Background The new Assessment of SpondyloArthritis international Society (ASAS) criteria classify axial spondyloarthritis (SpA) into human leucocyte antigen-B27 and/or imaging-based arms. To aid implementation, ASAS has proposed a definition of a positive MRI for active sacroiliitis.
Objective The authors aimed to test the diagnostic and predictive value of the ASAS criteria and definition of a ‘positive’ MRI.
Methods Baseline MRI scans on 29 patients with early inflammatory back pain and 18 controls were read independently by four experienced rheumatologists. Both arms of the criteria were tested against a ‘gold standard’ of physician diagnosis of SpA. MRI abnormalities were assessed according to a global assessment of MRI and the ASAS definition. Sensitivity, specificity and likelihood ratios for individual and concordant reader data were calculated for axial SpA diagnosis at baseline and the development of radiographic sacroiliitis, fulfilling the modified New York criteria at 8 years.
Results All patients were classified as having axial SpA, with more patients fulfilling the imaging arm (83%, n=24/29) than the human leucocyte antigen B27 arm (62%, n=18/29). Concordant reader data showed that the baseline MRI had high diagnostic utility for SpA according to global assessment (sensitivity/specificity: 66%/94%, LR+ (positive likelihood ratio) 11.8, LR− (negative likelihood ratio) 0.4) and ASAS definition (sensitivity/specificity: 79%/89%, LR+ 7.1, LR− 0.2). Likewise, a positive baseline MRI had 100% sensitivity for subsequent radiographic sacroiliitis by either assessment, although specificity was lower (56% for global assessment and 33% for ASAS definition).
Conclusion Both arms of the ASAS criteria have good diagnostic utility in early SpA, although they are of limited value for the prediction of radiographic progression. This may be due to the definition of a positive MRI for sacroiliitis that lacks specificity at baseline.
Statistics from Altmetric.com
A major challenge in the management of axial spondyloarthritis (SpA) is the substantial delay in diagnosis, which consistently averages 7–9 years.1,–,3 This reflects the lack of specificity of presenting clinical features of the disease, the lack of sensitivity of laboratory markers and the slow rate of radiographic progression in the sacroiliac joints (SIJs) despite ongoing symptoms of inflammatory back pain (IBP).4 This formed the key rationale for a complete reappraisal of the classification criteria for SpA that was undertaken by the Assessment of SpondyloArthritis international Society (ASAS) in a series of data-driven exercises that demonstrated that MRI had the capacity to substantially reclassify patients presenting with back pain into different diagnostic categories.5 6 A new set of criteria for axial SpA that incorporated MRI as an imaging criterion in addition to radiography was then formulated.6 Testing of these criteria in patients presenting with back pain for the first time to a rheumatologist demonstrated a sensitivity of 66% and a specificity of 97%.6
To help the widespread implementation of the new criteria, it became clear that there was a need to develop a definition of a positive MRI for sacroiliitis in axial SpA that would be feasible and reliable for use in clinical research.7 Recently, such a proposal was formulated through consensus opinion by a working group under the auspices of ASAS and OMERACT (Outcome Measures in Rheumatology).7This is based on the finding of bone marrow oedema (BMO) on the short τ inversion recovery (STIR) MRI sequence or osteitis on gadolinium-enhanced T1FS (T1-weighted fat-suppressed) MRI sequence in subchondral or periarticular bone marrow of the SIJ. According to this definition, an MRI is considered positive or ‘highly suggestive of SpA’ if two BMO lesions are present on the same coronal slice through the SIJ or if a single BMO lesion is present on two consecutive coronal slices. This definition now requires validation, preferably in early axial SpA cohorts followed up prospectively. In particular, the validation process should first address to what extent this ASAS MRI definition captures the diagnostic utility compared to the overall assessment of the MRI. Second, it is important to determine to what extent a baseline MRI that is considered positive according to the ASAS definition correctly predicts the development of radiographic sacroiliitis on long-term follow-up. Our aims were therefore twofold: first, we wanted to assess the diagnostic and prognostic validity of both arms of the new ASAS classification criteria in our cohort of patients with IBP, and second, we aimed to explore the utility, both for diagnosis and for prognosis of the ASAS definition of a positive MRI for sacroiliitis by comparing it to global assessment.
