Do musculoskeletal ultrasound and magnetic resonance imaging identify synovitis and tenosynovitis at the same joints and tendons? A comparative study in early inflammatory arthritis and clinically suspect arthralgia

Patients that newly presented with early IA or CSA between May and October 2017 at the Leiden rheumatology outpatient clinic were studied. They were consecutively included in either the Early Arthritis Clinic (EAC) cohort or the CSA cohort. Requirements for inclusion in both cohorts are described in reference and supplementary [8, 17].

Both cohort studies were approved by the local Medical Ethical Committee. All patients provided informed consent.

All patients underwent unilateral contrast-enhanced MRI of metacarpophalangeal (MCP), wrist, and metatarsophalangeal (MTP) joints and musculoskeletal US at the same day < 2 weeks after first presentation. According to the protocol, imaging was done before disease-modifying anti-rheumatic drug (DMARD) initiation (including glucocorticoids) in patients with IA. DMARDs were not prescribed to patients with CSA. All patients were asked to stop NSAIDs 24 h before imaging. More details are provided supplementary.

All patients were scanned on the same scanner (an MSK Extreme 1.5 T extremity MR system (GE Healthcare, Wisconsin, USA)). Unilateral MRI scans of wrist, MCP (2–5) and MTP (1–5) joints were made of the most affected side, or the dominant side in case of equally severe symptoms. Sequences acquired were coronal pre-contrast T1-weighted fast spin-echo (FSE) and coronal and axial post-contrast T1-weighted FSE with frequency-selective fat suppression of MCP and wrist, and post-contrast coronal and axial sequences of the MTP joints. More details are provided in reference and supplementary [17].

Each MRI-scan was scored according to RA MRI scoring (RAMRIS) method by two experienced readers (inter-reader intraclass correlation coefficients (ICC) > 0.94) [18, 19]. MRI scores for joints (synovitis) and tendons (tenosynovitis) ranged from 0 to 3. Mean scores of two readers were calculated and lesions were considered absent in case it was scored by only one reader.

A high-end US machine was used (GE Logiq E9, Genova, Italy) with a linear array transducer of 6–15 MHz. US examinations were performed bilaterally in GS and PD mode according to a standardised protocol. The same locations that were scanned by MRI were studied here. PD was assessed with a pulse repetition frequency of 0.8 kHz, and gain was set to a level until background signal was removed.

The presence of synovitis was assessed on a semi-quantitative scale (0–3) for GS/PD according to Szkudlarek et al. [20], and synovial effusion and hypertrophy were combined (called ‘modified Szkudlarek method’) [21]. Tenosynovitis was examined on a semi-quantitative scale (0–3) for GS/PD according to OMERACT [22]. A detailed US-scoring protocol is provided supplementary. All US scores per joint/tendon ranged from 0 to 3.

During the study, the newly developed EULAR-OMERACT-scoring method for synovitis was published [16]. To explore if the results changed when this definition was used, the static images of US were re-scored for GS synovitis by two examiners (ICC 0.92) and mean scores were calculated. The different scoring methods are described in Additional file 1: Table S1 and in the supplementary methods.

Imaging results were not communicated to clinicians at any time point.

We compared semi-quantitative scores of US-detected synovitis and tenosynovitis to MRI-detected synovitis and tenosynovitis scores (each on a scale from 0 to 3), respectively, for each location using spearman’s correlation coefficients. For the primary analyses, we used the method according to the EULAR-OMERACT for GS synovitis. After analysing (semi-)quantitative data, US and MRI scores were dichotomized. For US, different cut-offs were studied: ≥ 1 and ≥ 2 for GS synovitis, ≥ 1 for PD synovitis and ≥ 1 for GS/PD tenosynovitis. Additionally, GS synovitis and tenosynovitis scores ≥ 2 or PD ≥ 1 were combined. MRI scores were dichotomized with ≥ 1 as cut-off and also on a cut-off based on findings from symptom-free volunteers, which has been published previously [23]. Then, an MRI was considered positive if synovitis or tenosynovitis was seen in < 5% of age-matched healthy controls. We calculated test characteristics for US with MRI as reference. Analyses were done on individual joint/tendon level and firstly presented on joint-group level (wrist, MCP, MTP joints) for reasons of clarity. Sub-analyses included stratification for patients presenting with IA and CSA and presentation of data at individual joint/tendon level. Finally, for GS synovitis, the ‘modified Szkudlarek method’ was compared to the EULAR-OMERACT method and also compared to MRI [16, 18, 21]. IBM SPSS (New York, USA) v23 was used.

Seventy patients newly presenting to the rheumatology outpatient clinic (40 with recent-onset CSA, 30 with early IA) were included. Table 1 presents their baseline characteristics. The majority was female; mean age was 45 for patients with CSA and 57 for patients with IA (Additional file 1: Table S2). In total, 840 joints and 700 tendons were examined.

