High-Sensitivity C-Reactive Protein & Cardiovascular Disease
by William R. Ware, Ph.D.Emeritus Professor of Chemistry, University of Western Ontario
There is a rapidly growing body of evidence suggesting that atherosclerosis is in part an inflammatory condition . C-reactive protein (CRP) is a serum marker for inflammation, which is easily and inexpensively measured and assays have recently been developed that provide very high sensitivity. Numerous epidemiologic studies have consistently shown that high-sensitivity C-reactive protein (HSCRP) levels provide a strong and independent indication of risk of future heart attacks, ischemic stroke (due to obstruction, not bleeding), and peripheral arterial disease, even among individuals who are thought to be free of vascular disease . This article reviews the arguments both for and against the proposal that its measurement might advantageously be included in blood work associated with physical examinations or cardiovascular risk assessments, even for individuals with no indication of cardiovascular disease (CVD) and those thought to be a low risk on the basis of conventional parameters.
The term acute-phase response is used to describe the greatly increased synthesis and secretion of certain plasma proteins including CRP, principally by the liver, following trauma, tissue necrosis (tissue death, which for example may occur as a result of a heart attack), infections, and the acute effects of inflammatory diseases . Inflammation releases cytokines, and the cytokine Interlukin-6 is thought to be largely responsible for triggering the production of CRP, which is thus generally viewed as a marker of inflammation, with very high values associated with acute inflammation. Chronic, low-level inflammation can result in near normal but nevertheless elevated values.
The intense interest in CRP in the context of coronary heart disease, peripheral vascular disease and stroke is quite recent, although the protein was discovered about seventy years ago. Its original use in differential diagnosis and for following the course of treatment, which has always been controversial, appears to be limited today, at least in North America. The index of Harrison's "Principles of Internal Medicine", 14th edition published in 1998, contains no mention of CRP. Until recently, the assay used for CRP had very low sensitivity, with the result that values near the limit of detection, i.e. 3-10 mg/L, were considered normal, which was not a problem, since the acute-phase response is generally accompanied by very large increases in serum CRP levels, with values well over 100 mg/L common. As the cause of the inflammation resolves, the level usually declines over a few weeks to the normal range.
As with cholesterol levels, which are sometimes expressed in mg/dL and sometimes in mmole/L, a lack of standardization is also evident in the case of CRP. Some researchers express the value in mg/L while others use mg/dL. Fortunately, the conversion is easy. One-tenth of the value in mg/L gives mg/dL. Because of the potential for confusion, it is essential that the unit always be specified when referring to CRP values. In fact, there are errors in a few CRP papers because of inadequate attention to this problem.
Today the determination of serum CRP is generally accomplished with an assay of high sensitivity (HSCRP). The range of "normal" values found in one study  of over 5000 presumably healthy individuals varied between a low of 0.1-0.7 mg/L (found in 20% of the population) and a high of equal to or greater than 3.8 mg/L (found in 20% of the population). The median CRP (CRP will be used to indicate HSCRP in what follows) was 1.6 mg/L. Similar values have been obtained for presumably healthy individuals in other studies using modern high-sensitivity assays. These numbers clearly are very low compared to the old "normal" value of <10 mg/L, a value that still appears on some clinical laboratory printouts as a "reference range" even for the high sensitivity assay.
Chronic inflammatory diseases such as rheumatoid arthritis, as well as surgery and trauma, and infections, both chronic and otherwise, can cause CRP levels to be elevated well above the 3.8 mg/L. . In addition, low-level chronic inflammatory conditions that may produce no noticeable symptoms can mildly elevate CRP values. Hormone replacement therapy has been found to increase CRP levels , and women, especially older women, on average appear to have slightly higher normal values as compared to men. Overweight, obese and diabetic individuals typically have elevated serum CRP  and elevated CRP levels have been found to predict the development of adult-onset (type 2) diabetes . Consistent with this is the observation that individuals with insulin resistance syndrome typically exhibit elevated CRP levels . Living in an environment where air pollution is high can raise CRP values . It is well known that smoking elevates serum CRP. At least two studies have shown that patients with atrial fibrillation have higher than normal CRP levels [10,11].
