Boston, MA - A new analysis of data from the Framingham study has suggested that 10 common contemporary biomarkers, which have attracted interest as possible predictors of cardiovascular risk, actually do not add much to traditional risk factors for the assessment of future risk of cardiovascular events in healthy individuals [1].

The study, published in the December 21, 2006 issue of the New England Journal of Medicine, is believed to be the largest analysis to compare and combine several contemporary biomarkers for cardiovascular risk assessment in healthy individuals. Lead investigator Dr Thomas Wang (Massachusetts General Hospital, Boston) commented to heartwire: "Even in combination, all 10 biomarkers together only moderately predicted risk when used in addition to traditional risk factors for individual patients. They don't add a whole lot. So our results do not provide support for screening large populations of healthy people for raised levels of these biomarkers."

But he cautioned that these results are applicable only to healthy individuals. "We did not look at patients who already had heart disease, and our conclusions do not exclude the possibility that these biomarkers could be useful in selected groups—such as to more accurately stratify patients at intermediate risk as determined by traditional risk factors," he said.

Wang added that the study had proven the value of traditional risk factors for assessing cardiovascular risk and highlighted a need to identify new biomarkers that may be better in terms of cardiovascular risk assessment in healthy individuals. "The biomarkers that we focused on are not good enough. There are many other possible candidates that have been identified, and it's possible that some of these may do a better job, so we need more studies to find out," he said.

But CRP advocate Dr Paul Ridker (Brigham and Women's Hospital, Boston, MA) objects to the conclusions drawn by Wang et al and questions their method of assessing the clinical utility of the biomarkers.

In the study, 10 biomarkers were measured in 3209 participants attending a routine examination cycle of the Framingham Heart Study. The biomarkers were: C-reactive protein (CRP), B-type natriuretic peptide (BNP), N-terminal proatrial natriuretic peptide, aldosterone, renin, fibrinogen, D-dimer, plasminogen-activator inhibitor type 1, homocysteine, and the urinary albumin-to-creatinine ratio.

During a median 7.4 years of follow-up, 207 participants died and 169 had a first major cardiovascular event. After adjustment for conventional risk factors, five biomarkers were found to predict the risk of death and two biomarkers were found to predict the risk of cardiovascular events, with BNP and the urinary albumin-to-creatinine predicting both the risk of death and the risk of cardiovascular events.

"The best of the biomarkers we examined was BNP, but even with BNP the data were not strong enough to support routine measurement in healthy people for future cardiovascular risk assessment," Wang commented to heartwire.

He noted that even when all the biomarker scores were combined into a "multimarker" score, this was not much more helpful in predicting risk than conventional risk factors alone. "Although high multimarker scores were associated with fairly large increases in relative risks—a fourfold increase in the relative risk of death and a twofold increase in the relative risk of cardiovascular events—these statistics are not very helpful when applied to the individual patient," Wang said. "These relative risks give us the increased risk in one group of people compared with another group. But a doctor wants to know the risk of an individual person sitting in the office, and these relative risk values do not necessarily apply in that situation," he added.

He explained that the improvement in the ability to predict risk with new markers when added to conventional risk factors is measured by an index known as the C statistic. "The C statistic for a test can vary from 0.5, which would signify no use in prediction, to 1.0, which would be the perfect test. How much these biomarkers increase the C statistic will tell us how much they add to risk assessment. Traditional risk factors such as blood pressure, cholesterol, and smoking status give a C statistic of around 0.76—that is pretty good. If we add the information from the 10 biomarkers we looked at, the C statistic increases only by about 0.01, to 0.77," Wang commented to heartwire.

In an accompanying editorial [2], Dr James H Ware (Harvard School of Public Health, Boston, MA) reemphasizes this point, noting that despite the strong association between a risk factor and the disease outcome, it does not follow that the risk factor provides a basis for an effective prediction rule for individual patients. He says that the group of biomarkers studied by Wang et al "makes a substantial contribution to the proportional-hazards model for predicting death from any cause, but it is of limited value for the risk stratification of individual patients. This scenario has unfolded repeatedly as we have discovered new biologic variables that lie on the complex pathway leading to chronic disease and death. The work of Wang and colleagues, however, shows us how difficult it is to achieve effective risk stratification with respect to multifactorial disease processes. Much work remains to be done before biomarkers of the type the authors consider here can provide a basis for prognostic evaluation of the individual patient."

