Bone erosions and bone marrow edema as defined by magnetic resonance imaging reflect true bone marrow inflammation in rheumatoid arthritis




To investigate the pathologic nature of features termed “bone erosion” and “bone marrow edema” (also called “osteitis) on magnetic resonance imaging (MRI) scans of joints affected by rheumatoid arthritis (RA).


RA patients scheduled for joint replacement surgery (metacarpophalangeal or proximal interphalangeal joints) underwent MRI on the day before surgery. The presence and localization of bone erosions and bone marrow edema as evidenced by MRI (MRI bone erosions and MRI bone marrow edema) were documented in each joint (n = 12 joints). After surgery, sequential sections from throughout the whole joint were analyzed histologically for bone marrow changes, and these results were correlated with the MRI findings.


MRI bone erosion was recorded based on bone marrow inflammation adjacent to a site of cortical bone penetration. Inflammation was recorded based on either invading synovial tissue (pannus), formation of lymphocytic aggregates, or increased vascularity. Fat-rich bone marrow was replaced by inflammatory tissue, increasing water content, which appears as bright signal enhancement on STIR MRI sequences. MRI bone marrow edema was recorded based on the finding of inflammatory infiltrates, which were less dense than those of MRI bone erosions and localized more centrally in the joint. These lesions were either isolated or found in contact with MRI bone erosions.


MRI bone erosions and MRI bone marrow edema are due to the formation of inflammatory infiltrates in the bone marrow of patients with RA. This emphasizes the value of MRI in sensitively detecting inflammatory tissue in the bone marrow and demonstrates that the inflammatory process extends to the bone marrow cavity, which is an additional target structure for antiinflammatory therapy.

Chronic synovitis in the context of rheumatoid arthritis (RA) leads to pathologic changes in adjacent structures, such as the articular cartilage, the cortical bone surface, and the underlying bone marrow. Knowledge of this complex destructive process is predominantly driven by findings of radiographic examinations, which have identified local bone erosions as well as joint space narrowing as key monitoring parameters in RA (1, 2). From these findings it is apparent that inflamed synovial tissue has the capacity to invade neighboring structures such as cartilage and bone. It has been particularly difficult, however, to unravel the histopathologic nature of these changes, since usual methods to assess and/or to surgically treat synovitis, such as biopsy or synovectomy, target the synovial tissue itself but do not yield insight into changes in cartilage, bone, or bone marrow. It is therefore not surprising that information about the structural and functional correlates of radiographic findings in RA is scarce and is driven by findings in specimens obtained at joint replacement surgery, which usually occurs late in the disease course. Thus, key features of local bone erosions in RA have only recently been described, revealing that these lesions are populated by osteoclasts, which have the capacity to degrade bone (3, 4).

Improvements in imaging strategies, in particular, technical developments in magnetic resonance imaging (MRI), have provided insight into the complexity of joint destruction in RA (5, 6). Thus, MRI scanning has extended our knowledge of RA by allowing direct visualization of synovial inflammation and by depicting the invasion of inflammation into bone and bone marrow very early in the disease. Importantly, MRI scans show signal changes, which extend into the bone marrow cavity and are linked either to cortical bone destruction (“MRI bone erosion”) or to more diffuse changes in the bone marrow space (“MRI bone marrow edema” or “MRI osteitis”). The latter lesions have also been described in osteoarthritis, ankylosing spondylitis, and systemic lupus erythematosus (7–9). Both lesions exhibit signal characteristics consistent with increased water content (5), distinguishing these lesions from the fatty bone marrow of the extremities.

The morphologic nature of bone marrow changes in RA, however, has not been well investigated. The lack of easy access to bone marrow from RA patients is a logistic challenge and has so far prevented clear definition of the structural correlates of MRI changes. However, recent histologic investigations using specimens obtained at joint replacement surgery have shown that fat-rich bone marrow can indeed be focally replaced by inflammatory synovial tissue, which invades the cortical bone, penetrates the cortical barrier, and exposes the bone marrow to inflammatory triggers, leading in particular to B cell–rich cellular aggregates (10).

