Abstract
Objective To assess the effect of increasing age on the frequency of inflammatory and structural magnetic resonance imaging (MRI) lesions in the sacroiliac joints (SIJ) in 3 independent cohorts of healthy individuals and patients with nonspecific back pain (NSBP).
Methods We assessed MRI SIJ lesions in 3 cohorts (A, B, and C) of healthy individuals (cohort A, n = 79; cohort B, n = 78) and patients with NSBP (cohort A, n = 87; cohort C, n = 46) aged ≤ 45 years referred with back pain suspicious of axial spondyloarthritis (axSpA). MRI lesions were recorded on consecutive slices in SIJ quadrants or halves through the cartilaginous SIJ. Lesions were ascertained by 2-7 central readers according to standardized lesion definitions. Lesions recorded concordantly by the majority of readers were analyzed according to age categories (18-29, 30-39, and 40-50 yrs) and previously reported data-driven MRI cutoffs indicative of inflammatory or structural lesions typical of axSpA.
Results Only 3.8% (in both cohort A and cohort B) of healthy individuals and 5.7% (cohort A) and 4.3% (cohort C) of patients with NSBP had erosion in ≥ 1 SIJ quadrant, and progressive increases of erosion with age categories were not evident. None of the healthy individuals and 2.3% and 4.3% of cohort A and cohort C, respectively, of the patients with NSBP showed erosion in ≥ 3 SIJ quadrants, the cutoff indicative of axSpA; not a single individual met this cutoff in the highest age category. Fat metaplasia was slightly increased with age among healthy individuals and patients with NSBP in cohort A, but not in cohorts B or C.
Conclusion SIJ MRI data from healthy individuals and NSBP controls did not indicate progressive increases in structural lesions with increasing age categories when standardized definitions for axSpA lesions were adopted. MRI cutoffs for structural lesions denoting axSpA discriminated equally well between axSpA and NSBP across all age categories.
Previous controlled studies of patients with axial spondyloarthritis (axSpA) have demonstrated that certain magnetic resonance imaging (MRI) structural lesions in the sacroiliac joints (SIJ), notably erosions, have high specificity for axSpA, especially when compared to bone marrow edema (BME).1,2 Contextual assessment of concomitant structural and active MRI lesions is key to enhance diagnostic utility of SIJ MRI. BME—probably strain-related—was demonstrated in up to one-third of healthy subjects, patients with mechanical back pain, athletes, and postpartum patients, whereas erosion as a highly specific lesion type could reliably be detected in 60% to 90% of patients with incipient axSpA as soon as 2 years after symptom onset.1,2 However, a previous study has reported a progressive increase in MRI structural lesions as assessed by 2 readers in a sample of 95 healthy individuals with increasing age categories; erosions were detected in 13.9% of healthy subjects aged 20 to 29 years and in as many as 39.3% aged ≥ 40 years.3 If replicated in other cohorts, this could have a major effect on the diagnostic interpretation of MRI lesions in the SIJ.
We aimed to assess the effect of increasing age categories on the frequency of the spectrum of inflammatory and structural MRI lesions in the SIJ in 3 independent cohorts of healthy individuals without back pain and patients diagnosed with nonspecific back pain (NSBP) assessed by 2 to 7 readers per cohort.
METHODS
Subjects. The analysis was based on SIJ MRI lesions of control subjects in 3 independent cohorts published between 2009 and 2019.4-7 The 3 prospective cohorts comprised consecutively recruited patients with nonradiographic axSpA (nr-axSpA) or radiographic axSpA (r-axSpA) along with patients with NSBP and/or healthy volunteers, who served as controls. The main objective of 2 of these cohorts was to evaluate the diagnostic utility of MRI in axSpA (MORPHO and Scientific Investigation of MRI and Biochemical Markers in Patients With Axial Spondyloarthritis, Back Pain of Other Reasons, Subjects With Strain in the Sacroiliac Joints and Healthy Subjects [MASH]), whereas the third cohort (Assessment of SpondyloArthritis international Society classification cohort [ASAS-CC]) was designed to develop and validate the ASAS classification criteria for axSpA.4-7 Evaluation of the SIJ MRI was standardized in each of these cohorts and included both a global contextual assessment and a granular assessment of individual lesion types. All MRIs in the 3 cohorts were read independently in random order by 2-7 readers blinded to demographic and clinical characteristics and to the classification of the study subjects. The control group of patients with NSBP and healthy volunteers served to determine the specificity of MRI findings. The study design of all 3 cohorts included standardization of technique for acquiring and reading MRI, application of consensus definitions for MRI lesions in the SIJ, development of consensus reference images of MRI lesions in the SIJ according to these definitions, use of a standardized calibration/training module by readers at different sites, and use of a customized online data entry module based on a standardized approach to recording MRI lesions in the SIJ.
