Article Text

Download PDFPDF

Extended report
MRI versus conventional measures of disease activity and structural damage in evaluating treatment efficacy in juvenile idiopathic arthritis
  1. C Malattia1,2,3,4,
  2. A Consolaro1,
  3. S Pederzoli1,
  4. A Madeo1,
  5. A Pistorio3,
  6. M Mazzoni1,
  7. C Mattiuz2,
  8. GM Magnano2,
  9. S Viola1,
  10. A Buoncompagni1,
  11. E Palmisani1,
  12. R Hasija1,
  13. N Ruperto1,
  14. A Ravelli1,2,3,4,
  15. A Martini1,2,3,4
  1. 1Istituto Giannina Gaslini, Pediatria II, Genova, Italy
  2. 2Istituto Giannina Gaslini, UO Radiologia, Genova, Italy
  3. 3Istituto Giannina Gaslini, Servizio di Epidemiologia e Biostatistica, Genova, Italy
  4. 4Dipartimento di Scienze Pediatriche, Università degli Studi di Genova, Genova, Italy
  1. Correspondence to Clara Malattia, Pediatria II, Istituto Giannina Gaslini, Largo G Gaslini 5, Genova 16147, Italy; claramalattia{at}


Objective To compare the American College of Rheumatology paediatric (ACRp) response criteria and conventional radiography with MRI findings in a cohort of patients with juvenile idiopathic arthritis.

Methods Forty consecutive patients (30 girls, 10 boys; median age 10.8 years) with arthritis of the wrist starting treatment with disease-modifying antirheumatic drugs or biological agents were recruited. At 1-year follow-up the treatment response was assessed by ACRp criteria and radiographic progression using the adapted Sharp/van der Heijde method. Wrist MRIs were evaluated using both the paediatric-MRI and the OMERACT rheumatoid arthritis MRI scores. Sensitivity to change of clinical and imaging variables was assessed by standardised response mean (SRM) and relative efficiency (RE) was used to compare SRMs.

Results ACRp90 responders showed a significantly higher decrease in MRI synovitis score (median change −4) than non-responders (median change 0), ACRp30–50 responders (median change 0) and ACRp70 responders (median change −1) (p=0.0006, Kruskal–Wallis test). Non-responders showed significantly higher radiographic progression than ACRp90 responders (pB=0.016). The MRI synovitis score showed a greater responsiveness to change (SRM 1.69) compared with the majority of ACR core set of variables. MRI erosion scores were less responsive than conventional radiography in detecting destructive changes (RE <1). MRI follow-up revealed no signs of inflammation in four out of 24 wrists with clinically inactive disease.

Conclusion Only ACRp90 responders showed a significant decrease in synovitis and the halting of structural damage, suggesting that levels of response higher than ACRp30 are more appropriate for assessing drug efficacy. The excellent responsiveness of MRI and its ability to detect subclinical synovitis make it a promising outcome measure.

Statistics from


Juvenile idiopathic arthritis (JIA) is the most common chronic paediatric rheumatic disease and an important cause of acquired disability.1 Recent advances in therapeutics have made available novel immune response-modifying drugs, such as biological agents,2,,6 that have dramatically changed the outcome of children with JIA and significantly decreased the risk of permanent structural damage. The trend towards early introduction of these effective structure-modifying treatments has shifted the focus away from radiographically detectable structural damage and has generated the need for alternative imaging modalities that are more sensitive in detecting pre-erosive inflammatory changes in order to stratify patients for treatment and to monitor treatment efficacy more effectively.

Response to treatment is currently assessed using validated outcome measures and is calculated as the percentage of change of a core set of clinical variables.7 The American College of Rheumatology paediatric (ACRp) 30 definition of improvement has been accepted by regulatory agencies for drug registration as the primary outcome in clinical trials in JIA.8

By directly imaging synovitis, contrast-enhanced MRI has an advantage in assessing response to treatment over clinical outcome variables which are all surrogate markers of synovium inflammation. In recent years MRI has increasingly been used as an outcome measure in clinical trials in rheumatoid arthritis (RA) as well as in clinical practice.9,,12 Even though a large amount of evidence has accumulated on the value of MRI in monitoring the effect of treatment, this has not been explored in JIA. Standardised and validated methods of evaluating MRI findings in JIA have progressed and validated paediatric scales are available.13 ,14 MRI measures of synovitis, bone marrow oedema and bone erosions have been shown to be valid and reliable for the assessment of disease activity and damage in cross-sectional and observational studies,13 ,15 but less is known about the responsiveness of these measures during an interventional study.

