High prevalence of symptomatic enthesopathy of the shoulder in ankylosing spondylitis: Deltoid origin involvement constitutes a hallmark of disease




To examine the prevalence and characteristics of shoulder involvement in ankylosing spondylitis (AS). To analyze the sensitivity and specificity of shoulder lesions defined by magnetic resonance imaging (MRI) in patients with AS.


Prevalence of shoulder involvement was ascertained by chart review of 400 AS patients. One hundred of these patients and 285 controls were selected for clinical evaluation. AS patients with a clinically defined shoulder disorder (n = 15) and a control group of 91 patients (94 shoulders) with nonspecific shoulder pain were studied with MRI. Fifty-four MRI-defined variables per shoulder were analyzed by 2 observers. A third cohort of patients with AS (n = 76) was prospectively evaluated by clinical exam for AS-specific shoulder lesions identified on MRI.


Shoulder pain was recorded in 3.5% of patients by chart review. Shoulder involvement by clinical evaluation was noted in 24.7% of patients versus 14.2% of controls (odds ratio [OR] 8.17, 95% confidence interval [95% CI] 3.14–21.28, P < 0.001). Rotator cuff tendinitis was significantly more prevalent in patients (15.1%) than controls (3.5%; OR 8.17, 95% CI 2.66–25.14, P < 0.001). Acromioclavicular joint arthrosis was the most common lesion observed in AS shoulders (94%), although specificity was low (32%). Bone marrow edema at any entheseal site was noted in significantly more AS shoulders (70.6%) than in control (19.1%) shoulders (P < 0.001, corrected P = 0.02). Erosion of the greater tuberosity with or without adjacent bone edema had the best combination of sensitivity (58–65%) and specificity (86–92%). Intense acromial entheseal edema at the deltoid origin was observed only in AS shoulders (41.2%; P < 0.001). Evaluation of a prospective cohort of patients with AS showed that 22.4% had rotator cuff enthesopathy.


Shoulder lesions in AS are common and characterized clinically by rotator cuff tendinitis and on MRI by intense bone edema localized to the supraspinatus/greater tuberosity and deltoid/acromial entheses. Intense acromial bone edema at the deltoid origin is a hitherto undescribed and highly specific feature of AS. Enthesopathy of the rotator cuff is underrecognized in AS and should be incorporated into instruments measuring enthesitis.


Among inflammatory rheumatic diseases, the spondylarthropathies form a prominent subgroup (1). The characteristic clinical features of ankylosing spondylitis (AS) have been well documented over the years with efforts to clinically classify the condition among the other spondylarthropathies (2). Shoulder involvement as part of the spectrum of joint disease in AS has been described (3, 4), as has shoulder pain as a first clinical presentation in AS patients in whom a diagnosis had not been made (5, 6). Reported prevalence based on clinical assessment has varied from 7% to 33%, although these studies were neither controlled nor conducted using standardized and validated clinical assessment instruments (3, 4, 7, 8). A prevalence of 11.7% for shoulder pain has been reported in 1 community-based survey (9).

The cause of shoulder pain in AS patients has, on clinical grounds, been presumed to be due to synovitis, bursitis, or structural joint damage; however, no studies have systematically examined the etiology of shoulder pain in a controlled format. Several cross-sectional plain radiographic surveys have described joint space narrowing, erosion, and bony proliferation in the acromioclavicular joint, the glenohumeral joint, and around the rotator cuff insertion (3, 4, 10). Limitations of these studies include the lack of controls and systematic evaluation together with the limited sensitivity of plain radiographic assessment. In one study, radiographs were reported as normal in 23% of patients with restricted shoulder movement (4).

Since the term enthesopathy was first used, there has been increasing evidence in support of enthesitis being a prominent feature of AS (11, 12). Enthesitis has been demonstrated in various peripheral locations including the Achilles tendon, pes anserinus, greater trochanter, ischial tuberosities, and the knee (13–16), but it has not yet been formally evaluated in the shoulder. Limited success has been achieved in demonstrating enthesitis in patients with seronegative arthropathy using high-definition ultrasound and scintigraphy. One study reported abnormal echogenicity in 46% of patients with symptomatic plantar fasciitis and spondylarthropathy (17) and bone scintigraphy identified 94% of lesions in a group of patients with heel enthesitis secondary to Reiter's syndrome; specificity was not examined (18).

