Baseline levels of c-reactive protein and prediction of death from cardiovascular disease in patients with inflammatory polyarthritis : A ten-year followup study of a primary care–based inception cohort




To test the hypothesis that the C-reactive protein (CRP) concentration at baseline is an independent predictor of death from cardiovascular disease (CVD) in newly diagnosed patients with inflammatory polyarthritis (IP).


Patients with IP (n = 506) who were recruited from the Norfolk Arthritis Register between 1990 and 1992 were followed up to the end of 2001, and complete data on mortality were obtained. At baseline, subjects underwent a structured interview and joint examination and completed a Health Assessment Questionnaire (HAQ). Blood was obtained and analyzed for rheumatoid factor (RF) and CRP concentration. Cox regression was used to calculate hazards ratios (HRs) for risk of death from CVD.


The median followup was 10.1 years (interquartile range 9.3–10.8). There were 104 deaths, 40 of which were the result of CVD. Elevated CRP levels (≥5 mg/liter) predicted death from CVD in univariate analyses: HR 3.9 (95% confidence interval [95% CI] 1.2–13.4) for men, and HR 4.22 (95% CI 1.4–12.6) for women. After adjusting for age and sex, the CVD mortality association was strongest in the subgroup of patients who were RF positive at baseline (adjusted HR 7.4 [95% CI 1.7–32.2]). Multivariate analysis revealed that elevated CRP levels remained a significant independent predictor of death from CVD, even after adjusting for age, sex, smoking status, HAQ score, RF positivity, and swollen joint counts (HR 3.3 [95% CI 1.4–7.6]).


The CRP concentration at baseline is an important predictor of subsequent death from CVD in patients with new-onset IP and is independent of other indicators of disease severity. This supports the theory that CRP may play a direct role in the pathogenesis of CVD.

There is now a substantial body of evidence linking rheumatoid arthritis (RA) with cardiovascular disease (CVD) (1–3) and, in particular, with increased death from cardiovascular causes (4–6). We have previously observed that within 8 years of the onset of arthritis symptoms, patients with rheumatoid factor (RF)–positive inflammatory polyarthritis (IP) have increased CVD mortality rates (7). In RF-positive women with IP, all the excess deaths observed could be attributed to cardiovascular causes.

Although there are a number of possible explanations for this link, one key question concerns the role of inflammatory activity. Atherosclerosis is now accepted to be a chronic inflammatory process (8). The concentration of C-reactive protein (CRP), both in the general population and in patients with known CVD, shows a strong association with subsequent events attributable to CVD (9–12). CRP is an acute-phase protein that is produced in the liver in response to an elevation in interleukin-6 levels. There is evidence that CRP has direct effects at the vessel wall that may promote atherosclerosis. These direct effects include stimulating the production of cellular adhesion molecules by vascular endothelial cells (13), facilitating the adhesion and migration of monocytes through the vessel wall, mediating the uptake of low-density lipoprotein cholesterol by macrophages (14), and causing complement activation (15).

Substantial evidence linking CRP concentrations and subsequent CVD risk in the general population now exists (16). Most of the studies that show such a link (17, 18) have utilized high-sensitivity CRP assays, which estimate the value of CRP within what is considered to be the normal range. This might not be relevant to the situation in RA, since most patients with this disease have an elevated CRP level. Thus, it is unclear whether the CRP level will be a useful predictor of subsequent CVD events in patients with chronic inflammatory joint conditions such as IP and RA.

We therefore examined whether the CRP concentration measured early in the IP disease process predicts future death from CVD. We also investigated whether this effect is independent of other markers of disease activity.


Study design

This was a prospective followup study of an inception cohort of individuals with new-onset IP who had CRP measured at baseline, together with other aspects of disease status. The Local Research Ethics Committee approved the study. Subjects were followed up for an average of 10 years to identify the role of CRP and other baseline factors in predicting subsequent death from cardiovascular causes.

