Magnetic Resonance Imaging of Shoulder Arthropathies

The role of magnetic resonance imaging in evaluating shoulder arthropathies is evolving. This article reviews 4 of the major arthropathies: septic arthritis, rheumatoid arthritis, calcium pyrophosphate dihydrate (CPPD) deposition disease, and hydroxyapatite disease (HAD), with special attention to their magnetic resonance imaging features. Comfort with identifying these entities allows appropriate and prompt treatment, which is critical for joint preservation in the case of infection, for maximal therapeutic efficacy of disease-modifying drugs in the case of rheumatoid arthritis, and for expediting symptomatic relief in the cases of CPPD deposition disease and HAD.

Magnetic resonance (MR) imaging is well established for assessment of the shoulder. The integrity and morphology of the rotator cuff tendons, musculature, and capsular structures, and any associated findings, help the orthopedist to determine the appropriateness and type of any surgical intervention that may be necessary after injury. In cases of osseous or soft tissue neoplasm, the sensitivity of MR imaging for marrow changes and soft tissue contrast can direct biopsy planning and help determine appropriate surgical resection margins. The role of MR imaging in the assessment of various shoulder arthropathies is less defined and is evolving. In this review, the clinical and MR imaging features of 4 of the major shoulder arthropathies are addressed: septic arthritis, rheumatoid arthritis (RA), calcium pyrophosphate dihydrate (CPPD) deposition disease–related arthropathy, and hydroxyapatite disease (HAD).

Septic arthritis remains primarily a clinical diagnosis based on symptoms, physical examination findings, and laboratory analysis of joint fluid aspirate. A missed or delayed diagnosis can lead to articular destruction, concomitant osteomyelitis, and eventually sepsis. Conventional radiographic manifestations of disease occur late and once they are present, often only salvage procedures can be performed. The much greater sensitivity of MR imaging for early joint infection, although not a replacement for fluid sampling or operative management, helps narrow the clinical differential diagnosis, determine the extent of infection, and assess for associated osteomyelitis and soft tissue or intraosseous abscesses. Findings also provide a baseline for assessment of treatment response.

This sensitivity of MR imaging for even subtle levels of inflammatory activity is also of benefit in the case of RA. The classic features of periarticular osteoporosis, soft tissue swelling, osseous erosions, and loss of joint space are well assessed on radiography, which rightfully remains a fundamental part of disease surveillance. However, as RA begins with inflammatory changes at the synovial level, MR imaging offers a unique opportunity to identify disease at its earliest stages. Similar to the situation with septic arthritis, it has become a clinical necessity to assess for changes before there is articular destruction. Enhancement patterns of synovium and associated changes in bone marrow can distinguish between different levels of disease activity. Proper timing in the initiation of newer disease-modifying drugs is critical to obtain a good therapeutic response, and the ability to identify RA stage and activity is therefore invaluable. Perhaps even more important is to assess, by functional musculoskeletal MR imaging, the effectiveness of these expensive and often toxic medications.

CPPD disease and HAD are usually readily identified on conventional radiography, as the soft tissue calcifications typical of each entity are the cardinal finding. With the increased use of advanced imaging for other indications, however, these entities may be encountered on MR imaging and can be confused with other arthropathies such as RA or primary osteoarthrosis. In the case of HAD especially, reactive bursitis and regional inflammation related to resorbing intratendinous calcifications can be confused for infection or rotator cuff tear. Although such confusion can often be resolved by correlation with radiography, films are not always available at the time of interpretation.

Septic arthritis

Background, Demographics, and Presentation

The shoulder is a relatively uncommon site of septic arthritis. In one retrospective review, Leslie and colleagues reported on 18 cases over 18 years based on all records from 2 large North American hospitals. Another earlier study by Lossos and colleagues identified 11 cases from 6 major hospitals in Israel over a 9-year period. More recent reviews from both Europe and the United States have reported similarly low incidence among their patient populations, with rates such as 21, 17, and 23 cases over 8-, 11-, and 15-year periods, respectively. As the shoulder is the second largest joint in the body, with commensurately large synovium, it seems peculiar that infection is infrequent. The incidence is higher if infection in the setting of joint arthroplasty is included. For instance, during a 5-year prospective study of consecutive patients with or without arthroplasty and with a diagnosis of septic glenohumeral arthritis, Kirchhoff and colleagues enrolled 43 patients.

However, the incidence may be increasing because of the increasing longevity of the population, resulting in more individuals with comorbid risk factors for infection being exposed for longer periods. Multiple studies have observed at least one major comorbidity in 80% or more of patients. In general, septic shoulder occurs in older adults, with a mean age of 60 to 65 years.

