Rheumatoid arthritis (RA) is a chronic inflammatory disease with increased incidence of cardiovascular morbidity in both men and women.1^–3 Traditional risk factors such as age, hypertension and dyslipidaemia are associated with an increased cardiovascular risk in RA. Nevertheless, these risk factors alone do not explain the excess risk of cardiovascular disease (CVD) in patients with RA.4 5 It is suggested that systemic inflammation plays a major role in the aetiology of CVD in RA.6 7 Systemic inflammation in RA may accelerate atherogenic processes, either by the accentuation of known pathways of plaque formation or by the onset of additional immune pathways.6 Inflammatory cytokines mediate numerous metabolic effects, such as the generation of insulin resistance, endothelial dysfunction, haemostatic changes and dyslipidaemia.8 9 The lipid profile in RA depends on disease activity. Higher disease activity leads to depressed levels of total cholesterol. However, high-density lipoprotein (HDL) cholesterol levels are even more depressed, resulting in a more unfavourable atherogenic index.10^–14 HDL is not only reduced in patients with RA, but it also seems to dysfunction, losing its ability to protect low-density lipoprotein (LDL) cholesterol from oxidation.15 In turn, anti-inflammatory treatment increases HDL levels.10 13 16

Because systemic inflammation seems to have an important effect on the generation of atherosclerosis and CVD in RA, it may be expected that the extent of inflammation is associated with the occurrence of CVD. Many studies have found that patients with RA have a greater intima media thickness (IMT) of the common carotid and/or femoral arteries than healthy controls.1 17 18 Interestingly, there is ample direct evidence for a relationship between inflammation and the occurrence of CVD in RA.17 We therefore performed a case-control study in a cohort of patients with RA to examine whether the level of disease activity is associated with the risk of myocardial infarction (MI). We chose MI from the cardiovascular disease spectrum because its pathogenesis and traditional risk factors are well defined and it has a relatively high prevalence in the population.

METHODS

Patients

This was a case-control study nested in a large prospective cohort of patients with RA that started in 1985.19 Consecutive patients who fulfilled the 1958 (later, 1987) American College of Rheumatology criteria for the diagnosis of RA, with disease duration <1 year and without prior disease-modifying antirheumatic drug (DMARD) use, were included in the study. Trained personnel regularly assess the course of the disease and data are continuously updated and stored in an electronic database. Comprehensive information on co-morbidities (including CVD) and medication use is regularly collected by review of the medical records. We had access to the medical files of each patient including the periods before and after diagnosis of RA, so the registration of co-morbidities and medical events was complete. Cases for this study were patients who either developed a first MI after the diagnosis of RA or developed a first-time episode of unstable angina (UA). Cases were selected if the diagnosis of MI or UA was made by a cardiologist. The presence of ischaemic symptoms, ECG changes and raised cardiac enzyme levels were verified from the patient’s medical records. Cases with UA were selected if the presence of multiple coronary vessel disease was confirmed by angiography indicating a need for bypass surgery or percutaneous transluminal coronary angioplasty (PTCA). Controls were selected randomly from the cohort of patients with RA, aiming to have similar disease duration (exposition time to inflammation) for cases and controls. All patients could serve as a control up to the moment that they experienced their first event or up to the end of their observation time. This is an accepted way of control sampling in epidemiology, with the advantage that the odds ratio (OR) becomes a good estimate of the relative risk.20 No further matching on MI risk factors was applied, because matching is statistically inefficient and unnecessary if it is corrected for confounders in a statistical model.20

Assessment of disease activity

In the cohort, disease activity was prospectively assessed using the DAS28, which is calculated from the 28-tender joint count (TE28), 28-swollen joint count (SW28), erythrocyte sedimentation rate (ESR) and the patient’s global assessment of disease-related general health on a visual analogue scale (GH).21 Disease activity markers, including C-reactive protein (CRP), were collected 3-monthly. A time-averaged DAS28 was calculated for cases and controls using the area under the curve of the DAS28 divided by time, which is similar to an “area under the curve” approach but easier to interpret. The median of the time-averaged DAS28 was used to divide the patients into two equally sized groups with “low” (below the median) and “high” (median and higher) time-averaged disease activity. Time-averaged DAS28 as a continuous variable was also examined, which usually gives more power. To study the influence of time (ie, whether early inflammation or rather a rise in inflammation later contributes to the occurrence of MI), the analysis was also performed using the time-averaged DAS28 of the first year after diagnosis, as well as using the last year before event or censoring. As DAS28 is a pooled index, the effect of time-averaged ESR and CRP were also examined.

