Article Text
PDF
Abstract
Objectives To determine the accuracy of ultrasonography (US) for bone erosion detection in different areas of rheumatoid arthritis (RA) metacarpophalangeal (MCP) joints with multislice CT as the reference method. Second, to establish the necessary bone volume loss on CT for US to reliably detect it as an erosion, and finally to compare two semiquantitative US-erosion scoring methods.
Methods The 2nd–5th MCP joints of 49 patients with RA were examined by CT and US, and evaluated for the presence of bone erosion in each MCP joint quadrant. On CT, erosion volume was scored according to the OMERACT-RAMRIS score (bone volume loss in 10% increments of original bone volume). US erosions were scored 0–3 according to the Szkudlarek and Scoring by UltraSound Structural erosion (ScUSSe) systems, respectively.
Results Seven hundred and eighty-four MCP joint quadrants were examined. Erosions were detected by CT in 259 quadrants and by US in 142 quadrants. Sensitivity/specificity/accuracy of US was overall 44%/95%/78% compared with 71%/95%/90% in areas with good US accessibility (radial 2nd MCP, ulnar 5th MCP and all dorsal/palmar aspects). US detected 95% of erosions with bone volume loss >20%. In US accessible areas, 63% of erosions with 1–10% bone volume loss and 94% of erosions with >10% bone loss were detected. The two US scoring systems agreed well on large erosions, whereas the smallest erosions (Szkudlarek grade 1, of which 86% were confirmed by CT) were not scored by ScUSSe.
Conclusion In accessible areas, US was highly accurate for detection and semiquantitative assessment of RA bone erosion. Even the smallest erosions, only detected in one plane, were generally confirmed by CT.
Statistics from Altmetric.com
In rheumatoid arthritis (RA), the presence of radiographic bone erosions is important for the classification of RA according to both the American College of Rheumatology (ACR) 1987 classification criteria for RA and the ACR/EULAR 2010 criteria.1,–,3 Erosions are a common finding in patients with RA, and the majority of patients develop erosions during the first 2 years of disease.4 The prevention of bone damage is an essential goal in the treatment of RA and therefore, accurate and early diagnosis of RA, including early detection of erosive joint changes, is imperative. Ultrasonography (US) has been demonstrated in several studies to be more sensitive than conventional radiography in detecting bone erosions in metacarpophalangeal (MCP), proximal interphalangeal (PIP) and metatarsophalangeal (MTP) joints.5,–,7 However, US is inferior to MRI in the ability to follow erosive changes over time in the wrist and hand.8 The reason for the lower sensitivity has partly been explained by the fact that some areas of the joints are inaccessible. In accordance with this, US performs better at the 2nd and 5th MCP compared with the 3rd and 4th, where access to the lateral aspect of the joints is impaired.6 ,9 Scoring systems for US bone erosions have been developed6 ,10 ,11 to determine the severity of destructive changes and to monitor progression. The existing scoring systems, which are all semiquantitative, deal with different aspects of erosions. The scoring system by Szkudlarek et al10 evaluates the presence and extent of erosions, whereas, the scoring systems by Wakefield et al and Sommier et al 6 ,11 are related to the size of the erosions.
US is increasingly being used in daily routine care to diagnose and monitor patients with arthritis, and it is therefore of great interest to evaluate the reliability and performance of US for erosion detection. MRI is often used as the gold standard for US in detection of erosive bone changes; however, CT has been shown to be even more sensitive for detecting erosions than MRI, and therefore, seems a better reference method.9 ,12,–,14
The aims of the present study were (with multi-slice CT as the reference method) to evaluate the accuracy of US in different areas of MCP joints and to estimate the minimal bone volume loss necessary for high-resolution US to detect the erosion. Furthermore, we wanted to compare two semiquantitative US scoring methods of erosion.
Materials and methods
Baseline imaging data from 49 patients with RA included in an investigator-initiated longitudinal imaging study were used for comparing erosion detection by CT and US in RA MCP joints. Patients who were enrolled were to commence adalimumab and methotrexate combination therapy. The inclusion and exclusion criteria have been described previously.15 In brief, the inclusion criteria were: active RA with a disease activity score (DAS28(3)-C reactive protein) >3.2, indication for anti tumour necrosis factor therapy according to the treating rheumatologist, methotrexate treatment for ≥4 weeks prior to inclusion, no previous biological treatment and a minimum of two low-grade erosions (Larsen grade 2–3) in the wrist of one hand or MCP joints.