Study cohorts
• Cohort A: MORPHO cohort part 1 and part 2. Comprised 79 healthy individuals and 87 patients with NSBP aged ≤ 45 (MORPHO part 1)4 and ≤ 50 years (MORPHO part 2)5 recruited in Edmonton, Canada, and Zurich, Switzerland. The absence of back pain in healthy individuals was confirmed using the Nordic Questionnaire8 and the absence of clinical features indicative of axSpA by standardized questionnaires; patients with back pain referred for assessment were classified as having NSBP based on clinical grounds and pelvic radiography. The recruitment/readout periods were 2007-2009 and 2010-2012 for MORPHO parts 1 and 2, respectively.
• Cohort B: MASH cohort. Comprised 78 healthy individuals aged 18-45 years recruited from the Copenhagen area—including among the staff of a university hospital in Copenhagen, Denmark—consisting of hospital cleaning staff (n = 26), long-distance runners (n = 23), and healthy male individuals (n = 29), defined by the Nordic Questionnaire and absence of clinical features indicative of axSpA.6 Further, 25 patients with compressive radiculopathy of ≥ 2 months by lumbar disc herniation confirmed on spinal MRI were also included. The recruitment/readout period was 2013-2016.
• Cohort C: ASAS-CC. Comprised 237 patients with SIJ MRI available in electronic format or as hardcopy images, aged ≤ 45 years, and referred to a rheumatologist with undiagnosed back pain of > 3 months duration suspicious of axSpA, of whom 69 were classified as NSBP upon comprehensive assessment by local ASAS experts; 46/69 patients with NSBP had complete electronic SIJ MRI scans available.7 Recruitment and readout periods were 2006-2008 and 2017-2018, respectively.
MRI protocol and evaluation of SIJ MRI scans. Subjects with complete SIJ MRI scans in electronic format consisting of semicoronal T1-weighted spin echo (T1SE) and semicoronal short-tau inversion recovery (STIR) sequences were enrolled. MRI scans of the 3 independent cohorts were acquired on devices of different brands with similar technical MRI variables across study sites and scanner manufacturers.
Senior radiology and rheumatology readers blinded to subject identifiers and diagnosis independently evaluated SIJ MRI on electronic workstations for the 3 cohorts separately. SIJ MRI of patients with axSpA of various symptom duration and disease activity were read in random order concomitantly with scans from non-axSpA control subjects. MRI lesions were assessed on an online interface and recorded in SIJ quadrants (for BME, erosion, and fat metaplasia) or halves (for ankylosis) on consecutive semicoronal slices through the cartilaginous SIJ compartment. SIJ quadrant cutoffs refer to the total affected quadrants across all slices. Lesions were ascertained by 5 and 4 central readers for MORPHO parts 1 and 2, respectively, 2 central readers for MASH, and 7 central readers for ASAS-CC, according to standardized lesion definitions on SIJ MRI. The MORPHO module4 was used for the MORPHO cohort, Spondyloarthritis Research Consortium of Canada system9 for the MASH cohort, and ASAS definitions10 for ASAS-CC (Supplementary Text S1, available with the online version of this article). Pretest reader calibration was performed in all 3 cohorts. Lesion definitions and reference image sets used in MORPHO parts 1 and 2 and MASH were only slightly modified to generate the ASAS definitions to guide ASAS-CC readers.
