Article Text

Extended Report
The specificity of ultrasound-detected bone erosions for rheumatoid arthritis
  1. Ahmed S Zayat1,
  2. Karen Ellegaard2,
  3. Philip G Conaghan1,
  4. Lene Terslev3,
  5. Elizabeth M A Hensor1,
  6. Jane E Freeston1,
  7. Paul Emery1,
  8. Richard J Wakefield1
  1. 1Leeds Institute of Rheumatic and Musculoskeletal Medicine, University of Leeds & NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds, UK
  2. 2Department of Rheumatology, Parker Institute, Frederiksberg Hospital, Copenhagen, Denmark
  3. 3Center for Rheumatology and Spine Diseases, Copenhagen University Hospital at Glostrup, Copenhagen, Denmark
  1. Correspondence to Dr Ahmed S Zayat, Leeds Institute of Rheumatic and Musculoskeletal Medicine (LIRMM), Second Floor, Chapel Allerton Hospital, Leeds LS7 4SA, UK; aszayat{at}


Background Bone erosion is one of the hallmarks of rheumatoid arthritis (RA), but also seen in other rheumatic diseases. The objective of this study was to determine the specificity of ultrasound (US)-detected bone erosions (including their size) in the classical ‘target’ joints for RA.

Methods Patients fulfilling the diagnostic criteria for RA, psoriatic arthritis, osteoarthritis or gout in addition to healthy volunteers were included. The following areas were examined by US: distal radius and ulna, 2nd, 3rd and 5th metacarpophalangeal (MCP), 2nd and 3rd proximal interphalangeal (PIP) and 1st and 5th metatarsophalangeal (MTP) joints. All joints were scanned in four quadrants using both semiquantitative (0–3) and quantitative (erosion diameter) scoring systems.

Results 310 subjects were recruited. The inter-reader and intrareader agreements were good to excellent. US-detected bone erosions were more frequent but not specific for RA (specificity 32.9% and sensitivity 91.4%). The presence of erosions with semiquantitative score ≥2 in four target joints (2nd, 5rd MCP, 5th MTP joints and distal ulna) was highly specific for RA (specificity 97.9% and sensitivity 41.4%). Size of erosion was found to be associated with RA. Erosions of any size in the 5th MTP joint were both specific and sensitive for RA (specificity 85.4% and sensitivity 68.6%).

Conclusions The presence of US-detected erosions is not specific for RA. However, larger erosions in selected joints, especially 2nd and 5rd MCP, 5th MTP joints and distal ulna, were highly specific for and predictive of RA.

  • Ultrasonography
  • Rheumatoid Arthritis
  • Osteoarthritis
  • Psoriatic Arthritis
  • Gout

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Bone erosions are commonly considered to be the hallmark of rheumatoid arthritis (RA).1 They represent a destructive consequence of untreated synovitis and osteitis. One of the major goals in the treatment of RA is the arrest of erosion development in order to prevent irreversible functional disability.2 Until recently, the presence of radiographic bone erosions was one of the American College of Rheumatologist (ACR) 1987 classification criteria for RA.3 The new ACR/European League Against Rheumatism (EULAR) 2010 criteria allow the diagnosis of RA in the presence of typical erosions even without fulfilment of the other criteria within the scoring system.4

New more sensitive imaging techniques such as ultrasound (US) are increasingly used in disease assessment. Although US is more sensitive than conventional radiography (CR) for detecting bone erosions,5 there are limited data regarding the specificity of US-detected bone erosions for RA in comparison with other arthritides. There are also limited data about the effect of location, size and severity of erosion on US specificity. The objective of this study was to investigate the discriminative ability of US by studying the specificity of US-detected bone erosions for RA. We aimed to determine: (a) if the frequency of US-detected bone erosions in RA is significantly higher than other arthritis groups and normal controls, (b) the specificity of US-detected erosions in suggested target joints for RA and (c) the effect of erosion size and location on their specificity for RA.

Patients and methods

The study was conducted at two sites, Leeds and Copenhagen.


