C-reactive protein (CRP) is the most extensively studied inflammatory risk marker, and a series of prospective epidemiologic studies have shown that higher levels of CRP are associated with incident cardiac and vascular events in healthy men and women (1- 9). Recent clinical practice recommendations from the Centers for Disease Control and Prevention (CDC) and the American Heart Association (AHA) support the use of CRP testing in selected patients to help assess cardiovascular (CV) risk (10), and trials are underway to explore therapeutic strategies based on CRP levels (11).

Current recommendations for CRP testing suggest uniform CRP thresholds to characterize the relative risk of CV events based on approximate tertile values in several populations (10). Because these populations predominantly consist of white subjects, it is unclear whether the recommended thresholds appropriately reflect CRP distributions for black men and women. Only limited comparisons of CRP levels between black subjects and white subjects have been performed (12- 14), and few data are available regarding the prognostic utility of CRP values in black subjects.

Identical CRP thresholds are also suggested for men and women (10). To date, comparisons of CRP levels between men and women have largely been performed across heterogeneous clinical trial cohorts that may not accurately reflect the general population (15). Only limited data are available directly comparing the distribution of CRP levels between men and women within the same study population (13,16- 17).

Given the paucity of available data regarding CRP distributions in different race and gender groups and the increasing interest in CRP for risk prediction, we sought to determine whether race and gender differences exist in the distribution of CRP levels using data from the Dallas Heart Study, a large, multiethnic, population-based sample.

The Dallas Heart Study is a multiethnic, population-based, probability sample of 6,101 subjects in Dallas County, and was designed to study CV disease. Details of the study design and characteristics of the enrolled cohort (18) and variable definitions (19) have been described previously. Briefly, a random probability sample of Dallas County residents ages 18 to 65 years old was obtained from a pool of 841,943 eligible subjects using the U.S. Postal Service Delivery Sequence File, with deliberate oversampling of black subjects. All participants provided informed consent to participate in the study, and the protocol was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center. The initial visit for all 6,101 participants included a detailed in-home interview for demographic and health-related data, as well as measurements of weight, heart rate, and five sequential blood pressure measures. All subjects between the ages of 30 and 65 years who completed the initial visit were invited to participate in a second visit to collect fasting venous blood and urine samples. No significant differences were noted in demographics, medical history, blood pressure, or body mass index (BMI) among subjects participating in the home interview and the phlebotomy visit (18). Sampling weights, reflecting the different probabilities of selection for participants and sample attrition between visits, were constructed to generate unbiased estimates of population frequencies (18). The present analyses were restricted to white and black subjects who completed the second visit (2,749 of 3,398 total subjects completing the second visit) because sample sizes in other race and ethnic groups were limited. Race/ethnicity was determined by subject self-report.

Blood samples were obtained after an overnight fast in ethylenediamine tetra-acetic acid tubes and were stored for ≤4 h at 4°C before processing. Plasma aliquots were frozen at −80°C until assays were performed. High-sensitivity CRP measurements were performed on thawed samples using the Roche/Hitachi 912 System, Tina-quant assay (Roche Diagnostics, Indianapolis, Indiana), a latex-enhanced immunoturbidimetric method (20). The minimal detectable range of this assay is 0.1 mg/l, and the upper limit is 20 mg/l. Clinical validation of this assay has been described previously (21).

Categorical data are reported as proportions and continuous data as mean values with standard deviations. Baseline demographic variables and CV risk factors were adjusted for sample weights and compared across categories of CRP using the chi-square trend test for categorical variables and the test for trend across ordered groups for continuous variables. The CRP values were adjusted for sample weights and compared between race and gender groups using the analysis of variance test for four group comparisons and the t test for two group comparisons. The results of these analyses were unchanged when log-transformed CRP values were used, and only the non-transformed data are shown. Log-transformed CRP values were also used in multivariable linear regression models in which CRP was the dependent variable. Subjects were categorized according to the CDC/AHA cut points for CRP into three relative risk groups: low-risk (<1 mg/l), intermediate-risk (1 to 3 mg/l) and high-risk (>3 mg/l) groups. Comparisons of the frequency of high-risk CRP levels for each race and gender group were performed using the chi-square test, with white men as the referent group. Adjusted odds ratios for CRP values >3 mg/l for each race/gender group (compared with white men) were determined using multivariable logistic regression models that included traditional CV risk factors as covariates, and using sampling weights. Analyses were performed using Stata version 8.2 (Stata Corp. LP, College Station, Texas) and SUDAAN version 9 (Research Triangle Institute, Research Triangle Park, North Carolina).