In the present study, we aimed to define the nature of bone marrow edema in RA. We studied RA patients scheduled for total joint replacement of the proximal interphalangeal (PIP) and metacarpophalangeal (MCP) joints. These joints were scanned by MRI on the day before surgery and subsequent histologic processing. This allowed investigation of the histopathologic nature of bone marrow changes in RA as depicted by MRI.



Twelve different joints (heads of 2 second metacarpal, 3 third metacarpal, 2 fourth metacarpal, 2 fifth metacarpal, 1 second proximal phalangeal, 1 third proximal phalangeal, and 1 fourth proximal phalangeal bone) from 3 patients (all women, ages 43, 56, and 61 years) with longstanding RA (disease duration 8, 14, and 24 years) were assessed. All patients fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) criteria for the classification of RA (11). Patients were scheduled for joint replacement surgery because of chronic and persistent pain, joint swelling, and impaired range of motion in the target joint. All patients were being treated with methotrexate (15 mg/week) and low-dose glucocorticoids (5 mg/day). Surgical procedures were performed according to the methods described by Swanson (12), consisting of resection of the affected metacarpal or proximal phalangeal heads followed by implantation of a silicone spacer (NeuFlex; DePuy Orthopaedics, Warsaw, IN) (13). Written informed consent was obtained for all procedures.

Magnetic resonance imaging.

MRI was performed with a 1.5T MR scanner (Philips, Nijmegen, The Netherlands), using a flexible surface coil (flex medium) to obtain coronal STIR sequences. The acquisition parameters were as follows: repetition time (TR) 1,200 msec, echo time (TE) 14 msec, 2 averages, field of view (FOV) 200 mm, matrix 1,926 × 256 pixels, slice thickness 3 mm, interslice gap 0.3 mm, scan time 2 minutes 18 seconds. In addition, coronal T1-weighted sequences were obtained (TR 450 msec, TE 13.8 msec, 2 averages, FOV 200 mm, matrix 1,926 × 256 pixels, slice thickness 3 mm, interslice gap 0.3 mm, scan time 2 minutes 18 seconds). MRI bone erosions were defined based on hyperintensity on STIR sequences and hypointensity on T1-weighted sequences, in direct contact with cortical bone and with well-defined margins and apparent destruction of the cortical bone barrier. MRI bone marrow edema was identified as hyperintense lesions on STIR sequences, with less clearly defined margins and intact trabecular structures (5).

Histologic examination.

After resection, the localization (MCP or PIP joint; second, third, fourth, or fifth digit), side (left or right), and dorsal-palmar plane of each joint were documented. To ascertain an orientation of the histologic sections identical to that of the MRI, the 3-dimensional orientation of the joint had to be documented (Figure 1). In accordance with the MR images, joints were cut in the coronal axis. This was ascertained by defining the distal-proximal orientation by the resection rim and the cartilage, respectively, the lateral-medial orientation by the side of the joint (left or right hand), and the dorsal-palmar orientation by marking the dorsal rim with a suture.

Figure 1.

Ascertainment of coronal-plane magnetic resonance images (MRIs) and histologic sections. Metacarpophalangeal and proximal interphalangeal joints were analyzed in the coronal plane by MRI scanning, as well as by histologic examination of serial sections (black bars). To precisely define the orientation of sections, all 3 dimensions were documented: the distal-proximal orientation was based on the distal localization of the articular cartilage, the dorsal-palmar axis was identified through labeling with a suture placed at the dorsal rim of the joint head (red box) directly after explantation, and the lateral-medial axis was defined based on knowledge of whether the explants came from the left or the right hand. Serial sections were obtained at intervals of 50 μm, allowing identification of the exact localization of the respective section within the dorsal-palmar axis.