Patients fulfilling the diagnostic criteria for RA, psoriatic arthritis (PsA), osteoarthritis (OA) and gout were included in the study.3 ,6 ,7 During the screening process, patients with mono/oligo-articular PsA or knees/hip OA alone were not included as we aimed for polyarticular disease. Subjects who had undergone arthroplasty in their hands or feet, those with a diagnosis of more than one type of arthritis, undifferentiated arthritis or those who did not meet the diagnostic criteria of any of the included diseases were excluded. A group of healthy controls (HCs) without a history suggestive of arthritis was also included. The study was approved by the local ethics committees at both centres and informed written consent was obtained from all patients and healthy volunteers before study enrolment.

Ultrasonographic examination

Ultrasonography was performed at both sites using a General Electric Logiq E9 US machine (General Electric, Wauwatosa, Wisconsin, USA) with a multi-linear 6–15 MHz probe. Two investigators (ASZ and KE) trained in musculoskeletal US performed all examinations either at Chapel Allerton Hospital, Leeds, UK or Copenhagen University Hospital at Glostrup, Copenhagen, Denmark, respectively. The investigators were blinded to clinical diagnosis and findings.

The following joints or bony areas were examined bilaterally in each patient; distal articular radius and ulna including caput and styloid processes, 2nd, 3rd and 5th metacarpophalangeal (MCP), 2nd and 3rd proximal interphalangeal and 1st and 5th metatarsophalangeal (MTP) joints. Joints were examined both proximally and distally to the joint margin. All joints were scanned in quadrants (figure 1). Each scan was performed in longitudinal and transverse planes. An erosion was defined as a discontinuity of the cortical bone surface that was visualised in at least two perpendicular planes according to the OMERACT definition.8 Each quadrant was scored using a transverse view according to a 0–3 semiquantitative scoring system: 0=no erosions, 1=erosions covering less than a third of the surface of the quadrant, 2=erosions covering between one- to two-thirds of the surface of the quadrant and 3=erosions covering more than two-thirds of the surface of the quadrant (figure 2). A total score of 0–3 was then given to each joint (proximal or distal) by summing the joint quadrants scores and dividing by four. The diameter of the largest erosion in each joint (proximal or distal) was recorded mostly on longitudinal view unless maximum diameter was provided by transverse view. The inter-observer and intraobserver reliability were tested by rescoring 50 static US images by both observers in both centres in two occasions blindly.

Figure 1

Scoring method of bone erosion. Each bone surface was divided into ‘virtual’ quadrants on transverse view. The degree of bone affected determined the score given. 0=no erosions, 1=erosions covering <⅓, 2=erosions covering ⅓–⅔ and 3=erosions covering >⅔ of the surface of the quadrant.

Figure 2

Erosion in 5th metatarsophalangeal joint. (A) Erosion in longitudinal view and (B) transverse view. PL, plantar lateral; PM, plantar medial; MT, metatarsal; P, Phalanx; JS, joint space. This was scored 2.

Statistical analysis

Statistical analysis was performed using Stata V.12.1 and WinPEPI 11.4. Sample size was based on attaining a desired 95% CI of ±5% around an estimate of specificity of 90%. When counting the number of joints with erosions present, the proximal and distal regions of joints were counted separately. To assess the overall differences in the number of erosions between disease groups, negative binomial regression with robust SEs was used. Heteroskedasticity-robust quantile regression was used to compare the erosion scores and maximum erosion diameters between groups. Models were adjusted for age and centre, which was treated as a fixed factor. Receiver Operating Characteristics (ROC) analysis was used to identify cut-offs for erosion diameter which concurrently maximised sensitivity and specificity and/or maximised the Youden index J (sensitivity+specificity − 1) for different diagnoses. Analysis was performed twice, once with normal subjects included and once when they were excluded. The inter-observer and intraobserver reliability of the semiquantitative scores was assessed using quadratic-weighted κ (κw) statistics.9


Patient demographics

A total of 310 subjects were recruited including 70 RA, 60 PsA, 60 gout, 60 OA and 60 healthy volunteers. In all, 188 (61%) were women. The RA group consisted of 28 patients with early RA and 42 patients with established RA (table 1). In all, 80% of the patients were positive to either rheumatoid factor or anticitrullinated protein antibodies.

Table 1

The demographic and clinical characteristics of the different disease groups and normal volunteers in the study

US assessment


The inter-centre agreement was excellent (κw (95% CI) 0.88 (0.76 to 1.00), p<0.001). The percentage exact agreement (PEA) was 88% (44/50) while the percentage close agreement (PCA; within 1 category) was 98% (49/50). The intraobserver agreement values for each centre were also excellent (κw (95% CI) 0.87 (0.79, 0.96), PEA 82%, PCA 100% for Leeds and 0.82 (0.69 to 0.95), PEA 80%, PCA 98% for Copenhagen).

