ARTHRITIS & RHEUMATISM Vol. 48, No. 3, March 2003, pp 585–589 DOI 10.1002/art.10819 © 2003, American College of Rheumatology
Arthritis & Rheumatism An Official Journal of the American College of Rheumatology www.arthritisrheum.org and www.interscience.wiley.com
EDITORIAL
Magnetic Resonance Imaging in the Evaluation of Bone Damage in Rheumatoid Arthritis: A More Precise Image or Just a More Expensive One? Raphaela Goldbach-Mansky,1 James Woodburn,2 Lawrence Yao,2 and Peter E. Lipsky1 1 year later, patients with higher baseline MRI scores have significantly more subsequent radiographic erosions (7,8). These and other studies have suggested that MRI is superior for detecting osseous changes in joints early in the course of RA and that it may detect bony lesions up to 6 months or more before CRs. When individual MRI lesions were tracked over 2 years, only 1 of 4 “erosions” detected on MRI progressed to become radiographic erosions (9). This finding has raised a question about the true nature of “erosions” detected by MRI. Since the pathophysiologic basis of these “erosion-like lesions” has not been determined, it is unclear which MRI lesions are destined to become “radiographic erosions” and what the significance is of those MRI lesions that do not progress to become radiographic erosions. Notably, one study has failed to document changes in erosions on MRI in the metacarpophalangeal joints, even though radiographic scores of the hands showed progression (10), highlighting the difficulty in interpreting the nature of early MRI lesions and in using MRI in the clinical evaluation of patients. Moreover, because high-resolution MRI examinations are limited to the evaluation of specific body areas, an unresolved question is the extent to which changes in a single joint or group of joints reflect a systemic, polyarticular disease process. The interpretation and comparison of the various MRI studies have been confounded by differences in imaging sequences, in changes selected to be scored, and in scoring systems (11). Attempts to standardize image acquisition, nomenclature, and scoring systems for MRI have been summarized by a group of experts at the Outcome Measures in Rheumatology Clinical Trials 5 (OMERACT 5) consensus conference (12). The recommendations included assessment of the carpal bones as
The development of bone erosions early in the course of rheumatoid arthritis (RA) is a poor prognostic sign (1). Traditionally, erosions have been assessed by serial radiographs of the joints of the hands and feet. Conventional radiographs (CRs) are used in clinical practice and in clinical trials to follow the progression of joint damage over time and the response to therapy. The quantitative approach to assessment of joint destruction of the hands and feet (2,3) has permitted the development of correlations of radiographic scores that assess both bone erosions and joint space narrowing with long-term outcomes (4). Radiographs are thought to be insufficiently sensitive in imaging early cartilage and bone changes, however, and the development of these pathologic changes in joint structure may precede their detection by radiographs by months or even years (5,6). Magnetic resonance imaging (MRI) is a sensitive means of visualizing articular structures. MRI can assess osseous changes, but it also extends the capacity to visualize the surrounding soft tissue with good contrast and high spatial resolution. Synovitis, the primary inflammatory lesion in RA, can be detected and monitored longitudinally. Other soft tissues, such as the tenosynovium, tendons, entheses, joint effusions, and ligaments, can also be visualized. When MRI scores of the wrist and metarcarpal bones of patients with RA are compared with radiographic scores of the hands and feet 1 Raphaela Goldbach-Mansky, MD, Peter E. Lipsky, MD: National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH, Bethesda, Maryland; 2James Woodburn, PhD, BSc, SRCh, Lawrence Yao, MD: Warren G. Magnuson Clinical Center, NIH, Bethesda, Maryland. Address correspondence and reprint requests to Raphaela Goldbach-Mansky, MD, National Institute of Arthritis and Musculoskeletal and Skin Diseases, Building 10, Room 9S205, 10 Center Drive, Bethesda, MD 20892. E-mail:
[email protected]. Submitted for publication October 22, 2002; accepted in revised form November 18, 2002.
