SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Objective

To compare the diagnostic utility of T1-weighted and STIR magnetic resonance imaging (MRI) sequences in early spondylarthritis (SpA) using a standardized approach to the evaluation of sacroiliac (SI) joints, and to test whether systematic calibration of readers directed at recognition of abnormalities on T1-weighted MRI would enhance diagnostic utility.

Methods

Six readers independently assessed T1-weighted and STIR MRI scans of the SI joints from 187 subjects: 75 ankylosing spondylitis (AS) and 27 preradiographic inflammatory back pain (IBP) patients, and 26 mechanical back pain and 59 healthy volunteer controls ages ≤45 years. The exercise was repeated 6 months later on a random selection of 30 AS patients and 34 controls after calibration directed at lesions visible on T1-weighted MRI. Specific MRI lesions were recorded according to standardized definitions. In addition to deciding on the presence/absence of SpA, readers were asked which MRI sequence and which type of lesion was the primary basis for their diagnostic conclusion.

Results

Structural lesions were detected in 98% of AS patients and 64% of IBP patients. A diagnosis of SpA was based on T1-weighted or combined T1-weighted/STIR sequences in 82% of AS patients and 41% of IBP patients. Calibration enhanced the diagnostic utility of MRI in the majority of readers, especially those considered less experienced; the mean positive and negative likelihood ratios (of 6 readers) were 14.5 and 0.08 precalibration, respectively, and 22.2 and 0.02 postcalibration, respectively.

Conclusion

Recognition of structural lesions on T1-weighted MRI contributes significantly to its diagnostic utility in early SpA. Rheumatologist training directed at detection of lesions visible on T1-weighted MRI enhances diagnostic utility.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

There is increasing acceptance that magnetic resonance imaging (MRI) has high diagnostic utility in early spondylarthritis (SpA). Most prior studies on the diagnostic utility of MRI in early SpA patients (1–5) and a recent consensus-driven Assessment of SpondyloArthritis international Society (ASAS) proposal to define sacroiliitis by MRI (6) have focused only on the presence of bone marrow edema (BME) on STIR sequences or on the presence of osteitis on the T1-weighted gadolinium-augmented sequence as the principal diagnostic features. But an approach that is restricted to active sacroiliitis may ignore the potential contribution to diagnostic utility of early structural abnormalities characteristically observed on T1-weighted sequences. Information on the occurrence and relative importance of structural changes in the sacroiliac (SI) joints observed on T1-weighted MRI early in the disease course of SpA is, however, scarce. Furthermore, structural abnormalities such as erosions may be difficult to diagnose on MRI, whereas sclerosis and marrow fat infiltration have an uncertain diagnostic significance. It is therefore unclear whether structural lesions occur sufficiently frequently in early SpA to warrant diagnostic scrutiny, to what degree their assessment contributes to the diagnostic utility of MRI in SpA, and whether diagnostic utility of individual readers can be improved by reader training to recognize abnormalities on T1-weighted MRI.

In this international multireader MRI standardization, calibration, and reading exercise of 187 patients with early SpA and age- and sex-matched controls, we analyzed the frequency of structural lesions of the SI joints in early SpA and assessed the contribution of T1-weighted MRI to diagnostic utility in early SpA using a standardized approach to the evaluation of SI joints that included the following: 1) development of standardized definitions of structural lesions of the SI joints on T1-weighted MRI by the Canada-Denmark MRI Working Group, 2) development of a reference SI joint MR image set by consensus among study investigators based on these definitions (online at www.arthritisdoctor.ca), 3) calibration of readers using an online training module (online at www.arthritisdoctor.ca) followed by videoteleconferences, and 4) development of a customized online data entry module based on a standardized approach to recording abnormalities in the SI joints (online at www.arthritisdoctor.ca). We also conducted a second reading exercise to determine whether training of less experienced readers directed specifically at detection of structural abnormalities on T1-weighted scans enhances the overall diagnostic utility of MRI in early SpA.

SUBJECTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Subjects.

Six readers (4 rheumatologists and 2 radiologists) from 3 international centers blinded to patient and diagnosis independently assessed T1-weighted and STIR sequences of SI joint MRI scans from 187 subjects ages ≤45 years who were recruited in 2 rheumatologic university clinics specializing in SpA. The subjects consisted of 75 ankylosing spondylitis (AS) patients who met the modified New York classification criteria (7) and with a mean disease duration of ≤10 years (mean age 31.1 years, mean symptom duration 6.1 years, 72% men, 59 [83%] of 71 HLA–B27 positive); 27 patients with inflammatory back pain (IBP) and clinically diagnosed with preradiographic SpA with a mean symptom duration of 29 months (mean age 29 years, 67% men, 23 [92%] of 25 HLA–B27 positive); 26 age- and sex-matched controls ages ≤45 years with nonspecific back pain (NSBP) diagnosed on clinical grounds; and 59 age- and sex-matched healthy controls ages ≤45 years defined by the Nordic questionnaire (8). Treatment with biologics within 6 months prior to the SI joint MRI was an exclusion criterion in the AS and IBP groups. Pelvic radiographs of both SpA patient groups were independently assessed and categorized according to the modified New York criteria (7) by 2 readers at each site. The local ethics committees approved the study protocol, and written informed consent was obtained from all of the participants.

MRI protocol.

Scans from both centers included coronal T1-weighted turbo spin-echo (T1SE) and STIR sequences angled parallel to the SI joint. The scan parameters for all of the sequences were: 15–19 slices, 4-mm slice thickness, 0.4-mm interslice gap, and field of view of 280–300 mm. For the T1-weighted sequence, repetition time (TR) was 423–450 msec, echo time (TE) was 12–13 msec, echo train length (ETL) was 3, and matrix was 512 × 256 pixels. For the STIR sequence, the values were: TR 3,700–4,930 msec, inversion time 145–150 msec, TE 50–69 msec, ETL 7–9, and matrix 256–384 × 256 pixels. These are the usual sequences and scan parameters for routine MRI evaluation of patients with SpA in the involved institutions.

