Dallas, TX - The question of whether C-reactive protein (CRP) should be measured in routine risk-factor screening has been a topic of much discussion over the past few years. So far, consensus has been difficult to achieve, although all factions have the same data available to them.

CRP was the subject of a lively debate at the American Heart Association (AHA) meeting last November.

Prof Wolfgang Koenig

Presenting the argument in favor of measuring CRP, Prof Wolfgang Koenig (University of Ulm, Germany) noted that traditional risk factors were not enough to assess heart disease risk and that CRP has been shown many times over to be helpful in this regard. But opposing this idea, Prof Gordon Lowe (University of Glasgow, Scotland) claims the "spin" on CRP has been overdone and that routine measurement of CRP would not be a good value for the money.

Participating in the discussion, Dr Paul Ridker (Brigham and Women's Hospital, Boston, MA), who has been at the forefront of CRP research for over a decade, strongly disagrees with Lowe. "CRP has repeatedly been proven to predict risk over and above that achievable with global prediction models, and the cost-effectiveness appears equal to or better than that of cholesterol screening," he comments. [For full disclosure, Ridker notes that his hospital holds patents that relate to the use of inflammatory biomarkers in cardiovascular-disease therapies.]

Rounding off the debate at the AHA meeting, Dr Scott Grundy (University of Texas Southwestern Medical Center, Dallas) took the middle ground. He said he was happy to recommend that CRP be used to motivate patients to adopt healthier lifestyles, but he was not sure that it was appropriate to use CRP to guide treatment decisions at this point. "I think we can say that obese people with raised CRP are at higher risk of heart disease than those with lower levels, and if you tell me that you want to measure CRP as a tool to motivate such patients for lifestyle therapy, I can accept that without any doubt. I will do anything to get you to use lifestyle therapy more. But more difficult is the question of whether we can use it as a guide as to who should be treated with drugs such as statins and aspirin. We are moving in that direction, but perhaps we are not there yet," he commented.

At present, the US seems to have taken a more positive stance on CRP screening than Europe. In 2003 the American Heart Association and the Centers for Disease Control issued a statement advising that CRP testing may be useful as a discretionary tool for evaluating people with moderate risk. The European Society of Cardiology (ESC), however, has not recommended CRP screening, pending further evidence.

New marker needed to further stratify intermediate risk

On the issue of needing better risk stratification for heart disease, Koenig pointed out that patients at risk are often missed when just traditional risk factors such as blood pressure, cholesterol, and smoking are examined. He explained that currently the Framingham calculation is used to assess risk using these traditional risk factors and that a probability of developing CHD over the next 10 years of less than 10% is defined as low risk, between 10% and 20% is defined as intermediate risk, and over 20% is defined as high risk. "While there are clear instructions for high-risk and low-risk individuals, the large group at intermediate risk, which makes up about 40% of the population, needs further tests to try to determine their actual risk more accurately. So a new marker is desirable," he noted. Ridker agrees, noting, "Studies worldwide confirm that half of all heart attacks and strokes occur among apparently healthy men and women with normal cholesterol levels, and one event in five occurs among those with no major traditional risk factor at all."

Koenig said a new marker must fulfill the following requirements:

• Give independent information on risk or prognosis.
• Account for a clinically significant proportion of disease.
• Be reliable and accurate.
• Have good sensitivity and selectivity.
• Be available and practical for widespread applications at reasonable cost.

As inflammation is believed to be a precursor to CHD, many new emerging risk markers are measures of the inflammatory process. Noting that there are many markers that signal the presence of elevated inflammation, Koenig asked: "So why are we so focused on CRP?"

He said it was because it fulfilled most of the above criteria. "By and large, next to lipids and blood pressure, CRP gives us the strongest suggestion of future cardiovascular events. And beyond any doubt, CRP is the first of all the inflammatory markers to have good evidence of clinical utility." So why is there controversy about this?

Controversy stems from Reykjavik study

Koenig believes that the controversy has mainly arisen from a paper published last year in the New England Journal of Medicine [1]. This was the largest study so far to evaluate CRP as a risk factor for heart disease and was conducted by a group led by Prof John Danesh (University of Cambridge, UK). Researchers analyzed data from 6500 men and women from Reykjavik, Iceland, without a history of MI, who were followed for 17 years. They found that CRP was associated with a risk of future heart disease, but after adjustment for other risk factors, the hazard ratio (top-to-bottom tertile) was 1.45, somewhat lower than the hazard ratio of 2 previously reported. And in a receiver-operating-characteristic (ROC) analysis, looking at the area under the curve—a statistical approach favored by some to determine the incremental value of a new risk marker—Danesh et al found that CRP did not offer much incremental benefit after the traditional risk factors had been taken into account.

