Dr. del Rincón is the recipient of a Lyndon B. Johnson Award from the American Heart Association, Texas affiliate
Association between carotid atherosclerosis and markers of inflammation in rheumatoid arthritis patients and healthy subjects
Article first published online: 2 JUL 2003
Copyright © 2003 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 48, Issue 7, pages 1833–1840, July 2003
How to Cite
del Rincón, I., Williams, K., Stern, M. P., Freeman, G. L., O'Leary, D. H. and Escalante, A. (2003), Association between carotid atherosclerosis and markers of inflammation in rheumatoid arthritis patients and healthy subjects. Arthritis & Rheumatism, 48: 1833–1840. doi: 10.1002/art.11078
- Issue published online: 2 JUL 2003
- Article first published online: 2 JUL 2003
- Manuscript Accepted: 27 MAR 2003
- Manuscript Received: 3 OCT 2002
- Arthritis Foundation
- NIH. Grant Numbers: K23-HL-004481, R01-HD-37151, K24-AR-47530, R01-HL-24799, R37-HL-36820, M01-RR-01346
To examine the relationship between markers of systemic inflammation and carotid atherosclerosis in patients with rheumatoid arthritis (RA) and healthy controls.
Carotid artery intima-media thickness (IMT) and carotid plaque were measured using high-resolution B-mode ultrasound in 204 patients with RA, ages 40–85, and 102 age- and sex-matched healthy persons. No subject in either group had ever smoked cigarettes. Erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) were used to measure systemic inflammation. The relationship of the carotid artery IMT and carotid plaque to inflammation markers was examined, adjusting for age, sex, RA versus control status, and the cardiovascular (CV) risk factors hypercholesterolemia, systolic blood pressure, diabetes mellitus, and body mass index (BMI).
A significant linear trend for increased carotid artery IMT was associated with increasing ESR and CRP categories (r = 0.16, P = 0.004 for ESR, and r = 0.13, P = 0.02 for CRP). These trends did not differ among RA cases and controls, and were independent of age, sex, and CV risk factors. The difference in carotid artery IMT between the lowest and highest categories of ESR was 0.221 mm (95% confidence interval [95% CI] 0.767–1.020, P = 0.02). The difference between extreme CRP categories was 0.275 mm (95% CI 0.039–0.509, P = 0.02). Both remained significant after CV risk factor adjustment. Carotid plaque displayed a similar relationship to markers of inflammation.
Increased carotid artery IMT and the presence of carotid plaque are associated with markers of systemic inflammation in patients with RA and in healthy subjects. This observation is consistent with hypotheses that assign a role to systemic inflammation in atherosclerosis, and may have implications regarding RA and other chronic inflammatory diseases.
Chronic joint inflammation characterizes rheumatoid arthritis (RA). Systemic inflammation occurs as well, as evidenced by elevated levels of inflammation markers in the blood (1), and by extraarticular involvement. Patients with RA experience cardiovascular (CV) events more often than expected (2), and their mortality attributable to CV causes is increased (3–8).
The mechanism of the increased CV event rate in RA is not well understood. In healthy subjects, there is an association between inflammation markers and CV disease (9–12). This raises the possibility that the severe systemic inflammation seen in RA may bring about its high CV mortality by causing accelerated atherosclerosis (13,14). In the present study, the extent of carotid atherosclerosis in patients with RA and in healthy subjects was measured, and its relationship to inflammation markers in the blood was examined.
PATIENTS AND METHODS
Patients with RA.