Image acquisition and assessment
The 2nd–5th MCP joints of one hand of 49 patients with RA were examined by CT and US, and evaluated for the presence of bone erosion in each MCP joint quadrant (radial and ulnar part of the metacarpal head and phalangeal basis, respectively). US and CT images were assessed independently (UMD and MØ, respectively), blinded to clinical and other imaging data.
CT was performed on a Philips Mx8000IDT multidetector unit (Philips Medical Systems, Cleveland, Ohio, USA). Parameters used were voxel-size 0.4 mm×0.4 mm×0.4 mm, pitch 0.4 mm, slice spacing 0.4 mm, overlap 50%, 90 kV and 100 mAs. Images were reconstructed in the coronal and axial planes with a slice thickness of 0.4 mm. CT images were scored for erosions according to the definitions and principles of the OMERACT RA MRI scoring method (RAMRIS).16 ,17 Erosions were defined as a sharply marginated bone lesion, with correct juxta-articular localisation, visible in two planes, and with a cortical break seen in at least one plane. Erosions were scored on a 0–10 scale based on the proportion of eroded bone compared with the ‘assessed bone volume’ judged on all available images: 0=no erosion; 1=1–10% of the bone eroded; 2=11–20% of the bone eroded; 3=21–30% of the bone eroded, and so on. The assessed bone volume covered the area from the articular surface to a depth of 1 cm.
US examinations were performed by UMD using a General Electric LOGIQ9 unit (General Electric Medical Systems; Little Chalfont, Buckinghamshire, UK), equipped with a 14–9 MHz linear array transducer. The lateral and axial resolution of the used probe was 0.33 mm (17 mm depth (−6 dB (mm)) and 0.16 mm (36 mm depth (−6 dB (mm)). Joints were examined with longitudinal and transversal scans. Representative images of all joints were stored as DICOM files and assessed at a later time-point.
The definition of US bone erosions was a change in the bone surface of the area adjacent to the joint, as suggested by Szkudlarek.10 For reliability purpose, we also applied the more strict OMERACT definition of US bone erosions, that is an intra-articular discontinuity of the bone surface that is visible in two perpendicular planes.18 In addition to the assessment of presence/absence of erosion in each joint quadrant, erosions were scored on a 0–3 scale according to Szkudlarek et al (grade 0=regular bone surface; grade 1=irregularity of the bone surface without formation of a defect seen in two planes; grade 2=formation of a defect in the surface of the bone seen in two planes and grade 3=bone defect creating extensive bone destruction)10 and on a 0–3 scale according to the Scoring by UltraSound Structural erosion (ScUSSe) scoring system (erosions scored by the maximal length: 0=no erosions, 1=<2 mm, 2=2–3 mm, 3=>3 mm or multiple erosions).11 All Szkudlarek grade 1 erosions were not included in the comparison with the ScUSSe system, as they are not considered a definite erosion according to the ScUSSe.
Statistics
With CT as the reference method, the sensitivity, specificity and accuracy of US for erosion detection in RA MCP joints were calculated. This was done for all areas and repeated for areas with good access for US examination (dorsal and palmar aspects of all MCP joints and, additionally, the radial surface of the second and ulnar surface of the fifth MCP joint). The analysis was repeated using only Szkudlarek grades 2 and 3 erosions, thereby including only erosions fulfilling OMERACT definition. The agreement between CT and US was additionally evaluated by κ-values and by the proportion of US erosions confirmed by CT. Reliability analysis of CT and US assessment has been described previously15 ,19 The Statistical Package for the Social Sciences (SPSS) V.15.0 was used for the statistical analysis.
Results
The patient demographics are presented in table 1.
Patient demographics
The total number of examined quadrants was 784. Erosions were detected by CT in 259 quadrants and by US in 142 quadrants. With CT as the reference method, US had an overall sensitivity of 44%, specificity of 95% and accuracy of 78% for the detection of bone erosions in RA MCP joints. Values differed according to assessed joint areas, with the highest sensitivity of US at the second metacarpal head (table 2).
Sensitivities, specificities, accuracies and κ-values of erosion detection by ultrasonography, with CT as the reference method
Out of the 142 quadrants with erosions on US, 115 (81%) were confirmed by CT (see figure 1 for an example of agreement between CT and US).
