Article Text
Abstract
Objective To assess the incremental diagnostic value of spine MRI evaluated separately from and combined with sacroiliac joint (SIJ) MRI in non-radiographic axial spondyloarthritis (nr-axSpA) compared with SIJ MRI alone.
Methods The study sample comprised two independent cohorts A/B of 130 consecutive patients aged ≤50 years with back pain, newly referred to two university clinics, and 20 healthy controls. Patients were classified according to clinical examination and pelvic radiographs as having nr-axSpA (n=50), ankylosing spondylitis (n=33), or non-specific back pain (n=47). Four readers assessed SIJ and spine MRI separately 6 months apart, and 1–12 months later both scans simultaneously using standardised modules. Readers recorded presence/absence of SpA and their level of confidence in this conclusion on a 0–10 scale (0=definitely not; 10=definite). We analysed differences between SIJ MRI versus spine MRI alone, and SIJ MRI alone versus combined MRI, descriptively by the number/percentage of subjects according to the mean of four readers.
Results In cohorts A/B, 15.8%/24.2% of patients with nr-axSpA having a negative SIJ MRI were reclassified as being positive for SpA by global evaluation of combined scans. However, 26.8%/11.4% of non-specific back pain controls and 17.5% of healthy volunteers with a negative SIJ MRI were falsely reclassified as having SpA by combined MRI. Low confidence in a diagnosis of SpA by SIJ MRI increased to high confidence by combined MRI in 6.6%/7.3% of patients with nr-axSpA.
Conclusions Combined spine and SIJ MRI added little incremental value compared with SIJ MRI alone for diagnosing patients with nr-axSpA and enhancing confidence in this diagnosis.
- Spondyloarthritis
- Magnetic Resonance Imaging
- Ankylosing Spondylitis
Statistics from Altmetric.com
Introduction
The definition of a positive MRI scan as major criterion in the Assessment of SpondyloArthritis International Society (ASAS) classification criteria for axial spondyloarthritis (SpA)1 is based on MRI of the sacroiliac joints (SIJ) alone.2 It is not known whether a combination of spinal and SIJ MRI may enhance diagnostic certainty over and above SIJ MRI alone.
Several controlled studies explored the diagnostic utility of spinal MRI lesions alone without taking into account concomitant SIJ MRI lesions in patients with ankylosing spondylitis and non-radiographic axial SpA (nr-axSpA).3–6 The lesion type found to be most characteristic for SpA was bone marrow oedema (BMO) of the vertebral corners, corner inflammatory lesion (CIL).7 Another lesion type, corner fat lesion (CFL), seemed to be less specific for SpA.6 BMO in the lateral and posterior compartments of the spine7 proved to be highly specific for SpA, but their use for diagnostic purposes was limited by low sensitivity.5 ,8 A recent systematic literature review by an ASAS and Outcome Measures in Rheumatology working group generated a consensus statement that ≥three CILs or several CFLs may constitute a positive spine MRI for SpA.9
Patients with non-specific back pain (NSBP) and healthy subjects may also show spinal MRI lesions suggestive of SpA such as CIL or CFL.4–6 There is limited data about their frequency and whether they may result in a false positive classification of non-SpA controls as having SpA. Moreover, there are no studies that systematically evaluate concomitant SIJ and spine MRI lesions in patients with SpA and non-SpA controls and their contributions to diagnostic utility. A simultaneous assessment of spine and SIJ MRI may enhance diagnostic certainty in the clinically challenging nr-axSpA group over the traditional approach with SIJ MRI alone, particularly if SIJ MRI appears normal or inconclusive. The limited sensitivity of SIJ MRI alone is illustrated by only 38%10 and 52%11 of patients with nr-axSpA being identified, depending on the gold standard used, that is, histopathology of SIJ biopsies10 or clinician expert opinion.11
Our objectives were to assess: (1) the incremental diagnostic value of spine MRI evaluated independently from SIJ MRI and then simultaneously with SIJ MRI in two SpA inception cohorts compared with the diagnostic value of SIJ MRI alone; (2) the frequency of false positive classification as SpA in NSBP and healthy controls according to spine MRI alone and combined MRI; (3) which MRI spinal lesion type had the highest impact on correct/incorrect classification.