Septic shoulder is quite uncommon in neonates as well as in healthy adults, but septic arthritis (of any joint) is very rare between later infancy and young adulthood. This process is thought to relate to changes in vascularity within developing bone. Diaphyseal vessels in the newborn traverse the physis, allowing hematogenous agents ready access to the epiphysis and joint. Beginning at approximately 8 to 18 months of age, the diaphyseal vessels instead terminate in sinusoidal lakes situated in the metaphysis, effectively obliterating any hematogenous pathway to the epiphysis. This condition accounts for the metaphyseal predilection of infections such as Brodie abscesses in adolescence. After closure of the growth plate in adulthood, infection can again more easily extend to the epiphysis and joint. Even in the neonatal age group, incidentally, the knee and hip are much more common sites of infection than is the shoulder.

As with any septic arthritis, there are 3 major routes of seeding: hematogenous spread, direct inoculation, and contiguous spread. Hematogenous seeding is the most common route, and may occur with bacteremia of any cause. Direct routes include intraarticular injection, joint surgery, and penetrating trauma; contiguous routes include regional infection such as osteomyelitis, tenosynovitis, soft tissue abscess, cellulitis, and septic bursitis. Of the direct routes, intraarticular injection is most common, typically when performed therapeutically (corticosteroid injection). Because contrast is bacteriostatic, septic arthritis after diagnostic injection for arthrography is exceptionally rare. Postinjection septic arthritis occurs in approximately 1 per 1000 cases according to some reports; other studies report lower rates such as 1 per 3500, all joints included. Postsurgical septic arthritis has been reported in as many as 2% of constrained arthroplasties and fewer than 1% of unconstrained systems. Infection may present anywhere from months to years after the surgery. In the last 2 decades, with improvements in operative technique and infection control, acute postarthroplasty infections have significantly decreased; however, the rate of delayed infections has remained stubbornly constant.

By far the most common causative agent is Staphylococcus aureus , accounting for 40% to 70% in some series. In Kirchhoff’s study, the next most common agents were Staphylococcus epidermidis and Staphylococcus agalactiae . In recent years, an increasing number of S aureus infections have been attributable to methicillin-resistant S aureus (MRSA) strains. In Cleeman and colleagues’ series of 23 cases of glenohumeral infection, for instance, 70% of cases were due to S aureus and of these 17% were MRSA. Gram-negative bacilli ( Escherichia coli , Klebsiella pneumoniae , Proteus mirabilis ) have accounted for up to 20% of infections in some series. Streptococcus ( S pyogenes , S pneumoniae ) and Gonococcus are other potential agents. Specific types of patients have a greater susceptibility to otherwise rare causative agents. In particular, patients with sickle cell anemia are prone to Salmonella infection and immunocompromised patients (AIDS, immunosuppressive therapy) are at increased risk for Mycobacterium tuberculosis . However, in both these groups Staphylococcus is still the most common causative organism.

As already noted, one of the major risk factors for infection is the presence of comorbidities. Preexistent arthritis is a significant risk factor, and the risk of secondary septic arthritis increases with the severity of the primary arthropathy. The arthropathy with the highest secondary risk of infection is rheumatoid, followed by gout and even osteoarthritis. Theoretically, this is related to the hyperemia these arthropathies engender, increasing the risk of bacterial emboli. Diabetes mellitus is another major risk factor for septic arthritis, usually from focal spread. Additional factors include underlying malignancy, cirrhosis, immunosuppression (human immunodeficiency virus/AIDS, chemotherapy, stem cell/solid organ transplant recipients), obesity, chronic obstructive pulmonary disease, alcoholism, hyperuricemia (even in the absence of gout), and intravenous drug use.