Study subjects

Subjects were recruited from the Norfolk Arthritis Register (NOAR), a primary care–based incidence register of IP, the details of which have been described elsewhere (19). Briefly, all subjects residing within the NOAR catchment area who attend their primary care physician and have synovitis affecting 2 or more peripheral joints and lasting for 4 or more weeks are eligible for inclusion. Subjects with definite inflammatory synovitis that is not subsequently explained by another medical diagnosis (other than RA, psoriatic arthritis, postviral arthritis, or undifferentiated IP) are eligible for followup. In selecting subjects for inclusion, we did not attempt at baseline to distinguish between these latter 4 groups.

Baseline data collection

At the baseline interview, which was conducted by a research nurse, disease-specific data were recorded, including the date of symptom onset and the duration of morning stiffness. In addition, subjects were interviewed about their smoking history and were classified as never smoked, ex-smokers, or current smokers. Classification of socioeconomic status was based on the subject's current or most recent occupation or that of his or her spouse if the subject had no relevant occupational history. For analysis, subjects were categorized as those with manual and those with nonmanual occupations. As a measure of preexisting cardiovascular morbidity, subjects were stratified according to those taking none, 1, or >1 drug with a cardiovascular action, including drugs for hypertension and angina. The classification was based on drugs included in the cardiovascular chapter of the British National Formulary (20). All subjects completed the Health Assessment Questionnaire (HAQ) (21). Subjects were examined, and the swollen joint count, the tender joint count, and the presence of rheumatoid nodules were recorded. Blood was taken for RF analysis by the tube latex test. A positive result was defined as agglutination at a titer of ≥1:40.

Measurement of CRP levels

CRP levels were measured in previously frozen serum samples (stored at –70°C) using an end-point immunoturbidimetric agglutination method (Boehringer Mannheim, Lewes, UK). A Hitachi 917/911 analyzer (Hitachi, Danbury, CT) calculated the CRP concentration (in mg/liter). This traditional assay identifies elevated CRP levels in the range of 5–300 mg/liter. High-sensitivity CRP assays were not performed.

Followup on deaths and causes of death

Subjects recruited to NOAR between 1990 and 1992 were eligible for study. Followup on deaths and causes of death was obtained for all subjects by linkage with the Central Register of the UK National Health Service. This linkage provides the investigator with a copy of the death certificate. The analysis was based on the underlying cause of death, using rules developed by the World Health Organization Office of Population Censuses and Surveys (22), and causes of death were grouped according to the International Classification of Diseases, Ninth Revision (ICD-9) and the Tenth Revision (ICD-10) were used for deaths occurring after December 31, 2000 (22). Patients were followed up to December 31, 2001, and complete data on mortality were obtained for the cohort.

Statistical analysis

Subjects were stratified into 3 groups based on their CRP level. A concentration ≤4 mg/liter was considered normal and was used as the referent group. Subjects with CRP concentrations above this level were stratified into 2 groups of approximately equal sizes: those with a level of 5–15 mg/liter and those with a level ≥16 mg/liter. The relationship between the CRP level and subsequent death was explored using a Cox proportional hazards regression model. Initially, univariate analyses were performed and were stratified by sex and adjusted for age. Since the influence of CRP on mortality was the same in both sexes, data for all subjects were pooled for subsequent analyses and then adjusted for age and sex. Both cardiovascular causes of death and all causes of death were analyzed. Stratified analyses were then performed based on whether or not at baseline the subjects were RF positive or met 4 of 6 of the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) classification criteria for RA (23) (excluding radiographic data, since radiographs were not taken at the first assessment). Finally, a multivariate analysis was performed, adjusting for age, sex, swollen joint count, smoking status, comorbid CVD (based on CVD drug usage), and RF status, to determine the independent effect of an elevated CRP level on death from cardiovascular causes.


A total of 539 consecutive subjects in the NOAR had a CRP measurement, based on their inclusion in previous analyses of factors predicting disease outcome (24). At the baseline assessment, approximately one-half of the cohort satisfied the ACR criteria for RA. Data, including demographic, clinical information, and RF status, were incomplete for 33 participants. The study was therefore restricted to the remaining 506 patients with complete data.