The baseline characteristics of the four different race and gender groups are shown in (Table 1). Compared with white subjects, black men and women had a significantly higher prevalence of diabetes and hypertension. Black women had a higher BMI and used estrogen less often than white women. Black men had the highest prevalence of tobacco use.

Subjects were unequally distributed across CRP risk categories (CRP <1 mg/l, n = 546 [20%]; CRP 1 to 3 mg/l, n = 848 [31%]; CRP >3 mg/l, n = 1,355 [49%]) (Table 2). As summarized in (Tables 2, 3), increasing CRP levels were associated with a number of CV risk factors in unadjusted analyses. Importantly, female gender and black race were both associated with higher CRP levels. The correlation between CRP and BMI was stronger than the correlation between CRP and any other continuous variable, including age and all measured lipid risk factors, and was greater in women compared with men (Table 3).

Black subjects had higher CRP levels than white subjects (median, 3.0 vs. 2.3 mg/l; p < 0.001) and women had higher CRP levels than men (median, 3.3 vs. 1.8 mg/l; p < 0.001) (Figure 1). In multivariate analyses adjusting for traditional CV risk factors, BMI, and estrogen and statin use, race and gender were independently associated with log CRP levels (p < 0.05 for each). Black men, white women, and black women all had higher CRP levels compared with white men, with the highest levels observed among black women (Figure 2). Excluding users of 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins) and oral estrogen had little effect on the CRP differences between black and white subjects (median, 2.6 vs. 2.1 mg/l; p < 0.001) or the differences between women and men (median, 2.7 vs. 1.8 mg/l; p < 0.001). In a similar sensitivity analysis, exclusion of these subjects only slightly attenuated the differences in median CRP levels among the four demographic groups (white men, 1.8 mg/l; black men, 2.0 mg/l; white women, 2.6 mg/l; black women, 3.2 mg/l; p < 0.001).

The proportion of subjects with high-risk CRP levels according to the CDC/AHA practice recommendation was estimated for each race and gender group (Table 4). Black men, white women, and black women were significantly more likely to have CRP values in the high-risk range compared with white men, with almost two-thirds of black women having CRP levels >3 mg/l. These differences persisted after excluding statin and estrogen users and after sample-weight adjustment (Table 4). Adjustment for traditional CV risk factors, BMI, statin and estrogen use, and sampling weights attenuated the differences between race and gender groups, yet an elevated CRP level remained more common in white women and black women than in white men (odds ratio [OR] 1.6, 95% confidence interval [CI] 1.1 to 2.5, p < 0.05; and OR 1.7, 95% CI 1.2 to 2.6, p < 0.05, respectively) (Figure 3). In contrast, the differences between black men and white men were no longer statistically significant after multivariate adjustment (OR 1.3; 95% CI 0.8 to 1.9).

An increasing BMI was associated with higher CRP levels in each race and gender group (p < 0.001 for each), but the increase in CRP levels with obesity was greater for women than for men (p value for interaction of obesity and gender on log CRP levels <0.001) (Figure 4).

In a large, multiethnic, population-based sample, we observed that black subjects have significantly higher CRP levels than white subjects, and women have significantly higher CRP levels than men. These findings persisted after exclusion of subjects taking statins and oral estrogen, and after adjustment for traditional CV risk factors and BMI. The associations were robust whether CRP was considered as a continuous or as a categorical variable (CRP >3 mg/l). Given the increasing use of CRP as a component of CV risk prediction, and the potential future use of CRP in clinical decision-making, our findings have important public health implications.