585
586
well as the metacarpal bases and the distal radius and ulna. “Erosion,” “bone defect,” and “bone edema” are each scored on a scale of 1–10 (1 ⫽ 10% bone involvement and 10 ⫽ 100% involvement). Synovitis is graded on a scale of 0–3 at the radiocarpal joint and intercarpal joints. Of note, standardization of the image acquisition is only loosely defined, and the terminology used to describe image abnormalities is somewhat subjective and derived in part from nonvalidated pathology correlates. A major problem with any scoring system has been the large degree of interrater variability. It was recently reported (13) that 5 international centers scored sets of images obtained by slightly differing techniques using different scoring methods. The interrater agreement was moderate at best, indicating the need to standardize image acquisition technique, develop a standard scoring system, and improve training to achieve interrater reproducibility sufficient to assess synovial and osseous changes accurately. To reduce the problem of interrater variability, quantitative, rather than semiquantitative, methods to assess bone erosions and synovial changes have been explored. In this issue of Arthritis & Rheumatism, Bird and colleagues (14) report the use of a semiautomated segmentation tool to outline and measure the volume of “erosions” (hypodense areas on the T1-weighted images) and the inflamed synovium (enhancing areas on the postgadolinium images). MR images of the wrist were obtained at 3 mm from 12 patients with rheumatoid factor–positive RA. Erosion and synovial volumes were calculated by summing all erosions and the area of the synovium outlined on each image and then multiplying the sum by the slice thickness to generate a volumetric index. The same observer scored erosions and synovitis and measured erosion and synovial volumes of the same wrist of the same patient 48 hours apart. Scoring and volumetric measurements of each image were repeated 72 hours apart. The investigators demonstrated that “erosion volumes” were reproducibly stable within a 2-day period. Synovial volumes and scores showed more variability. It is not clear whether the difference in synovial scores reflects true changes in synovial volumes within short intervals. In this study, the intrarater variability of measuring bony defects and synovial volume of the same image 72 hours apart was very small. However, the confidence intervals were relatively wide. Bird et al conclude that this method of determining erosion and synovial volumes is feasible, reliable, and valid. If reproduced in other centers, this approach could facilitate the use of MRI to quantify articular lesions in
GOLDBACH-MANSKY ET AL
RA. However, there was only a single scorer; therefore, interrater variability, which is a critical concern if wider application of this approach is considered, could not be determined. In addition, wider application may be limited by the unwieldy process of manually outlining each lesion on every image. The technical challenges of outlining regions of interest, such as an erosion or synovial tissue, also introduce operator dependence. An operator needs to decide whether a “dark area” on the T1-weighted image is an erosion and then draw a line around this area to measure its volume. The process of outlining an area of interest on the computer screen is called segmentation. To overcome the possible subjectivity implicit in this approach, several automatic and semiautomatic segmentation methods have been developed for multidimensional image analysis of objects of interest (i.e., bone and bone erosions). Importantly, these methods differ in their degree of accuracy, precision, and efficiency (i.e., the extent of operator dependence involved). In studies conducted at the National Institutes of Health, we have employed a boundary-based assisted segmentation method using a “live wire” paradigm in which the operator offers continuous recognition help to an automatic delineation algorithm by steering a live wire close to the object boundary. This method has been shown to be more repeatable and more efficient than manual tracking methods (15), including the method used by Bird et al. Regardless of the method used to render bone volumes, subjectivity remains. For example, during segmentation, erosions present intermittently absent boundaries, and the larger the surface defect, the greater the user input necessary to define the spatial extent of the boundary. In addition, enhanced synovium is often difficult to distinguish from the surrounding joint capsule, muscles, and fat, raising the question of which sequences are most useful for estimating synovial volume. In view of these technical issues, a major question remains. How well does MRI depict bone erosions? The study by Bird et al assesses reproducibility and reliability (correlation coefficients and smallest detectable differences), but does not address the capacity of MRI to identify and assess relevant pathophysiologic changes. The latter is a question of validity. Validity is a term for how well an instrument or measurement procedure measures what it proposes to measure. This issue is particularly important, because mineralized bone does not give any signal on MRI, and is visualized as a
EDITORIAL
587
Figure 1. Volume renditions of the triquetrum and pisiform reconstructed in 3 dimensions by image processing techniques from A, computed tomography and B, magnetic resonance imaging scans of the carpal region in a patient with rheumatoid arthritis. In C and D (corresponding to A and B, respectively), the lesions are rendered and displayed. The bone and lesions (labeled 1–4; see Table 1) are assigned different opacity functions to show the surface area of each lesion (red) and its depth (margins defined by dashed lines) in 3 dimensions. Arrow in D indicates a lesion that does not correspond to any lesion in C.
