Introduction

Spondyloarthritis (SpA) comprises a group of diseases, which are phenotypically distinct although they share several clinical, biochemical and radiological characteristics. The clinical presentation of SpA is heterogeneous and no single symptom or clinical, imaging or laboratory finding is pathognomonic for the disease . During the last three decades, different classification criteria based on combinations of various symptoms, clinical, imaging and laboratory findings have guided clinicians to identify patients with SpA . Until recently, confidence in a diagnosis of axial SpA after clinical and biochemical examination was often primarily determined by the presence of sacroiliitis on radiography. This major role of pelvic radiographs for making a diagnosis of SpA was one of the principal reasons for a diagnostic delay of several years , because radiography captures only post-inflammatory structural changes and is unable to depict early inflammatory lesions in the sacroiliac joints (SIJs) . By contrast, magnetic resonance imaging (MRI) is more sensitive and superior to scintigraphy , computed tomography (CT) and radiography to detect active sacroiliitis in patients with inflammatory back pain who show normal SIJs on pelvic radiographs or equivocal SIJ changes not compatible with definite radiographic sacroiliitis.

The emergence of tumour necrosis factor-alpha (TNFα) inhibitors, a powerful treatment option in patients with ankylosing spondylitis (AS) , has called for new and more sensitive criteria to facilitate early diagnosis and treatment. This need to identify SpA patients early in the disease course was addressed by the Assessment of SpondyloArthritis international Society (ASAS) classification criteria for axial and for peripheral SpA , which were published in 2009 and 2011, respectively. In contrast to previous classification criteria for SpA, the ASAS criteria are the first to include findings on MRI of the SIJ. In these criteria, inflammatory MRI features are weighted equally with structural radiographic changes for detection of sacroiliitis, and imaging has become one of the two major criteria for axial SpA . Thus, MRI has become an important tool in daily practice for evaluating patients with early SpA suspected on clinical grounds.

The aims of this review are to: 1) describe the evidence for the use of MRI as a diagnostic aid in SpA, 2) to discuss the strengths and weaknesses of MRI as a diagnostic tool and 3) to summarise how to use MRI in daily routine to facilitate early diagnosis of SpA. This review is primarily based on studies that aimed to investigate the diagnostic utility of MRI in patients with early SpA including patients with back pain due to other reasons and healthy subjects, and it mainly focusses on studies published within the last 5 years.

The ASAS classification criteria for axial and peripheral SpA

According to the ASAS classification criteria for axial SpA , a patient with chronic back pain of more than 3 months’ duration and symptom onset before the age of 45 years can be classified as having axial SpA if either sacroiliitis is present on imaging (pelvic radiographs or SIJ MRI) and at least one additional clinical or laboratory feature is present (so-called ‘imaging arm’), or if human leucocyte antigen B27 (HLA-B27) is positive and at least two additional SpA-associated features are present (‘clinical arm’). The ASAS classification criteria for peripheral SpA comprise patients with predominantly peripheral symptoms. According to these criteria, a patient with arthritis, enthesitis or dactylitis and with an age of onset before 45 years can be classified as having peripheral SpA, if at least one additional SpA-associated feature including sacroiliitis on imaging is present or if at least two additional clinical or laboratory features indicative for SpA are present. In both sets of SpA classification criteria, sacroiliitis on MRI has been accorded as much weight as sacroiliitis on radiography. Sacroiliitis on MRI is defined by consensus as active inflammation as described below, whereas sacroiliitis on radiography is defined according to the modified New York criteria, that is, as bilateral grade 2 or a unilateral grade 3–4 sacroiliitis .

The ASAS classification criteria for axial and peripheral SpA

According to the ASAS classification criteria for axial SpA , a patient with chronic back pain of more than 3 months’ duration and symptom onset before the age of 45 years can be classified as having axial SpA if either sacroiliitis is present on imaging (pelvic radiographs or SIJ MRI) and at least one additional clinical or laboratory feature is present (so-called ‘imaging arm’), or if human leucocyte antigen B27 (HLA-B27) is positive and at least two additional SpA-associated features are present (‘clinical arm’). The ASAS classification criteria for peripheral SpA comprise patients with predominantly peripheral symptoms. According to these criteria, a patient with arthritis, enthesitis or dactylitis and with an age of onset before 45 years can be classified as having peripheral SpA, if at least one additional SpA-associated feature including sacroiliitis on imaging is present or if at least two additional clinical or laboratory features indicative for SpA are present. In both sets of SpA classification criteria, sacroiliitis on MRI has been accorded as much weight as sacroiliitis on radiography. Sacroiliitis on MRI is defined by consensus as active inflammation as described below, whereas sacroiliitis on radiography is defined according to the modified New York criteria, that is, as bilateral grade 2 or a unilateral grade 3–4 sacroiliitis .

