Point/Counterpoint

David C.W. Lau, M.D., Ph.D. and Steven A. Gover, M.D., M.P.A.

Yes, CRP adds value to traditional risk factors

The inflammatory marker C-reactive protein, synthesized in the liver and fat tissue, is a powerful activator of the classic complement system. Plasma C-reactive protein levels increase with obesity and decrease with weight loss (Can. J. Cardiol. 2006;22[suppl. B]:85B-90B). Although the precise function of C-reactive protein is unknown, recent evidence suggests that it is a direct participant in the dysfunction leading to atherosclerosis (Am. J. Physiol. Heart Circ. Physiol. 2005;288:H2031-41).

More than 20 prospective studies suggest that elevated plasma C-reactive protein (CRP) is one of the strongest independent predictors of cardiovascular disease (CVD) risk in asymptomatic patients. Six cohort studies have shown that CRP adds prognostic value beyond the Framingham risk score, while eight additional cohort studies have linked CRP with the development of both metabolic syndrome and type 2 diabetes.

Among those data are a 2004 study by Ridker and Cook, in which CVD risk among 27,939 apparently healthy people increased linearly from the lowest to the highest levels of high-sensitivity CRP (hsCRP), even after adjustment for Framingham risk score (Circulation 2004;109:1955-9). In the Strong Heart Study, hsCRP was a predictor of CVD in a population with a high prevalence of diabetes and other risk factors (Circulation 2005;112:1289-95). And in a prospective German study, CRP was strongly associated with a first major coronary event among 936 initially healthy middle-aged men (Circulation 1999;99:237-42).

Taken together, the data suggest a relative risk for cardiovascular events of 1.8 to more than 4.0 for individuals in the highest quartile of CRP, compared with those in the lowest quartile.

Importantly, CRP adds value to the Framingham risk score, particularly among people identified by the score as being at moderate CVD risk. In the prospective follow-up of 27,939 apparently healthy women in the Women's Health Initiative trial, screening for CRP appeared to identify a different high-risk group than did LDL cholesterol, and independent effects were observed for CRP in an analysis adjusted for all components of the Framingham risk score (N. Engl. J. Med. 2002;347:1557-65).

Among 3,435 men in the German cohort, CRP measurement added clinically relevant information to the total cholesterol/HDL cholesterol ratio. For individuals in the Framingham risk score range of 10%-20% in particular, a CRP value of 3 mg/L or above was highly predictive of 10-year event rates (Circulation 2004;109:1349-53).

Two studies published in 2005 showed that patients with lower CRP levels after statin therapy have better clinical outcomes than those with higher CRP regardless of LDL cholesterol level, suggesting that levels of both should be monitored in patients taking statins (N. Engl. J. Med. 2005;352:29-38).

In 2003, the American Heart Association and the Centers for Disease Control and Prevention recommended hsCRP as an adjunct to major risk factors to further assess absolute risk for primary prevention in persons at moderate 10-year coronary heart disease risk (Framingham risk score 10%-20%), and as an independent marker for assessing the likelihood of recurrent events in patients with stable coronary disease or acute coronary syndrome (Circulation 2003;107:499-511). This year, the Canadian Cardiovascular Society issued a similar statement (Can. J. Cardiol. 2006;22:913-27).

Only half of all heart attacks can be explained by traditional risk factors, so clinicians need additional tools to predict coronary heart disease and guide its management. For that, CRP is an excellent candidate.

No, CRP testing is not worth the additional cost

We already have a set of risk factors that are highly predictive of a patient's future risk for cardiovascular events. The data do not support the notion that C-reactive protein adds significantly to our prognostic abilities.

In 1995, my colleagues and I published an analysis comparing the then-new National Cholesterol Education Program Guidelines with the previous ones. We found that, compared with the individual risk factors of total, LDL, and HDL cholesterol and triglycerides, the ratios of TC/HDL and LDL/HDL improved the sensitivity, from 45% to 70%, while only slightly reducing specificity, from 86% to 82% (JAMA 1995;274:801-6).

When all the Framingham risk factors were considered in a multivariate risk model, the model correctly discriminated between those who would and would not develop fatal coronary heart disease 85% of the time. Although several studies have shown that CRP may be an independent predictor of CVD, most of these studies could not completely adjust for the classical Framingham risk factors.

For instance, investigators compared risk factors including CRP at baseline with those at a mean of 12 years' follow-up in 379 participants enrolled in the Reykjavik prospective study. After adjustment for many established risk factors at baseline, the odds ratio for CHD was 1.45 among those in the top tertile of CRP vs. those in the bottom tertile--an odds ratio lower than what had been reported previously (N. Engl. J. Med. 2004;350:1387-97).

Only total serum cholesterol levels were measured in the study, rather than those of its subfractions, which have opposing effects on CHD risk. As we showed in 1995, this would be expected to underestimate the predictive ability of lipid levels and potentially overestimate the predictive value of CRP. The investigators acknowledged this possibility in their article.

Many factors modulate an individual's CRP level: Obesity, high blood pressure, cigarette smoking, diabetes, and infection all increase CRP, while moderate alcohol, physical activity, weight loss, and drugs such as statins, fibrates, niacin, and thiazolidinediones reduce it. The potential for confounding is huge.

In one of the few studies that adjusted for all the classical Framingham risk factors, 1,949 men and 2,497 women in the Framingham Heart Study who were free of CVD were evaluated. A little more than a third--35% of the men and 41% of the women--had CRP levels greater than 3.0 mg/L, while 38% and 35%, respectively, had CRP levels below 1.0 mg/L. Compared with the adjusted relative risks for major coronary heart disease events of 2.16 for diabetes and 1.59 for current smoking, the 1.22 risk for CRP greater than 3.0 mg/L was nearly the same as the 1.20 risk for the TC/HDL-C ratio (Arch. Intern. Med. 2005;165:2473-8).

Importantly, the "C-statistic," a measure of the discriminatory capability of the prediction models, was 0.78 for the traditional risk factors (systolic blood pressure, TC/HDL-C, diabetes, current smoking, and blood pressure treatment) in predicting major CVD events, with or without the addition of CRP. Similarly, the C-statistic was 0.80 for major coronary heart disease event, again identical whether CRP was included in the model or not.

These findings tell us that adding CRP makes little difference in our prognostic capability. This was true even for individuals in the middle (10%-19.9%) of the Framingham risk range.

Should we really be adding another clinical test to our practices for little, if any, incremental benefit? I don't think so.