Abbreviations used
 
 

APP, acute phase protein; APR, acute phase response; BMI, basal metabolic index; CRP, C-reactive protein; G-CSF, granulocyte colony-stimulating factor; IL-1, interleukin-1b ; IL-6, interleukin-6; a -MSH, a -melanocyte-stimulating hormone; NEFA, non-esterified fatty acids; PAI-1, M-CSF, macrophage colony-stimulating factor; plasminogen activator inhibitor-1; SAA, serum amyloid A; SAP, serum amyloid P; sPLA₂, secretory (nonpancreatic) phospholipase A₂; TNF, tumor necrosis factor-a ; tPA, tissue plasminogen activator.
 
 

Shortly after they discovered a protein in human serum that binds to the pneumococcal "C" polysaccharide, Avery and others found that the serum concentrations of this "C-reactive" protein (CRP) can rise dramatically during illness. Following an acute stimulus, the serum CRP concentration increases rapidly, peaks approximately 1 to 3 days later, and then returns to the individual's usual concentration range (1). Since the half-life of CRP in the circulation (~ 19 hrs) is not significantly influenced by illness or other variables, the major determinant of its serum concentration is the rate at which CRP is produced and released into the blood (1). The production rate is regulated by factors that, influenced by signals from circulating cytokines, control transcription of the CRP gene.

The dependency of the serum CRP concentration upon the nature and intensity of the activating stimulus has allowed CRP to compete with the erythrocyte sedimentation rate for purposes such as distinguishing "sick" from "not sick," discriminating bacterial from viral infections, and gauging response to therapy for rheumatoid arthritis or osteomyelitis. Recent improvements in assay technology have increased the sensitivity of CRP measurement by 100-fold or more; using these assays, investigators have found striking trends within the previously accepted "normal" range. Remarkably, many of these observations have pointed to a role for CRP in predicting, and perhaps contributing to, atherosclerotic cardiovascular disease.

This Internal Medicine Grand Rounds presentation begins with a general overview of the potential uses of CRP to predict atherothrombotic vascular disease. It then discusses the known activities of CRP in the context of the body's acute phase responses, noting that most of the other biochemical risk markers for atherothrombosis are also acute phase proteins. The possible etiologic links between common clinical phenomena (aging, smoking, obesity, metabolic syndrome X, chronic infection), low-level but long-term activation of the body’s responses to stress, and atherothrombosis are then considered.

CRP and the risk of atherosclerotic cardiovascular disease

Studies in patients with myocardial infarction (2-7)

Myocardial infarction, like other kinds of tissue necrosis, provokes systemic responses such as fever, leukocytosis, and acute phase protein synthesis. Serum CRP levels increase exponentially, with a doubling time of ~8.2 hours, then decay with a half-life of 19 hrs (1)(Figure 1, data from a single patient who sustained MI). Post-MI CRP levels can reach well over 100 mg/L. The time to peak CRP level is variable; intervals of 50 ? 100 hours after infarction are typical. Given the apparent relationship between serum CRP concentration and the extent of myocardial injury (as reflected in circulating concentrations of LDH (2), CK-MB (3,8), troponin I, and hydroxybutyrate dehydrogenase) (6)), it is not surprising that very high CRP levels can identify patients with lower ejection fraction (6,9) or post-MI complications (3). There is also interesting evidence that CRP may exacerbate myocardial damage by binding to infarcted myocardium and activating serum complement (10,11).

Studies in patients with angina(9,12-24)

The technological innovation that allowed these studies was the development of "high sensitivity" assays for CRP ? assays that can measure CRP levels well below the previous detection limit, 10 mg/L. The assays are highly reproducible and plasma CRP levels in most healthy people do not fluctuate markedly over time (25). Most laboratories use an international (WHO) CRP standard, so between-lab reproducibility is high.

It is also important to note that serum CRP levels in humans are not normally distributed. The distribution is skewed toward lower values. Using the Dade/Behring rate nepholometric method, the 50^th percentile values for adult men and women blood donors were 0.7 and 0.9 mg/L, respectively; the 75^th percentile values were 1.4 and 3.1 mg/L, respectively (26).

Several lines of evidence now suggest that inflammation within coronary arteries is an important component of unstable angina (27,28). When compared to patients with stable or atypical angina, those with unstable angina have increased expression of adhesion molecules on neutrophils and monocytes collected from coronary sinus blood (29) and greater tissue factor expression on circulating monocytes (30). Using a thermistor probe to measure the temperature of the vessel wall as close as possible to the offending intracoronary lesion, Stefanadis et al. found that the gradient [vessel wall temperature ? oral temperature] was elevated in patients with unstable angina or myocardial infarction, and that this gradient also correlated well with blood CRP levels (31).

