The mechanism underlying the high frequency of cardiovascular (CV) morbidity that occurs in rheumatoid arthritis (RA) is not completely understood (1–16). Earlier studies have indicated that the excess CV event rate in RA is not explained by an excess of established CV risk factors (17, 18). Those studies did not address whether these risk factors operate similarly in RA as they do in the general population.
Diabetes mellitus, hypercholesterolemia, hypertension, cigarette smoking, and obesity are powerful classifiers of CV risk (19). One or more established CV risk factors are present in at least 80% of people with symptomatic coronary artery disease in the general population (20). The prevalence is even higher, approaching 100%, among persons with fatal myocardial infarction (21).
People with and those without RA have similar CV risk factor profiles (17, 18, 22). The resemblance suggests that established CV risk factors may play an important role in the CV morbidity that occurs in RA, as they do in the general population. In earlier studies that compared the extent of atherosclerosis or CV event rates between RA patients and controls, there was adjustment for the potential confounding effect of CV risk factors (23–27). However, those investigations did not include attempts to estimate the relative contribution of CV risk factors and RA clinical manifestations to CV outcomes, nor did they examine how the two interact. Such information is needed in order to select the interventions most likely to succeed in retarding atherosclerosis in RA. In the present study, we estimated the relative contribution to atherosclerosis of the established CV risk factors and of the clinical manifestations of RA.
- Top of page
- PATIENTS AND METHODS
The ÓRALE sample included 779 RA patients recruited into the parent study of the disablement process in RA. We began the arterial assessments in February 2000. Sixty-six patients died and 32 moved away from the San Antonio area before an appointment could be scheduled. This left 681 patients still eligible to participate in the arterial assessments. Of these, we could not establish contact with 17, and 19 declined to participate. We could not perform the carotid ultrasound on 13 because they were evaluated at their residence. High-resolution carotid ultrasound was performed on 632 patients (93% of eligible patients). In 1 of these patients, the carotid image was not of sufficient quality to obtain an IMT measurement, and this patient was thus omitted from the analysis. The ÓRALE study visit at which the carotid ultrasound was performed was the fifth visit in 3 patients, the fourth in 44, the third in 214, the second in 366, and the first in 4. Joint counts included in these analyses were averaged per patient over a mean of 2.5 measurements (range 1–5), ESR was averaged over a mean of 2.4 measurements (range 1–5), CRP over a mean of 1.4 measurements (range 1–3), and blood pressure over a mean of 2.5 measurements (range 1–6). The median amount of time since enrollment in the ÓRALE longitudinal study of the disablement process in RA was 3.25 years (range 0–7.08 years).
The patients' characteristics at the time of the carotid ultrasound are shown in Table 1. Patients were grouped according to the presence or absence of carotid plaque. In an unadjusted comparison, the patients with plaque were significantly older and the frequencies of male sex and white race were higher. Among the manifestations of RA, longer disease duration, higher deformed joint count, and increased ESR or CRP were associated with an increased likelihood of carotid plaque. The cumulative glucocorticoid dose was significantly higher among patients with plaque as well. Among the established CV risk factors, diabetes mellitus, hypercholesterolemia, smoking, and hypertension were all associated with carotid plaque. Of note, BMI was inversely associated with plaque in this bivariate analysis (Table 1).
