Magnetic resonance imaging assessment of spinal inflammation in ankylosing spondylitisStandard clinical protocols may omit inflammatory lesions in thoracic vertebrae.код для вставкиСкачать
Arthritis & Rheumatism (Arthritis Care & Research) Vol. 61, No. 9, September 15, 2009, pp 1187–1193 DOI 10.1002/art.24561 © 2009, American College of Rheumatology ORIGINAL ARTICLE Magnetic Resonance Imaging Assessment of Spinal Inﬂammation in Ankylosing Spondylitis: Standard Clinical Protocols May Omit Inﬂammatory Lesions in Thoracic Vertebrae WINSTON J. RENNIE,1 SUHKVINDER S. DHILLON,2 BARBARA CONNER-SPADY,2 WALTER P. MAKSYMOWYCH,2 AND ROBERT G. W. LAMBERT2 Objective. Radiologic assessment of spinal inﬂammation in patients with ankylosing spondylitis (AS) relies primarily on magnetic resonance imaging (MRI), although little is known about the distribution of inﬂammatory lesions within the structures of the spine. Our objective was to compare the distribution of inﬂammatory lesions centrally and laterally within the thoracic and lumbar spine vertebral bodies. Methods. We studied 49 patients with AS who were scanned with STIR and T1-weighted spin-echo MRI of the whole spine. Scans were read by 2 musculoskeletal radiologists, with a third reader as the arbitrator. Controls included 6 age-matched individuals. We recorded bone marrow edema on STIR images from each vertebral body, separately identifying central and lateral slices. The latter were deﬁned as images that included or were lateral to the pedicle. Interreader reproducibility was assessed by kappa statistics. Results. Inﬂammation was present in 263 (45%) of 588 thoracic and 86 (35%) of 245 lumbar vertebrae; the mean number of affected thoracic and lumbar vertebrae per patient were 5.4 and 1.8, respectively. Inﬂammation was present in the lateral aspect of 219 (37%) of 588 thoracic vertebrae and 45 (18%) of 245 lumbar vertebrae (P < 0.001). Lesions were more common laterally than centrally for all thoracic vertebrae except for T7. Involvement of only the lateral slices was observed in as many as 19.6% of thoracic vertebrae. Conclusion. Evaluation of spinal inﬂammation by MRI may omit lesions in up to 20% of inﬂamed thoracic vertebrae if both scanning and image assessment do not include sagittal slices that extend to the lateral edges of all vertebrae. INTRODUCTION Magnetic resonance imaging (MRI) is a sensitive imaging modality for detection of inﬂammatory lesions in patients with ankylosing spondylitis (AS) (1). Detailed histopathologic analyses of the sacroiliac (SI) joints and spine have Mr. Maksymowych is recipient of a Scientist Award from the Alberta Heritage Foundation for Medical Research. 1 Winston J. Rennie, MB, FRCR: University Hospitals of Leicester, Leicester Royal Inﬁrmary, Leicester, UK; 2 Suhkvinder S. Dhillon, MB, FRCR, Barbara Conner-Spady, PhD, Walter P. Maksymowych, MB, FRCPC, Robert G. W. Lambert, MB, FRCPC: University of Alberta, Edmonton, Alberta, Canada. Mr. Rennie is employed for honorary consultancy and has received speaking fees (less than $10,000) from DePuy (UK). Address correspondence to Robert G. W. Lambert, MB, FRCPC, Department of Radiology and Diagnostic Imaging, University of Alberta, 2A2.41WMC, 8440-112 Street, Edmonton, Alberta, Canada T5T 1E3. E-mail: email@example.com. Submitted for publication April 15, 2008; accepted in revised form June 11, 2009. shown that inﬂammatory lesions are particularly evident within bone marrow beneath articular cartilage and adjacent to the attachment of ligaments and capsules to bone (2). These regions are typically not amenable to biopsy, which has restricted progress in our understanding of the pathophysiology of disease and therefore the development of new therapeutic agents. Biopsy of the costovertebral joint is not feasible, and arthropathy of this joint is particularly difﬁcult to identify with radiography, although it is a well-recognized entity in AS causing thoracic spine pain (3). Limitation in