Rheumatoid arthritis (RA) is a chronic, inflammatory disease of peripheral joints often associated with progressive joint destruction and functional loss. Modern treatment strategies include early diagnosis, close monitoring and aggressive use of disease-modifying drugs (DMARDs) aiming at inflammatory control immediately after diagnosis to prevent joint destruction.1^–4 However, despite an aggressive strategy in the treatment of early RA, 25–50% of patients progress radiographically within 1 year after diagnosis.1 5

Substantial efforts have been exerted to identify this subset of patients with a poor prognosis at the time of diagnosis and several promising prognostic markers have been identified. Traditionally, high disease activity as well as radiographic erosions and IgM rheumatoid factor (RF) positivity at disease onset have been associated with poor prognosis,6 but these variables cannot predict radiographic progression in individual patients. Newer prognostic markers such as antibodies against cyclic citrullinated peptide (anti-CCP) and magnetic resonance imaging (MRI) have shown promising results. A recent meta-analysis concluded that the presence of anti-CCP antibodies in patients with RA was associated with greater radiographic progression and that the risk of radiographic progression was greater for patients with anti-CCP antibody positivity than for those with IgM RF positivity.7 MRI bone marrow oedema at disease onset has been shown to be related to the progression of joint damage 1–6 years later.8^–10 However, in these small studies (24–31 patients), the relative importance of other new potential predictive factors, such as anti-CCP, compared with MRI bone oedema, was not investigated. A recent study included anti-CCP in the model, but in a non-standardised treatment setting and with only 1 year of follow-up.11 In addition, the treatment strategy and a number of genetic and environmental factors (age, gender, educational level, smoking and shared epitope) may affect the outcome of RA12 and therefore should be taken into account in studies of potential tools for prognostication in RA.

The aim of this study was to investigate whether early MRI pathology had prognostic value for the development of x-ray erosions. A pool of potential prognostic markers was examined to identify independent predictors of radiographic progression during 2-years’ follow-up in a randomised controlled clinical trial of patients with early RA, who were treated aggressively with conventional DMARDs.

PATIENTS AND METHODS

Patient population

The CIMESTRA trial was an investigator-initiated, multicentre, randomised, double-blind, parallel-group, placebo-controlled trial that included 160 consecutive patients with early active RA. The design and the clinical and radiographic results from the first 2 years of the study have been published previously1 13 and the study is only briefly outlined below.

All 160 patients in the CIMESTRA trial were offered the possibility of inclusion in the present MRI substudy and 130 patients were included. Thirty patients were not included for various reasons: absolute contraindications for MRI (n = 3), disease activity that did not allow positioning in the MR scanner (n = 8), anxiety/claustrophobia (n = 5), patient refusal (n = 7), unknown (n = 1), MRI data lost owing to technical problems with storage media (n = 6). This paper presents data from the 130 patients who participated in the MRI substudy.

Inclusion and exclusion criteria

At study entry, all participants fulfilled the American College of Rheumatology 1987 revised criteria for RA,14 were DMARD naïve, had active disease of <6 months’ duration with at least two swollen joints and were aged 18–75 years. Table 1 presents the baseline characteristics of the study group.

Treatment strategy

The patients were treated aggressively to achieve tight disease control with conventional DMARDs (step-up doses of either methotrexate (7.5–20 mg/week) and placebo-ciclosporin (2.5–4.0 mg/kg) or methotrexate (7.5–20 mg/week) in combination with ciclosporin (2.5–4.0 mg/kg)). In addition, any swollen joint was treated with intra-articular injections of betamethasone. During the second year, ciclosporin (placebo-ciclosporin) was tapered to zero, hydroxychloroquine was added to the treatment of all patients and methotrexate and intra-articular injections with betamethasone were continued.

Disease activity

Clinical evaluations were performed monthly by experienced, blinded rheumatologists. Disease activity was assessed by the Disease Activity Score, using a 28-joint score (DAS28).15 Functional status was assessed using the Danish version of the Stanford Health Assessment Questionnaire.16

Laboratory assessments

RF of IgM and IgA isotypes was detected by enzyme-linked immunosorbent assay (ELISA) as previously described with a few modifications.17 Cut-off levels were >16 IU/ml and >24 U/ml, respectively (∼95th centile of healthy subjects). Anti-CCP IgG antibodies were determined by a second-generation ELISA (Immunoscan RA kit, Euro-Diagnostica AB, Malmö, Sweden) in accordance with manufacturer’s instructions and with the recommended >25 U/ml cut-off point.7

Human leucocyte antigen (HLA)-DRB1 genotyping for shared epitope (SE) was performed by polymerase chain reaction-based, sequence-specific oligonucleotide probing, as described elsewhere.18 Here, we define the SE as the presence of HLA-DRB1*04 or HLA-DRB1*01, or both. Genomic DNA was isolated from EDTA-preserved blood cells, using a QIAamp Maxi Kit (Qiagen, Chatsworth, California, USA) in accordance with the manufacturer’s instructions and stored at −20°C before HLA–DRB1 tissue typing.