Standardized assessment of MR images.

The steps that were followed in chronological order prior to the formal evaluation of any of the MRI scans are described below.

Standardized definitions of MR lesions.

We adopted standardized definitions of active inflammatory changes on STIR images and structural lesions on T1-weighted sequences of the SI joints developed by the Canada-Denmark MRI Working Group (9). BME was defined as an increase in bone marrow signal in the SI joints on STIR images; the center of the sacrum at the same craniocaudal level was used as the primary reference for normal bone marrow signal (10). We defined joint erosions as full-thickness loss of dark appearance of either iliac or sacral cortical bone of the SI joints and change in normal bright appearance of adjacent bone marrow on T1-weighted images; adjacent bone marrow demonstrates altered signal intensity on T1-weighted images as compared with normal iliac marrow (for iliac erosions) or normal sacral marrow (for sacral erosions) on the same slice at the same craniocaudal level. We defined marrow fat infiltration as focal increased signal in bone marrow on T1-weighted images. We defined ankylosis as a bright signal on T1-weighted images extending across the SI joints. Figure 1 shows joint erosions and marrow fat infiltration on a T1-weighted image of the anterior part of the SI joints with minimal BME on the corresponding STIR sequence slice.

thumbnail image

Figure 1. Comparison of sacroiliac (SI) joint magnetic resonance imaging (MRI) lesions on T1-weighted and STIR sequences (corresponding slices of the same patient). A, T1-weighted image of an HLA–B27-positive female ankylosing spondylitis patient age 34 years with a symptom duration of 8 years, showing erosion at the sacral side of the right SI joint (arrow). The marrow around this erosion demonstrates fat infiltration, but there is signal loss at the erosion (arrow) compared with normal sacral marrow (meeting the definition of erosion). The upper iliac part of the right SI joint shows erosion at the threshold of definition (broken arrow), with loss of adjacent marrow signal compared with normal iliac marrow. The left SI joint displays pseudo-widening of joint space due to definite erosion in the left iliac bone (arrowheads), showing loss of adjacent marrow signal as compared with normal iliac marrow. B, STIR image. In contrast to advanced structural lesions on the T1-weighted image, the corresponding STIR sequence only shows a faint bone marrow edema (BME) lesion in the lower iliac part of the right SI joint (arrow). Areas demonstrating faint increase in STIR signal may create diagnostic uncertainty in routine practice, as in this example of minimal increase in STIR signal in the right ilium. On the basis of the STIR sequence alone, one could not confidently confirm a diagnosis of spondylarthritis by MRI. The T1-weighted sequence demonstrates erosions and sclerosis and illustrates how much the T1-weighted image can be diagnostically useful when confronted with a doubtful BME lesion. T1SE = T1-weighted turbo spin-echo.

Download figure to PowerPoint

Reference SI joint MR image set.

We developed a reference SI joint MR image set by consensus among the study investigators based on the definitions for these 4 types of lesions (online at www.arthritisdoctor.ca). Four videoteleconference sessions involving the 3 university centers focused on the application of the standardized definitions of SI joint MRI lesions and served to calibrate the reader team by conducting 2 test reading exercises on SI joint scans displaying active and structural lesions.

Standardized assessment of MR lesions of the SI joints.

We standardized the approach to assessing MR images of the SI joints by adopting the methodology outlined in the online training module (9).

Online data entry module.

We developed a customized online data entry module for recording MRI findings based on a standardized approach to recording abnormalities in the SI joints, which has 2 sections. The first section contains 3 questions that address global assessment of each scan: 1) “This MRI scan confirms the presence of SpA (agree/disagree),” 2) “Your conclusion is based on which MRI sequence (STIR, T1SE, both sequences),” and 3) “What is the primary MRI feature on which your diagnosis of SpA is based (bone marrow edema, bone erosion, fat infiltration, ankylosis, not applicable as SpA is not present).” The second section of the Web-based data entry module consists of a detailed scoring section where the SI joint is represented as a schematic with 4 quadrants (upper and lower ilium, upper and lower sacrum). Each lesion, except for ankylosis, is recorded as being present/absent on a dichotomous basis in each quadrant. Ankylosis is recorded in each half of the joint (upper and/or lower).

Reading exercises.

We conducted 2 reading exercises. First, the scans of all 187 individuals were evaluated in random order on electronic work stations in the institution of each reader. The same reading exercise was repeated 6 months later on a random selection of 30 AS patients with a symptom duration of ≤5 years and 34 controls (26 NSBP patients and 8 healthy volunteers) from the original study population. The mean ± SD age of the 30 AS patients was 28.6 ± 5.8 years, with a mean ± SD symptom duration of 3.1 ± 1.6 years (70% were men). They showed a mean ± SD Bath Ankylosing Spondylitis Disease Activity Index (11) score of 4.3 ± 2.0 and a mean ± SD Bath Ankylosing Spondylitis Functional Index (12) score of 2.8 ± 2.7; the mean ± SD C-reactive protein level (reference range ≤5 mg/liter) was 14.9 ± 27.2 mg/liter. Twenty-five (89%) of 28 AS patients tested for HLA–B27 were positive. Twenty-six NSBP patients and 8 healthy volunteers had a mean ± SD age of 33.8 ± 7.9 years and 34.3 ± 5.8 years, respectively, with 58% and 88% being men, respectively. Two videoteleconference training sessions between the 2 readouts were directed at recognition of structural abnormalities on T1-weighted images and served to refine a second reference MR image set developed by group consensus that focused on structural lesions. The aim of this second reading exercise was to determine whether training/calibration focused on structural lesions could improve diagnostic utility, particularly for the 2 least experienced readers.

