Article Text

Download PDFPDF

The effect of etanercept on osteoclast precursor frequency and enhancing bone marrow oedema in patients with psoriatic arthritis
  1. A P Anandarajah1,
  2. E M Schwarz2,
  3. S Totterman3,
  4. J Monu4,
  5. C Y Feng5,
  6. T Shao6,
  7. S A Haas-Smith6,
  8. C T Ritchlin1
  1. 1
    Department of Allergy, Immunology and Rheumatology, University of Rochester, New York, USA
  2. 2
    Department of Orthopedics, University of Rochester, New York, USA
  3. 3
    Virtual Scopics, Rochester, New York, USA
  4. 4
    Department of Radiology, University of Rochester, New York, USA
  5. 5
    Department of Biostatistics, University of Rochester, New York, USA
  6. 6
    Clinical Immunology Research Center, Allergy, Immunology & Rheumatology Research Division, University of Rochester Medical Center, New York, USA
  1. A P Anandarajah, Clinical Immunology Research Center, 601 Elmwood Avenue, Box 695, Rochester, New York 14642; allen_anandarajah{at}


Objective: The frequency of osteoclast precursors (OCPF) and the presence of bone marrow oedema (BMO) are potential response biomarkers in psoriatic arthritis (PsA). Previous studies suggest a central role for tumour necrosis factor (TNF) in the formation of osteoclast precursors. To better understand this association, the effect of etanercept on OCPF and BMO was analysed in PsA patients with erosive arthritis.

Methods: A total of 20 PsA patients with active erosive PsA were enrolled. Etanercept was administered twice weekly for 24 weeks. OCPF was measured and clinical assessments were performed at baseline, 2, 12 and 24 weeks. Gadolinium enhanced MR images were obtained at baseline and 24 weeks.

Results: Significant improvements in joint score (p<0.001), HAQ scores (p<0.001) and SF-36 parameters were observed after 6 months of therapy with etanercept compared to baseline. The median OCPF decreased from 24.5 to 9 (p = 0.04) and to 7 (p = 0.006) after 3 months and 6 months of treatment, respectively. MR images were available for 13 patients. The BMO volume decreased in 47 and increased in 31 sites at 6 months. No correlation was noted between OCPF, BMO and clinical parameters.

Conclusion: The rapid decline in OCPF and overall improvement in BMO after anti-TNFα therapy provides one mechanism to explain the anti-erosive effects of TNF blockade in PsA. Persistence of BMO after etanercept treatment, despite a marked clinical response, was unexpected, and suggests ongoing subchondral inflammation or altered remodelling in PsA bone.

Statistics from

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.

Psoriatic arthritis (PsA) is an inflammatory arthritis with diverse manifestations that reflect the involvement of several tissues. PsA can result in severe joint damage and subsequent disability.1 Indeed, in one study x ray changes were noted to have occurred within 2 years of diagnosis in almost half of PsA patients.2 Peripheral joint damage in PsA may manifest radiographically as large eccentric erosions, joint space narrowing, extensive tuft resorption and pencil-in-cup deformities.3

The critical mechanisms that underlie pathologic bone resorption in PsA are not well understood, but recent evidence underscores a central role for osteoclasts. The importance of osteoclasts in the initiation and progression of bone erosions was first noted in rheumatoid arthritis (RA).4 Osteoclasts arise from circulating CD14+ monocytes or osteoclast precursors (OCP). The receptor activator of NFΚB ligand (RANKL)/osteoprotegerin (OPG) pathway plays an important role in osteoclastogenesis.5 We previously described osteoclasts at the pannus–bone junction in patients with PsA, an observation not seen in the synovium of patients with osteoarthritis.6 RANKL and low levels of OPG protein, conditions that favour osteoclast differentiation, were noted in PsA synovial membranes. We also found an increased osteoclast precursor frequency (OCPF) in the peripheral blood of patients with PsA. Of relevance to this study was the observation that OCPF decreased significantly in PsA patients following treatment with tumour necrosis factor (TNF) antagonists.6

