Prevalence of high C-reactive protein levels in men and women based on coronary heart disease risk factors (cigarette smoking, borderline or elevated blood pressure, impaired fasting glucose level or diabetes mellitus, borderline-high or high cholesterol levels, borderline-high or high triglyceride levels, elevated body mass index, and low high-density lipoprotein levels, as specified in the “Methods” section). Asterisk indicates P<.001 within sex. Prevalence of high C-reactive protein level was significantly higher in women than in men (30.95% vs 19.7%; P<.001).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Miller M, Zhan M, Havas S. High Attributable Risk of Elevated C-Reactive Protein Level to Conventional Coronary Heart Disease Risk Factors: The Third National Health and Nutrition Examination Survey. Arch Intern Med. 2005;165(18):2063–2068. doi:10.1001/archinte.165.18.2063
C-reactive protein (CRP), a marker of systemic inflammation, is predictive of coronary heart disease (CHD) events. However, the extent to which high CRP levels (>3 mg/L) may be attributable to high cholesterol levels and other CHD risk factors has not been well defined.
The prevalence of high CRP levels in the third National Health and Nutrition Examination Survey (n = 15 341) was studied using CHD risk-factor cut points designated as abnormal (total cholesterol values, ≥240 mg/dL [≥6.22 mmol/L]; fasting blood glucose levels, ≥126 mg/dL [≥6.99 mmol/L]; blood pressure, ≥140/90 mm Hg; body mass index [BMI], ≥30 kg/m2; high-density lipoprotein cholesterol values, <40 mg/dL [<1.04 mmol/L] for men and <50 mg/dL [<1.30 mmol/L] for women; triglyceride levels, ≥200 mg/dL [≥2.26 mmol/L]; current smoking status) or borderline (total cholesterol values, 200-239 mg/dL [5.18-6.19 mmol/L]; fasting blood glucose levels, 100-125 mg/dL [5.55-6.94 mmol/L]; blood pressure, 120-139/80-89 mm Hg; BMI, 25.0-29.9 kg/m2, and triglyceride values 150-199 mg/dL [1.70-2.25 mmol/L], former smoking status), or normal.
Weighted multiple logistic regression analysis demonstrated that high CRP level was significantly more common with obesity (odds ratio [OR], 3.78; 95% confidence interval [CI], 3.28-4.35]), overweight (OR, 1.88; 95% CI, 1.62-2.18), and diabetes (OR, 1.91; 95% CI, 1.54-2.38) and that high CRP level was rare in the absence of any borderline or abnormal CHD risk factor in men (4.4%) and women (10.3%). Overall, the risk of elevated CRP level attributable to the presence of any abnormal or borderline CHD risk factor was 78% in men and 67% women.
These data suggest that elevated CRP levels in the general population are in large measure attributable to traditional CHD risk factors. Moreover, CRP level elevation is rare in the absence of borderline or abnormal risk factors. As such, CRP measurements may have limited clinical utility as a screening tool beyond other known CHD risk factors.
Inflammation has been implicated in the origination and progression of cardiovascular disease,1 and among the most actively studied biomarkers is C-reactive protein (CRP), an acute-phase reactant released predominantly by hepatocytes.2 While low-grade inflammation is predictive of coronary heart disease (CHD),3 even in the absence of elevated low-density lipoprotein cholesterol levels,4 the extent to which high CRP level is attributable to well-established CHD risk factors has not been well studied. This is an important issue to investigate in view of the numerous proponents for or against CRP level as a CHD screening tool5-8 and a recent Centers for Disease Control and Prevention and the American Heart Association scientific statement9 reserving CRP measurements for those patients at intermediate CHD risk (defined as a 10-year CHD risk range of 10%-20%). If elevated CRP level is intimately linked to conventional CHD risk factors, then routine screening of this biomarker would seemingly be less valuable to clinicians. Therefore, the present study was undertaken to evaluate the extent to which high CRP levels (>3 mg/L) may simply reflect expression of the pathobiologic changes induced by conventional CHD risk factors.