In keeping with these findings, several groups have reported that admission CRP levels correlate positively with risk of subsequent cardiac events (MI, unrelenting angina requiring revascularization, death) in patients admitted to intensive care units with unstable angina (and, by definition, with no evidence for ongoing infarction), (4,9,15-17,21,32). Other investigators have not found such associations, however (19,20); the basis for this difference is not clear. Although patients with chronic stable angina may have serum CRP levels that are higher than those in healthy controls (33,34), these "elevated" levels are lower than those reported in patients with unstable angina (Figure 2), and there does not seem to be a strong association between CRP levels and the severity of angiographically-diagnosed CAD (33,34).

Liuzzo et al. found that myocardial ischemia does not elevate CRP levels, in keeping with the notion that in patients with unstable angina the inflammation is in the vessel wall, not in the myocardium (18).

Population-based studies in healthy adults (35-45).

I reviewed 10 published studies. Almost all were retrospective analyses of population data that had been collected for other reasons. In these studies, CRP was measured only in the baseline serum sample; the follow-up period lasted from 2 to 17 years. A meta-analysis of 5 prospective studies performed prior to 1998 concluded that "comparison of individuals with CRP values in the top third with those in the bottom third at baseline yielded a combined risk ratio for CHD of 1.7 (95% CI, 1.4 - 2.1)"(46) . The estimated mean CRP values in these two groups were 2.4 and 1.0 mg/L, respectively. Subsequent studies have reached similar conclusions. In most of the studies reviewed, CRP remained a strong predictor of subsequent cardiac events even when total cholesterol, HDL cholesterol and other common risk factors were taken into account. A recently published analysis of CRP and other markers in predicting CAD in postmenopausal women suggested that CRP measurement may provide prognostic information in addition to that obtained from total cholesterol levels (42). In particular, women with a high baseline CRP level had a high relative risk of CAD even when the LDL cholesterol was below 130 mg/dL. In women with the lowest LDL cholesterol levels (mean = 104 mg/dl), the adjusted relative risk of CAD events increased almost 40% with each increasing quartile of CRP.

Although high-normal CRP levels have also been predictive of stroke (13,37,39,47,48) they have not been consistently associated with high risk of peripheral vascular disease (13,49).

CRP and Cardiovascular Risk
 
 

The published studies suggest the following schema (Figure 2):

• Many years before cardiovascular disease becomes clinically apparent, inflammation may occur in coronary and peripheral vessels. The inciting stimulus is not known; many observations point to an important role for infectious agents such as C. pneumoniae and/or CMV. Inflammation or infection in non-vascular sites can also increase CRP and other acute phase reactants (see below). In either case, the CRP levels remain below 10 mg/L, the traditional threshold value for diagnosing ongoing inflammation (Figure 2).
• In some individuals, the intracoronary inflammatory process is particularly active. These individuals are more likely to develop unstable angina. In patients with unstable angina who are admitted to an intensive care unit, CRP levels correlate rather well with risk of subsequent myocardial infarction and death. There is good evidence that the inflammatory process in these patients is within the coronary vessels, not in the myocardium, since transient myocardial ischemia does not elevate CRP or other indicators of acute inflammation.
• Myocardial infarction (necrosis) does trigger inflammation within the damaged tissue. CRP levels rise rapidly and correlate quite well with the extent of myocardial damage. CRP can bind to cardiac muscle fibers and activate complement, possibly contributing to local leukocyte infiltration, further myocardial injury, and infarct extension (10,11).

Do drugs lower that prevent atherosclerosis also lower CRP?

Aspirin and NSAIDS. Aspirin has both anti-platelet and anti-inflammatory activities. In men participating in the Physicians' Health Study, taking aspirin (325 mg qod, Bufferin‰ ) was associated with a statistically significant reduction in the relative risk of myocardial infarction. The reduction was insignificant in those individuals in the lowest quartile of baseline CRP values, 33.4% in the second quartile, 46.3% in the third quartile, and 55.7% in the fourth quartile (37). Although this analysis has been criticized (46), it suggests that at least part of the beneficial effect of aspirin may be due to its anti-inflammatory potency. Does aspirin lower CRP levels? When it was given to healthy men for 7 days, 81 or 325 mg aspirin did not lower pre-exercise or post-exercise CRP levels (50). Although a 6-week course (300 mg/d) lowered CRP levels in patients with chronic stable angina (33), the net reduction was small. those individuals in