Statistical analysis.

The frequency of SpA patients and controls with specific MRI abnormalities, all 3 structural lesions combined, a structural lesion in the absence of BME, and vice versa, was analyzed descriptively according to single readers and the mean (range) for all 6 readers. For both SpA groups, the contribution of T1-weighted versus STIR sequence MRI and of BME versus structural lesions toward the diagnosis of SpA by MRI was also analyzed descriptively according to single readers, the mean (range) for all readers, and stratified by reader specialty, i.e., radiologist or rheumatologist.

The diagnostic utility of MRI for SpA was determined by calculating sensitivity, specificity, and positive and negative likelihood ratios (LRs) for individual reader data, for diagnoses recorded concordantly by all of the readers, and for diagnoses recorded concordantly by all of the rheumatologists and both of the radiologists. We compared the diagnostic utility before and after training/calibration of readers to recognize structural lesions on T1-weighted MRI. The 6-month postcalibration readout included a descriptive assessment of the relative contribution of T1-weighted versus STIR sequence and of active versus structural MRI lesions to the diagnosis of SpA, which was compared with the precalibration readout.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Descriptive analysis..

Frequency of structural lesions in both SpA and control groups.

Erosions and fat infiltration were recorded by the 6 readers in 80.4% (range 56–95%) and 87.3% (range 85–89%), respectively, of the 75 AS patients (Table 1). Radiologists recorded a higher frequency of erosions compared with rheumatologists (88.7% versus 76.3%). At least 1 of the 3 structural lesions was recorded in 98.2% (range 93–100%) of AS patients, whereas BME was recorded in 81.6% (range 72–89%). Structural lesions in the absence of BME were found in 18.2% of AS patients, whereas BME without structural lesions was virtually absent (1.6%).

Table 1. Frequency of spondylarthritis patients and controls with specific magnetic resonance imaging abnormalities as assessed by 6 readers*
 Any SLErosionFat infiltrationAnkylosisBMESL+/BME−SL−/BME+SL−/BME−SL+/BME+
  • *

    Values are the number (percentage) unless otherwise indicated. SL = structural lesion; BME = bone marrow edema; AS = ankylosing spondylitis; IBP = inflammatory back pain; NSBP = nonspecific back pain.

  • Erosion and/or fat infiltration and/or ankylosis.

  • Less experienced reader.

AS patients (n = 75)
 Radiologist 175 (100)71 (95)64 (85)17 (23)59 (79)16 (21)0 (0)0 (0)59 (79)
 Radiologist 274 (99)62 (83)65 (87)24 (32)64 (85)11 (15)1 (1)0 (0)63 (84)
 Rheumatologist 170 (93)42 (56)65 (87)11 (15)64 (85)11 (15)5 (7)0 (0)59 (79)
 Rheumatologist 274 (99)62 (83)65 (87)13 (17)67 (89)7 (9)0 (0)1 (1)67 (89)
 Rheumatologist 374 (99)60 (80)67 (89)27 (36)59 (79)16 (21)1 (1)0 (0)58 (77)
 Rheumatologist 475 (100)65 (87)67 (89)14 (19)54 (72)21 (28)0 (0)0 (0)54 (72)
 All 6 readers, mean no. (%)73.7 (98.2)60.3 (80.4)65.5 (87.3)17.7 (23.6)61.2 (81.6)13.7 (18.2)1.2 (1.6)0.2 (0.2)60.0 (80.0)
 2 radiologists, mean no. (%)74.5 (99.3)66.5 (88.7)64.5 (86.0)20.5 (27.3)61.5 (82.0)13.5 (18.0)0.5 (0.7)0 (0)61.0 (81.3)
 4 rheumatologists, mean no. (%)73.3 (97.7)57.3 (76.3)66.0 (88.0)16.3 (21.7)61.0 (81.3)13.8 (18.3)1.5 (2.0)0.3 (0.3)59.5 (79.3)
IBP patients (n = 27)
 Radiologist 121 (78)20 (74)10 (37)1 (4)16 (59)6 (22)1 (4)5 (19)15 (56)
 Radiologist 216 (59)14 (52)7 (26)0 (0)19 (70)3 (11)6 (22)5 (19)13 (48)
 Rheumatologist 111 (41)8 (30)9 (33)0 (0)18 (67)1 (4)8 (30)8 (30)10 (37)
 Rheumatologist 217 (63)13 (48)12 (44)0 (0)21 (78)3 (11)7 (26)3 (11)14 (52)
 Rheumatologist 317 (63)10 (37)10 (37)0 (0)19 (70)3 (11)5 (19)5 (19)14 (52)
 Rheumatologist 421 (78)16 (59)13 (48)3 (11)19 (70)5 (19)3 (11)3 (11)16 (59)
 All 6 readers, mean no. (%)17.2 (63.6)13.5 (50.0)10.2 (37.7)0.7 (2.5)18.7 (69.1)3.5 (13.0)5.0 (18.5)4.8 (17.9)13.7 (50.6)
 2 radiologists, mean no. (%)18.5 (68.5)17.0 (63.0)8.5 (31.5)0.5 (1.9)17.5 (64.8)4.5 (16.7)3.5 (13.0)5.0 (18.5)14.0 (51.9)
 4 rheumatologists, mean no. (%)16.5 (61.1)11.8 (43.5)11.0 (40.7)0.8 (2.8)19.3 (71.3)3.0 (11.1)5.8 (21.3)4.8 (17.6)13.5 (50.0)
NSBP controls (n = 26)
 Radiologist 19 (35)6 (23)5 (19)1 (4)3 (12)6 (23)0 (0)17 (65)3 (12)
 Radiologist 24 (15)3 (12)2 (8)0 (0)7 (27)2 (8)5 (19)17 (65)2 (8)
 Rheumatologist 17 (27)0 (0)7 (27)0 (0)6 (23)4 (15)3 (12)16 (62)3 (12)
 Rheumatologist 28 (31)2 (8)7 (27)0 (0)12 (46)3 (12)7 (27)11 (42)5 (19)
 Rheumatologist 35 (19)3 (12)2 (8)1 (4)3 (12)3 (12)1 (4)20 (77)2 (8)
 Rheumatologist 410 (38)7 (27)6 (23)1 (4)3 (12)8 (31)1 (4)15 (58)2 (8)
 All 6 readers, mean no. (%)7.2 (27.6)3.5 (13.5)4.8 (18.6)0.5 (1.9)5.7 (21.8)4.3 (16.7)2.8 (10.9)16.0 (61.5)2.8 (10.9)
 2 radiologists, mean no. (%)6.5 (25.0)4.5 (17.3)3.5 (13.5)0.5 (1.9)5.0 (19.2)4.0 (15.4)2.5 (9.6)17.0 (65.4)2.5 (9.6)
 4 rheumatologists, mean no. (%)7.5 (28.8)3.0 (11.5)5.5 (21.2)0.5 (1.9)6.0 (23.1)4.5 (17.3)3.0 (11.5)15.5 (59.6)3.0 (11.5)
Healthy controls (n = 59)
 Radiologist 120 (34)13 (22)9 (15)0 (0)7 (12)16 (27)3 (5)36 (61)4 (7)
 Radiologist 29 (15)5 (8)6 (10)0 (0)11 (19)5 (8)7 (12)43 (73)4 (7)
 Rheumatologist 18 (14)1 (2)7 (12)0 (0)14 (24)3 (5)9 (15)42 (71)5 (8)
 Rheumatologist 225 (42)7 (12)19 (32)0 (0)19 (32)15 (25)9 (15)25 (42)10 (17)
 Rheumatologist 313 (22)3 (5)10 (17)0 (0)4 (7)12 (20)3 (5)43 (73)1 (2)
 Rheumatologist 419 (32)13 (22)11 (19)0 (0)8 (14)13 (22)2 (3)38 (64)6 (10)
 All 6 readers, mean no. (%)15.7 (26.6)7.0 (11.9)10.3 (17.5)0 (0)10.5 (17.8)10.7 (18.1)5.5 (9.3)37.8 (64.1)5.0 (8.5)
 2 radiologists, mean no. (%)14.5 (24.6)9.0 (15.3)7.5 (12.7)0 (0)9.0 (15.3)10.5 (17.8)5.0 (8.5)39.5 (66.9)4.0 (6.8)
 4 rheumatologists, mean no. (%)16.3 (27.5)6.0 (10.2)11.8 (19.9)0 (0)11.3 (19.1)10.8 (18.2)5.8 (9.7)37.0 (62.7)5.5 (9.3)

