Objectives Investigating the utility of adding structural lesions seen on MRI of the sacroiliac joints to the imaging criterion of the Assessment of SpondyloArthritis (ASAS) axial SpondyloArthritis (axSpA) criteria and the utility of replacement of radiographic sacroiliitis by structural lesions on MRI.
Methods Two well-calibrated readers scored MRI STIR (inflammation, MRI-SI), MRI T1-w images (structural lesions, MRI-SI-s) and radiographs of the sacroiliac joints (X-SI) of patients in the DEvenir des Spondyloarthrites Indifférenciées Récentes cohort (inflammatory back pain: ≥3 months, <3 years, age <50). A third reader adjudicated MRI-SI and X-SI discrepancies. Previously proposed cut-offs for a positive MRI-SI-s were used (based on <5% prevalence among no-SpA patients): erosions (E) ≥3, fatty lesions (FL) ≥3, E/FL ≥5. Patients were classified according to the ASAS axSpA criteria using the various definitions of MRI-SI-s.
Results Of the 582 patients included in this analysis, 418 fulfilled the ASAS axSpA criteria, of which 127 patients were modified New York (mNY) positive and 134 and 75 were MRI-SI-s positive (E/FL≥5) for readers 1 and 2, respectively. Agreement between mNY and MRI-SI-s (E/FL≥5) was moderate (reader 1: κ: 0.39; reader 2: κ: 0.44). Using the E/FL≥5 cut-off instead of mNY classification did not change in 478 (82.1%) and 469 (80.6%) patients for readers 1 and 2, respectively. Twelve (reader 1) or ten (reader 2) patients would not be classified as axSpA if only MRI-SI-s was performed (in the scenario of replacement of mNY), while three (reader 1) or six (reader 2) patients would be additionally classified as axSpA in both scenarios (replacement of mNY and addition of MRI-SI-s). Similar results were seen for the other cut-offs (E≥3, FL≥3).
Conclusions Structural lesions on MRI can be used reliably either as an addition to or as a substitute for radiographs in the ASAS axSpA classification of patients in our cohort of patients with short symptom duration.
- Magnetic Resonance Imaging
- Ankylosing Spondylitis
- Low Back Pain
Statistics from Altmetric.com
A commonly used imaging method for the assessment of sacroiliitis is conventional radiography. Radiographic sacroiliitis is obligatory in the modified New York (mNY) criteria for ankylosing spondylitis (AS).1 However, it has been shown that it is very difficult to reliably detect sacroiliitis on conventional radiographs. Substantial observer variation exists for both radiologists and rheumatologists, which can lead to substantial misclassification of patients.2 Moreover, it has been shown that training in the reading of these radiographs does not lead to improvement.3
Another downside of conventional radiography is that only structural damage in the sacroiliac joints can be detected, frequently appearing after several years of disease, which hampers early detection. MRI can reliably detect inflammation of the sacroiliac joints at an early stage before damage on radiographs can be detected.4 Sacroiliitis on MRI plays an important role in the imaging arm of the Assessment of SpondyloArthritis (ASAS) axial SpondyloArthritis (axSpA) criteria.5
Besides inflammatory lesions, structural lesions such as erosions, fatty lesions, sclerosis and ankylosis are visible on MRI. Therefore, MRI has great potential for the assessment of both active inflammatory lesions and structural damage by means of one single-imaging technique. In the recently published European League Against Rheumatism (EULAR) recommendations for the use of imaging in the diagnosis and management of axSpA in clinical practice, it was advocated to take structural lesions (such as bone erosion, fat infiltration, sclerosis and new bone formation) into account in addition to active inflammatory lesions.6 This is augmented by accruing worries about radiation exposure coupled with conventional radiographs in some parts of the world.
By this study, we aim to further clarify the role of these structural lesions and to investigate their usefulness with regard to the ASAS axSpA classification of patients. First, we determine the agreement between the presence of sacroiliitis on radiographs (mNY criteria) and structural lesions seen on MRI. Subsequently, we evaluate what impact the use of structural lesions on MRI-SI could have on the performance of the ASAS axSpA criteria. Two scenarios are investigated: the addition of structural lesions seen on MRI to the definition of ‘sacroiliitis on imaging’ (scenario 1) and the replacement of radiographic sacroiliitis by structural lesions on MRI (scenario 2). For both scenarios, the impact on the classification of patients according to the ASAS axSpA criteria is assessed.
