Although it is clear that preventing structural damage is the primary goal of therapy in rheumatoid arthritis (RA) (1), the most effective means of documenting progression of joint damage is still a contentious issue (2). This is especially the case in patients with established disease, in whom existing erosions and joint malalignment complicate the detection of ongoing damage (3). Furthermore, whereas aggressive treatment is widely accepted to prevent erosion progression in early disease (4, 5), the efficacy of intensive intervention in patients with established disease is less clear. In addition, intervention studies utilizing new therapies often recruit “mixed” cohorts of patients with early and established disease (6–10), understanding the importance of establishing the sensitivity and reliability of available imaging methods in various clinical situations.
Radiographic imaging is widely accepted as the gold standard for the assessment of disease progression in RA. Radiographic scoring systems, such as the Larsen and Sharp scores (11) and their modifications (12, 13), are the standard methods for the determination of joint damage and its progression. However, despite considerable effort to reduce the intrinsic limitations of radiographs, their perceived lack of sensitivity to erosive change in early RA remains a major issue. This has led to the investigation of newer imaging techniques such as magnetic resonance imaging (MRI).
The advantages of MRI of the wrist and metacarpophalangeal (MCP) joints when compared with radiographic imaging are well documented in cross-sectional studies of patients with early disease (disease duration of <12 months) (14–19). In applying MRI to patients whose disease is established, the important issue is whether the promising results in early disease can be reproduced in cohorts of patients with more advanced disease, that is, whether the performance characteristics of MRI are applicable when significant joint damage is already present.
The aim of the present study was therefore to evaluate the progression of erosive damage in the MCP joints of patients with established RA, over a 24-month period using MRI, and to compare these results with the findings on radiographs of both hands, obtained over the same period.
- Top of page
- PATIENTS AND METHODS
The mean age of subjects was 59 years and the mean disease duration of the group was 5.1 years (range 0.5–29 years). Ten patients (21%) had a disease duration of <2 years at study entry.
MRI damage progression was initially defined as any progression in the MRI damage score from the MRI damage score at baseline. Using this criterion, MRI evidence of erosion progression was identified in 30 patients (64%). Progression was then defined more stringently by applying the SDD based on the intraobserver scores (±3.25 units). Using this definition, 11 patients (23%) showed MRI evidence of erosion progression that was greater than the SDD (Table 1).
Table 1. Damage progression scores as determined by magnetic resonance imaging (MRI) compared with radiography*
|Imaging modality||OMERACT MRI RA score of MCPs 2–5 of dominant hand||Radiographs, total Larsen score of both hands||Radiographs, Larsen score of MCPs 2–5 of dominant hand|
| % of actual maximum score||4.22||0.78||6.2|
|Number of damage progressors with >SDD†||11||19||7|
| Study identification numbers of damage progressors||5, 15, 20, 30, 35, 41, 49, 50, 51, 53, 55||5, 8, 13, 14, 15, 17, 18, 20, 26, 30, 35, 36, 41, 45, 47, 49, 50, 51, 53||18, 26, 30, 41, 47, 49, 53|
For the radiographs, the SDD based on the intraobserver total Larsen scores was 0.77 units and the SDD for the dominant-hand MCP joints 2–5 was 1.55 units. On the basis of these results, radiographic progression was noted in 19 patients (40%) by the total Larsen score and 7 patients (15%) by the dominant-hand MCP Larsen score (Table 1).
Comparisons between MRI progression and total Larsen score radiographic progression results showed that a similar group of patients were identified as having progression (damage progressors) by both methods, but the radiographs of both hands identified more progressors overall than did MRI with limited field of view (Table 2). For the 9 patients who showed progression of joint damage on radiographs that was not detected by MRI, the majority (66%) exhibited this radiographic progression outside the dominant-hand MCP joints. Three patients, however, did have evidence of radiographic progression in the dominant-hand MCP joints that was not detected by MRI. There were 7 patients who showed evidence of progression in the dominant-hand MCP joints on MRI that did not have evidence of progression when the limited (dominant-hand MCPs 2–5) Larsen score was applied to the radiographs.