These techniques, however, do not reveal abnormalities in the bone marrow adjacent to an enthesis. The advent of fat suppression sequences in magnetic resonance imaging (MRI) has allowed study of bone marrow changes in AS at various sites, with most attention being paid to the spine and the lower limbs (12, 19–22). McGonagle et al have demonstrated the value of fat suppression sequences in imaging of the knee in patients presenting to an early arthritis clinic (12). They were able to locate separate regions of entheseal and nonentheseal bone marrow edema (BME) in addition to synovitis. The presence of entheseal edema within the knee allowed differentiation of those patients who ultimately developed AS versus rheumatoid arthritis.

Despite the common occurrence of shoulder pain in AS, its etiologic basis remains to be defined. The humeral tuberosities, acromial insertion of the rotator cuff, and clavicular insertion of the deltoid muscle represent entheses that could be inflamed, but also include sites where pathology is commonly observed in the general population. The purpose of this study was therefore to examine the prevalence and characteristics of clinically defined shoulder disease in patients with AS compared with age- and sex-matched controls; furthermore, we aim to determine the sensitivity and specificity of MRI-defined lesions in the shoulders of AS patients compared with otherwise normal individuals with nonspecific shoulder pain presenting for MRI examination.


Patient cohort A.

The prevalence of shoulder disease in AS was first estimated by retrospective chart review of 400 AS patients (69.5% males, mean age 43.8 years [range 29–68 years], mean disease duration 18.4 years [range 2–29 years]) with an established diagnosis according to the modified New York criteria (23).

Patient cohort B.

One hundred patients from cohort A were randomly selected for systematic clinical evaluation. Those patients with clinical involvement of the shoulders, defined as current shoulder pain of ≥1 month's duration and pain within the shoulder-specific area of a manikin, were further evaluated by MRI.

Patient cohort C.

We examined a third cohort of AS patients that are being followed prospectively at the University of Alberta. We examined them for the same AS-specific shoulder lesions that were identified in cohort B patients through MRI. Systematic evaluation of this prospective cohort includes clinical evaluation for shoulder disorders and tenderness at peripheral entheses (supraspinatus insertion, achilles insertion, plantar fascia insertion, trochanter, patellar ligament insertions into the patella and tibial tuberosity).

Control cohort A.

The control group included consecutive patients (age >18 years) attending a primary care practice in the city of Edmonton for unrelated complaints.

Control cohort B.

To obtain controls for the MRI evaluation, a computer-generated list of shoulder MRI examinations performed over the previous 18 months from 4 local hospitals and clinics was created. To avoid selection bias due to the center location or seasonal variations, 101 examinations were randomly selected from the generated list to form a control group. Demographic and clinical data supplied at the time of the MRI examination were recorded.

The study protocol was approved by the research ethics board of the University of Alberta.

Clinical evaluation.

We used a validated structured examination schedule developed for the diagnosis and classification of musculoskeletal disorders of the upper limb (24) (Table 1). However, this schedule does not provide diagnostic criteria for acromioclavicular and glenohumeral joint disease. Pain was ascribed to a disorder of the acromioclavicular joint in the presence of pain on shoulder abduction and tenderness localized to this joint; glenohumeral joint disorder was defined according to pain on passive internal rotation of the shoulder with the elbow flexed and the arm held by the side (25). Assessments of patients and controls were made by a single trained research assistant after a high level of consistency was shown during a series of examinations carried out jointly with a rheumatologist (WPM) on healthy subjects and outpatients who had either soft-tissue complaints of the shoulder or systemic inflammatory joint diseases.

Table 1. Diagnostic criteria for disorders of the shoulder-the Southampton examination schedule (24)
DisorderDiagnostic criteria
  • *

    Not defined in the Southampton examination schedule. See reference 25 for description.