Baseline data from the DEvenir des Spondyloarthrites Indifférenciées Récentes (DESIR) cohort were used for this analysis. The DESIR cohort has been extensively described before.7 In short, the DESIR cohort is a French prospective longitudinal cohort study following 708 patients (aged >18 years and <50 years) over time with inflammatory back pain (IBP) for ≥3 months and <3 years, located in the thoracic spine, lumbar spine and/or buttock area. IBP was defined according to either the Calin or Berlin criteria.8 ,9 Patients were included only in case of a suspicion of SpA, according to the rheumatologist defined as a score ≥5 on a 0–10 scale (0: not suggestive of axSpA and 10: very suggestive of axSpA). All patients underwent a full diagnostic work-up including MRI and conventional radiographs of the sacroiliac joints (MRI-SI and X-SI, respectively), HLA-B27 testing and the assessment of all other SpA-features, in agreement with the descriptions provided by ASAS.5 Inclusion was performed in 25 centres across France and took place between December 2007 and April 2010. The database for the baseline data used for this analysis was locked on 30 October 2012. The study fulfilled Good Clinical Practice Guidelines and was approved by the appropriate medical ethical committees. Patients gave informed consent before start of the study.
Imaging and scoring methods
MRI of the sacroiliac joints (MRI-SI) was performed at baseline, in each participating centre, on 1.0–1.5 T machines. The acquired sequences were coronal oblique T1-weighted FSE and STIR with 12–15 semi-coronal slices of 4 mm thickness, parallel to the long axis of the sacrum. Plain radiographs of the pelvis (X-SI) were performed at baseline, in anteroposterior view. All available MRI-SIs and X-SIs were read by two well-calibrated readers (MRI: FT and RvdB and X-SI: GL and RvdB) independently, in different reading sessions. The reading process was completely blinded: readers had no insight in patient characteristics and other clinical and imaging data. The training and calibration process of the different readers was extensively described before.2
Sacroiliitis on radiographs was assessed according to the mNY criteria; radiographic sacroiliitis was defined as bilateral grade ≥2 or unilateral grade ≥3.1 Regarding the presence of inflammatory lesions on MRI-SI, the ASAS definition for a positive MRI-SI was used. An MRI-SI was marked positive if one bone marrow oedema (BME) lesion highly suggestive of SpA was present on ≥2 consecutive slices or alternatively if several BME lesions highly suggestive of SpA were visible on a single slice.4 The presence of structural lesions on MRI-SI was assessed using different types of lesions: fatty lesions, erosions, sclerosis, (partial) ankylosis.
We used a scoring system with similarities to the methodology outlined in the SPARCC online training module, as described by Weber et al in 2010, to assess MRI-SI on the presence of structural lesions.10 In contrast to the SPARCC method, the scoring method we used is purely based on counting the number of structural lesions. Each SI joint is divided into quadrants. The presence of structural lesions was assessed (present vs absent) in each quadrant on six consecutive slices through the SI joints; starting on the slice on which at least 1 cm of vertical height of the cartilage compartment is visible, from anterior to posterior, assessing the cartilaginous compartment of the SI joints and the anteroinferior portion of the SI joint. Structural lesions were only considered present if seen on at least two consecutive slices, resulting in a maximum score of 40 per lesion (5 per quadrant (as a lesion needs to be visible on two consecutive slices)× 4 quadrants× 2 sacroiliac joints) except for (partial) ankylosis. It was considered sufficient if (partial) ankylosis was seen on a single slice and since ankylosis involves always a sacral and an iliac part, the maximum score results in 24. More information about the scoring method for structural lesions is given in supplementary table S1.
Structural lesions on MRI-SI were assessed by two readers, without the use of an adjudicator. Therefore, results were analysed separately per reader. In contrast, regarding the presence of sacroiliitis on radiographs (mNY) and inflammation on MRI-SI (ASAS definition) adjudicator scores (MR) were available.