Table 1. Characteristics of 631 RA patients grouped by presence or absence of carotid plaque*
|Characteristic||Carotid plaque present (n = 328)||Carotid plaque absent (n = 303)||P|
|Demographic characteristics|| || || |
| Current age, years||64.5 ± 9.9||52.1 ± 12.0||≤0.001|
| Age at RA diagnosis, years||48 ± 13||40 ± 13||≤0.001|
| Women, no. (%)||206 (63)||250 (82)||≤0.001|
| White, no. (%)||137 (42)||75 (25)||≤0.001|
| Black, no. (%)||24 (7)||21 (7)||0.8|
| Hispanic, no. (%)||160 (49)||197 (65)||≤0.001|
|RA manifestations|| || || |
| Disease duration, years||16.0 ± 11.4||12.0 ± 8.8||≤0.001|
| Tender joint count||14.5 ± 13.4||13.7 ± 12.3||0.5|
| Swollen joint count||4.1 ± 5.0||4.3 ± 4.9||0.6|
| Deformed joint count||17.4 ± 12.8||13.9 ± 11.4||≤0.001|
| Subcutaneous nodules, no. (%)||158 (48)||135 (45)||0.4|
| ESR, mm/hour||43.4 ± 27.6||38.9 ± 24.3||0.03|
| CRP, mg/liter||17.7 ± 27.0||14.0 ± 17.7||0.05|
| RF positive, no. (%)||273 (83)||246 (81)||0.4|
| HLA–DRB1 SE positive, no. (%)||239 (73)||216 (71)||0.6|
| Cumulative glucocorticoid dose, gm||11.5 ± 15.6||8.7 ± 15||0.02|
|CV risk factors|| || || |
| Diabetes mellitus, no. (%)||79 (24)||42 (14)||0.001|
| Hypercholesterolemia, no. (%)||213 (65)||137 (45)||≤0.001|
| Smoking, no. (%)|| || ||≤0.001|
| Never||102 (31)||156 (51)|| |
| Past smoker||159 (48)||97 (32)|| |
| Current smoker||67 (20)||50 (16)|| |
| Hypertension, no. (%)||249 (76)||147 (49)||≤0.001|
| Systolic blood pressure, mm Hg||144.0 ± 21.6||134.2 ± 19.9||≤0.001|
| Diastolic blood pressure, mm Hg||76.0 ± 12.2||74.9 ± 10.9||0.2|
| BMI, kg/m2||28.4 ± 5.4||29.6 ± 7.1||0.02|
We tested two groups of multivariable models, one for the IMT and the other for carotid plaque. Because of colinearity between them, we included the ESR and CRP in separate models. Results of the IMT analysis, obtained using OLS regression, are shown in Table 2. In this model, age and sex were significantly associated with the dependent variable. All of the established CV risk factors were associated with the IMT, with the exception of BMI. Among the RA manifestations, disease duration and ESR were also significantly associated with the IMT in the multivariable model (Table 2).
Table 2. Multivariable model of factors associated with carotid intima-media thickness in 631 RA patients*
|Variable||Linear regression coefficient||95% confidence interval||Beta coefficient||P|
|Demographic characteristics|| || || || |
| Age (per 10 years)||0.172||0.141, 0.203||0.431||≤0.001|
| Age at RA onset (per 10 years)||0.168||0.136, 0.199||0.450||≤0.001|
| Female sex (1 = yes; 0 = no)||−0.235||−0.311, −0.158||−0.209||≤0.001|
| Hispanic versus white||−0.040||−0.116, 0.034||−0.040||0.3|
| Black versus white||−0.085||−0.216, 0.045||−0.043||0.2|
|RA manifestations|| || || || |
| RA disease duration (per 10 years)||0.204||0.160, 0.248||0.425||≤0.001|
| Tender joint count||0.000||−0.000, 0.003||0.008||0.8|
| Swollen joint count||−0.002||−0.009, 0.005||−0.023||0.5|
| Deformed joint count||0.002||−0.001, 0.004||0.039||0.3|
| Nodules||−0.049||−0.117, 0.019||−0.050||0.2|
| ESR (per 10 mm/hour)||0.015||0.001, 0.029||0.078||0.03|
| RF positive||0.031||−0.053, 0.117||0.024||0.4|
| HLA–DRB1 SE positive||−0.036||−0.107, 0.033||−0.032||0.3|
| Cumulative glucocorticoid dose (per quartile)||0.010||−0.017, 0.037||0.016||0.4|
|CV risk factors|| || || || |
| Diabetes||0.153||0.068, 0.236||0.119||≤0.001|
| Hypercholesterolemia||0.067||0.002, 0.132||0.067||0.04|
| Past smoker (versus never smoker)||0.094||0.021, 0.168||0.093||0.01|
| Current smoker (versus never smoker)||0.185||0.092, 0.277||0.143||≤0.001|
| Hypertension||0.086||0.012, 0.161||0.083||0.02|
| BMI (per 5 kg/m2)||−0.007||−0.34, 0.019||−0.019||0.5|
Table 3 displays the results of the second group of multivariable models, using logistic regression to analyze the presence of carotid plaque. Factors significantly associated with plaque in this model closely mirrored those found in the OLS model for IMT. However, the cumulative glucocorticoid dose was also associated with plaque (Table 3).