Table 2. Differences between imaging methods in the determination of damage progression*
|Imaging method||MRI progressors not identified by total Larsen score||Total Larsen score progressors not identified by MRI||Larsen score MCPs 2–5 progressors not identified by MRI||MRI progressors not identified by Larsen score MCPs 2–5|
|Number of patients||1||9||3||7|
|Study identification numbers of damage progressors||55||8, 13, 14, 17, 18, 26, 36, 45, 47||18, 26, 47||5, 15, 20, 35, 50, 51, 55|
- Top of page
- PATIENTS AND METHODS
This study compared MRI examination of the MCP joints of one hand with standard radiographic examination of both hands as assessment methods for damage progression over 2 years in patients with established RA. The results demonstrate that radiographs appear to be more responsive to progression of joint damage in this group of patients. Although both methods identified a similar group of damage progressors, the radiographs were more responsive, largely because of identification of progression in joints outside the field of view of the MRIs. As expected, MRI demonstrated greater sensitivity to damage progression in the MCP joints alone, but did not show an advantage when compared with the total Larsen radiographic scoring of both hands.
This study raises a number of issues that merit further comment. These issues include the role of observer error in scoring damage progression, the importance of the joint sample in evaluating damage progression, the influence of the MRI acquisition parameters, the content validity of the scoring methods, and the effect of the prolonged observation period.
With regard to the role of observer error, damage progression can be judged by changes in the raw score or, more rigorously, by changes that exceed the measurement error as indicated by the SDD. In this study, the SDD was measured using the intraobserver reliability, and the criterion for progression was defined as a change greater than the measurement error, i.e., greater than the SDD of the progression scores. It is important to note that SDDs calculated on progression scores, rather than on cross-sectional data, often lead to lower SDD values as a result of lower scores, since progression scores are usually smaller than cross-sectional scores. The SDD as a percentage of the actual maximum score was slightly higher for the MRI reader, but not sufficiently different to suggest that there was significantly greater intraobserver error in the MRI scoring.
However, because the SDDs for both the total Larsen score (0.77) and for the MCP joints alone (1.55) were small, it is relevant to speculate on the effect of greater variability in the radiograph reading on the study outcome. As stated in Patients and Methods, the reliability studies provided a 95% probability of true progression in the results presented. If an 80% probability of real progression is applied, the results, and therefore the conclusion of this study, do not change significantly. If a 67% probability of real progression is applied, the results alter to the extent that MRI demonstrates slightly more real progression than does the total Larsen score. Such a low probability of real change is based on a 33% error, which is unrealistically high for studies conducted by experienced investigators. Nonetheless, the point to note is that measurement error affects the results of comparative studies such as the present study, and not only is it important that the SDD is comparable in terms of its percentage of the actual maximum score (Table 1), but also it is essential that a rigorous application of the method is undertaken (i.e., 95% probability of true progression) to ensure that the results reflect real progression.
With regard to the importance of the joint sample in evaluating damage progression, adequate representation of progressive joint damage is important for patients with RA, at any stage of the disease process. Of the 19 patients identified as progressors by the total Larsen score, 9 were not identified as progressors by the limited-field MRI examination. Therefore, although both methods identified a similar group of patients with progressive disease, radiographs of both hands were clearly more responsive in this group of patients, largely due to progression in joints not included in the limited MRI study (Table 2). However, direct comparison between the dominant-hand MCP MRI and the Larsen score from the corresponding area showed that MRI was more sensitive in detecting erosive change in this subset of joints (Figures 1 and 2).
Figure 1. Baseline (A) and year 2 (B) radiographs obtained from patient 50 in a 24-month rheumatoid arthritis observational study during the period 1998–2002. Radiographs are of the dominant-hand metacarpophalangeal joints 2–5. At both time points, the Larsen radiographic damage score is 0. Although there appears to be an abnormality in the third metacarpal head (B), the defect does not qualify as an erosion using the Larsen scoring method, and there is no appreciable joint space narrowing.
