ROLE OF C-REACTIVE PROTEIN IN CORONARY RISK REDUCTION: FOCUS ON PRIMARY PREVENTION
Gotto AM
The American Journal of Cardiology 2007; 99:718-725


Ricerche sulla natura infiammatoria dell'aterosclerosi suggeriscono che alcuni marker dell'infiammazione possono essere utilizzati come predittori di eventi clinici; tra questi, la proteina c-reattiva è stata associata al rischio coronarico, ma non ci sono ancora evidenze sufficienti che supportano il suo utilizzo nella valutazione del rischio o come target nella terapia.

Studi su ampie coorti e trials clinici hanno trovato che almeno uno dei principali fattori di rischio coronarico (iperlipidemia, ipertensione, abitudine al fumo, diabete) è presente nell'80-90% dei pazienti che sperimentano un evento CHD. Tuttavia, anche il 70% dei soggetti che non sviluppano eventi CHD hanno almeno uno di questi fattori di rischio. Ciò indica che i maggiori predittori, sebbene siano sensibili fino al 90%, possono avere una specificità solo del 30%.
Di conseguenza, si evidenzia la necessità di individuare biomarker plasmatici addizionali che possano affinare le strategie correnti di valutazione del rischio ed essere valutati facilmente e senza costi elevati.
Un biomarker è una caratteristica, un processo o una sostanza biologica che è:
" osservabile o misurabile;
" associata alla malattia;
" utile nella predizione/stratificazione del rischio (in pazienti asintomatici e/o sintomatici), nella diagnosi, nel monitoraggio della progressione della malattia, nel monitoraggio della risposta ad un intervento e/o nella predizione dell'efficacia e della sicurezza terapeutica.

Oggi l'aterosclerosi è considerata una patologia infiammatoria in cui i meccanismi immunitari, stimolati dal danno endoteliale, interagiscono con fattori di rischio metabolici, portando ad avvio, propagazione e attivazione di lesioni nei vasi arteriosi. È quindi biologicamente plausibile considerare le proteine della risposta infiammatoria (molecole di adesione, citochine, proteine della fase acuta) come potenziali marker della progressione della malattia aterosclerotica o come predittori di eventi cardiovascolari.

La proteina C-reattiva è definita come un marker sensibile, ma non specifico, dell'infiammazione (proteina di fase acuta). Durante la risposta di fase acuta, i livelli di PCR aumentano di oltre 10 mg/L e possono raggiungere i 100 o 200 mg/L.
Nelle linee-guida stabilite congiuntamente dal Centers for Disease Control and Prevention e dall'American Heart Association, è raccomandata la valutazione opzionale dei livelli di PCR in aggiunta ai tradizionali fattori di rischio nei pazienti a rischio intermedio. Questa decisione è basata su evidenze epidemiologiche di un'associazione tra lievi innalzamenti della PCR e un incremento degli eventi coronarici. Queste evidenze indicano che aumenti dei livelli di PCR rispetto al basale predicono in modo abbastanza consistente l'incidenza di eventi CHD o CV in uomini e donne di mezza età senza precedenti patologie cardiovascolari. Spesso, l'associazione è lineare dopo correzione per diversi fattori di rischio tradizionali.
In una meta-analisi di 22 studi su base di popolazione svolti in soggetti con o senza storia di patologie vascolari, gli odds ratio di un evento coronarico incidente era 1,58 (IC al 95% 1,48-1,68) per coloro che avevano livelli di PCR nel terzile più alto della distribuzione vs il terzile più basso. Tutti gli studi inclusi nella meta-analisi, tranne 2, riportavano odds ratios corretti per i comuni fattori di rischio CHD. A causa di una significativa eterogeneità (chi-quadro 46, gradi di libertà 21, p=0,001) tra gli studi, i risultati di questa analisi devono essere interpretati con cautela.
Secondo dati recenti dal Dallas Heart Study (DHS), non c'è un'apparente correlazione tra PCR e gravità del quadro aterosclerotico subclinico (ad esempio, calcificazione o meno della placca) dopo correzione per i tradizionali fattori di rischio e per l'uso di estrogeni o statine. Una possibile spiegazione per questi risultati è che l'associazione tra aumenti di bassa entità dei livelli di CPR e incidenza di CHD possa essere l'espressione della composizione, della morfologia e della stabilità della placca.