“negative image” between surrounding tissues that contain mobile protons. Most of the “validation procedures” for MRI are based on the correlation of semi-
quantitative MRI measures with radiographic scores or other imaging modalities (such as semiquantitative scoring of the MRI lesions of the wrist compared with
588
GOLDBACH-MANSKY ET AL
Table 1. Volumes of lesions 1–4 shown in Figures 1C and D according to computed tomography (CT) and magnetic resonance imaging (MRI) Volume by CT Lesion (location) 1 2 3 4
(triquetrum) (triquetrum) (triquetrum) (pisiform)
Volume by MRI
Absolute, mm3
Relative, %
Absolute, mm3
Relative, %
88.8 1.0 8.9 66.5
8.5 0.1 0.9 11.2
113.4 7.7 9.1 182.4
15.2 1.0 1.2 47.2
radiographic scores of the hands and feet) (6–10). Although useful, these studies were not designed to relate MRI changes to pathophysiologic events. Ideally, physical measurement of the erosion in the bone itself and comparison with the volume measured by MRI would provide validation of the MRI technique for representing erosions. This kind of validation on cadaver samples has obvious limitations. Investigators in our group have tried to address this problem by comparing bone lesion volumes detected by MRI with computed tomography (CT) images of the carpal bones in patients with erosive RA (16). Since CT is considered to be the best way to assess bony changes, it was believed that this would provide a way of relating the MR images to the best “gold standard” available. An assisted livewire segmentation method was used as part of an image processing protocol to reconstruct in 3 dimensions the bone and erosions (or lesion) for both imaging modalities. Figure 1 and Table 1 represent characteristic findings from comparing these modalities for imaging the same carpal bone (triquetrum and pisiform on these images). This analysis yielded several surprising findings. As shown in Figure 1D, there is a lesion on the (T1weighted spin-echo) MRI that does not correspond to any lesion on the CT image. In addition, most of the MRI lesions are larger in volume than the corresponding lesions on the CT image. The degree of size difference varies between individual lesions. This difference is not a matter of inconsistently mapping corresponding lesions, since 1-mm cuts of the wrists were obtained (on MRI and CT) as opposed to the 3-mm cuts that are usually obtained. Multiple factors may contribute to the apparent size difference in lesions shown on CT and MRI. The MRI probably depicts bone marrow abnormalities surrounding or even filling an area of bone loss, and these marrow reactions may be variably depicted by other MRI sequences. Hence, caution must be used
when referring to these MRI lesions as “erosions,” and a change in their size should not be interpreted as a change in erosion volumes or even as “healing of an erosion” before more definitive validation has been completed. Additionally, the magnitude of signal changes on MRI is always influenced by the specific scan parameters; differences in MRI scanning methods must be strictly considered in applying any quantitative or semiquantitative MRI method. The development of highly effective treatments for RA, the need to monitor the effect of these therapies, and encouraging preliminary imaging results have stimulated interest in the use of MRI in the diagnosis, prognostication, and monitoring of RA (17). Nevertheless, major questions remain about the ability of this expensive imaging modality to add significantly to the quantitative assessment of articular lesions in RA to provide pathophysiologic information relevant to disease progression. It is clear that MRI provides detailed and potentially quantifiable images of articular structures, and varying changes on the MRI are seen during the course of the disease and in response to treatment. However, the enthusiastic embrace of this technology in clinical trials may be premature, given the challenges of analyzing MRI data and the uncertainty concerning the pathophysiologic meaning of these data. MRI generates a large amount of information about bone, the marrow space, and surrounding soft tissue, and it may be a powerful tool for improving our understanding of the pathophysiology of RA and other arthritides (18). Careful validation of MRI findings is required before the importance of these findings is established with respect to disease activity and prognosis. In particular, while MRI may ultimately prove useful in early prognostication of RA, a rigorous validation of MRI findings as surrogates for established radiographic outcome measures is warranted before the routine inclusion of MRI in clinical trials is accepted. Moreover, quantitative interpretation of MRI results and standards for MRI use must be predicated on well-defined MRI protocols. Nevertheless, as innovations in MRI technology evolve, our ability to obtain high-resolution images of joints and surrounding soft tissue pathology will improve. If properly validated, MRI may become a powerful tool for examining the physiology of bone damage and remodeling in RA and in other inflammatory diseases, as well as for evaluating the response to therapy.