Download figure to PowerPoint
Figure 2. Baseline (A) and year 2 (B) magnetic resonance imaging (MRI) of the dominant-hand metacarpophalangeal joints 2–5 from patient 50 in a 24-month rheumatoid arthritis observational study during the period 1998–2002. The coronal and axial MRIs show erosion (arrows in A) and progression of erosion (arrows in B) in the third metacarpal.
Download figure to PowerPoint
It should also be noted that the feet were not included in the radiographic analysis in this study. Perhaps inclusion of assessment of structural damage in the feet may have reinforced the results of this study, bringing radiographs into an even more favorable position by virtue of an increase in joint coverage.
Another consideration is whether the MRI sequence protocol was a factor in reducing the sensitivity of MRI. Slice thickness may be important in this regard; thinner slices would be expected to be more sensitive in detecting small erosions, but the trade-off is a reduction in the signal-to-noise ratio, which produces a less clear image so that, in theory, erosions may be more difficult to identify and measurement may be less accurate. In this study, 3-mm slice thicknesses with a 0.3-mm gap were utilized so that effectively there was a 3.3-mm gap between slices in the coronal plane. It is possible that this reduced the sensitivity of the MRI to detect small erosions, and may be a contributing factor in the apparent lack of sensitivity of MRI in this study.
A further issue is whether the scoring method used for MRI is comprehensive enough to document change. The scores for MRI include only erosions, whereas the Larsen score includes joint space narrowing as part of the total score; this may have contributed to the superior performance of the total Larsen score. This is suggested by the 3 patients with evidence of progression in the MCPs on radiographs that was not detected by MRI. Reanalysis of radiographs using the Sharp score confirmed this suspicion, with all 3 showing the majority of progression via change in joint space narrowing, rather than erosion progression.
Other methods of MRI erosion measurement are becoming available; for example, computerized measurement of erosive damage on MRI has been shown to be comparable with the OMERACT MRI RA score (27) and has shown excellent reliability in cross-sectional trials. However, in this study, the MR sequences were not adequate for computerized analysis because the slice thickness and interslice gap did not permit accurate computerized volume analysis in the MCP joints for the majority of patients.
Furthermore, the length of the observation period is important in judging the superiority of one imaging technique over another. We chose an interval that would allow for a period of change, in this observational study, sufficient to allow adequate comparison of the 2 imaging techniques. A shorter period would probably have skewed the results in favor of the MRI; the relatively long followup may have skewed the results in favor of radiographs. However, it is impossible to be certain without the benefit of serial measurements. Nevertheless, it is important to note that the results of our study pertain specifically to change over a 2-year period.
Finally, with regard to other longitudinal studies in established disease, comparisons of MRI and radiographs in longitudinal studies of patients with established RA are limited in number. Using 2 separate established disease cohorts, Ostergaard et al reported that MRI identified more erosions than did radiographs in 26 patients over 12 months (28) and 10 patients over a 5-year period (29). Erosions were counted, rather than scored, and in contrast to our study, enlargement of an existing erosion or the development of a new erosion in a previously eroded bone was not regarded as progression. Intraobserver reliability for the erosion scores was not recorded in those studies. These disparities illustrate the importance of the measurement method and the principle that the minimization of measurement error is a central issue in studies of damage progression.
The results of this study indicate that although MRI was more sensitive than radiographs within a restricted field of view, overall there is no clear advantage of MRI over radiographs in documenting damage progression in patients with established RA. However, this result must be interpreted strictly within the constraints of the project: a cohort with long-term disease, an observation period of 2 years, the MRI acquisition and scoring system used, and the assessment of the dominant-hand MCP joints rather than the carpus.
The conclusion that can be drawn from this study is not simply that radiographs are better than MRI or vice versa, but that careful further study and analysis is required to determine the optimal imaging method for the task, whether that be detection of erosive change over short periods in early disease or the assessment of more global damage progression over longer periods in established disease. To be able to select the most appropriate technique from the diverse array of methods now available, we need clear data on the validity, reliability, and sensitivity of each method in different disease stages and study settings. The results of this study contribute to the development of such a database.