Article Text

Download PDFPDF

Diagnostic evaluation of the sacroiliac joints for axial spondyloarthritis: should MRI replace radiography?
  1. Denis Poddubnyy1,2,
  2. Torsten Diekhoff3,
  3. Xenofon Baraliakos4,
  4. Kay Geert A Hermann3,
  5. Joachim Sieper1
  1. 1 Department of Gastroenterology, Infectiology and Rheumatology (including Nutrition Medicine), Charité - Universitätsmedizin Berlin, Berlin, Germany
  2. 2 Epidemiology Unit, German Rheumatism Research Centre, Berlin, Germany
  3. 3 Department of Radiology, Charité - Universitätsmedizin Berlin, Berlin, Germany
  4. 4 Rheumazentrum Ruhrgebiet, Ruhr University Bochum, Herne, Germany
  1. Correspondence to Professor Denis Poddubnyy, Department of Gastroenterology, Infectiology and Rheumatology (including Nutrition Medicine), Charité - Universitätsmedizin Berlin, 12203 Berlin, Berlin, Germany; Denis.Poddubnyy{at}charite.de

Abstract

The possibility of detection of structural damage on magnetic resonance imaging (MRI) of sacroiliac joints raises the question of whether MRI can substitute radiographs for diagnostic evaluation and to a further extent for classification of axial spondyloarthritis (axSpA). In this viewpoint, we will argue that it is time to replace conventional radiographs with MRI for the assessment of structural changes in sacroiliac joints. This message is based on current data on the following questions: (1) How reliable are conventional radiographs in the diagnosis of axSpA overall and radiographic axSpA in particular? (2) How does T1-weighted MRI compare to radiographs in the detection of sacroiliitis? (3) Are there now other (better) MRI sequences than T1-weighted, which might be more suitable for the detection of structural lesions? (4) Which MRI sequences should be performed for the diagnostic evaluation of the sacroiliac joints? (5) Do we have data to define sacroiliitis based on structural changes detected by MRI?

  • Spondyloarthritis
  • Spondylitis, Ankylosing
  • Magnetic Resonance Imaging

Statistics from Altmetric.com

Request Permissions

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.

Axial spondyloarthritis (axSpA) is a chronic inflammatory disease predominantly of the axial skeleton, but peripheral joints and entheses can also be affected. The condition usually starts in the sacroiliac (SI) joint, and isolated spinal involvement without affection of the SI joint is rare.1 In a study of patients with chronic low back pain and maximal symptom duration of 3 years, only 1%–2% of patients with axSpA had spondyloarthritis (SpA)-typical lesions in the spine without affection of the SI joints.2 This is the reason why in the Assessment of SpondyloArthritis International Society (ASAS) classification criteria of axSpA3 the presence of sacroiliitis on imaging took one of the central roles.

Like in any chronic inflammatory disease, axSpA has an early phase with bony inflammation and a later stage dominated by structural damage of the bone.1 Since the publication of the modified New York criteria for classifying ankylosing spondylitis (AS) in 1984,4 the detection of radiographic sacroiliitis became the cornerstone of the AS diagnosis. In the 1990s of the last century, it became evident that MRI is able to detect active inflammation early in the course of axSpA. Thus, in addition to radiographic sacroiliitis, active inflammation seen on MRI in fluid-sensitive, fat-suppressed pulse sequences such as short-tau inversion recovery (STIR) or T2 with fat saturation was added in the ASAS classification criteria for SpA in 2009.3 Also, the ASAS–European Alliance of Associations for Rheumatology (EULAR)5 and the American College of Rheumatology (ACR)–Spondyloarthritis Research and Therapy Network (SPARTAN)–Spondylitis Association of America (SAA)6 management recommendations for axSpA implemented the presence of sacroiliitis on MRI (as an alternative to radiographic sacroiliitis) as one of the starting points in the treatment algorithms.

The detection of active inflammation in the SI joints, that is, subchondral bone marrow oedema, is the major advantage of MRI. Especially in the early phase of the disease, it is of high diagnostic relevance and an important parameter of disease activity. MRI is able, however, to depict not only active inflammation but also structural lesions such as fat lesions of bone marrow, erosion, fat metaplasia in an erosion cavity or backfill, sclerosis and ankylosis. Both active and structural MRI lesions typical of axSpA seem to be relevant for the diagnosis, and active lesions should be interpreted in the context of structural damage.7 8 Indeed, subchondral bone marrow oedema without structural changes as detected by MRI seems to be less frequent in axSpA than previously thought.8–11

The possibility of detection of structural damage on MRI raises, therefore, the question of whether MRI can substitute radiographs in the diagnosis of axSpA and to a further extent in classification. In many countries and in many patients with possible axSpA, MRI is performed anyway as a part of a routine diagnostic approach that makes this question increasingly clinically relevant. At the same time, discussion of this question would not exclude radiography in a diagnostic process if MRI were unavailable, unfeasible or its application limited by its costs.