chest expansion due to costovertebral joint involvement is an important diagnostic criterion in the widely used modiﬁed New York criteria (4). Erosions, sclerosis, and bony proliferation have been reported within the costovertebral joint in AS, more pronounced on the vertebral side of the joint (5–7). The use of T2-weighted MR sequences that suppress the signal from marrow fat (STIR, T2 with fat suppression) allows detection of increased water signal, which is associated with active inﬂammation in bone marrow. These sequences are now routinely used to image the spine and 1187 1188 SI joint, providing a noninvasive approach to the evaluation of inﬂammatory lesions within bone marrow and soft tissues of the entire axial skeleton. Scoring systems have also been developed to quantify the extent of inﬂammatory change in the spine (8,9). In one report, MRI evaluation of inﬂammation in the spine showed that the majority of inﬂammatory lesions in patients with established disease were observed at the thoracic level (10). However, this evaluation focused on anterior spondylitis and spondylodiscitis, and foci of inﬂammation in the more lateral aspects of the vertebrae that include the costovertebral joint may not have been fully assessed. Complete nonvisualization of pathology at the lateral edge of vertebrae may occur when scoliosis is present because typical scanning protocols require that sagittal sequences depict the entire spinal canal and the majority of the vertebra/disc at each level, but do not specify the need to include the lateral edges of the vertebral bodies. This may have important implications for the approach to diagnostic evaluation as well as the assessment of treatment-refractory disease. In this study we aimed to assess the frequency of involvement of lateral structures of the spine in patients with AS by systematic examination of the extent and distribution of inﬂammatory lesions in the vertebral bodies using MRI. SUBJECTS AND METHODS Subjects. We studied 49 patients diagnosed with AS according to the modiﬁed New York criteria (4). All patients had been recruited to a prospective, longitudinal observational cohort, the Followup Research Cohort of AS study, in which clinical and laboratory data are systematically collected every 6 months, and plain radiographic imaging and MRI are obtained annually. Most patients (83%) received nonsteroidal antiinﬂammatory drugs and/or physical modalities of therapy. Of the 49 patients with AS, 35 (71%) were male, with a mean ⫾ SD age of 42.3 ⫾ 11.7 years (range 21–70 years) and a mean ⫾ SD disease duration of 19.4 ⫾ 11.4 years (range 2– 42 years). The mean Bath AS Disease Activity Index score was 4.7 (range 2.7– 8.8) and mean Bath AS Functional Index score was 4.1 (range 2.2– 8.4). We also included 6 controls, 3 of whom had a history of nonspeciﬁc back pain. The controls had a mean age of 36 years (range 24 – 47 years) and 4 were male. All controls were imaged with an MRI protocol identical to that for the patients with AS. The study was approved by the University of Alberta Ethics Committee and all patients provided informed consent. MRI. All MRI of the spine was performed with 1.5T systems (Siemens, Erlangen, Germany). Sagittal sequences were obtained with 3– 4 mm slice thickness and 11–15 slices acquired. Sequence parameters were 1) T1-weighted spin-echo (time to recovery [TR] 517– 618 msec, time to echo [TE] 13 msec), and 2) STIR (TR 3,000 –3,170 msec, time to inversion 140 msec, TE 38 – 61 msec). The ﬁeld of view was 380 – 400 mm and the matrix was 512 ⫻ 256 pixels. The spine was imaged in 2 parts: the upper half, comprising the entire cervical and most of the thoracic spine, Rennie et al Figure 1. The transverse anatomy of the mid-thoracic spine illustrated with superimposed lines representing the location of 11 slices in a typical sagittal STIR magnetic resonance imaging sequence. The drawing includes minimal asymmetry and rotation of the vertebra that is