Statins. The CARE trial of secondary prevention randomized post-MI patients to receive 40 mg/day pravastatin or placebo. In a retrospective analysis of this trial, there was an association between recurrent MI and high baseline levels of CRP and serum amyloid A (SAA). The proportion of recurrent coronary events prevented by pravastatin was 54% in the subgroup with inflammation (those above the 90th percentile of both CRP and SAA) and 25% in the subgroup without inflammation (51). In another analysis of the CARE data, the same group found that median CRP levels tended to increase over 5 years in placebo recipients, whereas median CRP levels decreased by 17% in those allocated to take pravastatin. At the 5 year analysis, the difference in median CRP levels (21.6%) and the absolute mean change in CRP (0.137 mg/dl) between placebo and pravastatin groups were highly significant (52). The differences persisted after stratification according to baseline lipid levels and other standard risk factors, and there was no obvious relationship between the magnitude of change in CRP and the magnitude of change in serum lipids. A smaller study from Finland also described decreases in CRP associated with statin use (53).

HMG CoA-reductase inhibitors have numerous potentially anti-atherogenic effects in addition to their cholesterol-lowering action; these include decreasing fibrinogen (and blood/plasma viscosity), decreasing tissue factor expression on macrophages, decreasing platelet aggregation and deposition, and increasing fibrinolysis (by reducing PAI-1 levels)(54). There seem to be some differences among the various statins on the market, however (54); the clinical significance of these differences is uncertain.

Since both aspirin and statins have multiple actions that may help prevent atherosclerotic disease, it is obviously impossible to know what portion of their beneficial action is due to inhibiting inflammation.

CRP: The Eyes of the Hippopotamus

CRP is one of the most dynamic acute phase proteins; its serum concentration can increase 1000-fold or more in response to various stimuli. It seems likely that, in addition to whatever direct role CRP may play in the pathogenesis of atherosclerosis, it is also a very sensitive marker for the body's systemic responses to stress. In the African metaphor, it’s the eyes of the hippopotamus ? much more of the beast lies below the surface of the water.

The Acute Phase Response (Table 1)

The term "acute phase response" (APR) refers to the body's systemic adaptations to stress (55). Stimuli that can trigger the APR include bacterial infection, trauma, neoplasms, bone fractures, tissue infarction, immunological reactions and diseases, and childbirth. Broadly defined, the APR is regulated by the hypothalamic-pituitary-adrenal axis (HPA), the sympathetic and parasympathetic nervous systems, the thermoregulatory center and cytokine-induced changes in protein synthesis. The term "acute" is obviously inaccurate when the same systemic responses persist, usually at low levels, for many years.The APR has been conserved from higher invertebrates to humans. It existed before animals evolved immunoglobulins, lymphocytes, or complement. It encompasses not only the "fight or flight" responses to danger but also many highly conserved mechanisms for defeating invading microbes. The body uses many of the same molecules (ACTH, cortisol, catecholamines, IL-6, others) to regulate its systemic adaptations to infection, injury, exercise, and many other stresses.

Acute phase proteins (APPs). The term "acute phase response" can also refer to the changes that occur in the blood (or intracellular) concentrations of several proteins. These proteins are sometimes called "acute phase reactants" or "acute phase proteins" (APPs). The "positive" APPs are those that become more abundant in the blood during the APR. The concentrations of some of these proteins increase as much as 50%, others may increase from 2- to 5-fold, while others may increase 1000-fold or more (Table 2). The concentrations of the "negative" APPs decrease; most prominent among these is serum albumin, although decreases in transthyretin and apolipoprotein AI can also be physiologically important.

Regulation of acute phase responses

The regulation of acute phase responses is exceedingly complicated and only partially understood. In brief, signals from the CNS (catecholamines, ACTH, glucocorticoids) and peripheral tissues (cytokines, prostanoids, others) participate in multiple interacting pathways; precisely how the non-linear interactions contribute to maintaining physiological stability is uncertain.

The most prominent stimulus for acute phase protein production is a cytokine, interleukin-6 (IL-6). IL-6 is made by most cells, which release it when they are injured; b -adrenergic agonists also can enhance IL-6 release from stimulated cells (56). Since mice that lack IL-6 have exaggerated inflammatory responses to various stimuli and are unable to control inflammation at local sites of infection, this cytokine is now thought to have predominantly anti-inflammatory or immunomodulatory actions (57). It is also attractive to think of IL-6 as an "SOS" cytokine that mobilizes systemic responses to injury or infection: in addition to its role in triggering acute phase protein production, it is a potent activator of the hypothalamic-pituitary-adrenal axis. Other cytokines, particularly IL-1, also play roles in APP regulation; the importance of IL-6 comes from its production by most cells, its broad stimulatory activity toward APP genes, and its ability to bind in blood to the soluble IL-6 receptor (sIL-6R), which prolongs its half-life and enhances its potency.

