Article Text

Concise report
Prediction of MRI erosive progression: a comparison of modern imaging modalities in early rheumatoid arthritis patients
  1. Pernille Bøyesen1,
  2. Espen A Haavardsholm1,
  3. Désirée van der Heijde1,2,
  4. Mikkel Østergaard3,
  5. Hilde Berner Hammer1,
  6. Sølve Sesseng4,
  7. Tore K Kvien1
  1. 1Department of Rheumatology, Diakonhjemmet Hospital, Oslo, Norway
  2. 2Department of Rheumatology, Leiden University Medical Centre, Leiden, The Netherlands
  3. 3Department of Rheumatology, Copenhagen University Hospitals at Hvidovre and Gentofte, Copenhagen, Denmark
  4. 4Department of Radiology, Diakonhjemmet Hospital, Oslo, Norway
  1. Correspondence to Dr Pernille Bøyesen, Department of Rheumatology, Diakonhjemmet Hospital, Box 23, Vinderen, N-0319 Oslo, Norway; pernilleboyesen{at}


Objectives To examine the associations between modern imaging modalities and joint damage measured as 1-year MRI erosive progression, in early rheumatoid arthritis (RA) patients.

Methods 84 RA patients with disease duration of less than 1 year were included in this inception cohort. Patients were evaluated at baseline, 3, 6 and 12 months by core measures of disease activity, MRI and ultrasound grey-scale (USGS) of inflammation, conventional radiography and digital x-ray radiogrammetry (DXR) bone mineral density (BMD) of cortical hand bone.

Results 53 of the 79 patients (67%) who completed the follow-up had MRI erosive progression (dependent variable). USGS and MRI bone marrow oedema (BME) were in multivariate analyses independent predictors of 1-year MRI erosive progression. There was a trend towards higher MRI synovitis score and 3-month DXR BMD loss in patients developing MRI erosions. On an individual level, USGS inflammation, MRI synovitis and MRI BME also somewhat better predicted outcome than rheumatoid factor, anticitrullinated protein antibodies and disease activity score 28.

Conclusions USGS inflammation and MRI BME were independent predictors of MRI erosive progression in early RA patients on a group level. The exact prognosis of the individual patients could not be determined by imaging alone.

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.

Rheumatoid arthritis (RA) is a chronic inflammatory disease characterised by synovitis, which may lead to bone damage. Modern imaging techniques have across several studies predicted joint damage in RA patients. MRI bone marrow oedema (BME) has been shown to be an independent predictor of joint damage.1 2 Furthermore, bone damage assessed as a change in digital x-ray radiogrammetry (DXR) cortical bone mineral density (BMD) has been shown to predict radiographic progression in RA patients.3 The predictive value of ultrasonography is less certain. However, ultrasound measured inflammation has been shown to be correlated with progression in radiographic damage, suggesting a possible predictive role for joint damage in RA patients.4 5

Despite the increasing knowledge of the associations between individual imaging methods and joint damage in early RA, little is known about their comparative value and how they perform on an individual level.

In this study we wanted to examine the associations between modern imaging modalities (MRI, ultrasonography, DXR) and joint damage measured as 1-year MRI erosive progression, in early RA patients, and to assess the individual predictive value of joint damage on a group and individual level.

Patients and methods


This study included 84 early RA patients from a previously described inception cohort.1 In short, the patients had disease duration of less than 1 year and fulfilled the American College of Rheumatology classification criteria for RA.6 Seventy-nine of 84 patients completed the 1-year follow-up. The patients were treated according to clinical practice. Visits were scheduled at baseline, 3, 6 and 12 months, and included 28 swollen and tender joint counts, patient-reported health status by questionnaires, blood samples and imaging procedures (MRI, ultrasonography and conventional radiographs). Erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), IgM rheumatoid factor (RF) and anticitrullinated protein antibodies (anti-CCP) were analysed.1 The disease activity score in 28 joints (DAS28) was computed based on 28 joint counts and ESR.7