Rotator cuff tendinitisHistory of pain in the deltoid region and pain on resisted active movement (abduction– supraspinatus; external rotation–infraspinatus; internal rotation–subscapularis)
Bicipital tendinitisHistory of anterior shoulder pain and pain on resisted active flexion or supination of forearm
Shoulder capsulitisHistory of pain in the deltoid area and equal restriction of active and passive glenohumeral movement with capsular pattern (external rotation > abduction > internal rotation)
Acromioclavicular joint disease*Pain on shoulder abduction and point tenderness localized to the joint
Glenohumeral joint disease*Pain on passive internal rotation of the shoulder with the elbow flexed and the arm held by the side

Systematic evaluation by MRI.

A range of structures was assessed using a prepared data sheet. These included the following: 1) rotator cuff tendons and their respective muscles for tear, tendinitis, and atrophy; 2) biceps tendon; 3) deltoid muscle; 4) the acromioclavicular and glenohumeral joints for the presence of changes in the cartilage and subchondral bone (arthrosis), synovial thickening, and effusion; 5) acromion morphology; 6) subacromial or subdeltoid bursae for fluid or thickening; 7) bones for bone marrow edema, the location and severity were noted; 8) humeral tuberosities for BME, erosion, and cyst formation; 9) miscellaneous pathology including paralabral cyst, Hill Sachs deformity, and os acromiale. Cuff tendons were assessed for partial or full-thickness tears. Cuff muscle atrophy or edema was noted. For this study we have referred to the supraspinatus and infraspinatus tendons as the “superior cuff” or “superior tendons.” This acknowledges their attachment to the greater tuberosity, which is partly conjoined. Joint arthropathy, tendinitis, effusion, and synovial thickening, were classed as absent, mild/moderate, or severe. The degree of articular surface change (arthrosis) was based on the presence of cartilage loss, joint space narrowing, osteophytosis, subchondral sclerosis, and subchondral cyst formation.

A complete listing of MRI definitions is provided in Table 2. Abnormal increased bone marrow signal on the fat suppression sequences that did not appear cystic or erosive on T1 spin-echo sequences was defined as BME. This was graded as either absent, mild/moderate, or severe. Although the presence of increased marrow signal on T2 short tau inversion recovery (STIR) sequences can represent inflammation and/or cellular infiltration as well as BME, we felt our arbitrary definition was suitable for this study. Entheseal BME was defined as bone marrow edema involving subcortical bone at a site of tendon, ligament, or muscle attachment. In the case of the acromion and clavicle, entheseal BME was documented as such only if it could be clearly identified as a separate entity from any subchondral BME within the acromioclavicular joint. Where this was not possible, any edema in these bones was assumed to be related to acromioclavicular joint pathology. Abnormal increased signal within a tendon that was not high enough on the T2-weighted sequences to constitute a tear was classified as tendinosis.

Table 2. Magnetic resonance imaging definitions
Rotator cuff tearAbnormal signal in rotator cuff tendon that is increased on all T2 images—similar to joint fluid.
Rotator cuff tendinosisAbnormal signal in rotator cuff tendon that is increased on all T2 images—less intense than joint fluid.
Tuberosity cystAbnormal signal in the tuberosity of the humeral head that is well-circumscribed and of increased signal on T2 sequences similar to joint fluid. The overlying cortex is intact and there is complete loss of marrow fat signal on the T1 sequence.
Tuberosity erosionAbnormal signal in the tuberosity of the humeral head that is well-circumscribed and of variable signal on T2 sequences; less intense than joint fluid. Overlying cortex is destroyed and marrow fat signal is completely lost on T1 sequence.
Bone marrow edemaAbnormal signal in bone marrow that is increased on T2 sequences and is best appreciated when fat suppression is employed. Overlying cortex is intact and there is partial loss of fat signal on T1 sequence.
Joint or bursal fluid/effusionIncreased volume of material in the appropriate space that is of signal character similar to joint fluid.
Joint or bursal inflammationIncreased volume of material in the appropriate space that demonstrates increased signal on T2 sequences but is not as bright as joint fluid.
ArthrosisHyaline cartilage thinning, labral abnormality, subchondral sclerosis, or subchondral cyst. In the acromioclavicular joint, this also includes capsular thickening.