The Acute Phase Response: Presumed Functions. The presumed functions of the acute phase response may be grouped for discussion into tentative categories. The actions of CRP illustrate four of these functions.

Anti-infective actions:

CRP is a planar molecule that has 5 identical subunits. It binds to phosphocholine moieties on the surfaces of microbes. Like some immunoglobulins, CRP can opsonize particles for phagocytosis; the major binding receptor for CRP on leukocytes is Fcg receptor IIA (58).) It can also activate complement via the classical pathway (C1q). CRP thus can facilitate microbial killing by at least two distinct mechanisms.

CRP protects mice from lethal pneumococcal infection (59,60). Serum bactericidal activity toward Haemophilus influenzae is also mediated by CRP, which binds to a phosphorylcholine moiety on the Haemophilus cell surface lipopolysaccharide (LPS) (61). Among Neisseria, the LPS of commensal strains is substituted with phosphocholine while that of Neisseria meningitidis is not (62). It seems likely that CRP may restrict tissue invasion by these human pathogens, all of which inhabit the oropharynx, to organisms that do not express LPS-linked phosphocholine. CRP can also bind a surface ligand on Leishmania (63).

CRP behaves very much like a broad-specificity antibody. It has been called an "ante" antibody, since it has been found in invertebrates that do not make either immunoglobulins or complement. It also is one of the key pattern recognition molecules that enable prompt host recognition of invading microbes (64): its "pattern" is determined by the configuration of phosphocholine moieties on membranes. (CRP is thought to bind to eukaryotic cell surfaces only after conversion of some of the membrane phospholipids to lysophospholipids by secretory phospholipase A₂ (sPLA₂).)

Other APP pattern-recognition molecules include mannose-binding protein (which can also activate complement) and LPS-binding protein.

Anti-inflammatory actions (57):

Several acute phase responses prevent systemic inflammation (i.e., neutrophil recruitment to, and activation within, tissues distant from a site of injury or infection). CRP increases L-selectin shedding from neutrophils and prevents neutrophil-endothelial cell adhesion (65); these actions, like those of epinephrine and cortisol, should favor neutrophil release from the marginated pool while preventing leukocyte adhesion to vessels in non-inflamed tissues. CRP also stimulates monocytes to release an important anti-inflammatory molecule, interleukin-1 receptor antagonist (IL-1Ra). In contrast, CRP can stimulate pro-inflammatory mediator production when it engages Fc receptors (66). In general, CRP seems to inhibit neutrophil activation whereas it activates monocyte-macrophages (67). Transgenic mice that overexpress rabbit CRP are protected from lethal reactions to endotoxin, IL-1b + TNF-a , and platelet activating factor (PAF) (68).

Other anti-inflammatory APPs include molecules that modulate the action of pro-inflammatory cytokines (IL-1Ra and soluble TNF receptors) as well as the numerous antiprotease and antioxidants that help confine the impact of potentially toxic proteases and oxidants to local sites of inflammation.

Procoagulant actions

Recent findings point to IL-6 as the major procoagulant cytokine in humans (69). It is thus not surprising that CRP, which is produced in response to IL-6, is able to increase the expression of tissue factor on monocytes (70). Other procoagulant APPs include fibrinogen and plasminogen activator inhibitor-1. Although clotting plays an important role in walling off infection at local sites (anticoagulation prevents animals from developing abscesses), a protective function for these systemic procoagulant activities is not apparent.

Scavenging actions

Although CRP does not bind to normal cell membranes, it can bind avidly to cells that are undergoing apoptosis or necrosis, possibly because it can recognize the lysophosphatidylcholine (lysoPC) that appears on the surfaces of dying cells. It then binds and activates complement (C1q), initiating a local inflammatory reaction that attracts neutrophils and monocytes to the scene. These activities may explain its ability to enhance myocardial injury in experimental animals (11,71). CRP also binds nuclear ribonucleoproteins and histones (thus, chromatin) under physiological extracellular ionic conditions, suggesting that it may be involved in scavenging nuclear antigens (72) and enhancing chromatin clearance from the blood into liver and spleen (73).