At least 1 structural lesion was recorded in 63.6% (range 41–78%) of IBP patients by the 6 readers, whereas BME was recorded in 69.1% (range 59–78%). MRI lesions were not recorded on either sequence in 17.9% (range 11–30%) of the 27 IBP patients. Erosions were recorded in 50% of IBP patients and radiologists also observed erosions more frequently than rheumatologists in this patient group (63% versus 43.5%), whereas BME was recorded more frequently by the rheumatologists (71.3% versus 64.8%). A structural lesion in the absence of BME and vice versa was observed, with a comparable frequency of 13% and 18.5%, respectively.

In both control groups, any structural lesion was observed in 27% of individuals. However, whereas BME was observed in 21.8% (range 12–46%) and 17.8% (range 7–32%) of NSBP patients and healthy controls, respectively, erosions were recorded in 13.5% (range 0–27%) and 11.9% (range 2–22%) of the corresponding controls. BME with and without structural lesions was recorded in 10.9% and 10.9% in the NSBP patients and in 8.5% and 9.3% of the healthy volunteers, respectively.

Relative importance of T1-weighted versus STIR sequence and BME versus structural lesions.

T1-weighted images alone or together with STIR images were indicated as the most important in concluding that the MRI was indicative of SpA in 97% of AS patients by the radiologists and in 74% by the rheumatologists (Table 2). For diagnosing SpA by MRI in the IBP group, the T1-weighted sequence, either alone or in combination with STIR, was still considered more important than the STIR sequence by radiologists for 67% of patients, whereas STIR alone was considered more important by rheumatologists for 70% of IBP patients. In regard to the relative importance of the 4 different MRI lesions, BME was considered most important in 26% of the AS patients by the radiologists and in 68% of the AS patients by the rheumatologists; erosions on T1-weighted images were recognized as most important in 46% by the radiologists and in 10% by the rheumatologists. In contrast, BME was indicated as the most important lesion in 81% of IBP patients by the radiologists and in 90% of IBP patients by the rheumatologists. Erosions were considered most important in 11% and 4% of the IBP patients by radiologists and rheumatologists, respectively.

Table 2. Relative contribution of T1-weighted versus STIR sequence MRI and the relative contribution of active versus structural MRI lesions in those patients diagnosed with SpA by MRI*
 No. of patients diagnosed as SpAMRI sequence considered most importantMRI lesion considered most important
STIRT1-weightedBoth STIR and T1-weightedBMEErosionFat infiltrationAnkylosisNA
  • *

    Values are the number (percentage) unless otherwise indicated. MRI = magnetic resonance imaging; SpA = spondylarthritis; BME = bone marrow edema; NA = not applicable (no lesion type predominant); AS = ankylosing spondylitis; IBP = inflammatory back pain; ND = not done.