The main purpose of this study was to compare conventional measures of evaluating treatment efficacy with MRI findings in a cohort of patients with JIA starting treatment with second-line agents (disease-modifying antirheumatic drugs (DMARDs) or biological agents). We also compared the responsiveness to change of MRI features with those of the core set of variables included in the ACRp criteria7 and with radiographic measures of structural damage.



This study included all consecutive patients with JIA according to the International League of Associations for Rheumatology revised criteria16 and wrist arthritis who started treatment with second-line agents between January 2009 and September 2010. Patients requiring general anaesthesia or those with a contraindication to MRI were excluded from the study.

The clinically more affected wrist was studied with MRI and conventional radiography together with a standard clinical examination immediately before starting new treatment and after 1 year. The study was approved by the Gaslini institutional review board and informed written consent was obtained from all participants.

Clinical assessment

The following ACRp core set of variables were recorded: physician's global assessment of overall disease activity; number of active joints; number of joints with limited range of motion; patient's functional ability as assessed by the childhood health assessment questionnaire17; parents' global assessment of the child's well-being; erythrocyte sedimentation rate (Westergren method); and C reactive protein level (nephelometry). The clinical response was assessed using the ACRp definition of improvement. ACRp307 was defined as an improvement of at least 30% in three of the six core set of variables, with no more than one of the remaining variables worsening by more than 30%. For ACRp50, ACRp70 or ACRp90, improvement had to be at least 50%, 70% or 90% respectively, with no more than one of the core outcome variables deteriorating by 30% or more. At the 1-year follow-up visit the imaged wrist was clinically evaluated for the presence of signs of active disease, defined by the presence of swelling or, in its absence, by the presence of pain or tenderness and limitation of motion.18 All clinical measures were assessed by two paediatric rheumatologists (SV and AB) with more than 20 years of experience in the field, who were blind to the results of the imaging procedures. Clinical assessment was performed on the same day as MRI and repeated at follow-up evaluation by the same rheumatologist for each patient.

Imaging assessment

Conventional x-rays of the wrist in the posteroanterior view were assessed by a paediatric rheumatologist (AR) with more than 20 years of experience in the field according to the adapted version of the Sharp/van der Heijde method.19 This score is based on the assessment of joint space narrowing and bone erosions on a 0–4 and a 0–5 point severity scale, respectively. The follow-up x-ray was scored without reference to baseline, and the radiographic progression was determined by calculating the change in the Sharp/van der Heijde score between baseline and follow-up x-rays (figure 1).

Figure 1

(A) Baseline and (B) follow-up wrist x-rays in an 8-year-old boy with systemic juvenile idiopathic arthritis showing radiographic progression over 1 year.

MRI was performed on a 1.5 Tesla MRI scanner (Achieva Intera, Philips Medical Systems Best, The Netherlands) using a Sense Flex Small Coil with a field of view from the distal radioulnar joint to the metacarpal bases. The imaging protocol consisted of a 3D TSE T1 (TR 600 ms, TE 10 ms, slice thickness/gap (mm) 1/0, matrix 176×175), a TSE T2 fat sat ‘SPAIR’ (TR 2000 ms, TE 70 ms, slice thickness/gap (mm) 3/0, matrix 272×217) and a 3D GRE ‘SPIR’ (TR 40 ms, TE 7 ms, slice thickness/gap (mm) 1, matrix 304×272), acquired immediately after the injection of 0.1 mmol/kg body weight of contrast medium (gadoterate meglumine, Gd-DOTA, Dotarem Guerbet, Villepinte, France).

Erosions were scored at 15 sites within the carpus (carpal bones, distal radius, distal ulna and metacarpal bases) using both the paediatric 0–4 scale (0=no erosion; 1=1–25% of the bone eroded, etc)13,,15 and the rheumatoid arthritis MRI scoring system (RAMRIS) 0–10 scale (0=no erosion; 1=1–10% of the bone eroded, etc).13 ,21 Bone marrow oedema was evaluated using the paediatric 0–2 scale (0=no oedema, 1=<50% of the bone involved, 2=≥50% of the bone involved) on the same site as for bone erosions. Synovitis was assessed on post-contrast images using a 0–3 scale based on the simultaneous assessment of the visual intensity of the synovium enhancement and the extent of inflammation at the distal radioulnar joint, radiocarpal joint, midcarpal joints and carpometacarpal joints 2–5.13 ,21

MRIs were read in unknown chronological order by a paediatric radiologist (C Mattiuz) and a paediatric rheumatologist (C Malattia) with 5 years expertise in musculoskeletal MRI and any disagreement was resolved by consensus. The readers were blind to the clinical status of the patient.