Table 3. Multivariable models of factors associated with carotid plaque in 631 RA patients*
|Variable||Odds ratio||95% confidence interval||Standardized coefficient||P|
|Demographic characteristics|| || || || |
| Age (per 10 years)||2.43||1.94, 3.05||0.644||≤0.001|
| Age at RA onset (per 10 years)||2.39||1.90, 3.00||0.665||≤0.001|
| Female sex (1 = yes; 0 = no)||0.55||0.34, 0.89||−0.150||0.01|
| Hispanic (versus white)||0.78||0.48, 1.24||−0.069||0.3|
| Black (versus white)||0.97||0.43, 2.21||−0.003||0.9|
|RA manifestations|| || || || |
| RA disease duration (per 10 years)||2.90||2.10, 3.98||0.613||≤0.001|
| Tender joint count||1.01||0.99, 1.03||0.095||0.1|
| Swollen joint count||0.93||0.92, 1.02||−0.070||0.3|
| Deformed joint count||1.00||0.97, 1.01||−0.014||0.8|
| Nodules||0.76||0.49, 1.17||−0.074||0.2|
| ESR (per 10 mm/hour)||1.12||1.02, 1.22||0.164||0.009|
| RF positive||1.22||0.71, 2.06||0.041||0.4|
| HLA–DRB1 SE positive||0.84||0.53, 1.32||−0.043||0.4|
| Cumulative glucocorticoid dose (per quartile)||1.27||1.06, 1.51||0.154||0.007|
|CV risk factors|| || || || |
| Diabetes mellitus||1.70||1.00, 2.89||0.109||0.04|
| Hypercholesterolemia||1.65||1.10, 2.47||0.136||0.01|
| Past smoker (versus never smoker)||1.75||1.10, 2.78||0.152||0.01|
| Current smoker (versus never smoker)||2.83||1.57, 5.09||0.228||0.001|
| Hypertension||1.98||1.24, 3.13||0.182||0.004|
| BMI (per 5 kg/m2)||0.89||0.75, 1.06||−0.078||0.2|
To estimate the relative contribution of demographic characteristics, CV risk factors, and RA clinical manifestations to carotid atherosclerosis, we reapplied the above models according to a predefined, hierarchical sequence. The results are shown in Table 4. The initial models included only demographic factors (age, sex, and ethnic group). The R2 for this group of variables was 0.36 for IMT and 0.21 for plaque. Next, we added RA clinical manifestations. This increased the R2 by 0.02 for IMT and 0.03 for plaque; both of these increases were statistically significant. We also tested a model that included demographic characteristics and CV risk factors without RA manifestations. The incremental R2 associated with CV risk factors was 0.05 for IMT and 0.04 for plaque, both statistically significant (Table 4). The final model included all 3 groups of variables, (demographic features, CV risk factors, and RA manifestations). The final R2 measures from this “full” model were 0.42 for IMT and 0.27 for plaque. After the addition of RA manifestations to a model that included only demographic features and CV risk factors, the incremental R2 associated with manifestations of RA was 0.01 for IMT and 0.02 for plaque (Table 4).
Table 4. Hierarchical regression models of carotid intima-media thickness and plaque in 631 RA patients*
|All ages|| || || || |
| Demographics + RA manifestations||0.38||≤0.001||0.24||≤0.001|
| Demographics + CV risk factors||0.41||≤0.001||0.25||≤0.001|
| Demographics + CV risk factors + RA manifestations||0.42||≤0.001||0.27||≤0.001|
|Age 21–54 years (lower tertile)|| || || || |
| Demographics + RA manifestations||0.205||≤0.001||0.177||≤0.001|
| Demographics + CV risk factors||0.181||0.003||0.098||0.01|
| Demographics + CV risk factors + RA manifestations||0.236||≤0.001||0.189||≤0.001|
|Age 55–65 years (middle tertile)|| || || || |
| Demographics + RA manifestations||0.158||≤0.001||0.049||0.008|
| Demographics + CV risk factors||0.231||≤0.001||0.097||≤0.001|
| Demographics + CV risk factors + RA manifestations||0.265||≤0.001||0.108||≤0.001|
|Age 66–90 years (upper tertile)|| || || || |
| Demographics + RA manifestations||0.168||0.005||0.096||0.03|
| Demographics + CV risk factors||0.234||≤0.001||0.084||0.03|
| Demographics + CV risk factors + RA manifestations||0.240||≤0.001||0.131||0.01|
Age in this patient sample ranged from 20 to 90 years, an interval that includes most of the adult lifespan, and during which most atherosclerosis accrues. This was reflected in the large proportion of variance in atherosclerosis explained by the demographic variables. We thus stratified the sample according to tertiles of age and repeated the above series of hierarchical models within the individual age strata. The findings are shown in Table 4. With age restriction, the variance in IMT explained by demographic characteristics was between 0.109 and 0.159. Adding CV risk factors without RA manifestations raised the R2 by 0.036–0.122. Adding RA manifestations in the absence of CV risk factors raised the model R2 by 0.009–0.06; in the presence of CV risk factors, the contribution of RA decreased slightly, to 0.006–0.055. Thus, in these age-stratified models, the contribution of demographic characteristics declined to less than half that in the unstratified sample, while the contribution of CV risk factors and RA manifestations increased. The logistic regression models for plaque revealed a pattern similar to that observed in the IMT models, with RA manifestations explaining up to 0.114 of the variance in plaque in the youngest age group (Table 4).