Assessment of structural lesions
Recently, definitions for a positive MRI for structural lesions (MRI-SI-s) using different cut-offs of various structural lesions were proposed by de Hooge et al.11 This was done in the SPondyloArthritis Early Cohort (SPACE): a cohort similar to the DESIR cohort, but slightly different since patients with chronic back pain (CBP) were included and not only patients with IBP and, more importantly, in SPACE also patients without a high suspicion of SpA were included.
The cut-offs were based on ≤5% false positives whereby the false positives were defined as structural lesions among patients with a very low likelihood of axSpA.11 These cut-offs are as follows: ‘erosions ≥3’, ‘fatty lesions ≥3’, ‘fatty lesions and/or erosions ≥5’. Prevalence of sclerosis and ankylosis was so low that there was not a cut-off that could clearly distinguish between SpA and no-SpA patients. Therefore, these types of lesions were not taken into account any further.
The ASAS axSpA criteria were used for classification of patients. Patients were grouped based on how they fulfilled the criteria: via the imaging arm of the ASAS axSpA criteria (either by fulfilling mNY criteria and/or positive MRI); via the clinical arm of the ASAS axSpA criteria; or via both. If patients fulfilled several categories, they were classified as such, resulting in seven possible combinations (tables 3 and 4). Patients not fulfilling the ASAS axSpA criteria were grouped into the no-axSpA group. It is important to realise that patients that have multiple axSpA features present but formally not fulfil the ASAS axSpA criteria (since they are mNY-negative, MRI-negative and HLA-B27-negative) are also included in this group.
Disease characteristics of patients included in the DESIR cohort were presented using descriptive statistics. Agreement on the absence or presence of structural lesions using both imaging modalities (X-SI and MRI-SI-s) was assessed by cross-tabulation and expressed as Cohen's κ. In order to disregard subjects labelled as negative by both readers, which can make agreement look artificially high, percentage agreement on the positive cases (positive agreement) was calculated.12 Both measures of agreement were also calculated for the concordance between radiographic sacroiliitis on conventional radiographs and a positive MRI-SI-s (per individual reader). Radiographic sacroiliitis (present vs absent) was based on the mNY criteria and MRI-SI-s for structural lesions was based on the proposed cut-offs for a positive MRI-SI-s.
Subsequently, patients were classified according to the ASAS axSpA criteria, using the various definitions of MRI-SI-s. First, MRI-SI-s was added to the imaging criterion of the ASAS axSpA criteria as an additional possibility to fulfil the imaging arm (scenario 1). Second, the mNY criterion was replaced by MRI-SI-s (scenario 2, as if only an MRI was available). For patients that newly fulfilled the ASAS axSpA criteria in scenario 1 and whom ASAS axSpA classification changed by scenario 2, the probability of having axSpA was calculated by the positive likelihood ratio (LR+) product. This was done by multiplying the individual LRs of all present SpA features in a patient.13 For example, an LR product of 200 results in a positive predictive value of 90% in patients with CBP with an assumed pre-test disease prevalence of axSpA of 5%.14 The analyses were performed in STATA V.12.0.
In total, 582 patients with complete imaging data (both MRI-SI and X-SI present), evaluated by two readers, were included in this analysis. Disease characteristics are depicted in table 1. The mean age was 31.5 years (SD 7.2 years) and 51% were male. Patients had a mean duration of back pain of 18.2 months (SD 10.6 months) and 338 patients (58.1%) were HLA-B27 positive. Fulfilment of the ASAS axSpA criteria was seen in 418 (71.8%) patients (while using adjudicated scores and based on the original definition of imaging).
As published before, agreement between the two readers regarding the absence/presence of radiographic sacroiliitis (mNY) is moderate: κ 0.54.2 Regarding the definition of a positive MRI-SI according to the ASAS definition, inter-reader agreement was much better: κ 0.73.15
Inter-reader agreement between readers 1 and 2 was calculated for the various structural lesions definitions. The agreement on the presence of ≥3 erosions was poor: κ 0.19. For the presence of fatty lesions and the combination of fatty lesions/erosions, it was somewhat better, κ 0.50 and κ 0.44, respectively. Subsequently, agreement between radiographic sacroiliitis (mNY criteria) and a positive MRI-SI-s (using the various cut-offs) was assessed. Agreement was rather low and varied between κ 0.21 and κ 0.44 (table 2).