From 1996 to 2000, patients who met the American College of Rheumatology (formerly, the American Rheumatism Association) classification criteria for RA (15) were enrolled into ÓRALE (Outcome of Rheumatoid Arthritis Longitudinal Evaluation), a study of disability in RA (16). Patients were recruited consecutively from 6 local rheumatology clinics: 1) a county-funded clinic; 2) a Veterans Administration clinic; 3) a private, university-based faculty practice; 4) a community-based, 7-rheumatologist private practice; 5) an Army medical center; and 6) an Air Force medical center. All patients resided in Bexar County, Texas, or nearby communities. Our sample has been described in previous publications (17). Between August 2000 and January 2002, all ÓRALE patients were invited to the University of Texas at San Antonio Health Science Center's General Clinical Research Center for an evaluation of carotid atherosclerosis. In the present analysis, only patients who had never smoked cigarettes and were 40 years of age or older were included.
Nonsmoking patients with RA were stratified into 18 categories defined by sex and 5-year age intervals from 40 to 85. Within these categories, 1 volunteer in good health who had never smoked was recruited per every 2 patients with RA. To recruit volunteers, advertisements were placed on billboards in our institution, in local churches and community centers, and in the local newspaper. Independent living facilities for older people were visited, and recruiting was done by word of mouth. Respondents whose age–sex category was not yet filled, who were in good health by self-report, had never smoked cigarettes, and who did not have RA or systemic lupus erythematosus were invited. Each control was paid $50.
Data collection procedures.
This study was approved by the Institutional Review Board. All subjects gave written, informed consent. A physician and a trained research associate evaluated the subjects.
CV risk factor ascertainment.
Blood pressure was measured and the body mass index (BMI) was calculated. Among patients with RA, diabetes mellitus and hypercholesterolemia were ascertained by self-report, accompanied by a medical record and medication review. These diagnoses were considered to be present if a physician had recorded the diagnosis in the medical records, or if a patient was taking antidiabetic or lipid-lowering medications. Among the controls, whose medical records were unavailable, participants were asked whether a physician had ever told them that they had diabetes, elevated blood lipids, or cholesterol. These conditions were considered to be present in the controls if they responded affirmatively or if they were taking antidiabetic or lipid-lowering medications.
Other data elements.
Age, sex, and race/ethnicity were ascertained by self-report (17).
Laboratory and radiographic studies.
The erythrocyte sedimentation rate (ESR) was measured using the Westergren technique, and the C-reactive protein (CRP) was measured by nephelometry (Quest Diagnostics, San Juan Capistrano, CA).
One technician performed a duplex scan of the carotid arteries in all patients and controls, following a standardized vascular protocol developed for the Multi-Ethnic Study of Atherosclerosis, for which she was certified (18). An ATL HDI-3000 high-resolution imaging machine with an L7-4 transducer (Philips Medical Systems, Bothell, WA) was used. The technician acquired 4 standardized B-mode images and a Doppler flow measurement from the right and left sides of the neck. The first image was of the distal common carotid artery, and the 3 others were centered on the site of maximum near and far wall thickening in the proximal internal carotid artery or carotid bulb. Results were recorded on Super VHS tape and mailed to a central facility (Ultrasound Reading Center, New England Medical Center, Boston, MA) for grading of the carotid artery intima-media thickness (IMT) and carotid plaque.
At the reading center, the recorded images were digitized at 30 frames per second, and fluctuations of the arterial diameter with the cardiac cycle were observed. Images were then selected and read by a single, certified reader who was blinded to all characteristics of the subjects (19). IMT measurements were obtained at each of the near and far walls of the right and left common carotid arteries, and anterior oblique, lateral, and posterior oblique views of the right and left internal carotid artery, for a total of 16 views per person. Results were summarized into 2 variables, 1 for the common carotid artery and 1 for the internal carotid artery. The maximal IMT of each of these was obtained by averaging the maximal measurements of the near and far walls and the right and left sides. Then a composite maximal IMT was calculated by averaging the common and internal carotid measurements. Results were expressed in millimeters. The images were also read for the presence of vessel plaque, identified as a discrete projection from the wall into the vessel lumen.