Number of joints with erosions

Compared with PsA, gout and OA, RA patients had twice as many joints with erosions (overall score >0) than patients with PsA (incident rate ratio (95% CI) 2.50 (1.87 to 3.35), p<0.001) or gout (2.28 (1.71 to 3.04), p<0.001), and five times as many as those with OA (5.41 (3.92 to 7.47), p<0.001). PsA patients did not differ from those with gout (0.91 (0.64 to 1.29), p=0.593) but had more eroded joints than those with OA (2.16 (1.49 to 3.14), p<0.001). The number of affected joints increased with age (incidence rate ratio per year 1.02 (1.01 to 1.03), p<0.001) and showed significant overdispersal (p<0.001). In a separate analysis, all four disease groups were found to have more affected joints than the HC group (RA 11.86 (7.81 to 18.01), PsA 4.74 (3.04 to 7.41), gout 5.17 (3.25 to 8.23), all p<0.001; OA 2.16 (1.33 to 3.53), p=0.002).

RA patients had more joints with extensive erosions (overall score ≥2) than PsA patients (4.31 (2.08 to 8.96), p<0.001), gout patients (6.67 (3.23 to 13.77), p<0.001) and OA patients (32.61 (11.76 to 90.46), p<0.001). None of the HC had any joints scoring ≥2 for erosion.

Size of erosions

The median total erosion score was higher in RA than PsA (adjusted difference (95% CI) 3.65 (1.86 to 5.44), p<0.001), gout (3.90 (2.09 to 5.71), p<0.001) and OA (5.90 (4.11 to 7.70), p<0.001) (table 2). In a separate analysis, patients with RA (5.03 (2.95 to 7.11), p<0.001), PsA (1.56 (0.75 to 2.38), p<0.001) or gout (1.35 (0.51 to 2.20), p=0.002) were found to have higher total erosion scores than the HC group, but OA patients did not differ from HC (−0.37 (−1.31 to 0.10), p=0.442). Total erosion score increased with age by 0.07 (0.05 to 0.10) units per year (p<0.001). The total erosion score increased with symptom duration by 0.014 (0.007 to 0.020) units per month (p<0.001).

Table 2

Ultrasound (US) erosion findings in different disease groups (RA, PsA, gout, OA) and healthy controls (HCs)

The median maximum erosion diameter recorded in any of the patient's joints was higher in RA patients than PsA (adjusted difference (95% CI) 1.79 mm (1.13 to 2.46 mm), p<0.001), gout (0.99 (0.12 to 1.86), p=0.026) or OA (2.84 (1.87 to 3.81), p<0.001). Patients with PsA had smaller erosions than those with gout (−0.80 (−1.49 to −0.12), p=0.021) but larger erosions than those with OA (1.05 (0.29 to 1.81), p=0.007). In a separate analysis, patients with RA (2.48 (1.80 to 3.17)), PsA (0.89 (0.42 to 1.35)) or gout (1.76 (1.08 to 2.44)) were found to have larger erosions than the HC group (all p<0.001), but OA patients did not differ from HC (−0.16 (−0.93 to 0.61), p=0.677).

Differences between early and established RA

Adjusting for age and centre, patients with early RA had significantly fewer joints with erosions than established RA (table 3). Although the median of the total erosion score was lower in the early RA group in comparison with established RA, there was no substantive difference in the median diameter of erosions. There were no differences in the proportions of patients with at least one joint scoring >0 for erosion, irrespective of the sites involved, but patients with established RA were more likely to have at least one joint with more extensive erosions (score ≥2), whether all joints were included, or a reduced set (table 3).

Table 3

Differences in erosion occurrence between early and established RA

Sensitivity and specificity for RA

The sensitivity of erosions for RA was initially calculated with the HC included. The presence of any joint with any degree of erosion (score >0) was not specific for RA (32.9%) but was sensitive (91.4%), whereas the presence of any joint with more extensive erosions (score ≥2) was specific for RA (89.2%) but not very sensitive (50.0%) (table 4). Excluding the HCs resulted in a reduction in specificity to a greater degree when the criterion was erosion score >0 (by ∼10%) than when the criterion was erosion score ≥2 (by ∼3%). Excluding the established RA patients had reduced the sensitivity of erosion score ≥2 (from 50% to 25%).