First, we investigated the effect of adding MRI-SI-s to the imaging criterion of the ASAS axSpA criteria (table 3). Using a cut-off value of 5 for the combination of ‘fatty lesions and/or erosions’, classification did not change in the majority of the patients: 556 patients (95.5%) and 522 patients (89.7%) for readers 1 and 2, respectively. Regarding the combination of ‘fatty lesions and/or erosions’, three or six patients (readers 1 and 2, respectively) would be additionally classified as axSpA if structural lesions on MRI were taken into account (two patients, identified by both readers). Only two of these in total seven patients presented with four SpA features in addition to IBP, which was present in all cases. None of these patients was HLA-B27 positive. The LR+ products varied between 3 and 968, respectively, corresponding to post-test probabilities of 13% and 98%. A post-test probability of >80% was seen in one patient only. Similar results were found when using a cut-off for ‘fatty lesions’ only and ‘erosions’ only (both cut-off values of 3) (see online supplementary data, S2 and S4).
More patients did not gain an ASAS classification by applying MRI-SI-s, but changed subgroups within the ASAS axSpA criteria. Regarding the combination of ‘fatty lesions and/or erosions’, 23 (reader 1) and 54 (reader 2) patients would be classified via different arms due to the presence of structural lesions on MRI-SI. Also, 16 or 40 patients (reader 1 or reader 2) were already classified via the imaging arm based on inflammatory lesions on MRI and also showed structural lesions on MRI (table 3) with or without fulfilment of the clinical arm. And, 7 or 14 other patients (reader 1 or reader 2) fulfilled the clinical arm only, but fulfilled the imaging arm too based on a positive MRI-SI-s. The same trends were seen when using a cut-off for ‘fatty lesions’ only and ‘erosions’ only (both cut-off values of 3) (see online supplementary data, S2 and S4).
Second, we assessed whether replacing radiographic sacroiliitis by structural lesions on MRI (table 4) had an impact on ASAS classification. Using the same cut-off values of 5 for the combination of ‘fatty lesions and/or erosions’, classification did not change in the majority of the patients (82.1% and 80.6% for readers 1 and 2, respectively). A similar result was seen using a cut-off value of 3 for ‘erosions’ and ‘fatty lesions’ only: classification did not change in 80.2% or 79.0% (readers 1 and 2) and 81.4% or 80.6% of the patients, respectively (table 5). Comparing results of the two readers, differences were seen among them, but these changes mainly involve shifts between the different arms within the ASAS axSpA criteria rather than changes in ASAS axSpA classification (yes/no) in general (table 5). Similar results were found when using a cut-off for ‘fatty lesions’ only and ‘erosions’ only (both cut-off values of 3) (see online supplementary data, S3 and S5).
In this scenario, of course, the same patients would be additionally classified as axSpA as in scenario 1. But in contrast to scenario 1, now patients can also no longer fulfil the ASAS axSpA criteria. Using the combination of ‘fatty lesions and/or erosions’, 12 and 10 patients (2.1% and 1.7% for readers 1 and 2, respectively) would not be classified axSpA anymore if radiographic sacroiliitis was replaced by structural lesions on MRI, for example, assuming that only an MRI was available. In total and between brackets for the individual readers separately, 89 (66+23) and 97 (43+54) patients changed arms within the criteria using the combined cut-off of ‘fatty lesions and/or erosions’ (table 4: yellow and orange boxes). However, all those patients were still classified axSpA.
The clinical phenotype of the patients that are no longer classified by the ASAS axSpA criteria was assessed. The same 10 patients were captured by both readers: all HLA-B27 and the number of present SpA features varied between 1 (only IBP) and 4 (IBP, peripheral arthritis, enthesitis, dactylitis). The LR+ products varied between 3 and 190, corresponding with disease probabilities of 13% and 90%, respectively. A post-test probability of >80% was seen in two patients. The two additional patients that no longer fulfilled the ASAS axSpA criteria by reader 1 only, showed both a good response to non-steroidal anti-inflammatory drugs and one presented with enthesitis as well (corresponding to disease probabilities of 45% and 74%). Similar results were found when using a cut-off for ‘fatty lesions’ only and ‘erosions’ only (see online supplementary data, S3 and S5).