Patients and methods
Subjects
The study sample comprised two independent cohorts A/B of 130 consecutive patients aged ≤50 years with back pain newly referred to two university outpatient clinics, in whom SIJ and spinal MRI were conducted. Forty-two patients with back pain of cohort A were referred by rheumatologists or general practitioners for further evaluation of suspected SpA. Twenty age-matched healthy controls were concomitantly recruited from hospital staff of the same clinic. Eighty-eight patients of cohort B presented with acute anterior uveitis (AAU) to a university ophthalmology department; all patients with AAU with past or present back pain were referred to the rheumatology department of the same university hospital for assessment of SpA. We used the Calin criteria to assess inflammatory back pain.12 Both cohorts were enrolled in observational protocols with structured questionnaires on SpA-related clinical features and with patient reported outcomes: a modified Outcome in Ankylosing Spondylitis International Study (OASIS) protocol for cohort A,13 and the Spondyloarthritis Research Consortium of Canada (SPARCC) protocol for cohort B.14 Two blinded readers at each site independently categorised pelvic radiographs according to the modified New York criteria15; discrepancies in radiographic assessment were resolved by consensus. As MRI features were the independent variables in this study, MRI could not be used concomitantly to classify the study subjects to avoid circuitous reasoning.16 In both inception cohorts, a classification of SpA was made based on the clinical opinion of one local rheumatologist (UW/WPM for cohort A/B). Patients were classified by clinical examination and pelvic radiographs as having nr-axSpA (n=19/31, for cohort A/B), ankylosing spondylitis (n=9/24) and NSBP (n=14/33). Ongoing or previous treatment with biologicals was an exclusion criterion. Local Ethics Review Boards approved the study protocol, and patients and controls gave written informed consent.
Evaluation of MR images
The technical parameters for Short Tau Inversion Recovery (STIR) and T1 spin echo sequences of semicoronal SIJ MRI and of sagittal MRIs of the entire spine have been described previously.5 ,8 ,11 The MRI scans were read and scored independently by four blinded readers (one radiologist: VZ; three rheumatologists: UW/WPM/ZZ). SIJ MRI alone and spine MRI alone were read independently 6 months apart, with another interval of 1–12 months to the combined assessment of SIJ and spinal MRI simultaneously. The films of each cohort were independently evaluated in random order on electronic work stations in each reader's institution. MRI data were entered into customised online modules for the SIJ and spine.
The MORPHO module served to assess the SIJ MRI.11 Spinal MRI scans were evaluated according to the Canada-Denmark MRI system.17 ,18 Both modules comprise two sections: global assessment of the MRI and detailed scoring of various MRI lesion types.
Global assessment: Readers recorded presence/absence of SpA and their level of confidence in this diagnosis on a 0–10 numerical rating scale (0=definitely not SpA; 10=definite SpA) according to global evaluation of T1 spin echo and STIR sequences simultaneously. Categories 8–10 and 0–2 were defined as constituting a high level of confidence in a classification of SpA and non-SpA, while categories 6–7 and 3–4 were defined as moderate confidence. A category of 5 was deemed equivocal. Readers indicated the single most important MRI lesion influencing their conclusion regarding presence/absence of SpA, separately for SIJ and spine scans.
Detailed scoring section: Readers reported the different MR lesions (BMO, fat infiltration, erosion) as present/absent according to SIJ quadrants (each half of the SIJ for ankylosis) according to the MORPHO module and standardised lesion definitions illustrated by a reference image set.11 ,19 Readers also assessed BMO and fat infiltration in the spine according to standardised lesion definitions in 23 discovertebral units from C2/C3 to L5/S1.17–19
Degenerative spinal lesions defined as loss of ≥50% of the normally expected disc height at this level recorded concordantly by ≥two of three readers (UW/VZ/WPM) were assessed in the cervical (C2/C3-C7/T1) and lumbar (T12/L1-L5/S1) spine.
Statistical analysis
Differences in demographic and clinical characteristics between cohorts A and B were assessed by Fisher's exact test for nominal and Wilcoxon test for continuous variables. The frequency of MRI lesion types was analysed descriptively on the subject level as the mean number of subjects (percentage) according to four raters who recorded the specific MRI lesion affecting ≥one SIJ quadrant and/or ≥one spinal location. The frequency of different MRI lesions in the SIJ and spine on a group level was analysed descriptively as mean value (range) according to the four readers.