Cytokines responsible for acute-phase protein production also are known to play significant roles in neuropsychologic function and dysfunction [12,13]. As Kushner points out , it is possible that detecting CRP elevation in some individuals is merely an indirect way of detecting depression, a known risk factor for coronary artery disease. Recently reported research supports this view . Thus CRP is clearly a non- specific marker of inflammation. As will be discussed below, the risk of future vascular disease increases continuously across the "normal" CRP ranges. There are a number of conditions, examples of which have been given above, that can either temporarily or chronically elevate CRP well above either the median or the lower limit of the highest quintile (the value of 3.8 mg/L given above); this complicates the task of interpreting serum CRP levels. In fact, the precise nature of the events that trigger increased CRP production that in turn predict future adverse cardiovascular (atherothrombic) events have not yet been identified .
CRP AS AN INDICATOR OF CARDIOVASCULAR DISEASE RISK
Studies carried out in the last few years on CRP and cardiovascular disease have involved both individuals at high and low risk of disease, but nevertheless considered disease free, as well as individuals with established disease, i.e. individuals who have experienced a heart attack or stroke, have stable or unstable angina, or who have peripheral vascular disease e.g. intermittent claudication. The studies discussed below are of particular interest in the context of this review, which is concerned with primary prevention rather than the use of CRP levels as a prognostic tool for individuals who have experienced an acute cardiovascular event.
One type of study, called a prospective or cohort study, attempts to correlate the incidence of a future health problem in an entire study population with parameters established at the start of the study, such as blood markers, smoking, body mass index, etc. Another type, the so-called case-control study, is frequently carried out within a prospective study, where a population of apparently healthy individuals provides the cases. That is, at a later time, individuals exhibiting a disorder of interest (the cases) are studied by comparing, for example, blood markers at baseline (at the start of the study) such as cholesterol or CRP, with those of a group selected from the same population who are thought to be free of the disorder (the controls).
A prospective study just published in the "New England Journal of Medicine"  compares CRP with LDL cholesterol levels in the prediction of first cardiovascular events in women. American women, deemed healthy, were enrolled between November 1992 and 1995 in the Woman's Health Study and followed for a mean of eight years for the first occurrence of myocardial infarction, ischemic stroke, coronary revascularization (angioplasty or coronary bypass), or death from cardiovascular causes, and these so- called endpoints were then correlated with the baseline blood levels of LDL, CRP, and as well, the calculated ten-year risk of coronary heart disease based on the Framingham risk score , which is commonly used for clinical evaluation.
After adjusting for confounding, the relative risks of a first cardiovascular event according to increasing quintiles (a quintile contains 20% of the study population) of CRP were 1.4, 1.6, 2.0 and 2.3 as compared to the first quintile. The corresponding numbers for LDL were 0.9, 1.1, 1.3, and 1.5. Thus CRP would appear to be a more significant marker than LDL. Also, a very interesting result was that 77% of all adverse cardiovascular events occurred among women with LDL levels below 160 mg/dL, and 46% occurred among those with LDL levels below 130 mg/dL. Thus a large proportion of the 27,939 women involved in the Woman's Health Study who suffered first adverse cardiovascular events had LDL levels at baseline that were below the threshold values for intervention and treatment under the current guidelines of the National Cholesterol Education Program (NCEP) .