Here we will argue that it is time to replace conventional radiographs with MRI for the assessment of structural SI joint lesions when available. Our message is based on current data on the following questions: (1) how reliable are conventional radiographs in the diagnosis of axSpA overall and radiographic axSpA in particular? (2) how does T1-weighted MRI compare to radiographs in the detection of sacroiliitis? (3) are there now other (better) MRI sequences than T1-weighted, which might be more suitable for the detection of structural lesions? (4) which MRI sequences should be performed for the diagnostic evaluation of the SI joints? (5) do we have data to define sacroiliitis based on structural changes detected by MRI?

How reliable are conventional radiographs in the diagnosis of (radiographic) axSpA?

Radiographic findings of the SI joints are difficult to interpret because the pelvic anatomy is complex, the SI joints have an oblique orientation, and superposition of bowel gas can hide or mimic structural bone changes. Therefore, the reliability of positive or negative findings might be better with cross-sectional imaging techniques such as MRI. An earlier study published in 2003 investigated the performance of radiologists (23 participants) and rheumatologists (100 participants) in detecting sacroiliitis12 in The Netherlands. The whole group performed a training in the interpretation of SI joint radiography and met again after 3 months. Gold standard (radiographic sacroiliitis yes or no) was defined by an expert panel. Sensitivity (84.3%/79.8%) and specificity (70.6%/74.7%) for radiologists and rheumatologists, respectively, were comparably low. These figures were similar at the second meeting after the training.12

In another study, there was a considerable difference reported when the results of local readings versus central readings of radiographs of SI joints in patients with axSpA with a disease duration less than 5 years (from the Devenir des Spondylarthropathies Indifférenciées Récentes (DESIR) cohort) were compared: 32 of 109 patients (29.4%) with bilateral obvious sacroiliitis or at least unilateral fusion (by local reading) were rated as negative by central reading and 68 of 579 patients (11.7%) were rated as positive by central reading, although the interpretation of the local readers was negative. Also, the inter-reader agreement even between the central readers was only moderate (kappa=0.54).13 Thus, applying SI joint radiographs for diagnosing axSpA is rather unreliable.

The problem becomes even larger at the early disease stage or in patients with suspected axSpA14 that is related to the definition of radiographic sacroiliitis grade 1 (suspicious changes leaving a big room for subjective interpretation) or grade 2 (minimal abnormalities) as compared with sacroiliitis grade 3 (advanced changes with joint space alteration) or 4 (complete ankylosis of the joint).

How does T1-weighted MRI compare to conventional radiographs in the detection of sacroiliitis?

Two studies compared radiographs with T1-weighted MRI for the detection of structural lesion in the SI joints. The first one took the radiographs as the reference method and reported a sensitivity of 84% and a specificity of 64% for fulfilling the modified New York criteria, with similar figures for the detection of ‘chronic sacroiliitis’ (=overall structural damage).15 This study has also demonstrated that the reliability (inter-reader variability) of the detection of structural lesions (erosions, sclerosis and joint space alteration) was better for MRI than for the radiography.15 In an analysis of the DESIR study, radiographs and MRI T1-weighted images of the SI joints were analysed and compared by two trained readers. Twelve (reader 1) or 10 (reader 2) patients would have been classified as axSpA by radiographs but not by MRI, and 3 (reader 1) and 6 (reader 2) patients were negative in radiography but MRI positive.16 Thus, taken together in these two studies, there was a moderate to good agreement between radiographs and MRI for scoring structural lesions at the SI joints.

