A detailed comparative study of high-resolution ultrasound and micro–computed tomography for detection of arthritic bone erosions

Authors


Abstract

Objective

To test whether bony lesions appearing on ultrasound (US) imaging are cortical breaks detectable by micro–computed tomography (micro-CT).

Methods

Twenty-six subjects (14 with rheumatoid arthritis, 6 with psoriatic arthritis, and 6 healthy controls) were assessed for bone erosions at the radial, palmar, and dorsal regions of the second metacarpophalangeal (MCP) joint and the palmar and dorsal regions of the third and fourth MCP joints. All patients underwent US and, for validation of the results, micro-CT scanning. The prevalence and severity of bone erosions as determined by US and by micro-CT were recorded and compared.

Results

Overall there was a good correlation between the severity of erosions as assessed by US and by micro-CT (r = 0.463, P < 0.0001). False-negative results (US negative/micro-CT positive) were obtained in only 9.9% of the joint regions and were mostly due to small erosive lesions at the dorsal sides of the MCP joints. False-positive results (US positive/micro-CT negative) were more frequent (28.6%) and were primarily based on vascular bone channels at the palmar sides of the MCP joints as well pseudo-erosions created by osteophytes.

Conclusion

These data show that the majority of bone lesions appearing on US are indeed bone erosions with a cortical break. The sensitivity of US for detecting bone erosions was high and there was a good correlation between the severity of bone erosions as assessed by US and as assessed by micro-CT. Specificity of US for bone erosions was substantially lower, suggesting that smaller lesions seen on US do not always represent breaks in the cortical bone surface.

Rheumatoid arthritis (RA) and psoriatic arthritis (PsA) are characterized by synovial inflammation and bone and cartilage destruction. While synovial inflammation is typically assessed by physical examination, measurement of bone and cartilage damage is a domain of radiography. In recent years, high-resolution ultrasound (US) has been increasingly used to assess synovial inflammation (1–4). With US one can quantitatively assess the morphologic (synovial thickness) and functional (blood flow) changes of joints during inflammatory arthritis (1–4). Several studies have shown that US is a sensitive instrument for depicting synovial inflammation in patients with RA (5, 6).

However, US has limitations for visualization of bone, as it allows detection of irregularities on the cortical bone surface but cannot penetrate bone. Pits in the juxtaarticular cortical bone surface visualized by US are considered “bone erosions” (2, 7). Recent comparative imaging studies in patients with RA have compared bone lesions detectable by US with bone erosions detected radiographically or by magnetic resonance imaging (8–11). Since resolution and overall image quality of US have improved considerably due to the use of high-frequency transducers, it is now possible to visualize even very small changes (<1 mm) in the cortical bone surface. Although bone lesions detected by US are highly suggestive of bone erosions, their exact nature remains somewhat speculative. Erosions are defined as localized cortical breaks, which allow communication between the bone marrow and the synovium (12). Whether a “pit” detected by US truly reflects a cortical break thus remains a matter of controversy, as the direct communication between the outside synovial membrane and the bone marrow cannot be proven directly.

We have recently developed a novel method to detect bone erosions in the metacarpophalangeal (MCP) joints of patients with RA and PsA, by using a micro–computed tomography (micro-CT) scanner allowing a resolution of 100 μm (13). Previous investigations have also shown that conventional CT scanning allows accurate detection of bone erosions (8, 9, 11, 14). Micro-CT, whose use is confined to the bones of the extremities, allows even higher spatial resolution with low radiation exposure (equivalent dose ∼10 μSv), similar to that obtained with conventional radiography of the hands (13).

In this study we aimed to validate whether bone changes observed by US correspond to bone erosions detected with micro-CT imaging. We performed a detailed comparative analysis of the use of high-resolution US and micro-CT to assess bone changes in patients with RA and PsA, as well as healthy individuals.

PATIENTS AND METHODS

Characteristics of the study subjects.