EDITORIAL
589
REFERENCES 1. Brennan P, Harrison B, Barrett E. A simple algorithm to predict the development of radiographical erosions in patients with early rheumatoid arthritis: prospective cohort study. BMJ 1996;313: 471–6. 2. Sharp JT. Assessment of radiographic abnormalities in rheumatoid arthritis: what have we accomplished and where should we go from here? J Rheumatol 1995;22:1787–91. 3. Van der Heijde D. How to read radiographs according to the Sharp/van der Heijde method. J Rheumatol 1999;26:743–5. 4. Scott DL, Pugner K, Kaarela K, Doyle DV, Woolf A, Holmes J, et al. The links between joint damage and disability in rheumatoid arthritis. Rheumatology (Oxford) 2000;39:122–32. 5. Gilkeson G, Polisson R, Sinclair H, Vogler J, Rice J, Caldwell D, et al. Early detection of carpal erosions in patients with rheumatoid arthritis: a pilot study of magnetic resonance imaging. J Rheumatol 1988;15:1361–6. 6. Foley-Nolan D, Stack JP, Ryan M, Redmond U, Barry C, Ennis J, et al. Magnetic resonance imaging in the assessment of rheumatoid arthritis: a comparison with plain film radiographs. Br J Rheumatol 1991;30:101–6. 7. McQueen FM, Stewart N, Crabbe J, Robinson E, Yeoman S, Tan PL, et al. Magnetic resonance imaging of the wrist in early rheumatoid arthritis reveals a high prevalence of erosions at four months after symptom onset. Ann Rheum Dis 1998;57:350–6. 8. Lee J, Lee SK, Suh JS, Yoon M, Song JH, Lee CH. Magnetic resonance imaging of the wrist in defining remission of rheumatoid arthritis. J Rheumatol 1997;24:1303–8. 9. McQueen FM, Benton N, Crabbe J, Robinson E, Yeoman S, McLean L, et al. What is the fate of erosions in early rheumatoid arthritis? Tracking individual lesions using x rays and magnetic resonance imaging over the first two years of disease. Ann Rheum Dis 2001;60:859–68.
10. Klarlund M, Østergaard M, Jensen KE, Madsen JL, Skjodt H, Lorenzen I, the TIRA Group. Magnetic resonance imaging, radiography, and scintigraphy of the finger joints: one year follow up of patients with early arthritis. Ann Rheum Dis 2000;59:521–8. 11. McQueen FM. Magnetic resonance imaging in early inflammatory arthritis: what is its role? Rheumatology (Oxford) 2000;39: 700–6. 12. Conaghan P, Edmonds J, Emery P, Genant H, Gibbon W, Klarlund M, et al. Magnetic resonance imaging in rheumatoid arthritis: summary of OMERACT activities, current status, and plans. J Rheumatol 2001;28:1158–62. 13. Østergaard M, Klarlund M, Lassere M, Conaghan P, Peterfy C, McQueen F, et al. Interreader agreement in the assessment of magnetic resonance images of rheumatoid arthritis wrist and finger joints—an international multicenter study. J Rheumatol 2001;28:1143–50. 14. Bird P, Lassere M, Shnier R, Edmonds J. Computerized measurement of magnetic resonance imaging erosion volumes in patients with rheumatoid arthritis: a comparison with existing magnetic resonance imaging scoring systems and standard clinical outcome measures. Arthritis Rheum 2003;48:614–24. 15. Falcao AX, Udupa JK, Samarasekera S, Sharma S. User-steered image segmentation paradigms: live wire and live lane. Graph Model Im Proc 1998;60:233–60. 16. Bedair H, Murphy M, Fleming D, Hill S, Schaub C, Thornton B, et al. A comparison of MRI and CT in detecting carpal bone erosions in early rheumatoid arthritis [abstract]. Arthritis Rheum 2001;44 Suppl 9:S222. 17. Østergaard M, Szkudlarek M. Magnetic resonance imaging of soft tissue changes in rheumatoid arthritis wrist joints. Semin Musculoskelet Radiol 2001;5:257–74. 18. McGonagle D, Gibbon W, Emery P. Classification of inflammatory arthritis by enthesitis. Lancet 1998;352:1137–40.