However, no CT as a gold standard for the structural damage detection was available in these studies. Therefore, based on these study data, the question on whether any of the two imaging methods for the detection of structural lesions in the SI joints would be superior to the other could not be answered. This question was addressed in the German SacroIliac Magnetic Resonance Computed Tomography (SIMACT) study.17 In this prospective study, 110 patients referred to the rheumatologist with chronic low back pain and possible axSpA were included. All patients underwent radiography, low-dose CT and MRI of the SI joints (a T1-weighted sequence was used for this analysis). All images were scored by three readers (radiologists), with the low-dose CT as standard of reference for the detection of structural SI joint lesions: erosions, joint space changes (including ankylosis) and sclerosis. MRI showed a better absolute agreements with CT compared with radiography for erosion (88.2% vs 70.9%), joint space changes (92.7% vs 80.9%) and overall positivity for the presence of structural damage (89.1% vs 70.0%), but not for sclerosis (83.6% vs 86.4%, respectively). Furthermore, the reliability (inter-reader variability) of structural damage detection was better on MRI for erosions and joint space alteration, but not for sclerosis.17 It should be noted that the study was performed in a specialised SpA centre with experienced readers; therefore, the performance of the discussed imaging methods might be different in non-specialised centres.

In another study from China, it was shown again that CT has the best sensitivity and specificity for the detection of structural lesions in the SI joints, but closely followed by MRI in patients with non-radiographic axSpA who were, by definition, negative on conventional radiographs.18

In a most recent study, the German group compared the value of different imaging approaches of the SI joints for the diagnosis of axSpA, using the clinical diagnosis by experts as reference standard: conventional radiographs, CT, MRI (both STIR and T1), conventional radiographs+MRI and CT+MRI.10 The obtained results confirmed radiography to be inferior to MRI: radiographs showed with 66.3% the lowest sensitivity compared with MRI (82%) and CT (76.4%) and also a lower specificity of 67.6% vs MRI (86.5%) and CT (97.3%). Also, the inter-rater reliability was lowest for radiographs (kappa=0.517), followed by MRI (kappa=0.665) and CT (kappa=0.875). In daily clinical practice, MRI is often ordered after radiography in case of negative or equivocal results. Consequently, both imaging modalities are assessed simultaneously. In the referenced study, the combination of radiography with MRI could not outperform MRI alone. An analysis of different scenarios showed that the current clinical standard with MRI done only in patients with no definite radiographic sacroiliitis according to the modified New York criteria for AS had the highest sensitivity (86.5%) but poor specificity (66.2%), indicating a substantial risk of overdiagnosis. Raising the positivity threshold of radiographs to at least grade 3 unilaterally increased the specificity of the approach to 81.1% but decreased the sensitivity to 79.8%. A strategy with MRI as the only imaging method had a sensitivity of 82% and a specificity of 86.5%.10

However, these studies also demonstrated that MRI (with T1-weighted and STIR sequences only) is not as good as CT, especially for the differentiation between sclerosis and erosions, raising the question whether the MRI technique can be improved for this purpose.

Are there now other (better) MRI sequences than T1-weighted, which might be more suitable for the detection of structural lesions?

Technical progress is moving fast, and it can be expected that new interesting imaging methods will emerge over the coming years. However, in the context of this discussion, erosion-sensitive three-dimensional gradient echo (3D-GRE) MRI sequences (such as volumetric interpolated breath-hold examination (VIBE)) are of special interest that are available on nearly all currently installed MR machines. Figure 1 presents examples of 3D-GRE images in comparison to T1-weighted images, CT and conventional radiography.

Figure 1

Depiction of structural damage in the SI joints by conventional radiography, MRI and CT. Patient 1: 28-year-old male patient with r-axSpA with bilateral sclerosis (black arrows) and erosion (white arrows) in radiography, MRI and CT. T1-weighted MRI depicts also backfill (arrowheads). 3D-GRE (VIBE) shows erosion in more detail. Patient 2: 46-year-old male patient with sclerosis (black arrows), possible erosion (white arrow) and suspected partial ankylosis (black arrowheads) in radiography. However, MRI and CT do not show structural damage on the SI joint besides osteophyte formation that causes superposition in radiography (open arrowheads). 3D-GRE, three-dimensional gradient echo; r-axSpA, radiographic axial spondyloarthritis; SI, sacroiliac; VIBE, volumetric interpolated breath-hold examination.

Two studies have been published comparing 3D-GRE and the T1-weighted MRI sequences, focusing on the scoring of erosions and using again low-dose CT as a gold standard. In the first study, patients from the SIMACT trial (110 subjects) were analysed by adding the 3D-GRE images.19 Reading of the images was done by two experienced radiologists. The 3D-GRE sequence had better sensitivity than the T1-weighted one (95% vs 79%, respectively) with a similar specificity (93% each). 3D-GRE identified approximately 20% more patients with erosion than the T1, 36 vs 30 of 38 patients with positive low-dose CT findings.