Because of the heterogeneity between the sexes in CV risk factor effects observed in the population, we were also interested in the extent to which CV risk factor effects varied between men and women in our RA study group. These findings are shown in Table 5. Age, cigarette smoking, and diabetes mellitus displayed a significantly stronger association with IMT in men than in women. The effect of RA disease duration, ESR, hypercholesterolemia, and hypertension did not vary significantly between men and women. The models for plaque did not reveal any significant heterogeneity by sex in the CV risk factor effect. Explained variance of the multivariable models was similar in men and women (Table 5).
Table 5. Association between markers of atherosclerosis and individual risk factors in 631 RA patients grouped by sex*
|Dependent variable, risk factor||Men||Women||P for interaction‡|
|Regression coefficient or odds ratio†||95% confidence interval||Regression coefficient or odds ratio†||95% confidence interval|
|Intima-media thickness (model R2 0.353 for men, 0.344 for women)|| || || || || |
| Age, per 10 years||0.282||0.233, 0.332||0.144||0.114, 0.175||≤0.001|
| Age at RA onset, per 10 years||0.260||0.210, 0.309||0.144||0.113, 0.175||≤0.001|
| Duration of RA, per 10 years||0.209||0.151, 0.268||0.205||0.165, 0.246||0.9|
| ESR, per 10 mm/hour||0.011||−0.010, 0.033||0.016||0.001, 0.030||0.7|
| Cumulative glucocorticoid dose, per quartile||0.005||−0.004, 0.005||0.11||−0.021, 0.042||0.8|
| Former smoker versus never smoker||0.318||0.150, 0.485||0.043||−0.038, 0.123||0.004|
| Current smoker versus never smoker||0.351||0.163, 0.539||0.142||0.038, 0.245||0.055|
| Diabetes mellitus versus nondiabetic||0.264||0.109, 0.419||0.083||−0.009, 0.176||0.049|
| Hypercholesterolemia present versus absent||0.136||0.020, 0.251||0.053||−0.023, 0.127||0.2|
| Hypertension||0.110||−0.063, 0.283||0.098||0.027, 0.170||0.6|
|Plaque (model pseudo-R2 0.245 for men, 0.254 for women)|| || || || || |
| Age, per 10 years||2.64||1.81, 3.84||2.57||2.03, 3.26||0.9|
| Age at RA onset, per 10 years||2.71||1.88, 3.89||2.42||1.93, 3.03||0.6|
| Duration of RA, per 10 years||2.62||1.71, 4.02||3.10||2.32, 4.14||0.5|
| ESR, per 10 mm/hour||1.11||0.96, 1.28||1.11||1.02, 1.21||0.9|
| Cumulative glucocorticoid dose, per quartile||1.21||0.89, 1.66||1.29||1.06, 1.57||0.7|
| Former smoker versus never smoker||4.46||1.56, 12.83||1.42||0.87, 2.32||0.055|
| Current smoker versus never smoker||3.91||1.23, 12.40||2.64||1.46, 5.22||0.6|
| Diabetes mellitus versus nondiabetic||0.70||0.26, 1.87||1.87||1.06, 3.29||0.09|
| Hypercholesterolemia present versus absent||1.46||0.69, 3.08||1.88||1.19, 2.96||0.6|
| Hypertension||2.18||0.94, 5.07||1.70||1.07, 2.71||0.9|
Hispanic patients were significantly less likely to have carotid plaque in the unadjusted comparison (Table 1). However, this difference was no longer significant after age adjustment (Table 3). We sought evidence of heterogeneity by race/ethnicity in the effect of CV risk factors. We found that among whites, smoking was associated with significantly greater IMT than it was among nonwhites (regression coefficient 0.241 [95% confidence interval 0.125, 0.358] versus 0.060 [95% confidence interval −0.018, 0.137]), and diabetes mellitus was associated with a greater difference in IMT in whites than in nonwhites (regression coefficient 0.382 [95% confidence interval 0.210, 0.554] versus 0.072 [95% confidence interval −0.018, 0.162]). In the case of carotid plaque, we found similar variation between whites and nonwhites in the effect of cigarette smoking, but not in the effect of diabetes mellitus. The effect of other CV risk factors on IMT or plaque did not vary significantly between whites and nonwhites. The R2 of the IMT model in whites was 0.368; in nonwhites, it was 0.415. In the plaque models, the pseudo-R2 was 0.297 in whites and 0.250 in nonwhites.