within normal limits. Central sagittal images include the spinal canal (solid lines C). Lateral images include the pedicle (solid lines L). In this study, one image on either side immediately medial to the completely visualized pedicle at each vertebral level was excluded from scoring, and images lateral to the vertebral body were not evaluated (dotted lines). Note that the left lateral edge of the vertebral body and the transverse processes would not be included on this sagittal sequence. and the lower half, comprising the lower portion of the thoracic spine and entire lumbar spine. The speciﬁc MRI parameters for acquiring spine images are provided on the author’s Website (http://www.arthritisdoctor.ca/spine). MRI reading exercises. A unique MRI study number was allocated to each patient and control, thereby ensuring blinding to all patient demographics. Allocation was done by a technologist unconnected with the study using computer-generated random numbers. Assessment was performed on a 3-monitor review station by 2 readers (WJR, SSD) using computer software that is optimal for this type of review (eFilm; Merge, Milwaukee, WI), and viewing conditions were standardized. Each subject was only identiﬁed by the MRI study number, and scans were read in random order by 2 fellowship-trained musculoskeletal radiologists, with a third musculoskeletal radiologist reader as arbitrator (RGWL). Readers were unaware of the number of control subjects included in the review. Deﬁnitions were ﬁnalized after training. This consisted of an initial pilot study of 15 scans of patients with AS read by 2 observers and by consensus. The pilot results were then discarded and these scans were not used in the main study. Increased bone marrow signal denoting inﬂammation in vertebral bodies was assessed on STIR sequences. Bone marrow that was positive for inﬂammation was deﬁned as an appearance where the bone marrow Assessment of Spinal Inﬂammation in AS with MRI 1189 Table 1. Frequency and distribution (by subject) of inﬂammatory lesions in the spine of 49 patients with AS and 6 controls* Group, spinal segment, and slice location Subjects with abnormalities, no. Effected levels per patient, mean ⴞ SD 16 0.6 ⫾ 1.1 0 (0–1) 0–5 37 32 34 5.4 ⫾ 4.3 3.0 ⫾ 3.2 4.5 ⫾ 4.1 5 (0.5–9) 2 (0–5.5) 4 (0–8.5) 0–12 0–11 0–12 28 26 21 1.8 ⫾ 1.9 1.5 ⫾ 1.7 0.9 ⫾ 1.4 1 (0–3) 1 (0–3) 0 (0–1) 0–5 0–5 0–5 0 2 4 0 0.3 1.2 Patients with AS Cervical, any Thoracic Any Central Lateral Lumbar Any Central Lateral Controls Cervical, any Thoracic, any Lumbar, any Effected levels per patient, median (IQR) Range 0 0 1 0 0–1 0–3 * AS ⫽ ankylosing spondylitis; IQR ⫽ interquartile range. signal was greater than the signal from the center of an adjacent normal vertebral body. We recorded inﬂammatory signal (as a dichotomous yes/no) on STIR images from each vertebral body, and data sheets were designed to separately identify inﬂammatory change from central and lateral slices (Figure 1). Abnormalities in the pedicles, other posterior elements of the vertebrae, or ribs were not evaluated. Only the presence of vertebral body inﬂammation was recorded, and scoring did not include assessment of lesion size, shape, or intensity of signal. Readers were instructed to not distinguish between various possible causes of inﬂammation. However, other lesions that may be bright on T2-weighted sequences, such as hemangioma, were excluded. At each level of the thoracic and lumbar spine, the sagittal images were deﬁned as either a lateral slice (the image of the vertebral body included or was lateral to the pedicle) or a central slice (the image of the vertebral body included the spinal canal or lateral recess). Often there is one slice at the medial cortex of the pedicle that is also partly through the lateral recess. To avoid variability of assignation of this transitional slice, one slice on the left side and one on the right side were discarded from the analysis, deﬁned as follows: the slice