Mean values of global MRI assessment over all four readers served to evaluate the diagnostic performance of SIJ MRI alone, spine MRI alone and combined MRI according to expert clinician opinion as gold standard for: (1) the number (percentage) of subjects with a positive SIJ MRI, a positive spine MRI and a positive combined MRI according to global evaluation of MR scans, and stratified by level of confidence for presence/absence of SpA; (2) comparison between SIJ MRI alone versus spine MRI alone, and between either SIJ MRI alone or spine MRI alone versus combined MRI to assess the incremental diagnostic value for presence of SpA; (3) the single most important MRI lesion responsible for correct classification of patients with SpA and misclassification of controls.
Results
Descriptive analysis
Study subjects. The two recruitment strategies to identify patients with nr-axSpA resulted in substantial demographic and clinical differences between the two back pain cohorts A/B with cohort B (AAU plus back pain) having longer disease duration and less severe disease for patients with nr-axSpA (table 1). This observed heterogeneity led us to analyse the two cohorts separately instead of pooling the data.
Frequency of MRI abnormalities on subject level. BMO in ≥one SIJ quadrants was recorded in 84.2%/49.2% of patients with nr-axSpA for cohort A/B, but also in 26.8%/27.3% of NSBP and 33.8% of healthy controls (table 1). Erosion in ≥one SIJ quadrants was rarely seen in controls (5.4%/13.6% in NSBP and 1.3% of healthy controls), but frequently in nr-axSpA with 63.2%/35.5%. The proportion of subjects having BMO or fat infiltration in ≥one spine location was 82.9%/88.7% for nr-axSpA and 91.1%/69.7% for NSBP in cohorts A/B, and 81.3% for healthy controls. For cohort A/B, the frequency of subjects with ≥one spinal MRI lesion, but normal SIJ scans was low for patients with nr-axSpA (5.3%/25.8%), but high for NSBP (53.6%/38.6%) and healthy controls (45.0%).
Frequency of MRI abnormalities on group level. The mean number of SIJ quadrants showing BMO in patients with nr-axSpA was 16.4/4.1 in cohort A/B, compared with NSBP controls with 0.9/1.3 SIJ quadrants. Patients with nr-axSpA had erosion in 4.3/2.6 SIJ quadrants, while erosion was rarely seen in NSBP controls (0.2/0.4 SIJ quadrants) and virtually absent in healthy controls. In patients with nr-axSpA, the mean number of spinal locations with BMO was 5.0/3.3 in cohort A/B, and 1.8/1.2 in NSBP controls. Fat infiltration was reported in 5.3/7.9 spinal locations in patients with nr-axSpA compared with 4.1/3.7 in NSBP and 3.0 in healthy controls. Due to the small frequency of degenerative spinal lesions in both cohorts no adjustment for degenerative lesions was necessary (data not shown).
Diagnostic value of SIJ MRI alone, spinal MRI alone and combined MRI
Comparison of SIJ MRI versus spinal MRI, and of SIJ MRI alone and of spinal MRI alone versus combined MRI. In the setting of a clinical classification of nr-axSpA but normal SIJ MRI, the combined MRI picked up 15.8%/24.2% additional patients with SpA in cohort A/B (table 2). However, comparable proportions of controls with negative SIJ MRI were misclassified by combined MRI as having SpA: 26.8%/11.4% of NSBP and 17.5% of healthy controls. A moderate level of confidence of 5–7 for a classification of SpA by SIJ MRI alone increased to high confidence of 8–10 by combined MRI in 6.6%/7.3% of patients with nr-axSpA.
Comparison of SIJ MRI, spinal MRI and combined MRI positivity for SpA with clinical classification as gold standard. In cohort A/B, 76.3%/46.0% of patients with nr-axSpA were correctly classified as patients with SpA by SIJ MRI alone, and 51.3%/48.4% by spine MRI alone (table 3). Combined scans increased the detection rate among patients with nr-axSpA to 89.5%/67.7% for cohort A/B. Falsely positive classifications as patients with SpA by SIJ MRI alone were observed in 7.1%/15.9% of NSBP and 7.5% of healthy controls, and by spine MRI alone in 33.9%/28.0% of NSBP and 26.3% of healthy controls, respectively. The high rate of false positive misclassification of controls by combined MRI (32.1%/24.2% of NSBP and 25.0% of healthy controls) was nearly identical to the misclassification rate of controls by spinal MRI alone, indicating that misclassification by combined MRI was mainly directed by spinal MRI.