Table 4

Sensitivity and specificity of ultrasound (US) erosions in distinguishing RA patients from patients with other arthritides or from healthy controls

When determining whether the size of erosion was predictive and specific for RA, ROC curve analysis (HC excluded) indicated that the diameter of the largest erosion in any of the observed joints was predictive of RA (area under the ROC curve (95% CI) 0.73 (0.65 to 0.80), p<0.001). The sensitivity and specificity were concurrently maximised at a cut-off of ≥2.5 mm (sensitivity 68.6%, specificity 68.3%).

Next, the specificity for the presence of erosions in certain individual joints for RA was examined. It is important that a diagnostic tool provide both reasonable sensitivity and specificity. We therefore decided that candidate joints needed to show at least 80% specificity and 30% sensitivity using a criterion of erosion score >0. Among the target joints, the US-detected erosions in the 2nd and 3rd proximal interphalangeal joints and 3rd MCP joint were highly specific for RA but none of these was sufficiently sensitive (table 4). Erosions in 1st MTP joint and distal radius were sensitive but not sufficiently specific for RA. When excluding these sites from the analysis, specificity improved, although the presence of any degree of erosion in the distal ulna, 2nd MCP, 5th MCP or 5th MTP joints was still was not highly specific for RA (68.8%). However, it was sensitive (81.4%) and this remained the case when established RA patients were excluded (75.0%). The presence of a more extensive area of erosion (score≥2) in the distal ulna, 2nd MCP, 5th MCP or 5th MTP joints was highly specific for RA (97.9%), but was moderately sensitive (41.4%). The sensitivities and specificities of erosions in RA when analysis was restricted to individuals with erosions are presented in the online supplementary material table S1.

The presence of any degree of erosion in the metatarsal head of the 5th MTP joint was specific for RA (85.4%) and was substantively sensitive (68.6%), remaining so after the exclusion of established RA patients (53.5%). Because the presence of erosions in the 5th MTP joint was moderately sensitive and highly specific for RA regardless of the score, we explored whether erosions tended to be more specific in certain quadrants. Sensitivity was higher in the lateral quadrants than the medial quadrants (40%–53% vs 14%–20%) while specificity was high (>90%) irrespective of quadrant location (see online supplementary material table S2). Sensitivity and specificity were nearly identical in the upper and lower lateral quadrants (erosion score >0 sensitivity 52.9%, specificity 92.5%; erosion score ≥2 sensitivity 40.0%, specificity 95.8% (upper) 97.5% (lower)).

When looking at what size of erosion was predictive and specific for RA, ROC analysis (HC excluded) indicated that the diameter of the largest erosion in the distal ulna, 2nd MCP, 5th MCP and 5th MTP joints was predictive of RA (area under the ROC curve (95% CI) 0.78 (0.71 to 0.84), p<0.001). Using a cut-off of ≥1.5 mm achieved the best trade-off between sensitivity and specificity (sensitivity 70.0%; specificity HC excluded=68.9%). To maintain consistency, we used the same cut-off used for all joints (2.5 mm; sensitivity 58.6%; specificity HC included=90.0%, HC excluded=86.7%; J=0.45) (table 4). Similar results were found when each joint was considered individually. Further results of maximum diameters providing the best trade-off between sensitivity and specificity including only early RA group are presented in the online supplementary table S3.

In the target RA-specific sites, erosions were more frequent at the proximal dorso-radial quadrant of 2nd MCP, proximal dorso-ulnar quadrant of 5th MCP joint and dorso-ulnar side of the distal ulna. In the 5th MTP joints, erosions were more frequent in the proximal dorso-lateral and plantar-lateral quadrants (see online supplementary material tables S4 and S5). In HCs, the most frequently involved joint quadrant was the proximal dorso-radial quadrant of the distal radius, which was affected in 11 patients (18.3%).


