Article Text

Download PDFPDF

Original research
MRI of shoulder girdle in polymyalgia rheumatica: inflammatory findings and their diagnostic value
  1. Martin Fruth1,2,
  2. Lucie Künitz2,
  3. Philipp Martin-Seidel1,
  4. Styliani Tsiami2 and
  5. Xenofon Baraliakos2
  1. 1Evidia Radiologie am Rheumazentrum Ruhrgebiet, Herne, Germany
  2. 2Rheumazentrum Ruhrgebiet, Ruhr-Universitat Bochum, Herne, Germany
  1. Correspondence to Dr Xenofon Baraliakos; Xenofon.Baraliakos{at}


Background Non-synovial inflammation as detected by MRI is characteristic in polymyalgia rheumatica (PMR) with potentially high diagnostic value.

Objective The objective is to describe inflammatory MRI findings in the shoulder girdle of patients with PMR and discriminate from other causes of shoulder girdle pain.

Methods Retrospective study of 496 contrast-enhanced MRI scans of the shoulder girdle from 122 PMR patients and 374 non-PMR cases. Two radiologists blinded to clinical and demographic information evaluated inflammation at six non-synovial plus three synovial sites for the presence or absence of inflammation. The prevalence of synovial and non-synovial inflammation, both alone and together with clinical information, was tested for its ability to differentiate PMR from non-PMR.

Results A high prevalence of non-synovial inflammation was identified as striking imaging finding in PMR, in average 3.4±1.7, mean (M)±SD, out of the six predefined sites were inflamed compared with 1.1±1.4 (M±SD) in non-PMR group, p<0.001, with excellent discriminatory effect between PMR patients and non-PMR cases. The prevalence of synovitis also differed significantly between PMR patients and non-PMR cases, 2.5±0.8 (M±SD) vs 1.9±1.1 (M±SD) out of three predefined synovial sites, but with an inferior discriminatory effect. The detection of inflammation at three out of six predefined non-synovial sites differentiated PMR patients from controls with a sensitivity/specificity of 73.8%/85.8% and overall better performance than detection of synovitis alone (sensitivity/specificity of 86.1%/36.1%, respectively).

Conclusion Contrast-enhanced MRI of the shoulder girdle is a reliable imaging tool with significant diagnostic value in the assessment of patients suffering from PMR and differentiation to other conditions for shoulder girdle pain.

  • Magnetic Resonance Imaging
  • Inflammation
  • Polymyalgia Rheumatica

Data availability statement

Data are available on reasonable request.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

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.


  • Extra-articular non-synovial inflammation around tendons and capsules represents an imaging feature in polymyalgia rheumatica (PMR) as depicted by 18fluordesoxyglucose positron emission tomography/CT and MRI, mostly accompanied by synovitis in bursal, tenovaginal and articular compartments.


  • Detection of non-synovial inflammation in terms of peritendinitis and capsulitis in shoulder girdle MRI dominates the inflammatory pattern in PMR with superior discriminatory effect compared with detection of synovitis.


  • Contrast-enhanced MRI is a reliable imaging tool in clinically equivocal cases with high discriminatory potential against other causes of shoulder girdle pain.

  • Specific addressing of the non-synovial inflammation in further imaging-based studies could improve our understanding of the aetiology of PMR.


Polymyalgia rheumatica (PMR) is a common rheumatic disease and the most frequent inflammatory rheumatic condition of the elderly.1 In contrast to rheumatoid arthritis (RA), PMR can behave like an autoinflammatory disease with sudden onset, remarkable good response to glucocorticoid treatment and is without an autoimmune antibody profile.2 While imaging plays a key role in finding the diagnosis and in the follow-up of many rheumatic diseases like RA or axial spondyloarthritis (axSpA), the diagnostic decision-making in PMR is still based on clinical grounds including patient history with shoulder and/or pelvic girdle pain, laboratory findings such as elevation of C reactive protein (CRP) and erythrocyte sedimentation rate (ESR) supported by rapid response to glucocorticoids.3 4 As imaging concerns, only ultrasound (US) has made its way to the European League Against Rheumatism (EULAR)/American College of Rheumatology (ACR) provisional classification criteria5 based on the detection of subacromial bursitis and tenosynovitis of the long head of the biceps brachii muscle, but it still contributes only little to these classification criteria since inflammatory changes due to degeneration and/or stress are also relatively common in these locations and the typical age group.6