On clinical examination, no finding is specific for septic arthritis. The initial presentation of septic arthritis can mimic that of any inflammatory arthropathy. Pain and limited range of motion are common presenting findings. Patients may also refer warmth, erythema, and swelling, but these are less reliable indicators. In a study by Ambacher and colleagues, these 3 findings were observed in only 60% of patients. The presence of fever and malaise is inconsistent. Evident signs of inflammation do not rule in infection, nor does their absence rule it out. Furthermore, preexisting joint abnormalities, especially concurrent arthropathies, can mimic or mask infectious symptoms. Adjacent inflammatory conditions such as bursitis and calcific tendinitis can also mimic septic arthritis. Immune status is also important, as compromised patients may not be able to mount an appropriate clinical response. Also, the precise causative agent must be considered. Tuberculous infection, in particular, most often has a more indolent course with subclinical manifestations for months to years, especially in older patients. Incomplete treatment with antibiotics, all too common in clinical care, can leave smoldering levels of infection while partially masking or blunting corresponding signs and symptoms. The nonspecific and variable presentations of septic arthritis can considerably delay diagnosis and treatment. The relative difficulty in palpating the shoulder joint because of its deep situation further limits assessment. In Leslie and colleagues’ series, diagnosis was delayed by more than 6 months in one-third of cases. In Kirchhoff and colleagues’ study population, only 44% of the patients were diagnosed or treated within 10 days of symptom onset. The remaining 56% were diagnosed or treated at a mean of 57 days from onset. In both of these studies, a longer delay in diagnosis correlated with poorer outcomes.

Laboratory analysis can imperfectly aid in the diagnosis of septic arthritis. Leukocytosis is an unreliable indicator, as it may be absent even in immunocompetent patients who are infected. The erythrocyte sedimentation rate (ESR) may also be increased. Several series have reported that 90% to 100% of subjects have increased C-reactive protein (CRP). However, increased levels are probably not sufficiently specific, as other inflammatory conditions such as RA would still need to be considered. A normal CRP may help to exclude infection.

Imaging Findings

Septic arthritis is ultimately a clinical diagnosis that hinges on appropriate synovial fluid analysis. Direct sampling of joint fluid is the single most important diagnostic step, and can be accomplished either by image-guided needle aspiration or intraoperative collection during irrigation. Nevertheless, imaging can offer significant information for both diagnosing and assessing the extent of infection. Furthermore, the posttreatment course can be monitored, albeit in a delayed fashion, with the aid of serial studies.

Radiographs are neither sensitive nor specific for early septic arthritis, but still offer useful information. Preexistent conditions such as osteoarthrosis and inflammatory arthropathy may be identified. The radiograph also offers a baseline by which the long-term treatment outcome may be monitored. In more advanced stages of infection, radiographs will show narrowing of joint space secondary to cartilage destruction, marginal erosions at the bare areas of the joint, or, rarely, periostitis and bone destruction if there is associated osteomyelitis. Inadequately treated disease may also demonstrate secondary arthrosis and/or bone destruction. Ankylosis, subchondral bone loss with reactive sclerosis, and periarticular calcifications may also be seen, the latter being more common in nonpyogenic infections.

Ultrasonography has a growing role in the evaluation of the septic shoulder. Both effusions and synovial hypertrophy can be well visualized, the latter typically appearing as hypoechoic intraarticular material that lacks compressibility and mobility and often demonstrates flow on Doppler analysis. The dynamic nature of sonography allows the entire glenohumeral joint to be scrutinized, which is useful, because an effusion may distribute unevenly. In one review of 30 glenohumeral joint effusions, fluid was consistently identified by ultrasonography in the posterior joint recess in 100% of the patients and in the biceps tendon sheath in 97%. Ultrasonography may also be used in this way to facilitate aspiration for diagnostic and therapeutic purposes. Accessing the joint without image guidance can be considerably less reliable. In one study, Sethi and colleagues found that only 26% of landmark-guided anterior glenohumeral joint injections by orthopedic surgeons were ultimately successful in reaching the intraarticular space.

Although the role for conventional radiography in evaluating and following septic arthritis should not be understated, MR imaging offers many advantages that make it a useful and important imaging step in assessing for joint infection. The presence and amount of joint effusion can be established, which may help direct aspiration. In the normal glenohumeral joint, almost no fluid should be present. In a review of 20 shoulder MR imaging studies from 12 asymptomatic patients, Recht and colleagues found joint fluid in 14 shoulders, but not exceeding 2 mL in any case. Furthermore, MR imaging allows at least a general grading of fluid volume in the abnormal joint. The following criteria have been proposed: grade 0 reflects scant fluid not distending any joint recesses; grade 1 demonstrates a small amount of fluid in the subscapularis recess, axillary recess (marked by a U-shaped inferior capsule), or biceps tendon sheath on at least 2 coronal-oblique images; grade 2 demonstrates distention of at least 2 of these recesses; and grade 3 demonstrates fluid in all 3 recesses.

The overall extent of soft tissue and osseous involvement in septic arthritis may also be assessed by MR imaging. Conversely, given the exquisite sensitivity for soft tissue and marrow changes that MR provides, the absence of joint effusion and of any other typical early findings for infection can exclude the diagnosis of septic shoulder in a way that radiography cannot.

T2-weighted and/or short-tau inversion recovery (STIR) sequences are particularly useful in identifying early stages of disease. Initial manifestations include glenohumeral effusion and synovitis. Reactive marrow edema is often present as well, within both the humeral head and the apposing portion of the glenoid. Postcontrast fat-suppressed T1-weighted images are helpful in demonstrating thick and/or frondlike rim-enhancing synovium ( Fig. 1 ). The authors recommend that these be done dynamically to quantify the extent of articular hyperemia. On postcontrast images, the acromioclavicular joint can be used as a standard of reference for normal enhancement in the absence of complete communicating rotator cuff tears. A periarticular abscess, if present, may be distinguished by characteristic thick-walled peripheral enhancement.

As infection progresses beyond mere synovitis and effusion, marginal erosions form at the bare areas of the joint and cartilage degradation may occur within days, leading to narrowing of the joint space. With protracted chondral loss, the infection may progress into the subacute phase, when subchondral edema and subchondral cyst formation occur. Patients are also at increased risk for rotator cuff tear at this time, as extension of joint fluid into the subscapularis recess facilitates invasion of the cuff tendons by inflamed outpouchings of synovium. In the setting of a cuff defect, a secondary acromioclavicular septic arthritis can develop. Acromioclavicular joint sepsis can also occur in the absence of any preexisting glenohumeral infection. However, this a rare occurrence; one recent review from France reported 5 cases over a 6-year period. The same review noted only about 20 reported cases in the literature.

In the chronic stages of glenohumeral septic arthritis, joint destruction progresses and ultimately can lead to ankylosis. Osteomyelitis may also occur, and can be difficult at times to distinguish from reactive marrow edema caused by the joint infection. One of the more reliable indicators of osteomyelitis is more confluent T1 hypointensity within the marrow, more overt than that usually seen in the setting of reactive edema. In neonates, incomplete red to yellow marrow conversion limits the usefulness of T1-weighted images for osteomyelitis detection, and in these cases the distinction between reactive edema and true osteomyelitis may be more challenging (see Fig. 1 ). The temporal and anatomic progression of red to yellow marrow conversion during normal maturation has been elaborated. Knowledge of these conversion patterns may help prevent false-positive and false-negative interpretations in the neonatal population. Also, if the contralateral joint is within the field of view, assessing the degree of marrow T1 hypointensity or T2 hyperintensity relative to the normal joint may be helpful.

Overall, the findings reported to correlate most strongly with septic arthritis are synovial enhancement, perisynovial edema, and joint effusion ( Fig. 2 ). Karchevsky and colleagues reported the presence of these findings in 98%, 84%, and 70%, respectively, of 50 consecutive subjects with joint infection (the study was not restricted to glenohumeral infection). Still, no one MR sign reliably confirms septic arthritis while excluding aseptic inflammatory arthritis. Graif and colleagues had demonstrated this in their assessment of multiple MR findings: joint effusion, fluid outpouching, fluid heterogeneity, synovial thickening, synovial periedema, synovial enhancement, cartilage loss, bone erosions, bone erosion enhancement, bone marrow edema, bone marrow enhancement, soft tissue edema, soft tissue enhancement, and periosteal edema. Their study did, however, demonstrate a strong trend toward significance for erosions indicating joint infection, as well as for erosions combined with bone marrow edema, synovial thickening, synovial periedema, bone marrow enhancement, or soft tissue edema.

When considering septic shoulder arthritis, it is important also to be mindful of nonpyogenic infections, especially those caused by Mycobacterium tuberculosis and other mycobacteria, as they can present with quite different clinical and imaging features. Skeletal tuberculosis (TB) is encountered in 1% to 3% of extrapulmonary cases of TB, and of these skeletal cases 1% to 10% involve the shoulder. Fever, erythema, and warmth around the joint are highly unreliable indicators on examination, as nonpyogenic infection can smolder in the joint for years. Richter and colleagues found an average 15-month delay from time of symptom onset to correct diagnosis of TB of the shoulder. Although MR imaging and radiography lack specificity for tuberculous shoulder infection, they demonstrate certain helpful findings. Generally, the cardinal features of mycobacterial infection are osteoporosis, marginal subchondral erosions (usually occurring later), and gradual rather than rapid cartilage destruction: the triad of Phemister. An appearance somewhat similar to chronic pyogenic osteomyelitis also can sometimes be seen, including sclerosis, periostitis, and synovial membrane thickening. Large effusion and osteolysis are other associated features. Even T2-intermediate intraosseous tubercles are sometimes encountered. Tuberculous bursitis has also been well described, having been encountered most commonly in bursae of the shoulder, hands, ischia, and gluteal muscles. Although also nonspecific, intrabursal rice bodies may be shed in the setting of TB or any chronic bursitis, appearing subcentimeric and isointense to muscle on both T1-weighted and T2-weighted images. Like pyogenic septic arthritis, nonpyogenic disease can result in significant bone and joint destruction in advanced stages of infection.

Assessment for potential infection of the postoperative shoulder poses unique challenges. Susceptibility artifact related to metallic hardware or to cement has been the major limitation to using MR imaging for diagnosing acute joint infection, as areas of interest may easily be obscured. Computed tomography (CT) is also subject to metal-related artifacts, which compounds the inherent low sensitivity of this modality for detecting subtle or early soft tissue changes. Radionuclide imaging has been the mainstay for evaluating the instrumented shoulder for septic arthritis, as it does not show deleterious artifacts from the presence of hardware and can be targeted for visualization of either soft tissue–centered or osseous-centered inflammation. Bone scintigraphy, although lacking in specificity, can almost entirely exclude osteomyelitis via a completely normal scan. By coupling bone scintigraphy with gallium 67 imaging, soft tissue inflammation that may occur in the earlier stages of septic arthritis can be detected, even before any associated osteomyelitis has developed. Accuracy for diagnosing or excluding joint infection in the setting of arthroplasty may be increased even further, to 90%, through the combination of bone scintigraphy with labeled-leukocyte imaging, as leukocytes demonstrate the highest sensitivity for detecting neutrophil-mediated processes. Some studies have reported on positron emission tomography (PET) imaging for diagnosing infection, but the clinical usefulness of this application remains unsettled.

Regardless of its limitations, MR imaging is still of value in the setting of joint arthroplasty with potential infection. Field distortion can be minimized by using lower field strength, wider bandwidths, smaller voxels, and/or higher gradients. Frequency-selective fat suppression and gradient echo techniques should be avoided. STIR and water excitation show less distortion. Although not yet widely used in clinical practice, there are also rapid advances in further minimizing artifacts through the use of fast spin echo (FSE) metal artifact reduction sequences (MARS), and newer multi-acquisition variable-resonance image combination (MAVRIC) and slice-encoding metal artifact reduction (SEMAC) sequences. Initial studies on patients undergoing shoulder, hip, and knee arthroplasty have demonstrated improved visualization of synovitis, periprosthetic bone, supraspinatus tendon fibers, and supraspinatus tendon tears with MAVRIC sequences.

RA

Background, Demographics, and Imaging Findings

RA is an inflammatory arthritis that begins at the synovial level. Certain autoimmune factors are important in the pathogenesis of RA. Naive B cells accumulate in synovium where select clones seem to be continuously activated. Synovial tissue T cells express transcription factors also important for maintaining an inflammatory response. The synovium, now congested with immune cells, becomes progressively more inflamed under the influence of monocyte and macrophage-secreted cytokines such as interleukins (IL)-1, IL-6, and IL-17, and tumor necrosis factor alpha (TNFα). Several of these cytokines also upregulate osteoclast activity and production of chondrolytic factors. The inflamed synovium can either return to a normal state or continue to hypertrophy into pannus that destroys articular cartilage, periarticular soft tissues, and bone under the influence of these inflammatory factors. Recent studies have also emphasized the importance of fibroblastlike synoviocytes (FLSs) that predominate in the synovium of patients with RA, especially as these may be the cell type most responsible for the spread of RA from one joint to another. Many studies have implicated the oral cavity bacterium Porphyromonas gingivalis in the pathogenesis of the disease, noting that patients with RA have high antibodies to the organism. It is thought that the bacterium’s ability to citrullinate enolase molecules at a site slightly different from that which is citrullinated physiologically may produce the autoantigen central to the inception of RA. Anticitrullinated protein antibodies (ACPAs) have been found in the serum of patients with RA and are thus considered a fundamental part of the disease pathway.

Recent epidemiologic summaries cite an approximate 1% prevalence of RA in the United States, England, and much of mainland Europe. Rates are slightly to considerably lower among the world’s remaining populations, with the exception of some Native American groups in which as many as 6% of individuals may be affected. The incidence rate of RA for the United States is approximately 0.02% to 0.07%. RA has a 3:1 female-to-male predominance, and a median age at onset of 30 to 50 years. Genetics and heavy smoking are additional risk factors.

RA can affect any synovial joint in the appendicular or axial skeleton, but favors the metacarpophalangeal, metatarsophalangeal, and proximal interphalangeal joints of the hands and feet, as well as the radiocarpal and radioulnar joints. However, the glenohumeral and acromioclavicular joints are frequently involved as well. Shoulder symptoms have been reported in 50% of patients within 2 years of disease and 83% within 14 years. Radiographic changes at the shoulder have been seen within 6 years of disease in more than 50% of patients and in 64% of patients within 19 years.

As opposed to septic arthritis, which rarely involves the acromioclavicular joint in isolation, RA of the shoulder frequently affects the acromioclavicular joint. In a study of 148 shoulders at 15 years of follow-up, Lehtinen and colleagues found erosive change in the acromioclavicular joint alone in 17% of the shoulders, in the glenohumeral joint alone in 6%, and in both joints in 42%.

RA can involve the bursae, especially the subacromial-subdeltoid bursa, resulting in marked, masslike distention that may be mistaken for a soft tissue neoplasm. The pain that results may prompt patients to limit motion, leading in the long term to tightening of the joint capsule, ie, adhesive capsulitis. Although ultimately a clinical diagnosis, adhesive capsulitis may be suggested by certain MR imaging findings in the appropriate patient population for which pretest probability is already increased. Classically, rigidity of the coracohumeral and superior glenohumeral ligaments may present as thickening on MR imaging. Concomitant findings include obliteration of the normal fatty signal intensity within the rotator interval and thickening of the joint capsule at the axillary recess. The tightened joint capsule makes motion even more uncomfortable, prompting chronic diminished use of the shoulder, which in turn leads to atrophy of the rotator cuff muscles. Then, gradual superior migration of the humeral head results in a narrowed outlet with rotator cuff impingement, leading to tear. Up to 80% of patients with RA have significant thinning of the rotator cuff; and up to 20% have full-thickness tears. Although cuff repair is an option, benefits are limited. One review of 23 repairs performed on RA shoulders over a 15-year period demonstrated significant improvements in pain and patient satisfaction after repair, but functional gains (defined as an increased range of abduction) were only obtained in the partial-thickness tear group.

Radiography remains a critical component in monitoring progression of joint compromise and destruction in RA. Osteoporosis, marginal erosions, narrowing of joint space, subchondral sclerosis and cyst formation, and soft tissue swelling are cardinal features seen in varying combinations depending on the stage of disease. As joint space is lost to destruction by synovial pannus, bone is also eroded, classically in a periarticular distribution. At the shoulder, this typically manifests at the superolateral aspect of the humerus, adjacent to the greater tuberosity, corresponding to the humeral bare area between the articular cartilage of the humeral head and the reflection of the joint capsule. An erosion may also develop opposite this site, in the medial aspect of the surgical neck of the humerus secondary to pressure from the glenoid. As the erosive process progresses, the greater tuberosity, anatomic neck, and apposing glenoid are destroyed and the resulting arthropathy can mimic a neuroarthropathy or crystal-associated arthropathy.

Such destructive changes are well assessed by conventional radiography and are part of the meter by which disease severity is scored. The low cost of radiographs, their widespread availability, the relative ease of establishing reproducible interpretations and assessment algorithms, and the ability to correlate findings with standardized scoring criteria are major benefits. For these reasons, radiographic findings are included in the American College of Rheumatology (ACR) classification criteria for RA and radiographs are recommended in clinical trials with duration of 1 year or longer.

MR imaging offers several advantages over conventional radiography. Because RA begins at the synovium, an isolated finding of subtle synovitis can suggest early disease in the appropriate clinical context. Standardized methods of scoring disease have been developed; most notably the rheumatoid arthritis MRI scoring (RAMRIS) system, developed as part of the Outcome Measures in Rheumatoid Arthritis Clinical Trials (OMERACT) international initiative. Generally, this uses standard field strength (1.5 T) contrast-enhanced MR imaging of the wrist and metacarpophalangeal joints to assign numeric scores for the severity of each of 3 findings: synovitis, marrow edema, and erosions. Studies have demonstrated good intrareader variability but less reliable interreader performance with this method. Other newer efforts include quantification of synovitis volume through segmentation and dynamic contrast-enhanced (DCE) MR imaging. DCE MR imaging has been particularly promising. A gadolinium dose of 0.05 to 0.3 mmol/kg is used and, typically, short repetition time, short echo time T1-weighted gradient echo images are acquired every few seconds over a period of minutes.

DCE MR imaging of the knees and wrist joints has yielded promising results. Cimmino and colleagues demonstrated that the enhancement rate of wrist synovium can be used to distinguish between active and inactive disease. Although the correlation with disease activity is probably the strongest advantage of DCE MR imaging at this point, other studies have shown that the early (within approximately the first 60 seconds after injection) enhancement rate of synovium correlates with erosions, pain, ESR levels, erosive progression, and treatment effects. DCE MR imaging findings have also correlated well with histopathologic findings. Active research is focusing on the potential role for DCE MR imaging in monitoring and helping to appropriately time RA treatment with newer disease-modifying antirheumatoid drugs (DMARDs) such as anti-TNFα. Response to more established agents such as corticosteroids and methotrexate are also under investigation. Because the shoulder tends to be affected later and less commonly than the hands and wrists in RA, these more recent MR scoring methods and dynamic enhancement protocols have not yet, to our knowledge, been scientifically applied to the glenohumeral and acromioclavicular joints. Similarly, no standardized protocols to assess treatment response via shoulder imaging have been developed. As RA tends to produce more pronounced changes in the dominant hand and wrist, multiple studies monitoring response to therapy use MR imaging of the dominant hand and wrist only. It would seem reasonable, then, to image the dominant shoulder if clinically indicated for treatment monitoring. Perhaps not surprisingly, whereas most of the sequelae of RA favor the dominant limb, adhesive capsulitis more commonly affects the nondominant shoulder.

Although dynamic and other advanced MR techniques are not in widespread use for the shoulder, conventional MR imaging is already a powerful tool for evaluating the extent of both early and late disease. The earliest findings of RA, namely, synovitis and joint effusion, are soon followed by erosions, all well demonstrated by MR imaging. As in the case of septic arthritis, synovitis can be appreciated by avid or thick enhancement, sometimes with a frondlike morphology as a more masslike pannus begins to develop. Even in the absence of intravenous contrast, synovitis can often be appreciated, especially if well outlined by a joint effusion. Over time, as RA reaches its advanced stages, portions of the synovium may even fail to enhance or may demonstrate relative hypoenhancement and T2 intermediate to low signal intensity, reflecting fibrous synovitis, although small amounts of fibrotic pannus can even be seen earlier in the disease course. Later on, the synovium will turn fatty. As with radiography, marginal erosions are found at the posterolateral aspect of the humeral head, and are often well-marginated T1-hypointense, T2-hyperintense foci within the bone, sometimes quite large ( Fig. 3 ). In Lehtinen and colleagues series of 148 glenohumeral joints, MR imaging revealed erosive changes in 71(48%) of the joints; and erosions were seen on the superolateral articular surface of the humeral head in 61 of these 71 joints. Glenoid involvement was only found in 28. Alasaarela and colleagues demonstrated the superiority of MR imaging for visualizing humeral head erosions. In their prospective multimodality study of 26 shoulders in 26 symptomatic patients with RA, MR imaging revealed humeral erosions in 25 shoulders, ultrasonography in 24, CT in 20, and conventional radiography in 19. All erosions were on the posterolateral aspect of the humeral head at the insertion of the rotator cuff. It was thus concluded that MR imaging is superior to other modalities in detecting small erosions of the humeral head.

MR imaging is also useful for assessing cartilage but this can be considerably more challenging than evaluating erosions and synovitis given that the mean widths of humeral head cartilage (1.24 mm) and glenoid cartilage (1.88 mm) are near the limit of spatial resolution for most MR imaging scanners. Nevertheless, focal regions of wear and destruction can often be appreciated, and visualization improves with high field strength (3T) scanners. Advances in structural and biochemical-based cartilage imaging in other parts of the body, particularly the hip and knee, could theoretically provide a more detailed picture of chondral compromise and areas of impending joint space loss in the shoulder. Recent advances include rotating frame T1 (T1 rho) mapping, Na mapping, T2 mapping, and delayed gadolinium-enhanced MR imaging of cartilage (dGEMRIC). However, these methods are not yet in widespread clinical use even for other joints.