In the second study with a comparable study design, the sensitivity of 3D-GRE was again higher than that for T1-weighted MRI (71.2% vs 63.4%, respectively) with a similar specificity of 87.3% vs 88%, respectively.20 Thus, based on these data, the 3D-GRE sequence comes close to CT in the detection of erosions in the SI joints and is superior to the T1-weighted sequence. However, for the detection of fat lesions and fat metaplasia in an erosion cavity representing repair fibrous tissue (also called backfill), the T1-weighted sequence is still indispensable.

While the strengths of 3D-GRE lie in a short acquisition time, thin slices and good contrast of erosion and bone, it still cannot depict the cortical bony surface directly, in contrast to CT. Furthermore, clear-cut definitions for erosions and their positivity in the clinical context are warranted to avoid overcalling of minor physiological or degenerative irregularities of the cortical bone and their misinterpretation as erosions related to axSpA.

Which MRI sequences should be performed for the diagnostic evaluation of the SI joints?

Four sequences for MRI of the SI joints have been recommended by the radiologists of the European Society of Skeletal Radiology Arthritis Subcommittee in 201521: (1) semicoronal oblique T1-weighted (for fat lesions, erosions and ankylosis); (2) semicoronal STIR (or another T2-weighted sequence with suppressed fat signal, for bone marrow oedema); (3) semicoronal cartilage (erosion sensitive) sequence (such as 3D-GRE/VIBE); and (4) a second T2-weighted semiaxial sequence with suppressed fat signal, also for bone marrow oedema. These sequences are also included in the international consensus developed by ASAS and SPARTAN that has been recently presented at the Annual European Congress of Rheumatology–EULAR 2022 (POS0989). Based on the evidence presented earlier, this would currently be the mandatory MRI sequence protocol for the detection of bone marrow oedema and structural lesions of the SI joints for the diagnosis of axSpA. Interdisciplinary work involving both radiologists and rheumatologists would be necessary to overcome barriers hampering the implementation of these recommendation in daily clinical practice.

Do we have enough data to define sacroiliitis based on structural changes detected by MRI?

The current ASAS definition of a ‘positive MRI of SI joints’ was intended for use as part of the ASAS classification criteria only, thus, not for diagnosis. Therefore, even the latest definition update relies on the presence of bone marrow oedema that should be highly suggestive of axSpA.7 However, structural damage is considered important contextual information for the interpretation of whether bone marrow oedema is suggestive of axSpA or not. In the clinical setting, when making the diagnosis, structural damage on MRI could be the leading finding resulting in a diagnosis of axSpA even if active inflammation was absent at the time point of the investigation. It is currently under debate if structural damage detected by MRI could substitute structural damage detected by conventional radiographs also in the context of classification criteria.

A systematic literature research on the performance of MRI in the diagnosis of axSpA was published in 2019 but included publications only until March 2017.22 Many studies have been performed in the last decade and many after this deadline. For example, in the study by Weber et al from 2015, erosions in >2 SI-joint quadrant performed best,23 and in the study by Baraliakos et al from 2021 analysing STIR and T1-weighted sequences, the combination of bone marrow oedema and erosions had the highest predictive value for the diagnosis of axSpA.11

In the original definition of a positive SI joint MRI by ASAS from 200924 and in the updated definition from 2016,25 structural lesions of the SI joints such as erosion, fat lesion, sclerosis and ankylosis were described, but it was felt by the ASAS group that more data were needed for a definition of a structurally positive or negative MRI. The same was the case for the EULAR recommendation for the use of imaging in the diagnosis and management of SpA in clinical practice.26 However, the ASAS MRI working group—including most of the researchers in this field, both rheumatologists and radiologists—published then in 2019 the following definitions for MRI SI joints lesions reflecting structural change: erosion is defined as a defect in subchondral bone associated with full-thickness loss of the dark appearance of the subchondral cortex at its expected location, with loss of signal on a T1-weighted non-fat suppressed sequence compared with the normal bright appearance of adjacent bone marrow. Moreover the definition for fat lesion was as follows: bright signal on a T1-weighted non-fat-suppressed sequence that is brighter than normal bone marrow, which meets the following requirement: (1) homogeneously bright, (2) located in a typical anatomical area (subchondral bone) and (3) sharply defined along its non-articular border with normal bone marrow.7 In a follow-up publication by the working group, structural damage on MRI T1-weighted of SI joints typical of SpA was defined as an ASAS-defined erosion (see aforementioned) in ≥3 SI joint quadrants or at the same location in ≥2 consecutive slices or by an ASAS-defined fat lesion in ≥5 SI joint quadrants or in ≥3 consecutive slices.8 In the latter study, in contrast to the majority of other publications, the positive predictive validity of the MRI definitions for a diagnosis of axSpA has been analysed after a mean follow-up of 4–5 years that strengthened the main study outcome: the diagnosis of axSpA.8