Data extraction from the original datasheets of the 3 cohorts. The detailed quadrant (joint half for ankylosis) scoring data for each subject and for each individual reader of MORPHO parts 1 and 2, MASH, and ASAS-CC were analyzed to determine whether it met the cutoff of ≥ 1 to ≥ 8 SIJ quadrants affected by BME, erosion, or fat metaplasia. A case was defined as meeting a given cutoff only if the majority of individual readers’ scores met or surpassed the defined cutoff. Results were summed up on the individual subject level.
Analysis of SIJ lesion data. The results of SIJ MRI assessment were analyzed according to (1) age categories 18-29, 30-39, and 40-45 or 40-50 years; (2) the number of subjects (%) showing ≥ 1 to ≥ 8 SIJ quadrants affected by BME, erosion, or fat metaplasia, and the number of subjects (%) showing SIJ halves affected by ankylosis; and (3) previously reported data-driven MRI cutoffs, expressed as the number of SIJ quadrants affected by a given lesion for a positive MRI indicative of inflammatory or structural lesions typical of axSpA according to the majority opinion of ≥ 4/7 ASAS-MRI working group central readers.11 These cutoffs were BME on ≥ 4, erosion on ≥ 3, and fat metaplasia on ≥ 5 SIJ quadrants on nonconsecutive SIJ MRI slices. Data were assessed separately for the 3 independent cohorts and for the 3 groups of healthy individuals, patients with NSBP, and patients with lumbar disc herniation.
Statistical analysis. Demographic and clinical characteristics of the study subjects were analyzed descriptively. SIJ MRI data were analyzed according to concordance by the majority of readers: ≥ 3/5 and ≥ 3/4 readers for MORPHO 1 and 2, respectively, 2/2 readers for MASH, and ≥ 4/7 readers for ASAS-CC.
RESULTS
Study subjects. Demographic and clinical characteristics of the 3 cohorts subdivided into 3 groups are summarized in Table 1. The mean age of healthy individuals from MORPHO and MASH was 31.4 (SD 6.9) and 34.2 (SD 6.8) years, respectively. In patients with NSBP, the mean age was 34.2 (SD 8.5) and 33.4 (SD 8.0) years in MORPHO and ASAS-CC, respectively. For patients with lumbar disc herniation from MASH, the mean age was 35.2 (SD 5.7) years.
Demographic and clinical characteristics of the 3 cohorts.
Proportions of subjects with MRI lesions. The frequencies of BME in healthy individuals in the MORPHO and MASH cohorts were numerically higher in the oldest age category (Table 2), although a progressive increase of BME for patients with NSBP with increasing age categories was not evident in the MORPHO and ASAS-CC groups (Table 3). In the MORPHO and MASH groups, only 3/79 (3.8%) and 3/78 (3.8%) healthy individuals had an erosion in ≥ 1 SIJ quadrant by the majority of readers (Table 2). Five of 87 (5.7%) and 2/46 (4.3%) patients with NSBP in the MORPHO and ASAS-CC groups, respectively, had an erosion in ≥ 1 SIJ quadrant by the majority of readers (Table 3). There was no progressive increase of erosion with age in healthy individuals or in patients with NSBP. More fat metaplasia lesions were observed in patients with NSBP than in healthy individuals in the MORPHO cohort, but a consistent progressive increase of fat metaplasia with age was not evident in healthy individuals and patients with NSBP across the MORPHO, MASH, and ASAS-CC groups. Patients with lumbar disc herniation showed no increase of BME across age categories, erosion was absent, and fat metaplasia increased across age groups, but only below the ASAS cutoff for fat metaplasia in ≥ 5 SIJ quadrants (Table 4). In our 3 cohorts, 13 subjects with non-axSpA (6 healthy individuals and 7 patients with NSBP) had ≥ 1 SIJ quadrant with erosion reported by a majority of readers. Ten of these also showed concomitant BME in ≥ 1 SIJ quadrant, as reported by a majority of readers, although this did not substantially change sensitivities and specificities in our analysis (data not shown).