The MRI definition of the ASAS classification criteria for axial and peripheral SpA

As sacroiliitis on MRI had not been defined formally before, an ASAS/OMERACT (Outcome Measures in Rheumatology Clinical Trials) MRI working group of radiologists and rheumatologists was established . The aim of the group was to identify and describe MRI findings in sacroiliitis and to reach a consensus on which features are essential for its definition. The group reviewed MRI scans from patients with axial SpA and healthy subjects, and provided a consensus statement ( Table 1 ). A ‘positive’ SIJ MRI was defined as showing active inflammation highly suggestive of SpA and located in typical areas, that is, in the subchondral or periarticular bone marrow of the SIJ . Inflammation should be clearly present as either bone marrow oedema (BME) on short-tau-inversion-recovery (STIR) images or hyper-intensive signal in the bone marrow on T1-weighted fat-saturated (T1wFS) post-gadolinium images (osteitis). To fulfil the definition, either one BME lesion should be seen on at least two consecutive MRI slices or at least two BME lesions should be present on a minimum of one MRI slice. The definition did not include other active inflammatory lesions in the SIJs such as synovitis, enthesitis or capsulitis, nor structural lesions such as fat infiltration, erosion, sclerosis or ankylosis nor findings on MRI of the spine.

Development and validation of the ASAS classification criteria for axial SpA

The ASAS classification criteria for axial SpA were developed by a two-step approach . In the first study, 20 ASAS experts reviewed standardised clinical records including clinical, biochemical and imaging data from 71 real-life patients referred to a SpA centre with the diagnosis of possible axial SpA . The reviews were performed twice in random order, first excluding information on SIJ MRI (i.e., active inflammation) and second including this information. In this study, a total of 27 (38%) patients had sacroiliitis on MRI. The availability of additional MRI information changed the diagnosis from non-SpA to SpA in 15 of 71 (21%) patients, and allowed classification of 11 of 23 patients who could not be classified based on clinical and biochemical information alone. Logistic regression analyses showed that SIJ MRI strongly contributed to the classification of patients with axial SpA.

In the second study, 25 rheumatology centres in Europe, Asia and North and Latin America enrolled 649 consecutive patients referred with possible axial SpA . All patients had radiographs of the SIJs, and 495 (76.3%) had MRI of the SIJs. Definite radiographic sacroiliitis and MRI inflammation in SIJs, as assessed by the local radiologists and/or rheumatologists, was present in 40.4% and 64.7%, respectively, of the 391 (60.2%) patients diagnosed with axial SpA according to the local rheumatologist. In contrast, they were present in only 3.1% and 2.6%, respectively, of the patients given a non-SpA diagnosis. However, interpretation of radiographs and SIJ MRI scans was performed un-blinded with all clinical and biochemical data available to the local radiologist and/or rheumatologist, who evaluated the images. The final ASAS criteria overall had a sensitivity of 82.9%, a specificity of 84.4% and positive and negative likelihood ratios of 5.3 and 0.20, respectively, when tested in the data set they were generated from. The imaging arm of the criteria had a sensitivity of 66.2% and a specificity of 97.3%, resulting in a positive likelihood ratio of 24.5 and a negative likelihood ratio of 0.35. Because information on active inflammation on MRI contributed substantially to the classification of patients with axial SpA in both studies, it was included as one of the cornerstones of the new classification criteria.

