What Imaging‐Detected Pathologies Are Associated With Shoulder Symptoms and Their Persistence? A Systematic Literature Review

Objective Shoulder symptoms are common, and imaging is being increasingly used to help with management. However, the relationship between imaging and symptoms remains unclear. This review aims to understand the relationship between imaging‐detected pathologies, symptoms, and their persistence. Methods A systematic review using Medline, EMBASE, Cochrane, and grey literature was conducted to April 2017. The cross‐sectional and longitudinal relationships between imaging‐detected abnormalities and symptoms were analyzed and associations qualitatively characterized by a best‐evidence synthesis based on study design, covariate adjustment, and the Grade of Recommendations Assessment, Development, and Evaluation (GRADE) methodology. Modalities included ultrasound, magnetic resonance imaging (MRI), radiographs, positron emission tomography (PET), bone scintigraphy, and computed tomography. Results A total of 6,569 abstracts was screened and 56 articles were included. In total, 50 studies did not adjust for covariates and 36 analyzed individual pathologies only. The majority of studies showed conflicting results. There was no significant association between most imaging features and symptoms among high‐quality, cross‐sectional studies. There was low‐quality evidence that enhancement of the joint capsule on MRI and increased uptake on PET were associated with symptoms in adhesive capsulitis. Based on high‐quality longitudinal studies, enlarging rotator cuff tears were associated with an increased incidence of symptoms. Conclusion There were conflicting results on the association of imaging features with shoulder symptoms and their persistence. The existing evidence was very low in quality, based on the GRADE methodology. Further high‐quality studies are required to understand the relationship between imaging and shoulder symptoms and to determine the appropriate role of imaging in care pathways.


INTRODUCTION
Shoulder pain is a very common musculoskeletal condition and a significant contributor to disability and morbidity (1,2). Recovery can be slow, with high rates of chronic pain; in a community-based cohort, only 49% of respondents reported a complete recovery at 18 months (3). Shoulder pain has a significant negative impact on quality of life (4)(5)(6). It also poses a significant economic burden, with costs estimated to be £310 million in the first 6 months following primary care contact (7).
Imaging modalities such as ultrasound and magnetic resonance imaging (MRI) can accurately detect soft-tissue pathologies such as rotator cuff (RC) tears, tendinopathies, and subacromial bursitis (8,9), and can detect pathology more accurately than clinical examination. To aid the diagnosis of shoulder pain, different imaging modalities have therefore been increasingly used. Despite this increase in imaging, the relationship of imaging findings to patient outcomes remains unclear. A systematic review on the accuracy of imaging has highlighted the fact that further studies are required to determine the extent to which diagnostic tests on shoulder pain ultimately inform patient management and affect outcomes (10). A report by the UK Academy of Medical Sciences has also highlighted the importance of rational, cost-effective diagnostic tests to improve patient care and reduce costs (11).
The relationship between imaging-detected shoulder pathologies and clinical symptoms may be complex. Imaging studies have shown that pathologies exist in asymptomatic individuals (12)(13)(14), whereas other studies have suggested that certain features may correlate with pain (15,16). Our aim was to systematically review the literature to determine what imaging features are associated with symptoms and their progression when common imaging modalities are employed.

MATERIALS AND METHODS
Search strategy and selection process. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses methodologies were followed and are described in Figure 1 Studies were included if they reported the relationship between structural abnormality on imaging and symptoms (cross-sectional) or progression/persistence of symptoms (longitudinal). Structures included RC tear, tendinopathy, subacromial bursitis, subacromial space, and acromion. Outcome measures included pain or function measures. Imaging modalities included radiographs, ultrasound, computed tomography, MRI, and positron emission tomography (PET). Exclusion criteria were postsurgical patients, systemic inflammatory conditions (such as polymyalgia rheumatica), neurologic disease, chronic pain syndrome, fibromyalgia, and nonhuman studies. There was no language restriction. Data extraction. The citations identified by a preliminary search were screened by 2 reviewers (GT and PC) and for references not identified by the preliminary search. Discordance in opinion was resolved by a third reviewer (SRK). Data extraction was performed by 2 reviewers (GT and PC). Articles meeting the inclusion/exclusion criteria were divided into longitudinal and cross-sectional articles and were evaluated for their relationship to shoulder symptoms and whether single or multiple pathologies were

Significance & Innovations
• The majority of studies show conflicting results on the association of imaging-detected features with symptoms.
• The majority of studies did not evaluate the role of multiple pathologies in shoulder symptoms.
• Of the possible individual structures to be associated with pain, enlarging rotator cuff tears may be associated with incident symptoms.
assessed. Extracted data were inclusion criteria and population, patient number/controls, patient demographics (age, sex, and body mass index), study design, aims, imaging feature, symptoms, whether pathology was defined, results with or without adjustment for confounders, and findings.
