Article Text

Extended report
Cardiovascular risk factors and not disease activity, severity or therapy associate with renal dysfunction in patients with rheumatoid arthritis
  1. D Daoussis1,
  2. V F Panoulas1,2,
  3. I Antonopoulos1,
  4. H John1,
  5. T E Toms1,
  6. P Wong1,
  7. P Nightingale3,
  8. K M J Douglas1,
  9. G D Kitas1,4
  1. 1Department of Rheumatology, Dudley Group of Hospitals NHS Trust, Russells Hall Hospital, Dudley, West Midlands, UK
  2. 2Department of Internal Medicine, School of Medicine, University of Ioannina, Ioannina, Greece
  3. 3Wolfson Computer Laboratory, University Hospital Birmingham NHS Foundation Trust, Birmingham, UK
  4. 4ARC Epidemiology Unit, Manchester University, Manchester, UK
  1. Correspondence to Professor G D Kitas, Department of Rheumatology, Dudley Group of Hospitals NHS Trust, Russells Hall Hospital, Pensnett Road, Dudley, West Midlands DY1 2HQ, UK; gd.kitas@dgoh.nhs. uk; g.d.kitas{at}


Objectives The present study aimed to evaluate the prevalence and associations of renal dysfunction in patients with rheumatoid arthritis (RA). It specifically addressed the hypotheses that renal dysfunction in these patients may associate with the presence of insulin resistance, dyslipidaemia, uric acid levels and/or current levels of systemic inflammation.

Methods Renal function was assessed by estimated glomerular filtration rate (GFR) using the modification of diet in renal disease equation in 400 consecutive RA patients for this cross-sectional, single-centre study. Risk factors for renal dysfunction were recorded/measured in all participants. Correlations between GFR and other variables were analysed by Pearson or Spearman test as appropriate. Linear regression was used to test the independence of the associations between GFR and other variables.

Results In this RA patient cohort, 67.75% of patients had a reduced GFR of less than 90 ml/minute per 1.73 m2 and 12.75% had a GFR of less than 60 ml/minute per 1.73 m2. Multivariable analysis revealed significant associations between GFR and age (β = −0.370, p<0.001), female sex (β = −0.181, p=0.002), total cholesterol (β = −0.112, p=0.022), serum uric acid (SUA) (β = −0.425, p<0.001) and the presence of extra-articular disease, apart from sicca and/or nodules (β = −0.084, p=0.040).

Conclusions Renal dysfunction in RA is quite common and associates with classic cardiovascular risk factors such as advanced age and dyslipidaemia, levels of SUA and the presence of extra-articular disease. Renal dysfunction was not related to other RA-related factors including disease activity and duration, disability and past or present use of nephrotoxic medications.

Statistics from

The cause of renal impairment in rheumatoid arthritis (RA) patients has been a subject of debate; previous studies have principally focused on the role of nephrotoxic medication1 although some investigators have highlighted the role of comorbidities or the inflammatory process per se.2 A possible association between insulin resistance and renal impairment has recently emerged from general population studies35 but this has not been specifically investigated in patients with RA despite the evidence for the presence of insulin resistance in RA.6 The aetiological role of inflammation in renal impairment has also received considerable attention but again not specifically in patients with RA.710 Hyperuricaemia is known to associate with impaired renal function and recent evidence supports the view that uric acid may not just be an innocent bystander but instead may be an active player in the pathogenesis of renal disease1113 in general, while we have shown that serum uric acid (SUA) independently associates with both hypertension14 and cardiovascular disease (CVD)15 in RA.

In this study we sought to evaluate the prevalence and associations of renal dysfunction in patients with RA. We also specifically addressed the hypotheses that renal dysfunction in these patients may associate with the presence of metabolic factors (such as insulin resistance and dyslipidaemia), SUA levels and/or current levels of systemic inflammation. To our knowledge, this is the first study in the literature to evaluate the potential association of renal dysfunction with insulin resistance and hyperuricaemia in RA.

Patients and methods

The study was approved by the local Ethics Committee and Research and Development Directorate and all participants gave their written informed consent according to the Declaration of Helsinki. Four hundred consecutive patients fulfilling the 1987 revised American College of Rheumatology criteria for RA were enrolled in this cross-sectional study.