In alcuni studi, l'intervallo di valori dei livelli di PCR può estendersi ben oltre i piccoli innalzamenti considerati utili per predire gli eventi CHD, e l'associazione con l'aumento del rischio può assumere significato solo a livelli di PCR >=10 mg/L. I pazienti con livelli di PCR >9,99 mg/L dovrebbero sottostare a controlli per valutare le condizioni infiammatorie, infettive o neoplastiche, alcune delle quali (ad esempio, infezioni da Chlamydia pneumoniae o artrite reumatoide) possono essere fattori confondenti nell'associazione tra PCR e CVD.

Dati epidemiologici: distribuzione della PCR nei singoli studi

CARE = Cholesterol and Recurrent Events;
CHS = Cardiovascular Health Study;
FHS = Framingham Heart Study;
HHP = Honolulu Heart Program;
MONICA = Monitoring Trends and Determinants in Cardiovascular Disease;
KIHD = Kuopio Ischaemic Heart Disease Risk Factor Study;
MRFIT = Multiple Risk Factor Intervention Trial;
PHS = Physicians' Health Study;
WHI = Women's Health Initiative;
WHS = Women's Health Study.

* studio caso-controllo innestato
† eventi CHD
‡ eventi cardiovascolari o morte cardiovascolare
§ prevenzione secondaria

Un'associazione tra un marker e una malattia non implica necessariamente una capacità predittiva nei singoli pazienti. Sulla base di misure statistiche sulla performance del test (sensibilità, specificità, calibrazione), la PCR è stata esclusa da molti modelli predittivi, sebbene possa trovare utilità in specifici sottogruppi della popolazione.
Infine, anche se la PCR potrebbe non modificare la tradizionale valutazione del rischio, può essere patologicamente e fisiologicamente correlata alle patologie coronariche e può servire da target della terapia. Alcuni trial sulle statine hanno correlato i livelli di PCR al rischio coronarico, portando all'ipotesi che la decisione di iniziare il trattamento, presa sulla base dell'innalzamento della PCR, possa migliorare gli esiti nei pazienti. Sebbene le statine siano state progettate per ridurre il colesterolo inibendo l'HMG-CoA reduttasi, l'enzima coinvolto nella sintesi de novo del colesterolo, il loro beneficio può essere mediato da diversi meccanismi, inclusa la riduzione della componente infiammatoria delle CHD misurata dalla PCR e da altri marker.

LAVORO ORIGINALE

Given the limitations of current risk assessment strategies, adjunctive markers are needed to improve the prediction of a first coronary event. Research into the inflammatory nature of atherosclerosis suggests that inflammatory-response proteins may serve as potential predictors of clinical events. One in particular, C-reactive protein, has been the focus of much attention. Epidemiologic studies have shown a fairly consistent independent association between high-sensitivity C-reactive protein (hs-CRP) elevations and coronary risk, although a causal relation has not yet been established. Given this association, current guidelines recommend the optional use of hs-CRP to predict enhanced absolute risk in selected patients. The use of a marker in general clinical practice should be based on statistical measures that show incremental benefit over established risk factors and on randomized clinical trials in which therapy initiated as a result of marker screening improves patient outcomes. Thus far, statistical evidence concerning the incremental benefit of hs-CRP is not conclusive. Justification for the Use of Statins in Primary Prevention: An Intervention Trial Evaluating Rosuvastatin (JUPITER) is now being conducted to compare the efficacy of statin therapy versus placebo in subjects considered to be at increased risk on the basis of hs-CRP elevations, despite low to normal levels of low-density lipoprotein cholesterol. In conclusion, although epidemiologic studies suggest that low-grade C-reactive protein elevations are independently associated with coronary risk, more complete evidence is needed to validate the use of hs-CRP as a risk assessment tool in general practice and as a target for therapy in individual patients.

