In rheumatoid arthritis (RA) of the shoulder, loss of cartilage and soft tissue degeneration coexists with pain and reduced range of motion. We evaluated the presence of bony and rotator cuff degeneration in RA of the shoulder joint and assessed their relationship with pain and loss of functioning. We hypothesized that rotator cuff degeneration plays an important role in the presence of pain and loss of functioning of the rheumatic shoulder.
We used a cross-sectional study to assess both bony and rotator cuff involvement using plain anteroposterior radiographs, ultrasound, and computed tomography images. Additionally, we used an electromagnetic tracking device and a force transducer to evaluate range of motion and maximum force of the shoulder muscles. Between January 2003 and July 2004 we included 26 consecutive patients (51 shoulders). Twenty-one shoulders showed no or slight joint destruction, 15 showed intermediate destruction, and 15 showed severe destruction.
Only 19 shoulders showed an intact rotator cuff. Proximal migration of the humeral head and fatty degeneration of the infraspinatus muscle especially showed a significantly strong correlation with increased pain and function loss (R2 = 0.36, P < 0.001). In a multivariate regression analysis, proximal migration and fatty degeneration of the infraspinatus muscle were related most significantly with pain and reduced functioning in the shoulder joint.
Rotator cuff degeneration plays an important role in the daily functioning of the rheumatic shoulder. Prevention of rotator cuff degeneration may therefore play an important part in the treatment of the rheumatic shoulder.
Rheumatoid arthritis (RA) affects ∼1% of the adult population and exhibits a chronic fluctuating course that often results in progressive joint destruction, deformity, and disability (1). In the etiology of glenohumeral arthritis, shoulder involvement generally occurs late in the disease process, usually after other joints have manifested arthritic change. Any of the 4 shoulder girdle articulations can be involved, but the glenohumeral joints are the most frequently symptomatic (2). Symptoms vary between patients, both in etiology and intensity, but swelling, stiffness, pain, decreased strength, and loss of range of motion are cited as being the most important (3). RA destruction of the shoulder is characterized by proliferative synovitis (pannus), which is capable of degrading bone and cartilage matrix within and around the joint capsule. It not only results in cartilage thinning and bone loss, but also in soft tissue detachment and destruction (e.g., muscle atrophy, fatty infiltration, and tendinitis). Recent studies have reported that 24–52% of all patients age ≥50 years with RA of the shoulder joint have at least 1 large rotator cuff tear (4, 5). This might explain the poorer functional results and significantly greater postoperative pain after shoulder arthroplasty observed in RA patients compared with osteoarthritis patients (6, 7). Furthermore, fatty degeneration of the rotator cuff proved to be a significant predictor for inferior functional results after surgical rotator cuff tear repair (8–10).
In RA, both bony and soft tissue involvement in the disease process have been related to increased pain and decreased range of motion and force in the shoulder joint (3, 7, 11, 12). However, to our knowledge there is no report available about the individual contribution of joint and soft tissue destruction to pain, motion, and force. Furthermore, physical function was scored mainly using the Health Assessment Questionnaire as a qualitative measure of pain, range of motion, shoulder function, and force (13, 14). No quantitative or individual measurements for pain, function, and force of the shoulder were found in the recent literature.
Although complex, this study was set up to evaluate the incidence and severity of joint destruction and rotator cuff degeneration in the rheumatic shoulder. Additionally, we set out to assess the relationship between shoulder joint degeneration and pain, range of motion, and force. We hypothesized that in addition to joint destruction, rotator cuff degeneration is also a relevant factor in the loss of function and force in the rheumatic shoulder. Furthermore, because rotator cuff degeneration and proximal migration of the humeral head were strongly correlated (5), we hypothesized that proximal migration causing subacromial impingement significantly relates to the amount of pain experienced.
MATERIALS AND METHODS
To test our hypotheses, bony and soft tissue involvement were assessed using plain anteroposterior (AP) radiographs, computed tomography (CT) images, and ultrasound (US) images. Shoulder motion was recorded by means of a 6 degrees-of-freedom electromagnetic tracking and a force transducer to accurately evaluate the range of motion and the net maximum force about the glenohumeral joint.