Images provided by MRI studies have provided an unparalleled opportunity to visualise the psoriatic joint.711 Perhaps one of the more striking features reported in psoriatic hand and knee joints is the presence of subchondral peri-entheseal and diffuse bone marrow oedema (BMO), findings that are not as prevalent in RA.12 13 The Outcomes Measures in Rheumatology (OMERACT) conferences defined BMO as a lesion within the trabecular bone, with ill-defined margins and signal characteristics consistent with increased water content.14 The increased water content results in a high intensity signal on T2-weighted fat-saturated and short T1 inversion recovery (STIR) images, and low signal intensity on T1 weighted images. BMO, a finding specific to MRI, has been shown to be a predictor of erosion formation.15 In longitudinal MRI studies of RA joints, McQueen et al estimated that the risk of developing erosion in an area with BMO at baseline, to be 6.5-fold greater than the probability of developing a new erosion in an area without preceding oedema.16 Thus, subchondral BMO may be reversible, whereas bone erosions constitute permanent structural damage with limited capacity for complete repair.17

The pathologic and clinical significance of BMO in PsA is less well understood. McGonagle et al have proposed that subchondral oedema in spondyloarthritis is an integral feature of enthesitis18 that may mark regions of severe bone destruction and/or new bone formation in the spondyloarthropathies.19 In support of this view are recent studies of zygapophyseal and sacroiliac joints in ankylosing spondylitis patients, which found a correlation between histopathological findings of surgical bone specimens in the axial spine and BMO lesions on MRI.20 21

Taken together, the studies outlined above provide data to support the hypothesis that serial assessments of OCP frequency and BMO are reasonable candidates for response biomarkers in PsA. To formally address this hypothesis, we designed a clinical trial with etanercept to assess the impact of anti-TNF therapy on OCPF and BMO in a cohort of PsA patients with erosive arthritis.


Patients and study protocol

This study was designed as a 24-week, single-centre, open label trial, and was approved by the Institutional Review Board at the University of Rochester Medical Center. Written consent was obtained from all participants in the study. Eligible patients were between the ages of 18–65 years, with a diagnosis of PsA based on the Moll and Wright criteria.22 Patients had to demonstrate active disease, defined as three or more tender and swollen joints on physical exam and at least one joint erosion on plain radiographs of hands, feet or knees. Patients with a history of infections, severe or progressive medical illness, substance abuse or demyelinating lesions were excluded from the study. Exclusion criteria also included patients who had been on biologic agents in the past and patients who were unable to undergo MRI for any reason.

A total of 20 patients were selected for the study. All patients were given subcutaneous injections of etanercept, 25 mg twice weekly for 6 months. Patients were instructed on how to inject etanercept by the study coordinator.

Outcome measures

Clinical assessment

A total of 68 joints were assessed for tenderness and 66 joints for joint swelling on a 0–3 scale. The joint count was performed by one of the two study rheumatologists at weeks 0, 2, 12 and 24. The sum of these scores for all joints provided the clinical joint score. Function and Quality of life were measured by the Health Assessment Questionnaire (HAQ) and the Short Form Health Survey-36 (SF-36) respectively. Visual analogue scale (VAS) scores for physician and patient global assessment were also recorded at weeks 0, 2, 12 and 24.

Osteoclast precursors

OCPF was measured at baseline, 2 weeks, 3 months and 6 months after etanercept therapy. The method used to determine OCPF has been described in detail previously.6 Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood, separated on Ficoll gradients and the unfractionated PBMCs were cultured. After 14 days in culture, slides were stained for tartrate-resistant acid phosphatase (TRAP). Slides were viewed by light microscopy, and TRAP positive cells with three or more nuclei were counted as osteoclasts.


The most active joint, on clinical examination that also manifested erosion(s) on plain radiographs, was selected for MR imaging. A gadolinium enhanced MRI was performed at the initial visit and then between weeks 25 and 26. MRI studies were performed using a 1.5 T superconductive magnet (GE Signa, Milwaukee, Wisconsin, USA). All patients were examined in at least one orthogonal plane using fast T2-weighted spin echo pulse sequences with fat suppression, 3D gradient recall echo (GRE), and gadolinium-enhanced images with fat suppression. The field of view varied according to the joint that was imaged. A musculoskeletal radiologist, blinded to the clinical scores, interpreted the MRI findings and analysed the images.

Assessment of BMO

BMO was identified as an area in the bone with low signal intensity on T1-weighted images and high signal on T2-weighted fat suppressed images that enhanced further after gadolinium. Using computer software, an experienced radiologist constructed a three-dimensional image of the joint and outlined the regions with BMO. The following numbers of sites were assessed at each joint: wrist (12), hand (17), knee (7) and foot (6). A volumetric analysis was performed on each BMO lesion at all sites. The number of sites with BMO and the volume of BMO lesions for each site as well for the whole joint were computed.