After explantation, the specimen was immediately placed into 0.9% NaCl, fixed in 4.0% formalin, and decalcified in 14% EDTA (Sigma, St. Louis, MO). Paraffin-embedded joints were then cut into 2 equal-sized pieces along the coronal plane. Both pieces were used to cut sequential sections (2 μm) every 50 μm, directed to the dorsal rim of the joint in 1 piece and to the palmar rim in the other. For each joint ∼70 serial sections were analyzed. All sections were stained with hematoxylin and eosin and analyzed quantitatively for the degree of bone marrow alterations, using a histomorphometric technique with an Axioskop 2 microscope (Zeiss, Marburg, Germany) and the OsteoMeasure Analysis System (Osteometrics, Decatur, GA) (14). The area covered by bone (cortical plus trabecular), normal bone marrow, and bone marrow with mild cellular infiltration (<50% inflammatory infiltrates per tissue area; intact trabecular structure) or severe infiltration (>50% inflammatory infiltrates per tissue area; trabeculae destroyed) were recorded. MRI and histomorphometric data were interpreted by 2 independent observers (EJ-B and IN-H), under blinded conditions.


Colocalization of cellular infiltrates with bone marrow edema seen on MRI.

To better understand the processes causing the pathologic changes seen on MRI in patients with RA, we performed a serial histologic analysis of sections from throughout the entire finger joint of patients who had undergone joint replacement surgery. STIR MRI sequences obtained on the day before surgery were compared with histologic sections. In all 12 joints analyzed, bone marrow changes were evident on MRI scans as well as in histologic sections. Bone lesions seen on MRI were designated as erosions when they were localized close to cortical bone and associated with synovitis, whereas the more diffuse lesions in the bone marrow were designated bone marrow edema or osteitis. Both types of lesion appeared bright on the STIR sequences but dark on the T1-weighted images, reflecting increased water content and decreased fat content.

Origin of bone erosions seen on MRI.

Analysis of corresponding histologic sections showed that bone erosions seen on MRI were due to localized replacement of bone marrow fat by accumulated inflammatory cells adjacent to a broken cortical bone barrier. Cortical bone is actually only a very thin barrier (∼0.25 mm in width) between the synovium and the bone marrow. A perforation of this layer enabled the accumulation of inflammatory tissue, in the form of either synovial inflammatory tissue or lymphocytic B cell–rich aggregates within the marrow space, appearing as bone erosions. In fact, only a small portion of the MRI lesion that was designated bone erosion represented true structural damage of bone, since inflammation affects the bone marrow after penetration through the cortical barrier.

Figure 2 shows an example of 2 histologic sections of a second metacarpal head from a patient with RA, as well as matched MR images with STIR and T1-weighted sequences. MRI lesions were characterized as a clearly demarcated zone of hyperintense signal within normal hypointense marrow on STIR images, and a hypointense signal on T1-weighted sections. The histologic correlate was identified as local bone marrow inflammation and accumulation of blood vessels at these sites, which were closely linked to a break in the adjacent cortical bone.

Figure 2.

T1-weighted (A) and STIR (B) magnetic resonance images (MRIs) and corresponding histologic sections at low and high magnification (C and D), of a second metacarpal head from a patient with rheumatoid arthritis. MRI bone erosion is defined based on penetration of cortical bone and localized bone marrow inflammation, depicted as a circumscribed area of hypointense signal at the medial circumference on the T1-weighted image in the upper row (arrowhead) within normal hyperintense bone marrow. This corresponds to findings on the STIR image in the upper row, where the erosion is seen as a clearly demarcated zone of hyperintense signal within normal hypointense marrow at this site (arrowhead). A blood vessel and inflammatory tissue next to the junction zone of the joint (arrowheads) are seen in the corresponding histologic images in the upper row. The MR and histologic images in the lower row show a more palmar view of the same joint, with clear signs of invasion of inflammatory tissue into subchondral bone and bone marrow seen on the STIR MRI (arrowhead). The corresponding histologic section shows cortical penetration at this site, with inflammatory tissue invading the bone marrow space and replacing fatty tissue (arrowheads). (Original magnification × 10 in C; × 50 in D.)

Origin of bone marrow edema seen on MRI.