The baseline demographic and clinical disease characteristics of the study cohort are shown in Table 1. Approximately 1 in 5 participants was taking a cardiovascular drug at study entry, including 37 (7%) who were taking 2 or more cardiovascular drugs. Two-thirds of the participants had ever smoked, and one-fourth of the participants were current smokers at the time of their first assessment. CRP concentrations ≤4 mg/liter were observed in 234 (46.3%) of the cohort. Of those patients with “elevated” CRP concentrations, 146 had levels in the range of 5–15 mg/liter and 126 had levels of 16 mg/liter and above.

Table 1. Baseline demographic and disease characteristics of the 506 subjects in the study cohort*
  • *

    IQR = interquartile range; HAQ = Health Assessment Questionnaire; RF = rheumatoid factor; CRP = C-reactive protein; RA = rheumatoid arthritis; ACR = American College of Rheumatology; NSAIDs = nonsteroidal antiinflammatory drugs.

Demographic variables 
 Sex, no. (%) male179 (35.4)
 Age at disease onset, median (IQR) years55 (42–68)
 Socioeconomic status, no. (%) manual occupation238 (47.2)
 Smoking status, no. (%) 
  Never smoked153 (30.2)
  Ex-smoker221 (43.7)
  Current smoker132 (26.1)
 Taking cardiovascular drugs, no. (%)93 (18.4)
Disease variables 
 Delay to presentation, median (IQR) months5.5 (2.9–12.0)
 Swollen joint count, median (IQR)7 (2–14)
 Tender joint count, median (IQR)9 (3–20)
 HAQ score, median (IQR)0.75 (0.25–1.50)
 RF positive, no. (%)173 (34.2)
 CRP concentration, median (IQR) mg/liter6 (2–15)
 Nodules, no. (%)50 (9.8)
 RA (by ACR criteria), no. (%)253 (50.0)
 Taking NSAIDs, no. (%)356 (70.4)

Patients were followed up to December 31, 2001, with a median followup period of 10.1 years (interquartile range [IQR] 9.3–10.8). Over this time period, there were 104 deaths (Table 2). The all-cause mortality rate was 21.9 (95% confidence interval [95% CI] 18.1–26.5) per 1,000 person-years of followup. The mortality rate was higher in men than in women. Cardiovascular deaths occurred in 40 patients, with death from ischemic heart disease being reported most frequently (Table 3). The CVD mortality rate was 8.4 (95% CI 6.2–11.5) per 1,000 person-years of followup.

Table 2. Causes of death in the study cohort*
Cause of deathMen (n = 179)Women (n = 327)All subjects (n = 506)
  • *

    Values are the number (%) of subjects.

All causes54 (30.2)50 (15.3)104 (20.6)
Cardiovascular causes20 (11.2)20 (6.1)40 (7.9)
Respiratory causes9 (5.0)5 (1.5)14 (2.8)
Neoplastic causes17 (9.5)14 (4.2)31 (6.1)
Table 3. Cardiovascular causes of death*
Underlying cardiovascular cause of deathICD-9 codesICD-10 codesNo. of subjects
WomenMenAll IP
  • *

    ICD = International Classification of Diseases (Ninth and Tenth Revisions); IP = inflammatory polyarthritis.

Hypertensive diseases401–405I10–I15101
Ischemic heart diseases     
 Acute myocardial infarction410I21, I22369
 Acute/subacute ischemic heart disease411I24112
 Chronic ischemic heart disease414I255914
 Total deaths from ischemic heart disease410–414I20–I2591625
Other vascular diseases     
 Pulmonary circulation415–417I26–I28213
 Other forms of heart disease420–429I30–I52101
 Cerebrovascular disease430–438I60–I69527
 Disease of arteries, arterioles, and capillaries440–448I70–I79213
Total deaths from cardiovascular diseases390–459I00–I99202040

In the univariate analysis, an elevated CRP (≥5 mg/liter) was associated with death from all causes, with a hazards ratio [HR] of 2.7 (95% CI 0.9–3.0) in men and 2.7 (95% CI 1.5–5.1) in women, and was strongly associated with death from CVD, with an HR of 3.9 (95% CI 1.2–13.4) in men and 4.22 (95% CI 1.4–12.6) in women. Figure 1 shows survival curves for death from CVD according to the CRP concentration. There was a trend toward an increasing risk of death with an increasing stratum of CRP for all-cause mortality and cardiovascular mortality, both in men and in women (Table 4). The associations were stronger for cardiovascular causes and were similar in both men and women.

