The main purpose of the present study was to investigate whether erosions detected by MRI and US could be confirmed by CT (that is, whether they are true erosions). With CT as the standard reference method, high specificity of MRI and US in detecting bone erosions in RA MCP joints was demonstrated, even when only radiographically normal MCP joints (that is, the joints with the most subtle changes) were considered. This study strongly indicates that radiographically invisible bone erosions detected by MRI and US are true erosive changes.
All patients were selected from former MRI studies and were eligible to participate in the study if they had at least one radiographically invisible MRI lesion, presumed to be an erosion. This selection was made in order to include only patients whose joints were not too severely damaged, that is, the patients in which MRI and US would be expected to have the greatest clinical value. Clearly, the sensitivity of radiography, MRI and US would have been higher if joints with extensive erosive changes had been included.
Previous studies [9
] have reported that radiography has poor sensitivity in detecting bone erosions compared with MRI. In the present study we also found that radiography had very poor sensitivity (19%) in detecting bone erosions in RA MCP joints compared with CT. This finding verifies that radiography, possibly because of its two-dimensional visualization of the joint, is insensitive in detecting the earliest stages of erosive bone changes in RA. In this material, radiography was unable to detect any erosions in nine out of 17 patients, whereas at least one erosion was seen on CT and MRI in all patients. When retrospectively reassessing radiographs of areas with erosions on CT, MRI and US, subtle changes (for example, changes in the trabecular pattern) may occasionally be recognized (for example, the 5th metacarpal head on Figure ). However, such changes are still not considered erosions by current criteria for radiographic erosions.
Signal on radiography and CT is based on attenuation of X-rays, and bone and other calcified tissues are easily depicted because of markedly higher X-ray attenuation by these tissues than by the surrounding soft tissues. The signal on magnetic resonance images is not dependent on X-rays but on presence of mobile protons in the tissue, and as the water content in bone is very low cortical bone is depicted as signal voids silhouetted against signal-emitting bone marrow and periosseous tissues. It has been argued that MRI is not well suited to visualizing lesions of calcified tissue, and the nature of bone erosions visualized with MRI, that are not visible on radiography, has been questioned [26
]. In this study MRI findings were in very good agreement with findings from the applied high-resolution three-dimensional tomographic X-ray modality (that is, CT findings), even in regions without radiographic erosions.
To our knowledge, no published studies have compared CT, MRI, US and radiography in small RA joints. In a recent study conducted by Perry and coworkers [12
], a comparison between CT and MRI in wrist joints of nine RA patients revealed an overall agreement between CT and MRI of 87% in detecting bone erosions. As in the present study, Perry and coworkers found more erosions with CT than with MRI. Whereas that study included joints with severe damage, the patients in the present study were selected on the basis of their having joints with MRI erosions that were radiographically occult, increasing the opportunity to demonstrate the specificity of radiographically invisible MRI erosions. The moderate sensitivities of MRI and, particularly, US obtained in the present study suggest the presence of more erosions than were detected by MRI (Figure ) and US. However, because the sensitivity of MRI and US in this and several other studies [9
] has been found to be much higher than that of radiography, we consider it acceptable that some minimal erosions are missed by MRI and US as long as the identified erosions are real. However, it should be emphasized that the sensitivities, specificities and accuracies reported in this study are study specific and not directly transferable to other patient cohorts. The moderate sensitivity of MRI found in this study suggests that the applied OMERACT RAMRIS definition of erosions [30
] does not overestimate the number of erosions; we consider this to be of major importance, and so we believe that the present study further supports the future use of the OMERACT RAMRIS definition of bone erosion.
Overall, the sensitivity of US in detecting bone erosions in the present study was lower than the sensitivity of MRI. Several other studies have reported that the sensitivity of US is best at the most easily accessible joints (that is, the 2nd and 5th MCP joints), where visualization of the joint is possible from three aspects [20
]. In this study we also achieved the highest sensitivity at these joints, but even in these joints there are certain bone surfaces that are not accessible to US assessment, contributing to the lower sensitivity of US. When looking at the joint surfaces accessible for US examination (that is, the dorsal and palmar aspects of all joints and the radial aspect of the 2nd and the ulnar aspect of the 5th MCP joint), US achieved markedly higher sensitivity (60%) compared with CT.
In two patients no erosions were detected on US, however, in all four healthy control individuals 'false positive' erosions were registered, whereas none of these were seen on CT, MRI and radiography. On US three control individuals had one erosion-like lesion each, whereas in the last control individual eight sites were registered with erosion-like lesions. It should be noted that the latter control individual developed a HLA-B27 positive arthritis one year later, which was treated with sulfasalazine. There was no tendency toward any specific regions in which erosion-like changes were seen on US. That erosion-like changes were observed in healthy control individuals in the present study is in agreement with previous US studies [20
], even though the frequency was markedly higher in the present study. Furthermore, small well defined bone defects at the dorsal aspect of the metacarpal head have been reported in 37% of healthy control individuals [32
]. Work on standardization of definitions of pathology in musculoskeletal US is being done in the setting of the European League Against Rheumatism (EULAR) Working Party for Ultrasound and OMERACT, and the definitions on erosions are not discordant with those used in the present study [33
When finding pathological changes in healthy control individuals, using any diagnostic procedure, it should be considered whether the method is too sensitive, and our findings question whether the definition of US erosions used in the study is optimal. However, there is an inherent, usually divergent balance between the sensitivity and specificity of a test. In clinical trials, in which the diagnosis is established, a high sensitivity is often of fundamental importance. However, in a diagnostic setting high specificity may have the highest priority, because the diagnosis has important implications for classification and treatment.
Despite descriptions in the literature of CT findings in RA peripheral joints [12
], CT is not a thoroughly validated method in RA, and use of CT as a standard comparator for better validated imaging methods such as MRI and US could therefore be questioned. However, high-resolution CT is the optimal radiographic method because it provides high-resolution tomographic direct visualization of calcified tissue, and CT is known from other skeletal conditions to be highly accurate. Thus, although CT findings may not represent the absolute truth, we found that comparison with CT provided very important validation of MRI and US findings.
Examination of the inter-reader agreement was not the objective of this study, and because previous papers have reported good inter-reader agreements for the readers of MRI (BE) [35
] and US (MC-P) [31
] involved in the present study, evaluations of magnetic resonance images and the US examination were done only once. CT, being less validated in RA, was evaluated by two readers (MØ and MH) in order to calculate the inter-observer agreement in reading CT images. The good inter-reader agreement of 90% is comparable with inter-observer agreements achieved in other studies with other imaging modalities [31
The present study suggests that CT may be a very sensitive method for detecting early bone erosions, possibly even more so than MRI and US, but further studies (for example, on validity) are needed before any general recommendations on the use of CT in RA can be given.