Assessment of other cardiovascular risk factors

Traditional cardiovascular risk factors include male gender, obesity, hypertension, diabetes mellitus (DM), smoking, positive family history of CVD, dyslipidaemia with high triglyceride (TG) levels, low serum HDL and high LDL levels and adopting an inactive lifestyle. In addition, other variables including rheumatoid factor (RF) status, disability and treatment with methotrexate (MTX), corticosteroids, tumour necrosis factor α antagonists (anti-TNFα) and non-steroidal anti-inflammatory drugs (NSAIDs) were regarded as risk factors and potential confounders.

Body mass index (BMI) was calculated from height and weight at baseline. DM was regarded as present if the diagnosis was made before the event or censoring. Hypertension was regarded as present if the diagnosis was made and antihypertensive medication was prescribed before the event or censoring. Information on smoking history and occurrence of CVD (coronary heart disease, heart failure, stroke or transient ischaemic attack, peripheral arterial disease) in first- and second-grade relatives is usually assessed in entry visits and was obtained from the medical records. If this information was missing, patients were approached by letter. Systematic information on active lifestyle was lacking for most patients with RA and so this was omitted.

Lipids could only be assessed from serum samples at baseline. Baseline lipids were measured in stored serum samples gathered within 6 months after diagnosis of RA. Serum levels of total cholesterol (TC), TG and HDL were determined on a Hitachi 747 analyser (Boehringer Mannheim, Germany). LDL serum levels were calculated according to the Friedewald formula.22 The atherogenic index (AI) was calculated as TC/HDL.

The presence of RF was assessed at baseline. Disability was also assessed at baseline using the disability index of the Health Assessment Questionnaire (HAQ). Medication use was assessed using a computerised register of information from the medical records. Total treatment duration, cumulative doses and number of treatment courses were calculated for MTX, corticosteroids, NSAIDs, anti-TNFα and folic acid use.

Study power

For the association between disease activity and MI, we expected an odds ratio (OR) of 3.0 based on the results of a pilot study. It was calculated that, with 40 cases and 180 controls, an expected OR of 3.0 for the low against high disease activity contrast and α = 0.05, the power exceeded 0.80.

Statistical analysis

Missing data on confounders were imputed using a single imputation method with multiple linear regression, conditionally on the data being “missing at random”.23

Cases of MI and controls were compared by disease activity variables and risk factors using the Mann-Whitney U test (for continuous variables) or the χ² test (for dichotomous variables).

The median of the time-averaged DAS28 was used to divide the patients into two equally sized groups with “low” (below the median) and “high” (median and higher) time-averaged disease activity. Potential confounders were identified among the available MI risk factors by univariate logistic regression analysis with time-averaged disease activity (below median or higher) as a dependent variable. Among these, serum lipids presumably act as an intermediate between systemic inflammation and MI.10 15 In this study, however, lipids were not measured over time but only at baseline, and baseline lipids were treated as a potential confounder because disease activity was measured after baseline.

To test the hypothesis that the level of disease activity over time is associated with an increased risk of MI, a conditional logistic regression model was used with occurrence of MI as the dependent variable and time-averaged DAS28 category (“low” or “high”) as the explaining variable. Potential confounders were added stepwise to the model, with a change by 10% or more in the beta coefficient of the DAS28 as selection criterion.

The data were analysed using SPSS statistical package Version 12.0.1 (SPSS, Chicago, Illinois, USA).

RESULTS

Patients

There were 544 patients with RA in the database at the moment of case and control selection. Forty-one patients were selected as cases, 38 had suffered MI and 3 had UA with multi-vessel disease requiring surgical intervention after the diagnosis of RA. One hundred and eighty-one controls were randomly selected.