Between January 2003 and July 2004, 26 consecutive patients with RA (51 shoulders) were included in the study. Patients were included initially after their treating physician had obtained bilateral AP radiographs in the assessment of their shoulder complaints. Final inclusion was based on the following criteria: 1) a clinical diagnosis of RA according to the 1987 American College of Rheumatology (formerly the American Rheumatism Association) criteria (15); 2) patient age >50 years, an age limit chosen to impose the smallest risk from radiation exposure (effective dose 1.6 mSv according to European Union guidelines ); 3) patient complaints of shoulder symptoms in ≥1 shoulder; and 4) no prior trauma or surgery to the shoulder. The study had prior institutional review board approval. All patients were informed and provided signed informed consent.
Six men and 20 women with an average age of 63 years (range 50–81 years) participated in the study. The mean Constant score was 68 (95% confidence interval [95% CI] 35–88) (17). One shoulder was excluded due to insufficient clinical data caused by a computer malfunction during range of motion and force measurements. Forty-one shoulders were symptomatic (objective pain and loss of function), with a mean Constant score of 65.2. Ten shoulders were asymptomatic, with a mean Constant score of 77.9. The mean interval between the diagnosis of RA and the CT scan was 13 years (range 1–40 years).
In order to assess the bony and cartilage involvement of RA, a standardized protocol AP radiograph was taken of each patient in the supine position, slightly turned to image side (20°), with the arm in external rotation, palm facing forward (18). Focus-to-film distance was measured at 115 cm, and a 15° craniocaudal tilt was used to project the undersurface of the acromion perpendicular. This created an optimal approximation of the true AP projection perpendicular to the glenohumeral joint (Figure 1). All radiographs were taken in a clinical setting in the presence of the principal investigator (MAJvdS) who controlled image quality and positioning (5).
Proximal migration, an indicator for fatty degeneration of the rotator cuff muscles (5), was measured using the Upward Migration Index (UI; UI = CA/R) (5, 19) as the distance between the center of the humeral head to the undersurface of the acromion (CA) divided by the radius (R) of the humeral head (Figure 1). Subacromial space measurement using the UI was validated with CT imaging (20). The mean ± SD absolute difference between the UI measured on AP radiographs and CT images was only 0.06 ± 0.07, providing a difference <5% of the mean UI measured on CT reconstructions. Medial displacement, an indicator for cartilage loss, was measured as the distance between the center of the humeral head to the most medial articular surface of the glenoid divided by the radius of the humeral head (19). We also calculated the medial displacement compared with the coracoid process as the distance between the center of the humeral head to the most lateral surface of the base of the coracoid process divided by the radius of the humeral head as a measure for bone loss (Figure 1) (21). All radiographs were scored for progression of rheumatic disease using the Larsen score (22, 23), which ranges from no to slight joint space narrowing (grade 0–1) to subchondral destruction (grade 3), to disappearance of original articular structure (grade 5).
Subsequently, all shoulders were scanned with a Toshiba Aquilion 16-slice CT scanner (Toshiba, Tokyo, Japan) using a constant protocol and calibration technique (24). Fatty degeneration was measured using the mean muscle density (MMD) (25). Individual rotator cuff muscles were outlined manually, carefully excluding pixels containing subcutaneous/intermuscular fat (Figure 2). A histogram was constructed from all voxels within the outlined region of interest in order to calculate the MMD of the rotator cuff muscles. The MMD is defined as the mean voxel intensity measured as the CT number within 1 outlined rotator cuff muscle in Hounsfield units. To correct for individual muscle-fat content, the MMD was divided by the body mass index (BMI) of the patient to find the normalized MMD (26). The MMD showed an interclass correlation coefficient (ICC) of 0.99 for repeated measurements and interobserver measurements. An ICC of 1 indicates that 99% of variation was caused by the difference between the patients (24). The teres minor and infraspinatus muscles were analyzed together because separation of these muscles has proven to be very difficult and unreliable (27).
All shoulders were examined for rotator cuff pathology by an experienced musculoskeletal radiologist using US (Table 1). All rotator cuff muscles were screened for the presence of tendinitis, a small tear, or a massive tear using standard US methods (28). Dinnes et al reported that the pooled sensitivities of US for diagnosing full-thickness and partial-thickness tears were 0.87 and 0.67, respectively (29).