The demographic details of the patients are presented in table 1. All 20 patients were Caucasian, with a diagnosis of PsA for a period ranging from 3–35 years. The 20 joints assessed by MRI included 6 wrists, 4 knees, 7 digits and 3 hands.

Table 1 Patient demographics

Treatment response

The tender joint score improved in 19 patients (p<0.001) and the swollen joint score improved in all 20 patients (p<0.001). The improvement in the clinical joint score was also significant (p<0.001). The HAQ improved from a mean of 1.21 at screening to 0.36 at week 24 (p<0.001). The physical component of the SF-36 improved from 31.2 at baseline to 46.3 at week 24 (p<0.001) while the mental component of the SF-36 improved from 46.1 to 53.9 (p = 0.017). The physician global assessment improved from 3.5 at screening to 0.7 (p = 0.001) after 6 months of etanercept therapy while the patient assessment improved from 3.5 to 1.3 (p = 0.001) (table 2). No adverse events were recorded.

Table 2 Change in clinical parameters Health Assessment Questionnaire (HAQ), and Short Form (SF) assessed at screening and following 6 months of etanercept therapy

Osteoclast precursor frequency

A striking decline in OCPF was noted as early as 2 weeks after starting therapy in 14 of 19 patients, with a decrease in median OCPF from 28 per 106 PBMC to 8 per 106 PBMC (p = 0.05). In one of five patients with no decline of OCPF at 2 weeks, the OCPF increased from 55 to 661 at 2 weeks, but then decreased to 57 after 6 months of therapy. After 6 months of etanercept therapy, the OCP frequency had decreased in 16 of 20 patients, increased in 2 and remained the same in another 2 patients. The median OCPF for the 20 patients decreased from 24.5 to 9 at 3 months (p = 0.04) and declined further to 7, after 6 months of etanercept therapy (p = 0.006) (fig 1).

Figure 1 OCP frequency for the 20 PsA patients at screening, 3 months and 6 months. The group medians are illustrated by bars.

BMO volumes on 3D MRI

MR Images were available for 13 patients. The 13 joints comprised of 6 wrists, 3 knees and 4 digits. Seven images had to be discarded for technical reasons including patient motion, non-matching longitudinal images and missing sequences.


In the six wrists assessed, the BMO volume for the individual joints decreased in five subjects and increased in one. The total BMO volume for the six subjects decreased from 11 230 mm3 at screening to 7596 mm3, after 6 months of etanercept therapy (p = 0.05). For the 6 wrists, a total of 47 sites were noted to have BMO, pre- and post-etanercept therapy. The extent of BMO decreased in 31 sites, while it increased in 16 (table 3). Interestingly in most patients, while a decrease in the extent of BMO was noted at some sites, an increase was noted at other sites in the same joint. Indeed, only one of the six patients demonstrated a decrease in BMO volume at all sites. In the one patient, in whom the overall BMO volume increased, an increase in BMO volume at five sites and a decrease in three sites were noted.

Table 3 Bone marrow oedema at the wrists of six patients, as measured by number of sites and volume


In the three knee joints, the BMO volume increased in two patients and decreased in one. The total BMO volume for the three patients, decreased from 143 539 mm3 to 142 379 mm3. A total of 16 sites were noted to have BMO. Eight sites showed a decrease in the extent of BMO and the other eight demonstrated an increase, after 6 months of etanercept therapy (table 4).

Table 4 Bone marrow oedema at the knees and digits of seven patients, as measured by number of sites and volume


In the four digits, the BMO volume decreased in three and increased in one. The total BMO volume for the four joints decreased from 13 205 mm3 to 6883 mm3. A total of 15 sites were noted to have BMO. Eight revealed a decrease while seven had an increase in the extent of BMO, after 6 months of etanercept therapy (table 4).

In summary, the overall BMO volume decreased in nine patients and increased in four. The mean BMO volume for all 13 patients, decreased from a mean of 2153.5 mm3 at baseline to 2010.9 mm3 after 6 months of therapy with etanercept. In all patients, BMO was recorded at 78 sites, and the extent of BMO decreased at 47 sites (60.3%) and increased in 31 sites (39.7%), after 6 months of etanercept therapy.