An initial sample set of 5 examinations from the computer-generated list was studied by 3 experienced observers to achieve consensus in the approach to image interpretation. Following this, examinations from the 2 groups were mixed in random order and read independently by 2 observers unaware of the patient's history. Disagreement was resolved with arbitration by a third observer.

MRI protocol.

All patients were subjected to the same MRI protocol: T2-weighted turbo spin-echo (TSE) sagittal and coronal, T1-weighted coronal, and proton-density weighted TSE axial. Patients also underwent either T2 turbo spin-echo with fat suppression (T2TSE/FS) or STIR imaging. Patients were examined with T2TSE/FS or STIR in both sagittal and coronal planes. All of the control and study examinations were performed in 1 of 6 centers with standardized protocols and sequence parameters. The examinations were all performed on 1.5T Siemens magnets installed since 1999, utilizing a dedicated shoulder array (Siemens 200 or 165 mm, Siemens Corp., Erlangen, Germany) with the following specifications: T2TSE repetition time (TR) 2,740–2,960 msec, time to echo (TE) 85–97 msec; T2TSE/FS TR 2,740–2,770 msec, TE 85 msec; STIR TR 3,970 msec, TE 45 msec, inversion recovery time 150 msec; T1 TR 360 msec, TE 15 msec; Proton density TR 2,000 msec, TE 16 msec. All sequences were performed with a 180-mm field of view, 256 × 512 matrix and 4-mm slice thickness. Acquisition times were 1.3–4.6 minutes.

Statistical analysis.

Data were analyzed using SPSS for windows, version 11.0 (SPSS, Inc., Chicago, IL). Multiple logistic regression was used to compute odds ratios plus 95% confidence intervals for the analysis of different shoulder disorders defined clinically in AS patients and controls matched for age and sex. Comparisons of proportions of MRI-defined categorical variables between AS and control shoulders were conducted by using chi-square or Fisher exact tests. The sensitivity and specificity of each variable were computed based on AS and control shoulders. P values were corrected for the number of comparisons. A P < 0.05 was considered statistically significant.


Clinical evaluation.

Patient cohort A.

Of the 400 charts that were reviewed, not a single patient with AS was recorded as having specific shoulder joint pathology, but shoulder pain was recorded in 14 (3.5%).

Patient cohort B.

Of the 100 patients with AS that were randomly selected to attend systematic evaluation, 73 responded and were examined (males 72.6%, mean age 44.6 years [range 29–66 years], mean disease duration 18.9 years [range 3–26 years]). The prevalence of shoulder disorders in these AS patients and consecutive controls (males 40.4%, mean age 51.2 years [range 22–78]) attending primary care practice for unrelated complaints (control cohort A) is provided in Table 3. Eighteen (24.7%) patients had clinically evident shoulder involvement, which was bilateral in 3 (4.1%). Rotator cuff tendinitis was significantly more common in patients with AS (11 patients, 15.1%) than in controls (10 patients, 3.5%; OR adjusted for age and sex 8.2, 95% CI 2.7–25.1, P < 0.001). Only 5 (6.8%) patients with AS had clinically evident acromioclavicular joint disease and there were no patients with clinically evident bicipital tendinitis or glenohumeral joint involvement. The clinical diagnosis was not determined in 2 patients. No significant differences were evident between patients with and without shoulder involvement with respect to age, sex, disease duration, disease activity (as measured by the Bath Ankylosing Spondylitis Disease Activity Index) (26), functional disability (as measured by the Bath Ankylosing Spondylitis Functional Index) (27), or prevalence of hip involvement (data not shown).

Table 3. Prevalence of shoulder disorders in AS patients (n = 73) and primary care practice controls (n = 285) as defined by the Southampton physical exam schedule*
Shoulder disorderAS patients No. (%)Controls No. (%)OR (95% CI)P
  • *

    AS = ankylosing spondylitis; OR = odds ratio adjusted for age and sex; 95% CI = 95% confidence interval; RCT = rotator cuff tendinitis; ACJ = acromioclavicular joint disease; BCT = bicipital tendinitis; GH = glenohumeral joint disease.