Cardiovascular disease (CVD) is a global health problem, accounting for 1/3 of all deaths worldwide. More than 7 million of the almost 17 million cardiovascular deaths each year are a result of coronary heart disease (CHD). According to a report issued by the World Health Organization, CHD is the leading cause of death in men and women aged ?60 years; it is surpassed only by human immunodeficiency virus/acquired immune deficiency syndrome in persons aged 15 to 59 years. Large cohort studies and clinical trials have found that ?1 major CHD risk factors (hyperlipidemia, hypertension, current smoking, diabetes) are present in 80% to 90% of patients who experience CHD events. However, approximately 70% of subjects who do not develop CHD also have ?1 major risk factors. This indicates that major predictors of risk, although up to 90% sensitive, may be just 30% specific. Consequently, researchers are engaged in efforts to identify additional blood-based biomarkers (Table 1) that can refine current risk assessment strategies and are easily and inexpensively evaluated in a primary care setting.

Table 1: Definition of a biomarker *

* Evidence to support the use of a biomarker in clinical practice must come from rigorously conducted research.
† "Association" denotes a statistical relation between 2 variables (i.e., a change in 1 appears to be related to a change in the other). Association is necessary for a causal relation to exist, but association alone does not indicate causation.

Markers of Inflammation
Originally considered a disorder of lipid metabolism, atherosclerosis is now regarded as an inflammatory disease in which immune mechanisms, triggered by endothelial injury, interact with metabolic risk factors, resulting in the initiation, propagation, and activation of lesions in the arterial tree. It is therefore biologically plausible to consider inflammatory-response proteins (e.g., adhesion molecules, cytokines, acute-phase reactants) as potential markers of atherosclerotic disease progression or predictors of cardiovascular events.

C-reactive protein (CRP)
First identified in 1930, CRP is defined as a sensitive, but nonspecific, marker of inflammation (an acute-phase reactant). An acute-phase reactant is a protein whose plasma concentration increases (or decreases) by ?25% in response to inflammation induced by trauma or metabolic, immunologic, infective, or other processes anywhere in the body. During the acute-phase response, CRP levels increase to >10 mg/L and may be as high as 100 or 200 mg/L.
Historically, CRP was thought to originate exclusively in the liver, where its production is stimulated by inflammatory cytokines (e.g., interleukin-6). More recently, however, investigators have found that other tissues, including human atherosclerotic lesions, coronary artery smooth muscle cells, aortic endothelial cells, and adipocytes, may also express CRP. This suggests that marked increases in CRP levels after a myocardial infarction may result from secretion by injured tissue. Moreover, persistent local production of CRP by atheromatous tissue or coronary artery smooth muscle cells may lead to chronic CRP elevations of 1.0 to 3.0 mg/L, which are detectable by the high-sensitivity CRP (hs-CRP) assay and may be potentially useful in predicting coronary risk. Although it is unclear whether CRP plays a causal role in atherosclerosis, a number of scenarios have been proposed. For example, monomeric CRP, which results from the conformational rearrangement of native CRP by means of an unknown mechanism, appears to induce adhesion molecule expression in human coronary artery endothelial cells. In 1 study, a significant increase in the attachment of neutrophils to monomeric CRP-activated endothelial cells was reported, suggesting that monomeric CRP may participate in the development of vascular inflammation. Although researchers know little about the monomeric CRP receptor(s) that participate in this process, they speculate that the Fc?RIII (CD16) receptor, which binds monomeric CRP on human neutrophils, may also be present on human coronary artery endothelial cells. Because it is involved in complement activation, CRP may also play a role in exacerbating tissue injury after a myocardial infarction. In vitro studies have found that CRP and complement are colocalized in infarcted myocardium and that CRP may contribute to further tissue damage through a complement-dependent mechanism. In 2003, the Centers for Disease Control and Prevention and the American Heart Association issued a statement identifying CRP as the inflammatory marker best suited for use in current clinical practice. 11 At the discretion of the physician, hs-CRP can be used to help gauge increased absolute risk in primary prevention patients with a 10-year CHD risk of 10% to 20%. Although a CRP level >3 mg/L (determined by averaging the results of 2 assays, 2 weeks apart, in metabolically stable patients) may guide decisions about the need for therapy or its intensity, the benefits of treatment based on this strategy are uncertain.
Recent guidelines issued by the Screening for Heart Attack Prevention and Education (SHAPE) Task Force recommend noninvasive imaging to detect subclinical atherosclerosis in asymptomatic men (aged 45 to 75 years) and women (aged 55 to 75 years). For patients whose imaging results place them in a moderately high risk category, CRP testing is an option. Those with CRP levels >4 mg/L may be reclassified as "high risk," a shift that would require more aggressive intervention.
These recommendations are based largely on epidemiologic evidence showing an association between low-grade CRP elevations and increased cardiovascular risk. The limitations of such evidence as a basis for clinical practice recommendations are discussed subsequently.