The third National Health and Nutrition Examination Survey (NHANES III)10 was conducted between 1988-1994 and used a stratified, multistage, probability sampling design to produce estimates generalizable to the US population. A total of 20 050 adults 18 years or older were screened for the study. A total of 15 341 subjects were included in the study, excluding subjects with any of the following data points missing from the record: CRP values (n = 3081), glucose levels (n = 1394), high-density lipoprotein cholesterol levels (n = 84), triglyceride levels (n = 80), body mass index (BMI, calculated as weight in kilograms divided by the square of height in meters) (n = 32), blood pressure values (n = 29), race designation (n = 8), and smoking status (n = 1). The dependent variable, serum CRP concentrations, was measured at the University of Washington Department of Laboratory Medicine using a Behring (Dade Behring, Deerfield, Ill) latex-enhanced nephelometer analyzer system; the assay was not of high sensitivity. We defined elevated levels of CRP as greater than 3 mg/L based on the recent Centers for Disease Control and Prevention and the American Heart Association scientific statement.9
Established cut points were used for the categorical variables in defining normal, borderline, or abnormal CHD risk factors. They included cigarette smoking (never, former, or current); blood pressure (normal, <120/80 mm Hg; prehypertension, 120-139/80-89 mm Hg; and hypertension, ≥140/90 mm Hg or receiving blood pressure medications)11; fasting glucose levels (normal, <100 mg/dL [<5.55 mmol/L]; pre–diabetes mellitus, 100-125 mg/dL [5.55-6.94 mmol/L]; and diabetes mellitus, ≥126 mg/dL [≥6.99 mmol/L] or receiving insulin or oral hypoglycemic agents)12; total cholesterol levels (normal, <200 mg/dL [<5.18 mmol/L]; borderline high, 200-239 mg/dL [5.18-6.19 mmol/L]; and high, ≥240 mg/dL [≥6.22 mmol/L] or receiving lipid-lowering medications)13; triglyceride levels (normal, <150 mg/dL [<1.70 mmol/L]; borderline high, 150-199 mg/dL [1.70-2.25 mmol/L]; and high, ≥200 mg/dL [≥2.26 mmol/L] or receiving lipid-lowering medications)13; high-density lipoprotein cholesterol values (normal, ≥40 mg/dL [≥1.04 mmol/L] in men and ≥50 mg/dL [≥1.30 mmol/L] in women; low, <40 mg/dL [<1.04 mmol/L] in men and <50 mg/dL [<1.30 mmol/L] in women)13; and BMI (normal, <25; overweight, 25-29; and obese, ≥30).14 We selected total cholesterol rather than low-density lipoprotein cholesterol because of the high number of missing values for the latter (n = 7829). Finally, we also classified physical activity into 4 categories: vigorous activity, moderate activity, light activity, and sedentary as previously defined.15 We also determined the use of estrogen replacement therapy in women who had surgical or natural menopause.
The association between CHD risk factors and the prevalence of elevated CRP levels was assessed using χ2 tests. Additional analyses were conducted using SAS statistical software, version 9 (SAS Institute Inc, Cary, NC). Because NHANES III oversampled minority populations, weighted analyses were performed to ensure that results were generalizable to US population demographics (SAS/STAT, version 9.1). The odds ratio of elevated CRP levels was determined in the presence (or absence) of each of the above CHD risk factors, and their 95% confidence intervals (CIs) were calculated by a multiple logistic regression model that used survey weights. The attributable risk of elevated CRP level due to one or more CHD risk factors was computed using the following formula: (PD/E – PD/Ē) NE/ND, where PD/E is the weighted percentage of high CRP levels with abnormal or borderline CHD risk factors, PD/Ē is the weighted percentage of high CRP levels without CHD risk factors, NE is the number of estimated individuals with abnormal or borderline CHD risk factors in the population, and ND is the estimated total number of cases in the population.
A total of 15 341 adult men and women were included in the analysis. Table 1 provides baseline and demographic information of the NHANES III cohort.
The frequency and prevalence of high CRP levels for each of the CHD risk factors are listed in Table 2. Blacks had higher CRP levels than whites. Overall, there were marked increases in the prevalence of elevated CRP levels especially associated with higher blood pressure, glucose levels, and BMI. Compared with never smokers, high CRP level was more common in both former and active cigarette smokers. A graded increase in the prevalence of high CRP levels was also apparent as blood pressure, glucose, cholesterol, triglyceride, and BMI cut points increased from normal to borderline and abnormal levels, with larger increases occurring between borderline and abnormal ranges. Similar changes were also observed in univariate analysis comparing sedentary and physically active subjects.