Another interesting finding of possible relevance in this context was reported recently when investigating which structural SI lesions on CT performed best for the diagnosis of axSpA.27 For this, the SI joint was segmented in ventral, middle and dorsal parts, and it was found that the presence of erosion and ankylosis differentiated best between axSpA and non-SpA; however, this was the case only for the middle and dorsal parts and not for the ventral regions of the joint that are typically prone to mechanical stress. Similar findings had been reported by the group previously for the location of bone marrow oedema, especially for the differential diagnosis of axSpA with osteitis condensans ilii.28 Thus, it should be further clarified whether these findings should be included for a positive definition of structural MRI lesions in the SI joint. Sclerosis was frequently present both in patients with axSpA and in patients without SpA in this study and did not differentiate between the two groups. Sclerosis was also not selected as a differentiating feature between axSpA and non-SpA by the ASAS MRI working group.7 8 In the dorsal part of the SI joint, the ligament insertion might result into a physiological variation of the contour of the cortical bone that should be taken into account when interpreting erosions in this location.

Based on the data presented here, structural SI joint lesions should be evaluated by MRI in T1-weighted and 3D-GRE sequences, while conventional radiographs should only be used if MRI is not available. Conventional radiography is neither specific nor sensitive for the diagnosis of axSpA and has a low rate of agreement among readers. (Low-dose) CT has good sensitivity and specificity for structural lesions; however, the radiation exposure (even if low-dose CT is comparable to that of conventional radiographs) and its lack for depicting bone marrow changes (ie, oedema and fat) does not recommend its use as first-line imaging. ASAS, together with SPARTAN, has embarked on a study for the re-evaluation of the ASAS classification criteria for axSpA: 1000 patients referred to a rheumatologist because of possible axSpA worldwide are being enrolled and undergo a thorough clinical, laboratory and imaging (MRI and radiographs in all patients) investigation. It is likely that recruitment will be finished in 2022. Whether it will be necessary to use these data for some fine-tuning of the recommendations made above or whether the available data we have now are already sufficient (yes, in the opinion of the authors) to redefine the role of radiographs and MRI in the diagnosis (and potentially also classification) of axSpA should be decided soon. There is a major clinical demand for this. The past decade has seen a continuous discussion and training regarding diagnosis and differential diagnosis of bone marrow oedema of the SI joints by STIR MRI. This training should now also include evaluation of structural lesions on MRI.

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.

References

Footnotes

  • Handling editor Josef S Smolen

  • Twitter @rheumaimaging

  • Contributors All authors contributed to the conception of the work, drafting and revising the manuscript and approved the final versions.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests DP: research support from AbbVie, Eli Lilly, MSD, Novartis and Pfizer; consulting fees from AbbVie, Biocad, Eli Lilly, Galapagos, Gilead, GlaxoSmithKline, Janssen, MSD, Moonlake, Novartis, Pfizer, Samsung Bioepis and UCB; speaker fees from AbbVie, Bristol-Myers Squibb, Eli Lilly, Janssen, MSD, Medscape, Novartis, Peervoice, Pfizer and UCB. TD: research grants to the institution from Canon MS and the Berlin Institute of Health; consulting and lecture fees from Eli Lilly, Novartis, MSD, Roche and Canon MS. XB: grants from Abbvie, BMS, Eli-Lilly, Galapagos, Janssen, MSD, Novartis, Pfizer, Roche, Sandoz, Sanofi and UCB; consulting fees from Abbvie, BMS, Eli-Lilly, Galapagos, Janssen, MSD, Novartis, Pfizer, Roche, Sandoz, Sanofi and UCB; honoraria from Abbvie, BMS, Eli-Lilly, Galapagos, Janssen, MSD, Novartis, Pfizer, Roche, Sandoz, Sanofi and UCB; meeting attendance support from Abbvie, BMS, Eli-Lilly, Galapagos, Janssen, MSD, Novartis, Pfizer, Roche, Sandoz, Sanofi and UCB; leadership roles as editorial board member of Annals of Rheumatic Diseases, ASAS president. KGAH: consulting fees from AbbVie and lecture honoraria from Novartis, Pfizer and MSD; cofounder of BerlinFlame GmbH. JS: honoraria for consultancy or being a member of a speaker’s bureau from Abbvie, Novartis, MSD and UCB.

  • Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.