As noted earlier, the prevalence of thinning and tearing within the rotator cuff tendons is substantial in the setting of RA ( Figs. 4 and 5 ). This may be related partly to the destructive effects of synovitis at and near the supraspinatus-infraspinatus footprints. MR imaging allows for characterization of the extent of tearing, facilitating surgical planning for repair. In more longstanding cases, identification of fatty muscle atrophy has important prognostic implications for the potential success of a primary repair. Multiple studies suggest that patients with advanced rheumatoid arthropathy benefit from either hemiarthroplasty or reverse shoulder arthroplasty, provided that either the rotator cuff is intact or can be well repaired at the time of arthroplasty. MR imaging is useful in assessing the integrity of the deltoid muscle, a key consideration when determining appropriateness for reverse shoulder arthroplasty. Results of a recent Cochrane review, however, highlight a relative paucity of research evidence to support decision-making about arthroplasty in patients with RA. At least one study by Soini and colleagues concluded that MR imaging before arthroplasty was of only minor importance in cases of severely destroyed rheumatoid shoulder. The investigators observed that the extent of scar tissue and inflammation in their series of 31 patients limited the accuracy of soft tissue analysis. However, these were advanced cases of disease and the same study demonstrated that an accurate determination of cuff damage and surrounding anatomy was difficult even during open inspection at time of surgery.

Beyond assessment of the cuff, analysis of surrounding soft tissue and osseous structures is important. The acromioclavicular joint, for instance, is of particular significance. In a review of 66 patients with RA with total shoulder arthroplasty and 75 patients with RA with hemiarthroplasty, erosions and cysts at the acromioclavicular joint were found to be associated with poorer postoperative scores on the Clinical Hospital for Special Surgery Inventory. Also, the status of the rotator cuff, and its repair at time of surgery, were predictive of the degree of postoperative improvement. Careful assessment of acromioclavicular joint integrity is therefore advisable, alongside characterization of any rotator cuff compromise when using MR imaging to examine the patient with RA preoperatively.

In a series of 49 patients with RA, Petersson assessed clinical and radiographic findings and noted radiographic changes to the acromioclavicular joint in 85% of cases. Pain and tenderness on examination were observed in one-third of the cases. Early in the course of disease, the acromioclavicular joint may demonstrate subchondral osteoporosis on radiographs, followed by erosions at the undersurface of the clavicle. On MR imaging, distention of the acromioclavicular joint capsule with extension of pannus into the joint may be seen at any stage of disease. As involvement progresses to the early stages of distal clavicular osteolysis, the distal end of the clavicle demonstrates subchondral marrow edema disproportionate to the acromion. Erosions enlarge over time to produce eventual osteolysis of the distal clavicle and even possible erosion of the distal acromion (see Fig. 4 ). Erosive changes often remain more pronounced at the caudal aspect of the distal clavicle. Radiographs and MR imaging demonstrate associated joint space widening, but dislocation and subluxation are uncommon. Lehtinen and colleagues reviewed 148 shoulders in 74 patients with RA after 15 years of follow-up. These investigators found a mean acromioclavicular joint distance (defined as the average of the joint distances measured from the cranial and caudal edges of the distal clavicle to the acromion) of more than 7 mm in 31% of the male shoulders and more than 5 mm in 15% of the female shoulders.

Although it is traditionally taught that the acromioclavicular joint is affected earlier and more severely than the glenohumeral joint in RA, this has not been our experience. However, as RA and septic arthritis can appear similar on many imaging modalities, when the acromioclavicular joint is involved as well, regardless of severity, it makes the diagnosis of RA more likely.

Synovial cyst formation is another complication of RA commonly encountered. The cysts develop within immediately surrounding soft tissues and frequently will dissect along tendon sheaths. Extension along the biceps tendon sheath is characteristic. This synovitis may dissect more than half way down the humerus (see Fig. 3 ). Synovial cysts may also develop under the subscapularis or around the axillary recess. However, the latter 2 locations are more common with septic arthritis. MR imaging allows enumeration and location of these cysts, which are of course not usually appreciable with conventional radiography. Rarely, a cyst may grow large enough to be masslike. MR also identifies subacromial-subdeltoid bursitis, which may manifest on physical examination as swelling about the shoulder in patients with longstanding RA. Regions of proliferative synovium can, over time, infarct and shed into the bursae or joint, appearing as small nodules of varying signal intensity (although generally isointense to muscle) known as rice bodies (see Figs. 4 and 5 ).

Gandjbakhch and colleagues recently published their analysis of 6 cohorts from 5 international medical centers, confirming that synovitis and marrow edema in patients with RA can still be appreciated on MR imaging even in patients who have achieved clinical remission. Thus, MR imaging, even without dynamic enhancement, provides a better gauge of disease activity than conventional radiography by demonstrating subclinical inflammation. This subclinical state of activity accounts for the observation that sometimes radiographic signs of RA progress despite a patient’s being in remission.