medial to the most medial image clearly depicting the full length of the pedicle. It was not possible to apply these deﬁnitions to the cervical spine, where the vertebral bodies are small with a proportionately wider spinal canal, and assessment of cervical vertebrae did not distinguish between central and lateral images. At each level, the number of central slices varied mostly according to the size of the vertebra, and the number of lateral slices varied according to the presence of scoliosis. The central sections consisted of a minimum of 3 and a maximum of 5 slices, and the lateral sections consisted of a minimum of 0 and maximum of 3 slices for each side, with a combined minimum of 2 and maximum of 6 lateral slices at each vertebral level. Statistical analysis. The distribution of inﬂammatory lesions according to spinal segment, vertebral level, and lateral versus central slices was analyzed descriptively and by Fisher’s exact test. Interobserver reliability was assessed using kappa statistics. Kappa values of ⬎0.75, 0.40 – 0.75, and ⬍0.40 were designated as representing excellent, fair to good, and poor reliability, respectively. Table 2. Relative frequency and distribution (by vertebra) of inﬂammatory lesions in central and lateral slices of the thoracolumbar spine in 49 patients with AS* Location of inﬂammation Thoracic vertebrae (n ⴝ 588) Lumbar vertebrae (n ⴝ 245) OR (95% CI) P (Fisher’s exact) Any slice Central slice Lateral slice Central and lateral slices affected Central slice only Lateral slice only No inﬂammation 263 (44.7) 148 (25.1) 219 (37.3) 104 (17.7) 44 (7.4) 115 (19.6) 325 (55.3) 86 (35.1) 72 (29.4) 45 (18.4) 31 (12.7) 41 (16.7) 14 (5.7) 159 (64.9) 1.50 (1.10–2.04) 0.81 (0.58–1.13) 2.64 (1.83–3.80) 1.48 (0.96–2.29) 0.40 (0.26–0.63) 4.01 (2.25–7.14) 0.67 (0.49–0.91) 0.01 NS ⬍ 0.0001 NS 0.0001 ⬍ 0.0001 0.01 * Values are the number (percentage) unless otherwise indicated. AS ⫽ ankylosing spondylitis; OR ⫽ odds ratio; 95% CI ⫽ 95% conﬁdence interval; NS ⫽ not signiﬁcant. 1190 Rennie et al RESULTS Descriptive data by patients. The number of patients with MRI evidence of spinal inﬂammation in any vertebral body was 39 (80%); 16 patients (33%) had inﬂammation in the cervical spine, 37 patients (76%) had inﬂammation in the thoracic spine, and 28 patients (57%) had inﬂammation in the lumbar spine. Of the patients who had inﬂammation in any vertebra, 41% involved the cervical spine, 95% involved the thoracic spine, and 72% involved the lumbar spine. There were somewhat more patients with lateral involvement (34 [70%] of 49) of the thoracic spine compared with central involvement (32 [65%] of 49). In contrast, more patients had involvement centrally (26 [53%] of 49) compared with laterally (21 [43%] of 49) in the lumbar spine. Four of the 6 controls had evidence of vertebral inﬂammation in either lumbar or thoracic vertebrae. Two controls had inﬂammation in the thoracic spine and 4 had involvement of the lumbar spine, although only a small number of vertebrae were affected (Table 1). All 3 of the controls with a history of nonspeciﬁc back pain demonstrated minor foci of inﬂammation in the lumbar spine. Descriptive data by vertebrae. A total of 343 (7 per each of the 49 subjects) cervical vertebrae, 588 (12 ⫻ 49) thoracic vertebrae, and 245 (5 ⫻ 49) lumbar vertebrae were assessed for inﬂammation. Inﬂammation was present signiﬁcantly more frequently in thoracic vertebrae (263 [45%] of 588) than in cervical vertebrae (30 [9%] of 343; P ⬍ 0.0001) or lumbar vertebrae (86 [35%] of 245; P ⫽ 0.01) (Table 2). Inﬂammation was also recorded in a total of 2 thoracic and 7 lumbar vertebrae from controls. The abnormalities identiﬁed in the control subjects were all small and were adjacent to either a small tear of the annulus ﬁbrosus or a degenerate disc. In the thoracic vertebrae of the AS group, involvement of lateral slices (37%) was more common than involvement of central slices (25%) (odds ratio [OR] 1.76, 