College of American Pathologists
Printable Version

  Using markers for a head start on heart risk




December 2007
Feature Story

Karen Titus

If, as Socrates posited, the unexamined life is not worth living, doubtless some wag has also noted that the overexamined life is a bore to everyone. Into the breach steps Robert Christenson, PhD, who urges pathologists to reconsider, but not wholly trash, what they thought they knew about biochemical markers of long-term cardiovascular risk.

Leaving few markers unturned, Dr. Christenson systematically examined the doubts, and data, surrounding current biomarkers, in an AACC conference, held last spring, on cardiac risk assessment, diagnosis, and management. In his talk, Dr. Christenson took an evidence-based poke at markers of primary prevention.

Cystatin C is a worthy example of how Dr. Christenson approaches matters. The marker may be unfamiliar to some. "If you haven't heard much about this protein, you probably ought to Google it," joked Dr. Christenson.

For those who've yet to do a search, Dr. Christenson provided a few basics. Cystatin C is a single-chain basic protein, nonglycosylated, with a molecular mass of 13,360 kD. It inhibits cysteine proteases and is synthesized by all nucleated cells. It demonstrates a constant production rate typical of so-called housekeeping genes—in other words, it's not influenced by acute-phase reactions. It's cleared by free glomerular filtration. Tubular reabsorption and rapid breakdown occur by proximal tubular cells; it is not secreted by tubules or eliminated by any extra-renal route. Neither muscle mass, nor food intake, nor body surface influence it.

It's this last feature that makes cystatin C particularly intriguing, said Dr. Christenson, and gives it a leg up on creatinine. "But," he cautioned, "it's more like a $5 test, rather than a 50-cent test."

One of the key investigations that put cystatin C on the map, said Dr. Christenson, was measurement in the cohort of the Cardiovascular Health Study (Shlipak MG, et al. N Engl J Med. 2005;352:2049-2060), which compared creatinine and cystatin C levels as predictors of mortality, from both cardiovascular causes and all causes, in the elderly. Cystatin C was found to be a substantially stronger predictor of risk of death and cardiovascular events. The researchers identified participants as being at low, intermediate, and high risk using cystatin C levels of <1 mg/L, 1-1.28 mg/L, and >1.29 mg/L, respectively.

Moreover, the study's authors note, cystatin C levels defined not only the top 20 percent of elderly people in the CHS cohort with a substantially higher risk of death, but also a large subgroup (the lowest 40 percent) at below-average risk of death. Cystatin C is likely associated with all-cause and cardiovascular mortality, Dr. Christenson said, and this information should remind physicians that patients with renal insufficiency are at higher risk for cardiovascular events. Some investigators suggest that cystatin C even has a direct (that is, not exclusively renal) impact on cardiovascular risk, he added.

The point of his discussion was not to spur a run on cystatin C. A good biomarker is only as good as what it brings to the bigger picture—think group photo, not head shots. How do established markers and emerging markers fit in with the other tools clinicians use? What is the right way to analyze them? In addition to searching for new markers, are there better ways of thinking about traditional ones?

The last 20 years have shown that cardiac incidents are motley events, involving the unhappy confluence of inflammation, impaired endothelial function, impaired plaque stabilization, formation of reactive species such as oxygen-free radicals, as well as the crowning event of thrombus formation. Thinking about markers in terms of their varied physiological processes probably is helpful, he said, and has given rise to multimarker strategies.

It might also be helpful for lab folks to step out of their domain, at least metaphorically, and think about how biomarkers fit into clinicians' worldview and their necessarily wider consideration of risk.

In his own lab at the University of Maryland School of Medicine, Baltimore, where he's professor of pathology and professor of medical and research technology, Dr. Christenson said he himself failed to fully realize the high risk to patients with impaired renal function, in part because the Framingham Risk Score does not contain a term for impaired glomerular filtration rate.

GFR is part of pathology's dialect, but FRS may not be. The Framingham Risk Score is a prediction tool for estimating 10-year risk of myocardial infarction or cardiovascular death. It considers factors both uncontrollable (age—"I wish we could do something about age," Dr. Christenson said with a laugh—and HDL) and controllable (smoking, blood pressure to some extent, LDL). With these cards on the table, the question changes slightly: What more can biomarkers do to add to the current ability to assess cardiac risk? "What we have to remember is the additive effect of what any biomarker brings. It can't be looked at by itself—it has to be looked at in the context of the way the clinician is going to use it."