A single ultrasonographer acquired, and a single reader measured, all the images in this study. Nevertheless, to assess the technique's reliability, the reader reread 50 of the recorded images, and a different reader read a separate set of 50 images from the study subjects. The intrareader intraclass correlation coefficient for carotid artery IMT was 0.99, and the interreader coefficient was 0.94. In the case of carotid plaque, the intrareader kappa statistic was 1.0, while the interreader kappa statistic was 0.94.
Preliminary data inspection revealed distinctly skewed distributions for the carotid artery IMT, CRP, and ESR. This was approached in 2 ways. In one analysis, the ESR and the CRP were divided into quartiles. However, levels of these 2 markers were considerably higher among patients with RA than among controls, which would have resulted in erroneously high cutpoints for the lower quartiles. Thus, the inflammation marker quartile boundaries were defined from their frequency distribution among the controls in a manner analogous to that in studies of the healthy population showing a graded increase in CV risk from lower to higher inflammation marker categories (10). Because a large proportion of patients with RA had ESR or CRP levels exceeding the highest control value, an additional, fifth category for these patients with RA was created. In a second analysis, the uncategorized carotid artery IMT, ESR, and CRP were transformed toward normality, using the natural logarithm of these values in these analyses.
The ordinary least squares (OLS) regression was used to compare the unadjusted and age-, sex-, and CV risk factor–adjusted carotid artery IMT between patients with RA and controls, and between inflammation marker categories. OLS regression and Pearson correlation coefficients were used to test for trends across the ordered inflammation marker categories. In separate OLS models, the degree of association between uncategorized inflammation markers and the carotid artery IMT were measured. RA × ESR and RA × CRP interaction terms were also tested to look for differences between patients with RA and controls in the carotid artery IMT–inflammation association. For plaque analysis, the individual carotid artery was used as the unit of analysis. Logistic regression was used to compare the presence or absence of carotid plaque between vessels from patients with RA and controls, and to examine the degree of association between carotid plaque and inflammation markers. For plaque analysis, robust standard errors were estimated, adjusted for within-person clustering (20). For results of analyses using transformed variables, standardized linear or logistic regression coefficients are shown. A 2-sided P value of 0.05 was considered to reflect statistical significance, but we provide actual values for any P ≤ 0.1. Intercooled Stata, release 7.0 (Stata Corporation, College Station, TX), and SAS 8.0 (SAS Institute, Cary, NC) were used for all analyses.
The original sample included 779 patients with RA recruited from January 1996 to April 2000. Of these, 707 were eligible for ultrasonography as of August 2000 because they were still alive and had not moved out of Texas. At the time of this writing, 608 of the eligible patients with RA (86%) had undergone ultrasonography. The 204 patients with RA who had never smoked cigarettes, by self-report, and who were 40 years of age or older were included in the present study.
For the control sample, 585 subjects who responded to the recruitment notices were screened. Of these, 247 were not eligible because they were current or past cigarette smokers, 19 because they were not in good health, 30 because their BMI exceeded 30 kg/m2, 6 because they were diagnosed as having RA, 9 because they had unspecified inflammatory arthritis, and 1 because of a history of lupus erythematosus. Among eligible respondents, 159 were not studied because their age–sex category was full when they responded. An additional 6 eligible healthy controls were scheduled but did not keep their appointments, and 6 more were screened, but could not be relocated to schedule an appointment. The final sample included 102 nonsmoking persons in good health by self-report who were matched by age and sex to the patients with RA.
Comparison between patients with RA and controls.