Development and validation of the ASAS classification criteria for peripheral SpA

The ASAS classification criteria for peripheral SpA were developed using two sets of pre-specified candidate criteria, which were developed by the use of cases from study 1 (described above) and clinical experience . These candidate criteria were tested and modified based on clinical, laboratory and imaging data of 266 consecutive patients referred with peripheral arthritis or enthesitis or dactylitis and suspicion of peripheral SpA. The local rheumatologist diagnosed 176 (66.2%) of the patients as having peripheral SpA. Radiography and MRI of SIJs had been performed in 227 (85.3%) and 60 (22.5%) patients, respectively. Definite radiographic sacroiliitis and inflammation on SIJ MRI were present in 32 (19.5%) and 22 (44.0%) of the patients with peripheral SpA, who had these examinations performed, whereas it was present in only 2 (3.2%) and 0 (0%) of patients with non-SpA. Again, all clinical, laboratory and imaging data were available to the local radiologist and rheumatologist when interpreting the images and making the diagnosis. In a multivariate logistic regression analysis, definite radiographic sacroiliitis was independently associated with SpA together with age, HLA-B27, positive family history, enthesitis and psoriasis when adjusted for gender and the other SpA features included in the criteria. MRI was not included in the analyses because of too few cases. The sensitivity and specificity of the classification criteria for peripheral SpA, when including sacroiliitis on radiography and/or MRI, were 77.8% and 82.2%, respectively, when tested in the data set they were generated from. As sacroiliitis on radiography in regression analyses independently contributed to the classification of SpA and as inflammation on MRI in univariate analyses was associated with SpA, both were included in the classification criteria for peripheral SpA.

Studies of the diagnostic utility of SIJ MRI in axial SpA

MRI of the SIJs

Diagnostic utility of the ASAS definition and subsequent proposals for a positive SIJ MRI

A series of studies in an international collaboration (the MORPHO studies, named after the similarity of light reflection patterns on the opened wings of rainforest Morpho butterflies with bright BME signal of sacroiliitis on MRI) aimed at assessing the diagnostic utility of SIJ MRI to differentiate patients with early axial SpA from patients with non-specific back pain (NSBP) and healthy subjects. The study also compared the diagnostic utility of the ASAS MRI definition with a global assessment of SIJ MRI and a new more comprehensive definition proposed by the MORPHO working group . The approach of a global assessment of SIJ or spinal MRI represents an evaluation of both T1-weighted and STIR sequences simultaneously, and readers conclude whether SpA is present or not after reviewing all scans in both sequences. The MORPHO definition of a ‘positive’ SIJ MRI requires either (Option 1) BME to be present according to the ASAS definition, ‘or’ (Option 2) erosions to be present in at least two SIJ quadrants on the same MRI slice or in a single SIJ quadrant on at least two consecutive MRI slices, ‘or’ (Option 3) BME and erosions to be present in at least one quadrant each (not necessarily on the same MRI slice . The SIJ MRIs were evaluated blinded by a group of SpA experts using standardised definitions of BME and erosion . The presence/absence of these lesions was dichotomously recorded in four quadrants of each SIJ on all MRI slices of the cartilaginous portion of the joint. The first MORPHO study included 187 subjects: 27 (14%) consecutive patients with non-radiographic axial SpA and 75 (49%) patients with AS, all with an age below 45 years, and a sex- and an age-matched control group of 26 (14%) patients with NSBP and 59 (32%) healthy subjects. All patients had been referred to two rheumatologic outpatient clinics with inflammatory back pain and/or clinical suspicion of axial SpA; the final diagnosis made by the local rheumatologist in both institutions based on clinical examination, laboratory values and pelvic radiographs served as clinical gold standard. The ASAS MRI definition had higher sensitivity and lower specificity for detection of patients with inflammatory back pain than patients with NSBP and healthy subjects, as compared to the MRI global assessment ( Table 2 ). The MORPHO definition, based on both BME and/or erosions, resulted in increased sensitivity compared to the ASAS MRI definition, whereas the specificity remained unchanged . Validation of the three definitions in a new inception cohort of 27 patients with inflammatory back pain, 26 patients with NSBP and 10 healthy subjects provided similar results, except for a higher sensitivity of the global assessment ( Table 2 ). A smaller study from Leeds , in which the diagnostic utility of the ASAS MRI definition and MRI global assessment were compared, reported sensitivities and specificities comparable to those found in the MORPHO studies ( Table 2 ). Furthermore, the MORPHO and Leeds studies demonstrated a high prevalence of BME (≥1 SIJ quadrant) not only in patients with AS (79–91%) and non-radiographic axial SpA (78%) , but also in patients with NSBP (23–27%) and healthy subjects (20–22%) ( Fig. 1 ). However, patients with SpA had a significantly higher number of quadrants with BME (mean (SD) 14.2 (16.5) per patient) as compared to controls (0.9 (2.3), p < 0.0001) . Erosions (erosions in ≥1 SIJ quadrant) were seen in the majority of patients with AS (93%), in half of the patients with non-radiographic axial SpA, but rarely in patients with NSBP and healthy subjects (0–8%) ( Figs. 2–4 ). Fat infiltration was seen with similar frequencies as BME, while ankylosis was exclusively present in patients with AS ( Fig. 3–5 ). Preliminary results from the MORPHO study showed that SIJ MRI fat infiltrations had high sensitivity but low specificity for SpA . This study focussed on the diagnostic utility of selected features of SIJ fat infiltration considered to be indicative for SpA, such as a distinct border of fat infiltration, homogeneity or proximity to subchondral bone. However, the diagnostic utility of these morphologic features of SIJ fat infiltration was largely associated with simultaneous occurrence of BME and erosion.