MRI lesions responsible for misclassification of controls
In cohort A, the single most important MRI lesion for correct classification of patients with nr-axSpA was BMO for SIJ and spine MRI (69.0% and 79.5%, respectively); misclassification of controls was also mainly based on BMO for SIJ MRI (50.0% for NSBP and 100.0% for healthy controls) and spine MRI (52.6% for NSBP and 52.4% for healthy controls) (table 4). In cohort B, correct classification of patients with nr-axSpA was mainly attributable to erosion (56.1%) on SIJ MRI and to fat infiltration (66.7%) on spinal MRI; misclassification of NSBP controls was principally based on BMO (52.4%) on SIJ MRI and on fat infiltration (62.2%) on spinal MRI. CIL and CFL contributed most to correct classification of patients with nr-axSpA as well as to misclassification of controls.
Discussion
This three-step reading exercise of SIJ MRI alone, spine MRI alone, and combined SIJ and spine MRI in two SpA inception cohorts by four readers showed that combined MRI added little incremental value for diagnosing patients with nr-axSpA and enhancing confidence in this diagnosis compared with the traditional imaging approach with SIJ MRI alone. In patients with clinically suspected nr-axSpA having a normal SIJ MRI, combined MRI identified 15.8% and 24.2% more patients with nr-axSpA in the two cohorts. However, 26.8% and 11.4% of NSBP controls and 17.5% of healthy volunteers were misclassified as having SpA by combined MRI, which neutralised the additional diagnostic value of combined MRI in patients with nr-axSpA. The lesion types mainly responsible for misclassification of controls were vertebral corner BMO and CFLs on spinal MRI scans.
A potential diagnostic benefit of combined spinal and SIJ MRI versus SIJ scans alone has to be weighed against the extra costs of spinal MRI and the additional discomfort to patients with longer scanning time. The crucial issue is the proportion of true versus false positive reclassification of cases and controls by combined spine plus SIJ MRI. For cohorts A/B, 15.8%/24.2% of patients with nr-axSpA considered having a negative SIJ MRI, were correctly reclassified as being positive for SpA by global evaluation of combined scans. However, also 26.8%/11.4% of NSBP controls and 17.5% of healthy volunteers, who showed a negative SIJ MRI, were falsely reclassified as having SpA by combined MRI. Given a negative SIJ MRI, the proportions of true positive reclassification of patients with nr-axSpA and of false positive reclassification of controls by combined MRI were very similar. SIJ MRI scans alone are less sensitive, but more specific than combined scans, where the additional spine MRI results in a comparable increase in sensitivity and loss in specificity. As specificity is the primary objective for classification purposes, and because currently no proven disease-modifying treatment is available for SpA, spinal MRI may add little incremental value to the traditional assessment with SIJ MRI scans alone in patients with clinically suspected nr-axSpA. Combined MRI did not even facilitate confidence in a classification of SpA. A moderate confidence level with a classification of SpA by global evaluation of SIJ MRI alone increased to high confidence by combined scans in only 6.6% and 7.3% of patients with nr-ax SpA in the two cohorts, and in 1.3% and 3.6% of controls.
NSBP and healthy controls were misclassified as having SpA by global assessment of spine MRI alone in 26.3%–33.9% and by combined MRI in 24.2%–32.1%, as opposed to SIJ MRI alone in 7.1%–15.9% (figure 1). Which spinal MRI lesions were the principal drivers to mislead the readers? In cohort A with shorter symptom duration and higher disease activity, the single most important lesion to falsely classify controls as having SpA was spinal BMO, for spine MRI alone and for combined MRI. For cohort B, which had longer symptom duration and less active disease, the single most important MRI lesion for incorrect classification was spinal fat infiltration.