If the results are viewed in terms of tertiles (a tertile contains 33% of the study participants) of CRP vs. tertiles of LDL, the relative risk associated with an LDL level greater than 160 mg/dL along with a CRP level greater than 3 mg/L was about 2.8. If a participant was in a high Framingham risk category (equal to or greater than 10% for the ten year risk of coronary heart disease-- see attachment for a Framingham risk calculation table), then the three tertiles of CRP alone (less than 1.0, 1.0-3.0, and greater than 3.0) gave adjusted relative risks of approximately 12, 21 and 24 (, Figure 4). These risk factors, which essentially combine CRP levels with age, smoking status, systolic blood pressure, as well as total and HDL cholesterol levels, are obviously huge and command attention. It is worth mentioning that for men the Framingham point scores for 10-year risk place considerable emphasis on age. For a man in the age range 70-74, a 10% 10-year risk of a heart attack or coronary death results from age alone, and can be reduced to 8% only by having a HDL value equal or greater than 60 mg/dL (1.55 mmole/L). Also interesting is that age alone for a woman in the same 70-74 range results in a 10-year risk of only 2%.
A case-control study where the patients were drawn from the Physicians' Health Study (PHS), was reported in 1997 . The PHS involved a cohort of over 22,000 male physicians between the ages of 40 and 84 years. Each participant who provided an adequate baseline plasma sample and who experienced an acute cardiovascular event (heart attack, stroke or venous thrombosis) was matched with one normal control selected randomly from the study participants who met the matching criteria of age, smoking history, and length of time since the beginning of the study. Men in the highest CRP quartile (greater than 2.1 mg/L) had three times the risk of a heart attack and two times the risk of an ischemic stroke when compared to men in the lowest quartile (less than 0.56 mg/L).
There are a number of other studies in the literature on the relationship between CRP levels and cardiovascular disease. Ridker  has reviewed nine such studies published between 1996 and 2000. Relative risks calculated by comparison of the top vs. bottom quartiles yield results that range between 2.3 and 4.6. The data summarized provide consistent and convincing evidence that CPR is a risk factor independent of the TC/HDL ratio (ratio of total cholesterol to HDL cholesterol) in both men and women, with very high relative risks for the combined effect of elevated CPR and a high TC/HDL ratio, in fact a risk ratio of 8 to 9 when the data is separated into quintiles. Of the markers of cardiovascular risk, i.e. CRP, LDL cholesterol, the TC/HDL ratio, homocysteine, and lipoprotein (a), the combination of CRP and TC/HDL provided the most powerful indicator of future risk.
Rifai and Ridker recently have proposed a cardiovascular risk assessment algorithm based on combined data from the Women's Health Study and the Physician's Health Study . Only three parameters, CRP, total cholesterol (TC) and HDL are used, with the latter two expressed as a ratio. Quintiles rather than quartiles are employed to obtain the overall risk as assessed from these two parameters.
According to this proposed assessment, a CRP level greater than 3.9 mg/L combined with a TC/HDL ratio greater than 5.5 (men) -5.8 (women) would confer a cardiovascular risk level 8.7 times greater than the risk for a person with a CRP level of 0.1-0.7 mg/L and a TC/HDL ratio of less than 3.4. The detailed relative risks by quintile of CRP and TC/HDL can be found in the attachment.
The numbers in the attachment clearly illustrate the continuous increase found in the risk of cardiovascular disease with increasing serum CRP, independent of the TC/HDL ratio, such that even if one is in the lowest quintile of TC/HDL, those in the highest CRP quintile have a relative risk of 2.2 times those in the lowest quintile. The same pattern is seen with the ratio of total cholesterol to HDL at constant CRP levels. Note that there is significant risk associated even with the "normal" population median CRP of 1.6 mg/L . It should also be remarked that other classical risk factors such as hypertension, smoking, LDL cholesterol and age are not use in this scheme of risk assessment, although total cholesterol is an approximate surrogate for LDL.
Rifai and Ridker suggest  that in clinical practice, CRP evaluations should be avoided if there has been recent infection or trauma. They suggest that two weeks is sufficient for a return to basal values. However, they go on to comment that CRP values greater than 15 gm/L indicate active or perhaps transient inflammation, and suggest two CRP measurements a month apart provide a better picture, especially if a CRP value greater than 5 is obtained on the first determination. The implication of the guidelines is that the clinical assessment requires a more or less stable value, be it 0.1 or 10 mg/L, i.e., it is desirable to avoid clinical decisions based on an abnormally high value that is transitory.