Up to date, direct imaging of the inflammatory musculoskeletal changes beyond US has not played a role in the assessment of patients with possible PMR. However, using 18fluordesoxyglucose positron emission tomography/CT (FDG-PET/CT) extracapsular inflammation of entheseal and bursal sites in pelvic and shoulder girdle sites, as well as lumbar and cervical interspinous inflammation, have been identified as characteristic for the myalgic pain symptoms reported by patients diagnosed by PMR.7–10 However, routine use of FDG-PET/CT is limited due to availability, costs and radiation dose. MRI studies have shown comparable performance to FDG-PET/CT, demonstrating peritendinous and pericapsular oedema and contrast enhancement11–13 as a correlate of inflammation at the same affected sites. Meanwhile, the broad availability of MRI makes it an appropriate modality for demonstrating the rather characteristic inflammation in PMR.13 14 In addition, due to its higher contrast and spatial resolution, MRI may allow for a more detailed look at the underlying pathoanatomy of inflammation in PMR.

Based on our experience with PMR and MRI diagnostics,14 15 we performed a retrospective case–control study to investigate whether and how these findings could contribute to its diagnosis in comparison to other causes of shoulder girdle pain.

Patients and methods

Study population and data collection

This is a retrospective study of patients who underwent a shoulder girdle MRI in our tertiary rheumatological university centre for evaluation of shoulder girdle pain from January 2017 to April 2022. Clinical, laboratory and demographic information from the time point of the MRI examination was available from all patients. MRIs were performed on indication from the treating rheumatologist for differential diagnostic considerations. Reference standard was the final diagnosis of an expert rheumatologist at the time of discharge from hospital without a waiting period. Cases were classified to PMR group and subgroups in non-PMR by an experienced rheumatologist. Patients with dominant peripheral arthritis of the hand accompanied by polymyalgic symptoms were given the diagnosis of PMR-like onset RA and classified into a discrete subgroup, the term PMR-like onset RA is used instead of elderly-onset RA (EORA) with polymyalgic onset. All patients were symptomatic regarding shoulder girdle pain at the time of the MRI scan and scans were obtained in less than 72 hours after registration. If patients had bilateral shoulder girdle pain, the clinically more affected side had been examined. If MRI scans of both shoulders were performed, only the scan of the left one was included assuming to be the non-dominant side. Non-contrast-enhanced MRI scans and scans degraded by artefacts were excluded from evaluation prior to analysis by an independent evaluator experienced in the interpretation of MRI examinations. The cohort included both patients with relapse of an established diagnosis of PMR and patients with a new diagnosis. If patients had repeated MRI scans of the shoulder during the study period, only the first scan was included, either during a relapse of an established disease or at the time of first diagnosis. All remaining qualified MRI scans were included in the image reading and statistical analysis. Flow chart of patient recruitment and case selection as well as clinical findings in PMR and non-PMR group are given in figure 1 and table 1, respectively.

Figure 1

Flow chart of patient recruitment and case selection.

Table 1

Demographic and clinical findings in PMR and non-PMR group

MRI evaluation

All investigated MRI scans were performed using a 1.5T MRI scanner (Siemens AERA) with a dedicated flexible surface coil covering the shoulder and the use of a clinical routine protocol, consisting of six sequences: oblique coronal T1w turbo spin echo (tse), oblique sagittal T2w tse, oblique coronal Pdw tse with spectral fat saturation, transversal Pdw tse with Dixon fat saturation, oblique coronal and transversal contrast-enhanced T1w tse sequences with Dixon fat saturation, adjusted to the individual anatomy. In all patients, weight-adapted gadoteric acid had been applied intravenously.

Investigated sites and image analyses

For imaging evaluation, six extracapsular non-synovial sites with optimal imaging reproducibility in two planes and three synovial sites were defined and considered based on data previously published13 16 and own clinical experience with imaging in PMR.