Table 2

Diagnostic utility of SIJ MRI assessed by ASAS definition, MRI global assessment of T1-weighted and STIR sequence and MORPHO definition.

	Sensitivity	Specificity	Positive likelihood ratio	Negative likelihood ratio
MORPHO 2010
ASAS MRI definition a	0.67	0.88	5.7	0.4
MRI global assessment a	0.52	0.95	9.8	0.5
MORPHO definition a	0.81	0.88	6.9	0.2
MORPHO 2011, Cohort B
ASAS MRI definition a	0.72	0.85	4.3	0.3
MRI global assessment a	0.75	0.96	20.2	0.3
MORPHO definition a	0.81	0.82	7.0	0.2
Leeds study
ASAS MRI definition b	0.79	0.89	7.1	0.2
MRI global assessment b	0.66	0.94	11.8	0.4

The results are for concordant reads for any reader pair. The ASAS proposal for a positive MRI is ≥2 BME lesions in 2 distinct SIJ quadrants on the same slice or ≥1 BME lesion extending across 2 SIJ quadrants or ≥1 BME lesion recorded on 2 consecutive slices in the same SIJ quadrant. The MRI global assessment represents an evaluation of both T1-weighted and STIR sequences simultaneously, and readers record whether SpA is present or not. The MORPHO proposal for a positive MRI is fulfilment of any of the following 3 criteria: (Option 1) BME according to the ASAS definition or (Option 2) erosion in ≥2 SIJ quadrants in the same slice or a single SIJ quadrant in 2 consecutive slices, or (Option 3) BME and erosion in any SIJ quadrant though not necessarily in the same quadrant. ASAS, Assessment of SpondyloArthritis international Society; BME, Bone marrow oedema; MRI, magnetic resonance imaging; SIJ, sacroiliac joint; SpA, spondyloarthritis; STIR, Short-Tau-Inversion-Recovery.

a Patients with IBP vs. NSBP and healthy subjects (MORPHO 2012, n = 112; MORPHO 2011 n = 69).

b Patients with SpA ( n = 29) vs. healthy subjects ( n = 9).

The MRI of the sacroiliac joints is from a 20 year old healthy woman, who has not been pregnant. The two consecutive MRI slices of the anterior cartilagenous portion of the sacroiliac joints show subchondral bone marrow oedema on the STIR sequence of the right sacrum (arrows) ( Fig. 1 C and D). These features meet the ASAS definition of a positive sacroiliac joint MRI. The corresponding T1 sequence is normal ( Fig. 1 A and B).

The MORPHO and Leeds studies were the first to investigate the diagnostic utility of the ASAS MRI definition outside the ASAS cohort, and to compare it with alternative definitions of a ‘positive’ SIJ MRI for SpA in controlled cohorts with inflammatory back pain patients, NSBP patients and healthy subjects. Both studies demonstrated that systematic assessment of SIJ MRI according to pre-defined lesion definitions has great diagnostic utility in patients with inflammatory back pain/non-radiographic axial SpA. BME and fat infiltration were frequent in patients with AS and non-radiographic axial SpA as well as in patients with NSBP and in healthy subjects, whereas erosions were commonly seen only in patient with AS and non-radiographic axial SpA but rarely in the two control groups. These findings may limit the diagnostic utility of BME as applied in the current ASAS classification criteria and support the inclusion of T1-weighted images focussing on erosions in a definition of a positive SIJ MRI in SpA. The very low frequency of patients with BME in the non-SpA (2.6%) patient group in the validation study of the ASAS classification criteria may reflect that some patients with non-SpA were misclassified as SpA, because of lack of knowledge of the prevalence of BME on SIJ MRI in patients with non-inflammatory back pain, or because the ASAS study population, despite inclusion of consecutive patients newly referred with inflammatory back pain/suspicion of SpA, has been selected before referral and therefore truly has a high frequency of patients with axial SpA and inflammation on SIJ MRI.