Number of healthy individuals with MRI SIJ lesions.
Number of patients with nonspecific back pain with MRI SIJ lesions.
Number of patients with lumbar disc herniation with MRI SIJ lesions.
Proportions of subjects with ASAS-MRI lesion thresholds. In the MORPHO and MASH cohorts, 3/79 (3.8%) and 1/78 (1.3%) of healthy individuals, respectively, and 4/87 (4.6%) and 0/46 (0%) in the MORPHO and ASAS-CC groups of patients with NSBP, respectively, had BME in ≥ 4 SIJ quadrants, the cutoff for inflammatory lesions denoting axSpA (Table 2 and Table 3). There was no evidence for an increase in study subjects reaching this threshold for BME with increasing age categories, both among healthy individuals and patients with NSBP. None of the healthy individuals and only 2/87 (2.3%) and 2/46 (4.3%) of patients in the MORPHO and ASAS-CC groups with NSBP, respectively, showed erosion in ≥ 3 SIJ quadrants, the cutoff for erosions indicative of axSpA. There was not a single case meeting the erosion cutoff in the highest age category. An age-dependent increase for ≥ 5 SIJ quadrants with fat metaplasia, the cutoff compatible with axSpA, was observed in healthy individuals (from 2.6% to 10%) and patients with NSBP (from 3.7% to 16%) in the MORPHO cohort only, which was not evident in the MASH and ASAS-CC cohorts (Table 2 and Table 3).
Specificity for controls as a whole group published previously and after stratification by age categories. MORPHO part 1 reported for healthy individuals vs nr-axSpA and r-axSpA specificities of 0.97 and 0.97, respectively, and for NSBP vs nr-axSpA and r-axSpA, specificities of 0.97 and 0.97, respectively.4 MORPHO part 2 reported, for ≥ 2 erosions, specificities for NSBP vs nr-axSpA in the Swiss and Canadian arms of 0.97 and 0.90, respectively, and for NSBP vs r-axSpA, specificities of 0.97 and 0.90, respectively.5 This study showed for pooled MORPHO healthy individuals stratified by age categories of 18-29, 30-39, and 40-50 years specificities of 0.97, 1.0, and 1.0, respectively, for ≥ 2 erosions and 1.0, 1.0, and 1.0, respectively, for ≥ 3 erosions. For pooled MORPHO patients with NSBP stratified by age categories of 18-29, 30-39, and 40-50 years, specificities of 0.96, 0.97, and 1.0, respectively, were shown for ≥ 2 erosions and 0.96, 0.97, and 1.0, respectively, for ≥ 3 erosions. MASH reported a specificity of 1.0 for healthy individuals for both lesion cutoffs of ≥ 2 or ≥ 3 erosions.6 This study showed for MASH healthy individuals stratified by age categories of 18-29, 30-39, and 40-50 years specificities of 1.0, 1.0, and 1.0, respectively, for both cutoffs of ≥ 2 or ≥ 3 erosions. ASAS-CC reported for NSBP vs axSpA specificities of 0.91 and 0.96, respectively, for ≥ 2 or ≥ 3 erosions.11 This study showed for ASAS-CC, NSBP stratified by age categories of 18-29, 30-39, and 40-45 years specificities of 0.87, 1.0, and 1.0, respectively, for both cutoffs of ≥ 2 or ≥ 3 erosions. Specificities for controls stratified by age categories in 3 independent cohorts were virtually identical to the specificity in the whole group of controls without decreasing specificity across progressive age categories.
DISCUSSION
Our data from 3 independent cohorts of healthy individuals, NSBP controls, and patients with lumbar disc herniation—collected in different sites all over the world in different time periods—did not indicate progressive increases in MRI erosions and BME in the SIJ with increasing age categories when standardized definitions for reporting axSpA lesions were adopted and applied by trained and calibrated readers. MRI cutoffs for BME and erosion denoting axSpA discriminated equally well between axSpA and NSBP across all age categories.