Non-synovial sites:

  • Around axillary recess of fibrous humeroglenoidal capsule (CAP).

  • Around tendinous origin of long head triceps brachii (LHT),

  • Around coracoclavicular ligament (CC).

  • Around intramuscular tendon of subscapularis muscle (SSC).

  • Around intramuscular tendon of infraspinatus muscle (IS).

  • Around tendinous origins of short head biceps brachii and coracobrachialis muscle (BBSH) at the tip of coracoid process.

Synovial sites:

  • Subacromial-subdeltoid bursa (SAB).

  • Humeroglenoidal joint (HGJ).

  • Tendon sheath of long head biceps brachii (LBTS).

Inflammatory lesions were defined for synovial compartments as hyperenhancing synovium with at least 1.5 mm thickness, visible in at least two consecutive images and corresponding second plane. Non-synovial inflammatory lesions were defined as contrast enhancement around ligament, around tendons including the intramuscular segments and around fibrous joint capsule visible in at least two consecutive images or two perpendicular planes. Exemplary images of inflammation of the six non-synovial sites are given in figure 2. Two experienced radiologists blinded to clinical and demographic information evaluated the presence or absence of inflammatory lesions at the predefined sites in a consensus reading, all included scans were read by each radiologist.

Figure 2

Exemplary images of inflammation at the six investigated non-synovial sites marked by white arrows. (A) Axillary recess of fibrous humeroglenoidal joint capsule. (B) Tendon origin of long head triceps brachii muscle. (C) Coracoclavicular ligament. (D) Intramuscular tendon of subscapularis muscle. (E) Intramuscular tendon of infraspinatus muscle. (F) Tendon origins of short head biceps brachii and coracobrachialis muscles.

Statistical analyses

Clinical and demographic data for the PMR and non-PMR group were analysed in descriptive manner. The prevalence of inflammation at the non-synovial and synovial sites was evaluated for both groups, PMR and non-PMR cases, including the diagnosis subgroups of the latter. Different combinations of the examined imaging features alone and together with clinical information were tested regarding their diagnostic ability to identify PMR and described as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under the receiver operating characteristic curve (AUC). Continuous variables are reported as arithmetic mean (M)±SD, compared by U-test. Categorical variables are reported as percentage, compared by χ2 test. A value of p<0.05 was considered statistically significant. Statistical analyses were performed by using SPSS V.23 (IBM).


A total of 708 patients who underwent an MRI scan of the shoulder girdle for evaluation of shoulder girdle pain were identified. 212 patients were excluded from evaluation due to non-contrast-enhanced MRI and overlaying artefacts mostly motion related, resulting in 496 patients qualified for final evaluation. Out of those, 122 patients had received the diagnosis of PMR based on all available clinical, laboratory and imaging information including 4 patients with concomitant giant cell arteritis (GCA), 104 were new diagnosed cases, 18 had a relapse of an already established diagnosis. The non-PMR group consisted of 86 patients with RA or a connective tissue disease (RA/CTD), 35 patients with PMR-like onset RA, 28 patients with psoriatic arthritis or peripheral involvement of spondyloarthritis (PsA/SpA), 11 with crystal-induced arthritis including 7 with calcium pyrophosphate deposition (CPPD)-induced arthritis and 4 with gout disease, 163 patients with degenerative joint disease (DJD) and 51 with fibromyalgia (FMS). Clinical and demographic characteristic as well as subgroups of diagnosis in non-PMR group are given in table 1.

Overall, 91.8% of PMR patients and 70.3% in the non-PMR group had bilateral shoulder pain. The non-PMR group consisted of significantly more females (66.6% vs 50%) and mean age was significantly younger in non-PMR group (64.9±10.5 vs 59.5±10.4 years, p<0.001). Inflammatory markers, CRP and ESR, were significantly higher in PMR group, 3.4±3.7 mg/dL vs 1.4±2.9 mg/dL p<0.001 and 36±25 mm vs 18±18 mm p<0.001, respectively. Whereas rheumatoid factor (RF) positivity and cyclic citrullinated protein antibody (CCP) positivity were significantly higher in the non-PMR group, 13.9% vs 23.8% p=0.013 and 6.6% vs 16.0% p=0.004, respectively. Only serum urate level did not differ significantly between the groups.