Summary: CRP is a dynamic component of the systemic response to infection, injury, and other stresses. Like many other APPs, it seems to have anti-infective, procoagulant and anti-inflammatory actions that promote host defense and tissue repair. Its ability to stimulate monocyte tissue factor expression could possibly promote atherosclerosis (thrombosis), while its role in clearing dead cells might account for its reported ability to exacerbate myocardial necrosis in experimental animals (71).

Other acute phase responses that may contribute to atherothrombosis:

1. IL-6. Elevated IL-6 levels occur following myocardial infarction (74) and during unstable angina (23). A decrease in plasma IL-6 during the first 48 hours of hospital admission for unstable angina was associated with an uneventful hospital course, whereas an increase occurred in most patients who had cardiac events (23). In apparently healthy men, baseline IL-6 levels predicted risk of subsequent MI; although there was a strong correlation between IL-6 and CRP levels, the relationship of IL-6 with subsequent risk remained after controlling for CRP (40). In other studies, CRP has been superior to IL-6 for estimating cardiac risk (42).

2. Procoagulants. Fibrinogen is the major coagulation protein in the blood; it also strongly affects blood viscosity and the erythrocyte sedimentation rate (ESR). In their meta-analysis of 12 population-based studies of CAD risk, Danesh et al. found a summary odds ratio of 1.8 (95% CI, 1.6-2.0) for individuals in the upper third of the fibrinogen concentration distribution compared to those in the lowest third (46) (in the same meta-analysis, the odds ratio for CRP was 1.7). Fibrinogen and CRP levels usually correlate strongly with each other (12,14,75). Thompson et al. (12) found that fibrinogen and CRP were good predictors of coronary events in adults with CAD, even in individuals with low total cholesterol levels. Even more striking was the ability of a low CRP level, when combined with a low fibrinogen level, to predict event-free survival in individuals with high total cholesterol (12).

Another important APP is plasminogen activator inhibitor-1 (PAI-1). It is perhaps the most important anti-fibrinolytic protein in plasma. Importantly, tissue plasminogen activator levels also increase during the APR; perhaps the increase in tPA is a compensation for higher PAI-1 levels. In any case, in studies in healthy adults, levels of tPA have also correlated with both CRP levels and CAD risk (76).

Blood levels of von Willebrand factor (vWf), another APP, also predict CAD risk; in some studies this risk prediction has been independent of, and stronger than, that associated with CRP (77). It is uncertain whether high vWf concentrations reflect the APR or diffuse endothelial activation, however; Jager et al. favor the latter interpretation (77).

2. sPLA₂. Secretory nonpancreatic type II phospholipase A₂ (sPLA₂) is not only expressed in atherosclerotic arterial walls, but it is also an APP; circulating levels can increase >100-fold during the APR. In addition to having antibacterial actions (it is the most potent anti-staphylococcal and anti-streptococcal molecule in acute phase plasma (78)), sPLA₂ is also involved in remodeling HDL and LDL and damaged membranes (see (71) and discussion of dyslipidemia below). A recent study in Japan found that sPLA₂ was superior to CRP for predicting subsequent cardiac events in a population of patients who underwent cardiac catheterization for chest pain or ischemic EKG changes (79).

3. Nutrient- and energy-redistributing changes. During the APR, the body adapts to use endogenous nutrients. As stated by Fernandez-Real and Ricart, "our genome is designed to fight against infection with minimal food intakes" (80). The known adaptations include an abrupt drop in albumin synthesis, which may conserve amino acids for the production of stress proteins. Glycolysis and glycogenolysis increase while glucose entry into skeletal muscle decreases (insulin resistance); these changes are thought to improve glucose availability for the brain and immune cells. Decreases in circulating triiodothyronine and transthyretin reduce tissue catabolism while increases in circulating leptin may decrease appetite during stress reactions. Although their adaptive functions are argued, several of these metabolic changes are risk markers for CAD. For example, the serum albumin concentration has a statistically significant inverse association with CAD (46). Although it is not easy to imagine how hypoalbuminemia might contribute directly to atherothrombosis, the changes that occur in circulating lipids seem clearly linked to risk. These will be discussed next.

4. Dyslipidemia and Serum Amyloid A (SAA)

Hypertriglyceridemia. Blood levels of triglyceride and VLDL increase during the APR; hypertriglyceridemia has been associated with increased risk of developing CAD (81). Hypertriglyceridemia is often attributed to decreases in lipoprotein lipase (LPL) and hepatic lipase. Free fatty acid concentrations also rise, due to increased lipolysis in adipose tissue and muscle and increased hepatic fatty acid synthesis. See (82) for a more complete review of these changes.