The best known of the biomarkers in this study is probably CRP, which did predict death but was not necessarily the best of the 10 biomarkers in this respect, Wang says. "Some other markers performed better than CRP in our analysis, but this does not mean that CRP does not have value in other situations. For example, proponents of CRP believe it is useful to further define risk in patients judged to be at intermediate risk from traditional risk-factor assessment. We did not look at this, so our results do not apply to this scenario. Also, CRP may be useful for identifying people who need statin treatment, which the ongoing JUPITER trial is testing. And our results have no bearing on that either."

The results of this study are similar to those of a previous study of CRP by Danesh et al published in 2004 [3], which analyzed CRP data from the Reykjavik Study and showed that people with a raised CRP level had an odds ratio of developing heart disease of 1.45 after adjustment for conventional risk factors, lower than had been reported in previous studies. Wang commented: "The two studies used different methods of defining raised biomarker levels—Danesh et al assessed risk based on the tertile of biomarker, whereas we looked at risk with each standard-deviation increase, but we reached similar conclusions about CRP."

But Ridker objects to the conclusions drawn by Wang et al, saying that the C statistic is not the best way of assessing clinical utility of new biomarkers.

"Given that many statisticians feel the C statistic is an ill-advised method for analyzing prospective cohort data and selecting variables for prediction-model inclusion, I am quite surprised that the authors would continue to rely on this technique," he commented to heartwire. "It can be shown within the Framingham data themselves that if the C statistic were used as the criterion for selecting 'clinically useful' risk predictors, we would be forced to abandon LDL cholesterol, HDL cholesterol, and blood pressure from our risk-prediction models. It thus simply does not make sense to claim that a failure to improve the C statistic tells us much, if anything, about clinical utility," he added.

Statistician Dr Nancy Cook (Brigham and Women's Hospital), who works with Ridker, agrees. She told heartwire: "In the paper by Wang et al, the multimarker score has a very large effect on the predicted risk of death and would clearly have a clinical impact since it could lead to different treatment decisions for individual patients. This paper and this field in general seem to rely only on the C statistic for evaluating risk-prediction models. This measure is insensitive and misused and should not be the sole basis for selecting models for risk stratification."

Ridker continues: "The crucial clinical question is not 'What does my C statistic look like?' but 'How many patients at intermediate risk by traditional Framingham scoring are correctly reclassified by the new biomarkers as being at higher or lower risk?' This is the only question that matters for clinical care. A continued failure to correctly analyze these kinds of data at a national guideline level has the potential to set prevention back a decade."

In response, Wang defends his use of the C statistic, saying it is "a widely used and understood metric, which has been employed for decades to evaluate diagnostic and screening tests, including the original Framingham risk score."

He also makes the point that an important rationale for including traditional risk factors in the Framingham risk score is that they play a causal role in the development of heart disease and that treating elevated levels of these risk factors has been shown to reduce cardiovascular risk in multiple clinical trials. "But we do not know whether treating elevated levels of novel biomarkers reduces the risk of heart disease. Dr Ridker and colleagues are performing such trials, and we look forward to the results. In the meantime, however, the only clinical reason for measuring novel biomarkers routinely for prevention would be if they added to our ability to predict risk. Our data suggest that they add only modestly, when considering a community-based population of people with heterogeneous risk," he says.

Wang agrees with Ridker that CRP and other biomarkers may be useful for the reclassification of certain patients but notes that "this still relies on the premise that treatment of the individuals who are reclassified would result in better outcomes, and data from randomized clinical trials are needed to address this point."

Wang concludes: "The point of the article is not that novel biomarkers should be abandoned. They have added and will continue to add important insight into our understanding of the origins of heart disease and potential future therapies. However, they are not ready for prime time when it comes to routine clinical use in the preventive setting. Our findings also emphasize the importance of research on newer biomarkers that can add to what we already have."