CRP AS A SCREENING TEST FOR CVD
The ideal screening test consists of results giving two non-overlapping distributions of the measured value of the diagnostic, with one distribution for those who are disease free, and the other for those with the disease . Distribution overlap produces false positives and false negatives. In the case of CRP it does not appear to have meaning to view the test in this fashion, since the CRP distribution, expressed in quartiles or quintiles, which is being proposed for assessing risk [4,19] is simply a single, smooth distribution. By the same token, one can question the use of the terms normal, healthy, and disease free (in the context of cardiovascular disease) to describe the subjects who generated baseline CRP distributions quoted in the literature. To take an extreme view, one could argue that while there were no observable indications of disease, a significant fraction of the whole "normal" group (presumably quartiles 2 to 5) carried an elevated future risk factor for CVD, independent of their blood lipid profile, and were thus not necessarily disease free. One could also imagine the possibility that this is consistent with the widespread early onset of atherosclerosis [21,22], which at the beginning can only be detected at autopsy or with invasive techniques. In one such study using intravascular ultrasound, an amazing 17-21% of individuals aged 13 to 19 had evidence of atherosclerosis, and the numbers jumped to between 37% and 60% for the age group 20 to 29 .
A related question is simply, why does the so-called normal, healthy population have the observed distribution of CRP values in the first place ? There does not appear to be a detailed and useful answer. There is also the very important question of whether the circulating CRP is a risk factor per se, or just a marker, which can lead to confounding in the context of cardiovascular risk assessment by some inflammatory conditions but not others. Put another way, if an individual has a CRP value of, let's say, 5 mg/L which is reproduced to within 10-20% a month later in a second measurement, is this high CRP indicating some inflammatory condition which might have no direct connection to the risk of cardiovascular disease in this isolated case, or is any sustained source of inflammation significant, in the context of CVD, for anyone? Unfortunately, while much research has and is being done, and some answers have been proposed to these questions, these answers are really incomplete, hypothetical or so-called hypothesis generating. These questions in fact define a significant portion of the research frontier in this field, and the absence of evidence-based answers has caused some experts to take a position of "wait and see" in connection with the use of CRP for general screening.
Studies that attempt to define the distribution of CRP in the "normal, healthy" population appear to contain a hidden variable, i.e. age. Kushner  has pointed out that a presumably non-inflammatory cause of CRP elevation is simply biologic aging. Wu et al , in a small study of 171 females and 183 males, found mean CRP levels of approximately 0.6 and 0.32 mg/L for males and females, respectively, in the age range of 20- 29, whereas in the age range of 70-79, the comparative mean values were approximately 1.3 and 1.7 mg/L. As a function of age, the values were roughly constant between 20 and 49 (mean 0.43 ± 0.42) and then jumped dramatically at about 50 years to give the much higher values (mean 1.3 ± 1.27), with women now higher than men, on average. There is considerable additional evidence that a minimal acute phase response may be a marker of biologic aging .
If only one CRP measurement is made, a reasonable question to ask regards the variation one might expect if instead, a series of measurements were made over an extended period. Clearly, the ideal test would give a set of values with, hopefully, a small enough variation so that one would remain in a given quintile, or at least show a variation of less than the average width of the lower quintiles. Because the measurement of CRP is being seriously proposed as a component of risk assessment for CVD, there have been several studies that address this obvious and important question [26,27,28,29,30,31]. In one study  of 10 men and 10 women, deemed healthy, aged 24-58, a series of CRP measurement over six months showed substantial intra-individual variations. If we use Rifai and Ridker's  quintiles, then 20/20 individuals exhibited a variation that covered two quintiles, and 17/20 covered three quintiles. Six individuals had a variation that spanned all five quintiles (approximate values taken from figures, references [27,28]. In a recent study by Ockene et al , of 113 individuals studied over a year, the within-subject standard deviation was 1.2 mg/L. Thus, if one had an average of say 2.0 mg/L, then just one standard deviation up or down would yield values of 0.8 and 3.2 mg/L, which covers three Rifai-Ridker quintiles.
