Early diagnosis and prompt initiation of appropriate treatment have been recognised as essential factors for improved clinical outcomes in early rheumatoid arthritis (RA).1 In the earliest phases of the disease the currently used criteria (ie, the American College of Rheumatology (ACR) 1987 classification criteria)2 are frequently not fulfilled,3 4 with an ensuing risk of a delay in diagnosis and treatment initiation. Magnetic resonance imaging (MRI) is known to be more sensitive than conventional clinical examination and radiography for detection of inflammatory and destructive joint changes in RA joints,5 6 and all the early features of the disease, including synovitis and bone erosion at its earliest stages, can be directly visualised.5 6 However, most of these studies have been performed in patients with known diagnoses and only a few studies have been in patients with undifferentiated arthritis.7 8 The differential diagnostic value of MRI is therefore unknown.

Bone scintigraphy has been investigated as a method for objective assessment of joint involvement in RA.9 10 The results on its differential diagnostic value vary.6 10 Despite a low specificity and poor resolution, bone scintigraphy may—through its ability to reveal the pattern of joint involvement—be of value in the differential diagnostic process, alone or as a supplement to MRI which provides detailed information on few joints.

The aim of the present study was to investigate the value in clinical practice of hand MRI and whole body bone scintigraphy in the differential diagnosis of patients with unclassified arthritis.

METHODS

Patients and design

Forty-one patients with arthritis (⩾2 swollen joints, >6 months’ symptom duration) and subjective symptoms in the hand (pain and/or swelling) who remained unclassified despite conventional clinical (including assessment of tender and swollen joints), biochemical (including serum C-reactive protein, erythrocyte sedimentation rate and IgM rheumatoid factor) and radiographic examinations (hands, wrists and feet) were included in the study. The characteristics of the patients are given in table 1. Patients who fulfilled the ACR 1987 criteria for RA2 or had radiographic bone erosions were excluded.

Contrast enhanced MRI of the wrist and metacarpophalangeal (MCP) joints of the most symptomatic hand and whole body bone scintigraphy were performed. Two rheumatologists subsequently agreed on the most likely diagnosis and the patients were treated accordingly. A final diagnosis was made by another specialist review 2 years later.

Imaging procedures

MRI of the wrist and 2nd–5th MCP joints of the most symptomatic hand was performed at baseline using a 0.2 Tesla Artoscan system (ESAOTE Biomedica, Genova, Italy), a dedicated extremity MRI unit. The imaging protocol comprised a coronal short T1 inversion recovery (STIR) sequence followed by coronal and axial T1-weighted spin echo images obtained before and after intravenous injection of gadodiamide (Omniscan, Nycomed, Amersham, UK), 0.1 mmol/kg body weight, via a peripheral vein (see Lindegaard et al11 for details on the MRI sequences). Whole body scintigraphy was performed using a Philips (ADAC) gamma camera (Vertex or Skylight) equipped with a low energy high resolution collimator (LEHR), with acquisition 2 h after injection of 700 MBq ^99mTc-labelled hydroxymethylene-diphosphonate (HDP, Oxidronate, Mallinckrodt, Chesterfield, UK).

Image evaluation

Radiographs of hands, wrists and feet were evaluated according to the Larsen method12 and for the presence of other abnormalities. As patients with radiographic erosions were excluded, only Larsen grades 0 (normal) and 1 (presence of joint space narrowing, soft tissue swelling and/or juxtaarticular osteoporosis) could be registered.

MR images were evaluated for the presence of synovitis and bone erosions and each patient was classified according to predefined patterns of MRI synovitis and MRI bone erosions as follows:

• MRI synovitis patterns: compatible with RA: at least one joint, not 1st carpometacarpal (CMC1) joints; compatible with osteoarthritis (OA): CMC1; compatible with RA/OA: mixed involvement of CMC1 and other joints.

• MRI erosion pattern: compatible with RA: at least one joint, not CMC1; compatible with OA: CMC1; compatible with RA/OA: mixed involvement of CMC1 and other joints.

Similarly, each patient was classified based on their joint involvement pattern on bone scintigraphy, as follows:

• Scintigraphic patterns: compatible with RA: several joints, but not distal interphalangeal joints (DIP) and CMC1; compatible with OA: CMC1 or proximal interphalangeal (PIP) and DIP joints; compatible with RA/OA: mixed involvement of CMC1/DIP joints and other joints, eg, MCP/wrist joints.

These patterns were assigned in order to be able to calculate the sensitivity, specificity, accuracy and predictive value of different MRI and scintigraphic appearances in the differential diagnosis of undifferentiated arthritis (see below).

Statistical analyses

The sensitivity, specificity, accuracy, positive predictive value and negative predictive value of bone scintigraphy and MRI parameters, and of selected biochemical and radiographical findings, were calculated.

RESULTS

Forty-one patients were included in the study. Their baseline characteristics are shown in table 1. The conventional radiography was normal in 19 patients and showed Larsen grade 1 changes (compatible with RA) in 5 patients, minimal OA changes in 13, and calcinosis in 4. MRI assessment for synovitis was normal in 15 patients, and “show” patterns compatible with RA and OA in 23 and 2 patients, respectively. One presented with tenosynovitis but no signs of synovitis. With regard to erosions, 27 MRI scans were normal while 14 had erosions compatible with RA. Bone scintigraphy was normal in 14 patients, showed appearances compatible with RA and with OA in 10 and 10 patients, respectively, and compatible with RA/OA in 4 patients.