Table 1. Average clinical results for rotator cuff pathology*
The range of motion was measured using the Flock of Birds 6 degrees-of-freedom electromagnetic tracking device (Ascension Technology Inc., Burlington, VT). This system consists of a transmitter emitting an electromagnetic field in which position and orientation of several receivers can be tracked. Prior to the measurements, a field calibration was performed (30). Four receivers were applied around the patient's shoulder: 1 was taped to the sternum, 1 to the upper arm, and 1 to the wrist. The fourth receiver was mounted on the flat upper surface of the acromion, in the most laterocaudal corner (31, 32). The full active range of movement of the shoulder was measured. The skin-fixed method was found suitable for dynamic recordings of scapular rotations because its intraobserver root mean square error was only 5°.
The maximum force of the shoulder muscles was assessed using a 6 degrees-of-freedom force transducer (AMTI-300; Advanced Mechanical Technology, Watertown, MA). Subjects were seated with the right arm in a splint and the elbow in 90° of flexion. The humerus was elevated 60° in the scapular plane (30° angle to the frontal plane). The forearm was positioned in a splint, which was attached to a 3-dimensional force transducer measuring the task force in an x-y plane, perpendicular to the longitudinal axis of the humerus. The arm was suspended in order to compensate for gravity. The force transducer was mounted on a sled so that it could move freely in the z direction, parallel to the humeral longitudinal axis. Axial rotation of the humerus was mechanically not restricted to prevent the subjects from generating supplementary moments (33, 34). The force exerted on the force transducer was displayed to the subject by a cursor on a video screen. The subjects were asked to generate a maximum force in 12 equidistant directions (30o apart) by moving the cursor along the displayed spokes of a wheel that denoted the force direction; concentric circles denoted the force magnitude. The maximum force that could be exerted in all 12 directions was measured (e.g., forward flexion 0o, abduction 120o). Force measurements were normalized for body weight.
Student's t-test was used to evaluate the differences in shoulder joint abnormalities within subgroups for the parameters presented above. All parameters were checked for outliers and verified to have a reasonably symmetric distribution. The modified t-test for unequal variances was used in case the Levene's test for equal variances was significant.
Multivariable linear regression analysis, Pearson's correlations, and Spearman's correlations were used to evaluate the relationship between the clinical parameters and bony and soft tissue involvement. Pearson's correlation coefficient was used on variables measured on a ratio scale using actual values; Spearman's rank correlation was used only for ordinal or ranked data. For each outcome (function, force, and pain) a multivariable linear model was constructed starting from a full model incorporating radiodiagnostic parameters and using backward elimination of nonsignificant predictors. The resulting model is then the smallest model for prediction of the outcome in the sense that deletion of any of the remaining predictors would significantly reduce the predictive ability of the model.
By dividing each slope of the regression line (i.e., each beta) by the SD of the associated independent variable in the data set, the coefficient measures the effect on the outcome of 1 SD (i.e., the unit of measurement becomes the SD). Significances are not influenced by this linear transformation. This results in mutually comparable regression lines and emphasizes the individual reliability of the radiodiagnostic measures. Additional influences of age, sex, dominance, arm side, and duration of rheumatic disease were included in the analysis. All analyses were performed using SPSS for Windows, version 14.0 (SPSS, Chicago, IL). Using a Bonferroni adjustment for t-tests and correlation analysis, P values < 0.005 and < 0.02, respectively, were considered significant.
Results for pain, range of motion, and force are subdivided for bony and rotator cuff degeneration and are presented in Tables 1 and 2.
Table 2. Average clinical results subdivided for radiodiagnostic parameters*
Values are the mean ± SD. N/kg = newtons per kilogram.
Larson scale 0–2.
Larson scale 3–5.
