SSR white paper: guidelines for utilization and performance of direct MR arthrography

Direct magnetic resonance arthrography (dMRA) is often considered the most accurate imaging modality for the evaluation of intra-articular structures, but utilization and performance vary widely without consensus. The purpose of this white paper is to develop consensus recommendations on behalf of the Society of Skeletal Radiology (SSR) based on published literature and expert opinion.

Materials and methods

The Standards and Guidelines Committee of the SSR identified guidelines for utilization and performance of dMRA as an important topic for study and invited all SSR members with expertise and interest to volunteer for the white paper panel. This panel was tasked with determining an outline, reviewing the relevant literature, preparing a written document summarizing the issues and controversies, and providing recommendations.

Results

Twelve SSR members with expertise in dMRA formed the ad hoc white paper authorship committee. The published literature on dMRA was reviewed and summarized, focusing on clinical indications, technical considerations, safety, imaging protocols, complications, controversies, and gaps in knowledge. Recommendations for the utilization and performance of dMRA in the shoulder, elbow, wrist, hip, knee, and ankle/foot regions were developed in group consensus.

Conclusion

Although direct MR arthrography has been previously used for a wide variety of clinical indications, the authorship panel recommends more selective application of this minimally invasive procedure. At present, direct MR arthrography remains an important procedure in the armamentarium of the musculoskeletal radiologist and is especially valuable when conventional MRI is indeterminant or results are discrepant with clinical evaluation.

Arthrographic Technique Pitfalls

MRI of the Hip: What the Surgeon Wants to Know

Article 04 May 2017

Society of Skeletal Radiology 2017 annual meeting summary

Article 22 September 2017

Introduction

The technique of direct magnetic resonance arthrography (dMRA) was first described in 1987 [1]. Despite the element of invasiveness that was introduced to an otherwise non-invasive imaging modality, several benefits were readily apparent. These included improved delineation of the surfaces of intra-articular structures related to joint distension, improvements in signal-to-noise as well as contrast-to-noise ratios as a result of the T1 shortening effects of dilute gadolinium, and added benefit of indirect arthrographic signs in the absence of direct depiction of the responsible defect (e.g., extra-articular leakage of contrast material and abnormal filling of bursae or adjacent joint compartments). Using a sample of cadaveric shoulders, wrists, knees, and ankles imaged on a 1.5-T system, the authors suggested higher accuracy for the diagnoses of several intra-articular abnormalities compared with conventional magnetic resonance imaging (cMRI) [1]. Clinical indications rapidly expanded over the subsequent years, often in parallel with surgical advancements such as arthroscopy [2,3,4,5,6,7], and dMRA became a routinely utilized and integral part of most clinical practices. After several decades of widespread use, dMRA is now accepted as a highly accurate imaging modality for the evaluation of intra-articular structures.

The drawbacks of dMRA have remained largely unchanged through the years, mostly related to the minimally invasive arthrogram component. Since the introduction of dMRA, there have been significant advancements in MRI technology, including widespread availability of 3-T scanners and better surface coils, that have vastly improved image quality. As a consequence, the advantages of dMRA over cMRI have evolved leading to variability in its utilization for different clinical indications and in different practices. In recognition of this, the Standards and Guidelines Committee of the Society of Skeletal Radiology (SSR) commissioned the current white paper. Twelve SSR members with expertise in dMRA were selected to form an ad hoc white paper panel and were tasked with reviewing the published literature and provide recommendations on the utilization of dMRA based on consensus expert opinion.

It should be noted that in the acute setting, a joint effusion may create an arthrographic effect. Some authors have suggested dMRA of the shoulder is unnecessary in this scenario [8, 9] (Fig. 1). This principle can be extrapolated to include all joints. Similarly, in the presence of moderate to severe arthritis, the value of dMRA is decreased, and if magnetic resonance (MR) imaging must be performed, the authorship panel recommends cMRI. A retrospective study in patients 50 years and older referred for hip pain found that 100% of patients with moderate to severe osteoarthritis (Tonnis grade 2–3 or joint space width ≤ 2 mm) demonstrated acetabular labral pathology on dMRA. Since arthroscopic surgical repair is limited in this setting, the authors concluded that dMRA may not be indicated [10]. We stress that panel recommendations are generalized for routine clinical practice and refer specifically to the utilization of cMRI or dMRA as the initial type of MR imaging exam. Notably, in the setting of an indeterminate cMRI exam, dMRA may be both an invaluable tool and appropriate for confirming or excluding intra-articular pathology that may alter management. We also emphasize that a practitioner may appropriately choose to supersede a panel recommendation based on their expertise and experience in a specific clinical scenario. Consultation with the referring physician, who may have a strong preference for or against dMRA in a particular patient or clinical scenario, is important.