After collection of the required additional data, a small amount of data remained missing. BMI could not be calculated in 14.9% due to missing data on height and/or weight. Data were also missing on smoking habit (1.8%), family history of CVD (7.2%), HAQ (1.8%) and serum lipid values (14.9%) due to missing serum samples. These data were found to be “missing at random” and were imputed. After imputation, average values of these variables remained the same for cases and controls.

Table 1 shows the characteristics of cases and controls. Of the classical risk factors, cases were significantly older, more often male, had higher BMI, TC, LDL, TG and AI and lower HDL serum levels than controls. However, none of the disease activity variables showed a significant difference between cases and controls. The time-averaged DAS28 in cases was 4.00 and in controls 3.94. The average DAS28 of the year before event (or censoring) and the average DAS28 of the first year after diagnosis were also calculated and were not different between cases and controls (data not shown). Use of MTX, folic acid, oral prednisone and methylprednisolone, anti-TNFα, NSAIDs, COX-2 inhibitors and naproxen was not different between cases and controls (table 2). Dosages and treatment durations were also analysed but no differences were found. Cardiovascular medication was used more often in patients who became cases (table 2).

Potential confounders

In univariate logistic regression analysis, high time-averaged disease activity (median DAS28 >4.0) was significantly associated with age, gender, disease duration, BMI, HDL, AI, HAQ and oral prednisone (OPR) use. Smoking, diabetes, hypertension, RF, family history of CVD, serum total cholesterol, LDL and TG were not associated with the DAS28 category (table 3) and were therefore not treated as potential confounders. Colinearity between the potential confounders was unlikely to be present as the highest correlation coefficient was 0.31 (Spearman correlation between age and AI).

Relationship between disease activity and MI

The raw OR of DAS28 level and occurrence of MI was 1.2 (95% CI 0.6 to 2.4). The OR corrected for age, gender, BMI, HDL and baseline HAQ-DI was 0.91 (95% CI 0.39 to 2.12) (table 4). The AI had a similar effect in the model as HDL. Inclusion of baseline lipids in the model did not remove an apparent effect, which might have occurred if it acted as an intermediate variable.

When the time-averaged DAS28 was used as continuous variable in the multiple logistic regression model, the raw and corrected ORs were 1.05 (95% CI 0.78 to 1.41) and 0.91 (95% CI 0.62 to 1.34), respectively. The OR here means the effect of an increase of 1 point in the time-averaged DAS28.

The results of analyses of the time-averaged DAS28 in the first year after diagnosis and of the time-averaged DAS28 in the last year before event or censoring did not lead to different results (not shown). Disease duration did not appear to be an effect modifier; the relation between DAS28 and MI was not different for patients with different disease duration.

Instead of time-averaged DAS28 we also used the time-averaged ESR and CRP; however, the results were similar (not shown).

DISCUSSION

In this case-control study we examined whether the level of disease activity over time is associated with the risk of MI in patients with RA. We expected to find a positive relationship based on proposed pathways.6 However, according to the results of this study, patients with RA who developed an MI after the diagnosis of RA had more classical risk factors but not higher time-averaged disease activity than patients without MI. As CRP may perform better in estimating the risk of MI than DAS28, we performed post hoc analyses using CRP instead of DAS28, but there was again no relationship with occurrence of MI. Also, the use of different cut-off points in time-averaged DAS28, a continuous DAS28 and looking at different periods of exposition time did not lead to other conclusions.

Notably, these findings are not in contradiction with the hypothesis that systemic inflammation increases the risk of CVD. Instead, it supports the hypothesis that low levels of inflammation are sufficient for accelerated atherogenesis and excess risk of CVD in RA.6 In the general population, already low levels of CRP (1–3 mg/l) are associated with a twofold risk of CVD.24 In RA, the magnitude and duration of the systemic inflammation may be a strong driving force of atherogenic processes, even when disease activity is low.6

Few studies have investigated the effect of disease activity in RA on the development of a first cardiovascular event. In these studies, a higher risk of cardiovascular events with more inflammation was found, although the effect was small.