The FRS remains the recommended starting point for exploring cardiac risk (, said Dr. Christenson, who performed the calculation using himself as a model ("a fella in his late 40s—OK, 53," he conceded) to arrive at a risk of five percent. By making additional, hypothetical adjustments, he bumped up his risk—if he were a smoker, for example, he'd more than double his risk, possibly making him a candidate for treatment with a statin.

"Any biomarker that we come up with must help with this risk," he said. "If it doesn't increase what we already can do in the clinics, then the marker is of modest value at best." Viewed from the standpoint of physiology, markers must be able to help clinicians determine which patients might be progressing toward atherosclerosis, plaque erosion and rupture, activation of the coagulation cascade, and the like. On the backside, markers might be able to help clinicians sort through various outcomes, such as coronary artery disease, peripheral artery disease, AMI, stroke, and heart failure.

Inflammation is certainly worth thinking about. It's fundamental in the pathogenesis of atherosclerosis, and it predicts risk of coronary heart disease in primary and secondary prevention settings, as well as the efficacy of some medications. There's no shortage of markers for monitoring inflammation, but how do we know which ones are best? "Why is it that you've seen hs-CRP on the cover of Time magazine?" but not fibrinogen or SAA, Dr. Christenson asked. "Part of the answer lies in stability of the biomarker, standardization, and, in short, our ability to report on the marker."

It's also partly because high-sensitivity C-reactive protein puts laboratorians on solid footing with clinicians. "We have to walk the walk and talk the talk," he said. That means using the same criteria and grading evidence clinicians rely on. These are derived from American College of Cardiology/American Heart Association classifications and summary of indications (Pearson TA, et al. Circulation. 2003;107:499-511). Classification I is the sturdiest, backed by evidence/general agreement that a given procedure or treatment is useful and effective, while level A offers the best weight of evidence, girded by data derived from multiple randomized clinical trials involving large numbers of patients. The Framingham Risk Score, for example, is a classification: 1, level of evidence: A. This leaves little room for quibbling.

One reason for hs-CRP's popularity is slap-your-head simple: It's easy to measure. "We've done a great job in achieving a level of standardizing the measurements," Dr. Christenson says. Because of this, it's easy to talk about hs-CRP with clinicians with a green, yellow, and red analogy. Levels below 1 mg/L are consistent with low risk; between 1-3 mg/L, average risk; and above 3 mg/L, high risk. Very high concentrations, greater than 10 mg/L, suggest an inflammatory process and require re-measurement at a later time.

A review of assays of inflammatory markers from the aforementioned Circulation article looked at hs-CRP as well as soluble adhesion molecules (for example, E- and P- selectin, intracellular adhesion molecule-1, and vascular cell adhesion molecule-1), various cytokines (for example, interleukin-1 -6, -8, and -10, and tumor necrosis factor-alpha), and the acute-phase reactants fibrinogen and serum amyloid A. Only hs-CRP displayed all four hallmarks for abetting clinical use: stability, many available assays, available standards from the World Health Organization, and good interassay precision (it has a CV of <10 percent). Like the Framingham criteria, hs-CRP enjoys an ACC/AHA classification of 1 and a level of evidence of A. It has, said Dr. Christenson, the most appropriate analyte and assay characteristics of the current inflammation markers for clinical use. "There's a lot of good evidence," he said.

The next step is to evaluate markers in terms of treatment. Though hs-CRP indicates risk, the evidence for its use in directing further evaluation and therapy in the primary prevention of cardiovascular disease is dodgy—in terms of the ACC/AHA guidelines, it's ranked as class IIa, meaning the weight of evidence/opinion is in favor of usefulness/efficacy. That may sound good, but in guideline-speak, it's far from a ringing endorsement.

A study from the Annals of Internal Medicine (Cook NR, et al. 2006;145:21-29) shows what even a good test is up against. It looked at the impact of hs-CRP on the risk classification of nearly 27,000 nondiabetic women. Nearly 88 percent of the women had a baseline 10-year FRS risk of less than five percent. In that group, there's no need to measure hs-CRP (classification: 1; level of evidence: A). In this group, said Dr. Christenson, risk can't sink any lower. "So getting an hs-CRP measurement to lower their risk is absurd." In the highest risk group (women with a baseline FRS risk of 20 percent or higher), adding hs-CRP is also unlikely to add useful information because treatment for that group will most likely be implemented regardless of the result. "You're already at high risk and a strong candidate for treatment, so what good is the hs-CRP going to be?" Dr. Christenson asked.