95% conﬁdence interval [95% CI] 1.37–2.27; P ⬍ 0.001), and lateral involvement was observed in a higher proportion of thoracic (37%) than lumbar (18%) vertebrae (P ⬍ 0.0001). Involvement of only the lateral slices was observed in as many as 19.6% of thoracic vertebrae versus only 5.7% of lumbar vertebrae (P ⬍ 0.0001). In contrast, involvement of only the central slices was more common in lumbar vertebrae (P ⫽ 0.0001). Moreover, the frequency of both lateral and central slice inﬂammation in the same thoracic vertebra (17.7%) was signiﬁcantly more frequent than central slice inﬂammation alone (7.4%) (OR 2.66, 95% CI 1.83–3.86; P ⬍ 0.0001), but was not more frequent than lateral slice inﬂammation alone (19.6%). The distribution in the lumbar spine was different where involvement of central slices (29.4%) was more common than lateral slices (18.4%) (OR 1.85, 95% CI 1.21–2.83; P ⫽ 0.006), and where the frequency of both a lateral and central slice lesion in the same lumbar vertebra (12.7%) was similar to the frequency of central slice inﬂammation alone (16.7%) and signiﬁcantly greater than the frequency of lateral slice inﬂammation alone (5.7%) Figure 2. Sagittal STIR magnetic resonance image in a 40-year-old man with clinically proven ankylosing spondylitis and mild scoliosis. The lateral aspect of the lower thoracic vertebrae demonstrates a typical oval area of inﬂammation at the posterolateral corner of the T9 inferior end plate and the T10 superior end plate that constitutes the vertebral side of the tenth costovertebral joint (long arrow). Inﬂammation is noted to extend posteriorly through the pedicle. Intense inﬂammation is present in the head of the eighth rib (short arrow). (OR 2.39, 95% CI 1.24 – 4.62; P ⫽ 0.01). As a proportion of all levels demonstrating inﬂammation, involvement of lat- Assessment of Spinal Inﬂammation in AS with MRI 1191 eral slices was recorded in 219 (83.3%) of 263 thoracic and in 45 (52.3%) of 86 lumbar vertebrae. Typical lesions are illustrated in Figures 2 and 3. Inﬂammation in lateral slices was evident throughout the thoracic spine, being most frequent at T8 and T11 thoracic vertebrae (data not shown). Inﬂammation in central slices was most frequent at T7 in the thoracic spine and was broadly distributed in the lumbar spine. Reliability. The mean kappa for thoracic lateral and central slices was 0.56 (range 0.35– 0.74) and 0.38 (⫺0.30 to 1.00), respectively (Table 3). The kappa values for assessment of thoracic lateral slices were fair to good (0.40 – 0.74) for 11 of the 12 thoracic vertebrae. Percentage agreement ranged from 67– 87% and from 53–100% for assessment of thoracic lateral and central slices, respectively. Mean kappa values for lumbar lateral and central slices were 0.72 (range 0.44 –1.00) and 0.20 (0.05– 0.44), respectively. The kappa values for assessment of lumbar lateral slices were excellent (⬎0.75) for 2 of the 5 lumbar vertebrae. Percentage agreement ranged from 73–100% and from 47–73% for the assessment of lumbar lateral and central slices, respectively. DISCUSSION Our systematic evaluation of inﬂammatory lesions observed on MRI in thoracic and lumbar vertebrae of subjects with established AS shows that such lesions in the lateral slices occur in approximately one-third of all thoracic and one-sixth of all lumbar vertebrae, that such lesions are more common in lateral versus central slices in the thoracic but not the lumbar spine, and that approximately one-ﬁfth of all thoracic vertebrae display lesions only on the more lateral images. Consequently, it is possible that a diagnostic examination or an MRI scoring system that does not include all the appropriate lateral images of the spine or does not systematically analyze the lateral images could omit nearly half (44%) of all inﬂamed vertebrae in the thoracic spine and 16% of all inﬂamed vertebrae in the lumbar spine. The inﬂammatory abnormalities identiﬁed in the central slices were usually typical Romanus