The classification of patients as having NSBP but who could actually have nr-axSpA is an important potential source of bias which could potentially account for the misclassification on MRI. However, in cohort B, the more likely bias is that the expert clinicians who assessed patients with AAU would have been aware of the clinical background and human leucocyte antigen B27 (HLA-B27) status of the patients referred from the AAU clinic and therefore more inclined to make a diagnosis of inflammatory back pain rather than mechanical back pain. Consequently, it is also possible that the primary bias in clinical assessment is directed towards false positive clinical diagnoses in the nr-axSpA population. In this circumstance, the expert clinician is only likely to diagnose NSBP if there is a strong clinical basis for deciding against an inflammatory basis to the patient's symptoms. This bias for the clinical diagnosis would have the effect of reducing sensitivity for MRI in the nr-axSpA group and increased specificity in the NSBP group. Consequently, the misclassification of spinal MRI in NSBP in the AAU cohort could in fact be regarded as a conservative estimate. The observation that similar misclassification was observed in both inception cohorts derived through different recruitment methods strengthens the validity of the findings. Further strengths of this study are the evaluation of MRI scans by radiologist and rheumatologist readers from three international sites, and the inclusion of healthy controls.20
The principal limitation inherent to imaging studies in nr-axSpA is the selection of the gold standard.21 To avoid conceptual circularity, MRI should not be used for classification when it is also being evaluated for its diagnostic utility.16 Gold standard for classification of SpA was therefore physician expert opinion which is consistent with the gold standard used in several classification systems in SpA (Amor,22 European Spondylarthropathy Study Group (ESSG),23 ASAS1 criteria) over the past two decades, and early SpA cohorts such as DEvenir des Spondyloarthropathies Indifférenciées Récentes (DESIR).24 Patients were also enrolled in standardised protocols to assess inflammatory back pain and SpA-related clinical features. HLA-B27 has limited diagnostic value in cohort B recruited by the key symptom AAU as about 50–60% of patients with AAU are HLA-B27 positive because of their AAU.25 Assessment of laboratory values in controls of cohort A was not permitted by the local ethics committee due to concerns that incidental positive findings may have a potentially negative psychological impact.
Our approach with physician expert opinion as gold standard for a MRI study in nr-axSpA does not exclude the possibility of false positive and false negative clinical assignments. A further gold standard could be re-evaluation of the initial classification by longitudinal follow-up. However, longitudinal studies are limited by patients lost to follow-up, particularly in disorders with slow progression such as SpA, and it has to be recognised that some patients may never develop radiographic features.
In conclusion, combined spine and SIJ MRI added little incremental value for diagnosing patients with nr-axSpA and enhancing confidence in this diagnosis compared with the traditional imaging approach with SIJ MRI alone. Vertebral corner lesions on spine MRI were the main drivers towards misclassification of controls. Caution is warranted if a classification of SpA is based on spinal MRI alone, where an additional assessment of the SIJ may be advisable.
Acknowledgments
The authors thank the patients and the healthy volunteers for their participation; Tracey Clare, Clinical Research Manager, Department of Radiology, University of Alberta, and Paul Filipow, Data Manager, Edmonton, Canada for coordinating the Web-based MRI scoring module; and Christian Streng, Head of Medical Documentation, Balgrist University Hospital, Zurich, Switzerland, for his technical assistance with the figures. Walter P Maksymowych is a Scientist of Alberta Innovates Health Solutions.
References
Footnotes
Handling editor Tore K Kvien
Contributors KR, MØ, RGWL, SJP, UW and WPM drafted the study design. SMC, UW and WPM acquired the clinical data. UW, WPM, VZ and ZZ were MRI readers. UW and KR performed the statistical analysis. All authors take responsibility for interpretation of data. UW drafted the manuscript with important contributions from all authors. All authors read and approved the final manuscript. UW had full access to all data in the study and takes responsibility for the integrity of the data.
Funding The Canadian Arthritis Society: National Research Initiative Award. Alberta Innovates Health Solutions. Walter L and Johanna Wolf Foundation, Zurich, Switzerland.
Competing interests Dr Rufibach is founder and owner of Rufibach rePROstat and is an employee of F Hoffmann-La Roche, Basel, Switzerland.
Ethics approval Ethics Committee of the University of Alberta, Edmonton, Canada. Kantonale Ethikkommission Zurich, Zurich, Switzerland.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement This is the first analysis of this dataset comparing combined SIJ and spine MRI data to SIJ MRI data alone. The dataset is available for all nine manuscript authors, but for nobody else. A scientific abstract from these data has been presented at ACR 2012 and at EULAR 2013; there were no further publications or public presentations of these data.