  • Less experienced reader.

  • Mean of 6 readers.

  • §

    Mean of 5 readers.

Combined SpA patient group (AS and IBP patients, n = 102)
 Radiologist 1874 (5)38 (44)45 (52)17 (20)52 (60)3 (3)14 (16)1 (1)
 Radiologist 2809 (11)9 (11)62 (78)42 (53)16 (20)11 (14)11 (14)0 (0)
 Rheumatologist 17838 (49)22 (28)18 (23)NDNDNDNDND
 Rheumatologist 28617 (20)2 (2)67 (78)63 (73)7 (8)4 (5)12 (14)0 (0)
 Rheumatologist 37638 (50)15 (20)23 (30)55 (72)9 (12)0 (0)12 (16)0 (0)
 Rheumatologist 49220 (22)15 (16)57 (62)65 (71)6 (7)8 (9)12 (13)1 (1)
 Mean no. (%) 21.0 (25.3)16.8 (20.2)45.3 (54.5)48.4 (57.5)18 (21.4)5.2 (6.2)12.2 (14.5)0.4 (0.5)
AS patients (n = 75)
 Radiologist 1741 (1)36 (49)37 (50)8 (11)49 (66)3 (4)14 (19)0 (0)
 Radiologist 2663 (5)9 (14)54 (82)29 (44)16 (24)10 (15)11 (17)0 (0)
 Rheumatologist 16426 (41)21 (33)17 (27)NDNDNDNDND
 Rheumatologist 2728 (11)2 (3)62 (86)50 (69)6 (8)4 (6)12 (17)0 (0)
 Rheumatologist 36226 (42)14 (23)22 (35)42 (68)8 (13)0 (0)12 (19)0 (0)
 Rheumatologist 4719 (13)14 (20)48 (68)47 (66)6 (8)6 (8)12 (17)0 (0)
 Mean no. (%) 12.2 (17.8)16 (23.5)40 (58.7)35.2 (51.0)17 (24.6)4.6 (6.7)12.2 (17.7)0 (0)
IBP patients (n = 27)
 Radiologist 1133 (23)2 (15)8 (62)9 (69)3 (23)0 (0)0 (0)1 (8)
 Radiologist 2146 (43)0 (0)8 (57)13 (93)0 (0)1 (7)0 (0)0 (0)
 Rheumatologist 11412 (86)1 (7)1 (7)NDNDNDNDND
 Rheumatologist 2149 (64)0 (0)5 (36)13 (93)1 (7)0 (0)0 (0)0 (0)
 Rheumatologist 31412 (86)1 (7)1 (7)13 (93)1 (7)0 (0)0 (0)0 (0)
 Rheumatologist 42111 (52)1 (5)9 (43)18 (86)0 (0)2 (10)0 (0)1 (5)
 Mean no. (%) 8.8 (58.9)0.8 (5.6)5.3 (35.6)§13.2 (86.8)1 (6.6)0.6 (3.9)0 (0)0.4 (2.6)

Impact of calibration for T1-weighted abnormalities on diagnostic utility of MRI.

Precalibration, the mean sensitivity and specificity of MRI for a diagnosis of SpA in 30 AS patients chosen randomly from the original cohort and 34 controls (26 with NSBP and 8 healthy) were 92% and 94%, respectively, whereas the mean positive and negative LRs were 14.5 and 0.08, respectively (Table 3). Postcalibration, the mean sensitivity and specificity of MRI improved to 98% and 96%, respectively, whereas the mean positive and negative LRs improved to 22.2 and 0.02, respectively. In the 4 experienced readers, sensitivity increased in 1 reader at the cost of a slight decrease in specificity. In 1 of the 2 less experienced readers, sensitivity markedly increased from 77% to 97%, with a slight decrease in specificity (100% versus 94%); the second reader with limited experience improved in both sensitivity (93% versus 100%) and specificity (79% versus 97%).

Table 3. Comparison of diagnostic utility of MRI for SpA before and after training/calibration of readers to recognize structural lesions on T1-weighted MRI*
 PrecalibrationPostcalibration
SensitivitySpecificityPositive LRNegative LRSensitivitySpecificityPositive LRNegative LR
  • *

    MRI = magnetic resonance imaging; SpA = spondylarthritis; LR = likelihood ratio; NC = not calculable (sensitivity or specificity of 1.0).

  • Postcalibration MRI readings were performed on a subgroup of 30 ankylosing spondylitis patients and 34 controls (26 patients with nonspecific back pain and 8 healthy volunteers).

  • Less experienced reader.

Radiologist 10.97 (29/30)0.97 (33/34)32.90.030.97 (29/30)1.0 (34/34)NC0.03
Radiologist 20.90 (27/30)1.0 (34/34)NC0.100.97 (29/30)0.97 (33/34)32.90.03
Rheumatologist 10.97 (29/30)0.94 (32/34)16.40.040.97 (29/30)0.88 (30/34)8.20.04
Rheumatologist 21.0 (30/30)0.91 (31/34)11.3NC1.0 (30/30)0.97 (33/34)34.0NC
Rheumatologist 30.77 (23/30)1.0 (34/34)NC0.230.97 (29/30)0.94 (32/34)16.40.04
Rheumatologist 40.93 (28/30)0.79 (27/34)4.50.081.0 (30/30)0.97 (33/34)34.0NC
Mean (range) for all 6 readers0.92 (0.77–1.0)0.94 (0.79–1.0)14.50.080.98 (0.97–1.0)0.96 (0.88–1.0)22.20.02
All 6 readers concordantly0.67 (20/30)0.76 (26/34)2.80.440.90 (27/30)0.82 (28/34)5.10.12

Concordance for the diagnosis of SpA by any 2 readers was 100% for both the pre- and postcalibration readout. Agreement regarding the diagnosis of SpA among all 6 readers in the precalibration readout was 67% (20 of 30), and for the absence of SpA in the control group was 76% (26 of 34). Postcalibration agreement for the diagnosis of SpA among all 6 readers increased to 90% (27 of 30 AS patients), and for the absence of SpA in controls increased to 82% (28 of 34).