Ultrasonography in longitudinal view (A) and corresponding CT image in the coronal plane (B) of the second metacarpophalangeal joint of a patient with rheumatoid arthritis. Erosions at the radial aspect of the metacarpal head (white arrow) and the base of the proximal phalanx (open arrow) were detected by ultrasonography and confirmed with CT. The erosion at the phalanx was grade 1 according to the Szkudlarek system, illustrating that even small cortical irregularities not confirmed in two planes, often represent true erosive changes.
When only areas with good accessibility were included, the sensitivity of US increased to 71%, whereas, the specificity was 95% and the accuracy 90%. Applying the OMERACT definition of US erosions, the overall sensitivity, specificity and accuracy was 38%, 95% and 76%, respectively, while in accessible areas it was 60%, 95% and 88%, respectively. In 27 out of the 142 quadrants with US erosions, the US erosion was not confirmed by CT. These were primarily located at the second (7) and fifth (11) metacarpal head. When reassessing these areas, different reasons for the disagreement between US and CT appeared. Cortical irregularities were found on all US images, retrospectively originating from osteophytes, notches at the metacarpal neck and subcortical bone cysts. However, in seven cases, erosions were found on reassessment of CT, while indentations in the bone surface were found in five cases.
Bone volume loss on CT versus detection by US
In order to evaluate the minimal bone volume loss necessary for US to detect the erosion, we used the CT OMERACT-RAMRIS score. US detected 80% of quadrants with erosions where the CT-determined bone volume loss was above 10% irrespective of the localisation. When only areas accessible to US were included, the proportion increased to 94% (table 3).
Erosion detection by ultrasonography (US), stratified by estimated percentage bone volume loss on CT
Of note is that two large CT erosions (grade 3 and 9) were not detected by US, and these were subsequently investigated. When CT images of the grade 3 erosion were reassessed it was evident that no cortical break was seen in areas accessible to US (figure 2). The grade 9 CT erosion was located at the proximal phalanx and expanded over two neighbouring joint quadrants in an extensively damaged joint. This erosion was only documented as present in the adjacent quadrant. If these two erosions were excluded from the analysis, all joint quadrants with a bone volume loss above 20% could be correctly identified by US, irrespective of localisation and, in areas accessible to US, all those with a bone loss above 10%.
Coronal (A) and axial (B) CT images of the second to fifth metacarpophalangeal joints of a patient with RA. The fourth metacarpal head shows a grade 3 erosion (white arrow), which was not detected by ultrasonography (US). This illustrates that even large erosions may be missed on US, if the cortical break is located at a position not accessible for US examination. Note the large subcortical area of loss of trabecular bone just below the palmar surface (open arrow).
Comparison between Szkudlarek and ScUSSe US scoring systems of erosions
The distribution of erosion grades in the two different scoring systems is shown in figure 3. As the Szkudlarek grade 1 erosion is not considered an erosion according to the ScUSSe system or OMERACT definition, we also investigated the proportion of those that could be confirmed by CT. We found that 86% of the Szkudlarek grade 1 erosions were confirmed by CT, the majority of those were located at the base of the proximal phalanx. The corresponding values for Szkudlarek grades 2 and 3 were 62% and 94%, respectively.
Distribution of erosion scores according to the scoring by ultrasound structural erosion (ScUSSE) system11 and corresponding scores by the Szkudlarek system.10 Note that practically all Szkudlarek grade 3 erosions are also ScUSSe grade 3, whereas the Szkudlarek grade 2 erosions are equally divided on ScUSSe grades 1–3.
Discussion
Ultrasonography is increasingly becoming a routine clinical tool for detection of signs of joint inflammation. During US examination, information about the bone surface in accessible areas can also be obtained. It is therefore of interest to know how reliable US detection of erosions is. The present study showed an overall moderate sensitivity of 44% and very high specificity of 95% for US-detected erosions when using CT as the reference method. However, the sensitivity improved considerably when only areas with good US accessibility were included (71%). Our data are in agreement with the findings in a smaller study (17 patients with RA and 4 healthy controls) published earlier by our group, in which US had an overall moderate sensitivity of 42% for erosion detection in MCP joints.9 In a study by Alasaarela et al from 1998, US showed more erosions than CT in the humeral head of 26 patients with RA.20 The difference from our data is possibly explained by different joint types and differences in equipment. Wakefield et al found complete agreement between MRI and US for detecting bone erosion at the radial surface of the second MCP joint in 25 patients with RA.6 However, Døhn et al in the above-mentioned study,9 investigating all MCP joint areas, reported a considerable difference in sensitivities between MRI and US (68% vs 42% in RA MCP joints). Limiting the analysis in that study to very easily accessed areas could have led to similar results, as in the study by Wakefield et al.