Figure 1(a–c) presents the scores for GS-detected synovitis (EULAR-OMERACT method) versus MRI-detected synovitis (OMERACT-RAMRIS method). Analyses were performed on individual joints and tendons (i.e. MCP-2 of US versus MRI) and presented per joint group (MCPs, wrist, MTPs). All scores within joint groups were significantly correlated (Additional file 1: Table S3). In MTP joints, MRI scores of 0 infrequently coincided with scores of 1 for US (Fig. 1c); this is in contrast to findings on MCP and wrist level (Fig. 1a, b). In line with this observation, the corresponding test characteristics showed a high specificity (> 90%) for GS synovitis of wrist and MCP joints and a somewhat lower specificity of 80% for MTP joints. The sensitivity was poor for MCP and wrist (29–39%) and higher (75%) for MTP joints with MRI as reference (Table 2)

Subsequently, PD synovitis scores were compared to MRI. Also here, increased US scores were accompanied by increased MRI scores, and correlations were statistically significant (Additional file 1: Table S3). As presented (Fig. 1d–f), PD scores were only rarely ≥ 1 when MRI-detected synovitis scores were 0. Furthermore, we observed regularly that PD scores were 0 for joints that were scored ≥ 1 by MRI. These observations were reflected by the test characteristics, which showed a high specificity for PD (97–99%) for all locations (MTP, MCP, wrist) with only a low to moderate sensitivity (30–54%, Table 2).

Test characteristics when US positivity was defined by a combination of GS scores ≥ 2 or PD ≥ 1 are provided in Table 2. The combined scores showed a high specificity (> 92%) accompanied by an increased sensitivity for the MCP and wrist joints (30–54%) but not for the MTP joints (68%) in comparison to GS/PD alone.

Figure 2(a–c) presents the data of GS-detected tenosynovitis versus MRI-detected tenosynovitis scores. MRI scores were significantly correlated to GS scores (Additional file 1: Table S3). However, scores ≥ 1 for MRI were also often accompanied by US scores of 0. Test characteristics were in line with these observations, with a specificity of 80% for the extensor wrist tendons and > 89% for the other tendons (flexor wrist, flexor MCPs), and a moderate sensitivity (50–78%, Table 2)

Figure 2(d–f) shows the data of PD tenosynovitis versus MRI. PD signals were infrequently increased. MRI detected 113 tendons with tenosynovitis (out of 700), while for PD this was only 45. The corresponding test characteristics in Table 2 showed a high specificity (98–100%) with a low to moderate sensitivity (19–58%)

Defining tenosynovitis as a combination of GS ≥ 2 or PD ≥ 1 slightly improved the test characteristics for the extensor and flexor tendons of the wrist, but not for the flexor tendons of the MCPs, compared to the separate ultrasound features (GS/PD) (Table 2).

To investigate whether the excess of increased MRI-detected scores compared to US scores could be explained by the definition of positivity for MRI, we also applied a cut-off based on findings from symptom-free volunteers [23]. This resulted in a slightly increased sensitivity and AUC for GS-detected (teno)synovitis, while the specificity remained high compared to the main analyses. For PD, it only caused small differences (Additional file 1: Table S4).

In Additional file 1: Figures S1–S4, we provided the data of the US synovitis and tenosynovitis scores (GS and PD) versus MRI for patients with CSA and IA separately. As expected, synovitis and tenosynovitis were less frequently present and/or less severe in patients with CSA than IA. However, the pattern of concordance between MRI and US was similar. We also calculated test characteristics for patients with CSA and IA separately (Additional file 1: Table S5). The sensitivity for US with MRI as reference was lower in CSA than in patients with IA. The specificity was similar in both populations.

For clarity, the main results were presented on joint-group level, although analyses were performed on joint/tendon level. However, as findings on different joints/tendons might be different and these differences cannot be seen by presentation the joint-group level, we also provided test characteristics for each joint/tendon separately (Additional file 1: Tables S6, S7). In general, results were similar with a low to moderate sensitivity and high specificity. Remarkably, for the flexor tendons of the wrist (GS), all flexors had high specificity (88–100%). However, the sensitivity varied broadly: 71% for the FPL, 55% for the FCR and only 17% for the FDS/FDP. Also for PD, the sensitivity for tenosynovitis was generally low (17–64%).

Examples of MRI-detected (teno)synovitis versus GS/PD are illustrated by Fig. 3.

Due to recent advances in scoring methods for GS, two methods were applied and test characteristics were also determined for GS by the modified Szkudlarek method [21] with MRI as reference (Table 3). The modified Szkudlarek method had a higher sensitivity of 68–91% and lower specificity of 52–71% than the EULAR-OMERACT method (sensitivity 29–75%, specificity 80–98%) compared to MRI. Thereafter, we compared the scores of the two scoring methods for GS for each joint. The modified Szkudlarek method generally had higher scores than the EULAR-OMERACT method (see Additional file 1: Table S8–S10).