lesions, central end plate erosion with edema or diffuse vertebral inﬂammation, and, not surprisingly, occurred with similar frequency in the thoracic and lumbar spine. At levels where inﬂammation was seen on both central and lateral images, often these abnormalities were clearly separate lesions. However, frequently contiguous pathology was noted, and then it was not possible to specify the precise source of inﬂammation. Lateral inﬂammation was identiﬁed much more frequently in the thoracic spine and at many levels was particularly common in, or immediately adjacent to, the vertebral portion of the costovertebral joint. At a few levels, inﬂammation appeared to be arising from the posterior elements and extending through the pedicle into the vertebral body. However, this is not likely to be the primary source of lateral inﬂammation in the vertebral bodies, and it would not explain the striking difference in lateral predominance between the thoracic and lumbar Figure 3. From the same scan as Figure 2. Inﬂammation arising from the eleventh costovertebral joint involves both adjacent vertebral end plates (short white arrow ). Inﬂammation from the twelfth costovertebral joint involves only the T12 vertebra (long white arrow). The anatomy of this rib is different, and the head of the twelfth rib does not articulate with the T11 inferior end plate. Some edema in the pedicles of the lumbar spine is due to facet joint inﬂammation with minimal or no extension into the vertebral bodies (black arrow). Posterior element changes were not scored in this study. spine. We also noted in some cases of scoliosis (with additional MRI slices on one side) that the typical pattern 1192 Rennie et al Table 3. Interobserver reliability between 2 readers of detection of inﬂammatory lesions in the thoracic and lumbar vertebrae of 49 patients with AS* Location of inﬂammation Thoracic Any slice Central Lateral Lumbar Any slice Central Lateral† Kappa, mean (range) Excellent: >0.75 Fair to good: 0.40–0.75 Poor: <0.40 Agreement, mean % (range) 0.59 (0.35–0.74) 0.38 (⫺0.30 to 1.00) 0.56 (0.35–0.74) – 1 (8) – 10 (83) 3 (25) 11 (92) 2 (17) 8 (67) 1 (8) 79 (67–87) 73.8 (53–100) 77.8 (67–87) 0.33 (0.22–0.48) 0.20 (0.05–0.44) 0.72 (0.44–1.00) – – 2 (50) 2 (40) 1 (20) 2 (50) 3 (60) 4 (80) – 62.6 (53–73) 54.6 (47–73) 88 (73–100) * Values are the number (percentage) unless otherwise indicated. AS ⫽ ankylosing spondylitis. † No variation for 1 rater at L2. of edema in the posterolateral aspects of 2 adjacent end plates was associated with bone edema in the head of the corresponding rib (Figure 2). Clinical involvement of the costovertebral joint is readily recognized as pain and stiffness in the thoracic spine with restricted chest expansion. Histopathologic studies of this joint reveal the presence of a ﬁbrocartilaginous disc, and the presence of ﬁbrocartilaginous tissue at this location might explain the frequency with which inﬂammation is observed in the costovertebral joint (6). In patients with AS, there is a predilection for inﬂammation at sites rich in ﬁbrocartilage, and it has been hypothesized that a constituent of ﬁbrocartilage may induce autoimmunity (11). Our observations would be consistent with this hypothesis. However, even bony abnormalities of the costovertebral joint are not readily evident radiographically due to the complexity of the anatomy and overlapping structures that obscure these joints. Bone marrow inﬂammation was seen in control subjects, although the lesions were generally small and few in number. They occurred predominantly in the lower lumbar spine and in association with disc degeneration. Of course, some patients with AS also had disc degeneration and in those cases, the vertebral inﬂammation observed may not necessarily have been due to the spondylitis. However, the pattern of bone edema is sometimes quite characteristic for one or the other, and in the thoracic spine in particular, the presence of vertebral inﬂammation at