We estimated the effect of the presence of multiple CV risk factors in combination with the 2 RA manifestations that were found to be significantly associated with carotid ultrasound findings in the multivariable models of both plaque and IMT, i.e., ESR and disease duration. First, we generated a variable for the number of CV risk factors associated with the carotid outcomes that were present in each patient. These included diabetes mellitus, hypercholesterolemia, hypertension, and current or past smoking (Tables 2 and 3). We chose these variables because they were significantly associated with both carotid outcomes. We then generated product terms for the number of CV risk factors present × ESR or disease duration. We found that the product term for the number of CV risk factors × ESR was significantly associated with the carotid IMT (P = 0.03). This suggested that the effect of ESR on IMT varied according to the number of CV risk factors. We then tested the ESR–IMT association within strata defined by the number of CV risk factors. This revealed that the association of ESR with IMT was significant only when CV risk factors were present, and not in the absence of CV risk factors. Figure 1 shows the age- and sex-adjusted mean IMT and plaque probabilities, with stratification by ESR quartile and number of CV risk factors. Variance inflation factors were <10 for all variables tested in the multivariable models, indicating that there was no significant multicolinearity between the variables we tested.
Figure 1. Age- and sex-adjusted mean carotid intima-media thickness (IMT) and probability of carotid plaque, as functions of the number of cardiovascular (CV) risk factors and erythrocyte sedimentation rate (ESR) quartiles (mm/hour). The number of CV risk factors and the ESR were each significantly associated with both dependent variables (P for trend ≤ 0.001 for each). There was a significant interaction between the ESR and the number of CV risk factors (P for the product term ESR × number of CV risk factors ≤ 0.001). The interaction suggested that the ESR's effect on IMT varied according to the number of CV risk factors. In the case of carotid plaque, the ESR's effect was significant, without evidence of interaction.
Download figure to PowerPoint
- Top of page
- PATIENTS AND METHODS
CV morbidity and mortality in RA occur at rates greater than would be expected from the profile of established CV risk factors (17, 18, 39). This has stimulated interest in identifying the additional risk factors that may explain the excess seen in RA. A number of potential markers have indeed been linked to atherosclerosis and CV events in RA, including molecules involved in the immune response, markers of inflammation, and therapeutic agents (23–27, 40–42). However, it should be noted that the great majority of CV events in the general population are attributable to established CV risk factors (20, 21). A similar connection between the established risk factors and CV disease may prevail in RA, with novel risk factors accounting only for the proportion that exceeds the expected risk. An important first step in evaluating this possibility is to establish the extent to which the atherosclerotic burden in RA is explained by the established CV risk factors.
We used a hierarchical modeling approach to address this question. We focused on the coefficient of determination, or R2, as a measure of the proportion of variance in atherosclerosis explained by the independent variables or predictors. We tested variables grouped a priori as demographic variables (age, sex, and ethnic group), CV risk factors (diabetes mellitus, hypercholesterolemia, smoking, hypertension, and obesity), and RA clinical manifestations (RA duration, joint examination findings, subcutaneous nodules, RF, ESR, CRP, HLA–DRB1 SE, and cumulative glucocorticoid dose). This technique permitted us to model these variables grouped according to our predefined criteria, rather than through an automatic stepwise approach. Thus, we were able to estimate the aggregate contribution of the established CV risk factors separately from that of the RA manifestations. The disadvantage of this approach is that the contribution of individual variables can be considered only in combination with other variables in the group being tested.
We found that most of the variability in carotid atherosclerosis was explained by demographic characteristics. This should not be surprising. Age is the major contributor to the extent of atherosclerosis in the population (43, 44). Our sample included patients whose age varied from 20 years, when little or no atherosclerosis would be expected, to 90 years, when maximal accrual of atherosclerosis would be expected. In the general population, the difference in IMT between young and old persons is greater than the difference between old persons with and those without coronary artery disease (45, 46).