Characteristics of the study subjects are shown in Table 1. As expected, there was no difference in age and sex between patients with RA and controls. The measured blood pressure was equivalent between the 2 groups. The patients with RA were less likely to have hypercholesterolemia, but were more likely to have diabetes mellitus, and had higher BMI. Also as expected, ESR and CRP were higher in patients with RA, as were the frequencies and titers of rheumatoid factor and antinuclear antibodies (Table 1).
|RA (n = 204)||Controls (n = 102)||P|
|Age, mean (range)||59.6 (40–83)||59.7 (40–81)||0.9|
|Male, no. (%)||23 (11)||12 (12)||0.9|
|Ever smoked cigarettes, no. (%)||0 (0)||0 (0)||–|
|Hypercholesterolemia, no. (%)||27 (13)||39 (38)||≤0.0001|
|Plasma cholesterol, mg/dl, mean ± SD||193 ± 46||192 ± 33||0.8|
|Systolic blood pressure, mm Hg, mean ± SD||144 ± 23||141 ± 23||0.2|
|Diabetes mellitus, no. (%)||36 (18)||1 (1)||≤0.0001|
|BMI, kg/m2, mean ± SD||29.6 ± 6.8||26.2 ± 4.5||≤0.0001|
|ESR, mm/hour, mean ± SD||41 ± 26||20 ± 13||≤0.0001|
|CRP, mg/liter, mean ± SD||12.3 ± 15.7||4.4 ± 5.4||≤0.0001|
|History of glucocorticoid use||≤0.0001|
|Currently using||91 (44)||0 (0)||–|
|Formerly used||35 (17)||0 (0)||–|
|Never used prednisone||80 (39)||102 (100)||–|
|Rheumatoid factor positive, no. (%)||157 (77)||1 (1)||≤0.0001|
|Rheumatoid factor titer, median (range)†||2 (0–9)||0 (0–1)||≤0.0001|
|Antinuclear antibody positive, no. (%)||162 (79)||40 (39)||≤0.001|
|Antinuclear antibody titer, median (range)†||3 (0–8)||0 (0–6)||≤0.0001|
Carotid artery IMT.
The carotid artery IMT increased with increases in the levels of inflammation markers (Tables 2 and 3). A significant linear trend was found for increasing carotid artery IMT as the level of inflammation markers increased. The Pearson correlation coefficient between the carotid artery IMT and ESR categories was 0.16 (P = 0.004) (Table 2), while that with the CRP categories was 0.13 (P = 0.02) (Table 3). The trend was independent of age, sex, and the measured CV risk factors (Tables 2 and 3). The unadjusted average difference between the lowest and highest ESR categories was 0.221 mm (95% confidence interval [95% CI] 0.767–1.020, P = 0.02). This difference increased to 0.236 mm (95% CI 0.067–0.404, P = 0.006) after age and sex adjustment, but decreased slightly to 0.175 mm (95% CI 0.006–0.344, P = 0.008), when CV risk factors were added to the adjustment model.
|ESR, mm/hour||No. of subjects||Mean carotid IMT, mm†||95% CI|
|CRP, mg/liter||No. of subjects||Mean carotid IMT, mm†||95% CI|
A similar pattern was observed with respect to the CRP categories (Table 3 and Figure 1). The unadjusted difference between the lowest and highest CRP categories averaged 0.275 mm (95% CI 0.039–0.509, P = 0.02). This difference remained at 0.271 (95% CI 0.068–0.475, P = 0.009) after age and sex adjustment, and increased slightly to 0.286 (95% CI 0.085–0.485, P = 0.005) with CV risk factors added.
Findings in the second analysis, using the uncategorized ESR and CRP, mirrored the above results. Significant unadjusted, and age-, sex-, and CV risk factor–adjusted associations were found between the carotid artery IMT and both the ESR and the CRP (Table 4). The association was modest in strength, with bivariate Pearson correlation coefficients <0.20 for both variables (Figure 1). The standardized regression coefficients shown in Table 4 represent the degree of change in the carotid artery IMT or carotid plaque associated with change in inflammation marker levels, expressed in SD. Thus, from Table 4, a 1-SD increase in the ESR is associated with an increase in carotid artery IMT equivalent to 0.19 SD, adjusted for age and sex.
|Dependent variable||Independent variable|
|Carotid artery IMT|
|Age- and sex-adjusted||0.19§||0.13†|
|Age-, sex-, and CV risk factor–adjusted||0.14†||0.10¶|
|Age- and sex-adjusted||0.17†||0.18†|
|Age-, sex-, and CV risk factor–adjusted||0.14¶||0.19§|
The mean (±SD) carotid artery IMT among the 204 patients with RA was 1.034 ± 0.434 mm, while among the controls the carotid artery IMT was 0.979 ± 0.449 mm (P = 0.30 for main effect). The size of the observed difference in carotid artery IMT between patients with RA and controls remained unchanged at ∼0.055 mm, even after adjusting for age, sex, and CV risk factors.