C-Reactive Protein: Epidemiology
Epidemiologic evidence indicates that baseline CRP elevations are a fairly consistent predictor of incident coronary or cardiovascular events in middle-aged men and women without previous CVD. Compared with the bottom segment of the distribution, higher CRP levels are associated with relative risk increases of up to threefold. Often, the association is linear after adjustment for a variety of traditional risk factors. In a meta-analysis of 22 population-based studies involving subjects with and without a history of vascular disease, the odds ratio of an incident coronary event was 1.58 (95% confidence interval 1.48 to 1.68) for those with CRP levels in the top third versus the bottom third of the distribution. All but 2 of the studies included in the meta-analysis reported odds ratios that were adjusted for established CHD risk factors. Because of significant heterogeneity (chi-square 46, degrees of freedom 21, p = 0.001) among the studies, however, the results of this analysis should be interpreted with caution.
The Framingham risk score, which comprises 5 major risk factors, is a widely accepted means of estimating a subject's 10-year likelihood of having a CHD event. 38 Within each category of estimated risk, higher CRP concentrations are generally associated with an increase in the incidence of CHD or CVD. Moreover, despite minimal correlations between CRP and most individual components of the Framingham risk score, there appears to be an association between increasing CRP concentrations and progressively higher risk scores. This implies that CRP may reflect an aspect of risk not captured by the variables on which the Framingham risk score is based.
According to recent data from the Dallas Heart Study (DHS), there is no apparent relation between CRP and subclinical atherosclerotic burden (i.e., calcified or noncalcified plaque) after adjustment for traditional risk factors and the use of estrogen and statins. A possible explanation for this finding is that the association between low-grade CRP elevations and incident CHD may reflect plaque composition, morphology, and stability.

Limitations of the Evidence
Association versus causation
An independent association between CRP and CHD, as represented by the odds ratio, is relevant only for the variables included in a given model. Because no prediction model can incorporate all possible risk factors, an independent association could be negated by residual confounders (i.e., unmeasured or imprecisely measured predictors that are associated with the marker and with the outcome). For example, an inverse association between older employees of a company and salary level does not necessarily mean that the firm discriminates on the basis of age, because a third unexamined variable (e.g., educational level) may lead to lower productivity, and thus lower wages, in older workers.
The confounding effect of an imprecisely measured variable can be understood by considering an example in which gray hair appears to be a risk factor for CHD. If age is unknown, a gray-haired man is more likely to have a myocardial infarction than a man without gray hair. However, if 2 men are the same age, but only 1 has gray hair, it is not likely that the gray-haired patient will be at increased risk. The initial association is spurious because gray hair, a proxy for age, is no longer an independent risk factor when age is added to the model.
The results of the Rotterdam Study illustrate the potential effect of a multivariate model in an elderly cohort. Initially, there was an independent association between CRP and myocardial infarction after adjustment for age and gender, but this largely disappeared after further adjustment for 7 risk factors routinely assessed in clinical practice.
Such examples demonstrate why the odds ratio is not a reliable basis for causal inference. To be valid, causal inference requires a scientifically supported theory that can explain the relation between 2 variables, together with empirical evidence, the highest level of which (Table 2) is provided by randomized, controlled trials.

Table 2: Categories for the evaluation of inflammatory markers as predictors of cardiovascular disease

* Level C evidence is based on the consensus of experts.