In one patient (patient 8), a third MRI of the knee obtained after 18 months of etanercept therapy showed a decrease in BMO at six of seven sites, compared to a decrease in five of seven sites after 6 months of therapy (fig 2). The BMO volumes were 44 431 mm3 at baseline, 31 204 mm3 at 6 months (a 30% improvement) and 13 911 mm3 at 18 months (69% improvement from baseline). Interestingly, at one of the sites where BMO had increased after 6 months of etanercept, no oedema was noted after 18 months of therapy. In contrast, at another site, BMO increased after 6 months of etanercept treatment and enhancement was even higher after 18 months of therapy.

Figure 2 Etanercept reduces bone marrow oedema in PsA. Reconstructed images of a series of 3D MRI from patient 8 at baseline, 6 and 18 months after continuous etanercept therapy are show with anterior–posterior view (top) and lateral–medial view (bottom).

Correlation between OCP frequency and other measures

No significant correlation was detected between the change in OCPF and the change in BMO volume (r = 0.15; p = 0.6), or between the change in OCPF and the change in clinical parameters (r = 0.05; p = 0.8), after 6 months of therapy. In the two patients noted to have an increase in OCPF at 6 months, the BMO volumes also increased. Of the two patients in whom the OCPF did not change after 6 months of etanercept therapy, the BMO volume increased in one patient but decreased in the other. In the one patient in whom the clinical joint scores worsened, the OCPF decreased and no post-therapy MRI was available. No significant correlation was noted between the change in tender, swollen and clinical joint scores and the change in BMO volumes.


We found that PsA patients with erosive arthritis demonstrated significant clinical improvement along with a rapid and significant reduction in the OCPF following etanercept therapy. The imaging data also revealed a reduction in BMO particularly at the wrist. An unanticipated finding was that the degree of BMO fluctuated following anti-TNF therapy even in patients who exhibited a marked clinical response. Taken together, these findings suggest that the rapid decline in OCP, coupled with the decrease in BMO after anti-TNF-α therapy may provide an additional mechanism to explain the protective effects of TNF blockade on inflammatory bone loss in PsA.

Osteoclasts are the pivotal cells in the pathology of erosions.22 In this trial, we were able to confirm, in a larger cohort of PsA patients, that the OCPF declines with anti-TNF therapy. The rate of decrease in OCP as early as 2 weeks after initiation etanercept therapy was striking and supports previous animal and in vitro studies that have found a central role for TNF in the potentiation of osteoclastogenesis.23 It is important to note, however, that despite the decline in the OCPF with etanercept therapy, the mean OCP number after 6 months of therapy was 11.1 per 106 PBMC. This is somewhat higher than the mean (SD) OCPF for healthy controls (3.7 (1.1) per 106 PBMC) as noted in our previous study.6 The higher number of circulating OCPs at the end of 6 months of etanercept therapy in the PsA patients may therefore suggest continued osteoclastogenic activity, albeit suppressed, in OCP populations.

BMO is thought to signify a pre-erosive lesion in this disease.15 16 A recent study suggests that the presence of BMO may also be predictive for development of erosions after a year in other types of inflammatory joint disease, including PsA and reactive arthritis.24 Comparatively, fewer studies than in RA, have reported on BMO in PsA patients.2528 Two of these studies describe possible distinct patterns for BMO in PsA and SpA.26 27 that differ from distribution of BMO in RA.

Marzo-Ortega et al recently reported on an open label trial of 18 patients with PsA, who exhibited a dramatic improvement in BMO, after four infusions of infliximab.29 In this study, MRI images of 18 joints (12 hand and 6 knees) identified 9 with BMO. Complete resolution of BMO was seen in seven of nine patients, an improvement in one and only one patient had no change after 14 weeks of therapy. These patients were on concomitant methotrexate, the extent of baseline BMO was not reported, and scoring was based on a scale used to analyse serial images. In another study by the same group, treatment with etanercept resulted in a dramatic decrease in BMO at sites of axial and peripheral entheseal lesions in patients with spondyloarthropathy.30

Several possibilities may explain the differences between our findings and those reported in the two studies by Marzo-Ortega. We found BMO in all our patients but only patients with erosions on plain radiographs were recruited. In addition, we applied a quantitative method of analysis to measure BMO in PsA patients before and after etanercept. Lastly, the differences in BMO may related to differences in therapies (etanercept vs methotrexate plus infliximab) but this seems less likely given that both treatment regimens have been shown to effectively inhibit progression on bone erosions on plain radiographs.31 32

An overall improvement was noted in the extent of BMO, as measured by the number of sites, BMO volumes for individual joints and for total volume of BMO in all joints, after 6 months of etanercept therapy. The improvement, as measured by volume, reached significant levels at the wrists. The lack of significant improvements for the other individual joints was not surprising given that the study was not powered to detect these changes. The lack of statistical significance for improvement in BMO for all joints is not clear. Again, this may have been due to the small number of patients in the study. Alternately, as the MRI findings in patient 8 (fig 2) suggests that a longer duration of anti-TNF therapy is required to demonstrate a significant decrease in BMO.