As was observed for the carotid artery IMT, carotid plaque displayed a positive association with the level of inflammation markers in patients with RA and controls combined. There was a significant trend for the proportion of vessels with plaque to increase as the level of ESR increased (P = 0.01), which remained significant after adjusting for age and sex (P = 0.02), but not after adjusting for CV risk factors (P = 0.07) (Table 5). In the case of CRP, there was a dramatic rise in the proportion of vessels with plaque between the first and second CRP categories, with a high proportion in all subsequent CRP categories. This remained significant even with age, sex, and CV risk factor adjustment (P ≤ 0.001) (Table 6). Carotid plaque was found in 121 of the 405 RA vessels imaged (30%), and in 51 of the 203 control vessels (25%) (odds ratio 1.27 [95% CI 0.82–1.96, P = 0.28]). This did not change with adjustment for age, sex, and CV risk factors.
|ESR, mm/hour||No. of carotid vessels imaged||Carotid vessels with plaque, no. (%)†||Odds ratio||95% CI|
|CRP, mg/liter||No. of carotid vessels imaged||Carotid vessels with plaque, no. (%)†||Odds ratio||95% CI|
A significant association was found between laboratory markers of systemic inflammation and the carotid artery IMT and carotid plaque, both of which are noninvasive measures of atherosclerosis. This observation is important because of its implications regarding the pathogenesis of atherosclerosis in RA and other rheumatic diseases (13, 14). Several of our findings support hypotheses that ascribe a causative role to systemic inflammation in atherosclerosis (21).
An association between carotid plaque and the ESR and CRP had not been previously noted in RA. In the case of the carotid artery IMT, published information about its association with the ESR is conflicting. One earlier report described a significant bivariate association between the ESR and the common carotid artery IMT (22), while other studies indicated that the two were not significantly associated (23, 24). The present study extends and clarifies prior studies by exploring the carotid wall–inflammation marker association in more detail, adjusting for the potential confounding influence of age, sex, and other CV risk factors, and using a more generalizable population sample.
The inflammation marker–carotid wall thickness association was significant, from both statistical and clinical standpoints. The carotid artery IMT in the highest ESR category exceeded that in the lowest category by an average of ≥0.200 mm, and the odds of vessel plaque between the 2 categories were nearly halved. Likewise, there was an 8-fold decrease in the odds of plaque between highest and lowest CRP categories. The potential clinical importance of differences of this magnitude is well illustrated in the Cardiovascular Health Study (CHS), a population-based study of atherosclerosis in the US that used vascular imaging studies very similar to ours (25). In that study, a difference in carotid artery IMT of 0.200 mm was associated with a doubling of the risk of CV events. In the CHS, the bivariate correlations (r) between the internal carotid artery IMT and age, systolic blood pressure, and smoking were 0.20, 0.15, and 0.12, respectively (26). This helps place the magnitude of the correlations we found between the carotid artery IMT and the ESR and CRP, shown in Table 4, in the proper perspective. Although the absolute size of the correlations appears modest, it may be comparable with those of other important risk factors.