Dr. Christenson suggested hs-CRP adds value for the three percent of women whose baseline FRS risk falls into the 10 to less than 20 percent range, where clinicians might be unsure about prescribing preventive therapies such as a statin. In this group, adding hs-CRP might reclassify patients into either a higher or lower risk category, defogging things a bit (1/A). About 13.7 percent would be reclassified downward, and their risk, after recalculation with hs-CRP, would be about 6.8 percent. Five percent would be reclassified at higher risk, with a recalculated risk of nearly 19.9 percent. For clinicians and patients wavering about treatment, recasting the risk may make their decision about whether to treat more clear-cut.

If the baseline FRS risk is between five and 10 percent, close to 10 percent of patients might be reclassified to a higher risk group if an hs-CRP measurement is added. It's not clear if moving women into this group has clinical benefit, especially in terms of longer-term, lifetime risk prediction (IIa/B). The authors suggested, however, that women with FRS risk between five and 20 percent might benefit from adding hs-CRP measurements.

Following the money trail is also a good way to evaluate what new markers might bring to the table, said Dr. Christenson, who not only walked the walk and talked the talk, but gave a decent impression of an actuary to boot.

He started by reviewing the estimated cost-effectiveness of various LDL-lowering therapies. If the 10-year FRS risk is 35 percent, even a higher-end drug—one costing, say, $3/day, or roughly $1,000/year—is "a pretty good deal," he says, noting the cost-effectiveness (cost per quality-adjusted life years gained) of this strategy is $10,000.

In the 25 percent FRS risk category, the cost per QALY gained is still a pretty easy decision: $25,000 for the high-end drug.

What about a 10 percent FRS? "Would you treat all those patients?" Dr. Christenson asked. If the treatment runs $3/day, "you have to be prepared to make some decisions," he said, noting this carries a tab of $100,000 cost per QALY gained. At five percent, the $3/day drug translates into $200,000 cost per QALY gained. "You can see where this is going," he said.

The 10 percent figure, he added, is considered the financial tipping point, and the reason why intermediate risk is defined as being 10-20 percent. "So when we look at these emerging markers, cost must be a very important consideration."

Maybe measuring hs-CRP will motivate patients to quit smoking, he acknowledged. "But nobody has done a trial" to see if this is true. The evidence suggesting a link between hs-CRP levels and lifestyle changes is scant (classification: IIb, level of evidence: C).

How did hs-CRP hold up to Dr. Christenson's scrutiny? Perhaps not as well as Time magazine would have it.

It has predictive power, but is not proven to be directly causative, Dr. Christenson said. It's influenced by body weight and highly correlated with metabolic syndrome. The assay's technical characteristics are good, but—and this is important—it's not yet integrated into the FRS assessment. At this point, he said, its clinical utility appears to be as an add-on risk marker.

And other biomarkers?


Homocysteine's star may be dimming a bit, said Dr. Christenson. It has positive association with CVD risk, though it's not integrated into global risk assessment. It can be reduced by folate and vitamins B12 and B6, and clinically, high levels are treated with vitamins. But, Dr. Christenson noted, the vitamin trials have generally been negative. This is because lowering homocysteine by 25 percent—the figure used in most analyses—doesn't lower risk, according to a meta-analysis published in JAMA (2002;288:2015-2022). "If you lower it and there is no improvement in outcome, it obviously can't be used as a target for therapy."

Despite its tangible link to ischemic heart disease, he said, homocysteine doesn't add much to the FRS, and measuring it isn't warranted for primary prevention and assessment of cardiovascular risk (III/C). Homocysteine levels in high-risk patients may have fairly modest ability to predict future cardiovascular events in those with multiple risk factors, such as renal insufficiency, hypertension, or metabolic syndrome (IIb/C).


While admiring the heart's stellar pumping attributes, Dr. Christenson admits to a fondness for its role as a hormone-producing organ. BNP and NT-proBNP affect natriuretic resistance, with a resulting impact on kidney as well as basal dilation. "I think of this as a marker of heart stress," such as increased ventricular load, he said. "This has an advantage over hs-CRP, although it's a whole different mechanism."