Quality assessment. The quality of each observational study was independently assessed by 2 reviewers (GT, PC) (see Supplementary Table 2 (17,18), was adapted to assess the following components: study population, imaging feature, pain or function outcome, study design, and analysis and data presentation. A score of 1 or 0 was allocated for each question according to whether the study fulfilled the criteria or not. Any discordance in opinion was recorded, and where consensus could not be achieved, a third reviewer (PGC) was consulted. Quality scores were converted to percentages of the maximum scores for each class of article (cross-sectional, case-control, or cohort study). A study was considered to be high quality if it exceeded or equaled the mean score in its class.
Meta-analysis was inappropriate due to heterogeneity in study populations and imaging modalities. A narrative analysis of the evidence for features and their associations with symptoms was provided, based on the study design adequacy of adjustment for covariates, using a best-evidence synthesis approach (19). Comparisons were made for crosssectional studies and longitudinal studies. The research synthesis results were interpreted using the Grade of Recommendations Assessment, Development and Evaluation (GRADE) framework (20).
RC tear was associated with pain in 4 ultrasound studies (16,21,32,35), 3 MRI studies (16,52,62), and 1 bone scintigraphy study (72). These studies were unadjusted, and the majority were of low quality. Larger RC tear size (mean size 22.7 mm) was associated with pain in 1 low-quality study (35). One study reported no association of RC tear size and symptoms, although in symptomatic tears >175 mm 2 , pain was correlated with tear size (30). Two high-quality studies reported no association between RC tear size or location with pain, one of which was well-adjusted (45,48). The type of RC tear (partial or full) was not associated with severity of pain (32).
RC tear was associated with disability in 2 high-quality MRI studies (16,54). RC tear was not associated with functional disability in 1 high-quality ultrasound study (16) and 2 high-quality MRI studies, one of which was well-adjusted (45,49). One study reported an association of RC tears with disability on MRI but not on ultrasound (16). RC tear was associated with worse composite scores in 2 ultrasound studies (26,32). There was no association between RC tears and composite scores in 2 ultrasound studies (29,32) and 5 MRI studies (48,49,54,56,58). In summary, in 1 high-quality, well-adjusted study, RC tears were not associated with pain or function (45). The other studies were of mixed quality, unadjusted, and reported conflicting findings.
One high-quality ultrasound study reported a relationship between tendinopathies and disability (28). One highquality ultrasound study (16) and 1 high-quality MRI study (49) reported no relationship.
One ultrasound study reported a relationship between RC tendon thickness (≥0.8 mm) and symptoms, which were undefined (28). One low-quality MRI study found that only a high-stage tendinopathy, defined by complete Shoulder Symptoms and Imaging-Detected Pathologies  Shoulder Symptoms and Imaging-Detected Pathologies Shoulder Symptoms and Imaging-Detected Pathologies    Shoulder Symptoms and Imaging-Detected Pathologies disruption of supraspinatus tendon, was associated with symptoms (64). There was no association between RC thickness and symptoms of impingement (pain with functional impairment) between patients in 1 ultrasound study, but a significant difference in RC thickness of >1.1 mm was seen between affected and unaffected shoulders of the same patient (23). Three MRI studies of mixed quality (48,49,58) reported no relationship with tendinopathy and symptoms. In summary, high-quality but unadjusted studies found a conflicting relationship between pain, disability, and tendinopathy (16,28,32,48,49). No studies adjusted for covariates. Subacromial bursal pathology. Ten studies (16,21,24,25,31,49,51,53,54,61) evaluated the relationship between the subacromial bursa (SAB) and symptoms (Tables 1, 2, and 3). Five mixed-quality ultrasound studies (16,21,24,25,31) and 2 MRI studies of mixed quality (16,61) reported an association between SAB and pain. In 1 study, peribursal fat and fluid and bursal thickness, but not bunching, were associated with pain (24). One study reported an association with pain and SAB when seen alongside power Doppler within calcific deposits (31). One study reported that the location of bursa pathology was important (61). One ultrasound study (16) and 1 MRI study (16) reported that SAB effusion/thickening was associated with reduced function. Two high-quality MRI studies reported no association (49,53). One high-quality MRI study reported an association between bursitis and symptoms (54). Two highquality MRI studies reported no relationship between SAB enhancement and composite score (49,51). In summary, 2 high-quality, unadjusted studies found no relationship between shoulder symptoms and subacromial pathology (51,53). No studies adjusted for covariates.