Serum C-reactive protein (mg/l) and erythrocyte sedimentation rate (ESR; mm/1st h) were measured using standard laboratory methods.

MRI

Contrast-enhanced MRI and intra-articular betamethasone injections were performed before the start of treatment in the 130 patients who entered the MRI substudy, in conjunction with the clinical, laboratory and radiographic assessments at baseline.

MRI covered the non-dominant wrist in all 130 patients and for 89 patients, in whom the field of view allowed it, also the second to fifth metacarpophalangeal (MCP) joints. In patients from the hospitals at Graasten (n = 61), Odense (n = 21) and Herlev (n = 9) a 0.2 T dedicated extremity MRI unit (Artoscan, Esaote Biomedica, Genoa, Italy) equipped with a dual phased array coil was used. In Hvidovre (n = 17) and Aarhus (n = 22), respectively, 1.0 T and 1.5 T whole-body MRI units (Siemens Impact and Siemens Vision, Erlangen, Germany), both equipped with circular polarised transmit–receive extremity coils, were used. MRI sequences included coronal and axial T1-weighted images (slice thickness 3 mm; matrix 192×192–384) before and after intravenous gadolinium-contrast injection (0.1 mmol gadolinium-DTPA-BMA/kg bodyweight (Omniscan, Amersham Health, Copenhagen, Denmark)) and a coronal short tau inversion recovery sequence (slice thickness 3 mm, matrix size 144–182×192–256).

MRI evaluation

The MR image sets were assessed for bone erosions, synovitis and bone marrow oedema according to the OMERACT (Outcome Measures in Rheumatology) MRI scoring system (RAMRIS)19 by an independent rheumatologist (BE), who was trained in the evaluation of MR images of RA joints. The reader was blinded to the treatment group assignment, clinical, biochemical and radiographic results.

Conventional radiography (CR)

Conventional radiographs of the hands (posteroanterior and Nørgaard20 projections), wrists (posteroanterior and lateral projections) and forefeet (anteroposterior view) were obtained at baseline and at 6 months, 1 and 2 years. The radiographs were scored by an independent senior musculoskeletal radiologist (AV) using the Sharp–van der Heijde scoring method.21 The reader was blinded to treatment group assignment, clinical, biochemical and MRI data, but not to the chronological order of the images.

Ethical considerations

All patients gave their written informed consent to participate. The protocol was approved by the national health authorities and ethics committees (reference number M-1959-98). The trial was performed in accordance with the Declaration of Helsinki and the International Conference on Harmonisation 1996 revised Guidelines for Good Clinical Practice in the European Community. The trial was registered at http://www.clinicaltrials.gov (reference number NCT00209859, accessed 21 November 2008).

Statistical analysis

Comparisons between groups were made with the Fisher exact test for dichotomous responses and the Mann–Whitney U test for non-dichotomous responses. Changes over time were analysed with the McNemar test for dichotomous responses and the Wilcoxon signed rank sum test for non-dichotomous responses.

Possible predictors of radiographic progression were initially tested in univariate analyses, secondarily in a multiple regression analyses. In addition to the significant (p<0.05) variables in the univariate analysis (MRI variables and Total Sharp Score (TSS)), the disease activity (DAS28), gender, age, shared epitope as well as anti-CCP, smoking habits (ever/never smoker) and their interaction were forced into the model because of the previously reported significance of these variables.7 When two covariates correlated significantly, only one was included (anti-CCP included, rheumatoid factors excluded; DAS28 included, tender and swollen joint count, ESR and patient pain and global scores, doctor’s global score excluded). The potential predictors were included as explanatory covariates in a multiple regression analysis with delta-TSS as the dependent variable. Backward stepwise selection was performed. Covariates that were excluded either initially or during backward selection were reintroduced one by one into the final model, which gave similar results. The immunological covariates were tested both as dichotomous and as continuous variables and dichotomous variables are presented. Model assessment was made using standard techniques. In addition, multivariate logistic regression analysis was performed with the same covariates in the model and with radiographic progression at 2 years (delta-TSS >0 as cut-off point) as the dependent variable. Analysis was by intention to treat. Completers’ analysis was also performed and gave similar results. Data are reported as the mean (SD) for variables with normal distribution; otherwise as the median and interquartile range (IQR).