Renal function assessment was made by glomerular filtration rate (GFR) estimation using three different predictive equations (the modification of diet in renal disease (MDRD), the abbreviated MDRD formula and the classic Cockcroft Gault formula). Data shown refer only to MDRD.

Risk factors for renal dysfunction recorded for each patient included: (1) hypertension; (2) fasting lipid profile; (3) insulin resistance evaluated using the homeostasis model assessment of insulin resistance (HOMA IR) and the quantitative insulin sensitivity check index (QUICKI); (4) obesity, by calculating body mass index (BMI) and (5) medication, especially potentially nephrotoxic drugs. Patients with extra-articular disease were categorised as follows: patients with sicca only; patients with rheumatoid nodules (with or without sicca) and patients with nailfold or systemic vasculitis (including neuropathy, ulcers purpura, etc), pulmonary fibrosis, serositis, scleritis or Felty's syndrome. CVD was defined as the presence of coronary heart disease, cerebrovascular accident (either stroke or transient ischaemic attack) or peripheral vascular disease. Coronary heart disease was defined as having had any of the following: myocardial infarction, angioplasty, coronary artery bypass grafting or angina diagnosed by a physician or elicited by the use of the Rose questionnaire.

To assess inflammatory activity we measured several markers including C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), fibrinogen, ferritin and white blood cells. More details are presented in a supplemental file (available online only).

Statistical analysis

Statistical analysis was performed using the SPSS software, version 14. Data are presented as mean (SD), median (inter-quartile range) or percentages as appropriate. Correlations between GFR (continuous variable) and other variables were analysed by Pearson or Spearman test as appropriate. Linear regression was used to test the independence of the associations between GFR and other variables. Significance was defined as p<0.05 (two-tailed).


Mean (SD) age was 61.6 (12) years. Patients were predominantly women, 292 (73%) and of Caucasian origin, 388 (97%). The mean (SD) of the GFR for the total population was 81.9 (21.2) ml/minute per 1.73 m2. There were 129 (32.25%) patients with GFR greater than 90 ml/minute per 1.73 m2, 220 (55%) patients with GFR between 60 and 90 indicating mild renal impairment and 47 (11.75%) patients with GFR less than 60, indicating moderate renal impairment. There were only four (1%) patients with GFR less than 30, ie, severe renal impairment in this cohort.

Only a small percentage of patients had evidence of either proteinuria or haematuria (5.1% and 4%, respectively, as defined by positive dipstick testing).

Correlations of GFR with demographic and RA characteristics

There was a strong negative association with age (correlation coefficient r = −0.469, p<0.001) and disease duration (r = −0.129, p=0.010). Smoking status was associated with GFR; post hoc analysis of variance revealed significantly poorer renal function among ex-smokers when compared with current smokers (−9.81 ml/min per 1.73 m2 (SD 3.15), p=0.006). The presence of rheumatoid nodules and/or sicca was not associated with GFR but patients with extra-articular disease other than nodules and sicca had significantly worse renal function compared with those without (81.3 (SD 19.9) vs 86.1 (SD 22.2), p=0.022).

Correlations of GFR with risk factors for renal dysfunction and CVD

A strong negative correlation of GFR with the presence of hypertension was observed (normotensive vs hypertensive 88.6 (SD 18.8) vs 79.1 (SD 21.6), p<0.001). This association remained significant when we used a lower threshold (125/80) for the definition of hypertension (normotensive vs hypertensive 87.8 (SD 21.14) vs 80.63 (SD 21.03), p=0.008). The main association was with systolic blood pressure (r = −0.185, p<0.001) when systolic and diastolic blood pressure were analysed as continuous variables. Analysis of the lipid profile revealed negative associations of total cholesterol and triglycerides with GFR (r = −0.100, p=0.046 and r = −0.182, p<0.001, respectively). The presence of diabetes mellitus was not associated with renal dysfunction, but a significant negative association was observed with insulin resistance as evaluated by HOMA IR and QUICKI (r = −0.146, p=0.004 and r = 0.144, p=0.005). SUA significantly correlated to GFR (r = −0.449, p<0.001). Renal function was inversely associated with BMI but this association did not reach statistical significance. The presence of CVD was significantly correlated to GFR (r = −0.127, p=0.011).