Patients and methods
Patients and controls
The Leeds IBP inception cohort has been previously described.8 Briefly, the cohort is composed of consecutive patients with IBP according to the Calin criteria9 with a median symptom duration of 24 weeks at presentation and a high clinical suspicion of a diagnosis of SpA. Subjects were recruited from regional early arthritis and IBP clinics and were available for follow-up after a mean of 7.7 years. Ethics approval was obtained, and all patients gave informed consent. For the current study, a full set of imaging data composed of baseline radiographs and MRI and follow-up radiographs of the SIJ was available in 29 patients. The majority of patients (n=21) were at the preradiographic stage at baseline, with 17 patients having grade 0 or grade 1 abnormalities and 4 patients having grade 2 unilateral changes. The rest (n=8) fulfilled the modified New York criteria (mNYC) for ankylosing spondylitis (AS) at baseline, with two patients having grade 2 bilateral changes and six patients having grade 3 bilateral changes in at least one SIJ. The intrareader reliability for change in radiographic status from baseline to follow-up was acceptable (κ=0.68, substantial agreement).10
All patients were treated with non-steroidal anti-inflammatory drugs, 17 (43%) received disease-modifying antirheumatic drugs and 5 (17.24%) received anti-tumour necrosis factor (TNF) treatments for short periods of time after baseline assessment (for a median of 1 year (range 1–3 years)). At follow-up, 11 patients (nine men, 90.9% human leucocyte antigen (HLA)-B27 positive) had AS according to the mNYC,11 two patients had psoriatic SpA, five patients had reactive SpA and 11 patients had undifferentiated SpA. Additional demographic information at baseline is provided in table 1.
In addition, baseline MRI scans were available on nine controls with mechanical low back pain (mean age 34.5 years (range 23–40 years)) and on nine healthy controls (mean age 35.6 years (range 22–39 years)).
Standardised assessment of MR images
STIR and T1-weighted sequences of semicoronal magnetic resonance (MR) scans of the SIJ were assessed and scored independently and in random order by four readers blinded to patient characteristics. The methodological approach to assessing MR images of the SIJ was standardised by adopting the methodology outlined in the Spondyloarthritis Research Consortium of Canada online training module reported previously.12 13 Readers were calibrated with a video teleconference session and a prereading exercise aimed at setting the threshold for detection of BMO lesions. MRI findings in the SIJ were recorded online using a customised data entry module that first includes a question that addresses the global assessment of each scan: ‘This MRI scan confirms the presence of SpA (agree/disagree)’. Second, BMO is recorded as being present/absent on a dichotomous basis in each SIJ quadrant (upper and lower ilium, upper and lower sacrum) with additional weighting for depth and intensity.
The distribution of BMO in patients and controls was analysed descriptively according to single readers, mean (range) values for all four readers and concordant data recorded by at least two readers. We considered the ASAS definition of sacroiliitis as being met in the data entry module in two circumstances. The first would be when a score of at least 2 for BMO was recorded on the same coronal slice, indicating two BMO lesions in two distinct SIJ quadrants or a single lesion extending across two SIJ quadrants. The second circumstance would also be a score of at least 2 for BMO, but this would be due to a score of 1, indicating a single BMO lesion, being recorded on two consecutive slices in the same SIJ quadrant. Diagnostic utility of MRI for SpA according to global evaluation of both T1-weighted and STIR MR scans and according to the ASAS definition for a positive MRI was determined by calculating sensitivity, specificity, positive likelihood ratio (LR+) and negative likelihood ratio (LR−) for individual reader data, concordant data according to at least two readers and data recorded concordantly by all four readers. We also compared the predictive value of global evaluation of the baseline MRI with the ASAS definition of a positive MRI, for the future development of radiographic sacroiliitis fulfilling the mNYC (hereinafter radiographic mNYC sacroiliitis) after 7.7 years of follow-up. Expert physician opinion was considered the gold standard for the clinical diagnosis of SpA at baseline.
Baseline characteristics of patients can be seen in table 1. According to the new ASAS criteria, all patients could be classified as having axial SpA at baseline, with more patients fulfilling the imaging arm (83% (n=24/29)) than the HLA-B27 arm (62% (n=18/29)) of the criteria. A total of 16 patients (57%) fulfilled both arms.
Diagnostic validity of the ASAS definition of a positive MR signal
The presence of BMO was reliably detected among reader pairs with the mean (range) intraclass correlation coefficient of six reader pairs for the total number of SIJ quadrants with BMO per patient being 0.87 (0.69–0.98). Patients with clinical SpA had a significantly greater number of SIJ quadrants with BMO on baseline MRI (mean (SD) per patient=14.2 (16.5)) as compared to controls (mean (SD) per individual=0.9 (2.3)) (p<0.0001). When concordant MRI data from any reader pair were used as the basis for analysis, BMO was observed in at least one SIJ quadrant in 79% of patients with a clinical diagnosis of SpA at baseline and in a substantial minority of controls (22.2%) (table 2). BMO in at least two SIJ quadrants and meeting the ASAS proposal for a positive MRI was recorded in the same proportion of patients clinically diagnosed as having SpA but in a lower proportion of controls (11.1%). A receiver operating characteristic analysis confirmed that a cut-off of two SIJ quadrants with BMO had the best combination of sensitivity and specificity for the clinical diagnosis of SpA for individual readers as well as for concordant data (data not shown).