Table 4 shows the contribution of T1-weighted versus STIR sequence toward the diagnosis of SpA. Precalibration, the T1-weighted sequence, alone or in combination with the STIR sequence, was considered most important in 95% of AS patients by the radiologists and in 69% of AS patients by the rheumatologists. Postcalibration, this did not change for radiologists, but for rheumatologists, the T1-weighted sequence, alone or with STIR, was now considered most important in 85% of patients. This difference in approach to diagnostic ascertainment by rheumatologists was also reflected in the analysis of the type of MRI lesions considered most important (Table 4). Precalibration, structural lesions were considered most important in 53% of patients by radiologists compared with 23% of patients by rheumatologists. Postcalibration, structural lesions were considered most important in 72% of patients by radiologists versus 53% of patients by rheumatologists.

Table 4. Relative contribution of T1-weighted versus STIR sequence MRI and the relative contribution of active versus structural MRI lesions to the diagnosis of SpA by MRI before and after training/calibration of readers to recognize structural lesions on T1-weighted MRI*
 MRI sequence considered most importantMRI lesion considered most important
Pre STIRPre T1Pre bothPost STIRPost T1Post bothPre BMEPre erosionPre FIPre ankPre NAPost BMEPost erosionPost FIPost ankPost NA
  • *

    Values are the number (percentage). MRI = magnetic resonance imaging; SpA = spondylarthritis; pre = precalibration; post = postcalibration; BME = bone marrow edema; FI = fat infiltration; ank = ankylosis; NA = not applicable (no lesion type predominant); ND = not done.

  • Less experienced reader.

  • Four rheumatologists.

  • §

    Three rheumatologists.

  • Six readers.

  • #

    Five readers.

Radiologist 11 (3)10 (33)19 (63)1 (3)13 (43)16 (53)8 (27)18 (60)1 (3)3 (10)0 (0)6 (20)19 (63)0 (0)3 (10)2 (7)
Radiologist 22 (7)2 (7)26 (87)1 (3)1 (3)28 (93)18 (60)5 (17)3 (10)2 (7)2 (7)8 (27)17 (57)0 (0)4 (13)1 (3)
Rheumatologist 114 (47)7 (23)9 (30)9 (30)9 (30)12 (40)NDNDNDNDNDNDNDNDNDND
Rheumatologist 23 (10)0 (0)27 (90)3 (10)0 (0)27 (90)23 (77)2 (7)3 (10)2 (7)0 (0)20 (67)7 (23)1 (3)2 (7)0 (0)
Rheumatologist 315 (50)3 (10)12 (40)5 (17)3 (10)22 (73)23 (77)2 (7)0 (0)2 (7)3 (10)18 (60)9 (30)0 (0)2 (7)1 (3)
Rheumatologist 45 (17)3 (10)22 (73)1 (3)6 (20)23 (77)23 (77)2 (7)2 (7)3 (10)0 (0)3 (10)23 (77)0 (0)4 (13)0 (0)
Radiologists3 (5)12 (20)45 (75)2 (3)14 (23)44 (73)26 (43)23 (38)4 (7)5 (8)2 (3)14 (23)36 (60)0 (0)7 (12)3 (5)
Rheumatologists37 (31)13 (11)70 (58)18 (15)18 (15)84 (70)69 (77)6 (7)5 (6)7 (8)3 (3)41 (46)39 (43)1 (1)8 (9)1 (1)§
All readers40 (22)25 (14)115 (64)20 (11)32 (18)128 (71)95 (63)29 (19)9 (6)12 (8)5 (3)55 (37)75 (50)1 (1)15 (10)4 (3)#

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Our systematic and controlled evaluation of lesions visible on MRI in patients with early SpA has demonstrated several findings of clinical relevance to the diagnostic assessment of SpA by MRI. First, structural lesions such as erosion and fat infiltration are observed relatively frequently in early SpA, even in patients with a duration of symptoms of only 2 years. Second, information on structural lesions evident on T1-weighted sequences contributes substantially to diagnostic utility and comparably with BME on the STIR sequence. Third, rheumatologists rely primarily on the diagnostic information provided by the STIR sequence and appear to be less familiar with the contribution of T1-weighted abnormalities such as erosions, which radiologists consider important to the diagnosis of SpA by MRI. Fourth, rheumatologist training directed specifically at detection of MRI lesions on T1-weighted scans improves the diagnostic utility of MRI.

Since first reports in the early 1990s on the use of MRI to detect sacroiliitis (13), MRI has gained widespread acceptance as the most sensitive imaging modality for diagnostic evaluation of patients with suspected early axial SpA. Most prior studies on the diagnostic utility of MRI in early SpA patients (1–5) have focused on active sacroiliitis lesions found on fat-suppressed or contrast-enhanced T1-weighted sequences. A study of 68 patients with recent-onset IBP showed structural SI joint lesions as defined by erosion (irregularly delineated joint space on T1-weighted sequence), sclerosis (low signal intensity on T1-weighted, STIR, and T2 fast spin-echo sequences, without gadolinium enhancement), or ankylosis (disappearance of the joint space in all sequences) in 16% of the patients as opposed to inflammation in 32% (14). The concordance rates between 2 calibrated readers for structural lesions and for inflammation were 0.85 and 0.82, respectively. Fat infiltration was not addressed in this report. However, compared with our work, the structural lesions were assessed globally for the entire right and left SI joints and there was no control group. In an evaluation of 41 early SpA patients, erosions were observed in 80%, sclerosis in 71%, and fat deposition in 63% of the patients, with an interobserver agreement between 2 senior radiologists of 71–84% for detection of these structural lesions (15). This study was based on a quadratic approach, separating the cartilaginous and ligamentous portion of the SI joints, but there was no control group. In our study, erosions were observed in 50% of 27 preradiographic IBP patients with a mean symptom duration of 29 months. This finding indicates that structural damage of the SI joints that is not detectable by radiography is detectable by MRI and may start very early in the disease course.