Prevalence and discriminatory effect of synovial inflammation

Synovitis in HGJ, SAB and LBTS had been detected frequently in PMR and non-PMR, but the average number of inflamed synovial sites was significantly higher in the PMR group, 2.5±0.8 vs 1.9±1.1, p<0.001 (figure 3). In PMR, most frequently SAB 90.2% and LBTS 86.9% showed inflammation while synovitis in all three synovial sites was detected in 68.9%. In the non-PMR group, synovitis in SAB and LBTS were found in 74.1% and 66.6%, respectively, and inflammation of all three synovial sites in 43.0%. The prevalence of synovial inflammation at all individual synovial sites was significantly higher in PMR group compared with non-PMR, p<0.001 at each synovial site. No significant differences in the detection of synovitis were found in the subgroups PMR-like onset RA and crystal induced arthritis when compared with the PMR group. The detection of synovitis in at least two of the three synovial sites differentiated PMR from non-PMR group best with a sensitivity and specificity of 86.1% and 36.1% respectively, and an AUC of 0.649, improving to 72.1% and 73.5% and AUC of 0.755 if additional bilateral shoulder pain and elevated inflammatory markers were clinically present (figure 4).

Figure 3

Prevalence of synovial and non-synovial inflammation in PMR and non-PMR including subgroups as heat map and table. Circles for synovial sites and quadrates for non-synovial sites superimposed at their approximate anatomic location on X-ray. Continuous variables were expressed as the arithmetic mean and standard deviation (SD), compared with Mann-Whitney U test. Categorical variables as percentage, compared by χ2 test. A value of p<0.05 was considered statistically significant. BBSH, origins of short head biceps brachii and coracobrachialis muscles; CAP, axillary recess of fibrous humeroglenoidal capsule; CC, coracoclavicular ligament; CTD, connective tissue disease; DJD, degenerative joint disease; FMS, fibromyalgia syndrome; HGJ, humeroglenoidal joint; IS, intramuscular tendon of infraspinatus muscle; LBTS, tendon sheath of long head biceps; LHT, long head triceps brachii; PsA, psoriatic arthritis; PMR, polymyalgia rheumatica; RA, rheumatoid arthritis; SAB, subacromial-subdeltoid bursa; SSC, intramuscular tendon of subscapular muscle; SpA, peripheral involvement in spondyloarthritis.

Figure 4

ROC curves for discriminatory effect of different imaging features and their combinations with clinical parameters: (1) prevalence of non-synovial inflammation alone, (2) prevalence of non-synovial inflammation in presence of synovitis in subacromial bursa (SAB) or long head biceps tendon sheath (LBTS), (3) prevalence of synovitis alone and (4) prevalence of synovitis in presence of bilateral shoulder pain and elevation of inflammatory markers. Performance is given as area under the curve. ROC, receiver operating characteristic curve.