RA and PsA patients from the rheumatology outpatient clinic at the University Clinic of Erlangen as well as healthy individuals were invited to undergo US and micro-CT examination of the hand. Twenty-six subjects agreed to participate in the study and provided written informed consent: 14 patients with RA, 6 patients with PsA, and 6 healthy individuals. Two additional RA patients and 2 additional PsA patients initially agreed but did not provide written informed consent and were thus not included. Of the 14 RA patients, 11 were female and 3 were male. The mean ± SD age of the RA patients was 53.6 ±10.8 years, and the mean ± SD disease duration was 8.9 ± 9.4 years. Of the 6 PsA patients, 2 were female and 4 were male. The mean ± SD age of the PsA patients was 56.6 ±14.2 years, and the mean ± SD disease duration was 13.3 ± 9.2 years. The age-matched healthy control subjects (4 female and 2 male; mean ± SD age 52.5 ±16.3 years) were included because small cortical bone lesions can occasionally be found in healthy individuals (13). All controls were healthy employees of the University Clinic of Erlangen with no current or previous joint symptoms. All subjects underwent micro-CT and US of the clinically more affected hand (or the right hand in the case of healthy control subjects). The study was performed in accordance with the Declaration of Helsinki. Approval from the local ethics committee and national radiation safety agency (Bundesamt für Strahlenschutz) was obtained for the study.

Imaging procedures.

Micro-CT of the second through the fourth MCP joints was performed with an XtremeCT (Scanco Medical) at a resolution of 82 × 82 × 82 μm voxel size. For scanning, the hand was positioned in stretched posture and padded. Scanning was performed within a region of 80 slices distal and 242 slices proximal of the top of the third metacarpal head. Micro-CT scanning was performed by a single investigator (SF) trained in the technique, who was not involved in the US scanning and not aware of the results obtained with US scanning. All patients and controls also underwent high-resolution US of the second through the fourth MCP joints, using a gray-scale US mode (18 MHz) (MyLab70XV Gold, Esaote) by a rheumatologist trained in US (MB), who was not aware of the results obtained with micro-CT. The second through the fourth MCP joints were sonographically examined in a standardized manner (2) and were chosen for comparative analysis since micro-CT imaging at these sites has been evaluated previously (13). Moreover, only those sites accessible for US, i.e., the radial, palmar, and dorsal regions of the second MCP joint, the palmar and dorsal regions of the third MCP joint, and the palmar and dorsal regions of the fourth MCP joint, were used for further analysis.

Evaluation of images and semiquantitative scoring procedure.

MCP joints were evaluated for the presence and size of bone erosions in a total of 322 2-dimensional micro-CT slices. Erosions as seen on micro-CT were defined as a juxtaarticular break within the cortical shell. Bone erosions detected by micro-CT in each joint region were semiquantitatively scored on a scale of 0–3 (0 = no erosion; 1 = 1 small erosion [≤2 mm]; 2 = erosion of >2 mm; 3 = multiple large erosions with destruction of joint integrity) (13). Erosive lesions seen on US were scored on a scale of 0–5 (0 = no erosion; 1 = <1 mm; 2 = 1–2 mm; 3 = 2–3 mm; 4 = >3 mm; 5 = multiple lesions >3 mm).

Statistical analysis.

For assessing the degree of congruence between micro-CT and US in the detection of erosive lesions, we computed the sensitivity, specificity, positive predictive value, and negative predictive value of US compared to micro-CT for each of the 7 regions noted above. This analysis was done by using two different cutoffs of US grading, one that included all lesions detected on US (score ≥1) and another including only more severe lesions (score ≥3). Additionally, to investigate the relationship between US and micro-CT grading we calculated Spearman's rho. Rho coefficients were calculated with a 2-sided testing procedure, using SPSS version 18.0.

RESULTS

Prevalence, localization, and severity of erosive lesions as visualized by US and micro-CT.

We first assessed the prevalence and distribution of bone erosions. The radial side of the second MCP joint was most commonly affected by bone erosions, with a prevalence of 73% as seen on micro-CT and 81% on US. The radial region was also the predilection site of small lesions found in healthy controls, both by US and by micro-CT. Both techniques also showed that erosive lesions were more prevalent in the second and third MCP joints than in the fourth MCP joint. By micro-CT analysis, the palmar region (27%, 54%, and 23% of the second, third, and fourth MCP joints, respectively) and dorsal region (46%, 38%, and 15%, respectively) were equally affected by erosive lesions. By US, in contrast, erosions were detected predominantly at the palmar region (77%, 69%, and 58% of the second, third, and fourth MCP joints, respectively) as compared to the dorsal region (50%, 42%, and 34%, respectively).