Results of the age-stratified models shown in Table 4 suggested that a substantial portion of the variance in atherosclerosis is explained by risk factors that may be potentially modifiable. The RA manifestations were most strongly associated with atherosclerosis in the youngest age group, where they explained 6% of IMT variance and 11.4% of plaque variance over that explained by demographic characteristics. In this age group, CV risk factors explained only 3.6% and 3.5% of the variance in IMT and plaque, respectively, over demographic factors. After accounting for demographic and CV risk factors, RA manifestations still explained 5.5% of the IMT variance and 9.1% of that of plaque. The proportion of atherosclerosis explained by RA manifestations decreased in the older age groups, suggesting that the systemic inflammation of RA exerts its effect early. This notion is supported by the observation of an increased CV event rate in RA patients prior to disease diagnosis (39). In contrast, the proportion of atherosclerosis variance explained by the established CV risk factors was greater in the older age groups, surpassing the effect of the RA manifestations. Thus, the established CV risk factors seemed to lag behind systemic inflammation in the time course of their effect, but ultimately had effects that were quantitatively stronger.
It is noteworthy that the number of CV risk factors present in each patient had an additive effect on carotid atherosclerosis, displaying a biologic gradient, or “dose-response” effect: the greater the number of CV risk factors, the greater the extent of atherosclerosis (Figure 1). We also noted a potentially important effect modification: higher ESR values were associated with greater IMT only in the presence of CV risk factors. In the patients who did not have any CV risk factors, the ESR was not significantly associated with IMT. This suggests that inflammation provides positive modulation of the established risk factors, but is likely not sufficient to cause the disease independently. It was not possible to discern the biologic substrate of this interaction in the present study. It is possible that fibrinogen, the concentration of which primarily determines the ESR, may accelerate the atherogenic effects of the other CV risk factors (47). In vitro studies targeting these interactions would be of great interest.
Of interest, body mass was not associated with the IMT or plaque in this RA cohort. In the general population, obesity is a major predictor of atherosclerosis and CV events (48). Its lack of association with the carotid IMT or plaque in the present study is consistent with the finding of a paradoxical effect of body mass on CV and all-cause mortality in RA (49, 50).
Some caution in interpreting our findings is warranted. The outcome measures we used, obtained with high-resolution carotid ultrasound, are markers of subclinical atherosclerosis, not of clinically verified disease (51). However, their clinical relevance is underscored by their ability to identify individuals at high risk of atherosclerotic complications such as myocardial infarction or stroke (43, 52). Moreover, carotid and coronary atherosclerosis are highly correlated, and thus, findings in the carotid arteries likely reflect findings in the coronary arteries (53). Ultrasound and histologic images of the IMT correlate highly as well (54). When performed by a single sonographer using an established protocol, and read by a single expert reader, as in this study, the technique is reliable. Thus, our findings likely reflect clinically significant CV risk.
Atherosclerosis is a process that takes place over time, as is the effect of the risk factors we tested. The cross-sectional nature of this study may have limited our ability to fully capture the temporal relationships in the atherosclerosis process. To counteract this potential limitation, we used time-averaged values for the ESR, in an effort to capture the effect of this variable over time. With certain exposure variables, it was also possible to generate indicators that reflected a change in status over time. Thus, we considered current cigarette smokers separately from those who had smoked in the past but quit. We found that indeed, current smokers had a greater IMT than did past smokers, consistent with the notion that those who stop smoking ultimately develop less atherosclerosis. Both smoker groups, however, had worse findings on carotid ultrasound than did nonsmokers.
We also disaggregated chronological age into two variables, age at RA onset and disease duration. This avoids counting the disease duration twice, as would occur if the full chronological age is included together with disease duration in the same model. It is of interest that both the regression coefficient and the odds ratio for disease duration were greater than those for age at disease onset. This could suggest that the slope for increase of atherosclerosis over time was steeper during the years that patients had RA. Longitudinal data are needed to thoroughly explore this possibility.
Our findings suggest that, after accounting for the effects of age and sex, both established CV risk factors and RA manifestations account for a significant proportion of atherosclerosis in RA. Factors related to RA may have a greater influence on the extent of atherosclerosis in young patients. The presence of established CV risk factors may be necessary for systemic inflammation to promote atherosclerosis. Further research is needed to understand the mechanisms whereby established CV risk factors and inflammation markers interact in the atherogenic process.