Prognostic utility of SIJ MRI inflammation

Short-term changes of MRI inflammation in the SIJs have recently been investigated in the Leiden SPondyloArthritis Caught Early (SPACE) cohort, which included patients with back pain for more than 3 months and less than 2 years with onset before the age of 45 years . The first results published in abstracts showed that of 90 patients with SpA or possible SpA, 24 (27%) patients had MRI SIJ inflammation at inclusion or at follow-up 3 months later. In half of these patients (12), inflammation was present in the same SIJs at both time points, in four patients it had resolved and in three patients it was reduced to involve one joint only. Five patients showed inflammation at follow-up, which was not present at baseline. Changes in inflammation did not seem to be related to changes in medication, as most patients were treated with non-steroid-anti-inflammatory drugs (NSAIDs) in stable dose . The strongest predictor for a positive SIJ MRI according to the ASAS definition was a positive MRI at baseline, but HLA-B27 positivity and male gender were also strongly associated with a positive MRI 3 months later . The Early Spondyloarthritis Cohort (ESpAC) study comprised 62 patients with inflammatory back pain according to the Calin criteria, that is, age at onset before 40 years, insidious onset, more than 3 months’ duration, association with morning stiffness and improvement of back pain with exercise, who had MRIs performed at baseline and then annually for 2 years . In this study, HLA-B27 negative patients with a ‘negative’ SIJ MRI (according to the ASAS MRI definition) at baseline had a likelihood of 95% for a ‘negative’ MRI over the next 2 years . HLA-B27 positive patients with a ‘positive’ SIJ MRI at baseline had a likelihood of 88% for a ‘positive’ MRI in the two subsequent scans. These results are in consistency with the results from a study from Leeds, which showed that being HLA-B27 positive was associated with more severe SIJ inflammation on MRI, higher number of lesions and persistence of lesions after 1 year .

The Leeds group has investigated the predictive value of inflammation on SIJ MRI and HLA-B27 status regarding future development of AS. The first study comprised 40 patients from the Leeds early inflammatory back pain cohort . Seven patients developed AS according to the modified New York criteria after a mean of 7.7 years. All patients had extended SIJ MRI inflammation and were HLA-B27 positive. These patients had a high positive likelihood ratio of 8.0 for the future development of AS, whereas patients with mild sacroiliitis showed a low positive likelihood ratio of 0.4 regardless of their HLA-B27 status . In a second study, the group assessed the predictive validity of the ASAS definition for a positive SIJ MRI and of the global assessment of SIJ MRI scans by evaluating both T1w and STIR sequences simultaneously . The study population comprised 29 patients from the inflammatory back pain cohort, 9 patients with mechanical back pain and 9 healthy subjects. The ASAS definition had a very high sensitivity (100%), low specificity (33%) and low positive likelihood ratio (1.5) for the future development of AS, whereas the global assessment of SIJ MRI had the same sensitivity (100%), but higher specificity (56%) and positive likelihood ratio (2.3). Three patients, who developed AS at follow-up, had more MRI quadrants with inflammation at baseline (mean (SD) per patient 23.5 (3.1)) than 18 patients who did not develop AS (8.0 (13.9), NS) and the 18 controls (0.9 (2.3), p < 0.0001) .

In summary, longitudinal studies on SIJ MRI inflammation in patients with clinically suspected SpA having a symptom duration of less than 2 years showed that the majority of patients (65–79%) showed no SIJ MRI inflammation at baseline. Among patients without MRI inflammation at baseline, only a minority (7–14%) displayed SIJ BME on follow-up scans . The association between extensive inflammation on SIJ MRI and a positive HLA-B27 test may help to identify early SpA patients with a more severe disease course. However, no long-term follow-up studies have been performed, and the independent predictive values of a ‘negative’ and a ‘positive’ SIJ MRI have not been addressed. The studies from Leeds suggest that the severity of MRI inflammation is clinically relevant, and that inclusion of structural changes in the assessment of SIJ MRI may be even more important, to identify patients at risk of developing AS. However, the study population was small and only few patients developed AS.