All disorders18 (24.7)12 (4.2)8.2 (3.1–21.3)< 0.001
RCT11 (15.1)10 (3.5)8.2 (2.7–25.1)< 0.001
ACJ5 (6.8)0  
Not determined20  

MRI evaluation.

Fifteen AS patients from patient cohort B with clinically evident shoulder disease had an MRI; the study group comprised 12 males and 3 females, mean ± SD age was 47.6 ± 9.2 years, and mean disease duration was 17.6 years (range 3–26 years). Seventeen affected shoulders were scanned. The control group of 91 individuals with nonspecific shoulder pain (control cohort B) comprised 51 men (mean ± SD age 43.7 ± 14 years [range 16–68 years]) and 40 women (mean ± SD age 46.2 ± 8.4 years [range 31–61 years]). MRI studies (n = 101) were reviewed and 7 examinations were excluded from the control group for a variety of reasons: poor image quality (1), history of previous shoulder surgery (2), incomplete study (2), and established diagnosis of rheumatoid arthritis (2). This resulted in a final control cohort B of 91 patients in whom 94 shoulders were examined. Three patients had undergone a bilateral study. Patients had been referred from various specialties: orthopedic (56), general practice (30), sports medicine (1), and other (3). The indications for MRI examinations among the control group were pain (87) and previous minor injury (61). A total of 54 MRI variables were recorded for each examination. Frequencies of the most common features in patients and controls are shown in Table 4.

Table 4. Frequencies of the most common lesions observed on MRI of shoulders from patients with AS and nonspecific shoulder pain*
MRI featureAS patients (n = 17) No. (%)Controls (n = 94) No. (%)Sensitivity (%)Specificity (%)P
  • *

    MRI = magnetic resonance imaging; AS = ankylosing spondylitis; ACJ = acromioclavicular joint; BME = bone marrow edema; GHJ = glenohumeral joint; A/C = acromial and/or clavicular.

 Arthrosis16 (94.1)64 (68.1)
 Subchondral11 (64.7)37 (39.4)64.760.60.07
 Effusion1 (5.9)13 (13.8)5.986.20.69
 Synovitis2 (11.8)3 (3.2)11.896.80.17
 Arthrosis2 (11.8)4 (4.3)11.895.70.23
Greater tuberosity     
 Erosion11 (64.7)13 (13.8)64.786.2< 0.001
 Erosion plus BME10 (58.8)8 (8.5)58.891.5< 0.001
Rotator cuff     
 Tear7 (41.2)46 (48.9)
 Tendinosis12 (70.6)54 (57.4)70.642.60.42
Entheseal BME     
 Any site12 (70.6)18 (19.1)70.680.9< 0.001
 Intense BME9 (52.9)2 (2.1)52.997.9< 0.001
 A/C BME8 (47.1)1 (1.1)47.199.0< 0.001

Acromion, clavicle, and acromioclavicular joint.

The commonest abnormalities in AS shoulders were observed in the acromioclavicular joint. Although arthrosis was noted in 16 (94.1%) AS shoulders, it was also observed in 64 (68.1%) control shoulders (P = 0.04, corrected P [Pcorr] = not significant [NS]). Subchondral BME was observed on one or both sides of the joint in 11 (64.7%) AS and 37 (39.4%) control shoulders (P = 0.07, Pcorr = NS; Figure 1). In the AS group, subacromial inflammation was seen in only 2 (11.7%) shoulders, whereas in the control group, subacromial fluid was present in only 16 (17.0%), inflammation in 5 (5.3%), and both were observed in 5 (5.3%) shoulders.

Figure 1.

Magnetic resonance image of right shoulder of 35-year-old male control subject with shoulder pain.A, Coronal turbo spin-echo T2 image. Note small partial-thickness tear of supraspinatus tendon at its insertion (black arrow). B, Coronal short tau inversion recovery image. Abnormal bright signal is seen in and around the acromioclavicular joint with bone marrow edema in the clavicle better seen on the fat suppression sequence (long white arrow). Note the subtle focus of bone marrow edema in the greater tuberosity of the humerus (short white arrows) associated with the supraspinatus tendon tear. HH = humeral head; DM = deltoid muscle; SS = supraspinatus; ACR = acromion process.