It should be mentioned that only MCP joints were investigated in the present study. Compared with MRI (and maybe also radiography), US is expected to perform worse with respect to erosion detection when it comes to more complicated joints such as the wrist. Overall, our findings indicate that detection by US of bone erosions is valid, and suggest that US can be used for diagnosing erosive disease with a high specificity and, in areas with good access for US assessment, a sensitivity comparable with that of MRI. Even though erosions on US are not currently accepted as a criterion for classification of RA,1,–,3 the high specificity of US for bone erosions suggests that US-detected bone erosion may play a role in the future for classification of RA.
In the present study, we wanted to establish how substantial the bone volume loss needed to be before US with certainty can detect it as an erosion. This was done by comparison with CT, scored according to the OMERACT-RAMRIS. It should be mentioned that the OMERACT-RAMRIS scoring system was originally made for scoring erosions on MRI, but has also been applied for scoring erosions on CT.12 ,15 ,21 We found that even minor erosive changes (1–10% of the assessed bone volume (OMERACT-RAMRIS grade 1) as judged on CT, were detected by US in 37% of cases, whereas, larger erosions (grades 2 and 3) were detected in 70% and 89% of cases, respectively, and after correction for one mis-registration, all OMERACT grades 4–10 CT erosions (≥31% of the assessed bone volume) were detected. Furthermore, when only US-accessible areas are considered, all quadrants with erosions of OMERACT grade 2 and higher (bone volume loss above 10%) could be detected by US.
In a recent study by Koski et al,22 a bone erosion phantom was developed by drilling round holes in bovine bones. Subsequent US evaluation revealed that 1.0 mm-deep erosions were detected as accurately as 4.0 mm-deep erosions. Our findings also support that US reliably detects even small bone defects/erosions.
Several scoring systems have been proposed, which all assess the erosive changes semiquantitatively.6 ,10 ,11 ,23,–,25 The present study is the first to compare different US erosion scores, and for the purpose, we used the methods described by Szkudlarek and the ScUSSe.10 ,11 The systems agree on larger erosions, whereas, the smallest (Szkudlarek grade 1) were not scored by the ScUSSe system. The fact that the majority (86%) of these Szkudlarek grade 1 erosions were confirmed by CT supports the reliability of US, even for the detection of the most subtle bone changes, and for assessment in areas such as the base of proximal phalanges, where confirmation of erosions in two planes can be difficult. For US, we chose to use published scoring systems and corresponding definitions, despite being different from the used CT scoring system. This may have influenced the agreement between the two methods. However, we found that using existing scoring systems would make the results more clinically relevant and more widely applicable. From the perspective of feasibility in routine clinical practice, these results are positive, as assessing the extent of the changes without performing measurements of erosion size, as required by the ScUSSe system, seems less time consuming. The previously reported good inter-observer agreement for scoring by the Szkudlarek method (ICC of 0.78 and unweighted κ value of 0.68) also supports its broader use.10 The fact that grade 2 erosions according to the Szkudlarek system included all sizes of erosions according to the ScUSSe system, may suggest that scoring systems using measurements of erosion size could provide a higher sensitivity to change; however, this needs further investigation.
In conclusion, our study demonstrates that US is a reliable and, in areas with good accessibility, sensitive method for detecting bone erosions in RA MCP joints. US reliably detect bone erosions which involves a CT-determined bone volume loss above 10%. The existing semiquantitative assessment of bone erosions seems comparable despite assessing different aspects of the pathology. A high CT-confirmation rate of the smallest US erosions supports that even such small erosions are registered by US and are true erosions. The present encouraging results support efforts to further develop and validate methods for erosion assessment by US, including studies of sensitivity to change.
Acknowledgments
The authors acknowledge financial support by Abbott Denmark, the Danish Rheumatism Association, and the Aase and Ejner Danielsen Foundation. Professor Henrik S Thomsen from Department of Diagnostic Radiology, Copenhagen University Hospital at Herlev is acknowledged for important scientific input and providing all CT scans.
References
Footnotes
Funding ABBOTT Denmark; The Danish Rheumatism Association; Aase and Ejnar Danielsens Foundation.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.