This study explored the diagnostic discriminative ability of US by investigating the specificity of US-detected bone erosions in RA. The results of this study demonstrate that although the overall presence of US-detected bone erosions are more frequent in RA compared with other common joint diseases, they are not necessarily a disease-specific finding. However, the presence of erosions with score ≥2 in any of the four suggested target joints (2nd and 3rd MCP, 5th MTP joints and distal ulna) was highly specific for RA. The presence of any erosion in the 5th MTP joint was both sensitive and specific for RA. The predominance of erosions in some quadrants of the target joints in our data can be supported by previous radiographic, CT and MRI data.10–13 A focused examination of the four suggested target areas with understanding of importance of erosion score and size might help reduce the time required for US examination and minimise the potential for false-positive results. This may help in avoiding repeated investigations and preventing unnecessary follow-up when potential false-positive erosions are found in the absence of supporting symptoms and signs.

This study has explored the specificity of US-detected bone erosions by comparing different diseases groups. There were extensive data comparing US with different imaging modalities.14 US has previously been reported to be more sensitive than CR for the detection of bone erosions.5 ,15–18 The sensitivity increases in joints that are easily accessible for US (eg, 2nd and 5th MCP, 5th MTP joints and distal ulna) but low in joints that are harder to access due to acoustic window (eg, 3rd MCP joints).19

The specificities of US-detected bone erosions in RA were different according to joint area and erosion size. This could potentially mean that erosions in RA might be different from other arthritides with respect to location and size. There have been different mechanisms of erosion development described previously related to different arthritides which can partially explain our results.20–23 We showed that small size erosions have low specificity. Previous data suggested that false-positive results were usually caused by small sized bone defects which represented physiological vascular bone channels rather than break in the bone cortex.24 We showed that a cut-off point of ≥2.5 mm maximises sensitivity and specificity. Previous data compared US with micro-CT revealed that high US specificity was obtained when more stringent criteria for erosions were used including considering cortical lesion of ≥2 mm.24

We have demonstrated the presence of erosions in HC. The presence of bone erosion in HC was demonstrated by previous MRI17 ,18 ,25 and US studies.17 ,26 ,27 We realised the importance of identifying aspects of erosions that affect their specificity in RA. A recent EULAR task force tried to define erosive disease in light of the 2010 ACR/EULAR RA classification criteria.28 They have suggested that erosions (detected by CR) have to be seen in at least three separate joints in RA. Our study has identified other aspects of erosions detected by US that can potentially affect their specificity for RA.

This study has a number of limitations. The specificity of erosions was mildly lower when normal subjects were excluded; however, this may better represent the real-world clinical situation. Despite attempts to recruit patients within similar age ranges, there were some differences in the mean age between the groups. The OA group was marginally older than other groups. Comparisons between the groups were therefore adjusted for age. Similarly, there was mild variance in the duration of symptoms. OA and gout did have mildly longer mean symptoms duration than RA and PsA groups despite attempts to standardise. This reflects the natural history of these diseases where most OA patients present years after initial symptoms and gout has relapsing remitting nature.

We were able to describe a new scoring system for US-detected bone erosions. Previously, different non-validated scoring systems for US-detected bone erosions have been used.17–19 ,29–33 We believe that the proposed scoring system in this study is simple and reproducible. This scoring system we devised could be used in other studies after being tested in longitudinal studies. Surprisingly, the inter-observer (inter-centre) agreement over scores was slightly higher than the intraobserver reliability at either centre. However, the differences were not substantive.

In conclusion, US does have the discriminative ability to differentiate RA from other diseases if only large erosions in certain target sites are considered. The target sites which are most specific for RA are the distal ulna, 2nd MCP, 5th MCP and 5th MTP joints. However, the presence of US-detected bone erosions per se is not specific for RA. The erosions of RA were more extensive and generally larger than other diseases.


We would like to thank Dr S Ayden and Dr S Das for their contribution by clinically assessing study patients.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • Correction notice This article has been corrected since it was published Online First. The spelling of the second author's surname has been corrected.

  • Contributors Study design: ASZ, RJW and PGC. Data acquisition: ASZ, KE, LT, SA, SS. Data analysis: LH. Manuscript preparations: ASZ, RJW, PGC, PE. All authors have approved the final manuscript.

  • Funding This project was performed in the Leeds institute of Rheumatic and musculoskeletal Medicine, Parker Institute and Copenhagen University Hospital at Glostrup without external funding requested.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Leeds East Ethics committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.