Statistical analyses

Non-normally distributed continuous data or ordinal variables were summarised as medians and first and third quartiles. Comparisons of quantitative variables among groups were made using the non-parametric analysis of variance (Kruskal–Wallis test); the Mann–Whitney U test adjusted according to Bonferroni's correction (pB) was chosen to assess the statistical significance of differences between pairs of patient groups.

For the assessment of discriminant validity, patients were divided by their maximum level of improvement into four mutually exclusive groups: non-responders, ACRp30–50, ACRp70 and ACRp90 responders; ACRp30 and ACRp50 responders were grouped because of the small numbers in the groups. Responsiveness to clinical change was assessed by computing the standardised response mean (SRM), calculated as the mean change in score from baseline to endpoint divided by the SD of the individuals' change in score.22 According to Cohen, the threshold levels for SRM were defined as follows: ≥0.20=small; ≥0.50=moderate; ≥0.80=good.23

The relative efficiency (RE) in relation to swollen joint count (SJC) for the measures reflecting inflammation and disease activity was computed. The RE has been calculated as the ratio of the squared SRM for the outcome to the squared SRM for the SJC in the responders.11 ,22 RE values >1 imply that the outcome is more efficient than the SJC in detecting inflammatory changes. For the measures reflecting damage, the adapted Sharp/van der Heijde score was used as a reference for calculating the RE. All statistical tests were two-sided and p<0.05 was considered statistically significant. The statistical package used was Statistica (StatSoft, Tulsa, Oklahoma, USA).


A cohort of patients with JIA (N=40, median age 10.8 years, median disease duration 4 years) who started treatment with second-line agents (based on clinical indications) was included in the present study. Eleven patients (27.5%) had systemic arthritis, 15 (37.5%) had polyarthritis (1 rheumatoid factor-positive) and 14 (35%) had extended oligoarthritis (12 out of 13 were antinuclear antibody-positive). Fourteen patients (35%) started DMARDs (methotrexate) while 26 (65%) started biological agents (20 etanercept, 2 adalimumab, 2 anakinra, 2 tocilizumab); 9 of these 26 patients were already receiving treatment with methotrexate. Seven patients (17.5%) were also receiving systemic corticosteroid therapy while 17 (42.5%) were taking non-steroidal anti-inflammatory drugs. Table 1 summarises the results of clinical and imaging assessments. At 1-year follow-up, 28 of the 40 patients showed improvement while 12 were non-responders. The frequency of ACRp30 (N=28), ACRp50 (N=25), ACRp70 (N=24) and ACRp90 (N=14) response at 1 year was 70%, 62.5%, 60% and 35%, respectively.

Table 1

Baseline and 1-year follow-up demographic, clinical and imaging findings of the 40 patients with juvenile idiopathic arthritis

The changes in MRI synovitis score in relation to the ACRp response categories are shown in figure 2. ACRp90 responders (N=14) showed a significantly higher decrease in MRI synovitis score (median change −4) compared with ACRp30–50 responders (median change 0), ACRp70 responders (median change −1) and non-responders (median change 0) (p=0.0006, Kruskal–Wallis test). Post hoc analyses showed the ability of the MRI synovitis score to discriminate between ACRp90 and ACRp30–50 responders and between ACRp90 and non-responders (pB=0.048 and pB=0.002, respectively).

Figure 2

Changes in the MRI synovitis score between baseline and 1-year MRI follow-up evaluation in different categories of American College of Rheumatology (ACR) responders (non-responders, N=12; ACRp30–50, N=4; ACRp70, N=10; ACRp90, N=14). Black circles represent median values, boxes indicate first and third quartiles, whiskers indicate minimum and maximum values. The MRI synovitis score was able to discriminate between different levels of ACRp response (p=0.0011, Kruskal–Wallis test); in particular, the change in MRI synovitis score was greater in ACRp90 responders than in ACRp30–50 responders (pB=0.048) and in ACRp90 responders than in non-responders (pB=0.002).

