Article Text

Download PDFPDF

Dynamic contrast enhanced MRI of bone marrow oedema in rheumatoid arthritis
  1. R Hodgson1,
  2. A Grainger2,
  3. P O’Connor2,
  4. T Barnes2,
  5. S Connolly4,
  6. R Moots3
  1. 1
    MARIARC, Liverpool, UK
  2. 2
    Chapel Allerton Hospital, Leeds, UK
  3. 3
    Clinical Rheumatology, School of Clinical Sciences, University of Liverpool, Liverpool, UK
  4. 4
    Whiston Hospital, St Helens, UK
  1. R Hodgson, MARIARC, Pembroke Place, Liverpool L69 3GE, UK; RichardHodgson{at}


Aims: The aim of this work was to assess the feasibility of using dynamic contrast enhanced (DCE) MRI of bone marrow oedema, to compare it with conventional marrow oedema scoring systems, and to determine the effects of anti-tumour necrosis factor (TNF)α therapy.

Methods: The wrist and metacarpophalangeal (MCP) joints of 25 patients with rheumatoid arthritis were studied. A total of 14 were imaged before and 1–2 weeks after anti-TNFα therapy. T2-weighted fat-suppressed images were collected. A dynamic series of 24 3D spoiled gradient-echo images were acquired before, during and after the intravenous administration of gadolinium-based contrast medium. Oedema was scored using the conventional Rheumatoid Arthritis MRI Scoring (RAMRIS) system from T2-weighted images. The relative enhancement rate (RER) was calculated using the dynamic series from oedematous bone, bone adjacent to oedema and from an uninvolved bone.

Results: A total of 56% of patients showed bone marrow oedema. The RER was significantly increased in and adjacent to areas of marrow oedema. There was a significant reduction in the RER after treatment, but not in the RAMRIS score.

Conclusions: Dynamic contrast enhanced MRI of bone marrow oedema yields additional information to RAMRIS scoring and may be a more sensitive marker of inflammatory activity and response to treatment.

View Full Text

Statistics from

The role of subchondral bone changes in the pathogenesis of rheumatoid arthritis remains controversial.1 Various histological inflammatory changes have been demonstrated1 2 and tumour necrosis factor α (TNFα) has been implicated.3 It has been suggested that the bone marrow oedema, a common finding seen on MRI of rheumatoid arthritis patients,46 is closely linked to these inflammatory changes.1 2 Its extent may be measured semi-quantitatively on T2-weighed images, for example using the Outcome Measures in Rheumatoid Arthritis Clinical Trials (OMERACT) Rheumatoid Arthritis MRI Scoring system (RAMRIS).5 Marrow oedema is associated with an increased risk of erosive progression6 and has been shown to respond to treatment, including anti-TNFα therapy.7 It is therefore expected to be a good marker for disease activity.

Measurements from dynamic contrast enhanced MRI (DCE-MRI) of the inflamed synovium, in particular early enhancement rates, have been used to monitor disease activity. They have been shown to correlate with histological evidence of inflammatory activity,8 predict erosive progression9 and respond to treatment including anti-TNFα drugs.10

Bone oedema also enhances with intravenous gadolinium. The aim of this work was to investigate bone marrow oedema using DCE-MRI, to compare DCE-MRI measurements with conventional RAMRIS scoring and to determine the effects of anti-TNFα therapy on such measurements.


Patient population

A total of 25 patients with severe, long-standing, active rheumatoid arthritis who fulfilled the British Society for Rheumatology guidelines for anti-TNFα treatment were studied. Patients included had not received steroids for at least 6 weeks before starting the study or were receiving stable, low dose oral steroids (<10 mg/day). A total of 14 of the patients were imaged before and 1 and 2 weeks after starting anti-TNFα treatment. The remaining 11 patients were imaged once before treatment. Patients were assessed before treatment using the Disease Activity Score (DAS28). The study was approved by the Liverpool Ethics Committee and informed consent was obtained.


The wrist and metacarpophalangeal (MCP) joints of the more involved side were imaged using a Siemens Trio 3T scanner (Siemens, Erlangen, Germany). T2-weighed, pre-contrast, fat-suppressed, fast spin echo, coronal images were acquired from 2 mm slices with an in-plane resolution of 0.4 mm, repetition time 2.5 s, echo time 33 ms and echo train length 7. A dynamic series of sequential, spoiled, gradient-echo images were acquired with repetition time 4.5 ms, echo time 2.0 ms, flip angle 30° and a resolution of 0.8×0.8×1 mm3. A total of 24 images were acquired every 13 s before, during and after a bolus dose of intravenous contrast (approximately 0.1 mmol/kg Gd-DOTA (gadolinium tetraazacyclododecane tetra acetic acid meglumine)).

Image scoring

T2-weighted images were scored independently by two blinded, experienced musculoskeletal radiologists for bone marrow oedema according to the OMERACT RAMRIS score on a scale of 0–3 for each bone or region.11 The composite score for oedema of the wrist and MCP joints was calculated.

Image analysis

Regions of interest (ROIs) consisting of enhancing synovium were outlined manually on post-contrast images. In patients with bone marrow oedema, T2-weighted images were used to outline three ROIs (fig 1): (a) from the most involved part of the most oedematous bone (ROI a); (b) from non-oedematous bone adjacent to oedematous marrow (ROI b); and (c) from a control bone in which there was no oedema (ROI c).

Figure 1 Coronal T2-weighted image of the hand and wrist of a patient with long-standing rheumatoid arthritis. There is oedema of the lunate (L) and the capitate (C). The oedematous areas demonstrate high signal within the bone marrow with preservation of the cortical margins, necessary to make the diagnosis of bone marrow oedema. Regions of interest (ROI) are shown in the capitate consisting of oedematous bone (solid circle, ROI a) and of bone adjacent to oedema (dotted circle, ROI b).