Tentative diagnoses after MRI and bone scintigraphy were: RA in 13, OA in 8, other inflammatory diseases in 11, and arthralgias without inflammatory or degenerative origin in 9. Two years later, 11 of the 13 patients with an original tentative diagnosis of RA had fulfilled the ACR criteria for RA while 2 were reclassified (one to psoriatic arthritis (RF negative + psoriasis) and one to non-specific self-limiting arthritis). No patients classified as non-RA at baseline had fulfilled the ACR criteria for RA after 2 years. In other words, in 39 of 41 patients the baseline diagnoses of RA or non-RA assigned by the treating physician after MRI and bone scintigraphy were correct.

Table 2 shows the performance of key baseline conventional, MRI and scintigraphic findings for predicting a diagnosis of RA at 2 years’ follow-up.

DISCUSSION

This study shows the potential of MRI and bone scintigraphy in the differential diagnostic process of undifferentiated arthritis in routine clinical practice. When using fulfillment of the ACR 1987 RA criteria after 2 years of follow-up as the the “gold standard” reference, 95% of patients with polyarthritis could be correctly classified as RA or non-RA after baseline MRI and scintigraphy results were made available to the treating rheumatologist.

MRI is known to be more sensitive than conventional clinical examination and radiography for detecting inflammatory and destructive joint changes in RA.5 6 Small studies have shown that, in RA, MRI signs of inflammation are more frequent in the synovial membrane than at the insertions of ligaments and tendons (enthesitis), while the opposite is true for seronegative spondyloarthropathies.13 14 However, definite answers concerning the differential diagnostic value of MRI should obviously be achieved through longitudinal studies of patients with undifferentiated arthritis. Studies of this kind are scarce. In two small studies it has been suggested that incorporation of MRI signs of synovitis in the ACR criteria for RA would increase their accuracy, leading to an earlier diagnosis of some patients with RA.7 8 These findings have not been retested on other cohorts and, to our knowledge, no other data on the value of MRI in the diagnosis of RA have been published.

In the present study, patients with varying disease durations were seen in the routine clinical setting of a rheumatological outpatient clinic rather than in a strictly controlled scientific protocol, and both MRI and bone scintigraphy were applied. It should be noted that the exact basis of the assignment of the tentative diagnosis by the treating rheumatologist—that is, the extent to which the decision was based on one specific finding or a combination of several—was not registered. This could potentially have provided additional information. However, the advantage of the present design is that it reflects clinical practice and, consequently, supports a value in the practical clinical management of patients with undifferentiated disease. Ultrasonography, genetic typing and measurements of anti-cyclic citrullinated peptide antibody (anti-CCP) and experimental biomarkers were not included in the present study. This should be done in a future cohort.

The performance of the individual clinical, biochemical, scintigraphic and MRI findings for correctly classifying patients at baseline as RA or non-RA were investigated (table 2). To achieve specificities and positive predictive values of 100%, both MRI synovitis and MRI bone erosion as well as a scintigraphy pattern compatible with RA needed to be present. However, the simultaneous presence of MRI bone erosions and MRI synovitis also showed a high specificity for RA. In the present study, all patients with RA as the final diagnosis had MRI synovitis at baseline.

The fact that MRI and scintigraphy in combination had a better diagnostic performance than either method alone could be explained by the fact that the advantages and disadvantages of MRI and scintigraphy are very different. MRI provides detailed information on inflammatory as well as destructive changes in the examined joints but provides limited information on the joint involvement pattern as it only covers a limited number of joints. In contrast, bone scintigraphy shows the joint pattern but provides no detailed information on the local joint pathology.

The fact that rheumatologists perform better when integrating the MRI and scintigraphic data in the overall impression of the patient than when the sensitivities and specificities of the scintigraphic and MRI parameters are analysed separately is in accordance with the diagnosis of RA as a multifactorial complex process integrating different types of information. It also suggests that, even though MRI or scintigraphy will probably never alone be able to assign specific diagnoses, it can be a very useful addition in the differential diagnostic process. Our current knowledge strongly encourages further testing in patients with suspected or unclassified arthritis.

In future studies of undifferentiated arthritis, the value of MRI and/or bone scintigraphy should be compared with the contributions of other potential diagnostic determinants such as anti-CCP, clinical/biochemical disease activity measures, radiographic erosions and promising biomarkers, eg, by logistic regression analysis with the aim of developing the best possible prediction model as previously performed (without incorporating MRI or scintigraphy) by Visser and colleagues.15

In patients with arthritis of varying disease durations which was unclassified despite conventional clinical, biochemical and radiographic examinations, MRI and scintigraphy allowed correct classification as RA or non-RA in 39 of 41 patients when fulfilment of ACR criteria 2 years later was considered the standard reference. This strongly encourages further testing of these methods in patients with suspected or unclassified arthritis.