Pain at rest, 0–100
21 ± 2.3
15 ± 2.0
16 ± 1.9
39 ± 2.5
20 ± 2.2
26 ± 1.8
18 ± 2.2
29 ± 2.4
Pain during activities, 0–10
35 ± 2.8
25 ± 2.6
33 ± 2.7
52 ± 2.8
21 ± 2.2
50 ± 2.7
30 ± 2.5
48 ± 3.2
109 ± 25.4
113 ± 16.9
118 ± 18.0
82 ± 36.4
123 ± 14.8
115 ± 22.6
113 ± 25.0
99 ± 30.7
External rotation, degrees
54 ± 25.1
55 ± 23.9
56 ± 15.4
49 ± 11.4
39 ± 11.7
55 ± 14.2
58 ± 16.2
47 ± 14.6
Forward flexion, degrees
105 ± 24.2
103 ± 24.6
113 ± 16.8
89 ± 33.8
119 ± 15.9
118 ± 29.8
108 ± 24.7
98 ± 25.4
Maximum abduction force, N/kg
0.52 ± 0.20
0.52 ± 0.10
0.58 ± 0.20
0.43 ± 0.22
0.52 ± 0.19
0.51 ± 0.20
0.58 ± 0.20
0.50 ± 0.19
Maximum forward flexion force, N/kg
0.52 ± 0.20
0.52 ± 0.19
0.58 ± 0.21
0.39 ± 0.13
0.58 ± 0.20
0.52 ± 0.21
0.60 ± 0.21
0.47 ± 0.18
Radiographic evaluation of bone destruction and cartilage loss.
Twenty-one shoulders showed no or slight joint destruction (Larsen score 0–1), 15 were intermediate (Larsen score 2–3), and 15 were severe (Larsen score 4–5). The mean ± SD Medial Migration Index, a relative measure for cartilage loss (19), was 1.08 ± 0.17 (median 1.05, 1st quartile to 3rd quartile [IQR] 0.95–1.18) relative to the glenoid, and was 1.14 ± 0.19 (median 1.16, IQR 1.06–1.28) relative to the coracoid process.
The mean ± SD Proximal Migration Index, a relative indicator for rotator cuff pathology (5), was 1.31 ± 0.07 (median 1.32, IQR 1.28–1.36). In 13 shoulders no proximal migration was observed (Proximal Migration Index >1.35). In 26 shoulders moderate proximal migration (1.25–1.35) was present, and in 12 shoulders severe proximal migration (1–1.25) was present (5).
CT analysis of rotator cuff quality.
The MMD ± SD divided by BMI (normalized MMD), a quantitative measure for fatty degeneration for the supraspinatus and infraspinatus muscles, was subsequently 1.30 ± 0.9 and 1.60 ± 0.47, respectively (5). There was no significant (cross) relationship between the BMI and the clinical parameters (pain, range of motion, and function of the shoulder). Fatty degeneration of the rotator cuff muscles is presented in 3 severity groups: none (>1.74 HU/BMI), mild (0.82–1.74 HU/BMI), and severe fatty degeneration (<0.81 HU/BMI) (5).
US evaluation of the rotator cuff tendons.
The majority of shoulders showed rotator cuff pathology on US images. Tendinitis was found in 20 supraspinatus tendons, 22 infraspinatus tendons, and 13 subscapularis tendons. Six small tears were found in the supraspinatus tendon, and 1 additional tear was found in the infraspinatus and subscapularis tendons. A massive tear was diagnosed in 6 supraspinatus tendons and 2 infraspinatus tendons.
The relationship between radiographic parameters and functional results.
Coefficients of determination (R2) between radiodiagnostic and clinical parameters are presented in Table 3. Proximal migration of the humeral head and especially fatty degeneration of the infraspinatus muscle showed the strongest correlation with increased pain and decreased range of motion (abduction/forward flexion). The maximum range of abduction (R2 = 0.25) and forward flexion (R2 = 0.16) were also related to the amount of pain experienced.
Table 3. Coefficients of determination (R2) between clinical parameters and bony/soft tissue abnormalities
The partial correlation between range of motion and fatty degeneration of the infraspinatus muscle, controlled for pain, remained relevant and significant (R2 = 0.25, P < 0.01). Bony deformation (Larsen score) also correlated with the perception of pain, but not as strongly. The presence of an infraspinatus tear showed a stronger negative correlation with range of motion and pain, compared with an isolated supraspinatus tear. The combined correlation coefficients for all parameters presented in Tables 2 and 3 were R2 = 0.44 for pain, R2 = 0.43 for abduction, and R2 = 0.66 for abduction force (P < 0.001).