The relative early enhancement rate (RER) was calculated from the synovitis and each of the bone marrow ROIs. The RER was defined as the maximum relative increase in enhancement over 26 s, ie (S26–S0)/S0. The RER was measured over 26 s (twice the image interval) to improve reproducibility while maintaining sensitivity to vascularity.

Statistical analysis

The inter-reader interclass correlation coefficient (ICC) was calculated for the RAMRIS scores. The RER from the oedematous marrow (ROI a) was compared with that from the control marrow (ROI c). The RER from bone adjacent to oedematous marrow (ROI b) was also compared to that from the control marrow (ROI c). Pre- and post-treatment RER values were compared for each of synovitis and marrow oedema (ROI a) to look for a significant reduction in RER after treatment. Pre- and post-treatment RAMRIS scores for marrow oedema were also compared. Significance testing was performed using the Wilcoxon matched-pairs signed-rank test. In patients with marrow oedema, correlations were sought between the bone oedema RER and synovitis RER as well as between the bone marrow RER and the RAMRIS oedema score by calculating the rank correlation coefficient.


The average pre-treatment DAS28 score of all patients was 6.44.

Figure 1 shows an example of a T2-weighted image of the wrist and MCP joints of a patient with bone marrow oedema of the lunate and capitate. Of the 25 patients studied, 14 demonstrated bone marrow oedema on T2-weighted images (56%). These included 7 of the 14 patients imaged before and after anti-TNFα therapy. The mean pre-treatment RAMRIS oedema score was 9.3. The inter-reader ICC of the RAMRIS scores was 0.93.

The RER in bones with oedema (ROI a) was significantly greater than in those without oedema (ROI c) (mean value 0.71 vs 0.07, p<0.001). The RER adjacent to oedematous bone (ROI b) was also significantly greater than in bones without any oedema (ROI c) (mean value 0.10 vs 0.07, p = 0.007).

Table 1 Effects of treatment on RER of marrow oedema and synovitis

There was a significant decrease after treatment in the RER of the bone marrow oedema of 50% (p = 0.016) and of the synovitis of 35% (p = 0.008). There was a small increase in the RAMRIS bone marrow oedema score after treatment (0.2) that was not significant.

There were no significant correlations between the bone marrow oedema RAMRIS score and the RER or between marrow and synovial RERs. All patients demonstrated enhancing synovitis.


This study demonstrates dynamic contrast enhanced MRI of bone marrow oedema is feasible. The results show that bone in and adjacent to areas of marrow oedema are associated with significantly increased RERs, while control areas from non-oedematous bone show little enhancement. Marrow oedema is thought to represent osteitis, which has been shown to be associated with increased vascularity,12 which is consistent with the greater RER. The higher RER may also be partly due to an increase in capillary permeability caused by pro-inflammatory cytokines.2 13

A significant decrease in the RER of the oedematous bone marrow was seen after 2 weeks of anti-TNFα treatment. This rapid response is consistent with histological evidence of TNFα involvement in bone marrow inflammation.3 13 It also suggests the possibility of using DCE-MRI of marrow oedema as a marker of treatment response in clinical trials.

No significant reduction in the RAMRIS bone oedema scores was seen after anti-TNFα therapy. There have been conflicting reports of the response after 4 weeks of treatment,7 14 with the larger study showing a significant improvement.7 The failure to show a reduction in the RAMRIS score in the current study may reflect the smaller effect after only 2 weeks of treatment, the smaller number of patients or the greater disease severity. However, the demonstration of a significant reduction in the RER suggests this may be more sensitive to treatment than the RAMRIS score. The RER depends on the bone marrow vascularity and capillary permeability, which might be expected to be more responsive to treatment than the RAMRIS score, which depends on the amount of bone involved. This difference may also explain the failure to detect a significant correlation between marrow oedema RER and RAMRIS scores.

A significant reduction in the RER of the enhancing synovitis was observed after anti-TNFα therapy, as previously reported after 4 weeks of treatment.10 However, no significant correlation was observed between the RER of bone marrow oedema and of synovitis. This is perhaps not surprising as only 56% of patients showed marrow oedema (although enhancing synovitis was present in all). This is consistent with values of 45–68% from the literature.46 15

One important disadvantage of using oedema rather than synovitis for DCE-MRI is that it is only present in approximately half of rheumatoid arthritis patients.46 15 However, since these are the patients at greater risk of developing erosions6 it may still be an appropriate measure for assessing erosive progression. Furthermore, the RER in bone marrow adjacent to oedematous bone was significantly greater than that in bones without oedema. This suggests that the vascularity is increased in these areas although conventional RAMRIS scoring is not sensitive enough to detect the difference. This raises the possibility that DCE-MRI may reveal abnormalities before marrow oedema is apparent; it may even be possible to detect a response to treatment in bone marrow without frank oedema.

The preliminary results presented here are limited by the relatively small number of patients. Further work will be needed to determine the value of DCE-MRI of bone marrow oedema in predicting outcome in terms of DAS28 score or erosive progression and for assessing treatment response in individual patients.

In conclusion, dynamic contrast enhanced MRI of bone marrow oedema is technically feasible. It reveals more extensive bone marrow change than conventional T2-weighted images. DCE-MRI has shown a significant reduction in the RER of marrow oedema after treatment and may be more sensitive for detecting changes in inflammatory activity than conventional RAMRIS scoring systems.


View Abstract


  • Funding: This work was funded by a Fellowship Grant from the Royal College of Radiologists, UK.

  • Competing interests: None

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.