It does not seem practical to base a risk assessment protocol intended for general use on a large number of repeated tests. Two tests, spaced a month or two apart, would seem to be the practical limit in the normal clinical setting. The study by Ockene et al  also looked at the agreement between just two measurements three months apart. They observed that while about 63% of the participants remained in the same quartile (note, not quintile), 28% moved up or down a quartile, and 7% moved up or down two quartiles. These results are disturbing from the standpoint of using CRP as a tool for risk assessment, although if a patient tests in the lowest or highest quintile or quartile, this information can be used with more confidence than if the result was in a middle quartile or quintile, where the intra-individual variation could conceal the fact that the patient could possibly be in fact at high risk or low risk. Intra-individual variation may prove to be a major stumbling block in the use of CRP for CVD risk assessment, and highlights the difference between looking at the general picture presented by studies involving a large number of subjects and the problem of interpreting the casual, single measurement for a given individual. The intra-individual variation also underscores the fact that CRP is a non-specific marker of inflammation and that a high value can be caused by an inflammatory condition that may be transitory or asymptomatic.
The key question, however seems to be--is elevated CRP from any cause a risk factor for CVD? While a direct pro-inflammatory effect of CRP on human endothelial cells has been observed in the laboratory , and while CRP has been found to mediate LDL uptake of macrophages implicated in atherosclerosis , studies like these and others [34,35,36] represent just the initial stages of research in this area, and no one knows the final answer.
A number of other papers, short communications and editorials that have appeared in the medical-scientific literature in the past several years have expressed concerns regarding the proposed use of CRP as a screening tool for CVD [37,38,39,40]. The consensus, aside from the problems discussed above, concerns the absence of a detailed understanding of the role of CRP on a biologic or mechanistic level, especially in the context of the action of CRP per se in the initiation and promotion of CVD and whether inflammation unrelated directly to atherosclerosis, as compared to actual arterial inflammation, is relevant.
PROS & CONS OF TESTING
Is there enough evidence to justify requesting or agreeing to a CRP test? There does not appear to be a simple answer. However, in view of the large number of consistent studies indicating that it is an independent risk marker for CVD, and in view of the significantly elevated risk associated with high CRP and LDL on the one hand or elevated CRP and a high TC/HDL ratio on the other, it is difficult to dismiss or ignore the potential importance of this test. Consider two extreme situations:
REDUCTION OF CRP
It is natural to conclude that since CRP appears to be a strong and significant risk factor for CVD, then reducing the blood level of this protein is obviously beneficial, but this is a fallacious line of reasoning. In fact there are no large intervention studies specifically linking CRP reduction with adverse cardiovascular events as endpoints, where there are no questions regarding other actions associated with the intervention that might reduce events independent of the reduction of CRP. For example, the statin class of cholesterol lowering drugs reduces CRP, but this does not prove that the reduction in observed adverse cardiovascular events is thus directly due even in part to this reduction in CRP. In fact, it is thought that the statin drugs have a multiplicity of effects, both good and bad (the various side effects, some of which can be serious or life-threatening). The actual direct benefit derived specifically from lowering the circulating level of CRP is presumably unknown, since the drug in question could produce the observed reduction in adverse events by a mechanism independent of CRP lowering. Nevertheless, it can be argued that there is nothing to loose and perhaps much to be gained by reducing elevated CRP levels, provided the actions have little or no risk or side effects.
There are a number of non-pharmaceutical interventions or actions with the potential for reducing CRP levels. Several studies indicate regular exercise can lower levels [41,42], and that in fact there is an inverse correlation between fitness, as measured with treadmill tests, and CRP levels in both men and women [43,44]. Since CRP levels correlate with body mass index  ( BMI--weight in kg divided by the square of height in meters), a program of weight reduction, especially for those with BMI over 26, would be indicated. Smoking is strongly correlated with CRP, which provides one of many reasons for smoking cessation.