Test performance
In evaluating the clinical utility of a biomarker, investigators must determine whether it improves the predictive accuracy of traditional models (i.e., the standard of care). It has been argued that statistical associations, particularly when the odds ratio is ?3 (the size often seen in epidemiologic studies), do not necessarily provide this information. Instead, it is recommended that measures of discrimination and calibration be considered together. Discrimination is defined as the ability of a test to differentiate between those who will develop a disease and those who will not; model calibration, accompanied by goodness-of-fit statistics, indicates how the risk predicted by a biomarker (or a model that includes the biomarker) relates to the observed risk in a reference population.
The receiver-operating characteristic curve, which plots sensitivity and specificity, assesses discrimination. A marker that perfectly discriminates between subjects with and without a disease (or the likelihood of a disease) has an area under the curve of 1.0 (a 100% true-positive rate), whereas an area under the curve of 0.5 means that the discriminatory capacity is no better than chance.
Typically, traditional CHD or CVD prediction models have areas under the curve of 0.64 to 0.81, and the addition of CRP generally does not increase the curve area. Moreover, in the Atherosclerosis Risk in Communities (ARIC) study, graphs representing the predicted 5-year probability of a CHD event across deciles of risk were essentially the same for the basic model of traditional risk factors and the basic model plus log CRP. On the basis of these data, the investigators do not expect CRP to provide substantial improvement when added to traditional models for identifying patients at risk for a first coronary event.
In an analysis involving 15,048 initially healthy subjects from the Women's Health Study (WHS), the investigators basically used discrimination (i.e., the c-index, which is equivalent to the area under the curve), information criteria, and calibration to compare cardiovascular risk prediction models with and without hs-CRP. Although the c-index was minimally affected by the inclusion of hs-CRP after adjustment for age, smoking, and systolic blood pressure, prediction models that included hs-CRP were better calibrated than models without hs-CRP according to several global measures of fit. In particular, the adjusted likelihood ratio statistic indicated that hs-CRP improved fit more than did total or low-density lipoprotein (LDL) cholesterol. On the basis of these data, the investigators concluded that hs-CRP was stronger than any of the lipid measures as an individual predictor of risk in initially healthy women aged ?45 years, particularly those originally classified as having ?5% risk.
In weighing this conclusion, several caveats are necessary. Whereas the causal link between LDL cholesterol and CHD is well established, and clinical trials have clearly shown that lowering cholesterol levels reduces coronary risk, the association between CRP and CHD is based on observational evidence. As in the case of hormone replacement therapy to reduce cardiovascular risk, the medical literature contains examples of clinical trials that have failed to confirm hypotheses derived from observational evidence. Therefore, until the results of randomized controlled trials are available, it seems premature to conclude that CRP is stronger than LDL cholesterol as a predictor of risk, despite the extensiveness of the WHS data.
The limitations of sensitivity, specificity, and the likelihood ratio must also be considered. Because populations tend to be heterogenous, one cannot assume that these values are for every individual and across all subgroups. In the WHS, for example, the likelihood ratio statistic (and other goodness-of-fit measures) indicated better calibration for models that included hs-CRP, which was interpreted to mean that hs-CRP improves the prediction of "actual" or "true" risk in the study sample. However, we do not know whether the risk level in the reference group (women aged ?47 years from the Framingham Heart Study) was "true" for all subgroups in the study sample (i.e., whether the characteristics of the reference and study groups were sufficiently comparable to warrant this assumption). Therefore, although hs-CRP may have adjunctive value in women aged ?45 years, particularly those at ?5% risk, its value as a predictor of "true" risk in this group cannot be ascertained from the WHS data.
Sensitivity and specificity require that test results be classified as positive or negative. This means that patients with markedly elevated results are grouped with those whose results are slightly higher than the cut point. In the case of hs-CRP, however, simply knowing that a patient's level is >3 mg/L is not sufficient, because a substantial elevation (?10 mg/L) represents an acute-phase response that could indicate the presence of an infective, inflammatory, or neoplastic condition. In contrast, likelihood ratios, which are calculated using sensitivity and specificity data, can assign a specific value to each level of abnormality, thereby capturing the degree of abnormality of test results. In primary care populations, however, values at the far ends of the distribution, although potentially offering the most helpful diagnostic information, are least precise because of very wide confidence intervals due to sparse data at the extremes.
Every quantitative method requires different assumptions for optimal performance, and there is no universally accepted mathematical tool that is valid in all cases. When evaluating the clinical utility of a biomarker, it is important to consider the totality of the evidence within a study, the accumulated evidence from multiple independent data sets, and the limitations of each statistical measure.

Other limitations
In some studies, the CRP range may extend well beyond the low-grade elevations considered useful for predicting CHD events (Table 3), and the association with increased risk may acquire significance only at CRP levels >=10 mg/L. Patients with CRP levels >9.99 mg/L should be checked for inflammatory, infective, or neoplastic conditions, some of which (e.g., Chlamydia pneumoniae infection, rheumatoid arthritis) may confound the association between CRP and clinical or subclinical CVD if they are associated with both. In general, these studies do not report the exact CRP level at which the risk increase becomes significant for the highest versus the lowest segment of the distribution. If the association acquires significance at CRP levels >=10 mg/L (indicative of an acute-phase response), it may be confounded by the presence of other conditions (eg, inflammatory, infective, metabolic) that are associated both with CRP elevations and with CVD.