In a retrospective study, high disease activity significantly increased the risk of CVD.25 However, only the last ESR value before the event was independently and significantly related to the cardiovascular risk. A recent prospective study did find a slightly increased hazard ratio for mean ESR and CRP when investigating the risk factors for a first cardiovascular event and for cardiovascular mortality in an RA cohort.26 Joint counts or compound measures of disease activity were not assessed and patients were not followed up from time of diagnosis. Another study conducted in male veterans found that the DAS28 was associated with an increased likelihood of experiencing a cardiovascular event, although the effect was small. The DAS28 used in this analysis was the score obtained within 6 months before the event for cases and the first score recorded in the registry for patients without an event.27 In this study, only male patients were involved and it is unclear whether the relationship between CVD risk and inflammation is the same for both genders.

There is some evidence for the hypothesis that low levels of systemic inflammation may increase the cardiovascular risk in patients with RA. Young patients with low disease activity, measured as a DAS28 <3.2, had an increased IMT of the carotid artery compared with patients without inflammatory rheumatic disease.28 Other studies have shown that young to middle-aged patients with RA, free of cardiovascular risk factors or cardiovascular disease, with low disease activity have an altered endothelial reactivity measured by brachial flow-mediated vasodilatation.29^–31

It is possible that, in RA, the magnitude of systemic inflammation does not play a major role in causing CHD because the level of inflammation is sufficiently heightened even if disease activity is “low”. Disease duration is probably more important than the magnitude of the inflammation in RA. Longer disease duration increases patient exposure to the deleterious effects of systemic inflammation, which may increase the risk of CHD. Studies that failed to demonstrate a significant increase in disease activity markers in patients with RA with atherosclerosis and/or a higher incidence of CVD did find that an increased risk of CVD was associated with disease duration and/or damage as a result of longstanding disease.1 4 18 32 33 We could not detect an effect of disease duration in our study because cases and controls had similar disease duration by design, which is appropriate for studying the effect of the level of disease activity over time on the risk of MI.

An advantage of this study is that data from an RA cohort could be used, with prospective and systematic collection of exposure, outcome and most confounding factors. The number of missing values was reasonably low. By the design of the study, it was arranged that MI cases and controls had similar disease duration so that both groups had a similar time of exposure to disease activity. All other confounders were corrected for by statistical modelling, which is as effective as (and more efficient than) matching on risk factors (eg, age and gender). A limitation of our study is that some of the cardiovascular risk factors that could act as confounders, such as smoking and BMI, had to be gathered by re-examination of the patients’ medical records. Systematic information on adopting an active lifestyle was not available for most patients, but the level of disability was. Lipids were measured using stored serum samples. The age of the samples did not differ substantially between cases and controls, allowing the same processes of degradation to take place in the lipid concentration of both groups. Despite aging samples, we were able to show that cases had higher levels of TC, LDL and TG and a lower HDL concentration than controls. We were not able to measure the lipid profile of patients with RA over time, which is also of interest because the effect of inflammation on MI may be mediated through the actions of lipids.10 Risk factors that were actual confounders such as BMI and HAQ were assessed at baseline (RA onset). However, it is unlikely that better measurement of confounders (eg, time-averaged BMI instead of BMI at baseline) would have a large influence on the results.

The treatment of systemic inflammation may also contribute to the risk of MI. In our study we did not find significant differences between the RA treatment of cases and controls. NSAID use, including COX-2 inhibitors, was equal in the two groups, providing no explanation of the occurrence of cardiovascular morbidity in the cases. Large-sized studies have shown that MTX use exhibited a protective effect against the cardiovascular risk in RA.34 35 Furthermore, the use of oral prednisone is suggested to be associated with a higher cardiovascular risk.36 37 However, this study was neither designed nor large enough to reproduce these results.

In conclusion, this study indicates that the level of disease activity, measured by the time-averaged DAS28, does not increase the risk of MI in patients with established RA. In view of the currently available evidence, it is likely that traditional cardiovascular risk factors, as well as the chronic inflammatory processes, mediate the cardiovascular risk in patients with RA. The relative degree of inflammation seems not to contribute to the risk. As the results of this study were a priori unexpected, replication by other studies would be valuable.