Materials and methods

The white paper authorship panel first convened on March 14, 2022, at the Society of Skeletal Radiology Annual Scientific Meeting in San Diego, CA, USA, with an additional 7 virtual or blended in-person/virtual meetings over the following 13 months. The manuscript outline was first delineated, and six subgroups established, each pairing a senior and more junior panel member. Subgroups were assigned section(s), including one major joint (shoulder, elbow, wrist, hip, knee, or ankle/foot), and tasked with completing an exhaustive literature review, preliminary draft, and presentation of evidence to the authorship panel for discussion, deliberation and development of panel recommendations by consensus agreement. In an effort to substantiate panel member support of individual recommendations, an anonymous poll was conducted prior to the final meeting and recommendations that were not unanimous were further deliberated and a final panel recommendation established. Panel recommendations are provided as one of three categories (dMRA recommended, dMRA or cMRI recommended, cMRI recommended), with comments added for clarification as needed. All recommendations were made by consensus. For transparency, the level of agreement is classified as unanimous (12/12), supermajority (≧ 10/12) or majority, in order to reflect dissenting opinions on the consensus recommendation [11].

Subgroups completed their primary MEDLINE/PubMed searches between April 30, 2022, and June 28, 2022, using the terms, “MR” and “arthrography” and the specific joint(s) they were assigned. All abstracts were reviewed. Scientific articles, review articles, and systematic reviews that included MR arthrography and were available online or through the local institution’s interlibrary loan program were obtained. Articles were reviewed and any additional references from these articles that were not originally captured in the initial search were also obtained and reviewed. In addition, limited but dedicated searches were performed to ensure a comprehensive review of available literature. Articles evaluating the utility and/or diagnostic performance of dMRA and those directly comparing cMRI and dMRA were favored, particularly those with surgical confirmation.

The following is a summary of the literature and recommendations from the panel, organized by (1) technical considerations in the performance of dMRA; (2) pathology-specific, joint independent indications; and (3) joint-specific considerations in the shoulder, elbow, wrist, hip, knee, and ankle/foot regions. We conclude by discussing controversies and gaps in the literature. Of note, the focus of this white paper is on dMRA and not the other forms of arthrography, including indirect MRA and CT arthrography.

Technical considerations

Image guidance

Image-guidance for the arthrogram component of the exam is recommended for higher accuracy compared to blind injections. Fluoroscopy and ultrasound (US) guidance are the most commonly used modalities [12], but computed tomography (CT) and MR imaging may be useful when no other imaging modalities are available [13,14,15]. When radiation is used, the “as low as reasonably achievable” (ALARA) principle applies to reduce dose to both the patient and operator. Operators must be properly trained and routinely use all available dose reduction techniques [16, 17]. Ultrasound-guided arthrographic injection eliminates the exposure to ionizing radiation which may be especially pertinent in teenagers and young adults [18], but a drawback is the limited ability to document flow into neighboring compartments. Joint-specific considerations are discussed in the respective sections below.

General procedure

As with all procedures, an initial assessment for potential contraindications should be performed. The main contraindications for dMRA are suspected peri-articular or joint proper infections, reflex sympathetic dystrophy, severe coagulopathy, and allergic reaction to any of the injected components. Positioning of the patient depends on the approach. In our experience, taking the necessary time to optimize positioning to ensure ease of joint access and patient comfort is critical. After the skin start point has been selected and marked, the area should be sterilized with a cleaning solution and then draped [19]. Local anesthesia may be provided to the skin and underlying subcutaneous tissues using 1% lidocaine, with sodium bicarbonate buffer at the preference of the proceduralist [19]. Twenty to 25-gauge needles are used to access the joint, with the length depending on the particular joint and approach [12, 20]. If present, pre-existing joint fluid could be aspirated to prevent dilution of the injectate. Introduction of gas should be avoided by limiting the number of syringe exchanges, clearing bubbles from any extension tubing, and filling the needle hub with injectate prior to making connections. After the joint is accessed, pure iodinated contrast may be injected in small increments (e.g., 0.2 mL) via extension tubing while obtaining fluoroscopy spot views to ensure proper intra-articular flow. If resistance is met, rotating the needle may aid in penetration through the joint capsule [19]. In the case of an US-guided injection, if fluid accumulates around the tip of the needle, the needle is likely periarticular and needs to be adjusted accordingly. After confirmation of intra-articular needle positioning, a mixture of diluted gadolinium-based contrast agent, normal saline, iodinated contrast, and/or anesthetic may be injected. Too little injected volume will result in inadequate expansion of the joint, whereas too large of a volume will create iatrogenic leakage from the joint space that may be interpreted as a local tear or a lesion [21]. For the appropriate injection volume for each joint, the reader may refer to the respective sections in this article and Table 1. The needle is subsequently removed, and a bandage should be placed over the insertion point. Image acquisition should be performed as soon as possible after the injection to maximize capsular distention and minimize absorption of contrast [22, 23], ideally within 30 min. Clear instructions to the patient regarding remaining still during MRI acquisition is crucial to prevent deleterious motion-related image degradation.

Table 1 Recommended injection volumes for the major joints

Full size table

Injectate and safety

Most practitioners prefer to include gadolinium-based contrast agents (GBCAs) for dMRA, but some use a saline-only technique [24], where reduced contrast reaction risk and potential cost savings could be realized. Some authors have made head-to-head comparisons in the shoulder and found equivalent performance for evaluation of glenoid labral and rotator cuff tears, as well as in the detection of acetabular labral tears and cartilage lesions in the hip [25,26,27].

The intra-articular injection of GBCAs is not approved by the United States Food and Drug Administration (FDA) and is performed as an off-label indication. The small volume administered for dMRA in clinical use is widely considered safe [28], but data from some in vitro studies have suggested that chondrocytes may be adversely affected [29, 30], while others have found no adverse effects [31]. The gadolinium ion does not dissociate from the contrast agent [32], so any ill effects may be attributed to the intact gadolinium chelate [29]. In recent years there has been concern about the deposition of gadolinium into the organs of patients receiving intravenous gadolinium. In one pre-clinical study using rats, detectable levels of gadolinium were present in joint tissues, bone marrow, and/or kidneys following intra-articular injection of both linear and macrocyclic GBCAs, though the clinical significance of this remains unknown [33]. Intra-cranial gadolinium deposition has not been shown after intra-articular administration of GBCAs at clinical doses in either pre-clinical models [33] or on patient brain MRI exams [34, 35].

GBCAs can be diluted over a wide concentration range (0.7–3.4 mmol/L) and still yield acceptable results for dMRA [36], but a concentration range between 1.25 and 2.5 mmol/L is considered ideal for optimum signal-to-noise ratio [37]. To achieve this dilution, one must be aware of the concentration of their particular GBCA, which is summarized in Table 2. The most common commercially available concentration is 0.5 mol/L but may range between 0.25 and 1 mol/L [38]. The addition of iodinated contrast to the mixture allows continuous confirmation of appropriate needle position during the procedure if fluoroscopic guidance is utilized and, if in sufficient amounts (e.g., 25–50% of the total injectate), allows for conversion of dMRA to CT arthrography if necessary [39, 40].

Table 2 Concentrations of commercially available gadolinium-based contrast agents

Full size table

It should be noted that the addition of iodinated contrast results in T1 and T2 signal shortening, and optimum concentrations of gadolinium are lower as a result [36, 37]. This effect is exaggerated at 3 T because the peak signal-to-noise ratios for iodinated contrast dilutions are slightly lower than at 1.5 T [36, 37, 41]. Intra-articular administration of iodinated-based contrast agents is FDA-approved and considered safe, but in vitro studies have suggested chondrotoxicity [30] and transient increases in cartilage stiffness, which may result in an increased risk of tissue or cell damage during weightbearing [42].

Some practitioners include anesthetics in the injectate to provide comfort to the patient during image acquisition [43, 44] or for diagnostic purposes [45], but a growing body of pre-clinical literature points to chondrotoxicity of local anesthetics [46,47,48,49]. Limiting intra-articular delivery with use of less chondrotoxic anesthetics (e.g., ropivacaine) may be warranted [46].

Complications

Although image guided injection of the joints is generally reported to be safe and tolerable by patients [28, 50], a significant number of patients (up to 66%) may experience delayed onset pain in the hours to days following a dMRA [51,52,53]. Patients under the age of 30 have been reported to experience more pronounced pain, though this can be expected to resolve within a week of the injection [54]. Anaphylactoid reactions may occur after arthrography, including hives (0.4%), with severe anaphylaxis being exceedingly rare (0.003%) [53]. The incidence of joint infection is reported to be 0.003% [55]. Although vasovagal reactions have been reported at a very low rate with dMRA (0.015%), our collective experience is that the frequency is closer to that reported for arthrography in general (1.4%) [53]. Neurovascular complications are exceedingly rare [53], but could occur depending on the chosen needle path. A recent systematic review and meta-analysis has shown that it is safe to perform joint injections in patients on warfarin without routine testing of the international normalized ratio (INR) [56]. Another recent prospective study including 5080 musculoskeletal procedures found no clinically significant bleeding events with warfarin use, continuation of direct oral anticoagulants, or with concomitant antiplatelet or combination antiplatelet therapy [57]. All of these potential risks are important to discuss with patients and document in the consent process.

Imaging protocols

dMRA has been successfully performed at a variety of field strengths, most commonly 1.5 T. As with cMRI, dedicated extremity coils should be used when available and imaging parameters should be optimized [58, 59]. As with all post-contrast MR imaging at any field strength, short tau inversion recovery is ideally avoided as gadolinium signal may be inadvertently nulled.

The most common sequences that are used with dMRA are T1-weighted fat-suppressed fast or turbo spin echo (FSE/TSE) in all three anatomic planes [3, 60,61,62,63,64,65,66,67,68,69]. To better evaluate the bone marrow and soft tissue fat planes, some choose not to fat-suppress one of the planes, while others may add an additional non-fat-suppressed T1-weighted sequence. A fluid-sensitive sequence, such as intermediate-weighted or T2-weighted FSE/TSE with fat-suppression in at least one plane, is also typically included to aid in the evaluation of marrow signal abnormalities and extra-articular structures [54, 60,61,62,63, 66,67,68,69,70,71,72,73,74,75,76,77,78]. Others recommend 3D gradient (spoiled or steady state) or FSE/TSE sequences with thinner slices, which allow for multi-planar reformats [66, 79,80,81,82,83,84,85]. Particularly in the shoulder, some authors have explored the ability to exchange 3D sequences for conventional 2D sequences [86,87,88,89,90,91,92,93,94,95,96,97]. A recent meta-analysis showed 3D dMRA had similar pooled sensitivity and specificity to 2D dMRA for diagnosing rotator cuff tears and labral lesions, however 3D FSE/TSE sequences demonstrated higher sensitivity than 3D gradient sequences [98].

Metal artifact reduction techniques may be applied with any pulse sequence, including high receiver bandwidth, view angle tilting, and multi-spectral imaging techniques [99,100,101]. When spectral fat-suppression fails due to non-isocentric positioning or metallic implants, T1-weighted pulse sequences may be obtained without fat-suppression or Dixon techniques may be used.

“When the dMRA goes wrong”

Unexpected issues may arise with improper arthrographic technique. Inadvertent injection of GBCAs outside of the optimal concentration range (e.g., 0.7–3.4 mmol/L [36]) will result in lower signal of the injectate. For concentrations that are too low, standard fluid-sensitive sequences from a cMRI protocol could be used to salvage the exam. For concentrations that are too high, a “black” contrast effect can be seen on the MR images as a result of T2 shortening [102,103,104]. There is no danger to the patient, and for the larger joints, repeat MR imaging after a few hours can potentially salvage the exam in some instances [102] (Fig. 2). The success of delayed imaging in these scenarios depends on the balance between trans-synovial diffusion of gadolinium and loss of joint distention [23, 105].

The unintentional introduction of gas may potentially simulate intra-articular bodies. These bubbles can typically be recognized as they migrate to the nondependent regions of the joint and often demonstrate a characteristic dipole field pattern artifact, with adjacent signal pile-up in one direction and signal loss in the other [61] (Fig. 3). In select cases where pathology is simulated, the patient can be reimaged in a prone position to mobilize the foci of gas to another nondependent location.

There are instances where the injectate may be in an unexpected location. For this reason, it is prudent for the radiologist to review the initial images (i.e., after the localizers or first series is completed), or if possible, request a notification from the technologist for any unusual patterns of contrast distribution. For cases where the injectate is entirely outside of the joint, such as into a bursa (e.g., subacromial-subdeltoid bursa in the shoulder or iliopsoas bursa in the hip), extra-synovial fat pad, or in a recess that happens to be separated from the joint (e.g., cases of complete suprapatellar plica), completing the MRI exam using a routine non-contrast protocol would be advised. It is optimal to empathetically disclose this ‘maloccurrence’ to the patient at the time of imaging. If the patient agrees and if time allows, a repeat injection could be performed prior to completing the MR examination. Alternatively, the patient can be requested to return for a repeat procedure.

Rarely the dMRA exam may not be completed due to MRI scanner failure, or a scan may be severely compromised or even aborted due to patient motion and/or intolerance. In these instances, conversion to CT arthrography may be considered if iodinated contrast was included in the injectate. In one study where CT arthrography of the knee was performed an average of 100 min following a dMRA injection (a scenario that mimics the delay that may be experienced with an aborted exam), accuracy for meniscal tears remained high [62].

Pathology-specific, joint independent indications

Chondral and osteochondral abnormalities

Direct MR arthrography may be useful to diagnose and stage chondral and osteochondral injuries [28]. The differentiation between a stable and unstable nondisplaced in-situ fragment is important for surgical decision-making. Overlapping imaging appearances exist between stable and unstable in situ osteochondral fragments, but fragment instability can be confidently diagnosed when fluid (either native or introduced via dMRA) is seen extending along the entirety of the interface between the fragment and underlying tissue [28] (Figs. 4 and 5).

There is a paucity of head-to-head comparison studies between cMRI and dMRA in the shoulder, elbow, wrist, and ankle regions. In the shoulder, dMRA has shown moderate diagnostic performance, due to reduced sensitivity, for detecting glenohumeral cartilage lesions with fair to moderate interobserver agreement [106,107,108]. In the ankle, reported accuracies for diagnosis and grading of osteochondral lesions of the talus appear comparable between cMRI and dMRA [109,110,111,112].

In the elbow, published algorithms for evaluating osteochondral lesions in adolescent athletes include radiographs and cMRI, particularly for establishing lesion stability and to guide surgical management [113, 114]. Despite the lack of studies directly comparing cMRI and dMRA, several studies have demonstrated a high sensitivity of cMRI for the detection of unstable osteochondral lesions, relying primarily on the visualization of a high, fluid-signal intensity rim undermining the interface between the lesion and underlying bone and a full-thickness articular cartilage defect [115, 116]. One study comparing cMRI to a gold standard of arthroscopy in 52 osteochondral lesions found 100% sensitivity and 80% specificity for unstable lesions and close correlation in 94% of cases with the International Cartilage Repair Society classification for lesion instability [117]. In contrast, a recent meta-analysis concluded that cMRI criteria for instability are adequate for assessing adult osteochondral lesions of the elbow, knee, and ankle, but performed less well for predicting stability in pediatric osteochondral lesions [118]. This meta-analysis included 3 studies using cMRI for the assessment of pediatric elbow osteochondral lesions. The collective sensitivity, specificity, and accuracy from these studies can be calculated as 94%, 66%, and 84% respectively.

At the hip, dMRA performs more favorably compared with cMRI. cMRI of the hip is known to have limited sensitivity but high specificity in the detection of chondral abnormalities [119]. A meta-analysis of the accuracy of MR imaging in the detection of chondral lesions in the setting of femoroacetabular impingement (FAI) showed sensitivity, specificity, and accuracy of 76%, 72%, and 75% respectively for cMRI, compared to 75%, 79%, and 83% for dMRA [120]. This analysis included studies which utilized magnetic field strengths ranging from 1 to 3 T [120]. However, a subsequent study suggested that 3T cMRI may be superior to 1.5 T dMRA for the detection of cartilage lesions, but the differences were small and should be interpreted with caution [121]. In that same study, when specifically assessing cartilage delamination, 3 T MRI was equivalent to 1.5 T MR arthrography [121]. Another study compared 3T cMRI to dMRA in the detection of acetabular chondral defects in the same patient population and suggested greater sensitivity with dMRA (sensitivity/specificity of the two readers: 65%/100% and 59%/100% for cMRI compared to 81%/91% and 71%/82% for dMRA), although the differences were not significant [122]. A study which specifically addressed chondral delamination in the setting of FAI found a sensitivity of 6%, specificity of 98%, NPV 27% and PPV 91% with 1.5 T dMRA [123]. Attempts have been made to improve the conspicuity of hip cartilage lesions, including delamination, using leg traction at the time of dMRA [124]. Although preliminary experience suggested that traction does indeed aid in detecting surface cartilage lesions and cartilage delamination, the literature has shown mixed results [123, 125, 126]. In addition, the time involved and fear of potential patient discomfort have dissuaded most sites from employing hip traction at the time of MR imaging.

At the knee, dMRA has performed favorably compared with cMRI. In one meta-analysis, dMRA was found to be superior to conventional MRI for detection of patellofemoral chondral lesions [127]. For studies that have compared dMRA to cMRI for cartilage abnormalities in the knee, sensitivities have ranged from 69 to 93% and specificities have ranged from 98 to 100% for dMRA, while sensitivities have ranged from 25 to 81% and specificities have ranged from 50 to 99% for cMRI [65, 84, 128,129,130,131]. It is noteworthy that both dMRA and cMRI are less accurate for grade I compared with grade IV cartilage abnormalities [65, 128, 130], but delayed image acquisition by a few hours may improve the performance of dMRA for grade I lesions [22, 84, 132]. For osteochondral lesions in 25 knee MRI exams, one study found that correct staging could be performed in 100% of cases on dMRA and 57% of cases on cMRI [133].

Recommendation: dMRA or cMRI recommended. The use of dMRA may be valuable when symptoms are discrepant with cMRI results (consensus recommendation, unanimous).

Post-operative evaluation of chondral and osteochondral abnormalities

A variety of chondral and osteochondral restoration techniques exist, including marrow stimulation, osteochondral transplantation (autologous or allogeneic), autologous chondrocyte implantation, and allogeneic particulate cartilage fragment implantation [134]. Direct MRA may be useful to delineate defects at the cartilage interface following all types of repair or graft/host bone junction following osteochondral repairs [135,136,137], but the protocol should include fluid sensitive and non-fat-suppressed sequences to assess the subchondral marrow and trabeculae. Authors have used dMRA to evaluate patients after matrix-induced autologous chondrocyte implantation in mid- and long-term follow-up studies [63, 74], but one study found that the dMRA findings correlated poorly with clinical outcomes [74].

Other authors have reported effective evaluation of chondral repair tissue without the use of dMRA [138, 139] and, in fact, cMRI is most widely used to assess the features deemed the most important after chondral repair or osteochondral transplantation [140,141,142].

Recommendation: dMRA or cMRI recommended (consensus recommendation, supermajority (10/12)).

Intra-articular bodies

Intraarticular bodies may present without any history of prior injury and may cause limited range of motion, pain, catching or locking [143, 144]. Imaging can confirm the diagnosis and provide useful surgical planning information [145, 146]. When ossified, intraarticular bodies can be visualized by radiography, however radiography is frequently inadequate [144, 147, 148]. cMRI often detects an intraarticular body, especially with a joint effusion; however, dMRA may prove useful in detection of small bodies [28, 60, 143, 145, 149].

There is a paucity of head-to-head comparison studies between cMRI and dMRA in the shoulder, elbow, wrist, and ankle regions. In the hip, one study showed high diagnostic accuracy for the detection of intra-articular osteochondral bodies using dMRA, both without and with traction [150]. Using osseous and cartilaginous bodies placed inside 16 cadaveric knees, one study showed 92% accuracy for detection using dMRA compared with 57–70% using cMRI [151].

Recommendation: cMRI recommended (consensus recommendation, supermajority (11/12)).

Joint-specific indications

Shoulder

Technical considerations

The recommended target for fluoroscopic glenohumeral joint injection varies, with many advocating the medial upper third of the humeral head and others advocating the middle or lower third [152,153,154,155]. One study reported an 85% first attempt success rate utilizing the rotator interval approach for resident trainees and 100% success rate with reduced fluoroscopy time for musculoskeletal radiologists [153]. Additional techniques using a posterior approach have been proposed to avoid traversing the anterior stabilizing structures and confounding imaging findings related to inadvertent extra-articular contrast leakage [156, 157]. Comparison of US-guided rotator interval and posterior approach injections found that both techniques are successful and well tolerated by patients, but the posterior technique resulted in decreased extra-articular leakage rate [158, 159].

The volume of injectate used for direct shoulder MR arthrography varies in the literature, ranging from 8 to 15 ml, with some authors titrating to perceived resistance to the injection while monitoring distention with imaging [152, 155]. A patient with chronic capsular laxity may have a higher injection capacity, whereas a patient with adhesive capsulitis may tolerate a smaller volume injectate [160,161,162,163]. In work with cadavers, a volume of 15 mL of intraarticular fluid has been described as optimal for dMRA, but little data exists comparing adequate joint distention with rates of contrast leakage [164]. The authors recommend a minimum of 8 ml of injectate. After shoulder arthrography, neither internal derangements nor history of prior surgery had an apparent effect on the post-injection pain course, and post-injection exercise prior to MRI does not improve image quality or the depiction of rotator cuff or labral tears [165, 166].

Provocative maneuvers

In specific clinical instances, an additional fat suppressed T1-weighted abduction external rotation (ABER) sequence may be performed to obtain an oblique axial plane with regard to the glenohumeral joint [167]. The ABER position changes the capsular dynamics by creating traction on the anterior band of the inferior glenohumeral ligament and axillary pouch structures, which may increase the conspicuity of anteroinferior and posterosuperior labral lesions [167,168,169] (Fig. 6). The total additional time for patient re-positioning and scanning ranges from approximately 5–15 min for the ABER sequence [170,171,172]. Patients may be apprehensive to assume the ABER position due to shoulder pain or concern for dislocation resulting in increased motion artifact, but one study reports a 95% success rate through patient education combined with lidocaine in the intra-articular injectate [167, 172, 173]. However, some centers elect to forgo the ABER sequence due to practical concerns about throughput as well as consistency and reproducibility across sites. A recent survey of European Society of Musculoskeletal Radiology (ESSR) members reported the ABER sequence is used in about half of cases, with no differences between general (60%) and orthopedic (65%) hospitals [174].

The ABER sequence has comparable sensitivity and specificity to conventional dMRA for anteroinferior labral lesions [170, 171, 175, 176]. Several studies have shown no significant differences in interobserver agreement between the neutral and ABER position [170, 171]. A recent meta-analysis showed increased diagnostic accuracy of routine axial plus ABER dMRA (pooled sensitivity 95.7%, pooled specificity 94.5%) compared with dMRA without ABER (pooled sensitivity 81.5%, pooled specificity 88.8%). However, results are interpreted with caution since the 95% confidence intervals overlapped, there was a high degree of heterogeneity among the studies, and there was publication bias for axial plus ABER dMRA studies [172]. Furthermore, of the 9 articles included in the meta-analysis, 3 articles were at least 20 years old, and 2 additional articles included cases that were over 20 years old. All considered, the superiority of ABER over routine axial positioning, when scanning using modern imaging technology would benefit from further study, and ideally would consider relative time on the MRI scanner/cost considerations. When evaluating only studies performed at 3 T, another meta-analysis showed improved sensitivity for the diagnosis of anterior and posterior labral lesions with dMRA over cMRI, including improved sensitivity but reduced specificity with ABER positioning for anterior labral tears [177] . In particular, the ABER sequence may add value in the detection of the Perthes variant of anteroinferior labral lesions and after Bankart repair [171, 178, 179]. However, ABER sequences have been reported to have decreased accuracy in the characterization of anterior labroligamentous periosteal sleeve avulsion (ALPSA) lesions [171, 180]. For SLAP lesions, there is no definitive data to support increased accuracy with inclusion of the ABER sequence [181, 182], though it has been suggested to be useful in characterizing posterosuperior labral peel back lesions in throwing athletes [183]. Assessment of combined redundancy signs during ABER positioning has been advocated to differentiate patients with atraumatic multi-directional instability (MDI) from clinically stable shoulders with 81-90% sensitivity and 94% specificity [184].

Additional specialized provocative maneuvers have been described, but thus far, have not been adopted at most centers for routine clinical use. Weighted traction and FADIR (flexion, adduction, internal rotation) positioning may increase the conspicuity of SLAP tears and posteroinferior labral lesions respectively [185, 186]. ADIR (adduction internal rotation) positioning has been reported to improve characterization of ALPSA lesions, but data is limited [180].

Recommendation: Provocative maneuvers are not recommended in routine practice. Selective application of imaging in the ABER position may be advantageous in some patients with anterior instability. This decision is ideally made when an anteroinferior labral abnormality is not identified on review of the routine axial images in a patient with known anterior instability (consensus recommendation, majority (9/12)).

Clinical indications

Instability

Anterior instability is one of the most common indications for dMRA [174, 187]. One study demonstrated 88% sensitivity, 91% specificity, and 89% accuracy with regard to dMRA and detection of anteroinferior labral tears [188] (Fig. 7). More recently, pre-operative dMRA has been reported as a predictor of recurrent instability based on calculation of “off-track bone loss” [189].

Using arthroscopy as the reference standard, dMRA showed a sensitivity of 82% and specificity of 98% in the identification of superior labrum anterior and posterior (SLAP) tears [190] (Fig. 8). Multiple meta-analyses have shown that dMRA is more accurate than cMRI for the diagnosis of SLAP tears [181, 182, 191, 192], with higher diagnostic accuracy at 3 T over 1.5 T with or without intra-articular contrast material [181]. One meta-analysis evaluating 3 T studies showed that dMRA was similar in sensitivity to cMRI (0.84 vs 0.83, p=.575), but less specific (0.99 vs 0.92 p < 0.0001) for SLAP lesions [177].

Direct MR arthrography has been proposed as a method to aid the complex clinical diagnosis of multidirectional instability (MDI) (Fig. 9). Prior studies have shown an increased capsular volume on dMRA in MDI patients [160, 193]. In particular, greater inferior and posteroinferior axillary recess depths have been accurate and reproducible in differentiating MDI from control groups [161, 193,194,195]. The rotator interval width may be greater in patients with other forms of instability but remains debated in regard to diagnostic utility for MDI [160, 161, 184, 193, 195,196,197].

Recommendation: dMRA or cMRI recommended. Direct MRA has a compelling role in the assessment of younger individuals with suspected instability, when subtle labroligamentous abnormalities may have profound influences on shoulder function, management, and prognosis (consensus recommendation, unanimous).

Humeral avulsion of the glenohumeral ligament (HAGL)

Direct MRA may be helpful in diagnosing inferior glenohumeral ligament (IGHL) injuries. Extra-articular contrast material leakage and the J-sign are described imaging signs [198, 199] (Fig. 10). However, iatrogenic extra-articular contrast leakage at dMRA is not uncommon and should be interpreted with caution [200, 201]. A recent study examined several dMRA features that can be used to distinguish iatrogenic extra-articular contrast leakage and true IGHL tears [202]. There are no studies directly comparing the diagnostic accuracy of cMRI and dMRA for HAGL injury.

Recommendation: dMRA or cMRI recommended (consensus recommendation, unanimous).

Thrower’s shoulder

High-performance athletes may represent a subgroup of patients for whom initial dMRA is indicated as it potentially yields more diagnostic information over cMRI, considering the greater sensitivity for partial-thickness articular-sided rotator cuff tears and posterosuperior labral pathology [203,204,205] (Fig. 11). In throwing athletes, dMRA may also assist in characterizing biceps pulley lesions resulting in anterosuperior impingement as well as internal impingement and glenohumeral internal rotation deficit (GIRD) [206, 207]. It should be noted that these cited studies were mostly published prior to 2007 with imaging performed at 1.5 T.

Recommendation: dMRA or cMRI recommended (consensus recommendation, unanimous).

Recurrent instability and the post-operative labrum

Shoulder imaging following surgery for instability can present as a challenge secondary to scar tissue, and these patients are likely to benefit from dMRA (Fig. 12). Using arthroscopy as the gold standard, one author found that sensitivity for labral and supraspinatus tears after repair ranged from 71 to 84% with cMRI and increased to 100% with dMRA at 3 T [40]. In another study using dMRA at 1.5 T, the accuracy for labral tears (anterior, posterior and SLAP tears) was 92% and comparable to dMRA of the native shoulder [208]. Another study that compared dMRA (neutral and ABER) at 1T to second look arthroscopy correctly confirmed the structural integrity of the repaired glenoid labrum with an accuracy of 95% [179].

Recommendation: dMRA recommended (consensus recommendation, unanimous).

Rotator cuff tear and re-tear

Several meta-analyses and systematic reviews have concluded dMRA as the most sensitive imaging method for detection of rotator cuff tears when including all field strengths [209,210,211,212], but the diagnostic advantage is less pronounced for full-thickness tears [209,210,211]. dMRA at 3 T shows similar sensitivity to dMRA at 1.5 T as well as similar specificity to 3 T cMRI for the diagnosis of full-thickness tears, but improved sensitivity for partial-thickness tears [209, 213]. As expected, dMRA performs slightly better than cMRI for articular sided partial-thickness tears (sensitivity 74% and specificity 90% for dMRA versus sensitivity 67% and specificity 82% for cMRI) but shows similar sensitivity (0.75 dMRA versus 0.73 for cMRI) for detection of bursal-sided tears [209, 214]. However, it should be noted that both of these studies specifically addressing bursal surface tear detection included indirect MRA in the subgroup analysis for partial-thickness tears [209, 214].

Regarding subscapularis tears, there is limited data on the best diagnostic method. Both dMRA and cMRI have shown relatively lower accuracy in the diagnosis of subscapularis tears, due to decreased sensitivity, particularly for partial thickness tears [209, 213, 215,216,217]. Studies have suggested that dMRA may be marginally better for subscapularis tears at 3 T [213, 216]. 2D and 3D dMRA at 3 T appear to be equivalent for all types of rotator cuff tears [98, 213]. Regarding recurrent rotator cuff tears, there are relatively few studies assessing the diagnostic accuracy of dMRA when surgical correlation is used as the reference standard [218,219,220]. One study reported high sensitivity (88%) and specificity (90%) for full-thickness tears, but overall moderate sensitivity (72%) and specificity (77%) for partial-thickness tears and only fair interobserver agreement for subscapularis tears (k = 0.20) [218]. A recent meta-analysis had insufficient data to separate dMRA from cMRI in subgroup analyses to determine the most accurate imaging method [220].

The ABER sequence has been reported to have similar sensitivity and specificity for full-thickness superior rotator cuff tears with good interobserver reproducibility [170, 221] but may improve the sensitivity for detection and characterization of partial-thickness articular surface tears [169, 222,223,224].

Recommendation: cMRI recommended as the differences in sensitivity and specificity between dMRA and cMRI for the diagnosis of rotator cuff tear and retear are small (consensus recommendation, supermajority (11/12)).

Adhesive capsulitis (AC)

Several MR imaging signs of AC have been reported [225]. The majority of studies examining the utility of dMRA for the diagnosis of AC compare clinically symptomatic patients to asymptomatic control subjects or those with other shoulder pathology [163, 226,227,228,229]. Few studies report surgical or arthroscopic confirmation in at least a portion of patients [71, 230, 231]. Studies comparing patients with a clinical diagnosis of AC to controls have found thickening of the rotator interval joint capsule and coracohumeral ligament as suggestive for AC [71, 227]. Reliance on visualization of axillary pouch abnormalities on dMRA alone is unclear as there is conflicting data with respect to the utility of axillary recess capsular thickening and diminished filling capacity [71, 226,227,228,229, 231,232,233]. Complete obliteration of the subcoracoid fat triangle has poor sensitivity but high specificity for adhesive capsulitis; however, coracohumeral ligament and IGHL thickening have a higher correlation with range of motion impairment [71, 227, 232]. The utility of measuring the rotator interval dimensions on dMRA is also limited with conflicting evidence [163, 230, 232]. On 3D volumetric assessment, decreased joint capacity has also been reported with greater frequency in AC patients versus control subjects [163]. One systematic review and meta-analysis recommended cMRI over dMRA in AC as the sensitivity and specificity of IGHL thickening on cMRI and dMRA were not significantly different from one another [234].

Recommendation: cMRI recommended (consensus recommendation, unanimous).

Long head biceps tendon (LHBT) and pulley lesions

Similar to cMRI, dMRA is insensitive for LHBT tendinopathy, particularly when compared with histopathology [235,236,237,238,239,240]. When images are evaluated in two planes, dMRA is reasonably accurate for diagnosing biceps rupture [240]. Comparison of 1.5 T cMRI and dMRA in 199 patients who underwent arthroscopy, the authors found no significant difference between the two methods for the detection of intra-articular LHBT tendinosis and tears [238]. At 3 T, dMRA was insensitive for biceps partial-thickness tears and performed similar to CT arthrography [239]. Diagnosing LHBT instability on static images can be challenging in the absence of frank tendon dislocation. Individual imaging findings have a limited role in diagnosing LHBT instability, but the accuracy of dMRA can be improved by combining imaging findings and assessing the integrity of the biceps pulley structures [235, 241,242,243,244,245]. In cadavers, dMRA was superior to cMRI for evaluation of the rotator interval structures [246]. For considerations specific to throwing athletes, please refer to the “Thrower’s shoulder” section.

Recommendation: cMRI recommended (consensus recommendation, supermajority (10/12)).