Results for the univariable regression analysis of pain, function, and force (dependents) with bony and soft tissue parameters (independents) are shown in Table 4. Using backward multivariable regression analysis for all above parameters, proximal migration and the MMD of the infraspinatus muscle were calculated as the strongest and most significant predictors for the degree of pain and dysfunction in the shoulder joint. Proximal migration presented the most significant influence on the amount of pain experienced (beta −1, P = 0.002). The MMD of the infraspinatus muscle was the most significant predictor for the amount of range of abduction (beta 9.9, P = 0.008) and forward flexion (beta 12.5, P < 0.0001). Further input of age, sex, arm side, and duration of rheumatic disease in this multivariable regression analysis revealed no confounders for the regression and correlation presented above.
Table 4. Univariable regression analysis between clinical and radiodiagnostic parameters*
We found a significant relationship between the Larsen score and fatty degeneration of the rotator cuff muscles (beta 0.37; R2 = 0.15, P < 0.001), and also between the duration of the disease and pain in rest (beta 1.2 [0.2/1.9]). However, individually neither the Larsen score nor the duration of rheumatic disease was a relevant predictor for shoulder dysfunction in a linear regression analysis. We found a significant difference for the mean amount of joint destruction, according to the Larsen score, between early (0–2 years) and progressed (>2 years) rheumatic disease. A significant difference for the mean amount of fatty degeneration of the rotator cuff muscles in early and progressed rheumatic disease (mean difference 0.36; P = 0.003) was found also.
Functional disability has been associated with pain, joint destruction, and rheumatic disease activity (12, 35). In the early stage of the disease, functional ability may be influenced more by disease activity than by joint destruction (36). Reports have shown a significant relationship between joint destruction and functional impairment later in the disease process (35). However, in these studies pain and function loss were measured either subjectively or by use of a questionnaire (12, 37). Therefore, a quantitative comparison between pain, function (range of motion and force), and shoulder joint destruction has added value. We set out to evaluate the incidence of bony and soft tissue abnormalities diagnosed using radiographs, CT, and US in the rheumatic shoulder. We also sought to determine the relationship between these radiodiagnostic changes and clinically relevant parameters such as pain, range of motion, and force.
Our results indicate a diverse involvement of bony and rotator cuff degeneration in the rheumatic shoulder. Although shoulder dysfunction and pain were related to multiple factors, we observed that rotator cuff quality in the rheumatic shoulder predicted a distinct increase in pain and a significant decrease of range of motion and force. Involvement especially of the infraspinatus and teres minor muscles showed a relevant and significant relationship with these clinical parameters. Although abduction and forward flexion force are mainly the result of deltoid muscle contraction, shoulder muscle imbalance caused by fatty degeneration of the infraspinatus muscle causes the adductors to compensate for lost downward force, resulting in a decreased total upward moment. We believe that pain and fatty degeneration of the infraspinatus and/or teres minor muscles can both induce proximal migration due to imbalance of shoulder muscle forces, and can cause secondary pain due to impingement or shoulder instability (5, 38). Because fatty degeneration is thought irreversible, early referral and treatment of shoulder involvement in rheumatic disease may protect patients with RA from this downward spiral (12, 38, 39).
We hypothesized that pain caused by RA (e.g., synovitis, swelling, or cartilage loss) leads to shoulder immobilization and disuse. This may initiate fatty degeneration of the rotator cuff muscles, causing shoulder muscle imbalance, dysfunction, and secondary subacromial shoulder pain (38, 40, 41). Clinical reports on shoulder arthroplasty in patients with RA have stated that they are referred too late for surgical treatment due to advanced shoulder joint destruction (6, 37). Kelly hypothesized a relationship between this late referral and the clinical course of rheumatic disease (37). Also, it has been stated that rheumatic shoulders with a more rapid and progressive joint destruction showed significantly more rotator cuff abnormalities and that these were related to superior subluxation of the shoulder joint (19, 37). In accordance with these findings, we found a significant relationship between the bony destruction and rotator cuff involvement, and a significant correlation between proximal migration of the humeral head and fatty degeneration of the rotator cuff. In particular, infraspinatus and teres minor degeneration were related to proximal migration (6). As we found no significant relationship between the duration of rheumatic disease and pain or function (range of motion and force), it seems more likely that the severity of RA, not disease duration, influences shoulder pain and function.