Osteoarthritis (OA). Four studies of mixed quality (49,58,62,67) evaluated the relationship between shoulder OA and symptoms (Tables 1, 2, and 3). One low-quality combined MRI and radiographic study reported an association between subacromial osteophytes in patients with impingement (62). One high-quality radiographic study reported no relationship between acromioclavicular joint (ACJ) or glenohumeral joint (GHJ) OA and pain (49), and another no relationship with GHJ space width (67). One radiographic study reported that an increased size of osteophytes, but not joint space, was correlated with reduced range of motion (67). Two MRI studies of mixed quality reported no relationship between pain, function, and ACJ arthrosis (49,58). In summary, 1 high-quality, unadjusted study found no relationship with symptoms and features of ACJ or glenohumeral OA (49). There were no adjusted studies.
Calcification. Three studies evaluated the association between calcification and pain (21,22,31), and 2 studies evaluated the association with pain and function (49,68) (Tables 1, 2, and 3). Two ultrasound studies of mixed quality (22,31) showed that calcification was associated with pain, and 1 low-quality study found no association (21). Larger fragmented calcifications (mean dimensions: longitudinal 1.64 cm and transverse 1.39 cm) were associated with pain (31), as was morphology and color Doppler (22). Two high-quality radiographic studies reported no association with pain or function (49,68). In summary, 2 high-quality, unadjusted studies found no relationship with calcification and symptoms (49,68). There were no adjusted studies.
One high-quality combined radiographic and ultrasound study (27) reported that the acromial index (lateral extension of the acromion relative to humeral head) was associated with pain, whereas 2 high-quality radiographic studies (53,68) reported no association with pain or function. Those with full-thickness cuff tears had an increased prevalence of type 3 acromion compared to controls and patients with surgical impingement (47). No relationship existed between scapuloacromial angle (57), subacromial distance, or acromion shape (62) and impingement. One MRI study in patients with clinical impingement reported a reduction in the subacromial space during activity (50), and another reported decreased coracohumeral distance (62).
One high-quality, adjusted ultrasound study reported displacement in the coracoacromial ligament in symptomatic patients (34). A difference in distance (2.1 mm) between the inferolateral edge of the acromion and the apex of the greater tuberosity of the humerus was observed in affected shoulders in a high-quality study (23). Two studies of mixed quality reported an association with acromial humeral distance (AHD) and symptoms (46,55), but another high-quality study found no association (56). One study showed no relationship with acromial shape (55), whereas another was positively correlated to the intrinsic acromial angle (46).
In summary, high-quality, unadjusted studies found conflicting results on the relationship between symptoms and the acromial index (lateral extension of the acromion relative to humeral head) (27,53,68) and AHD (55,56). No relationship was found between scapuloacromial angle or acromion shape in high-quality, unadjusted studies (55,57).
Adhesive capsulitis. Six studies (15,60,65,71,73,74) evaluated the relationship between adhesive capsulitis and pain (Tables 1, 2, and 3). Two high-quality MRI studies reported enhancement of the joint capsule in the axillary recess, and RC interval was associated with pain intensity (15,60). One MRI study showed that capsular thickening was associated with decreased external rotation (15). Two PET studies of mixed quality showed increased uptake of 18 F-labeled fluorodeoxyglucose ( 18 F-FDG) in the RC interval, anterior joint capsule, or axillary recess (73,74). One low-quality bone scintigraphy study reported no difference (71). One low-quality study using bone scintigraphy and radiographs showed increased technetium uptake but no association between passive range of motion or recovery (65). In summary, high-quality, unadjusted studies have shown imaging features associated with symptoms in adhesive capsulitis (15,60,73).
Other features. Several studies evaluated other pathology imaging features. None of these were adjusted. Radiographically, 1 high-quality study showed that reduced upward and axial rotational tilts of the scapula were impaired in shoulder pain (66). There was no relationship in abnormal scapular planar glenohumeral motion measured using radiographs in patients with RC tears and pain in a low-quality study (69). One high-quality MRI study reported that the presence of glenohumeral effusion was not related to pain (59). One high-quality radiographic study found that in symptomatic patients with full RC tears >175 mm 2 , pain was correlated with humeral migration (30). One low-quality study reported an association between the absence of subacromial fat in patients with impingement (62). One high-quality study reported no association between acromioglenoid angle, supraspinatusglenoid angle, and pain (53). Glenoid-labral tear was associated with disability on MRI in a high-quality study (16). In another high-quality MRI study, glenoid-labral tears or cartilage damage were not associated with pain or functional impairment (49). In 1 high-quality ultrasound study, coracohumeral interval was significantly narrower in symptomatic shoulders (P < 0.0001) (33).