The baseline MRI had high diagnostic utility for the clinical diagnosis of SpA at baseline according to global assessment of T1SE (T1-weighted spin echo) and STIR sequence images (mean (range) for four readers, sensitivity/specificity: 67.3% (62–72%)/94.3% (89–100%)). Analysis of concordant MRI data from any reader pair showed that sensitivity was 66% and specificity was 94% for the clinical diagnosis of SpA (LR+ 11.8, LR− 0.4) (table 3). Analysis of MRI data concordant for all four readers showed a sensitivity of 59% and a specificity of 100% for a clinical diagnosis of SpA. When testing the ASAS definition of a positive MRI for diagnostic utility in clinical SpA, sensitivity was noted to be higher but specificity was lower (mean (range) for four readers: 79% (62–83%) and 89% (78–100%)). Analysis of concordant MRI data from any reader pair using the ASAS definition showed that sensitivity was 79% and specificity was 89% for the clinical diagnosis of SpA (LR+ 7.1, LR− 0.2).
Prognostic utility of ASAS criteria and definition of positive MR signal
Eight patients with baseline evidence of radiographic mNYC sacroiliitis were excluded from the predictive analysis. Of the rest (21/29), three patients developed mNYC AS at follow-up (2 HLA-B27 positive). Overall, the imaging arm of the criteria showed 100% sensitivity and 22% specificity, whereas the HLA-B27 arm had 67% sensitivity and 56% specificity. Ten of 21 patients (47.6%) had an increase in the radiographic sacroiliitis scores at 8 years, indicating radiographic progression, although not enough to fulfil the mNYC. Nine of these patients (90%) had an abnormal MRI at baseline and 7 (70%) were HLA-B27 positive.
Subjects who developed AS (n=3) tended to have a significantly greater number of SIJ quadrants with BMO on baseline MRI (mean (SD) per patient=23.5 (3.1)) as compared to those who did not progress onto AS (n=18) (mean (SD) per patient=8.0 (13.9)) (p=0.08) or controls (mean (SD) per patient=0.9 (2.3)) (p<0.0001). This was confirmed when patients who had radiographic mNYC (AS) at baseline (n=8) were included. At the group level, patients with radiographic AS had significantly higher scores of BMO at baseline compared to patients who did not have radiographic AS (at baseline and/or follow-up) (mean (SD)=24.214 vs 8 (13.9); p=0.005). Although it did not reach significance, patients who had any progression had higher scores of BMO (mean (SD)=21 (16.9) vs 10.6 (14.4); p=0.09).
A positive baseline MRI for SpA according to global assessment had 100% sensitivity for the subsequent radiographic mNYC sacroiliitis for three of the four readers and for concordant MRI data from any reader pair (table 4). Similarly, a positive baseline MRI for SpA according to the ASAS definition had 100% sensitivity for subsequent radiographic sacroiliitis in two out of four readers and for concordant MRI data from any reader pair. Specificity was considerably lower, particularly for baseline MRI scans considered positive according to the ASAS definition. This low specificity is due to the fact that half of the patients with clinical SpA who were recorded as having a positive baseline MRI for SpA by global assessment had still not developed radiographic mNYC sacroiliitis after 7.7 years of follow-up. Interestingly, on baseline MRI evaluation by any reader pair, there were 7/18 (39%) patients who had ≥5 SIJ quadrants with BMO and 6/18 (33.3%) patients who had ≥10 SIJ quadrants with BMO who had mild or no radiographic progression at follow-up at 7.7 years. When looking at these patients independently, three (two HLA-B27 positive and one HLA-B27 negative) had minor radiographic progression: one from grade 0 to unilateral grade 1 and two from grade 0 to unilateral grade 2. Two of these patients had received treatment with TNF blockers for a period of 2–3 years. The rest, including those who did not have any radiographic progression, had received treatment with non-steroidal anti-inflammatory drug agents only.