Recognition of erosion and other structural lesions was the key feature on pelvic radiographs in the traditional imaging assessment in axSpA for decades.12 The advent of SIJ MRI not only permitted a tomographic joint evaluation but also provided novel insights into the lesion signature of bone remodeling upon initial sacroiliitis visualized by BME. MRI of the SIJ set the basis for contextual assessment of concomitant structural and active changes by scrolling simultaneously across T1SE and STIR sequences of the cartilaginous joint compartment. Recognition of erosion on pelvic radiographs and on SIJ MRI is challenging. False positive assignments of structural lesions such as erosion on SIJ MRI can be minimized by adopting standardized lesion definitions.10
A major challenge in recognizing erosion on SIJ MRI is the paucity of data on the many physiological variations of joint anatomy and their potential change with increasing age, such as degenerative progression. The resulting negative effect on the reliability of nontrained readers in recognizing structural lesions on SIJ MRI in patients with new-onset inflammatory back pain has previously been reported by Jacquemin et al.13 In their study, reader 1 recorded erosion in 82/664 patients (12.3%), whereas reader 2 reported erosion in 256/664 (38.6%), leading to a > 3-fold interreader discrepancy regarding frequency of erosion in an inception cohort of patients with axSpA. The authors concede that physiological SIJ variants may have been considered as structural lesions, especially for erosion.
A multireader validation exercise by the ASAS-MRI working group, using 278 MRI scans from ASAS-CC, defined erosion 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 T1SE non–fat-suppressed sequence compared with the normal bright appearance of adjacent bone marrow.10 The optimal cutoffs for lesions to reach a positive predictive value of ≥ 95% for clinical diagnosis of axSpA were erosion in any of ≥ 3 SIJ quadrants and/or on ≥ 2 consecutive slices with erosion.11 Figure 1A shows extended erosion along the entire vertical extent of both iliac bones with concomitant sclerosis. Large iliac erosions with loss of cortical bone and adjacent marrow matrix meet the ASAS data-driven definition for erosion.
SIJ MRI showing erosion and noninflammatory lesions simulating erosion. (A) A 30-year-old male with axSpA. T1-weighted SIJ MRI demonstrates multiple large iliac erosions (arrows) with surrounding sclerosis. (B) A 26-year-old healthy female volunteer. A small erosion-like feature is visible in the right ilium (arrow) with focal loss of visualization of the iliac cortex. This normal variation occurs when the joint is not perpendicular to the acquisition plane. (C) The semiaxial image shows an abrupt change in orientation of the joint at this point (arrow). (D) A 35-year-old healthy female volunteer. Linear areas of low signal intensity (arrows) extend perpendicular to or at angles from the articular surface. When short, this normal variation can mimic erosion. The figures were derived from daily routine databases to match as closely as possible several signal changes on the reference figures by Renson et al.3 axSpA: axial spondyloarthritis; MRI: magnetic resonance imaging; SIJ: sacroiliac joint.
We hypothesize that the high prevalence of erosion and its progressive increase with age categories reported by Renson et al,3 which was not confirmed by our data, may be a result of physiological SIJ contour variations classified as erosion (Figure 3E in Renson et al). Figures 1B,C are similar to the lesion in the right ilium classified as erosion by Renson et al in their Figure 3E. Figures 1B,C demonstrate an apparent small erosion in the right ilium that is an artifact caused by partial volume averaging of a part of the joint that is not perpendicular to the semicoronal plane of acquisition by an abrupt change in orientation. Figure 1D displays lesions in the right sacrum and left ilium, similar to the observation designated as erosion by Renson et al in their Figure 3E. Linear areas of low signal intensity on T1SE MRI extend perpendicular to or at angles from the articular surface. When short, these can mimic erosion, but they are a normal variation on MRI. Neither of these 2 lesions meets the ASAS definition of erosion.10 The reports by Jacquemin et al13 and Renson et al3 emphasize the need to apply standardized lesion definitions for assessment of SIJ MRI to mitigate overcall of physiological SIJ variation or imaging artifacts.