Changes in HDL concentration and composition. Two related phenomena seem important. First, circulating HDL-cholesterol levels decrease during the APR, largely due to increased HDL clearance from the plasma (83). In some studies of healthy individuals, CRP has had significant negative correlations with apoA-I and apo-AII (38). Other studies have not found such correlations (44,84), but numerous analyses have documented an inverse relationship between CRP and HDL-cholesterol (42,84,85). The second phenomenon is that HDL particles undergo major remodeling.Early in the APR, HDL particles acquire triglyceride; this may partly reflect remodeling caused by secretory phospholipase A₂. Later they undergo even further change as they acquire serum amyloid A (SAA). During the APR, serum levels of SAA may increase 500- to 1000-fold. SAA is carried almost entirely on HDL₃; when it binds to HDL₃, it displaces apoA-I. When it is compared with normal HDL in vitro, acute phase (SAA-)HDL has a higher affinity for binding macrophages; it also surrenders cholesterol esters more rapidly to macrophages and, at least in some studies, is significantly less able to enable cholesterol efflux from these cells (86). These changes may contribute to the pro-atherogenic nature of SAA-HDL.

Other acute phase changes in HDL composition include decreases in cholesterol (87), platelet activating factor acetylhydrolase (PAF-AH)(88), and paraoxonase (which prevents LDL oxidation). The anti-oxidant and anti-inflammatory character of the particles is presumably diminished as they lose PAF-AH and paraoxonase. In addition, HDL gains ceruloplasmin, another acute phase protein, and is thought to become more pro-oxidant (88). See the review by Khovidhunkit et al. (82) for details.

5. Small, dense LDL. Of the subclasses of LDL, small dense LDL is thought to be more proatherogenic because it is more susceptible to oxidation and can penetrate the endothelium more effectively than the larger LDL particles. Although the appearance of small dense LDL as an acute phase reaction has received little study, small dense LDL has been associated with hypertriglyceridemia in patients with the acquired immunodeficiency syndrome (82). In healthy men, Lamarche et al. found an association between the presence of small LDL particles in baseline plasma and subsequent risk of ischemic heart disease (89). The small, dense LDL phenotype also seems to cluster with other features of metabolic syndrome X (see below)(90).

6. Leukocytosis. Although leukocytosis is not usually included in discussions of the APR, it may result indirectly from the actions of cortisol (which promotes demargination of leukocytes from postcapillary venules (91)) and CRP (which induces neutrophils to shed L-selectin, thus reducing neutrophil adhesion to vascular endothelium), as well as from marrow stimulation by G-CSF. In their 1998 meta-analysis, Danesh et al. found 19 prospective studies of leukocyte count and CAD. Individuals in the top third (mean = 8.4 x 10⁹/L) had a risk ratio of 1.4 (95% CI, 1.3-1.5) relative to those in the bottom third (mean = 5.6 x 10⁹/L). The white cell types were not specified.

Summary: almost all of the known laboratory-based risk markers for CAD are acute phase reactants (Figure 3). The list includes, in addition to those discussed above, heat shock protein 60 (92), intercellular adhesion molecule 1 (93), and macrophage colony-stimulating factor (33,94). Prominent exceptions include the plasma levels of homocysteine and lipoprotein (a). Low-level activation of the APR thus may play an important role in determining CAD risk. So what triggers the APR? Is it activated by ongoing inflammation within atherosclerotic vessels, is it activated by extravascular stimuli, or both? Similar questions may be asked regarding the role of infectious agents in the pathogenesis of atherothrombosis.

Associations between serum CRP levels and infection with specific microbial agents (42,84,92,95-100). Both intravascular and extravascular infections can provoke acute phase responses. Several groups of investigators have looked for associations between positive serologies (for C. pneumoniae, cytomegalovirus [CMV], hepatitis A, Herpes simplex, Helicobacter pylori, and E. coli lipopolysaccharide) and serum CRP. Some of these studies found statistically significant correlations between serological titers and serum CRP levels yet others did not. Given that so many variables influence the occurrence of infection with each microbial agent studied, it is not surprising that there is lack of consensus on this issue. Indeed, the results of two recent studies suggest that the "total burden" of infectious exposures may correlate best with CRP levels (95,97), and another investigative group found a significant association between "chronic infections," defined using clinical definitions, and risk of carotid atherosclerosis (92). Perhaps the relevant variable is whether or not an individual has one or more chronic infections that provide a long-term stimulus to the APR. In addition to C. pneumoniae and CMV, which can be found in atheromatous lesions, common extravascular infections such as bronchitis, periodontal disease, urinary tract infections, and gastritis also deserve further study.