The R software package22 was used for statistical analysis, which was performed by an independent statistician.

RESULTS

Patient characteristics

The two MRI populations (wrist-only, n = 130 and wrist+MCP, n = 89) had similar baseline and radiological status after 2 years (table 1). At baseline, the patients had active disease as judged by clinical, laboratory and MRI findings, a majority had erosive disease. By 2 years, disease activity had been almost completely suppressed with a decline in the median DAS28 score from 5.6 at baseline to 2.0 (in the wrist+MCP group) and 2.2 (in the wrist-only group), respectively. Thus, more than 50% of the patients were in DAS remission at 2 years. Twenty-six per cent (of 89) and 30% (of 130 patients) had progressed radiographically by 2 years (delta-TSS >0 unit). The average progression rate in TSS was 1.6–1.8 units/year (range: 0–43). The MRI and radiographic findings were similar in the two treatment groups, which were therefore pooled before statistical analyses.

Predictors of radiographic progression

TSS at baseline, MRI erosion score and MRI bone marrow oedema score were significantly associated with radiographic progression after 2 years (delta-TSS) in univariate regression analysis in both subgroups (0.001<p<0.02); the MRI synovitis score was significantly associated with radiographic progression in the wrist + MCP subgroup (p = 0.03) (table 2). Anti-CCP was borderline significant in the wrist-only subgroup (p = 0.09). No other variable was statistically significantly associated with radiographic progression.

Multiple linear regression analysis with backward selection showed that only bone marrow oedema remained in the final model in both the wrist-only MRI group and the wrist + MCP MRI group (table 3). MRI bone marrow oedema explained 41% of the variation in the progression of the TSS (Pearson’s r = 0.64) in the wrist + MCP subgroup. In the wrist-only subgroup, the results were similar, but weaker (25% explained, r = 0.50).

Sensitivity analyses (logistic regression analysis, completers’ analysis and all-subsets regression analysis) gave similar results (data not shown). Thus, in all models, MRI bone marrow oedema was the strongest predictor of radiographic progression after 2 years.

DISCUSSION

As far as we know, this is the first clinical trial with a standardised treatment protocol that investigates the predictive value of a variety of potential prognostic markers including both imaging (MRI and conventional x-ray examination), immunological (anti-CCP, IgM RF and IgA RF), environmental (smoking, educational level), genetic (shared epitope) and disease activity markers. Our main finding was that in this comprehensive model, MRI bone marrow oedema at presentation was the strongest predictor of radiographic progression 2 years later in patients with early RA.

A number of interesting imaging studies have demonstrated that MRI pathology (synovitis, bone erosions or bone marrow oedema) of the wrist at disease onset in early RA predicts radiographic erosions.8^–11 23^–25 Some looked at the short-term predictive value of MRI after 1 year,9 11 23 24 other studies had two,10 six8 or 10 years25 of follow-up and found MRI to be a highly significant predictor of radiographic erosions. A high baseline combined score of wrist joint MRI erosions and synovitis was the best predictor of radiographic erosive progression in the hand, wrists and feet after 10 years in 114 patients with early RA.25 In that study, however, bone marrow oedema was not assessed and the separate predictive values of MRI bone erosions and of MRI synovitis were not reported, which make direct comparison with our study difficult. Bone marrow oedema at baseline was the strongest individual predictor of MRI bone erosions at 1 and 2 years’ follow-up in two small studies of early RA,10 24 but conventional x-ray examinations, generally considered to be the “gold standard” for structural joint damage, were not carried out. A study of dedicated extremity MRI in 25 patients with early RA found a relative risk of 4.0 for x-ray erosions after 1 year, if erosions or oedema were present at baseline.9

All previous studies, except one,11 were imaging studies that did not include anti-CCP or take into account other potential prognostic markers such as smoking or shared epitope carriage. The latter study reported that bone marrow oedema and male gender (but not anti-CCP) were independent predictors of radiographic progression after 1 year. In contrast to our study, the cohort did not receive a standardised treatment regimen and it was, as for all other previous studies, carried out at a single-centre. The most significant predictor of disease outcome in RA is the treatment,12 but studies performed with a standardised treatment protocol have been lacking. In some of the previous studies of predictors in RA the clinical disease control was modest (as a result of previous treatment strategies), allowing severe joint damage to develop during the follow-up period.8

In our study, a relatively high proportion of patients had erosive disease at baseline. It is not known whether this might have influenced the results of the analysis of the predictive value of different variables.