More diffuse MRI signal alterations in the bone marrow of patients with RA are considered to indicate bone marrow edema or osteitis. As was seen with the above-mentioned lesions, they appeared bright on STIR sequences, indicating increased water but lower fat content. Similar to the findings in MRI bone erosions, the histopathologic correlate of MRI bone marrow edema/osteitis was infiltration of the bone marrow by inflammatory tissue. This lends more credence to the term “osteitis” rather than “bone marrow edema.” Thus, all lesions that appeared bright on STIR sequences (and dark on T1-weighted sequences) and were localized within the cortical bone layer were due to inflammatory infiltrates in the bone marrow, regardless of whether these lesions were attached to the endosteum and associated with cortical penetration (MRI bone erosion) or were more diffusely located within the marrow space (MRI bone marrow edema/osteitis).

Distribution of bone marrow changes.

Normal bone marrow is dominated by adipocytes, with occasional interspersed stromal cells. Mild infiltration of bone marrow was characterized by a decreased number of adipocytes in favor of hematopoietic cells infiltrating the bone marrow (<50% infiltrates/tissue area [grade I lesion]). Severe infiltration of bone marrow was recorded based on findings of either synovial pannus–like tissue within the cortical lining, lymphocytic aggregates, or blood vessels associated with inflammatory infiltrates almost completely replacing bone marrow fat (grade II lesion).

Most areas of the more diffuse lesions reflecting MRI bone marrow edema/osteitis were composed of grade I lesions, with some interspersed grade II lesions. In contrast, peripheral lesions reflecting MRI bone erosions were almost exclusively dense infiltrates corresponding to grade II lesions. In accordance with this, grade II lesions were localized peripherally at the dorsal and palmar rims of the bone marrow cavity, reflecting their close interaction with synovial tissue penetrating through the cortical barrier into the bone marrow. Grade II lesions were absent in the center of the bone marrow. Grade I lesions, in contrast, were localized at the center and palmar areas of the joint, colocalizing with lesions appearing as MRI bone marrow edema/osteitis. Areas of mild infiltration of the bone marrow (grade I lesions) were generally more prevalent (by ∼4-fold) than areas with more severe changes (grade II lesions) (Figures 3C and D). Our findings indicated that STIR MRI sequences can depict mild inflammatory infiltrates in the bone marrow, which are commonly termed bone marrow edema/osteitis, as well as dense bone marrow infiltrates associated with penetration of cortical bone, termed bone erosions.

Figure 3.

Different localization patterns of mild and severe bone marrow inflammation. Magnetic resonance imaging (MRI) bone marrow edema is defined based on bone marrow inflammation. A, STIR MRI of the third metacarpal head, showing signal enhancement reflecting bone marrow edema. The image on the right is a close-up of the boxed area in the image on the left. B, Histologic section corresponding to the boxed area in the right image shown in A, demonstrating inflammatory infiltrates in the bone marrow at the site of the MRI lesion. C, Higher-magnification views of the boxed areas in B, showing normal bone marrow containing adipocytes (left portion of C, corresponding to the boxed area in the upper right of B), mild infiltration with hematopoietic cells, reflecting a grade I lesion (middle portion of C, corresponding to the boxed area in the middle of B), and strong infiltration (grade II lesion) with almost complete replacement of fatty tissue by inflammatory tissue (right portion of C, corresponding to the boxed area in the lower left of B). D, Graph depicting the findings in >70 serial coronal sections from the metacarpal head, showing that grade II lesions are localized peripherally at the dorsal and palmar rims where bone erosions are present, whereas grade I lesions are distributed more centrally. (Original magnification × 20 in B; × 200 in C.)


Magnetic resonance imaging not only allows visualization and quantification of synovitis, but has also enabled more detailed viewing of the pathologic changes of neighboring bone, cartilage, and bone marrow. Localized MRI changes in close association with the cortical bone are termed bone erosions, whereas more diffuse changes in the bone marrow are termed bone marrow edema or osteitis (5, 6). These changes originate from focally increased water content in the bone marrow, suggesting that bone marrow fat is replaced by water, or structures containing more water and less fat than normal bone marrow. They appear dark (low signal intensity) on T1-weighted MRIs, whereas they are bright (high signal intensity) on STIR MRI sequences (5).