With respect to the severity of erosive lesions, the second MCP joint was the most severely affected and the fourth MCP joint the least severely affected, as assessed by both US and micro-CT (Figure 1A). Moreover, the radial region of the second MCP joint was consistently the most severely affected region. On micro-CT the severity of erosive lesions appeared similar at the palmar and dorsal regions of the second, third, and fourth MCP joints, whereas severity seen on US tended to be greater in the palmar than the dorsal region of the MCP joints. Moreover, in most of the joints studied (66.6%), the metacarpal heads appeared more severely affected than the phalangeal bases on US. The distribution of bone lesions was equal in the metacarpal heads and phalangeal bases in 23.6% of the joints, and only 9.3% showed more pronounced involvement of the phalangeal bases. On micro-CT the predilection for the metacarpal heads was even more marked: 84.6% of the joints exhibited more lesions in the metacarpal heads, 15.4% exhibited an equal distribution, and none exhibited more severe involvement of the phalangeal bases.

Figure 1.

Comparative analysis of bone erosion as assessed by high-resolution ultrasound (US) scanning and by micro–computed tomography (micro-CT). A, Severity of bone erosions in various regions of the metacarpophalangeal (MCP) joints, assessed by high-resolution US and by micro-CT. Values are mean ± SD. B, Scatterplot showing the correlation of erosion scores obtained with high-resolution US and with micro-CT. A total of 182 sites are depicted. C, Anatomic distribution of false-negative and false-positive bone erosion findings.

Correlation of erosive lesions as determined by US and by micro-CT.

We next analyzed the correlation between the severity of erosions detected by US and the severity of lesions detected by micro-CT, by comparing the scores of the matched individual joint regions. Overall, the correlation between the severity of erosions as assessed by US and by micro-CT was good, with a Spearman's rho of 0.463 (P < 0.0001) (Figure 1B). This suggests that the majority of negative and positive findings for erosions obtained by US were correctly classified and could be verified by micro-CT.

However, there were also misclassifications showing either false-negative (US negative/micro-CT positive) or false-positive (US positive/micro-CT negative) results. Eighteen of 182 results (9.9%) were false-negative. Importantly, erosive lesions that were not detected by US but were visible by micro-CT were small lesions of <2 mm, indicating that US indeed allows accurate detection of moderate to severe erosive lesions at those joint regions. More importantly, however, 52 of 182 erosions as recorded by US (28.6%) were false-positive, with surface irregularities suggestive of bone erosions seen on US but no underlying erosion seen on micro-CT. Interestingly, the score for some of these lesions was high by US, but this was not substantiated by micro-CT. When the localization of false-negative and false-positive results was assessed in more detail, we found that the majority of false-negative erosions were in the dorsal regions of the MCP joints (predominantly along the physiologic groove of the metacarpal head), whereas the majority of false-positive erosions were located in the palmar regions (Figure 1C).

Morphologic correlates of false-positive lesions.

Whereas false-negative findings of lesions by US were much less frequent and exclusively based on small lesions as seen on micro-CT, the rather high number of false-positive lesions attracted our attention. Interestingly, many of the false-positive findings of lesions by US were based on small cortical tunnels localized at the palmar grooves of the metacarpal heads and phalangeal bases, which allow entry of blood vessels and do not represent bone erosions (Figure 2). Most of the false-positive “erosive” lesions appearing on US, localized at the palmar regions, fell into this category. These findings applied to patients with arthritis as well as healthy controls. In addition, pseudo-erosions formed by 2 osteophytes assembled like a forceps also mimicked erosive lesions. Such changes are particularly frequent in patients with PsA and could thus prevent accurate assessment of the cortical bone surface, which is hidden underneath newly formed bone.

Figure 2.

Erosions based on micro-CT (left images) and US (right images). Top, Bone erosion at the radial surface of the second MCP joint, which is visible by both micro-CT and US. Middle, Pseudo-erosion at the palmar surface of the third MCP joint, appearing on US due to a larger forceps-like osteophyte forming a de novo cavity (asterisk on micro-CT image) distant from the cortical bone surface. Bottom, Small cortical breaks seen on US, corresponding to physiologic bone channels that enter the bone marrow through the palmar surface of the second MCP joint. Lesions are marked by arrows. See Figure 1 for definitions.

Effects of small US bone changes on the specificity for bone erosions.

Based on the substantial number of false-positive lesions, we assessed the impact of different cutoffs for scoring bone erosions by US. We analyzed the sensitivity and specificity for bone erosions using cutoffs of ≥1 (any lesion) and ≥3 (cortical lesion of ≥2 mm in size) (Table 1). When small lesions seen on US were classified as bone erosions the sensitivity for erosive lesions was good among all joint regions (83–100%) but specificity for erosions was sometimes only moderate, particularly at the palmar regions (26–55%). In contrast, rather high specificity was obtained at the dorsal regions of the MCP joints (77–88%). When more stringent criteria for erosions seen on US were used (cortical lesion of ≥2 mm), many more lesions were identified correctly and, in particular, the number of false-positive results at the palmar regions of the MCP joints declined substantially.