Interestingly, the improvement in the extent of BMO was often not widespread. The finding that changes in BMO at different sites were independent, even within a single joint, is intriguing. In fact, only two patients were noted to have improvement of BMO at all sites. Another notable finding was that despite a substantial decline, BMO persisted at many sites after 6 months of therapy. This observation supports the concept that dynamic changes continue to occur within the bone albeit at a less intense level, in patients treated with a TNF antagonist. Emery et al noted persistent synovitis and BMO measured by MRI in RA patients considered to be in clinical remission on DMARDs.33 We recently demonstrated the presence of BMO in the knees of TNF-transgenic mice with inflammatory erosive arthritis.34 A striking finding was the presence of inflammatory cell infiltrates in the subchondral bone of the mice that over-expressed TNF but not in the littermates without the transgene. The areas with the highest BMO signal had the most intense cellular infiltrates. In support of our data is a recent study that demonstrated inflammatory infiltrates in the bone marrow at sites of BMO in RA patients.35 Collectively, these results suggest that residual BMO in PsA patients treated with DMARDS or biologic agents represents ongoing bone marrow cellular infiltration and not just oedema. Alternatively persistent MRI signals may represent altered bone remodelling or enhancement derived from an enriched vascular supply.

A major limitation of our study was the small number of individual joints imaged. Regrettably, serial images from seven patients had to be excluded for technical flaws that precluded analysis. Serial MRI analysis of one specific joint (knee or wrist) in all subjects rather than the most involved joint clinically would have provided us with more power to detect significant changes. The identification of a single joint, however, is a formidable challenge in PsA, because unlike RA, joint inflammation is frequently asymmetric and oligoarticular. The ability to quantify BMO in a three-dimensional format, however, was a major advantage over traditional approaches that rely on subjective scoring of sites based on consecutive analyses of serial slices. Recent developments in medical imaging that utilise novel cross-sectional imaging techniques allow the reconstruction of 3D structures within the body and thereby permit precise measurements to be made.36 Moreover, this approach can more readily capture the entire field of oedema for quantification and have been applied to measure lesion burden in patients with multiple sclerosis, Alzheimer’s disease and malignancy.3739 Contrast enhanced 3D MRI has also been shown to be a sensitive method for detecting both soft tissue lesions and early erosions in arthritis.40 We used a similar 3D image to demarcate and measure the extent of BMO.

The poor correlation between OCPF and BMO may relate to the method that was used to identify the precursor population. It is well known that osteoclast nuclear numbers can range from 3–30 or more. Our scoring system only tabulates osteoclasts and does not take into account cell size or the number of nuclei. With this technique, patient PBMC cultures with different frequencies of circulating OCP in vivo may look very similar in culture depending on their propensity to form polykaryons. Some investigators have addressed this issue by measurement of total osteoclasts area.41

Our study demonstrated a rapid and significant reduction in OCPF with etanercept therapy. This study also showed an overall reduction in the extent of BMO, but with fluctuating BMO levels even within individual joints. The cellular events that underlie bone remodelling in PsA are not well understood, but the unanticipated finding that the extent of BMO fluctuates following anti-TNF therapy even in patients who exhibit a marked clinical response, suggests that bone remodelling is a dynamic process that is not always coupled to inflammation. We did not detect significant correlations between BMO, OCP and clinical parameters, and only a weak association was noted between the change in number of sites with BMO and change in OCP scores. Larger studies, possibly restricted to a single joint, however will be required to uncover any potential association between BMO, change in OCPF and clinical outcome measures. We conclude that OCPF and BMO are attractive candidates for markers of disease activity in PsA and that the rapid decline in OCP, coupled with the change in BMO after etanercept therapy may provide an additional mechanism to explain the protective effects of TNF blockade on inflammatory bone loss in PsA.



  • Funding: This study was funded by Amgen Inc., Thousand Oaks, California, USA

  • Competing interests: None declared.