Changes in the adapted Sharp/van der Heijde score were statistically different among the ACRp response categories (p=0.0065, Kruskal–Wallis test). Non-responders showed significantly greater erosive progression than ACRp90 responders (pB=0.016; figure 3). ACRp30–50 and ACRp70 responders did not differ significantly from non-responders in terms of erosive progression. Only patients who met the definition of improvement according to the ACRp90 criteria did not experience erosive progression and even showed a reversal of radiographic damage.

Figure 3

Changes in the adapted version of the Sharp/van der Heijde score between baseline and 1-year follow-up in patients grouped by American College of Rheumatology paediatric (ACRp) response categories (non-responders, N=12; ACRp30–50, N=4; ACRp70, N=10; ACRp90, N=14). Black circles represent median values, boxes indicate first and third quartiles, whiskers indicate minimum and maximum values. Changes in the Sharp/van der Heijde score were statistically different among the ACRp response categories (p=0.0065, Kruskal–Wallis test). Non-responders showed significantly greater erosive progression, as assessed by the adapted Sharp/van der Heijde score, than patients who achieved ACRp90 criteria (pB=0.016).

The MRI synovitis score showed an excellent responsiveness to change (SRM 1.69; figure 4). Figure 5 shows the REs in relation to the SJC for the MRI synovitis score and for the clinical measures reflecting disease activity included in ACRp response criteria. With the exception of the physician's global assessment of disease activity, the MRI synovitis score showed a higher responsiveness to change than the other ACRp core set of variables.

Figure 4

Contrast-enhanced coronal 3D GRE ‘SPIR’ T1-weighted with fat saturation in a 15-year-old girl with juvenile idiopathic arthritis and wrist involvement. (A) Presence of synovitis in the radioulnar, radiocarpal and intercarpal joint recesses. (B) Marked reduction of synovitis after 1 year of treatment with anti-tumour necrosis factor α.

Figure 5

Relative efficiencies (RE) of the American College of Rheumatology (ACR) core set of variables reflecting disease activity and of the MRI synovitis score with respect to swollen joint count (SJC) used as a reference value (RE=1; black column). The RE is calculated as the ratio of the squared standardised response mean (SRM) for the outcome variables to the squared SRM for the SJC. RE values >1 imply that the outcome is more efficient than the SJC in detecting inflammatory changes. C-HAQ, Childhood Health Assessment Questionnaire; ESR, erythrocyte sedimentation rate; LJC, limited joint count; NAA, number of active joints; SJC, swollen joint count, TJC, tender joint count.

Both the paediatric (range 0–4) and the RAMRIS (range 0–10) bone erosion scores showed a low responsiveness to change (SRM 0.5 and SRM 0.4, respectively). RE values of the paediatric and the RAMRIS bone erosion scores in relation to the adapted Sharp/van der Heijde score were <1, indicating that MRI is less efficient than conventional radiography in detecting destructive changes over 1 year.

At 1-year follow-up, 22 of the 40 patients (55%) showed no clinical signs of active arthritis at the imaged wrist. Clinical examination was positive in 18 (45%) and was associated with higher MRI synovitis score grades (median 3) compared with a negative clinical examination which was associated with lower MR grades (median 2.5; p=0.028). The 1-year follow-up MRI revealed no features of disease activity in only four of the 40 patients (10%).


This study is the first to show that the MRI synovitis score represents a promising imaging biomarker for measuring the therapeutic response in JIA. We found that a significant decrease in joint synovitis occurred only in patients who achieved the higher level of ACRp response criteria. Patients who improved according to the ACRp30–50 and ACRp70 criteria did not, in fact, differ significantly from those in whom the synovitis did not improve over time. This is in accordance with the results of Graham et al24 who found that the percentage change in wrist synovial volume over time in patients with JIA starting disease-modifying treatment did not differ significantly between ACRp30 responders and non-responders. A poor concordance between clinical and MRI measures of synovitis, with persistent MRI features despite improvements in active joint counts, has also been reported by Ostergaard et al in patients with RA starting treatment with anakinra.10

Notably, in the present study the persistence of inflammation was significantly associated with a higher risk of progression in radiographic damage. Only the ACRp90 responders did not experience erosive progression and even showed an improvement in their radiographic erosion score. A reversal in radiographic damage—most likely as a result of a combination of cartilage repair and halting of radiographic progression—was not entirely unexpected as it has previously been reported in patients with JIA treated with antitumour necrosis factor agents.25