Koenig says that this study has been widely interpreted as negative, but he adds: "There is no doubt in my mind that this is not a negative study." He explains that the odds ratio of 1.45 for the association between CRP and cardiac events is smaller than other studies, but this was over an average 17 years of follow-up. "The normal follow-up time for such studies is about 10 years, and longer follow-ups do appear to be associated with some attenuation in the hazard ratio. And if we look at the risk estimate at 10 years in the Reykjavik study, it is similar to those from other studies—around 2."

He also says that there are other limitations to this study that have not been discussed adequately. For example, the data come from an Icelandic population in the mid-1960s, which may differ considerably from the populations in the rest of Europe and the US today.

High cholesterol levels masked everything else?

"I would regard that as a historical population," he said, pointing out that their cholesterol levels were very high compared with modern-day values. "The fact that the cholesterol was unusually high in this population needs to be taken into account. This sort of cholesterol level is not normally seen today. Because cholesterol was so high it went in very strong as a confounding factor and it masked everything else. So the other variables did not offer very much in comparison. It is well known that if you have one very strong risk factor, it is difficult to show incremental benefits of other risk markers. According to this study, neither smoking nor systolic blood pressure contributed to the risk of CHD either, and we know that this is not the case."

Koenig further points out that the ROC analysis used by Danesh et al may not be the best way of judging the incremental value of individual risk markers. This view was also expressed in an editorial published on the subject just last month in the Archives of Internal Medicine, which says this approach has severe limitations [2].

A different interpretation

Prof Gordon Lowe

But Lowe disagrees with Koenig's arguments on the Danesh study, describing them as "red herrings." "Several other recent studies show similar findings to the Reykjavik study," Lowe says. "Okay, the Reykjavik study started during the 1960s when cholesterol levels were higher than they are today, but this is no reason not to consider its results. You can't just leave a study out because you don't like it, or else there is publication bias."

Lowe agrees that studies conducted during the 1990s suggested that high CRP levels were associated with about a doubling in risk of CHD. "An early meta-analysis with less than 2000 cases suggested that the odds ratio for the highest vs lowest tertile of CRP was around 2 [3]. But there were 11 other studies on the association of CRP and CHD published between 2000 and 2003, and they all showed similar hazard ratios—around 1.5 — similar to the Danesh result," he noted.

Lowe explained that as a hazard ratio of 2.0 means a 100% increase in risk and a hazard ration of 1.5 means a 50% increase in risk, the latest data show the association between CRP and CHD to be only about half as strong as was previously suggested. "What I expect is that studies in the early meta-analysis suffered from publication bias. As in all fields, the more positive results are always published first, as these are exciting. A new exciting risk factor comes along, and everybody jumps on the bandwagon. After 2000, a lot more studies were reported, which actually might have been available before but were not published, as they did not show such exciting results. So what we are seeing is a regression to the truth. This is a phenomenon we see again and again; it is not unique to CRP."

Dr Paul Ridker

But Ridker says Lowe is missing the big picture. Ridker believes the Danesh paper was actually a very good study but adds: "What Dr Lowe forgets is that the magnitude of predictive value of CRP in the Iceland study was statistically identical to that associated with hypertension and smoking. Given these data, the only logical conclusion one can draw is that Dr Lowe also does not believe that blood pressure and smoking are important risk factors, a position I doubt many would agree with."

Lowe fires back that it is Ridker who is missing the point here. "We know that blood pressure, smoking, and cholesterol are important causal risk factors, which collectively provide the current, useful risk scores. In contrast, CRP has not been proven to be a causal factor; and its additional value to the current risk scores has not been rigorously evaluated, nor has it been shown to change clinical management," he argues.

So what other evidence is there?

Koenig summarized some of the key studies that he believes support the addition of CRP to the usual risk-assessment model. These include a study from Ridker et al that used data from the Women's Health Study and showed that if the Framingham risk categories are further stratified for levels of CRP, there is a further variation in risk [4].

And there is Koenig's own study, published last year, which used data from the MONICA cohort and stratified Framingham risk by CRP level [5]. "All three statistical approaches employed showed great variation in risk in those in the intermediate Framingham category if further stratified by CRP levels," he reported.