The potential for important racial and ethnic differences in CRP levels has been shown in studies of other ethnic groups (22- 24). However, studies comparing CRP levels between black and white populations are limited (13- 14,17,25), and the major studies examining the relationship between CRP levels and CV events have included few if any black subjects (1- 9). The National Health and Nutrition Examination Survey 1999 to 2000 compared CRP levels in black and white men age ≥20 years, an age range younger than in the current study (14). Although the investigators concluded that CRP levels were similar between black and white men based on median values, the upper tertiles for black men ages 45 to 64 years and ≥65 years were 5.4 and 7.0 mg/l, respectively, compared with 3.2 and 3.7 mg/l in white men. A similar analysis in women was performed using the same database and showed that the median CRP level in black women was 3.5 mg/l compared with 2.5 mg/l in white women (13). These differences persisted after excluding users of hormone replacement therapy, but data on statin use were not available. Recently, one study explored ethnic differences in CRP levels in women and found that black women had higher CRP levels than white women (25). However, this study consisted of a clinical trial cohort of postmenopausal women with much higher rates of estrogen use, lower BMI measures, and lower CRP values than our population-based sample. The present population-based study showed that in Dallas County, which likely represents a typical multiethnic U.S. urban population, the median CRP level was 30% higher in black subjects than in white subjects.

Previously, CRP levels were thought to be similar between women and men (15,26). However, these conclusions were based on comparisons of CRP levels across different studies with heterogeneous study populations, rather than on direct comparisons between men and women within the same sample. Moreover, these studies were largely performed in volunteers, who may differ from the general population in important ways. No differences in CRP levels between men and women were found in a compilation of European population-based studies (27). However, direct comparisons in U.S. research populations suggest that there may be gender differences in CRP levels (13,16). In our study of a large, contemporary, U.S. population-based sample, we observed that the median CRP level was almost twice as high in women compared with men.

Our contemporary, urban, and unselected cohort has a higher prevalence of overweight and obese subjects than has been reported in previous studies of CRP from more highly selected research populations (4- 5,25). We found that increases in BMI were associated with much greater increases in CRP in women than in men (Figure 4), a finding that may in part explain gender-based differences in CRP levels.

The finding of race and gender-based differences in CRP levels has important implications for the clinical use of CRP. Our findings suggest that the CRP risk categories outlined in the CDC/AHA statement, which were derived from selected research populations, may not be reflective of a contemporary, predominantly urban, multiethnic population with very high rates of obesity.

Several different interpretations of our findings are possible. Higher CRP levels in black subjects, and in particular in black women, may portend an increased risk for future CV events. Black subjects have a greater prevalence of CV risk factors than white subjects (28- 29), and higher CRP levels may reflect this risk factor burden. Also, racial differences in inflammation may directly contribute to the higher coronary heart disease mortality rates in black subjects (30). On the other hand, it is also possible that reliance on CRP for risk assessment in black subjects may overestimate the risk for cardiac and vascular events, and could lead to overuse of resources. For example, if a CRP value >3 were used as a criterion for initiating a therapeutic intervention, approximately 50% of black subjects between 30 and 65 years of age would be eligible for this intervention in Dallas County. Because our study uses a cross-sectional design, the prognostic implications of the racial differences in CRP distributions cannot be determined: long-term outcome studies in black subjects are needed to determine the implications of these findings and to determine whether the CDC/AHA risk thresholds are appropriate.

The clinical implications of gender differences in CRP levels also require consideration. Several studies have shown a strong association between CRP levels and CV outcomes in women (4,6). It is not clear, however, that the threshold for defining high-risk CRP levels for research cohorts can be generalized to the population, in which obesity is highly prevalent and more than 50% of women have CRP values >3.

In this multiethnic population-based study, we found significant race and gender differences in the distribution of CRP levels, with women and black subjects having higher CRP levels than men and white subjects, respectively. Studies with long-term follow-up will be needed to determine whether these differences in CRP distribution contribute to differences in clinical outcomes and whether CRP risk thresholds should be adjusted for different race and gender groups.

The authors thank Jody Balko for performing the CRP measurements and Dr. Helen H. Hobbs for her critical review of the manuscript.