Key result
C-reactive protein testing provides minimal gains in discrimination and risk reclassification for coronary heart disease, leaving its incremental predictive and clinical utility equivocal.
Answers are grounded in this paper and cite their sources.
This commentary highlights that despite strong associations with CHD, CRP provides only modest incremental predictive utility over standard risk factors, emphasizing the need for rigorous evaluation of clinical usefulness for novel biomarkers.
The only relevant test of the validity of a hypothesis is comparison of prediction with experience Milton Friedman The last three decades have seen major advances in our understanding of the pathogenesis of coronary heart disease (CHD), with substantial improvements in our management of patients at risk of disease and of those who have suffered a coronary event. The multifactorial nature of coronary disease and its evolution as a life course disease1 due to the combined influences of genetic and lifestyle-related factors are firmly accepted concepts. We no longer argue about the role of cholesterol and its components as critical mediators of the disease, nor dispute whether the occlusive thrombus in the coronary tree in a patient with a myocardial infarction antedates or follows the clinical event. However, the debates in cardiology continue to rage with a shifting of the themes. One such contemporary controversy centres on if, when, and how novel biomarkers can be used incrementally over standard risk factors for predicting and preventing CHD.2 The key role of subclinical atherosclerosis as a precursor of overt coronary events is well established.3 There is substantial evidence to indicate that the sequence of events that mediate the transition from fatty plaques through the stage of vulnerable plaques to the complicated plaques (that characterize acute coronary syndromes) involve the activation of numerous biological pathways/processes: lipid metabolism, endothelial dysfunction and injury, vascular wall remodelling, thrombotic and haemostatic factors, neurohormonal activation, to name just a few.4 These diverse processes can be studied clinically using a plethora of biomarkers.2 The firm grounding of the concept that inflammation is a fundamental component of atherosclerosis5 has led to the expectation that measurement of circulating levels of one or more inflammatory markers could be potentially useful for predicting CHD risk. A lot of the work in the published literature has revolved around the C-reactive protein (CRP), the inflammatory marker most-written about. Yet, the literature on the utility of CRP for risk prediction has been equivocal. Indeed, an expert panel6 concluded recently: ‘Specifically, those patients at intermediate risk (e.g., 10% to 20% risk of CHD over 10 years), in whom the physician may need additional information to guide considerations of further evaluation (e.g., imaging, exercise testing) or therapy (e.g., drug therapies with lipid-lowering, antiplatelet or cardioprotective agents), may benefit from measurement of hs-CRP (Class IIa, level of evidence B).’ The equivocation stemmed largely from the lack of clarity based on review of the literature regarding what incremental utility CRP testing actually offered. Before a discussion on predictive utility of CRP can occur, there has to be a consensus on how best to assess the incremental predictive utility of any biomarker of CHD risk. It goes without saying that a pre-requisite for assessing such usefulness of a biomarker is the demonstration of a consistent and robust association with CHD,2 a precondition well satisfied by CRP. Figure 1 displays some of the common metrics used to assess the incremental predictive utility of biomarkers. It is readily evident that such assessment requires the use of several metrics, largely because each individual metric has its inherent limitations7 (see reference 7 for detailed review). Indeed, newer metrics (such as net reclassification or discrimination indexes) were formulated to address deficiencies of existing ones.8 Thus, increments to the c-statistic with addition of biomarkers tend to be quite modest,9 with a lack of clarity regarding what constitutes clinically meaningful increments. The calibration statistic is useful to compare predicted and actual risks for models with and without novel biomarkers, but again the clinical importance of statistically significant gains in calibration are not readily evident.7 The net reclassification index8 is an improvement on simple assessment of reclassification, yet is sensitive to choice of the categories used to define risk strata, and in some formulations may weight equally any shift between adjacent categories,7 a contestable premise because risk of disease is greater in the higher categories. Assessing the predictive utility and clinical usefulness of a new biomarker In this context, the investigation by Shah