Prevalence and discriminatory effect of non-synovial inflammation

Inflamed non-synovial sites were detected significantly more frequently in the PMR group 3.4±1.7 compared with non-PMR group 1.1±1.4, p<0.001 (figure 3). Most frequently peritendinitis around origin of LHT in 68.9% of PMR patients was found, followed by inflammation of CC 67.2%, SSC 65.6%, CAP 54.9%, IS 42.6% and BBSH 38.5%. In the non-PMR group, inflammation of CC was the most frequent non-synovial finding in 29.7% of patients, followed by CAP at 19.5%, SSC 17.1%, LHT 14.2%, IS 14.2% and BBSH 11.5%. The prevalence of inflammation was significantly higher at all individual non-synovial sites in PMR compared non-PMR group, p<0.001 for each site. In non-PMR subgroups the highest average number of inflamed non-synovial sites was found in the PMR-like onset RA subgroup with 2.6±2.0 inflamed sites, differing only slightly significant compared with the PMR group, p value of 0.042, while no significant difference in the prevalence of inflammation at individual non-synovial sites except for LHT was found in PMR-like onset RA subgroup compared with PMR group. In all other diagnosis subgroups, average number of inflamed non-synovial sites differed significantly compared with PMR, p<0.001 to 0.007. The prevalence of non-synovial inflammation alone had an excellent discriminatory power to differentiate PMR from other causes of shoulder girdle pain, AUC 0.840. If inflammation of at least 3/6 non-synovial sites was present, PMR was differentiated from non-PMR group with a sensitivity of 73.8% and specificity of 85.8%, PPV 0.629, NPV 0.909 (figure 4). A lower diagnostic threshold of at least 2/6 inflamed non-synovial sites still performed excellently with a sensitivity increased to 84.4% at the expense of a lowered specificity of 73.8%, PPV 0.512, NPV 0.936, AUC 0.840. The prevalence of non-synovial inflammation combined with detection of concomitant synovitis of LBTS or subacromial bursa discriminated slightly poorer but still excellent, AUC 0.837, changing sensitivity/specificity to 83.6%/74.6% if at least 2/6 inflamed non-synovial sites together with synovitis of LBTS or subacromial bursa were present.

The potential of the aforementioned tests to differentiate between PMR and PMR-like onset RA was restricted. The prevalence of non-synovial inflammation alone discriminated with a sensitivity of 73.8% but limited specificity of 51.4%, AUC 0.611, if at least 3/6 sites were inflamed. The presence of synovial inflammation alone or in combination with bilateral shoulder pain and elevated inflammatory markers was uninformative, AUC 0.495 and 0.538, respectively.

Within the PMR group, there was a significant difference in average number of inflamed non-synovial sites in newly diagnosed patients versus cases with relapse of an established PMR, 3.6±1.6 vs 2.4±1.9, p=0.006. But no significant difference in average number of inflamed synovial sites, 2.6±0.8 vs 2.2±0.9, p=0.098. The presence of non-synovial inflammation alone performed best to discriminate cases with a newly established diagnosis of PMR from non-PMR cases, AUC 0.864, with a sensitivity of 89.4% and specificity of 73.8% if at least two non-synovial sites were rated as inflamed. In cases of relapse of PMR, performance of this test lowered, AUC 0.705. For the best result, at least three non-synovial sites had to be rated as inflamed, sensitivity 55.6% and specificity 85.8%.


In this retrospective study of 496 consecutive patients with shoulder girdle pain, we describe inflammatory findings in PMR as shown by contrast-enhanced MRI in detail and compare them to findings in patients with other reasons for shoulder girdle pain. We found that non-synovial inflammation detected as contrast enhancement around tendons, ligaments and fibrous capsules is the striking musculoskeletal finding in patients suffering from PMR. In addition, we show that the detection of non-synovial inflammation has a robust discriminatory potential against other inflammatory and non-inflammatory causes of shoulder girdle pain and outperforms the discriminatory ability of synovitis detection alone. The detection of inflammation of at least three of six predefined non-synovial sites differentiated PMR from other causes of shoulder girdle pain with a sensitivity of 73.8% and specificity of 85.8%. This characteristic pattern of non-synovial inflammation can improve diagnostic certainty in clinically inconclusive cases or be used to objectively confirm the clinical aspects of PMR.

A lower threshold of two out of six inflamed non-synovial sites as a diagnostic test for PMR improved sensitivity to 84.4% but decreased specificity to 73.8%. When additional synovitis of subacromial bursa or of the long head biceps tendon sheath was present, specificity improved only slightly to 74.6% while sensitivity decreased to 83.6%, owed to the fact that one clinically diagnosed PMR case and three non-PMR cases exhibited non-synovial inflammation at least two sites but no concomitant bursal or tenosynovial inflammation. This fact underpins the assumption that synovial inflammation in shoulder girdle of PMR cases is not a conditio sine qua non but a very frequent finding since both tests, the prevalence of non-synovial inflammation alone or with concomitant presence of synovial inflammation in SAB or LBTS, performed very similar (figure 4).