Figure 1.

Kaplan-Meier survival estimates of death from cardiovascular disease, by baseline concentration of C-reactive protein (CRP).

Table 4. CRP concentration as a predictor of death, stratified by cause of death*
CRP concentrationAll causesCardiovascular causes
No. of subjectsMortality rateHR (95% CI)No. of subjectsMortality rateHR (95% CI)Mortality rateHR (95% CI)Mortality rateHR (95% CI)
  • *

    The mortality rate represents the number of subjects/1,000 person-years of followup. Hazards ratios (HRs) were adjusted for age. CRP = C-reactive protein; 95% CI = 95% confidence interval.

≤4 mg/liter6924.91.0 (referent)1658.51.0 (referent)4.71.0 (referent)2.41.0 (referent)
5–15 mg/liter6232.91.5 (0.7–2.9)8419.91.5 (0.7–3.1)16.53.7 (1.0–13.9)8.72.2 (0.6–7.6)
≥16 mg/liter4852.52.3 (1.0–3.7)7827.52.0 (1.0–4.0)21.04.0 (1.1–15.2)12.43.0 (0.9–9.8)

Data stratified by RF status and RA classification, pooled across both sexes and adjusted for age, are shown in Table 5. The results show a substantial association between a raised CRP level and death from cardiovascular causes in subjects who were RF positive, with a virtual absence of an association in subjects who were RF negative, and a suggestion of a strong association between the CRP level and death from CVD in patients who met the ACR classification criteria for RA at baseline (Table 5). There was no evidence of a dose-response effect, with similar levels of risk in subjects with a CRP level of 5–15 mg/liter and those with a level ≥16 mg/liter. We finally tested whether there was any dose-response effect at levels below 5 mg/liter, comparing subjects with CRP levels <3 mg/liter with subjects with levels in the range of 3–5 mg/liter, and no difference was found.

Table 5. CRP concentration as a predictor of death from cardiovascular disease, stratified by baseline RF and RA status*
CRP concentrationRF positiveRF negativeRA
No. of subjectsHR (95% CI)No. of subjectsHR (95% CI)No. of subjectsHR (95% CI)
  • *

    Hazards ratios (HRs) were adjusted for age and sex. CRP = C-reactive protein; RF = rheumatoid factor; RA = rheumatoid arthritis; 95% CI = 95% confidence interval.

≤4 mg/liter581.0 (referent)1761.0 (referent)961.0 (referent)
5–15 mg/liter588.0 (1.8–36.2)880.9 (0.2–3.8)7913.6 (1.7–107.1)
≥16 mg/liter576.4 (1.3–30.8)692.2 (0.7–7.0)7815.6 (2.0–120.7)
Elevated CRP (≥5 mg/liter)1157.4 (1.7–32.2)1571.5 (0.5–4.5)15714.7 (2.0–109.2)

Multivariate analysis was used to examine whether elevated CRP levels predicted death from CVD, after adjustment for demographic and disease-related potential confounders including age, sex, HAQ score, socioeconomic class, swollen joint counts, smoking status (current and ex-smokers), and RF status. Only RF positivity and CRP levels ≥5 mg/liter were observed to be both independent and strong predictors of death (Table 6). This association between CRP and death from CVD persisted after adjusting for comorbid CVD, as assessed by the use of cardiovascular drugs at baseline (Table 7). All models satisfied the proportional hazards assumption.

Table 6. Multivariate model for the prediction of cardiovascular death*
VariableHazards ratio (95% CI)
  • *

    95% CI = 95% confidence interval; CRP = C-reactive protein; RF = rheumatoid factor; HAQ = Health Assessment Questionnaire.