He also cites a contemporary study from the UK (the EPIC-Norfolk study) conducted between 1993 and 2003 in a large population of cases and controls [6]. "If all cases were considered, CRP was among the strongest variable predicting risk of CHD. And it was the strongest if only fatal cases were analyzed," Koenig noted. "This was the opposite way round from the Reykjavik study. CRP was the strongest marker, and the other risk factors had no incremental value," he said.

He adds: "If we look at the major prospective studies, all these show positive associations. And if we include CRP in the Framingham risk score, 20% to 40% of all people in the intermediate-risk category are classified more accurately."

What about the elderly and diabetics?

Noting that it has been suggested that CRP is not predictive in elderly and diabetics, Koenig cited two studies refuting this idea. In the Cardiovascular Health Study, conducted in an elderly population, a CRP over 3 mg/L was associated with a more than doubling in the risk of cardiovascular events [7]. And a US study by Dr Shaista Malik(University of California, Irvine) et al in 3800 diabetic patients also showed a twofold increase in risk with high CRP vs low CRP [8].

Studies available still confounded?

Lowe notes that other studies, including Framingham, have reported less favorable findings [9]. He agrees that many studies have shown that CRP is undoubtedly associated with CHD, but he claims this association is "largely coincidental" due to residual confounding. "Most studies will not have adjusted enough for standard risk factors such as smoking, blood pressure, and cholesterol, and the meta-analyses have not generally adjusted for other risk factors such as obesity, alcohol, exercise, and socioeconomic factors. You can adjust only for the data you have. And data on all these factors would not have been available in every study," he says.

He points out that in the British Regional Heart Study, in which 4000 men had lifestyle changes documented and inflammatory risk markers measured every five years, it was shown that smokers have a dose-dependent increase in CRP and other inflammatory markers, and when people quit smoking, it takes about 20 years for them to completely normalize [10]. "Therefore, adjusting for smoking habit in a multivariate analysis is complicated—you have to adjust for the dose (number of cigarettes smoked) and also for the time since they quit, for ex-smokers. And this may not necessarily be well defined," he explained to heartwire.

Lowe notes that in the NHANES study, three quarters of the risk of increased CRP was accounted for by conventional risk factors such as smoking, blood pressure, lipids, glucose, and body-mass index [11]. "Elevated CRP without other risk factors was observed in less than 10% of the population, and therefore CRP may have limited clinical utility as an additional screening tool for CHD risk," Lowe said. "So whether the hazard ratio for CRP is 1.5 or 2, it is still going to be confounded. To really make a difference to CHD risk prediction, we need an odds ratio of about 10. 1.5 is not going to be anywhere near enough. CHD is a multicausal disease—there will be no magic bullet for prediction," he adds.

However, Koenig countered that many of the major epidemiological studies have in fact adjusted for all major risk factors. He cites another study from Finland in 2321 men, in which the risk for cardiovascular mortality associated with CRP over 3 mg/L was still increased twofold, even after adjustment for conventional risk factors, factors related to diet and insulin resistance, and lifestyle factors [12]. And Ridker adds: "If you really believe that a risk factor makes a difference only if the odds ratio is 10 or greater, then you should never screen for cholesterol, blood pressure, or smoking. This kind of thinking is backward and is not going to help stem our epidemic of heart disease and stroke."

CRP no different from other inflammatory risk markers?

Lowe cited several studies suggesting that CRP was no different from other inflammatory risk markers in terms of CHD risk prediction. These included a table of meta-analyses for CHD risk of several inflammatory markers (indirect comparisons), which did not find CRP at the top of the league [14]. "Fibrinogen actually showed the strongest association, and CRP showed similar associations to other commonly used inflammatory markers such as viscosity, serum amyloid A, leukocyte count, etc. There is nothing special about CRP in this respect," Lowe commented.

Lowe also stated that CRP has a large short-term variability. He recently conducted a study in healthy people who had various inflammatory risk markers measured weekly for 15 weeks and found that the mean intraindividual variability of CRP was 50%, compared with just 5% for fibrinogen. "You need to take six blood samples to get a reliable CRP measurement—and the costs and hassle are mounting all the time," he commented to heartwire. Ridker disputes this, pointing out that all epidemiological studies are based on a single blood draw. "Even the controversial data from Iceland demonstrate that the decade-to-decade variation for CRP is identical to that of cholesterol, and that study involved thousands of patients, not a small handful," he noted. He also pointed out that the Centers for Disease Control has shown that for most patients a single CRP test is sufficient.