The carotid artery IMT and the probability of vessel plaque increased progressively with increasing levels of both the ESR and the CRP, in effect exhibiting a dose-response relationship. This represents a biologic gradient, which supports hypotheses of a mechanistic relationship between systemic inflammation and atherosclerosis (21). The ESR and the CRP each measure different components of the acute-phase reaction. The ESR is a nonspecific marker of systemic inflammation, performed with fresh blood. It primarily reflects an increased serum concentration of fibrinogen in response to systemic inflammation. Fibrinogen, immunoglobulins, and other plasma proteins modify the blood's electrostatic environment, causing red blood cell clumping and increased sedimentation (27). There is a complex interaction between the inflammation and coagulation systems that is of great importance in thrombogenesis (28). The association between fibrinogen, as measured by the ESR, and carotid artery IMT that we observed here suggests that inflammation–coagulation interactions may also have a role in atherogenesis. The ESR's longer half-life compared with the CRP and other inflammation markers (29) may underlie in part its stronger correlations with the carotid artery IMT and carotid plaque. CRP is produced by the liver in response to interleukin-6, an earlier inflammation mediator (30, 31). It can be found, in conjunction with terminal complement components, within early atheromatous lesions, suggesting a pathogenic role for CRP in atherothrombosis (32).
That these dissimilar markers should share an association with atherosclerosis represents an effect consistency that further supports the notion of a causative role for systemic inflammation in atherosclerosis (21). However, the exact mechanistic links between the two processes are not fully understood at present, nor are the causes of inflammation in atherosclerosis known. Atherosclerosis and RA share certain immune dysregulation characteristics (33), and the possibility that they could share etiologies, perhaps infectious, is intriguing and deserving of careful research.
High-resolution, B-mode ultrasound measurement of the carotid artery IMT is a widely used, noninvasive method to assess atherosclerosis in research. The carotid artery IMT's validity is firmly supported by 3 lines of evidence. First, histologic studies demonstrate a close correlation between carotid and coronary atherosclerosis, validating the carotid arteries as indicative of what is happening in the coronary arteries (34). Second, ultrasonographic measurements correlate highly with histologic measurements of carotid artery IMT (35). Third, the carotid artery IMT is associated with CV risk factors, prevalent CV disease, and atherosclerosis in other vascular beds (36), and is a strong predictor of incident myocardial infarction and stroke (25). When performed by a single sonographer who follows an established protocol and is read by a single expert reader, as was done in this study, the technique is highly reliable.
We accounted for the possibility of confounding by matching (age and sex), specification (smoking), and statistical adjustment (remaining CV risk factors) (37). People who had smoked cigarettes were excluded because there was considerable variation in smoking exposure level among the patients in the ÓRALE cohort, which would have made accurate matching on this variable impracticable. Moreover, smoking is both a risk factor for atherosclerosis and may be associated with increased inflammation markers. Further specification to handle potential confounding by other CV risk factors would have brought an unacceptable loss of generalizability. Among the patients with RA we studied, only 40 women and 5 men were free of all the CV risk factors we measured. Thus, we handled these risk factors statistically, including them as covariates in regression models. This did not efface the association between inflammation markers and atherosclerosis and, in the case of the association between CRP and carotid plaque, adjustment led to a strengthening of the effect (Table 4). Thus, the association we observed is unlikely to be due to confounding by these covariates.
A goal of our study was to compare the extent of carotid atherosclerosis between patients with RA and healthy subjects. We found that the carotid artery IMT among patients with RA exceeded that of the controls by an average of 0.055 mm, a difference that, with our sample's size and distribution of carotid artery IMT values, did not reach statistical significance. A difference of this magnitude may be important clinically, however. As a point of reference, it is useful to know that the difference in carotid artery IMT between smokers and nonsmokers in the CHS is 0.050 mm (36). Moreover, in the same study, an increase in carotid artery IMT of 0.050 mm was associated with a 59% increase in the incidence of CV events (25). With the carotid artery IMT distribution we found and our sample size, the probability of finding a statistically significant difference between patients with RA and controls was only 17%. We would need to study 2,000 patients with RA and 1,000 controls for a difference of the magnitude we found to be statistically significant. Therefore, the lack of statistical significance in our RA–controls comparison may represent a Type II statistical error.