The presence of rotator cuff dysfunction, caused by fatty degeneration or tears, has been related to inferior functional results and increased pain after shoulder arthroplasty (38, 42, 43). This may be explained by glenohumeral instability in rotator cuff-deficient patients. In patients with massive rotator cuff tear, compensatory co-contraction of the adductor muscles (i.e., teres major, latissimus, and dorsi) was measured (40). It was hypothesized that this co-contraction prevents superior subluxation due to deltoid muscle forces in order to decrease subacromial pain (40, 41).
We included both symptomatic and asymptomatic shoulders because earlier reports suggested a significant relationship between rheumatic disease activity and functional performance (44). Although we did not analyze patients for disease activity using the Disease Activity Score, we could not reproduce the significance of disease duration or clinical activity of the disease on shoulder dysfunction. In addition, we did not find any correlation between the duration of rheumatic disease and pain, shoulder function, joint destruction, or muscle involvement.
Only 2-dimensional AP radiographs were used to assess bony and cartilage destruction. Soft tissue degeneration, function, and force were measured with very high accuracy and reliability. Therefore, it could be argued that the relationship between bony destruction and clinical parameters is questionable. Further evaluation of cartilage and bone loss is needed to evaluate the relationship of those losses to synovitis, loss of function, and pain. A relationship between soft tissue degeneration and cartilage loss can also be hypothesized as the result of shoulder muscle imbalance and increased proximal joint loading. Although only assessed by qualitative measures, we did find a significant relationship between the Larsen score and fatty degeneration. More accurate measurement of cartilage and bone loss may provide better insight on their relationship with pain and loss of function.
Shoulder muscle force is thought to be related to pain (40). Our results supported this relationship; pain was significantly correlated to abduction and anteflexion force (R2 = 0.3). Although pain at rest and activities was evaluated before and after force and range of motion measurement, pain measurement remains subjective. This may explain the difference between pain during rest and pain during activities (Table 3).
Force measurement was only normalized for body weight, not for arm length (distance from glenohumeral joint to transducer on splint). Although we found decreased forces during abduction and forward flexion in patients with severe supraspinatus and infraspinatus degeneration, this relationship was not found for mild degeneration. One could argue that this difference originates from a less reliable measurement of force. However, we believe that this difference supports our hypothesis that severe fatty degeneration of the rotator cuff muscles is related to subacromial impingement pain due to muscle imbalance, and therefore relates to a decreased range of motion and loss of abduction force.
Although no causal relationships were found, our results support the hypothesis that subacromial pain is induced by incapacity of the depressor muscles of the rotator cuff; pain and range of motion are significantly related to proximal migration and fatty degeneration of the infraspinatus and teres minor muscles. Prevention of rotator cuff involvement in rheumatic disease of the shoulder may lead to better functional results after shoulder arthroplasty. In already severely rotator cuff-deficient shoulders, a transfer of the teres major muscle combined with shoulder arthroplasty (in either 1 or 2 sessions) may be used as a salvage procedure in order to give patients with RA sufficient pain relief and increased postoperative range of motion (4, 40).
We believe it is of great importance to screen patients with RA for shoulder involvement at an early stage, even if other joints are the principle cause for medical concern. Because proximal migration of the shoulder measured with the Upward Migration Index is strongly correlated with rotator cuff disease (R2 = 0.74), an AP radiograph provides an easy-to-use and reliable screening of the rotator cuff (5). This study underlines the importance of this measurement, as we also found it to be strongly correlated with shoulder pain and functional parameters. Measurement of proximal migration at an early stage can play an important role in the timely initiation of functional and medicinal treatment of RA, and it may present patients with better possible outcomes when shoulder arthroplasty is indicated.
Dr. van de Sande 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 design. van de Sande, de Groot, Rozing.
Acquisition of data. van de Sande, de Groot.
Analysis and interpretation of data. van de Sande, de Groot.