The overall prevalence of elevated CRP level was 25.7%. Compared with subjects who never smoked cigarettes and were normotensive, normolipidemic, euglycemic, and not overweight (n = 813), the presence of at least 1 borderline or abnormal CHD risk factor (n = 14 528) was associated with an approximate 3-fold higher prevalence of CRP level higher than 3 mg/L (8.7% vs 26.7%) (P<.001) (Figure). Only 4.4% of men and 10.3% of women with a favorable risk-factor profile had elevated CRP levels. In the presence of any borderline CHD risk factor, high CRP level increased 1.5- to 2-fold and approximately 3- to 5-fold with at least 1 abnormal CHD risk factor.
The prevalence of elevated CRP level was also higher among postmenopausal women receiving estrogen replacement therapy (51%; n = 473) than among estrogen nonusers or former users (37%; n = 3105) (P<.001). These data are consistent with previous studies that have shown higher CRP levels in women than in men16 and higher levels in women undergoing estrogen replacement therapy than in women not undergoing such therapy.17,18 However, because of the small number of women receiving estrogen replacement therapy, this variable was not included in our multiple logistic regression model.
The inclusion of moderate and/or vigorous physical activity sharply reduced the number of men and women without any CHD risk factors from 813 to 71 of 15 341 subjects. Because only a very small fraction of these normal subjects had elevated CRP levels (n = 7, or 0.05% of the total cohort), physical activity was not included in multiple logistic regression analysis.
Weighted multiple logistic regression analysis adjusted for age and race identified the factors that were most highly associated with elevated CRP level: excess weight, hypertension, female sex, diabetes, cigarette smoking, and low high-density lipoprotein cholesterol values (Table 3). Table 4 demonstrates that the attributable risk of high CRP level in men and women was primarily accounted for by the presence of at least 1 abnormal CHD risk factor. In NHANES III, the attributable risk of high CRP level for the presence at least 1 abnormal or borderline CHD risk factor was approximately 78% for men and 67% for women.
Until recently, many believed that only 50% of CHD was accounted for by traditional risk factors such as cigarette smoking, diabetes mellitus, hypertension, and high blood cholesterol level.19 This in turn led to intensive investigations to uncover other potentially important biomarkers influencing atherothrombosis and CHD risk. Among the numerous markers surveyed, CRP level has been the most thoroughly examined with more than 1900 entries in PubMed during the past 9 years. However, recent studies have affirmed that traditional risk factors account for the overwhelming majority of CHD cases in the United States.20-22
The present study reinforces the central role that traditional CHD risk factors play in atherothrombosis by demonstrating that a pivotal biomarker of systemic inflammation, elevated CRP level, is generally accompanied by borderline or abnormal CHD risk factors and rarely occurs in their absence. Thus, while these data in no way negate the important influence that inflammation plays in promoting or accelerating CHD,23 they underscore the likely role that well-established risk factors contribute to the inflammatory process. Several lines of evidence support a strong interrelationship between inflammatory biomarkers and CHD risk factors. For example, hepatic production of CRP is up-regulated by visceral adipocyte-mediated secretion of inflammatory cytokines.24 In addition to interleukin 6 and nuclear factor κ, other proinflammatory mediators released from adipocytes, notably angiotensin II, contribute to dyslipidemia, insulin resistance, and systemic hypertension.25,26 The present study reaffirms that elevated CRP level is strongly correlated with measures of adiposity27 as defined by BMI measurements for overweight and obesity. These close-knit relationships make it difficult to disentangle the distinctive impact of CRP level beyond well-established risk factors.
Moreover, the treatments commonly used to reduce conventional CHD risk factors also lower CRP levels.28-31 Therefore, given the tight interrelationship between systemic inflammation and CHD risk factors, and in view of the extremely low prevalence of high CRP level in the absence of the 8 risk factors in NHANES III (ie, 0.05%), it is difficult, if not impossible, to prove the “CRP hypothesis” (ie, reduction of CRP level independently diminishes CHD event rates). This would necessitate the development of targeted agents that would lower CRP levels and reduce CHD without affecting values for lipids, blood pressure, glucose, and visceral adiposity.
