Many different etiologies for post-stroke shoulder pain are postulated and include adhesive capsulitis, bicipital tendonitis, complex regional pain syndrome, brachial plexopathy, altered sensation, spasticity and rotator cuff disease.⁵ These clinical entities, however, are likely end products of a complex cascade of events initiated by the biomechanical compromise of the post-stroke glenohumeral joint.⁶ Although arthrograms have been used to evaluate structural abnormalities in the painful post-stroke shoulder,^7–12 the results of these studies have been inconsistent. For example, the prevalence of rotator cuffs tears was reported to be zero in one study¹¹ and 40% in another.⁷ Given these inconsistencies and the invasiveness of arthrograms, Turner-Stokes and Jackson wrote, “It should be noted that all these studies were undertaken using arthrography, before the era of MRI, and the question should now be readdressed in larger numbers, in the light of more advanced and less invasive technology.”(p 286)⁵

This report represents the first MRI based evaluation of structural abnormalities in painful hemiplegic shoulders. The primary objective was to describe the structural abnormalities of the painful hemiplegic shoulder among chronic stroke survivors exhibiting pain duration greater than 3-mo. The second objective was to relate these findings to clinical characteristics.

Subjects were further characterized at baseline with respect to physical examination findings by physician investigators. Inferior glenohumeral subluxation was documented based on “fingerbreadths” relative to the unaffected shoulder.¹⁶ Motor status was documented based on the MRC scale.¹³ Specific muscles/muscle groups that were measured included the upper trapezius, shoulder abductors, elbow extensors, wrist extensors, finger flexors and finger abductors. The summary score was defined as percent of maximum (Sum of all grades/maximum sum × 100). Light touch and pin-prick sensations were assessed with the volar surface of the finger tip and a sharp safety pin, respectively.¹⁷ Specific test areas included the skin overlying the acromio-clavicular joint, lateral epicondyle and medial epicondyle, and the finger pads of digits 1, 3 and 5. Sensation to light touch and pin prick were coded as “normal” or “abnormal.”

A total of 89 stroke survivors received MRI scan evaluations of their painful post-stroke shoulder. The mean age was 57.8±11.0-yrs (SD), 67% were men, mean BPI 12 score was 7.1±1.9, mean time from stroke onset to study entry was 61.1±79.3-wks and the mean duration of pain was 53.4±7.5-wks. Sixty-seven percent sustained nonhemorrhagic strokes. Sixty-five percent exhibited left hemiparesis. The mean motor score was 49±29%. The mean glenohumeral subluxation was 0.89 ±0.56 fingerbreadth relative to the non-hemiparetic side with 61% exhibiting at least ½ fingerbreadth of subluxation. Light-touch and pin-prick sensations were impaired in 51 and 53% of subjects, respectively.

Table 1 shows the prevalence of rotator cuff tears. There was 1 case of biceps tear and no case of deltoid tear. Thirty-one (35%) subjects exhibited a tear of at least one muscle. Fourteen (16%) subjects exhibited tears of multiple muscles. Table 2 shows the prevalence of rotator cuff, deltoid and biceps tendonopathy and atrophy. Fifty subjects (56%) exhibited tendonopathy of at least one muscle. Eighteen (20%) exhibited tendonopathy of multiple muscles. Twenty-four subjects (27%) exhibited atrophy of at least one muscle. All subjects who exhibited atrophy exhibited atrophy in multiple muscles. Labral ligamentous complex abnormality was seen in 9%, subacromial bursa fluid in 26% and acromio-clavicular capsular hypertrophy in 67%.

Table 1 Prevalence (%) of Rotator Cuff Tears

Table 2 Prevalence (%) of rotator cuff, deltoid and biceps tendonopathy and atrophy

Table 3 shows the specific MRI structural abnormalities, the significant clinical correlates based on bivariate correlations and the independent predictors of the structural abnormality based on logistic regression. Supraspinatus tears were more common among older subjects (Figure 1) and among those with better motor function (Figure 2). Supraspinatus and infraspinatus tears were more common among those with more recent strokes (Figure 3). Atrophy of the rotator cuff, biceps or deltoids was more common with more severe motor impairments (Figure 4), less severe shoulder pain (Figure 5), older subjects (Figure 6) and women (Figure 7). Subacromial bursa fluid was more common with increased motor strength.

Table 3 MRI structural abnormalities, significant clinical correlates based on simple correlations and independent predictors of the structural abnormality based on logistic regression.

Figure 1 Prevalence of tear of the supraspinatus according to age.

Figure 2 Prevalence of tear of the supraspinatus among those with motor status greater than or equal to the median value of 50 versus less than 50.

Figure 3 Prevalence of tears of the supraspinatus and infraspinatus among those with time from stroke greater than or equal to the median value of 27.5-wks versus less than 27.5-wks.

Figure 4 Prevalence of any atrophy of the rotator cuff, biceps or deltoid muscles among those with motor status greater than or equal to the median value of 50 versus less than 50.

Figure 5 Prevalence of any atrophy of the rotator cuff, biceps or deltoids muscles among those with BPI 12 score greater than or equal to the median score of 7.25 versus less than 7.25.

Figure 6 Prevalence of any atrophy of the rotator cuff, biceps or deltoid muscles according to age.

Figure 7 Prevalence of any atrophy of the rotator cuff, biceps or deltoid muscles according to gender.

Sensitivity analysis revealed minimal effect of entry criterion p-value. When p≤0.20 was used, minor changes related to gender were noted. Gender became a significant predictor of subscapularis tendonopathy, but was no longer a significant predictor of atrophy of the supraspinatus. The use of the stricter criterion of p≤0.05 led to the exclusion of 3 variables that were significant predictors based on the p≤0.10 criterion. Specifically, motor status with regard to supraspinatus tears, and age and gender with regard to supraspinatus atrophy were excluded. With the minor exceptions noted above, the results of the logistic regression were unaffected by the entry criterion.

Post-stroke shoulder pain is common and significantly impacts rehabilitation outcomes. However, the exact etiology remains unknown. A description of the structural abnormalities of the painful post-stroke shoulder is fundamental to elucidating its etiology. The present report constitutes the first MRI based description of the painful shoulder among chronic stroke survivors. Stroke survivors with shoulder pain demonstrate high prevalence of tears, tendonopathy and atrophy of various muscles around the glenohumeral joint, subacromial fluid collection and acromion-clavicular capsular hypertrophy. However, with the exception of atrophy, these structural abnormalities were not related to the severity of shoulder pain.

Rotator cuff pathology has been postulated as an etiology of post-stroke shoulder pain.⁵ In the present series, 34% of subjects exhibited a tear of at least one rotator cuff muscle. Most were partial tears and the supraspinatus was most commonly affected. The prevalence of 34% is consistent with the 40 and 30% reported by Najensen et al⁷ and Nepomuceno and Miller,⁸ respectively, but higher than the 0 to 20% reported by others.^{9–11, 19} Differences in prevalence may be due to the differences in patient selection. Among studies that reported low prevalence rates, the 2 studies with the largest sample sizes included all stroke survivors regardless of pain, whereas the present study included only those with pain.^{10, 19} The age dependent increase in the prevalence of supraspinatus tears is consistent with the profile observed in the able-bodied asymptomatic population.²⁰ Given that the reported prevalence of rotator cuff tears among older asymptomatic able-bodied persons range between 22²⁰ to 40%,²¹ it would be difficult to definitively demonstrate a causative relationship between rotator cuff tears and post-stroke shoulder pain.

Subjects with more recent strokes were more likely to exhibit rotator cuff tears. It is possible that shoulder pain in these groups represent different etiologies with different natural history. If rotator cuff tear is the etiology of shoulder pain in the “tear” group, data suggest that this type of pain resolves in the long-term. These data further suggest that post-stroke shoulder pain that persists well into the chronic phase of stroke recovery is less likely to be due to a rotator cuff tear. Subjects with greater motor function were more likely to exhibit rotator cuff tears. Greater motor function in this population is still abnormal motor function, which may place the rotator cuff at increased risk for impingement, other traumas and eventual tears.

Fifty-three percent of subjects exhibited tendonopathy of at least 1 rotator cuff, biceps or deltoid muscle. Tendonopathy or tendonosis refers to intratendinous atrophy and degeneration of the tendon.²² While the term “tendonitis” is often applied loosely to a variety of shoulder problems that lead to focal pain and tenderness, the histological findings in the involved tissue are more compatible with an ischemic or degenerative process than an inflammatory one.¹⁸ Thus, for the purpose of MRI assessment, the term tendonopathy is preferred. As with tears, the supraspinatus was most commonly involved, followed by infraspinatus and subscapularis. Since tendonopathy is likely a precursor to tears, this parallel is not entirely surprising.

The prevalence of tendonopathy in the stroke population has not been well documented. Joynt ²³ injected local anesthetic into the subacromial bursa of stroke survivors with shoulder pain. Fifty percent reported significant reduction in pain or improvement in pain-free range of motion. This would be consistent with the results of the present series. However, it is uncertain whether the response to local anesthetic injection means that the pain was due to supraspinatus tendonopathy, subacromial bursitis or both. In the only large scale ultrasound based study of post-stroke shoulder, Lee et al¹⁹ reported significantly higher prevalence of rotator cuff “tendonitis” in the paretic shoulder (17%) versus the nonparetic shoulder (0%). While these studies suggest a relationship between tendonopathy and shoulder pain the present study did not corroborate their findings.

The prevalence of atrophy was similar among muscles studied, ranging between 20 and 23%. The exceptions were somewhat lower prevalence for the teres minor and biceps at 14% each and 0% for the subscapularis. As one might anticipate, atrophy was most commonly associated with reduced motor strength. However, atrophy was also associated with less pain. One might anticipate that greater atrophy leads to greater mechanical instability and greater pain. However, it is possible that atrophic muscles are also less spastic. Spasticity and post-stroke shoulder pain have been strongly associated²⁴ and may represent a cause and effect relationship. This may also explain why subscapularis atrophy was not observed in our series. Many stroke survivors exhibit significant reduction in external rotation range of motion of the shoulder, which has also been associated with shoulder pain.^{12, 25–27} Some have speculated that this is due to adhesive capsulitis.^{12, 26} However, spasticity of the subscapularis is an alternate consideration. Braun et al²⁸ postulated that spasticity causes pain by traction on the periosteum at the muscle insertion. Spasticity may also present a greater risk for development of impingement.⁵ If the “spasticity hypothesis” is correct, then rendering the subscapularis “atrophic” via neurolysis might reduce post-stroke shoulder pain. Earlier experience with phenol neurolysis of the subscapularis^{29, 30} and results of a recent small randomized clinical trial of botulinum toxin injection to the subscapularis lend credence to this hypothesis.³¹

Subacromial bursa fluid was seen in 26% of cases and was associated with greater motor strength. Stroke survivors with greater motor strength are more likely to use their hemiparetic upper limb and thus, may be at increased risk for repetitive movement trauma, leading to supraspinatus impingement and subacromial bursitis. A large percent of subjects had acromio-clavicular capsular hypertrophy. However, this was not related to any clinical variables.

While the present series represents the first large scale MRI assessment of the painful post-stroke shoulder, the study has a number of limitations. First, control comparisons were not included. Scans of the unaffected shoulder may provide invaluable comparative data to facilitate assessment of causative relationships. Similarly, a comparison group of motor impairment matched stroke survivors without shoulder pain may be invaluable. A lack of a relationship between structural abnormalities and shoulder pain in the present study may be an artifact of all subjects having shoulder pain. A longitudinal analysis of stroke survivors at risk for developing shoulder pain may provide further insights into the structure-pain relationship governing post-stroke shoulder pain. Second, the clinical variables were not comprehensive. The study did not include pain-free range of motion,²⁵ a measure of spasticity,³² a valid and reliable measure of post-stroke motor impairment³³ or other specific physical examination assessments for shoulder pain.³⁴ The addition of these variables may provide greater insight regarding the relationship between structure and clinical manifestations. Third, there was a selection bias in the study sample. By necessity, not all stroke survivors with shoulder pain were included due to the constraints of the clinical trial and MRI restrictions. Since the included subjects do not represent a random sampling of all stroke survivors with shoulder pain, the study results may not be generalizable to the broader stroke population.

This work was supported by a grant from NeuroControl Corporation, North Ridgeville, OH, the National Institute for Child Health and Human Development (1K24HD054600) and the General Clinical Research Center (M01RR0080) from the National Center for Research Resource