Imaging assessments

MRI of the dominant wrist was obtained using a 1.5 T MRI scanner (General Electrics Signa, Milwaukee, Wisconsin, USA) with a dedicated wrist coil. The MRI acquisition and assessments have previously been described elsewhere.8 MRI synovitis, BME, erosions and tenosynovitis were scored according to the RA MRI score (RAMRIS) and MRI tenosynovitis was scored by a trained reader (EAH).9,,11

Ultrasound grey-scale (USGS) of the dominant wrist was performed by a trained user (HBH) using an 8–16 MHz linear array transducer on a Diasus machine (Dynamic Imaging, Scotland, UK) without power Doppler function. The radiocarpal joint was assessed in the dorsal midline, with a longitudinal scan for USGS synovitis defined as abnormal, hypoechoic, poorly compressible intra-articular material.12 The extensor tendons were viewed in radial (compartments 1–2), dorsal midline (compartments 3–5) and ulnar (compartment 6) transversal scans and the flexor tendons in a transversal palmar midline scan. The scanned tendon areas were assessed for USGS tenosynovitis defined as a hypoechoic, poorly compressible, thickened tendon.12 All findings were graded as 0=none, 1=mild, 2=moderate or 3=marked.13 The five assessed locations were dichotomised into present versus not present and summed into a USGS inflammation score ranging from 0 to 5. Only 70 (83%) patients had a baseline ultrasound examination due to logistical problems. Last observation backwards was used to replace missing ultrasound inflammation scores.

Based on digital conventional radiographs of the hands taken at baseline and 3 months follow-up, hand BMD was measured by DXR (Sectra, Linkjöping, Sweden). The images were analysed according to a computerised version of the traditional radiogrammetry, and mean values of both hands were used in the current analyses.14 15

Digital conventional radiographs of hands and wrists were scored according to the van der Heijde modified Sharp score by a trained observer (PB).16

Statistical analyses

All statistical analyses were performed using SPSS 14 statistics package. Baseline characteristics were described by median IQR for continuous variables and percentage (number) for counts. Group comparisons were performed by Mann–Whitney U test and χ2 tests.

MRI erosive progression was defined as a one or more unit increase in 1-year RAMRIS erosive change. Possible associations between imaging modalities (independent variables) and 1-year MRI erosive progression (dependent variable) were explored by logistic regression analyses. Age, sex and independent variables with univariate associations with a p value of 0.25 or less were included in the multivariate analyses. The final multivariate model with independent predictors of 1-year change in MRI erosions was obtained by stepwise exclusion of the least significant variable until only significant variables were left.

Receiver operating characteristic analyses were performed to define cut-offs for possible predictors reflecting MRI erosive progression. Furthermore, the predictive value at the patient level was evaluated by sensitivity, specificity, positive and negative predictive values, likelihood ratios of positive and negative tests and accuracy.

All tests were two-sided and p values of 0.05 or less were considered statistically significant. Standard diagnostic tests of model assumptions were routinely performed.


Median (IQR) age was 58 (47–67) years, disease duration 107 (70–188) days, 65 patients (77%) were female and 46 (55%) were anti-CCP positive (table 1). One-year median (IQR) change in MRI erosions was 1 (0–3) and 53 patients (67%) developed MRI erosive progression during the 1-year follow-up.

Table 1

Baseline characteristics of patients with and without 1-year MRI erosive progression, and associations between baseline characteristics and MRI erosive progression (univariate logistic regression analyses)

Levels of MRI BME and USGS inflammation were statistically significantly higher in the patients with progression in MRI erosions with OR of 1.24 and 2.15, respectively (table 1). There was a trend towards higher levels of MRI synovitis and DXR BMD loss in patients with 1-year MRI erosive progression.