Table 1. Sensitivity and specificity of ultrasound in detecting bone erosions*
 False-positive, %False-negative, %True-positive, %True-negative, %PPVNPVSensitivitySpecificity
  • *

    Results were considered to be false-positive if they were positive by ultrasound (US) but negative by micro–computed tomography (micro-CT) and false-negative if they were negative by US but positive by micro-CT. PPV = positive predictive value; NPV = negative predictive value; MCP2, MCP3, and MCP4 = second, third, and fourth metacarpophalangeal joints.

Cutoff = any cortical lesion on US        
 MCP2 palmar53.83.823.119.20.300.830.860.26
 MCP2 dorsal11.57.738.542.30.770.850.830.79
 MCP2 radial15.43.865.415.40.810.800.940.50
 MCP3 palmar23.17.746.223.10.670.750.860.50
 MCP3 dorsal7.73.834.653.80.820.930.900.88
 MCP4 palmar34.60.023.142.30.401.001.000.55
 MCP4 dorsal19.20.015.465.40.441.001.000.77
Cutoff = cortical lesion ≥2 mm on US        
 MCP2 palmar23.17.719.250.00.450.870.710.68
 MCP2 dorsal7.711.534.646.20.820.800.750.86
 MCP2 radial11.511.557.719.20.830.630.830.63
 MCP3 palmar15.419.234.630.80.690.620.640.67
 MCP3 dorsal0.07.730.861.51.000.890.801.00
 MCP4 palmar19.27.715.457.70.440.880.670.75
 MCP4 dorsal11.53.811.573.10.500.950.750.86

DISCUSSION

This study shows that a large proportion of bone lesions detected by US, though not all, can be verified by micro-CT. In general, the sensitivity for detecting erosions was high, with the rate of false-negative results, i.e., erosions escaping detection by US, being <10%. The false-negative findings were not based on failure to detect lesions at joint regions that are inaccessible for US, since only sites accessible for US were studied. Importantly, a previous study by Dohn and colleagues comparing US with CT showed that the sensitivity of US for detection of bone erosions was high in joints that are well accessible for US (the second and fifth MCP joints) but low in joints that are less well accessible (the third and fourth MCP joints) (9). Our data suggest that if US investigation is confined to well-accessible joint regions, its sensitivity for detection of bone erosions is indeed high (usually >85%). One reason some lesions still escape detection may be related to the specific shape of bone erosions. In particular, Ω-shaped lesions, with only a very small cortical break but a larger erosion underneath, may be barely detectable with an imaging tool that is confined to scanning of the bone surface only. Such lesions are particularly found along the dorsal groove of the metacarpal heads. Overall, however, the impact of false-negative lesions is rather small, which is likely due to the improvement in technology of US with higher spatial resolution.

False-positive US findings, however, are far more common and are based on the detection of bone channels and forceps-like osteophytes, both of which mimic bone erosions. Particularly in the case of small lesions (<2 mm) or lesions localized at the palmar grooves of the metacarpal heads and phalangeal bases, there is a risk of overinterpretation of US data. In the palmar grooves, nutrient blood vessels physiologically pass the cortical bone barrier through vascular bone channels and enter the bone marrow. Such lesions are indeed cortical breaks, but are physiologic and do not result from inflammation. Also, osteophytes can mimic bone erosions, especially when these lesions form forceps-like structures well above the cortical bone surface.

In summary, we have shown that most bone erosions detected by US are indeed erosive lesions as indicated by the confirmatory micro-CT analysis. As the spatial resolution of US has improved, false-negative results are relatively rare. In contrast, however, some of the erosive lesions detected by US do not correspond to cortical bone breaks seen on micro-CT. Focusing on more pronounced bone erosions, such as those with a cortical irregularity of >2 mm in width, reduces the risk of overinterpretation of US findings.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Schett had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Finzel, Ohrndorf, Stach, Schett, Backhaus.

Acquisition of data. Finzel, Ohrndorf, Messerschmidt, Schett, Backhaus.

Analysis and interpretation of data. Finzel, Ohrndorf, Engelbrecht, Stach, Schett, Backhaus.

Ancillary