Glenohumeral joint.

A mild to moderate effusion was seen in 1 (5.9%) AS and 13 (13.8%) control shoulders, synovial thickening was present in 2 (11.8%) AS and 3 (3.2%) control shoulders, and mild to moderate arthrosis in 2 (11.8%) AS and 4 (4.3%) control shoulders (all P = NS). Features possibly related to previous glenoid injury (bony defect, focal sclerosis, subchondral cyst) were seen in 1 (5.9%) AS and 2 (2.1%) control shoulders.

Greater tuberosity.

In the AS group, 11 (64.7%) shoulders demonstrated erosions of the greater tuberosity as compared with 13 (13.8%) control shoulders (P < 0.001, Pcorr = 0.02; (Figure 2). These were the same AS patients diagnosed as having rotator cuff tendinitis on clinical evaluation. Bone marrow edema adjacent to an erosion was seen in 10 of the 11 AS and in 8 of the 13 control shoulders. In the control group, an erosion with adjacent BME on the greater tuberosity only occurred in the presence of a superior cuff tear. This was not evident in the AS shoulders.

Figure 2.

Magnetic resonance image of the right shoulder of a 41-year-old man with a 9-year history of ankylosing spondylitis and a 6-month history of right shoulder pain diagnosed as rotator cuff tear on clinical evaluation. A large erosion in the greater tuberosity of the humerus demonstrates loss of marrow fat and is of relatively low signal intensity on both sequences (long arrows). A, Coronal turbo spin-echo T2 image. B, Coronal short tau inversion recovery (STIR) image. Intense inflammation/edema in bone surrounding the erosion is better seen on STIR acquisition (short arrow). HH = humeral head.

Rotator cuff.

Superior cuff tear (supraspinatus or infraspinatus) was seen in 7 (41.2%) AS shoulders, of which 1 was full thickness and 6 were partial thickness, and in 46 (48.9%) control shoulders, of which 22 were full thickness and 24 were partial thickness (P = NS). Two shoulders with full-thickness tears of the subscapularis tendon were identified only in the control group. No tear of the teres minor tendon was seen in either group. Tendinosis, whether diffuse or focal, in the superior cuff was seen in 12 (70.6%) AS and 54 (57.4%) control shoulders (P = NS).

Entheseal edema.

BME at any entheseal site (greater tuberosity, acromial, clavicular) was noted in significantly more AS (12 shoulders, 70.6%) than in controls (18 shoulders, 19.2%) shoulders (P < 0.001, Pcorr = 0.02; Figures 3 and 4). Furthermore, most AS patients had intense edema (9 of 12 shoulders) as compared with controls (2 of 18 shoulders). Greater tuberosity entheseal BME unrelated to erosion was present in 8 AS shoulders, of which 5 were intense and 3 mild (Figure 5). Two of the AS shoulders with intense BME had partial superior cuff tears and 1 of the shoulders with mild BME had a partial tear. Two control shoulders demonstrated intense entheseal BME, 1 with a full-thickness superior cuff tear and 1 with a partial-thickness tear. Mild entheseal BME was present in 15 control shoulders, of which 14 had full-thickness (7) or partial-thickness (7) superior cuff tears.

Figure 3.

Magnetic resonance image of the right shoulder of a 49-year-old man with a 21-year history of ankylosing spondylitis and bilateral shoulder pain of 1 year's duration. Sagittal short tau inversion recovery image at the level of the acromion process. Intense bright signal in bone marrow is present along the superior edge and in the posterior aspect of the acromion process at the origins of the deltoid muscle (arrows). Other images confirmed that this bone marrow abnormality was clearly separated from the acromioclavicular joint. ANT = anterior; POST = posterior; HH = humeral head; DM = deltoid muscle.

Figure 4.