And Koenig points out that while fibrinogen also has good epidemiological data, it is difficult to translate this into clinical use. "It is very easy to translate a CRP measurement into something useful for clinical practice, as there is an international WHO standard for CRP. The analytical side is very easy. Fibrinogen has no such international standard—there are 15 to 20 different assays, and it is difficult to compare values from different laboratories," he told heartwire. Ridker confirms this view: "The reason no one is measuring fibrinogen is that the assay hasn't been standardized. This was sorted out for CRP a long time ago," he said.

Emerging Risk Factors Collaboration will provide better estimate

All the speakers at the AHA session noted that all current data comparing the various risk factors come from indirect comparisons and different studies and so must be interpreted with caution. Direct comparisons are now being done by the Emerging Risk Factors Collaboration. This involves finding every study published that has measured risk factors such as CRP, fibrinogen, lipoprotein (a), albumin, etc, and asking the authors to pool their data on each of these risk factors. Lowe explained to heartwire that epidemiological studies publish many different papers—some of these data will already have been published, but some study groups may not have gotten around to publishing all their data on all the various risk factors. The Emerging Risk Factors Collaboration has sent letters to the various investigators asking which risk factors were measured in their study and whether they could send in the relevant data.

This has already been done for fibrinogen—in the Fibrinogen Studies Collaboration, on which Lowe was a contributor [15]. "The fibrinogen collaboration took five years from start to finish, but hopefully the CRP data will be available in about two to three years. It is a huge job and involves obtaining all measures of all risk factors from every study, but there is no shortcut. If we want a precise risk estimate for CHD for each risk factor, there is no alternative."

The Fibrinogen Studies Collaboration found that fibrinogen measurement picked up only an extra 4% of CHD cases. "If we measured fibrinogen in 1000 healthy individuals, we would only identify an additional 40 who would develop CHD that would not be picked up from traditional risk-factor screening. This is not a good public-health measure. I doubt that it will be much different for CRP," Lowe said.

Measuring CRP will detract from key health messages?

Lowe believes that even if CRP is shown to be a reliable marker, that doesn't automatically make the case for screening. "Very few risk markers are actually worth measuring in the general population—at present, just blood pressure and cholesterol. My strong feeling is that it will not be worth measuring CRP for the amount of benefit it will give us. It would be better to spend the money on educating people on what we know is good for them—healthy eating, exercise, and not smoking," he told heartwire.

"If CRP is adopted, we must ask how many patients with elevated CRP are going to be sent for full workups to detect underlying disease, and at what cost. Routine screening will cause patient concern and worry about insurance and will detract from key health messages. If everyone is screened and you find you are in the lowest CRP tertile you may think it is okay to eat unhealthily, smoke, and not exercise. We have to tell people to stop smoking and keep walking. The fact that you get some reductions in inflammatory markers due to these lifestyle changes is incidental," Lowe concluded.

Ridker claims this approach is "inconsistent," as CRP has been shown to be as strong a predictor as the other factors Lowe does advocate. "Remember that for many years a number of European investigators insisted that cholesterol screening was also a bad public-health policy. Regardless of how many studies show something works, there will always be people who believe all change is for the worse—even change for the better," he told heartwire.

But even Koenig and Ridker believe that CRP is valuable only if traditional risk factors are also measured. "My concern is that we are not even measuring traditional risk factors enough," Koenig stated. "Many doctors are not using Framingham scores. And new, easy-to-measure markers such as waist circumference and amount of physical activity have also been shown to be useful but are not being used enough. First we must get better at the traditional risk factors. But in my view the database that we have now does suggest that CRP is valuable on top of these," Koenig concludes.

CRP can guide statin therapy

Ridker also highlighted data from the PROVE IT-TIMI 22 trial [16] and the REVERSAL trial [17] showing that the best clinical outcomes and the least progression of atherosclerosis occur among statin-treated patients who not only reduce LDL below 70 mg/dL but also reduce CRP below 2 mg/L. "Very high-risk patients may need a 'dual-goal' strategy that monitors inflammation along with LDL cholesterol if we are to maximize the benefit for our patients," he noted. But he added that primary-prevention data on this issue are not yet available; they will come from the JUPITER trial now under way, which is evaluating statin therapy in individuals with low levels of LDL but high levels of CRP.