Is activation of the APR a mechanism by which risk factors such as age, smoking, and obesity accelerate atherothrombosis? We shall now consider the possibility that some of the major clinical risk factors for CAD accelerate atherothrombosis by activating the APR.

Age. In apparently healthy men, CRP levels increase with age (38). Similar results were found for both genders in another study (101). The mean levels were 1.90 mg/L in the 25-34 year age group vs. 3.03 in the 75-84 year group (101).

Smoking. Smokers have significantly elevated CRP levels (35,38,44). In some studies, smoking had such a striking relationship to CRP that smokers were considered as a separate group: even when this was done, CRP remained an independent risk factor for CAD. Smoking might trigger APR by inducing low-grade airway inflammation, predisposing to chronic bronchitis, or by other mechanisms; alveolar macrophages from smokers tend to make less IL-6 than do those from normals , however, raising the interesting possibility that the smoking-related stimulus to the APR is extrapulmonary (102).

Obesity. There is a strong correlation between BMI and CRP levels, particularly in women (103) but also in men (38,44). In a recent study by Visser et al., the prevalence of "clinically elevated" CRP (> 10 mg/L) was 4.0% in normal-weight women, 7.7% in overweight women (BMI 25 - 29 kg/m²), and 20.2% in obese women (BMI > 30 kg/m²)(103). Older women were less likely to have elevated or clinically elevated CRP than younger women. CRP levels, like CAD, also correlate directly with the waist/hip ratio. BMI is also strongly associated with plasma levels of leptin, an adipose tissue APP (104,105).

Metabolic syndrome X. Reaven (106) suggested that the increased risk of CAD in hypertensive patients, and the fact that this risk is not reduced with antihypertensive treatment, is due to the clustering of hypertension, hyperinsulinemia (insulin resistance), hypertriglyceridemia, and low HDL-cholesterol. More recently, studies have suggested an association between insulin resistance and CRP levels in healthy individuals (107). Pickup et al. (108,109) and Fernandez-Real and Ricart (80) have suggested that the APR makes an important contribution to non-insulin-dependent diabetes mellitus (NIDDM) and, in addition, to metabolic syndrome X (obesity, NIDDM, hypertension, atherosclerosis). Acute phase responses noted in patients with metabolic syndrome X include hyperinsulinemia, hypertriglyceridemia, high leptin, low HDL, high CRP, high complement and SAA, high fibrinogen, high PAI-1 and von Willebrand factor, high cortisol, low testosterone and zinc (108). Elevated levels of ferritin, another APP, may also be part of the "insulin resistance syndrome" (110), as may the production of small dense LDL (90). In obese individuals, hyperleptinemia is associated with high IL-6 and cortisol levels in addition to its strong correlation with BMI (111).

Summary: there is intriguing but inconclusive evidence that low-grade, chronic activation of the APR plays a contributory role in the atherogenesis associated with obesity, smoking, age, and metabolic syndrome X. What activates the APR in these diverse conditions? Yudkin (112) has proposed that IL-6 is the mediator that links the APR to metabolic syndrome X. A similar hypothesis has been suggested by McCarty (113). A slightly modified version of the hypothesis is diagrammed in Figure 4.

Is IL-6 the link that connects obesity, metabolic syndrome X and the APR? A summary of the evidence follows:

• IL-6 is a major trigger for the APR. This conclusion is supported by the observed consequences of gene disruption experiments in mice: IL-6 -/- mice do not mount an acute phase response (114), whereas TNF-a -/- mice are essentially normal in this regard. (IL-1 -/- mice also do not have a normal APR, pointing to important interactions between IL-1 and IL-6 in APR activation.)
• Age, smoking, obesity, and metabolic syndrome X are all associated with elevated blood IL-6 levels.
• Why IL-6 levels increase with age and smoking is uncertain. In contrast, the association of obesity with elevated blood IL-6 levels may be explained by the striking observation that approximately 10% to 35% of the body's IL-6 is produced in adipose tissue (115,116). Omental adipose tissue evidently produces significantly more IL-6 than subcutaneous adipose tissue (116), in keeping with the observation that waist-to-hip ratio is a risk factor, independent of BMI, for an elevated blood CRP level (103) and the "metabolic syndrome." (Adipose tissue also produces and releases TNF-a (117); blood TNF-a levels may correlate better with the waist-to-hip ratio than with BMI (117). Since TNF-a can trigger IL-6 release, one theory holds that it is adipose tissue TNF-a that actually is the "driver" behind the visceral obesity-APR-CAD connection (113). Evidence against this hypothesis is the finding that subcutaneous adipose tissue evidently releases IL-6, but not TNF-a , in vivo (115). On the other hand, blood levels of soluble TNF-a receptor (sTNF-R55, an indicator of TNF-a activation), correlate with blood leptin levels in both controls and diabetic subjects (104).)
• An allelic polymorphism in the IL-6 promoter has been associated with LPS- or TNF-induced transcription of the IL-6 gene in vitro and with blood IL-6 levels in vivo (118). Fernandez-Real et al. found that subjects who were homozygous for the C allele at position ?174 of the IL-6 gene had significantly lower fasting insulin levels and better performance on an oral GTT than did individuals who were carriers of the G allele (119). The same group reported that G carriers had significantly higher VLDL, higher free fatty acid levels, and lower HDL-cholesterol than did C homozygotes; a tendency toward higher IL-6 levels in the G carriers was also found (120) (see also (118)).
• Yudkin's analysis of 106 non-diabetic adults found a strong correlation (r = 0.59, p < 0.00005) between acute phase response molecules and insulin resistance syndrome variables (107).
• In atherosclerosis-prone mice, administering IL-6 can accelerate the rate at which atherosclerotic changes develop (121). In normal mice, IL-6 administration had no effect, in keeping with the conclusion that the APR can contribute to, but does not initiate, atherosclerosis (122).