multiple levels may be quite speciﬁc for spondylitis. The interobserver reliability was only moderate overall. The highest kappa values were recorded for readings of the lateral slices, which may be surprising to those not familiar with MRI interpretation. However, there are several factors that may explain this. All MRI scans are subject to artifact and for some sequences this may be considerable. The STIR sequence is very useful in this clinical setting because it is quite reliable, even with different patients and MRI units. A limitation of the sequence is its propensity to display phase-encoding artifacts caused by ﬂowing blood or cerebrospinal ﬂuid, and this spurious signal tends to be projected over the vertebral bodies near the midline, so the lumbar vertebrae may be particularly affected (12). This can make it quite difﬁcult to interpret small or faint abnormalities such as Romanus lesions. By comparison, involvement of the costovertebral joint tends to affect a broader area of the vertebral body and this may simplify interpretation even in the presence of artifact (Figure 2). The kappa values were particularly low in the central slices of the lower lumbar spine, inﬂuenced by the variability of marrow signal at levels that are so frequently affected by disc degeneration. Other contributing factors include difﬁculty in interpreting small end plate erosions versus Schmorl’s nodes, the small size of the upper thoracic vertebrae, and other MRI artifacts, although generally the image quality was good. The standard approach to MRI evaluation of the spine focuses on sagittal and transverse sequences and on the neurologic structures in the central spinal canal, lateral recesses, and neuroforamina. It is neither practical nor usually necessary to perform transverse imaging throughout the entire spine. So, regardless of indication, a typical protocol usually involves doing sagittal sequences to ﬁnd the pathology, and then performing additional transverse or other sequences through the identiﬁed levels of concern. The costovertebral joints are not generally the target for neoplastic, degenerative, or neurologic disorders and, therefore, neither are they the focus of routine clinical MRI. Scoliosis compounds the imaging problem and scanning instructions to the MRI technologists usually run along the lines of “spinal canal and lateral recesses must be included.” However, inclusion of the lateral aspects of all pedicles is not mandatory, let alone scanning lateral to the pedicles. Thus, with even minor degrees of scoliotic curvature, the pedicles are often only partially visualized, and the lateral edges of the vertebral bodies may not be seen at the middle or ends of the curvature. This problem is exaggerated when scanning at large ﬁelds of view and can only be overcome by adding more slices or making the slices thicker with corresponding increase in scan time or loss of resolution. Our results have signiﬁcant implications for the training of radiologists and rheumatologists in the interpretation of MRI from patients referred for both diagnostic evaluation and assessment of activity in established disease. It has, for example, been reported that at least one-quarter of patients with AS who are sufﬁciently symptomatic to be recruited to trials of anti–tumor necrosis factor ␣ (anti-TNF␣) therapies have no evidence of spinal inﬂammation on baseline Assessment of Spinal Inﬂammation in AS with MRI MRI scans (13). However, this assumption may have been based on a limited assessment of sagittal images. Furthermore, scores for spinal inﬂammation on MRI have been analyzed as both predictor variables for treatment response and structural damage and in correlation analyses with clinical, laboratory, and biomarker variables. Our data suggest that these analyses may be compromised if assessment of MRI scans does not include the entirety of the thoracic spine, extending to the costovertebral joints bilaterally. Abnormalities in the