In our study, we examined only one shoulder in all cases, though it is worth to be discussed whether an MRI scan of a single unilateral region is sufficient to diagnose a systemic disease which usually affects multiple musculoskeletal regions. However, our results show that an MRI-based definition of the underlying inflammatory pattern in only one clinically affected test region might be suitable to foster the diagnosis, especially in synopsis with clinical and laboratory data. Probably it would improve if additional regions like the contralateral shoulder or pelvic girdle were included, combining distributional and local pathomorphological pattern of inflammation.

For diagnostic and classificatory purpose, it might be helpful to include these non-synovial imaging findings as an additional criterion. An appropriate approach was made by Nakamura et al,16 who investigated 137 patients with a fraction of 58 patients suffering from PMR with shoulder girdle MRI and US in each case. They found enhancement of joint capsule and rotator cuff tendons in addition to joint effusion and focal bone marrow oedema (BME) the most significant differentiating findings in PMR compared with findings in non-PMR group. This MRI-based finding differentiated PMR from non-PMR group with a sensitivity/specificity of 76%/85% if two imaging findings from rotator cuff tendon enhancement, capsule enhancement and focal BME were positive. Whereas biceps tenosynovitis and subacromial-subdeltoid bursitis as detected by US performed with a shortened sensitivity/specificity of 50%/72%. By replacing the US criteria with these MRI criteria in the EULAR/ACR provisional classification they improved sensitivity from 86% to 93% while specificity remained constant.

Our methodical approach did not allow a comparison between the influence of non-synovial versus synovial inflammation on the EULAR/ACR provisional classification criteria since they were not obtained in our retrospective study. Therefore, a comparison of our findings with studies addressing the discriminatory influence of synovitis in SAB and LBTS as detected by ultrasound is not possible. But within the cohort studied, we could show a superior diagnostic performance for the detection of non-synovial inflammation over synovial inflammation, even if additional clinical features bilateral shoulder girdle pain and elevated inflammatory markers, which are commonly used in diagnostic and classification criteria, were included (figure 4).

In our study group, the prevalence of non-synovial inflammation was significantly lower in cases with a relapse compared with new established diagnoses of PMR, this led to a markedly decreased sensitivity of only 55.6% but high specificity of 85.8% if at least three out six non-synovial sites were rated as inflamed. Since we did not obtain information about treatment and symptom duration at the time of the MRI, we cannot conclude whether residual treatment or patients awareness of disease activity influenced the inflammatory pattern. Therefore, further longitudinal studies are needed to investigate the correlation between disease activity and MRI-based inflammatory burden to qualify MRI as an objective monitoring tool for disease activity.

Interestingly, PMR and subgroup PMR-like onset RA could not be differentiated by imaging since there were only little differences in their inflammatory pattern with slight statistical significance. This finding again highlights the question of whether PMR-like onset RA is a special type of RA or rather a course of PMR with peripheral involvement. The fact that PMR-like onset RA shares the imaging feature of dominant non-synovial inflammation with PMR, argues for the latter. Further imaging-based studies could help to reclassify this enigmatic inflammatory condition.

Similar to our prior findings in MRI scans of the pelvic girdle in PMR17 and findings from Owen et al,18 we could show that the target of inflammation in PMR is the outer covering of a tendon, the peritendineum. This peritendinous inflammation can propagate into the intramuscular tendon or even up to the perimysium of greater muscle bundles. This is an important feature when analysing the rotator cuff tendons because here the extramuscular tendon is lined by synovial layers on the bursal and articular side. Hence, inflammation of the peritendineum cannot be differentiated from accompanying bursal or articular synovitis until the inflammation of the peritendineum reaches the intramuscular tendon. Furthermore, this proximity of the peritendineum and synovial layers is also a plausible explanation for the high incidence of accompanying bursal, tenovaginal and articular synovitis, making it the most proximate secondary site for the spread of the inflammation rather than the primary target of inflammation in PMR. We also detected very frequently periligamentous inflammation of the acromioclavicular ligament, which is not surprising, since ligaments and tendons share a similar histological composition.