A recent meta-analysis of the predictive value of anti-CCP antibodies in RA concluded that anti-CCP-positive patients had greater risk of radiographic progression than anti-CCP-negative patients.7 The reviewed studies only included a limited number of variables besides anti-CCP antibodies and RF and none included MRI. One study of patients with RA (<1 year’s disease) included smoking status and HLA-SE in addition to anti-CCP.26 In multivariate analysis, the baseline erosive score was the most significant prognostic marker of radiographic progression after 2 years, followed by anti-CCP positivity and raised ESR. In another study of early RA (<2 years), anti-CCP and C-reactive protein were the only significant predictors of Larsen score after 10 years.27 Only some patients received DMARD treatment and methotrexate was used in only a few patients. The median increase in radiographic score was eight units in two years,26 and 46 units in 10 years,27 reflecting poorer disease control than achieved by today’s treatment strategies. In a recent study of 238 patients with RA (10 years’ follow-up), anti-CCP was the strongest independent predictor of radiographic progression.28 Also in that study, the progression rate was higher (2.9 units/year) than in our work.

The above-mentioned serological and biochemical risk factors for radiographic progression were not identified as such in our study. This may be owing to the tight disease control and low rate of radiographic progression obtained in our study, thus leaving little power to discriminate associations with radiographic changes.

HLA-SE positivity has been associated with the presence of anti-CCP antibodies in early RA, thus playing an indirect role as a risk factor for erosive disease.29 In a mortality study of 767 patients with early RA, those with the DRB1*0101/0401 genotype were found to develop significantly higher Larsen scores by the second year than all patients with genotypes other than DRB1*0401/*0401.30 In our study, carriage of HLA-DRB1-SE was not found to be a prognostic marker for erosive disease. This may be owing to the limited radiographic progression in our study cohort, although a previous study is in accordance with our results.31

Smoking is a well-established risk factor for the development of anti-CCP positive RA,32 but its influence on RA disease progression is unclear. Two recent studies support our findings: in 2000 patients with RA with variable disease duration, radiographic joint damage progressed at an equivalent rate in smokers and non-smokers,33 and in 200 patients with early RA, smoking status did not influence radiographic progression after 2 years.34

Our study reflects the modern treatment strategy aiming at inflammatory control immediately after diagnosis. This goal was largely achieved and this is reflected by a very low rate of radiographic progression after 2 years. With only a minority of patients progressing radiographically, the set-up constitutes a challenge when one attempts to identify the subgroup of patients who have aggressive disease with erosive potential. The fact that the vast majority of patients responded well to the treatment strategy may explain, at least in part, why IgM RF, anti-CCP and disease activity, which have been identified as predictive markers in other studies, were not predictive of joint damage in this study.

MRI bone marrow oedema has recently in established RA been shown to represent inflammatory infiltrates in the bone marrow.35 36 In contrast to radiographic erosions, which reflect bone damage that has already occurred, bone marrow oedema thus may represent an important part of the early immunopathological development in RA.35

Methodological studies have reported that the sensitivity for detecting bone marrow oedema may vary with different types of MRI units.37 Our study had five participating centres. More than half of the MRI scans (70%) were done using low-field dedicated extremity MRI units, which are less sensitive than high-field units. This indicates that MRI bone marrow oedema is a robust predictor of erosive progression. Expectedly, the predictive value was highest in the subgroup that had MRI of both wrist and MCP joints than in the wrist-only group. Regional MRI as a predictor of radiographical progression in other anatomical regions confirms previous studies.8

It can be considered a disadvantage of this study that different MRI units, including low-field (0.2 T) and 1.5 T units, were used. However, this was unavoidable owing to the multicentre design and lack of identical units at different hospitals. Further, only one reader read each image. This is, however, in accordance with previous MRI studies, and the MRI reader was experienced and had in previous studies shown a high agreement with other trained readers.38

The number of patients who progressed radiographically in this study was low owing to the tight disease control which was achieved. This may influence the results of the multiple regression analysis, which assumes a normal distribution of the dependent variable in the model. It is reassuring that our sensitivity analyses (logistic regression analysis, completers’ analysis and all-subsets regression analysis) confirmed the results of the multiple regression analysis model. This study indicates that MRI might be a useful supplement to the conventional examination programme in patients with early RA, in order to optimise the identification of patients at high risk of erosive progression.

In conclusion, this multicentre study of patients with early RA treated in a randomised controlled trial with DMARDs and intra-articular glucocorticoids aiming at maximal inflammatory control showed that the baseline MRI bone oedema score of MCP and wrist joints, as well as of wrist joints only, was a strong independent predictor of radiographic progression in hands, wrists and forefeet during the subsequent 2 years. DAS28, anti-CCP, smoking, shared epitope, MRI synovitis score, MRI erosion score and radiographic score at baseline were not independent risk factors.