Formulating a single standardized definition for a reliable assessment of erosions by MRI is complex. A similar definition of erosion has recently been published that also included nonenhanced T1-weighted fat-saturated and gradient echo sequences delineating the joint space from subchondral bone (16). These “cartilage sequences” may offer advantages, and they require further studies regarding their contribution to the diagnostic utility in SpA. Erosions detected in 11.9% of healthy controls in our study may be explained by these technical conditions or they may represent physiologic variants. The smaller the structural lesions meeting the definition of erosion, the more difficult the differentiation is from physiologic variants, e.g., the insertion of ligaments, and this may contribute to their detection in controls. The erosion definition for SI joints should be validated further, according to the principles suggested by the Outcome Measures in Rheumatology Clinical Trials, also investigating whether a definition requiring 2 planes would be useful.

Prior to specific calibration for detection of structural lesions, rheumatologists based their diagnostic conclusion for both AS and preradiographic IBP more frequently on the STIR images alone than radiologists (34% versus 8%) and deemphasized the contribution of abnormalities on the T1-weighted sequence, which the radiologists considered important to the diagnosis of SpA by MRI. This difference was associated with a greater recognition of the importance of erosions on T1-weighted images by the radiologists compared with the rheumatologists (41% versus 9%). This may be due to the fact that the principal focus of study of MRI for SpA in the rheumatology literature has been active lesions as recognized on STIR sequences. For example, several methods have been developed and validated to score the degree of SI joint and spinal inflammation based on the assessment of BME on STIR sequences (17, 18). Recent interventional studies in SpA describing pivotal trials of anti–tumor necrosis factor agents have focused almost entirely on their impact on active lesions (19, 20). In particular, diagnostic studies of the SI joints using MRI have focused primarily on active lesions with assessment of structural lesions being exploratory in nature, and not preceded by the development of lesion definitions that facilitate standardization and comparability across studies. The crucial importance of this preliminary step has been advocated in a recent report by the ASAS that described the spectrum of abnormalities observed in the SI joints on MRI (6). This group of experts achieved consensus on a definition for BME in the SI joints that then led to the formulation of a proposal for a positive MRI indicative of SpA based on the detection of BME. Our data-driven approach has shown that not only are structural lesions frequent, but also that they may occur in 13–18% of patients with early SpA in the absence of BME, and training focused on their recognition can improve the diagnostic utility of MRI.

The online recording module was designed to first capture the reader's diagnosis according to a global assessment of the MRI. It is possible that diagnostic ascertainment may have been different if the more detailed recording of active and structural lesions had been conducted first. However, this approach would lack external validity because clinical practice is based on global evaluation of STIR and T1-weighted scans. Subsequent questions in the data entry module were aimed at understanding what formed the basis for the reader's decision to score the MRI as indicative of SpA or not. The 2 senior musculoskeletal radiologists participating in our study both have longstanding experience in assessing MRI scans of patients with SpA, and it was therefore instructive that their diagnostic ascertainment was so markedly dependent on assessment of T1-weighted images and structural lesions. However, it is also likely that their emphasis on structural MR lesions of the SI joints may not be representative of the wider spectrum of radiologists who evaluate MRI scans for SpA. Nevertheless, MRI is increasingly recognized as the preferred imaging modality for making a diagnosis of early SpA; it will be essential for both specialties to develop their diagnostic skills in interpreting MRI of the SI joints. Our data show that training courses will need to focus on the recognition of structural as well as active lesions.

A limitation of our study design is that our data do not allow us to determine whether the enhanced diagnostic utility after training was simply due to improved overall interpretation of the MRI scans as opposed to improved recognition of structural lesions per se.

In conclusion, this systematic, standardized, and controlled evaluation of T1-weighted and STIR sequences demonstrates that structural abnormalities occur frequently in the SI joints of patients with early SpA, that they may occur in the absence of BME, and that they contribute substantially to the diagnostic utility of MRI in early SpA. Rheumatologist training directed at detection of structural lesions, and especially erosions, on T1-weighted scans improves the diagnostic utility of MRI. Further study in larger populations of patients with early SpA using this study design is necessary to clarify the diagnostic value of structural lesions on MRI. However, it is already clear that further educational interventions should be aimed at achieving a much greater understanding of the information provided by the T1-weighted scan.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Weber had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Weber, Lambert, Pedersen, Hodler, Østergaard, Maksymowych.

Acquisition of data. Weber, Lambert, Pedersen, Hodler, Østergaard, Maksymowych.