A major hindrance in the understanding of disease pathogenesis in AS has been the inability to achieve a timely or prompt diagnosis largely due to the reliance of the widely used diagnostic mNYC on the presence of advanced structural damage at the SIJs as shown on conventional radiography.11 The recently developed ASAS classification criteria represent a major step forward by allowing two different routes for patient classification. This is based on the presence of HLA-B27 positivity and ≥2 SpA features or ≥1 SpA feature and the presence of sacroiliitis as depicted by either radiographs or MRI as an imaging criterion. This is crucial as it is well documented that MRI is the most sensitive tool to identifying active inflammatory change occurring at the SIJ early on in the disease process.8 14
The availability of TNFα blocking agents and the expectation that these lead to an arrest in disease progression if administered early raise the issue of identifying patients who are likely to have a progressive disease. The ideal set of diagnostic criteria should therefore identify, with a high sensitivity and specificity, which patients with IBP have SpA and those whose disease will likely to evolve rapidly into AS since this represents the clinically challenging group in whom effective treatment ought to be targeted early in the disease course. The present study shows that in patients presenting with early IBP, the new ASAS classification criteria have a useful diagnostic value for both arms using the physician expert opinion as the gold standard comparison, as all patients fulfilled at least one arm of the criteria. Our results show that the imaging arm appeared more sensitive (83% vs 62%), suggesting the important role of MRI in diagnosis. Although the full validation of the ASAS criteria will not be possible until the longitudinal follow-up of the initial study cohort is completed,5 6 here we test for the first time the predictive value of both arms by applying it retrospectively in our patients with early IBP followed up for 8 years. Of the patients included in this paper, only three developed new AS according to mNYC, whereas a further eight had some radiographic progression of sacroiliitis. In this analysis, neither the ASAS classification criteria nor the ASAS definition of a positive MRI appeared to be of prognostic utility. The reasons for this are multifactorial, although the available follow-up time, which may be insufficient, could be an important contributing factor. Indeed, up to 35% of the patients appeared to have some radiographic progression in the follow-up time, although this was not enough to fulfil the mNYC definition that requires structural abnormalities to be at least grade 2 bilaterally or grade 3 unilaterally.11 Longer follow-up of this cohort and of other larger inception cohorts is needed to confirm these findings since it is expected that some individuals may not progress as rapidly as others.
According to the ASAS recommendation for defining an MRI as being positive for sacroiliitis, this requires the presence of either subchondral BMO on STIR or osteitis on postgadolinium T1FS that is highly suggestive of sacroiliitis.7 However, this definition does not take into account the severity of MRI findings, and mild BMO is therefore also used to diagnose sacroiliitis. By using this definition, all patients could be classified as having axial SpA at baseline, providing a sensitive tool but at the same time limiting its prognostic value. We have previously demonstrated that mild BMO on MRI had a sensitivity of 23% and a specificity of 38% for predicting radiographic sacroiliitis, whereas the combination of severe BMO and HLA-B27 positivity increased sensitivity to 62% and specificity to 92%.10 Furthermore, the presence of severe BMO independent of HLA-B27 had the same prognostic value.10 The current study failed to demonstrate the prognostic value of both arms of the ASAS classification criteria for the future development of radiographic AS, where mild MRI changes are included as a part of the imaging arm.10 Nevertheless, an important observation from our data is that a subgroup of patients with significant BMO of the SIJ affecting ≥10 SIJ quadrants did not progress radiographically after 8 years of follow-up, pointing to an important role for additional prognostic factors that have an influence independent of MRI.
In this study, semiquantitative assessment of MRI findings was restricted to inflammatory changes of BMO and structural changes were excluded. Nevertheless, structural changes seen on T1 images such as erosions or fat infiltration may represent a significant pathological process since they can occur as the result of previous inflammation. Recent studies suggest that structural lesions do have diagnostic relevance and can increase sensitivity without impairing specificity.15 Furthermore, structural damage may already be detectable earlier on in the disease process, and their value as part of the definition of a positive MR signal needs to be reappraised.16 Indeed, our analysis shows that the global assessment proposal has better specificity than the ASAS definition for the prediction of mNYC sacroiliitis, which may be due to the fact that it allows for the subjective concomitant assessment of factors such as the presence of structural damage or the severity of the BMO lesion.
In conclusion, both arms of the ASAS classification criteria have good diagnostic utility in this cohort of patients with early SpA, although they appear to be of limited utility for the prediction of radiographic progression including the development of mNYC AS. This may be due to the inclusion of mild BMO in the current ASAS definition for a positive MRI and an important role for additional prognostic factors that are independent of MRI. Our findings suggest that a full reappraisal of the definition of what constitutes a positive MRI signal for sacroiliitis in SpA, which has both diagnostic and prognostic relevance, is needed. This may need to incorporate calibration of the severity of the BMO signal and the inclusion of structural changes.
ASAS, Assessment of SpondyloArthritis international Society; LR+, positive likelihood ratio; LR−, negative likelihood ratio; mNYC, modified New York criteria; STIR, short τ inversion recovery; T1SE, T1-weighted spin echo.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the Local Ethics Committee, Leeds.
Provenance and peer review Not commissioned; externally peer reviewed.
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.