More obvious is the use of our findings for further imaging-based studies implying the informative benefit of non-synovial inflammatory findings over synovial inflammation in PMR. This is particularly of interest since imaging-based studies are needed to investigate the nexus between GCA and PMR. The recently enforced concept of GCA and PMR as a spectrum disease (GPSD)19 20 compels to illuminate not only the different phenotypes of PMR and GCA and their frequent overlaps but as well the borderlines of the disease spectrum. Here imaging, both MRI and FDG-PET/CT, will help as objective tools. Up to date, these findings mainly have been addressed by several studies using FDG-PET/CT. The high diagnostic value of this method in patients with suspected PMR has been recapitulated by van der Geest et al in a meta-analysis covering nine studies.21 Its substantial power for investigation of the GCA-PMR-spectrum is the excellent ability for detection of large vessel vasculitis (LVV) and PMR simultaneously in a panoramic view. In this topic, MRI-based studies lag behind although its usefulness is obvious regarding prompt availability, cost and radiation exposure. In a recent MRI-based study, Seitz et al22 found a high frequence of PMR-typical musculoskeletal inflammatory findings throughout the body trunk in a cohort of 90 patients with GCA using the scoring system introduced by Mackie et al.12 All patients with PMR symptoms and 93% of all patients showed at least one pathological musculoskeletal enhancement congruent with an inflammatory finding in PMR, supporting the concept of GPSD. Prospective longitudinal studies are needed to illuminate the different clinical phenotypes of this spectrum disease and could provide imaging-based markers to predict the course of disease. A highly informative contribution can also be expected from MRI-based studies addressing other diseases with polymyalgic clinical symptoms, which must be contextualised in the new concept of GPSD, especially PMR-like onset RA. Here, the superior capability of hand MRI in the detection and characterisation of synovial and non-synovial inflammation could bring out differences in inflammatory pattern compared with classic RA.

In this study of real-world data, we present a relatively large cohort of patients with shoulder girdle pain, but the significance of our findings is limited by the retrospective nature and concept of the study. Shoulder girdle pain at the time of examination was the only clinical criterion for disease activity, information regarding symptom duration and already initiated therapy were not obtained, therefore, the influence of these factors cannot be addressed. Our reference standard was expert rheumatological diagnosis only, which reflects the real-world scenario of our data but limits its accuracy. As well classification of the non-PMR cases to six subgroups represents a rough simplification of the often-complex clinical scenario. A preselection of the PMR cases with more cases referred to MRI with discrepancy between presumptive diagnosis and clinical appearance is also possible, in addition patients with GCA and concomitant PMR are underrepresented in our study population since those run through a different diagnostic algorithm in our institution focusing on vasculitis detection by FDG-PET/CT or MRI with detection of PMR-related inflammation en passant. Furthermore, non-synovial inflammatory findings are reported in detail in our institution and this could have influenced clinical decision-making, leading to a circular reasoning our study might be impaired. Therefore, these findings cannot be generalised and must be confirmed in prospective studies. Besides, our routine imaging protocol was suboptimal for the detection of peritendinous inflammation of intramuscular segments of rotator cuff muscles. Addition of a contrast-enhanced oblique sagittal T1w sequence could eliminate partial volume effects and improve detection of peritendinous inflammation.

In conclusion, PMR exhibits a rather characteristic inflammatory pattern in the shoulder girdle consisting of peritendinous, periligamentous and capsular inflammation. Bursal, tenovaginal and articular synovitis are frequent concomitant findings but with less discriminatory potential from other causes of shoulder girdle pain. MRI of the shoulder girdle is a reliable imaging tool with significant diagnostic value in assessment of patients suffering from PMR.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

The ethical committee body responsible for our hospital is the Ethical Committee (EC) of the Ruhr-University Bochum, Germany. Approval of the EC had been obtained, approval number 22-7568.



  • Contributors MF and XB designed and concepted the study; LK, ST and PM-S collected the data; MF and PM-S read the images; MF and XB analysed and interpreted the data and drafted the article; MF, LK, PM-S, ST and XB critically revised the article and approved the final version for publication; manuscript guarantor: MF.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

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