Analysis and interpretation of data. Weber, Lambert, Pedersen, Hodler, Østergaard, Maksymowych.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

The authors thank the patients and the healthy volunteers for their participation; Tracey Clare, Clinical Research Manager, and Paul Filipow, Data Manager, Department of Radiology, University of Alberta, Edmonton, Canada, for coordinating the Web-based Spondyloarthritis Research Consortium of Canada scoring index; Christian Streng, Medical Documentation, Balgrist University Hospital, Zurich, Switzerland, for his technical assistance with Figure 1; and Rudolf O. Kissling, MD, Department of Rheumatology, Balgrist University Hospital, Zurich, Switzerland, for scoring the SI joints (Balgrist patients) on pelvic radiographs.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  • 1
    Braun J, Bollow M, Eggens U, Konig H, Distler A, Sieper J. Use of dynamic magnetic resonance imaging with fast imaging in the detection of early and advanced sacroiliitis in spondylarthropathy patients. Arthritis Rheum 1994; 37: 103945.
  • 2
    Hanly JG, Mitchell MJ, Barnes DC, MacMillan L. Early recognition of sacroiliitis by magnetic resonance imaging and single photon emission computed tomography. J Rheumatol 1994; 21: 208895.
  • 3
    Bollow M, Braun J, Hamm B, Eggens U, Schilling A, Koenig H, et al. Early sacroiliitis in patients with spondyloarthropathy: evaluation with dynamic gadolinium-enhanced MR imaging. Radiology 1995; 194: 52936.
  • 4
    Blum U, Buitrago-Tellez C, Mundinger A, Krause T, Laubenberger J, Vaith P, et al. Magnetic resonance imaging (MRI) for detection of active sacroiliitis: a prospective study comparing conventional radiography, scintigraphy, and contrast enhanced MRI. J Rheumatol 1996; 23: 210715.
  • 5
    Bredella MA, Steinbach LS, Morgan S, Ward M, Davis JC. MRI of the sacroiliac joints in patients with moderate to severe ankylosing spondylitis. AJR Am J Roentgenol 2006; 187: 14206.
  • 6
    Rudwaleit M, Jurik AG, Hermann KG, Landewe R, van der Heijde D, Baraliakos X, et al. Defining active sacroiliitis on magnetic resonance imaging (MRI) for classification of axial spondyloarthritis: a consensual approach by the ASAS/OMERACT MRI group. Ann Rheum Dis 2009; 68: 15207.
  • 7
    Van der Linden S, Valkenburg HA, Cats A. Evaluation of diagnostic criteria for ankylosing spondylitis: a proposal for modification of the New York criteria. Arthritis Rheum 1984; 27: 3618.
  • 8
    Kuorinka I, Jonsson B, Kilbom A, Vinterberg H, Biering-Sorensen F, Andersson G, et al. Standardised Nordic questionnaires for the analysis of musculoskeletal symptoms. Appl Ergon 1987; 18: 2337.
  • 9
    Maksymowych WP, Dhillon SS, Chiowchanwisawakit P, Pedersen SJ, Martinez B, Ostergaard M, et al. Development and validation of web-based training modules for systematic evaluation of active inflammatory lesions in the spine and sacroiliac joints in spondyloarthritis. J Rheumatol 2009; 36 Suppl 84: 4857.
  • 10
    Maksymowych WP, Inman RD, Salonen D, Dhillon SS, Williams M, Stone M, et al. Spondyloarthritis Research Consortium of Canada magnetic resonance imaging index for assessment of sacroiliac joint inflammation in ankylosing spondylitis. Arthritis Rheum 2005; 53: 7039.
  • 11
    Garret S, Jenkinson T, Kennedy LG, Whitelock H, Gaisford P, Calin A. A new approach to defining disease status in ankylosing spondylitis: the Bath Ankylosing Spondylitis Disease Activity Index. J Rheumatol 1994; 21: 228691.
  • 12
    Calin A, Garrett S, Whitelock H, Kennedy LG, O'Hea J, Mallorie P, et al. A new approach to defining functional ability in ankylosing spondylitis: the development of the Bath Ankylosing Spondylitis Functional Index. J Rheumatol 1994; 21: 22815.
  • 13
    Ahlstrom H, Feltelius N, Nyman R, Hallgren R. Magnetic resonance imaging of sacroiliac joint inflammation. Arthritis Rheum 1990; 33: 17639.
  • 14
    Heuft-Dorenbosch L, Weijers R, Landewe R, van der Linden S, van der Heijde D. Magnetic resonance imaging changes of sacroiliac joints in patients with recent-onset inflammatory back pain: inter-reader reliability and prevalence of abnormalities. Arthritis Res Ther 2006; 8: R11.
  • 15
    Puhakka KB, Jurik AG, Egund N, Schiottz-Christensen B, Stengaard-Pedersen K, van Overeem Hansen G, et al. Imaging of sacroiliitis in early seronegative spondylarthropathy: assessment of abnormalities by MRI in comparison with radiography and CT. Acta Radiol 2003; 44: 21829.
  • 16
    Madsen KB, Jurik AG. Magnetic resonance imaging grading system for active and chronic spondyloarthritis changes in the sacroiliac joint. Arthritis Care Res (Hoboken) 2010; 62: 118.
  • 17
    Landewe RB, Hermann KG, van der Heijde DM, Baraliakos X, Jurik AG, Lambert RG, et al. Scoring sacroiliac joints by magnetic resonance imaging: a multiple-reader reliability experiment. J Rheumatol 2005; 32: 20505.
  • 18
    Van der Heijde DM, Landewe RB, Hermann KG, Jurik AG, Maksymowych WP, Rudwaleit M, et al. Application of the OMERACT filter to scoring methods for magnetic resonance imaging of the sacroiliac joints and the spine: recommendations for a research agenda at OMERACT 7. J Rheumatol 2005; 32: 20427.
  • 19
    Lambert RG, Salonen D, Rahman P, Inman RD, Wong RL, Einstein SG, et al, for the M03–606 Study Group. Adalimumab significantly reduces both spinal and sacroiliac joint inflammation in patients with ankylosing spondylitis: a multicenter, randomized, double-blind, placebo-controlled study. Arthritis Rheum 2007; 56: 400514.
  • 20
    Barkham N, Keen HI, Coates LC, O'Connor P, Hensor E, Fraser AD, et al. Clinical and imaging efficacy of infliximab in HLA–B27–positive patients with magnetic resonance imaging–determined early sacroiliitis. Arthritis Rheum 2009; 60: 94654.