et al.10 in this issue of the journal provides a comprehensive synthesis of studies evaluating the predictive utility of CRP for CHD, including the contribution of original data from the Northwick Park and Edinburgh Artery Studies. Two features that are striking about this landmark effort are the review of data from 31 studies, making this the largest body of data on CRP and its predictive utility, and the use of a comprehensive panel of metrics to evaluate such utility. Not surprisingly, the verdict continues to be equivocal: minimal gains in the c-statistic, very modest increase in calibration of unclear clinical significance, and small gains in risk reclassification that would alter treatment in very few people. The performance in terms of DR5 (detection rate at a 5% false-positive rate) is also disheartening. Two caveats should be noted when interpreting these data. One, it is unclear if there exists a biomarker that will fare better if assessed using this panel of performance metrics. Such an evaluation must be extended systematically to other promising CHD biomarkers such as the B-type natriuretic peptide, urine albumin excretion or mid-regional proadrenomedullin (to name just a few). Second, a couple of recently published studies11,12 could not be included in the present report, and these studies do seem to suggest modest gains in risk reclassification with the addition of CRP in one,12 and with the addition of CRP and parental history in the other;11 unfortunately, the data from the latter study11 did not permit an assessment of the incremental utility of CRP distinct from that of parental history since both were modelled conjointly. Iterative reviews with the addition of these and other new studies will likely continue to refine our knowledge. It is important to note that the ultimate goal of any biomarker evaluation is to assess its clinical utility. For this purpose, we would need data on not just associations and predictive utility, but also on clinical usefulness, i.e. does knowledge of a biomarker (biomarker-based strategy) change the management of a patient compared with the current standard of care (blinded to biomarker knowledge).2 Such an investigation would require a randomized clinical trial. Indeed, even if a biomarker provides incremental predictive utility, it is unlikely to be cost-effective unless it changes the management of the patient at risk of CHD. Are such data available for the clinical usefulness of a CRP-based strategy vs a conventional strategy in the primary care setting? In this context, the results of the recent JUPITER trial13 have been hailed as providing initial evidence of the clinical usefulness of CRP testing in patients with relatively normal LDL cholesterol levels, given the lower cardiovascular event rates associated with rosuvastatin use (compared with placebo) in people with a CRP ≥ 2 mg/l. However, an accompanying editorial14 cautioned that the JUPITER trial was a ‘statin’ trial, and not a ‘direct test’ of a CRP-based management strategy (as there were no trial arms that did not use CRP testing or that randomized people with relatively lower CRP levels). Further, nearly 80% of those screened did not meet the inclusion criteria for the actual trial,13 raising issues about the ‘number-needed to screen’7 with CRP to prevent one cardiovascular event if the test were to be used routinely in primary care settings based on this trial. Yet, undeniably JUPITER13 does provide indirect evidence for the potential clinical usefulness of CRP in a subgroup of individuals in this setting. It is important also to note that another recent landmark study questioned if CRP was causally related to CHD using a Mendelian Randomization approach.15 So, we are left with a biomarker that is a marker of CHD risk but may not be causally related,15 provides very modest (if any) gains in risk prediction,10 and with some indirect evidence of clinical usefulness based on a clinical trial.13Quo vadis? We are entering into an era where incremental predictive and clinical utility of biomarkers will be assessed using very stringent criteria and rigorous scientific design. The fact that a well-researched biomarker such as CRP should ‘stumble at the gate’ rings warning bells to all of us engaged in the quest for novel biomarkers that add to CHD risk prediction beyond the standard risk factors. National Institutes of Health/National Heart, Lung & Blood Institute (Contract N01-HC-25195 and 2K24HL04334 to R.S.V.).
No takes yet. Share an insight, caveat, or question.
Ramachandran S. Vasan (2009) conducted an editorial in Coronary heart disease. C-reactive protein (CRP) testing was evaluated. C-reactive protein testing provides minimal gains in discrimination and risk reclassification for coronary heart disease, leaving its incremental predictive and clinical utility equivocal.
Synapse has enriched 10 related papers. Consider them for comparative context: