Examining patients with shoulder problems can be challenging because (1) shoulder motion involves a complex interaction of movement of the scapula on the thorax, the humeral head on the glenoid, and the clavicle at the acromioclavicular and sternoclavicular joints [6, 13, 24]; and (2) physical examination tests for the shoulder are sensitive but not specific for one particular shoulder condition [4, 5, 10, 20, 23, 26, 29–31]. It is important clinicians understand the limitations of physical examination tests when evaluating patients with shoulder pain.

Although we have used the shrug sign as a nonspecific indicator of shoulder abnormality, we found only two studies in the English literature that refer to this test [1, 2]. Blevins et al. [2] reported a positive shrug sign in eight of 10 professional athletes with rotator cuff abnormalities. However, those authors did not report the distribution of the shrug sign by the type of abnormality (full tears, three; partial tears, five; isolated contusion, two).

Our study was designed to evaluate our clinical impressions using the shrug sign and to test the hypotheses that (1) a positive shrug sign would be insensitive and nonspecific for rotator cuff disease; (2) a shrug sign would be reasonably reliable; (3) individuals with a positive shrug sign would be more likely to have loss of range of motion (ROM) or weakness to manual muscle testing in the involved shoulder; and (4) no demographic or physical examination finding would be associated with an increased likelihood of a positive shrug sign.

Our data were obtained prospectively and entered into a database for patients having shoulder surgery at our institution [5, 17, 18, 23, 25–27, 31, 32]. Inclusion criteria were shoulder surgery performed by the senior author (EGM) from 1994 through 2006 and the presence of shrug test data (see subsequently). Of the 991 patients in the database, 982 consecutive patients met our inclusion criteria. We obtained approval by our Institutional Review Board.

Either the senior author (EGM) or a trainee under his direct supervision (the senior author observed the tests being done and recorded the data) performed a preoperative assessment on all patients within 4 weeks of surgery, including a standardized subjective questionnaire (demographic and historical shoulder data) and a physical examination. Each patient had subjectively evaluated symptoms (eg, pain at rest, night pain, activity-related pain, pain with arm overhead, and others) through a visual analog scale of 100 points [17, 18].

In all patients, both shoulders had been exposed and examined. The examination included active and passive ROM, manual muscle strength testing, an upper extremity neurologic evaluation, and determination of other physical examination signs, including the Neer impingement sign [13, 18, 27, 31, 32], Kennedy-Hawkins impingement sign [17, 18, 27, 29, 31, 32], and Gagey sign [12] (suggested as a measure of inferior capsular contracture). Weakness in abduction or external rotation with the arm at the side had been recorded. The strength grading system used was that initially described by Lovett and Martin [22] and modified by Hoppenfeld [14]. We used preoperative radiographs (obtained for all patients) to determine a final diagnosis but did not include measurements or other analyses as part of this study.

For the shrug sign test, we asked the patient to abduct both arms to 90° in the plane of the body and to hold that position briefly. A shrug sign was considered positive if the patient had to elevate the whole scapula or shoulder girdle (“shrug”) to lift the arm to 90° (Fig. 1). The magnitude of the shoulder shrug was defined as the angle between the arm and the horizontal point at which the shrug movement began (measured by a handheld goniometer) (Fig. 2).

Fig. 1 A shrug sign was considered positive if the patient had to elevate the shoulder girdle for the arm to reach 90° abduction. The right shoulder shows a shrug sign; the left shoulder is normal.

---

Fig. 2 The degree of shrug sign was measured with a handheld goniometer. The patient was asked to elevate the arm until the shrug sign began, and then the angle between the horizontal and the arm was measured as shown.

---

The final diagnoses, based on preoperative radiographs and operative findings, included full-thickness rotator cuff tear, 261 patients; shoulder instability (anterior, anterior-inferior, posterior, or multidirectional), 221 patients; glenohumeral arthritis (osteoarthritis, osteonecrosis, or rheumatoid arthritis), 169 patients; partial-thickness rotator cuff tears, 88 patients; symptomatic rotator cuff tendinosis (no rotator cuff tear, just impingement) [8], 75 patients; isolated acromioclavicular joint arthritis, 61 patients; massive rotator cuff tear (defined as multiple tendon tears, including complete tears of two or more rotator cuff tendons), 47 patients; superior labrum anterior and posterior lesions, 25 patients; adhesive capsulitis, 19 patients; and other (infection, isolated biceps tear, failed arthroplasty, pectoralis major rupture, and benign soft tissue tumors), 16 patients. We defined rotator cuff disease as symptomatic tendinosis, partial rotator cuff tear, or full-thickness rotator cuff tear [29].

To assess the interrater reliability of the shrug sign, 30 patients (60 shoulder evaluations) not included in the study who presented to the senior author’s clinic for various shoulder problems were examined by the senior author and by an experienced physician assistant (JK) independently on the same day with the technique described previously. This number of patients was determined on the basis of previous experience with testing interrater reliability. Each examiner was blinded to the results of the other. Agreement on the presence of the shrug sign between these two raters was assessed using Cohen’s kappa coefficient [9], believed to be a more robust measure than a simple percentage calculation because it accounts for agreement occurring by chance. In addition to testing the reliability of the binary interpretation of the presence of a shrug sign (positive or negative), it was important to estimate the interrater agreement of the magnitude of the shrug sign. This agreement was assessed with the Shrout-Fleiss intraclass correlation coefficient to account for chance agreement [34]. We assessed the strength of the observed agreement between these two raters with a kappa coefficient.

After reliability of the evaluation for the presence of the shrug sign was established, we computed the percentage of the patient population with a positive shrug sign in each diagnosis. The diagnostic usefulness of a positive shrug sign for the presence of rotator cuff disease and other diagnoses in the involved shoulder was determined through evaluation of the test’s sensitivity, specificity, negative predictive value, and positive predictive value. We considered individuals with primary diagnoses of tendinosis, partial-thickness rotator cuff tear, full-thickness rotator cuff tear, and massive rotator cuff tear as having rotator cuff disease. Individuals with a primary diagnosis of superior labrum anterior and posterior lesions, glenohumeral instability, glenohumeral arthritis, acromioclavicular joint arthritis, or adhesive capsulitis were considered as not having rotator cuff disease. To achieve this definition, the patients were stratified on the basis of each preoperative diagnosis into two groups: the study group with a positive shrug sign and the comparison group without a shrug sign. Then we calculated diagnostic values using the following equations: where TP = true-positive, FP = false-positive, FN = false-negative, and TN = true-negative.

To determine the association between the presence of a positive shrug sign and loss of ROM in the involved shoulder, patients were characterized as having a positive shrug sign and as having loss of ROM in the involved shoulder in the direction of active and passive elevation in flexion, active and passive elevation in abduction, active internal and external rotation with the arm abducted 90°, passive internal and external rotation with the arm abducted 90°, and active and passive external rotation with the arm at the side. To assess the association between these analog measures, we used bivariate analysis (Pearson correlation) to test the correlation between the magnitude of the shrug sign and the range of shoulder motion. To determine the relationship of weakness to a positive shrug sign, patients with a strength grade of 4 or less in abduction or in external rotation with the arm at the side was the independent variable and patients with normal abduction strength (ie, Grade 5) comprised the control group. A third variable studied for weakness was the “drop arm sign”; the inability to hold the arm against gravity when it was placed above 90° elevation was considered a positive sign [7, 31]. We tested the association between these three binary outcomes and the shrug sign using a chi square test of independence with one degree of freedom.

The final objective was to determine the association between demographic and clinical findings and the presence of a positive shrug sign in the involved shoulder. We compared demographic and clinical characteristics, subjective symptoms, and physical examination findings between patients with and without a positive shrug sign. Univariate analysis was performed with Student’s t-test for continuous variables and the chi square test for categorical variables. To estimate the likelihood of a positive shrug sign, given these characteristics, we used logistic regression analysis. The outcome of interest was presence of a positive shrug sign. Independent variables included demographic characteristics (age, gender), clinical characteristics (eg, involvement of dominant arm, high-level sports activity, history of trauma), subjective symptoms (eg, rest pain, activity pain, night pain, lift arm above shoulder level, overhead activity pain, loss of ROM, limitation in throwing, difficulty in styling hair, limitation in sports participation), and other physical examination findings (eg, the loss of ROM in the involved shoulder). We calculated the correlation between the degree of shrug sign and other variables in these initial analyses, the 95% confidence interval, and the odds ratios by using univariate logistic regression with an alpha of 0.20. To control for potential confounding variables, stepwise logistic regression analysis was performed. We selected variables with a p value < 0.20 in the univariate analysis as candidates for the multivariate model to determine which factors were associated with a shrug sign. We used Statistics Program for the Social Sciences, version 15 (SPSS, Chicago, IL) for all analyses.

The interrater agreement kappa coefficient was 0.833. The interrater agreement in shrug sign magnitude in degrees (Rater 1: 14.3 ± 1.9 [mean ± standard error of mean]; Rater 2: 14.2 ± 1.9 [mean ± standard error of mean]) had an intraclass correlation of 0.875.

Patients with increasingly large-angle shrug signs showed increasing loss of motion; the highest correlations were with loss of active flexion (r = −0.803), active abduction (r = −0.772), and passive flexion (r = −0.720). Patients with large-angle shrug signs also had more abnormal physical examination findings, including the Gagey sign (abduction) (r = −0.660, p = 1.130E−11). Patients with positive shrug signs were associated with less strength in abduction (p = 1.884E−5) and external rotation (p = 0.0002131) and with more (p = 3.381E−11) positive drop-arm signs than patients without a positive shrug sign. If a patient had weakness in abduction, weakness in external rotation, and a positive drop-arm sign, the odds ratio was 32.634 that they had a positive shrug sign. Patients with massive rotator cuff tears were weaker (p = 0.005) in abduction and external rotation strength and had a higher positive rate of shrug sign than patients with other rotator cuff abnormalities (tendinosis, partial tears, or full-thickness tears; odds ratio, 2.580).

The “shrug sign” generally has been associated with a diagnosis of rotator cuff disease but, based on our clinical experience, we believed it to be nonspecific. We therefore hypothesized (1) a positive shrug sign would be insensitive and nonspecific for rotator cuff disease; (2) a shrug sign would be reasonably reliable; (3) individuals with a positive shrug sign would be more likely to have loss of ROM or weakness to manual muscle testing in the involved shoulder; (4) no demographic or physical examination finding would be associated with an increased likelihood of a positive shrug sign.

These findings must be interpreted with an understanding of the limitations of our study. First, our study group included only patients undergoing surgical shoulder procedures; it did not include patients who might have had other diagnoses (eg, paralysis, thoracic outlet syndrome, or neurologic conditions such as stroke) that could produce weakness in elevation and a positive shrug sign. Our patients, derived from a primarily referral shoulder practice, may have characteristics different from those of patients seen by others. There was no control group of patients without shoulder problems, but it would be difficult to perform an examination, MRI, or arthroscopy to confirm the lack of or presence of shoulder abnormality in a control group of asymptomatic patients. Second, it is possible that some of the examination findings in our population are associated with age alone. To our knowledge, there are no studies that address the relationship of the shrug sign to age alone. However, it has been shown that there is decreased shoulder ROM with age, and it is possible that increasing age alone may have influenced our results [28]. Third, we found weakness was associated with a shrug sign, but we did not measure strength objectively, and manual muscle testing is prone to interobserver variability [3, 15, 21, 33]. More sensitive measures of shoulder weakness might result in a higher or lower correlation with the shrug sign. It also is possible that the shrug sign results in part from concomitant shoulder pain because there was a high degree of rest and night pain in our patients. Fourth, although we did measure interobserver reliability of the shrug sign, we could not determine the intraobserver reliability. We found acceptable interobserver reliability, but we examined patients preoperatively and did not perform a repeated measures analysis to determine intraobserver reliability. Fifth, we did not study anatomically the structures that might contribute to a positive shrug sign. Although the shrug sign did correlate highly with a Gagey sign, a suggested measure of inferior capsular tightness [11, 12], the patterns of capsular tightness that led to a positive shrug sign could not be determined by our study. Also, we did not study scapular positioning, which might affect the type of shrug sign [16, 24]. Finally, we did not assess the long-term outcome to determine if the surgical procedure resulted in changes in the presence of the shrug sign.

Our data showed that the shrug sign is a nonspecific finding in patients with shoulder disorders and that it is not specific or sensitive for rotator cuff disease. In patients with rotator cuff disease, the shrug sign can be seen more frequently with massive rotator cuff tears because of weakness. The shrug sign is sensitive for conditions in which there commonly is loss of motion because of stiffness, especially adhesive capsulitis and glenohumeral degenerative arthritis. Specifically, the shrug sign was associated with loss of shoulder elevation and loss of rotation when the shoulder was elevated 90°. As expected, the patients’ functional limitations in terms of using the arm above shoulder level were associated with a shrug sign of an increasing angle or severity. We also found a relationship between weakness in abduction and external rotation and a positive drop-arm sign. Therefore, a patient with a positive shrug sign should be evaluated for stiffness or weakness as a cause of that sign.

Our data show the shrug sign can detect shoulder abnormalities, especially those associated with loss of ROM or weakness on manual muscle testing. However, the presence of a shrug sign warrants additional evaluation of its cause. Although this sign previously was associated with rotator cuff disease, we found it more commonly was associated with glenohumeral arthritis, adhesive capsulitis, and massive cuff tears. Patients with a positive shrug sign have altered shoulder function, and the shrug sign can be used clinically as a nonspecific physical examination finding indicative of shoulder dysfunction.