can coronary artery calcium (CAC) testing identify JUPITER subgroups expected to derive the most, and the least, benefit from statin therapy?

We found that CAC discriminates risk up to 20-fold, identifying JUPITER-eligible patients both at very low risk and very high risk. Expressed as a predicted 5-year number needed to treat (NNT), we found that among JUPITER-eligible participants with CAC of zero, the NNT to prevent one heart attack is more than 500. These patients are unlikely to gain a net benefit from statin treatment within the next 6 years. Physicians of these patients with normal LDL levels could reasonably elect not to start a daily lipid-lowering medication for the rest of their lives. In contrast, those with measurable CAC, especially CAC scores of more than 100, have a highly favorable NNT of approximately 20 to 40. These patients can expect a reasonable benefit from statin therapy during the next decade. We believe that our findings have important potential public health implications for statin allocation in primary prevention.

CAC vs. hsCRP for risk prediction

The second major aim of our paper was to directly compare CAC vs. hsCRP for risk prediction among JUPITER-eligible participants — the sort of patients in whom we all struggle to make treatment decisions. We found that CAC is clearly superior to hsCRP for risk prediction. Although the simple presence of CAC increases risk more than fourfold, an hsCRP of at least 2 mg/L had no prognostic significance after multivariable adjustment. Importantly, CAC was an equally good predictor of risk regardless of the baseline hsCRP level. The analysis was a clear “win” for CAC and showed us that hsCRP of less than 2 mg/L may not provide much reassurance in the context of measurable subclinical atherosclerosis.

Why is CAC so much better than hsCRP for risk prediction? The answer is likely that CAC is a direct measure of subclinical atherosclerosis — the disease we all are seeking to identify and treat, the disease that is a necessary precursor to an atherosclerotic heart attack. In an asymptomatic adult with CAC of zero, very few heart attacks will occur in the coming 6 years. All-cause mortality in patients with zero CAC is approximately 1% during the ensuing 10 years. In contrast, an elevated CAC score reliably reflects the build-up of substantial atherosclerosis. Thus, CAC integrates the cumulative risk exposure during one’s lifetime, explaining its superior performance to individual risk factors measured at single time points. We find that most patients (and clinicians) understand this relatively simple concept, and many elevated CAC scores result in motivating to comply with more aggressive secondary prevention strategies.

The power of zero CAC

Although many risk factors, traditional risk scores and novel serum biomarkers are available for quantifying increased risk in asymptomatic patients, none of these have sufficient sensitivity to exclude clinically important CAD. In the primary prevention population, a CAC of zero stands alone as perhaps the most powerful “negative risk factor” for near-term development of a coronary event. A test such as hsCRP, when the result comes back low, provides clinicians very little reassurance against a future MI, as evidenced by our MESA results. When is the last time you were very reassured about a patient’s prognosis because the hsCRP came back as 1.7 mg/L? Since hsCRP and almost all other nonspecific markers cannot rule out disease, they may only be used to raise risk estimates and, thus, are often tied to more treatment and downstream cost.

Our Lancet study has several important implications. First, a CAC scan appears to have great value in patients with normal LDL in whom the decision to treat is uncertain. With a finding of zero CAC and a normal LDL, clinicians might feel comfortable emphasizing low-cost lifestyle therapies, prescribing less costly pharmacotherapies, and deferring repeat imaging or invasive tests in persons who develop atypical symptoms in the next few years. In patients with elevated CAC, particularly those with a score of more than 100, clinicians should be aggressive and treat similarly to those with established heart disease.

Are we advocating for a decrease in the overall number of patients taking statins? Although relying on hsCRP alone and following strict JUPITER criteria may result in overtreatment, our analysis leads to a more broadly applicable conclusion. Since our data suggest that CAC predicts risk equally well when hsCRP levels are low, we conclude that we simply need to be directing treatment at the right patients. This appears to argue for expanded use of CAC testing in low-intermediate risk patients without a clear indication for statin therapy. Indeed, the 2010 ACCF/AHA Guideline for Assessment of Cardiovascular Risk in Asymptomatic Adults gives CAC scoring a IIa recommendation in the intermediate-risk group, equivalent to the IIa recommendation for hsCRP in those meeting the exact JUPITER entry criteria, and a class IIb recommendation in those with a Framingham risk estimate for an MI/CHD death of 6% to 10%.

Considering a paradigm shift

Is it time to shift toward subclinical atherosclerosis testing as the primary means of risk prediction after accounting for traditional risk factors? There are still several roadblocks to this sea-change in clinical thinking. Foremost, a randomized controlled trial is probably necessary to show that statin allocation guided by CAC testing actually saves lives. The NIH is considering such a trial; however, such a large undertaking will be expensive and will likely take at least 5 years to complete.

More research is needed on the downstream implications of CAC testing. We still hear reports of patients referred for cardiac catheterization after an elevated CAC score. This should never happen. CAC tests are prognostic tests only and should not be considered diagnostic tests. Referral for invasive testing should be driven by symptoms or demonstration of at least moderate inducible ischemia. The issue of incidental pulmonary findings (non-calcified lung nodules), which may occur in 5% to 10% of scans, is also an issue that must be explored before more widespread use of CAC testing can be considered. Patients must be informed of the approximate 1 mSv of ionizing radiation, which is similar to one to two mammograms.

We enjoy the ongoing debate between those tied to traditional risk factors alone vs. risk factors plus either selective use of novel biomarkers or imaging as a means to refine CV risk prediction. Of note, the median hsCRP in MESA was more than 2 mg/dL, so the currently used threshold of 2 mg/L may not be the right one when we look at an older, multi-ethnic population with a high prevalence of hypertension. We hope our controversial paper provided more food for thought.

Michael J. Blaha, MD, is a cardiologist at The Johns Hopkins Ciccarone Center for the Prevention of Heart Disease in Baltimore.

Khurram Nasir, MD, is a radiology resident at the Yale University School of Medicine in New Haven, Conn.

Roger S. Blumenthal, MD, is professor of medicine and director of The Johns Hopkins Ciccarone Center for the Prevention of Heart Disease. He is also Section Editor of the CHD and Prevention section of the Cardiology Today Editorial Board.

For more information:

• Blaha M. Lancet. 2011;378:684-692.

Disclosures: Drs. Blaha, Blumenthal and Nassir report no relevant financial disclosures.