Unless these concerns can be assuaged, we do not recommend screening for elevated CRP levels. In fact, CRP screening may be counterproductive because finding normal levels (<1 mg/L) could dissuade patients and physicians from using intensive lifestyle therapy that might prove useful. For example, a low CRP level (<1 mg/dL) in an overweight, dyslipidemic, prehypertensive, or prediabetic subject might provide false reassurance that no further therapy is required. However, failure to institute lifestyle strategies aimed at normalizing body weight, lipid levels, blood pressure, and glucose values leads to greater disease progression and increased CHD event rates,32 regardless of CRP levels. In an era where obesity-related illnesses have grown to epidemic proportions,33 more intensive efforts should be directed toward eradicating the primary culprit of elevated CRP levels, visceral adiposity.
There are several limitations associated with the present study. First, to circumvent oversampling of selected groups in NHANES III (eg, blacks), weighted analyses were performed. Second, the highly sensitive CRP assay currently in widespread use, enabling detection of lower CRP levels (eg, <2.1 mg/L), was not used in NHANES III. While this shortcoming did not affect the primary question raised in the present study related to high CRP level, it restricted further analyses using moderate to high CRP cut points (ie, 1-3 mg/L).34 Third, 15% of the cohort had missing data for CRP levels, resulting mostly from refusal to have blood drawn or insufficient blood sampling. However, there were no material differences in demographic characteristics between the subjects with and without blood for analysis. Finally, the objective cut points used to define normal, borderline, and abnormal lipid levels, blood pressure, glucose values, and measures of adiposity were considerably easier to quantify in NHANES III compared with more subjective risk factors, such as mental stress.
Even though a strong association between CRP level and overall CHD risk exists, 1 cross-sectional study did not demonstrate that CRP level correlated with individual CHD risk factors.35 However, the size of that cohort was considerably smaller than NHANES III (n = 1666 vs 15 341), and measures of adiposity, the most powerful variable influencing CRP level, were not undertaken. Because the presence of multiple risk factors (as opposed to a single factor) increases the likelihood of CHD,36 focusing on individual parameters may underestimate overall CHD risk, especially in the United States, where the prevalence of multiple risk factors (eg, metabolic syndrome) is high.37 In the present study, elevated CRP level was very prevalent in the presence of at least 1 borderline or abnormal risk factor.
Based on the recent results of the INTERHEART study38 of about 30 000 people, where 90% of CHD could be accounted for by 9 measures of risk (abdominal obesity, alcohol intake, cigarette smoking, hypertension, diabetes, abnormal lipids [apolipoprotein B/apolipoprotein A-I ratio], low consumption of fruits and vegetables, physical inactivity, and psychosocial factors), it is tempting to speculate that a similarly high attributable risk relating CRP level and CHD risk factors might have been ascertained had NHANES III enrolled twice the number of subjects and been able to measure parameters such as physical activity with greater precision.
In summary, these data suggest that elevated CRP level is primarily attributable to traditional CHD risk factors. As such, measurement of CRP levels may have limited clinical utility as a screening tool for CHD risk assessment unless randomized clinical trials can demonstrate that lowering CRP levels offsets CHD events beyond well-established lifestyle and pharmacologic modalities. In the meantime, more intensified efforts should be aimed at promotion of smoking cessation, increasing physical activity, and lowering of elevated blood pressure, lipid levels, and visceral adiposity because all have proven impact on CHD reduction.
Correspondence: Michael Miller, MD, Division of Cardiology, Room S3B06, 22 S Greene St, Baltimore, MD 21201 (email@example.com).
Accepted for Publication: March 17, 2005.
Financial Disclosure: None.
Funding/Support: This work was supported in part by National Institutes of Health grant HL-61369 and a Veterans Affairs Merit Award to Dr Miller.
Previous Presentations: This study was presented in part at the American Heart Association Annual Scientific Sessions; November 8, 2004; New Orleans, La.
Acknowledgment: The authors acknowledge Laurence Magder, PhD, for helpful discussions related to statistical methods.