Correlations of GFR with medication

As expected, the majority of these patients were taking disease-modifying antirheumatic drug (DMARD) therapy (87.5%), either as monotherapy or combination therapy; the most frequently prescribed DMARD in this cohort was methotrexate (56% of all patients). Other DMARD prescribed were sulphasalazine (29.5%), hydroxychloroquine (20%), leflunomide (4%), azathioprine (1.8%) and ciclosporine (0.5%). No patients in this cohort were current users of either gold or penicillamine, but a limited number (19 for gold and 35 for penicillamine) had used these agents in the past; 11.5% of patients were on anti-TNFa biological therapy. The GFR was significantly higher among patients on DMARD compared with those not on such treatment (82.8 (SD 21) vs 75.4 (SD 21.7), p=0.021). Only sulphasalazine use was associated with significantly better renal function (87.2 (SD 21.6) vs 79.6 (SD 20.7), p=0.001).

Surprisingly, NSAID use was associated with a higher GFR compared with non-users (88.6 (SD 23.9) vs 80.2 (SD 20.2), p=0.002). Oral prednisolone use was associated with worse renal function (78.8 (SD 22.9) vs 83.4 (SD 20.3), p=0.043). No associations between renal function and coxib use were observed in this cohort. The use of angiotensin-converting enzyme inhibitors (73.2 (SD 22) vs 84.9 (SD 20.1), p<0.001), loop diuretics (65.3 (SD 22.6) vs 82.9 (SD 20.7), p<0.001), thiazide diuretics (76.4 (SD 19.2) vs 83.2 (SD 21.5), p=0.011) and statins (74.9 (SD 20.9) vs 83.6 (SD 21), p=0.001) was associated with a significantly reduced GFR in all cases. A history of the previous use of any DMARD (including penicillamine and gold) or NSAID was not significantly associated with GFR.

Correlations of GFR with markers of inflammation and DAS28 score

There was a trend towards a negative correlation of GFR with ESR (r = −0.094, p=0.060), haemoglobin (r = 0.093, p=0.063) and ferritin (r = <0.089, p=0.080). No associations were observed with CRP, fibrinogen, white blood cells and the disease activity in 28 joints (DAS28) score.

Multivariable analyses

Linear regression was used to evaluate whether the abovementioned associations were independent. A basic model was constructed to include all the risk factors for renal impairment that contributed significantly in the univariable analyses (age, disease duration, smoking status, systolic blood pressure, total cholesterol, triglycerides, HOMA IR, SUA, presence of extraarticular disease (other than nodules, sicca), CVD, use of DMARD, NSAID, oral prednisolone, angiotensin-converting enzyme inhibitors, diuretics and statins). We also included gender and BMI in the model as potential confounders. The R2 of the model was 0.408 and variables significantly associated with GFR were: age (standardised coefficient β = 20.370, p<0.001), female sex (β = −0.181, p<0.001), total cholesterol (β = −0.112, p=0.022), SUA (β = −0.425, p<0.001) and the presence of extra-articular disease (β = −0.084, p=0.040). When adding individually ESR, ferritin or haemoglobin in the model, these were not significantly correlated with GFR. When using stepwise linear regression the most significant variables that were independently associated with GFR were (presented in order of significance): age (β = −0.462 p<0.001), SUA (β = −0.366, p<0.001), female sex (β = −0.208, p<0.001) and the presence of extra-articular disease (other than nodules, sicca) (β = −0.093, p=0.023). R2 of the model was 0.4.

When data were analysed with GFR as a categorical variable (>90, 60–90, <60) the results were very similar (see table 1). Analyses based on other methods of GFR assessment, including the use of the Cockroft Gault, the abbreviated MDRD formula or serum creatine as the dependent variable, also produced very similar results (data not shown).

Table 1

Demographics, RA characteristics and risk factors for renal impairment in the cohort


In this study we aimed to determine the prevalence and main predictors of renal dysfunction in patients with RA. Advanced age, increased SUA levels, female gender, the presence of extra-articular disease and increased levels of total cholesterol were significantly and independently associated with renal dysfunction in RA.