Table 3: Epidemiologic data: C-reactive protein distributions in individual studies

CARE = Cholesterol and Recurrent Events;
CHS = Cardiovascular Health Study;
FHS = Framingham Heart Study; HHP = Honolulu Heart Program;
MONICA = Monitoring Trends and Determinants in Cardiovascular Disease;
KIHD = Kuopio Ischaemic Heart Disease Risk Factor Study;
MRFIT = Multiple Risk Factor Intervention Trial;
PHS = Physicians' Health Study;
WHI = Women's Health Initiative;
WHS = Women's Health Study.

? Nested case-control study.
† CHD events (definitions vary).
‡ Cardiovascular event or cardiovascular death.
§ Secondary prevention.

Although CRP levels are assessed once at baseline in most epidemiologic studies, data from a large prospective cohort suggest that 3 serial determinations are needed to correct for within-subject variation and achieve a reliability coefficient of ?0.75, which was the reliability of total cholesterol in this study.
Given this intrasubject variability, the Centers for Disease Control and Prevention and American Heart Association guidelines advise clinicians to average the results of 2 measurements, 2 weeks apart, when using hs-CRP to augment risk prediction.

C-Reactive Protein: Statin Therapy
Even if CRP does not augment traditional risk assessment, it may be pathophysiologically related to CHD and may serve as a target of therapy. 48 A number of statin trials have linked CRP levels with coronary risk, leading to the hypothesis that treatment initiated on the basis of CRP elevations can improve patient outcomes. The following discussion of clinical trial evidence, although focused on primary prevention, considers selected secondary prevention trials.
Since the first statin was approved in 1987, a total of 7 randomized controlled trials (n = 54,191) have shown that these cholesterol-lowering drugs safely reduce the relative risk for a cardiovascular event by 1/4 to 1/3 in subjects at varying risk levels. Although statins were designed to reduce cholesterol by inhibiting 3-hydroxy-3-methylglutaryl coenzyme A reductase, the enzyme involved in de novo cholesterol synthesis, their benefit may be mediated by multiple mechanisms, including the reduction of the inflammatory component of CHD as measured by CRP and other markers.

Preliminary (hypothesis-generating) trial
Conducted in primary prevention subjects with average total cholesterol and LDL cholesterol levels but less than average levels of high-density lipoprotein cholesterol, the Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS) found that lovastatin reduced the risk for a coronary event by 37% versus placebo. According to a post hoc analysis, the coronary event rate increased with increasing quartiles of baseline CRP (from <1.6 to >3.5 mg/L), so that the relative risks with placebo versus lovastatin were 1.0, 1.2, 1.3, and 1.7 (p = 0.01). After 1 year, the median CRP level with lovastatin had decreased by 14.8% (0.2 mg/L) from baseline, whereas placebo produced no change. Furthermore, lovastatin versus placebo reduced cardiovascular risk by approximately 45% in participants with LDL cholesterol levels >149 mg/dl at entry and in those with LDL cholesterol levels <149 mg/dl and CRP levels higher than the median. This suggests that statin therapy may lower CRP levels and decrease coronary risk in primary prevention patients with evidence of low-grade inflammation, even when hyperlipidemia is not present.