Age, per decade2.9 (2.1–4.1)
Female sex0.9 (0.5–1.8)
CRP level ≥5 mg/liter3.3 (1.4–7.6)
RF positive3.0 (1.6–5.8)
Swollen joint count, per joint1.0 (0.9–1.0)
Smoking status, ever smoked2.0 (0.8–5.1)
HAQ score, per unit0.9 (0.6–1.5)
Table 7. Alternative multivariate model with cardiovascular drug variables*
VariableHazards ratio (95% CI)
  • *

    95% CI = 95% confidence interval; CRP = C-reactive protein; RF = rheumatoid factor; HAQ = Health Assessment Questionnaire.

Age, per decade2.7 (1.9–4.0)
Female sex0.8 (0.4–1.6)
CRP level ≥5 mg/liter2.6 (1.1–6.3)
RF positive3.0 (1.6–5.9)
Swollen joint count, per joint1.0 (0.9–1.0)
Taking 1 cardiovascular drug1.5 (0.6–3.8)
Taking ≥2 cardiovascular drugs5.7 (2.6–12.5)
Smoking status, ever smoked1.7 (0.7–4.5)
HAQ score, per unit0.8 (0.5–1.2)


This analysis is the first to show that the baseline level of CRP is a predictor of all-cause mortality, and specifically CVD mortality, in both sexes in the 10-year period following the onset of IP. The major effect seems to be restricted to those who were RF positive or who met the ACR classification criteria for RA at baseline. Interestingly, other measures of disease activity at baseline, for example, tender or swollen joint counts or HAQ score, were not independently predictive of subsequent death from cardiovascular causes after allowing for the effect of the CRP level. This suggests that it is specifically CRP (or something closely associated with it) that is responsible for the link between RA and CVD. These data add to the studies showing that modest elevations in the CRP concentration in apparently healthy adults are associated with subsequent cardiovascular events (18, 25). The finding that a CRP concentration above the normal range is a significant predictor of death from CVD in groups of patients with inflammatory joint disease provides further evidence for the role of inflammation in promoting atherosclerotic disease.

This study had a number of strengths. It was a true prospective enquiry of a large inception cohort followed up for a substantial period of time and with complete followup data on mortality. The clinical and laboratory data collection were also standardized. However, there are some methodologic issues that need to be discussed. This study only measured CRP concentrations at the baseline NOAR assessment. It is possible that some patients would have received treatment of their inflammatory disease that may have modified their CRP concentration prior to this assessment. It is also possible that cumulative CRP concentrations during the period of followup would be even more predictive of subsequent death from CVD. However, despite this, it is of considerable interest that a single CRP measurement early in the inflammatory disease process proved so predictive of death from CVD. There is some evidence that the level of inflammation in RA is determined early in the disease course (26), and so, baseline CRP levels may be predictive of cumulative CRP levels. Baseline CRP levels have also been shown to predict subsequent radiologic progression in RA (27).

Another limitation of this study was that comorbid CVD was not recorded at baseline. The presence of comorbid CVD might have confounded the association between the CRP concentration and death from cardiovascular causes. Prescribed CVD drugs at baseline was used as a proxy measurement of baseline comorbid CVD. While we accept that this method is not ideal, the use of CVD drugs has previously been shown to be predictive of physician-confirmed CVD, with high specificity (95%) but low sensitivity (28). Therefore, it is likely that estimating comorbid CVD using this method will have introduced some misclassification and reduced the accuracy of our results. Patients who were receiving CVD drugs at baseline had higher baseline CRP levels (median 9 mg/liter [IQR 3–21]) than patients who were not receiving CVD drugs (median 5 mg/liter [IQR 2–14]) (P = 0.02). However, when we repeated the multivariate analysis excluding the patients who were taking CVD medications, elevated CRP concentrations were still associated with death from CVD (adjusted HR 2.9 [95% CI 1.0–8.2]). While it is possible that this group will still include a small number of patients with CVD comorbidity, we believe that it is more likely that the magnitude of CRP concentrations reflect the underlying inflammatory joint disease rather than undiagnosed and untreated comorbid CVD.