When sacroiliitis on radiographs was replaced by structural lesions on MRI, only minor changes in the classification according to the ASAS axSpA criteria were seen. Most patients change from one subcategory to another subcategory, rather than becoming ASAS axSpA criteria positive or negative. This adds to the robustness of the ASAS axSpA criteria as a whole.
Based on these data, if a T1-sequence MRI is available, but a pelvic radiograph is lacking, this MRI may suffice and there is no reason of always obtaining additional radiographs. MRI can be therefore a reasonable alternative for radiographs since this prevents radiation exposure. More generally, our results are in line with the recent published EULAR recommendations for the use of imaging in the diagnosis and management of SpA in clinical practice, stating that on MRI both active inflammatory lesions (primarily BME) and structural lesions (such as bone erosions, new bone formation, sclerosis and fat infiltration) should be considered.6 Nevertheless, it is important to stress the difference between that publication and the current study: the EULAR recommendations are aimed for diagnosis and in the current study we look at the ASAS axSpA classification.
As described earlier, the recognition of radiographic sacroiliitis is challenging. Unfortunately, agreement on structural lesions was only fair–moderate as well. However, it is reassuring to see that the same conclusions (effects on ASAS axSpA classification) can be drawn while comparing the data of the individual readers. This strengthens our findings and adds to the validity of the criteria itself since the results seem not too much affected to inter-reader variation in this respect. This in contrast to the mNY criteria that immediately change in case of discrepant readings. We know from earlier studies that training does not improve recognition of radiographic sacroiliitis.3 The question whether training ameliorates recognition of structural lesions on MRI was not addressed in this study and could be relevant for future studies.
To our knowledge, the replacement of conventional radiographs by structural lesions on MRI in itself, and the effects on the ASAS axSpA classification have never been investigated before. However, data on the recognition and reliability of structural lesions visible on MRI-SI in general are available. In 2009, ASAS experts defined structural damage lesions of the sacroiliac joints on MRI. Subsequently, reliability of their detection has been studied. Weber et al16 found that erosions can be reliably detected on MRI to a comparable degree of reliability as BME (κ: 0.72). Regarding the detection of fatty lesions, lower κ values were shown in the same paper (κ: 0.55).16 More recently, the SPARCC SSS score was developed by the Canada–Denmark study group.17 Recent data from a collaboration between two research groups (from Edmonton, Canada, and Leiden, the Netherlands) have shown that the presence and extent of erosions and fat metaplasia can be reliably assessed by readers from these different centres (intraclass correlation coefficient (ICC): 0.60 and 0.65, respectively), even without calibration.18 Interobserver reliability for status scores was good for fat metaplasia (ICC: 0.71–0.78) and moderate to good for erosions (ICC: 0.58–0.62) for different reader pairs. In this study, we found a lower agreement compared with data from other studies. However, a clear comparison between reliability data of different groups is hampered by variability in study design such as inclusion criteria of the cohorts and slightly different scoring methods. Hypothesising about this, it could be plausible that patients with a short symptom duration, like in our early cohort, usually have less structural damage and that recognition of structural lesions is more challenging in a cohort of patients with early disease. Therefore, the agreement on positive imaging presented in this study might be slightly worse than could be expected in more established disease.
In the current study, we investigated the effect of using different combinations of structural lesions (fatty lesions, erosions and a combination of fatty lesions and erosions) on the classification of patients with axSpA. We could debate on the choice for the ideal cut-off and corresponding structural lesion. In this study, we found worse agreement for erosions alone and better agreement for the combination of erosions/fatty lesions and fatty lesions alone. Despite similar effects on the ASAS classification of patients, personally we prefer using a combination of fatty lesions and/or erosions instead of fatty reasons alone for reasons of face validity. But a definition based on erosions only could be a viable option too.