USGS inflammation and MRI BME were in the final multivariate analyses independent predictors of MRI erosive progression on a group level, with OR of 2.01 and 1.28, respectively (table 2). The same final multivariate results were reproduced with MRI erosive progression defined as a two or more unit increase in 1-year RAMRIS erosive change and when only including the patients with baseline USGS inflammation scores (n=55). The results are illustrated in supplementary figure 1 (available online only).

Table 2

Independent predictors of 1-year MRI erosive progression (multivariate logistic regression analysis)

Furthermore, the predictive value of modern imaging modalities was tested on an individual level (table 3). USGS inflammation, MRI synovitis and MRI BME performed slightly better than anti-CCP, RF and DAS28. USGS inflammation performed best, with a positive likelihood ratio of 1.75 and an accuracy of 70%.

Table 3

Test performance of imaging modalities and measures of disease severity in identifying early RA patients at risk of developing subsequent MRI erosions


In this 1-year follow-up study of early RA patients, we compared MRI, USGS and DXR as predictors of joint damage. We found that USGS inflammation and MRI BME were independent predictors of 1-year MRI erosive progression on a group level. On an individual level, the imaging modalities were slightly, but not dramatically, better than the clinical and demographic variables.

Ultrasonography measured inflammation has previously been shown to be correlated with progression in radiographic damage, suggesting a possible predictive role in RA patients.4 5 As far as we know, this study shows for the first time that USGS inflammation is an independent predictor of subsequent joint damage in early RA even when including BME in the predictive model. USGS synovitis and USGS tenosynovitis did not separately influence the MRI erosive progression (OR 7.2; 95% CI 0.9 to 61.0 and OR 1.7; 95% CI 0.4 to 7.3, respectively). Limited statistical power in this small study emphasises the need for larger studies to evaluate their individual importance.

MRI findings of BME in RA has contributed to understanding the importance of bone marrow in the development of bone damage.17 In this study we confirm previous findings of MRI BME as an independent predictor of joint damage.1 2

Change in cortical hand BMD has previously been shown independently to predict subsequent radiographic damage in RA patients.3 These findings were not confirmed in our study, even though we found a trend towards larger 3-month DXR BMD loss in patients with MRI erosive progression. Analyses with 6 and 12 months DXR BMD change gave similar results as the 3-month change (data not shown).

Several factors, which have previously been identified as predictors of joint damage in RA (baseline radiographic erosions, ESR, CRP, IgM RF, anti-CCP and DAS28), did not reach statistical significance in this study and were excluded from the multivariate analyses. The number of included patients (n=84) is in this setting a large number taking into account the comprehensive data collection from imaging modalities. However, it still gives a rather low power to perform logistic regression analyses and might explain why known predictors were not confirmed in these analyses.

A major limitation of the study was the lack of ultrasound power Doppler assessment based on the previous findings of associations between ultrasound power Doppler and progression in radiographic joint damage, both in active and inactive RA disease.4 5 18 A second limitation was that USGS reliability was not built into the study design. In the event of reduced reliability, the statistical power decreases thus leading to an increased risk of type II error. As a consequence the reported results may be conservative estimates.

In this study we found that USGS inflammation and MRI BME were independent predictors of erosive progression on a group level. However, translating these promising findings into prognosis on the individual level proved to be more demanding.

Individual predictors are unlikely to reflect the entire multifaceted nature of RA. Future studies are needed to examine the prognostic role on an individual level of composite indices including modern imaging modalities.


The authors would like to thank research nurse Margareth Sveinsson for collecting clinical data, research coordinator Tone Omreng for organising the data collection and Dr Inge Olsen for statistical advice.


View Abstract

Supplementary materials


  • Funding This study was funded by the Eastern Norway Regional Health Authority, the Research Council of Norway, the Norwegian Rheumatism Association, the Norwegian Women Public Health Association, Grethe Harbitz Legacy and Marie and Else Mustad's Legacy.

  • Competing interests Johannes WJ Bijlsma was the handling editor for this manuscript.

  • Ethics approval This study was conducted with the approval of the regional ethics committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.