Magnetic resonance image of the right shoulder of a 31-year-old woman with a 9-year history of ankylosing spondylitis and right shoulder pain of 2 year's duration. Sagittal short tau inversion recovery image at the level of the acromion process. Abnormal bright signal is present in the posterior aspect of the acromion process at the deltoid attachment (arrow) and focally within adjacent soft tissue. ANT = anterior; POST = posterior; HH = humeral head; DM = deltoid muscle.

Figure 5.

Magnetic resonance image of the right shoulder of the same patient as in Figure 3. Coronal short tau inversion recovery image. Increased signal intensity in the greater tuberosity of the humerus is typical of edema/inflammation at insertion of supraspinatus tendon (arrow).

Acromial entheseal BME at the deltoid origin was present in 8 (47.1%) AS shoulders, of which 7 (41.2%) were noted as intense (Figure 6) and 1 as mild. Mild acromial entheseal BME was seen in only 1 control shoulder in a patient with a deltoid avulsion injury (P < 0.001, Pcorr = 0.02). High signal was also seen in the deltoid muscle adjacent to its acromial origin in 4 of the 7 AS shoulders with intense acromial entheseal BME. Mild entheseal BME was present in 2 AS shoulders at the clavicular origin of the deltoid. This finding was not seen in any of the controls.

Figure 6.

Magnetic resonance image of the right shoulder of a 48-year-old man with an 18-year history of ankylosing spondylitis and bilateral shoulder pain of 1 year's duration. Sagittal short tau inversion recovery image at the level of the acromion process. Intense inflammation in the posterior aspect of the acromion process is associated with marked increase in signal intensity in the deltoid muscle at its origin along the posterior aspect of the scapular spine/acromion process (arrow). (Radiofrequency type artifact is projected over the lower part of the illustration). ANT = anterior; POST = posterior; HH = humeral head; DM = deltoid muscle.

Validation in a prospective cohort of AS patients (patient cohort C).

Consecutive patients with AS (73.2% males, mean age 45.3 years [range 24–65 years]) mean disease duration 19.7 years [range 2–39 years]) were systematically evaluated by clinical exam in a prospective manner for AS-specific shoulder lesions identified on MRI. Rotator cuff enthesopathy was evident in 17 (22.4%) of 76 patients. This was the second most common site for enthesopathy evident clinically, the greater trochanter site being the most common (data not shown).


This work represents the first systematic, controlled evaluation by clinical exam and MRI of shoulder disorders in patients with AS. It shows that symptomatic disease is comparatively common when compared with other peripheral sites of involvement. Furthermore, the frequency of abnormal radiologic findings may represent a conservative estimate because MRI examination demonstrated lesions in the acromioclavicular joint of almost all AS patients. This is despite the absence of supporting clinical findings and despite the fact that AS patients with asymptomatic shoulders were not scanned.

The greater tuberosity was included in the first index that was developed to score enthesitis (28), but the comparative frequency of involvement has been addressed in only one study that attempted to develop a more simplified version of this index (29). In this latter study, the process of data reduction based on the frequency of involvement led to a more concise index that primarily included entheses located in the axial skeleton with only the Achilles tendon enthesis being included as a peripheral site. Achilles tendon involvement was much less common in our prospective cohort (data not shown), suggesting either referral bias or potential differences in disease phenotype as possible explanations. Yet another recent approach to scoring enthesitis does not even include the shoulder in the scoring system (30). In addition, a recent comprehensive evaluation of peripheral entheses by ultrasound in 164 patients with spondylarthropathy also did not evaluate the shoulder (31).

This study also demonstrates the major MRI findings in the shoulder of a group of AS patients with shoulder pain compared with a control group of symptomatic patients not known to have any inflammatory arthropathy. Intense entheseal BME, particularly in the acromion and in association with erosion at the greater tuberosity, was the most specific finding in AS shoulders. Greater tuberosity edema was evident in controls, where it was usually mild and almost always associated with cuff tear. In the AS group, 5 patients demonstrated intense humeral BME, 3 of which had no associated cuff tear and 2 had only a small tear. Entheseal BME at the acromial or clavicular origin of the deltoid was present in 10 shoulders in the AS group compared with 1 control shoulder with a deltoid muscle injury. The presence of entheseal BME, particularly intense BME in the absence of significant injury, appears to be a finding specifically associated with AS. It was interesting to note that in one of our patients, acromial entheseal BME at the posterior deltoid origin was only well demonstrated on the sagittal fat suppressed image. The coronal image made confident assessment for edema at the posterior aspect of the acromion difficult due to partial volume averaging at this location. In view of this observation, we feel a sagittal fat suppressed image is essential in the assessment of acromial edema.