The first hypothesis in the present study regarded the potential role of insulin resistance and dyslipidaemia in renal dysfunction in RA patients. A recent large-scale study identified a positive, strong association between insulin resistance and chronic kidney disease in non-diabetic patients, independent of other risk factors.16 The results of the present study suggest the presence of an independent association of renal function with hypercholesterolaemia but not with insulin resistance in RA patients. Of note, however, is that hypercholesterolaemia and uric acid, both of which were independent GFR predictors in this study are either directly or indirectly associated with insulin resistance. It may therefore be worthwhile exploring further the potential association of renal dysfunction with insulin resistance in RA before any definite conclusions are drawn.

The second most powerful predictor of GFR in the present study was SUA, confirming our second hypothesis. Clinical studies have suggested a potential role of uric acid in the progression of renal disease,17 18 but the most compelling evidence to support the potential role of uric acid in renal dysfunction comes from animal models in which induced hyperuricaemia in normal rats led to glomerular hypertension that was prevented by allopurinol treatment.19 This potential association of uric acid with renal dysfunction could be partly mediated by CVD risk factors and metabolic abnormalities, because uric acid has been associated with increased cardiovascular or renal risk,20 21 hypertension22 and insulin resistance.23

The third hypothesis of the present study concerned the potential role of inflammation in the development of renal dysfunction. The potential association between renal dysfunction and “low grade” inflammation has already been evaluated in diseases such as diabetes mellitus24 and hypertension,25 in which it has been shown that inflammatory markers, especially CRP, may correlate to some extent with a decline in GFR even after adjustment for traditional risk factors. The results of this study are not supportive of a direct association between the “high grade” inflammatory process of RA with renal impairment, but definite conclusions cannot be drawn because this is a cross-sectional study and inflammatory markers vary over time.

The presence of extra-articular disease, however, (other than sicca and nodules) was an independent predictor of renal dysfunction in this cohort. The presence of extra-articular disease has been linked to excess mortality in community-based cohorts26 and associates with vascular disease.27 Therefore, extra-articular disease could be a marker of vascular disease that subsequently leads to renal impairment. An alternative explanation may be the probable higher “disease burden” among RA patients with extra-articular manifestations,28 which renders them more likely to develop amyloidosis.

The major limitations of the present study are its crosssectional design, which does not allow inferences regarding the directionality of cause and effect relationships and the indirect measurement of GFR. Radioisotope methods with the use of chromium–EDTA are considered the gold standard for direct GFR measurement but are expensive, time consuming and not easily applied in large cohorts such as this. Conversely, 24-h urine collections aiming to determine creatinine clearance are inaccurate and are being abandoned.29 Estimated GFR from predictive equations is generally accurate and has been validated in very large cohorts.30 We chose to use the MDRD formula, even though not specifically validated for patients with RA, because this is the most widely used in the clinical setting and is considered the most accurate formula in general.31 Nevertherless, our results were similar when we used the Cockcroft Gault formula, which has been validated in RA,32 33 and this consistency enhances their strength.

In summary, this study shows that renal dysfunction is quite common in RA and appears to be mainly mediated through classical cardiovascular risk factors such as dyslipidaemia. Novel CVD risk factors in RA, such as SUA and extra-articular disease, may also be of importance, but not other RA characteristics, including duration, activity, severity or the use of antirheumatic medications. The potential role of renal dysfunction either as a predictor or a risk factor for CVD has been increasingly recognised in recent years. Previously, the CVD risk profile was considered discrete from renal dysfunction, but new clinical data have emerged indicating a strong link between the two.34 Prevention strategies for either renal dysfunction or CVD can be effectively the same and clinically important for both.

Notwithstanding this, we propose that all patients with RA should undergo regular monitoring of their renal function with the use of GFR predictive equations rather than the far less reliable creatinine, while specific subgroups, such as older patients with CVD comorbidities, extra-articular disease and/or high uric acid levels may need particularly intensive monitoring.


Supplementary materials

  • Web Only Data ard.2008.105049

    Files in this Data Supplement:


  • Funding This study was funded by the Dudley Group of Hospitals R&D Directorate cardiovascular programme grant. The Department of Rheumatology is in receipt of infrastructure support from the Arthritis Research Campaign (grant number 17682).

  • Competing interests None.

  • Ethics approval The study was approved by the local Ethics Committee and Research and Development Directorate.

  • Patient consent Obtained.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.