lateral aspects of the vertebrae appear to be more reliably detected, and scoring systems that include assessment of lesions in consecutive sagittal slices may possess advantages over those that focus on the scoring of individual lesions in the anteroposterior dimension. In particular, inclusion of ⬎1 slice in the scoring of lesions may improve responsiveness, and this may be advantageous in the evaluation of therapeutic agents that possess structure-modifying properties but are less effective as antiinﬂammatory agents than anti-TNF␣ therapies. In conclusion, this systematic analysis of inﬂammatory lesions evident on MRI in the thoracic and lumbar spine demonstrates that the majority of lesions in thoracic vertebral bodies are present in the more lateral aspect of the vertebrae and that lateral lesions in general are more reliably detected than those closer to the midline. Lateral lesions may not be fully assessed on the sagittal images of standard MRI protocols, especially in the presence of scoliosis, and so it is recommended that diagnostic evaluation and assessment of disease activity in spondylarthropathy should include assessment of sagittal images that extend to the lateral edges of all thoracic vertebrae as a minimum requirement. AUTHOR CONTRIBUTIONS 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 submitted for publication. Mr. Lambert 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. Rennie, Dhillon, Maksymowych, Lambert. Acquisition of data. Rennie, Dhillon, Maksymowych, Lambert. 1193 Analysis and interpretation of data. Rennie, Dhillon, ConnerSpady, Maksymowych, Lambert. REFERENCES 1. Maksymowych WP, Lambert RG. Spinal inﬂammation in ankylosing spondylitis: how and why should it be measured by MRI? [editorial]. Nat Clin Pract Rheumatol 2006;2:232–3. 2. Francois RJ, Gardner DL, Degrave EJ, Bywaters EG. Histopathologic evidence that sacroiliitis in ankylosing spondylitis is not merely enthesitis: systematic study of specimens from patients and control subjects. Arthritis Rheum 2000;43: 2011–24. 3. Pascual E, Castellano JA, Lopez E. Costovertebral joint changes in ankylosing spondylitis with thoracic pain. Brit J Rheumatol 1992;31:413–5. 4. 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; 27:361– 8. 5. Hart FD, Emmerson PA, Gregg I. Thorax in ankylosing spondylitis. Ann Rheum Dis 1963;22:11– 8. 6. Fassbender HG. Pathology of the rheumatic diseases. Berlin: Springer-Verlag; 1975. p. 221–58. 7. Dihlmann W. Current radiodiagnostic concept of ankylosing spondylitis. Skelet Radiol 1979;4:179 – 88. 8. Braun J, Baraliakos X, Golder W, Brandt J, Rudwaleit M, Listing J, et al. Magnetic resonance imaging examinations of the spine in patients with ankylosing spondylitis, before and after successful therapy with inﬂiximab: evaluation of a new scoring system. Arthritis Rheum 2003;48:1126 –36. 9. Maksymowych WP, Inman RD, Salonen D, Dhillon SS, Krishnananthan R, Stone M, et al. Spondyloarthritis Research Consortium of Canada magnetic resonance imaging index for assessment of spinal inﬂammation in ankylosing spondylitis. Arthritis Rheum 2005;53:502–9. 10. Baraliakos X, Landewe R, Hermann KG, Listing J, Golder W, Brandt J, et al. Inﬂammation in ankylosing spondylitis: a systematic description of the extent and frequency of acute spinal changes using magnetic resonance imaging. Ann Rheum Dis 2005;64:730 – 4. 11. Maksymowych WP. Ankylosing spondylitis: at the interface of bone and cartilage. J Rheumatol 2000;27:2295–302. 12. Maksymowych WP, Lambert RG. Magnetic resonance imaging for spondyloarthritis: avoiding the mineﬁeld [editorial]. J Rheumatol 2007;34:259 – 65. 13. Braun J, Landewe R, Hermann KG, Han J, Yan S, Williamson P, et al, for the ASSERT Study Group. Major reduction in spinal inﬂammation in patients with ankylosing spondylitis after treatment with inﬂiximab: results of a multicenter, randomized, double-blind, placebo-controlled magnetic resonance imaging study. Arthritis Rheum 2006;54:1646 –52.