Another concern relates to the potential for overinterpretation of ankylosis. Definitions of joint ankylosis for all imaging modalities require evidence of bone crossing the entire joint space width (Figure 2A).10 The bony prominence seen in the left SIJ in Figure 3D by Renson et al is typical for an osteophyte and is almost identical to the osteophytes in our example of SIJ osteoarthritis (Figure 2B).
SIJ MRI showing ankylosis and SIJ osteoarthritis. (A) A 38-year-old male with axSpA; T1-weighted MRI displays partial ankylosis of the left SIJ (arrow). Bone is bridging the upper joint space with bright bone marrow signal crossing the joint space in continuity with both ilium and sacrum. This is the definition of joint ankylosis on T1-weighted MRI. (B) A 32-year-old male with low back pain and SIJ osteoarthritis. Sessile osteophytes arise from the right ilium and left sacrum close to the anterior joint (arrows). The bright bone marrow signal within the osteophytes does not cross the joint space. Apart from anterior osteophytes, the MRI scan was normal and there was no clinical evidence of axSpA. The figures were derived from daily routine databases to match as closely as possible several signal changes on the reference figures by Renson et al.3 axSpA: axial spondyloarthritis; MRI: magnetic resonance imaging; SIJ: sacroiliac joint.
In a retrospective study, SIJ MRIs of 309 patients with chronic back pain diagnosed with axSpA or non-SpA were evaluated for BME, fat metaplasia, and erosion by 2 blinded readers.14 Consistent with our results, the number of SIJ quadrants affected by the 3 lesions showed no age-dependent progression. Even in a proportion of 36% of patients with non-SpA beyond age 50, for which there are hardly any data available for SIJ lesions, erosions showed no increase in frequency compared to the younger age categories.
T1SE sequences are recognized for their limitations in detecting structural lesions. Additional high-resolution sequences specifically used for assessment of the bone/cartilage interface may advance the field, yet the utility of these sequences has not yet been proven for MRI assessment in clinically suspected axSpA. A crucial requirement not met today is evidence that these sequences enhance reliability for detection of erosion. Prior reports have only dealt with their enhanced sensitivity compared to the gold standard of computed tomography of the SIJ. Concerns relate to false positive erosions compared to computed tomography, potentially leading to overcall of physiological SIJ variations because of thinner slices compared to traditional T1SE and increased risk of MRI artifacts.15
SIJ MRI features of uncertain significance seen in 1 orientation only should be verified on a second plane with a different, preferably orthogonal, orientation.10,16,17 Small solitary lesions should be interpreted with caution since they are more likely to be subject to simulation by anatomical variation or imaging artifacts. If in doubt, assignment of small lesions, which are hard to classify, may be influenced by their appearance on adjacent slices and by the presence of concomitant lesions.10 A standard MRI acquisition protocol to evaluate the SIJ should include at least 4 sequences: a T1SE sequence and a bone/cartilage interface sequence in the semicoronal plane, such as T1 fat-saturated or gradient echo sequences, and fluid-sensitive sequences in 2 perpendicular semicoronal and semiaxial planes, such as STIR or T2 fat-saturated sequences.17,18
Erosion almost never occurs as an isolated lesion but mostly with some other lesion types, such as sclerosis, as shown in Figure 1A. Beyond the contextual lesion signature, lesion size also matters. The data-driven ASAS proposal for lesion cutoffs aiming at high specificity stipulates that a lesion must be clearly observed in ≥ 3 SIJ quadrants and/or on ≥ 2 consecutive slices before readers can be confident that there is true erosion. Figures 1B-D represent but 2 examples of “fake erosion” among a broad spectrum of physiological SIJ variation. Imaging artifacts demonstrate that minor abnormalities that do not meet the ASAS criteria can occur for various reasons and that it is desirable to have standardized definitions that provide guidance for interpretation of SIJ MRI.