It has been suggested that traditional risk factors for CVD do not explain the increase in CVD events in RA (2). However, several studies have highlighted the association between CRP concentrations and traditional CVD risk factors (29, 30). In particular, smoking and body mass index (BMI) appear to be strongly and independently associated with CRP concentrations (31, 32). An important limitation of this study was that complete CVD risk factor data, and more importantly BMI, were not recorded at baseline. We did have data available from the simultaneous assessment of both BMI and CRP at 5 years in this cohort. At this time point, there was only a very weak association between these variables (rs = 0.08). Adjusting for current smoking at the baseline assessment did not alter the associations seen. Health-related behaviors do change over time, and it is possible that baseline smoking status does not reflect tobacco exposure during the followup period.

Recent population-based studies have used high-sensitivity assays that measure CRP in the range of 0.1–10 mg/liter (33). Such assays were not used in the current study. However, we were able to detect an association with death from CVD with only modest increases in the CRP concentration as measured using traditional assays. Treating CRP as a continuous variable in the multivariate model revealed that for every 1-mg rise in CRP concentration, we observed a 0.8% rise in the risk of death from CVD. However, as discussed above, stratification of CRP values <5 mg/liter did not reveal any trend in risk.

The subjects recruited for this study came from an inception cohort based on attendance in the primary care setting, and results may not be generalizable to patients with RA who are referred to a hospital. Approximately 50% of our subjects satisfied the ACR criteria for RA at baseline. However, we have previously demonstrated that these criteria are not stable in early disease (34) and, indeed, are probably not appropriate to use in that situation (35). Nonetheless, we found that CRP was a strong predictor of death from CVD within that group of patients meeting criteria for RA at baseline, although with wider confidence limits. More convincing was the association between elevated CRP and death from CVD in the RF-positive group, where a modest increase in CRP was associated with a 7-fold increased risk of death from CVD. A raised CRP was not associated with mortality in the seronegative subgroup. CRP concentrations in the seropositive patients (median CRP 9 mg/liter [IQR 3–22]) were higher than those in the seronegative patients (median CRP 4 mg/liter [IQR 1–13]). These data are consistent with our previous observation that compared with the general population, an increased risk of death from CVD was restricted to seropositive IP patients (7). It is possible that the mortality outcome in these seronegative patients is influenced by their lower inflammatory disease burden.

The diagnosis of death due to CVD was based on the underlying cause of death as recorded on the death certificate. The cause of death is not always known, and in some patients, CVD was recorded elsewhere on the death certificate but not as the underlying cause of death. Thus, the current analysis is restricted to those dying from CVD as opposed to those dying with CVD. It would also have been preferable if we had been able to verify the cause of death by inspection of the relevant medical charts, but the entire medical record was not available to us, particularly for subjects that had died some years prior to this analysis. These inevitable errors, which are typical of studies that rely on death certificates, should only have introduced random misclassification rather than systematic bias, since the extent of such errors is likely to be independent of CRP status.

A total of 33 subjects were excluded from this analysis because of incomplete data. There were 5 CVD deaths in this excluded group, and so, they were at a slightly higher risk of this end point than were patients included in the analysis. When the age-adjusted analyses were repeated, retaining the data for the excluded patients, stronger associations were observed between elevated CRP levels and death from CVD: HR 4.2 (95% CI 1.2–14.2) in men and HR 3.2 (95% CI 1.1–9.3) in women.

It is widely accepted that atherosclerosis is an inflammatory condition, and although initially, it was considered that CRP might just act as a marker of this inflammatory activity, it is possible that it plays a pathogenic role, promoting atherosclerosis. CRP levels in excess of 5 mg/liter have been found to be associated with increased vasoreactivity in patients with both stable and unstable angina (36). It may be that systemic inflammation associated with inflammatory arthritis will potentiate any underlying atherosclerosis and increase the risk of death from CVD.

In conclusion, we have found that an elevated CRP concentration measured early in the disease process is a powerful predictor of death from CVD in patients with IP and RA. The CVD outcomes in these high-risk patients may be improved by targeted interventions to reduce traditional risk factors for CVD and may also be improved by more aggressive suppression of their inflammatory joint disease.