Several aspects of our study merit additional discussion. We used a validated vascular imaging protocol to comprehensively assess carotid artery morphology. We obtained measurements from the near and far walls of both the common and internal carotid arteries, on both sides of the neck, imaging each common carotid artery from 2 separate projections, and each internal carotid artery from 6 projections, for a total of 16 projections per subject. The ultrasonographer and the reader worked independently, in geographically removed locations, the reader being blinded to any individual characteristic of the subject being scored. We recruited our sample of patients and controls from a wide variety of locations rather than from a single clinic, which may make our findings more generalizable.
Recent studies that have compared patients with RA with matched controls have found significantly greater carotid artery IMT in the patients with RA (22–24). Our study differs from those studies in several important ways, such as the sample selection criteria and target population. Subjects in our study had a high frequency of CV risk factors, resulting in carotid artery IMT values that were considerably higher and distributed over a broader range. Our study may be more generalizable to patients with RA seen by rheumatologists in North America.
Our findings have potential clinical implications. The significant association between atherosclerosis and systemic inflammation that we have documented should be added to the growing evidence linking the 2 processes. Indeed, it has been suggested that accelerated atherosclerosis may be an extraarticular feature of RA (14). Clinicians providing care for patients with RA and other diseases characterized by chronic systemic inflammation should be aware of the propensity to atherosclerosis that these patients share. Although specific screening guidelines cannot be derived from the current study, clinicians should maintain a high index of suspicion for CV pathology when caring for patients with RA and other inflammatory disorders.
The authors wish to thank the following individuals: Terumi Morgan, for outstanding work performing the carotid ultrasound scans; Kate O'Leary, for technical skill in reading the carotid images; Florencia Salazar, Samvel Pogosian, MD, and the rest of the ÓRALE staff for excellent performance in conducting this study; the Frederic C. Bartter General Clinical Research Center (San Antonio, TX); Laurie Funk, from the Ultrasound Reading Center (Boston, MA) for outstanding clinical and technical support; and Drs. Ramón Arroyo, Dan Battafarano, Rita Cuevas, Michael Fischbach, John Huff, Alex de Jesus, Rodolfo Molina, Mathew Mosbacker, Fred Murphy, Carlos Orces, Christopher Parker, Thomas Rennie, Jon Russell, Joel Rutstein, and James Wild for giving us permission to study their patients.
- 12C-reactive protein, a sensitive marker of inflammation, predicts future risk of coronary heart disease in initially healthy middle-aged men: results from the MONICA (Monitoring Trends and Determinants in Cardiovascular Disease) Augsburg Cohort Study, 1984 to 1992. Circulation 1999; 99: 237–42., , , , , , et al.
- 18NHLBI Communications Office. NHLBI launches 10-year study on early detection of heart disease. Accessed September 14, 2000. URL: www.nhlbi.nih.gov/new/press/sep14-00.htm.
- 20Sg17: regression standard errors in clustered samples. Stata Technical Bulletin 1993; STB-13: 19–23..
- 21Causation and causal inference. In: RothmanKJ, GreenlandS, editors. Modern epidemiology. Philadelphia: Lippincott-Raven; 1998. p. 7–28., .
- 27Erythrocyte sedimentation rate and the acute phase response. In: KelleyWN, HarrisEDJr, RuddyS, SledgeCB, editors. Textbook of rheumatology. Philadelphia: WB Saunders; 1989. p. 719–27..
- 29Validity of acute phase proteins as markers of disease activity. J Rheumatol Suppl 1994; 42: 9–13.
- 37Enhancing causal inference in observational studies. In: HulleySB, CummingsSR, BrownerWS, GradyD, HearstN, NewmanTB, editors. Designing clinical research. Philadelphia: Lippincott Williams & Wilkins; 2001. p. 125–38., , .