A previous study had examined the shoulder of 26 consecutive AS patients attending an in-patient physiotherapy program using plain radiography (4). Radiologic changes were common but usually minor, being evident in 31% of patients, the acromioclavicular joint being most commonly involved. Severe radiologic changes in the glenohumeral joints were noted in 9 (17.3%) shoulders, these being cystic changes, proliferation, and enthesopathy. Features suggesting rotator cuff degeneration were common and included cystic changes in the humeral head, erosion at the rotator cuff insertion, bony proliferation, and enthesopathy. A significant limitation of this latter study (4) was the absence of a control group. Our present study showed an even lower prevalence of glenohumeral lesions that was no different from that observed in controls. In addition, the lack of sensitivity is highlighted by the absence of radiologic changes in 26% of patients that had restricted shoulder movement. One additional study examined 52 AS patients and noted plain radiographic abnormalities in 29 shoulders, especially narrowing of the acromioclavicular joint (3). As highlighted in our report, acromioclavicular joint arthropathy is also commonly observed in other causes of shoulder pain. Sclerosis of the greater tuberosity and cystic changes in the humeral head were again noted, although as for the previous study, there was no control group.

McGonagle et al have demonstrated the presence of entheseal BME in sites around the knee joint and differentiated this from the findings in patients with rheumatoid arthritis (11). In their AS population, 6 (60%) of the 10 knee examinations demonstrated entheseal BME at an entheseal site. This is in concordance with our own findings in which 70.6% of the examinations in AS patients demonstrated entheseal BME (53% intense) either at the humeral tuberosity, acromial enthesis, or clavicular enthesis. However, they demonstrated perientheseal edema in all 10 knee examinations in their AS group, which was seen less commonly in our group. It is unclear whether this difference is due to image interpretation or a real variation in the course of the disease in the shoulder and the knee. This group has also shown that BME is common in patients with mechanical enthesopathy, namely, plantar fasciitis (32). However, this was more pronounced in spondylarthropathy patients with concomitant plantar fasciitis who were HLA–B27 positive. This concurs with our results demonstrating the specificity of intense entheseal edema for AS.

The variable sensitivity and specificity of clinical examination techniques has been previously noted (33, 34). One limitation of our study was the absence of a subacromial lidocaine injection test, although this may not discern the presence of symptoms arising from entheseal bone edema. As a consequence, it was unclear from this study whether the presence of entheseal BME and enthesitis was the predominant cause of pain in the AS patients in whom it was observed. A further limitation was that MRI examination was not performed on the control group that was also evaluated clinically (control cohort A).

Our data clearly show that primary glenohumeral joint involvement is not a feature of AS. This concurs with the careful pathologic examination by Romanus and Yden (35). Previous reports of narrowing of this joint likely reflect the elevation of the humeral head in relation to the glenoid, indicating rotator cuff disease associated with impingement and secondary osteoarthritis of the glenohumeral joint.

Early detection of AS is now becoming more important with the promise of disease-modifying therapy (30, 36). BME, rotator cuff tear, and entheseal inflammation can be examined by the single modality of MRI. As MRI examination becomes more widespread, the need to recognize features that may indicate the presence of AS becomes more important.

Our results lead us to conclude that in the absence of a significant rotator cuff injury, the presence of entheseal BME and, in particular, intense entheseal BME or erosive change with adjacent BME strongly suggests the presence of AS. The comparatively high frequency of rotator cuff enthesopathy also emphasizes the importance of recording this site both in routine clinical evaluation and in instruments for scoring enthesitis in clinical research. These findings will need to be further validated in other types of spondylarthropathy.


Walter P. Maksymowych is a Senior Scholar of the Alberta Heritage Foundation for Medical Research.