Comparable to the report by Renson et al, absolute numbers of healthy individuals and patients with NSBP are relatively small in our 3 cohorts. The reason for a consistent decline in numbers toward the oldest age category is that in the study by Renson et al and in our MORPHO and MASH cohorts, control subjects were age- and sex-matched to patients with axSpA, in which symptom onset is more likely to occur in the younger age categories than in the category of 40-49 years. However, despite small numbers in the older age categories, the lack of an age-dependent increase in SIJ MRI lesions was consistent across our 3 independent cohorts. Further, we observed this lack of age-dependent increase in SIJ MRI lesions not only in healthy individuals but also in patients with NSBP, a group of greater clinical relevance in daily routine.
There are several study limitations. The study design was different among the 3 source cohorts. MORPHO and MASH were designed to explore the diagnostic utility of MRI in axSpA by enrolling patients with nr-axSpA and r-axSpA, patients with NSBP (the clinically most challenging differential diagnostic condition), and healthy controls. ASAS-CC recruited consecutively new patients with chronic (≥ 3 months) back pain of unknown origin that began before the age of 45 years aimed at classifying patients as having or not having axSpA following clinical investigations. Patients with NSBP in ASAS-CC underwent a full clinical examination, whereas patients with NSBP and healthy volunteers recruited to MORPHO and MASH filled in questionnaires about inflammatory back pain, the Nordic Questionnaire, and a questionnaire about conditions potentially related to axSpA. Acquisition of SIJ MRI in MORPHO and MASH was performed on 1.5 T devices of the same brand, whereas there were various brands in use for the global recruitment of patients in ASAS-CC. However, the MRI protocol was similar across the 3 cohorts. Lesion definitions and reference image sets were slightly modified between the first readout time period of MORPHO and the last readout time period of ASAS-CC. Arguably, the largest change was related to fat metaplasia, where additional items were included in the lesion definition to enhance specificity such as a distinct border, homogeneous T1-weighted fat signal, and location adjacent to subchondral bone. This is the most likely explanation for the inconsistent results for fat metaplasia seen only in the earliest cohort of MORPHO, since these additional descriptions for fat metaplasia did not yet apply at the time. We did not analyze backfill in all our cohorts, given this lesion is among the least common of the structural lesions related to axSpA. Moreover, sclerosis is well-known to become more evident with increasing age. Our MRI data with an upper age limit of 45 or 50 years cannot be applied to patients with suspected axSpA presenting beyond age 50.
In conclusion, multireader assessment of SIJ MRI from 3 independent cohorts of healthy individuals, patients with NSBP, and patients with lumbar disc herniation did not show progression in structural lesions with increasing age categories when standardized definitions for axSpA lesions were applied. Data-driven cutoffs indicative of structural MRI lesions typical of axSpA discriminated equally well between axSpA and patients with NSBP across all age categories. Structural lesions on SIJ MRI had high diagnostic specificity for axSpA also in older age categories.
Footnotes
CONTRIBUTIONS
UW: writing – first draft. All authors were involved in the conception and design of the study and in data analysis and interpretation. All authors contributed to revising the manuscript critically for important intellectual content, and all authors approved the final version to be published. UW had access to all data and takes responsibility for the integrity of the data and the accuracy of the data analysis.
FUNDING
No specific funding was received from any bodies in the public, commercial, or not-for-profit sectors to carry out the work described in this article.
COMPETING INTERESTS
The authors declare no conflicts of interest related to this article.
ETHICS AND PATIENT CONSENT
Local ethical committee approval had been obtained for all cohorts, and all study subjects had provided written informed consent at enrollment. ASAS-CC was considered to not require ethical approval for the post-hoc MRI analysis. MASH received approval from the Ethical Committee of the Capital Region of Denmark (H-2-2013-019). MORPHO received approval from the Kantonale Ethikkommission Zurich, Zurich, Switzerland and Health Research Ethics Board of the University of Alberta, Edmonton, Canada.
DATA AVAILABILITY
The authors confirm that the data supporting the findings of this study are available within the article.
- Accepted for publication January 16, 2025.
- Copyright © 2025 by the Journal of Rheumatology
REFERENCES
SUPPLEMENTARY DATA
Supplementary material accompanies the online version of this article.