Dr. Christenson's enthusiasm comes, in part, from a "wonderful" prospective study published in the New England Journal of Medicine (Wang TJ, et al. 2004;350:655-663) that looked at 3,346 patients without heart failure. The patients who fell into the highest third of BNP scores still had relatively low baseline scores: >12.8 pg/mL for men, and >15.8 pg/mL for women. Noted Dr. Christenson: "You have to be prepared, if you use this in primary prevention, that very low values are what we're going to use for our risk stratification decisions." Plasma NPs predicted risk of death and cardiovascular events after adjustment for traditional risk factors, with excess risk apparent at levels well below current thresholds used to diagnose heart failure, he said.

Citing several other studies, he called NT-proBNP and BNP "death markers." Virtually all studies have shown that if you have high values, you're at increased risk.

Is the evidence all in? Certainly not. The ACC/AHA scorecard rates the markers III/B: Benefits of therapy based on marker measurements are uncertain, and measurement for primary prevention and assessment of cardiovascular risk is not advised. But, said Dr. Christenson, "I think it offers another piece of that physiological evidence, along with inflammation, and maybe unstable plaque, that there's heart stress going on."


Another "old" marker came in for a new look—microalbumin, which, says Dr. Christenson, is another powerful risk factor that reflects a renal component and perhaps increased pressure at the level of the glomerulus. A number of studies make that point. But, he warned, labs still don't have a good handle on the standardization, appropriate units for measurement—the studies bounce between milligrams/millimole and micrograms/minute—and other measurement issues.


To round out the picture, Dr. Christenson turned to another study by Thomas Wang, MD (N Engl J Med. 2006;355:2631-2639), this one looking at multiple biomarkers for predicting first major cardiovascular events and death. While the study did not have primary prevention as its focus, Dr. Christenson suggested this study was still worth a close look.

Ideally, he said, it would be great if markers could be used to help steer people into low-, intermediate-, and high-risk groups. "Wouldn't that be a useful addition to the Framingham score?" he asked.

Alas, they couldn't pull it off, at least not in this study, which looked at 10 emerging biomarkers: hs-CRP, BNP, N-terminal pro-atrial natriuretic peptide, aldosterone, renin, fibrinogen, D-dimer, plasminogen-activator inhibitor type 1, homocysteine, and urinary albumin-to-creatinine ratio.

Adding the multimarker approach to what clinicians already use was of little benefit in predicting death. The ROC curve area (also referred to in the study as C statistic) for age, sex, and conventional risk factors as predictors was .80; for age, sex, and multimarkers, the ROC curve area was .79. Adding all predictors produced little added benefit—the ROC curve area was .82.

Doing the same type of analysis for major CV events has yielded similar, slightly disappointing results. Age, sex, and conventional risk factors translated into an ROC curve area of .76. The ROC curve area for age, sex, and multimarkers was .70. Using all predictors was .77—again, adding multimarkers made only a small dent.

Numbers such as these explain the muted enthusiasm for cardiovascular biomarkers, Dr. Christenson said. But, he added, this approach—ability to shift the ROC curve—"is a fair way to look at markers. I think there is a science right now that I hope some of us are turning to."

"Certainly the ROC curve is a very high mark to jump, and it's a tough one," he continued. In this study, the markers added moderate value to the standard risk factors, the researchers concluded, a statement Dr. Christenson called "pretty kind," given that the markers had only a small increase in the area under the ROC curve. "Even a great enthusiast like me about these markers has to be cautious after examining this sort of study."

Where does this leave laboratorians? In pursuit of novel biomarkers and perhaps with a handful of questions, suggested by Dr. Christenson as a way to evaluate them:

  • Is it predictive of risk?
  • Is it an independent risk factor?
  • Can the emerging risk factor be used as an add-on global risk assessment?
  • Can it be incorporated into a global risk-assessment algorithm? If not, why not?
  • Is the marker an innocent bystander, or is it involved in physiologic process? If it's involved, could it be a target of therapy?
  • What is the cost? What is the availability of testing?

Until these questions are answered, perhaps no emerging marker should earn an answer of "yes" to another question: Does it belong on the cover of Time?

Karen Titus is CAP TODAY contributing editor and co-managing editor.

Related Links