Also note that the available data do not exclude the possibility that smoking, age and obesity directly or indirectly enhance vascular inflammation and that it is the inflammation in the vessel wall, not elsewhere, that triggers the APR. Only for obesity, metabolic syndrome X, and non-vascular infections is there strong evidence for activation of the APR from an extravascular site.

Concluding comments

CRP is an important element of the acute phase response. Although its direct contribution to atherothrombosis is uncertain, in some epidemiological studies it remained an independent predictor of CAD risk even when numerous other variables were included in multivariate analysis. As one of the most dynamic and easily measured APPs, CRP is currently the best laboratory marker for the APR, which produces most of the known biochemical risk factors for atherothrombosis.

It is also important to note that the various components of the APR are not so highly orchestrated that they play the same tune in response to every kind of stimulus or in all individuals. There are both inherited and stimulus-specific differences. In some circumstances, CRP is not a useful way to monitor the APR ? a prominent example is systemic lupus erythematosus, in which the ESR is much more responsive to disease activity than is CRP.

Finally, Yudkin et al. found that, among the residents of a single area in India, those living in rural communities had significantly lower blood levels of several APPs and cytokines than did those living in the city (123). If chronic, low-level activation of the APR is a price that humans have paid for urbanization and the evolution of modern lifestyles, accelerated atherosclerosis may be the bill. Learning how to prevent chronic activation of the APR without interfering with its beneficial functions could reap great rewards for science and medicine.

Using CRP blood levels in clinical practice

Among patients with unstable angina, those who have both a low CRP and a low troponin I level have a very low risk of progression to MI, need for revascularization, or death (4,17). At the other extreme, a high CRP level in a patient with unstable angina or myocardial infarction should heighten concern for subsequent adverse events. Clinical trials are needed to find out if the CRP level can be combined with other clinical information to guide the length of hospitalization, the need for other diagnostic tests, and the use of preemptive interventions to prevent adverse outcomes.

Although some studies suggest that a high CRP (or SAA, or IL-6) level can identify patients at high risk of CAD despite normal (or even low) total cholesterol levels, randomized prospective clinical trials are also required to determine whether a drug intervention (e.g., with ASA or statin) can reduce both CRP and CAD in such individuals.

Some of the other unanswered questions:

• How often (or how many times) should CRP be measured in a healthy adult in order to establish his or her risk category?
• What is the best "cut point" for defining increased CAD risk in different patient populations ? for each age group, gender, ethnic group, clinical condition?
• Does measuring other risk factors (e.g., fibrinogen or von Willebrand factor) increase the ability to predict CAD risk sufficiently to warrant measuring these blood levels in addition to CRP? Is CRP superior to the risk stratification provided by newer imaging techniques?
• Is a high CRP an indication for aspirin or statin treatment in someone whose total or LDL cholesterol level is low? (And what is "high" and "low" in this context?) What if the patient's total cholesterol/HDL ratio is less than 5 (42)?
• Should anti-inflammatory therapy be part of the routine management of MI with elevated CRP? If so, with what?
• What is the value of measuring CRP in patients with known CAD who are already taking a statin and/or aspirin?

Acknowledgments. I am grateful for support from the Jan and Henri Bromberg Chair in Internal Medicine and from grants AI18188 and AI38596 from the National Institute of Allergy and Infectious Diseases.