This study raises a number of issues that deserve further comment. To date, the primary endpoint in major studies evaluating the efficacy of conventional DMARDs26 or biological agents2 ,3 ,5 ,27 has been ACRp30. Since only the higher levels of clinical response were associated with a significant decrease in synovitis and the halting of structural damage, we question whether it is time to move towards higher levels of response to assess drug efficacy. Other authors have recently argued that the ACRp30 is too low a benchmark of improvement to serve as a primary efficacy outcome, especially since therapeutic advances have increased the expectations of treatment benefits28 with disease remission now becoming a realistic goal of treatment.29,,32

MRI, by directly imaging synovitis, has an intuitive advantage over clinical surrogate measurements of inflammation for the assessment of treatment efficacy, allowing a comparison of the absolute response between single patients and groups of patients. Notably, MRI measurement of synovitis, which provides information limited to the inflammatory status of a single joint, was more responsive than most ACRp outcome variables which reflect global inflammation. This observation, together with the encouraging correlations between MRI measurement of inflammation of the wrist and conventional measures of disease activity,13 ,24 ,33 argues for the potential utility of changes in the inflammatory status of the wrist as a marker for evaluating the systemic response. Since one of the major limitations of MRI is its ability to explore only a few joints per session, the impact of such an ‘informative’ joint is intuitive.

Unlike synovitis, the responsiveness to change of the MRI erosion scores was poor, as expected in an intervention study with disease-modifying drugs aimed at halting the progression of structural damage. Although previous studies in JIA have shown that MRI has a higher sensitivity in detecting early erosive damage than other imaging modalities,15 we did not find a clear advantage in the use of MRI compared with conventional radiography when evaluating structural damage progression over 1 year, substantiating previous concerns in RA that MRI is not sensitive to long-term progression.12 ,34 Since we did not collect imaging data at 3 or 6 months after the beginning of treatment, we cannot exclude the potential value of MRI versus conventional radiography in detecting erosive changes over short periods, as suggested in RA studies.12

The fact that cartilage assessment, which is included in the radiographic scoring system but not in the MRI scoring system, has a main role in the progression of structural damage in JIA19 might explain the higher sensitivity of conventional radiography in detecting structural changes, emphasising the need to include cartilage assessment in MRI scoring systems.

Finally, in line with previous studies,35,,37 our results confirm the high sensitivity of MRI in detecting synovitis in the clinically inactive wrist (subclinical disease) and provide additional evidence to support the use of imaging to ensure accurate disease activity assessment, particularly in patients with low disease activity state. In this respect, ultrasound has also been shown to be promising in identifying subclinical synovitis in patients with RA and JIA.35 ,38 However, its application in monitoring JIA is hampered by the absence of standardised definitions of ultrasound pathologies and of validated scoring systems in JIA.

The lack of data from age-matched healthy controls is a potential limitation of this study. Data on the MRI appearance of the synovium in healthy children are needed to elucidate the prognostic significance of subclinical synovitis in JIA and to provide a consistent definition of imaging remission, in view of the fact that studies on adults have reported low-grade synovitis in healthy controls.39 Moreover, the high prevalence of bony depression, bone marrow oedema and joint effusion recently described in healthy children40 further emphasises the need to include age-matched healthy controls in MRI studies in JIA. We did not evaluate a control group since the study was not powered to detect differences between treatments but to compare imaging and clinical evaluations for the assessment of treatment response. Finally, the results of the present study may not be applicable to all patients with JIA but only to those with wrist involvement. This joint is usually involved in patients with a polyarticular course who are more likely to develop progressive destructive disease.41 Further studies designed to analyse data stratified according to treatments or JIA subtype are needed to better define the value of MRI in a clinical trial setting.

In conclusion, we report data that strongly suggest that ACRp30 can no longer be considered a sufficient therapeutic response. The excellent sensitivity of MRI in detecting inflammatory changes makes this imaging technique a promising outcome measure for the assessment of treatment efficacy. A more comprehensive MRI scale including cartilage assessment might increase the sensitivity of MRI for testing the disease-modifying potential of new antirheumatic drugs. Our results confirm that clinical criteria are insufficiently sensitive to detect a low level of inflammation and provide further evidence of the need to include imaging for a more accurate definition of remission state.


The authors wish to acknowledge the technical staff of the Radiological Department of the G Gaslini Institute, in particular Mrs Francesca Maiuri, Mr Paolo Del Mirto, Mr Stefano Franceschi and Mr Marco Ciccone, who supervised the MRI scans.



  • Competing interests None.

  • Ethics approval Ethics approval was obtained from the Gaslini institutional review board.

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

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.