The fact that MR techniques have revealed profound changes in a previously uncharacterized compartment of the rheumatoid joint, beneath the inflamed surface, is of particular interest. MRI is increasingly used in the monitoring of RA patients who are receiving immunomodulatory therapies, including biologic agents, in both clinical trials and daily clinical practice. Among the radiographic changes observed on the MRIs, the anatomic basis of synovitis is very well characterized and the structural nature of local bone erosion has been recently defined (3, 10, 15, 16). In contrast, little information has been available on the nature of bone marrow changes found in RA joints but not in normal joints. This is due to 1) the apparent difficulty in assessing this particular joint region, which is not accessible via synovial biopsy or synovectomy, and 2) the scientific focus on joint pathology at the outside, but not the inside, of the cortical bone barrier (10).

The cortical bone barrier, which separates the synovial compartment from the bone marrow compartment, is only a very thin layer. The vast majority of the lesion termed bone erosion on MRI scans and the whole lesion termed bone marrow edema/osteitis is clearly localized within this cortical bone layer. This suggests that MRI can depict pathologic changes in the bone marrow beneath the inflamed joint. The present study reveals that these lesions are due to the replacement of bone marrow fat by an inflammatory infiltrate resembling a sterile “osteitis” or “osteomyelitis,” rather than a true edema.

Dense bone marrow infiltrates were found at the periphery of the bone marrow, where adjacent cortical bone had been fenestrated by synovial inflammatory tissue (MRI bone erosions). These lesions were composed of dense infiltrates consisting of 1) synovial inflammatory tissue invading the bone marrow, 2) lymphocytic infiltrates emerging at the interface between synovial tissue and bone marrow fat, and 3) blood vessels close to inflammatory infiltrates. Thus, MRI bone erosions not only show penetration of the cortical barrier, but are largely due to inflammatory changes in the neighboring bone marrow. MRI bone marrow edema was also due to inflammatory infiltrates, but infiltration was less severe and localized to more central regions of the bone marrow.

This study had limitations due to the small number of patients investigated and the focus on late-stage disease. The number of patients was small due to the low frequency of surgical replacement of finger joints, the improved control of disease by pharmacologic methods, and the complexity of the histologic analysis (>70 sections per joint for histomorphometric analysis). However, the fact that MRI lesions corresponded to histologic signs of bone marrow inflammation in all 12 joints investigated is a strong indicator that true inflammation is the cause of MRI lesions in the bone marrow. The second limitation of the study, lack of inclusion of patients with early disease, was unavoidable since, for obvious reasons, finger joint replacement surgery is not the treatment of choice in early arthritis. Thus, we cannot exclude the possibility that MRI bone marrow changes in early disease, which can be reversible, have a different structural correlate than the lesions found in advanced disease as investigated in this study.

In summary, the present results show that MRI bone erosions as well as MRI bone marrow edema/osteitis reflect bone marrow inflammation. This indicates that, in addition to the synovial membrane, juxtaarticular parts of the bone marrow are inflamed in RA, suggesting active involvement of this compartment in the inflammatory process. These findings reveal a previously uncharacterized component of the pathophysiology of RA.


Dr. Schett had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Jimenez-Boj, Nöbauer-Huhmann, Dorotka, Wanivenhaus, Kainberger, Tsuji, Smolen, Schett.

Acquisition of data. Jimenez-Boj, Nöbauer-Huhmann, Hanslik-Schnabel, Dorotka, Wanivenhaus, Trattnig, Axmann, Hermann, Schett.

Analysis and interpretation of data. Jimenez-Boj, Nöbauer-Huhmann, Hanslik-Schnabel, Kainberger, Trattnig, Axmann, Hermann, Smolen, Schett.

Manuscript preparation. Jimenez-Boj, Dorotka, Tsuji, Smolen, Schett.

Statistical analysis. Jimenez-Boj, Schett.

Operations. Hanslik-Schnabel, Dorotka, Wanivenhaus.


We thank Ivana Mikulic and Birgit Tuerk for excellent technical assistance.