Hypothesis-testing trials
The Pravastatin Inflammation/CRP Evaluation (PRINCE) study sought to determine whether pravastatin 40 mg/day has an effect on CRP levels at 12 and 24 weeks in either a double blind, placebo controlled, primary prevention cohort or an open-label, secondary-prevention cohort.
After 24 weeks, pravastatin had lowered CRP levels by a median of 0.02 mg/dl from baseline (approximately 14%) in the 2 treatment arms, whereas placebo produced no change. Primary prevention subjects also experienced significant decreases in CRP with pravastatin versus placebo, both overall (16.9%, p <0.001) and in several prespecified subgroups. In an intent-to-treat analysis that included 311 subjects with missing data, assigning them a value of zero for the change from baseline, the reduction in CRP was 7.1% (p <0.001) versus placebo. As in AFCAPS/TexCAPS, these CRP reductions were quite small, albeit statistically significant.
All pravastatin-treated subjects had notable improvements in LDL cholesterol and other lipid levels from baseline, while the change in CRP at 24 weeks was minimally associated with changes in each lipid parameter. Moreover, change in LDL cholesterol level was not predictive of change in CRP. This suggests that pravastatin lowers CRP levels in a manner largely independent of its effect on LDL cholesterol.
A clinical trial to determine whether CRP reduction improves patient outcomes must await the development of an agent whose primary benefit is the reduction of vascular inflammation. However, in an a priori analysis conducted as part of the 2-year Pravastatin or Atorvastatin Evaluation and Infection Therapy (PROVE-IT) trial, patients with acute coronary syndrome and CRP levels <2 mg/L after statin therapy were at significantly lower risk for a primary end point event (all-cause mortality, myocardial infarction, unstable angina requiring rehospitalization, revascularization, or stroke) than those with CRP levels ?2 mg/L, regardless of achieved LDL cholesterol level. Similar to the PRINCE trial, the correlation between achieved LDL cholesterol and CRP levels in PROVE-IT was minimal.
Overall, there was a 16% reduction in risk for all-cause mortality or recurrent coronary events with high-dose atorvastatin versus standard-dose pravastatin. This benefit began to emerge after 30 days, perhaps in part because of a significantly greater decrease in CRP levels with intensive versus standard therapy at that early time point. In contrast, clinical trials of statin therapy versus placebo in subjects with stable CHD required 1 to 2 years for a separation of the event curves. Although high-dose therapy was more effective in reducing LDL cholesterol and CRP concentrations, there was little evidence of an improved outcome with either agent once target levels of both parameters were reached.
The use of composite end points in time-to-event trials leads to higher event rates, making it possible to design studies with smaller sample sizes or shorter follow-up periods. However, the selection of end-point components is crucial. In PROVE-IT, a dissection of the primary end point reveals that the overall benefit was driven by significant reductions in 2 of the outcomes: revascularization and rehospitalization for unstable angina. Revascularization, in particular, may be subject to bias, because it is dependent on clinical judgment. Although PROVE-IT is a secondary prevention trial, with results that are specific to subjects after acute coronary syndrome, it illustrates the importance of critically analyzing the composite end points of all trials.
Currently, Justification for the Use of Statins in Primary Prevention: An Intervention Trial Evaluating Rosuvastatin (JUPITER) is being conducted to determine whether statin therapy will reduce the risk for a first major cardiovascular event, including revascularization, in subjects at high risk based on hs-CRP elevations, despite low to normal LDL cholesterol levels. This hypothesis is derived in part from the AFCAPS/TexCAPS analysis discussed previously. Specifically, JUPITER will evaluate the use of rosuvastatin 20 mg/day versus placebo in 15,000 men (?50 years old) and women (?60 years old) with LDL cholesterol levels <130 mg/dl, no history of cardiovascular events or coronary risk equivalents (according to current guidelines), and hs-CRP levels ?2 mg/L. Subjects will be followed for a mean of 3 to 4 years, the estimated time needed to accrue the 520 cardiovascular end points for which the study is powered.

Conclusion
Atherosclerosis, the principal cause of CVD, has been identified as an inflammatory disease. This suggests that some markers of inflammation (e.g., CRP) may serve as predictors of disease progression or clinical events. In guidelines issued jointly by the Centers for Disease Control and Prevention and the American Heart Association, the optional use of hs-CRP is recommended as an adjunct to traditional risk factors in intermediate-risk patients. This is based on epidemiologic evidence of an association between low-grade CRP elevations and an increase in coronary events. However, an association between a marker and a disease does not necessarily indicate predictive capacity in individual patients. According to statistical measures of test performance (e.g., sensitivity, specificity, calibration), hs-CRP has not been established for routine inclusion in most prediction models, although it may have utility in specific subgroups. Before CRP testing is routinely adopted in clinical practice, evidence is needed from rigorously conducted randomized controlled trials in which interventions based on low-grade CRP elevations are shown to improve patient outcomes. Regardless of the type of evidence, a critical assessment is necessary, with attention to potential sources of bias and confounding.