Although our data look promising, we should be cautious with generalising conclusions and results should be placed into perspective carefully. Before any decision can be taken on eventually leaving conventional radiographs behind, more evidence is needed since this is a decision with possibly far-reaching consequences. MRI is an expensive tool and might not be available throughout all parts of the world, which can lead to feasibility problems. Besides this, rheumatologists and radiologists worldwide are familiar with the mNY criteria. We should especially be cautious to generalise results of these data to patients with long-standing disease since we only tested this hypothesis in one cohort with early disease. The ideal cut-off could be different in patients with longer symptom duration and more structural lesions in the sacroiliac joints. Therefore, more data are warranted and confirmation of these results is needed first, especially in cohorts of patients with established disease.
Replacement of sacroiliitis on radiographs by structural lesions on MRI is resulting in a loss of a small number of patients (scenario 2). And these patients showed a low likelihood of axSpA based on the SpA features present. Nevertheless, for the reasons explained above, we favour (for the time being) addition of structural lesions on MRI to the ASAS axSpA classification (scenario 1) instead of replacement of conventional radiographs.
Strengths of our study are the intensive scoring process by multiple readers, which adds to the credibility of our findings and the cohort itself. The DESIR cohort is a cohort of early disease, with a substantial number of patients. In general, the urge for sensitive and specific imaging tools is an emerging issue in the field of early axSpA. Besides strengths of the study, we would like to address some limitations. An important limitation of this study is the lack of a gold standard to assess structural changes in the sacroiliac joint by means of CT. Another limitation of this study is that only patients with a short disease duration are included in the DESIR cohort and that we in general see a limited number of patients with structural lesions. Therefore, hesitation is needed in order to prevent drawing definite conclusions about the use of MRI-SI-s for the definition of sacroiliitis on imaging in the ASAS axSpA criteria.
To conclude, assessment of structural lesions on MRI instead of or in addition to conventional radiographs does not lead to a different ASAS axSpA classification in the large majority of patients with IBP. Although these findings are promising, replication in other cohorts is awaited.
The DESIR cohort is conducted under the control of Assistance Publique-Hopitaux de Paris via the Clinical Research Unit Paris-Centre and under the umbrella of the French Society of Rheumatology and Institut National de la Santé et de la Recherche Médicale. The database management is performed within the department of epidemiology and biostatistics (Professor Jean-Pierre Daurès, D.I.M., Nîmes, France). An unrestricted grant from Wyeth Pharmaceuticals was allocated for the first 5 years of the follow-up of the recruited patients. We also wish to thank the different regional participating centres: Pr Maxime Dougados (Paris—Cochin B), Pr André Kahan (Paris—Cochin A), Pr Olivier Meyer (Paris—Bichat), Pr Pierre Bourgeois (Paris—La Pitié-Salpetrière), Pr Francis Berenbaum (Paris—Saint Antoine), Pr Pascal Claudepierre (Créteil), Pr Maxime Breban (Boulogne Billancourt), Dr Bernadette Saint-Marcoux (Aulnay-sous-Bois), Pr Philippe Goupille (Tours), Pr Jean-Francis Maillefert (Dijon), Dr Xavier Puéchal (Le Mans), Pr Daniel Wendling (Besançon), Pr Bernard Combe (Montpellier), Pr Liana Euller-Ziegler (Nice), Pr Philippe Orcel (Paris—Lariboisière), Pr Pierre Lafforgue (Marseille), Dr Patrick Boumier (Amiens), Pr Jean-Michel Ristori (Clermont-Ferrand), Dr Nadia Mehsen (Bordeaux), Pr Damien Loeuille (Nancy), Pr René-Marc Flipo (Lille), Pr Alain Saraux (Brest), Pr Corinne Miceli (Le Kremlin Bicêtre), Pr Alain Cantagrel (Toulouse), Pr Olivier Vittecoq (Rouen).
Handling editor Tore K Kvien
Contributors PB performed the statistical analyses and writing of the manuscript. DvdH gave methodological advice and supervised all procedures. All authors contributed to the acquisition and interpretation of data, read, revised and approved the final manuscript.
Funding The DESIR cohort is financially supported by unrestricted grants from both the French Society of Rheumatology and Pfizer France. Neither funding source had any role in designing the study; collecting, analysing or interpreting the data; writing the manuscript; or deciding to submit the manuscript for publication.
Competing interests None declared.
Ethics approval Local medical ethical committees (of all participating centres).
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.