Large daily doses of vitamin E (1200 IU) have been found to dramatically reduce CRP levels . In fact, the percentage decrease observed with vitamin E exceeds that obtained by most pharmaceutical interventions . Unfortunately, the study quoted did not investigate dose dependence, and some would view 1200 IU/day as high. The omega-3 essential fatty acids have been found to be both anti-inflammatory and to decrease serum CRP [47,48,49]. Dietary sources of omega-3 fats include fish, fish oil, canola oil, various nuts—especially walnuts, and ground flax seeds or flax seed oil.
Alcohol and possibly other constituents of alcoholic drinks have an anti-inflammatory effect when consumed in moderation. Both non-drinkers and heavy drinkers have been found to have higher CRP levels than moderate drinkers, who in fact derived cardiovascular protection from alcohol along with the reduction in CRP levels . Avoiding stress and taking action to resolve causes of depression are important, since there is a significant link with both [12,13,51].
The connection of CVD risk (and thus frequently elevated CRP) with infections is complex and the subject of much current research . It can be argued that it may be wise to deal with dental infections, although the link between CHD and periodontal disease is far from clear [53,54]. However, research just published  based on an analysis of data derived from the Health Professionals' Follow-Up Study, found an increased risk of ischemic stroke to be associated with periodontal disease and tooth loss. In general, suspected chronic infections should be brought to the attention of the individual's physician and treatment discussed, at least in part, in the context of the potential for an enhanced CVD risk, even though the strength of the link between infections and CVD risk is still very much the subject of debate and research.
Finally, in one large study aspirin was found effective in reducing both CRP and adverse cardiovascular events, especially for men with the highest CRP levels . This observation in fact suggests that aspirin may have benefits over and above its well known anti-platelet effect, which involves interference with thrombus formation by platelets, and is the rationale for its widespread promotion and use in preventive medicine, in spite of the small but significant risk of adverse gastrointestinal bleeding and hemorrhagic stroke. Unfortunately, the effect of aspirin on CRP levels is uncertain, since the data from available studies are in fact inconsistent [56,57]. In addition, the current U.S. consensus on the use of aspirin for primary prevention of cardiovascular events  concludes that the balance of benefit vs. harm is most favorable in patients at high risk of CHD (those with a 10-year risk equal to or greater than 6%, presumably calculated from the Framingham study data  ).
Dr. Stephen Sinatra, a practicing cardiologist, faculty member of the University of Connecticut Medical School and author of the recent book "Heart Sense for Women" , recommends a natural approach to reducing CRP which is described on his web site . Sinatra was one of the early advocates of measuring CRP as part of the blood work-up for both heart patients and individuals presumed disease free .
High CRP levels have been associated with an increased risk of future heart attacks, ischemic stroke, and peripheral arterial disease. However, there is no clear consensus as to whether a single screening test for CRP is useful as a diagnostic tool. The diagnostic value of CRP screening can be improved by performing two tests a month or so apart and by including both CRP and the cholesterol level ratio (TC/HDL) in the final risk assessment, perhaps in conjunction with a Framingham risk calculation.
The lifestyle modifications that favorably impact CRP levels will be recognized by many readers as those that have also been widely discussed and advocated as part of a general approach to good health  and to decreasing the risk of cardiovascular disease. Some as well are frequently recommended as part of a non-pharmaceutical approach to reducing the risk associated with an unfavorable lipid profile [59,61].
It is common knowledge that about half of all individuals experiencing heart attacks have normal blood lipid profiles. A good example was described above . It has been frequently suggested in the literature that one of the missing pieces in the puzzle is in fact CRP, independent of whether or not it acts as merely a marker, or is active per se, or both. Closely related to this hypothesis is the growing evidence that atherosclerosis is, at least in part, an infectious, inflammatory and autoimmune disease [1,62]. Research now in progress will almost certainly help clarify the picture.