In 4 studies of mixed quality, an increase in RC tear was associated in the incidence of pain (39,40,43,44), although this was not shown in 2 other studies (41,42). Two highquality, unadjusted studies showed function worsened with increasing RC tear (39,44). An increase in RC tear and tear type from partial to full thickness was associated with the incidence of symptoms, measured using a composite score (39), although this did not reach statistical significance in 2 other studies of mixed quality (41,42). Overall, the high-quality studies suggested that increasing size of tears was associated with symptom incidence (39,40). These studies were unadjusted.
Tendinopathies and SAB pathology. One low-quality, unadjusted MRI study reported that patients with tendon edema and inflammation were more likely to achieve complete recovery with conservative treatment compared to those with fibrosis or tears (P = 0.038) (63). One highquality, unadjusted ultrasound study found that pain was associated with SAB thickness 1 week after a marathon swim (P = 0.032) (37).
Calcification. One low-quality, unadjusted ultrasound study showed that vascularity and shape were associated with resorption of the calcium deposit and improved pain (P < 0.001) (36). One low-quality, unadjusted radiographic study reported no association between calcium deposition and progression of pain or function (70).
Other features. One high-quality, unadjusted ultrasound study reported a reduction of the AHD narrowing on abduction correlated with improvement of symptoms (38).
The rate of progression to advanced fatty muscle degeneration on MRI and the long head of the biceps on ultrasound was associated with an increased odds of symptom incidence, measured using a composite score in a highquality, unadjusted study (41) but not in another study (40). One study reported no relationship between the incidence of pain and progression of fatty degeneration (40). Another low-quality, unadjusted study found supraspinatus atrophy was associated with worse strength and composite scores (44).

DISCUSSION
This systematic review is the first to comprehensively examine the relationship of imaging features with shoulder symptoms. The majority of studies reported conflicting results and evaluated single-shoulder pathologies. RC tendons contain nociceptors, and it would be rational to expect that as tear size increased, patients would be more likely to report pain and functional impairment. However, studies evaluating RC tears reported conflicting results, and the majority of studies were unadjusted and of low quality. A cross-sectional, high-quality, adjusted study did not find any association with symptoms, although in high-quality, unadjusted longitudinal studies, increasing size of tears was associated with symptom incidence. Inflammation may also be a possible cause of pain. The relationship between imaging and adhesive capsulitis was only evaluated in cross-sectional studies. Although these studies were unadjusted, enhancement of the joint capsule on MRI and increased uptake of 18 F-FDG in the RC interval, acromioclavicular joint, or axillary recess on PET may be associated with symptoms. There were conflicting results on RC tendinopathy, SAB pathology, and calcific tendinopathy and symptoms. These studies were unadjusted and of mixed quality, and therefore further high-quality studies are required to determine whether any relationship exists.
Studies have shown that numerous pathologies commonly coexist in the same symptomatic individual, although most studies in this review did not compare multiple pathologies with symptoms (75,76). Only 1 unadjusted, low-quality study evaluated shoulder pain with multiple pathologies in this review, and the authors found that SAB effusion may be associated with pain. Recent work suggests that these multiple pathologies may cluster into groups that could contribute to different outcomes, further confounding structure-pain analyses (75).
In 9 of the 56 studies, the pathologies being studied were not defined, and multiple studies used varying definitions Shoulder Symptoms and Imaging-Detected Pathologies  for the same pathology. These differences in nomenclature have further added to the confusion in diagnosing and treating shoulder pain (10). There was also heterogeneity between study populations, and this may be a reason for conflicting findings. Structure-pain relationships are complex. There is the possibility that there may be no relationship between imaging findings and symptoms, and imaging findings need to be considered as part of a wider pain construct. Other factors that may be associated with musculoskeletal symptoms include age, sex, body mass index, activity, mental health, and central sensitization (77). Only 6 studies adjusted for age and sex when evaluating the relationship between shoulder pain and imaging (15,26,27,35,51,54), and none adjusted for psychological factors. Other adjustments included occupation (54), arm dominance (15,27,54), and comorbidity (15).
There were several limitations to this work. Observational studies were rated relative to the overall mean quality scores, which may have artificially rated studies as high quality. However, the distribution of quality scores indicated a broad range of quality. Publication bias could not be assessed with a funnel plot as there were insufficient results for odds and relative risk ratios. Meta-analysis was not performed due to the heterogeneous nature of the measures of the features and pain or function outcomes.
Shoulder imaging is increasingly used for assessment of shoulder pain. Despite this rise, the relationship of imaging with symptoms and its role in informing management remain unknown. Understanding the relevance of imagingdetected pathologies and their role in the shoulder management pathway is essential to improving care and reducing costs. This review found that increasing RC tear and certain imaging features found in adhesive capsulitis may be associated with symptoms. Further high-quality, adjusted prospective studies evaluating the role of multiple imaging pathologies and other extrinsic factors are required to understand the role of imaging in shoulder pain care pathways.