Differentiation between early rheumatoid arthritis patients and healthy persons by conventional and dynamic contrast-enhanced magnetic resonance imaging

Differentiation between early rheumatoid arthritis patients and healthypersons by conventional and dynamic contrast-enhanced magneticresonance imaging

MB Axelsen1,2, BJ Ejbjerg3, ML Hetland1,2,4, H Skjødt1, O Majgaard3, UB Lauridsen1, K Hørslev-Petersen5, M Boesen6,7,O Kubassova8, H Bliddal6, M Østergaard1,2

1Copenhagen Center for Arthritis Research, Center for Rheumatology and Spine Diseases, Glostrup University Hospital, 2Department ofClinical Medicine, Faculty of Health and Medical Sciences, Copenhagen University, 3Department of Rheumatology, Slagelse UniversityHospital, 4DANBIO Registry, Glostrup University Hospital, 5King Christian X Hospital for Rheumatic Diseases, Graasten, 6The ParkerInstitute, Frederiskberg University Hospital, 7Department of Radiology, Frederiksberg University Hospital, Denmark, and 8ImageAnalysis Ltd, London, United Kingdom

Objectives: To identify the magnetic resonance imaging (MRI) parameter that best differentiates healthy persons andpatients with early rheumatoid arthritis (RA), and to investigated responsiveness to treatment of various MRI parameters.Method:Conventional MRI and dynamic contrast-enhanced (DCE)-MRI of the hand were performed once for 26 healthypersons, and before and after 6 and 12 months of disease-modifying anti-rheumatic drug (DMARD) treatment for 14 earlyRA patients, using a 1.0-T MRI unit. One-slice DCE-MRI was analysed using Dynamika version 4.2. The number ofenhancing voxels (Nvoxel), the initial rate of enhancement (IRE), the maximum enhancement (ME), ME�Nvoxel, andIRE�Nvoxel were calculated for wrist and 2nd–5th metacarpophalangeal (MCP) joints. Conventional MR images wereevaluated using the RA MRI scoring system (RAMRIS) synovitis score.Results: Using DCE-MRI, enhancement was demonstrated in 61.5% of healthy persons and in 91.7% of RA patients atbaseline, with a median Nvoxel of 3 and 362, respectively. At baseline, all parameters were higher for patients than forhealthy persons (all p� 0.003). Only one patient had a baseline RAMRIS synovitis score below the 95th percentile of thehealthy persons. The corresponding number of patients was 3 for Nvoxel, ME�Nvoxel and IRE�Nvoxel, and 10 for IRE andME. The RAMRIS synovitis score and IRE showed the highest responsiveness, with a standardized response mean (SRM)of –1.00 and –0.88, respectively.Conclusions: The RAMRIS synovitis scoring of conventional MRI, and to a lesser extent the one-slice DCE-MRIparameters of synovial volume, differentiated early RA patients and healthy persons. The decrease in RAMRIS synovitisscore, Nvoxel, and IRE showed sensitivity to change during treatment.

Magnetic resonance imaging (MRI) is a more sensitivetool than conventional clinical evaluation for detectingsynovitis in rheumatoid arthritis (RA) patients. Moreover,signs of inflammation visualized by MRI are frequentlyfound in patients in areas where disease activity is notclinically evident (1).The semi-quantitative RA MRI scoring system

(RAMRIS), which was developed by the OutcomeMeasures in Rheumatology Clinical Trials (OMERACT),is widely used to score synovitis, bone marrow oedema(BME), and bone erosions in clinical trials. The RAMRISsynovitis and BME scores are reliable and responsive

outcome measures, but evaluation takes time and trainingof readers is necessary (2). The RAMRIS system is there-fore rarely used in the routine management of RA patients,and the development of quicker and easier methods for theevaluation of inflammation visualized by MRI would beuseful.

Dynamic contrast-enhanced MRI (DCE-MRI) can beused to analyse synovial enhancement following intrave-nous injection with a gadolinium (Gd)-containing contrastagent. The early enhancement rate in the synovium accu-rately reflects histological synovial inflammation (3–5) andis responsive to treatment (6–10). Traditionally, DCE-MRIhas been analysed for manually delineated areas ofenhancement, regions of interest (ROIs), covering smallexact areas within the synovial area with the most intenseenhancement. Use of these manually delineated ‘exactROIs’ is time-consuming and potentially compromisesintra- and inter-reader agreement (11, 12).

Mette Bjørndal Axelsen, Copenhagen Center for Arthritis Research,Center for Rheumatology and Spine Diseases, Glostrup UniversityHospital, Nordre Ringvej 57, DK-2600 Glostrup, Denmark.E-mail: [email protected]

Accepted 8 July 2013

Scand J Rheumatol 2013;iFirst article:1–10 1

© 2013 Informa Healthcare on license from Scandinavian Rheumatology Research Foundation

DOI: 10.3109/03009742.2013.824022 www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

The Dynamika software (www.imageanalysis.org.uk;Image Analysis Ltd, London, UK) has been developedfor a more user-independent analysis of DCE-MRI (13),and data can be extracted for a manually delineated ‘roughROI’, which includes the joint, but excludes enhancingvoxels due to large blood vessels or large artefacts. This‘rough ROI’ method is highly correlated to histologicalsigns of synovitis (14) and to RAMRIS synovitis and BMEscores (15). Furthermore, the intra- and inter-reader agree-ment is high for knee (14) and wrist (16) joints.Our hypothesis was that DCE-MRI, used with semi-

automated software, would be a feasible method thatbetter differentiated RA patients and healthy personsand was more responsive to treatment than conventionalMRI, using the RAMRIS system for quantification ofinflammation. The aim was to identify the conventionalor DCE-MRI parameter that would best differentiatehealthy persons from early RA patients. We also investi-gated the responsiveness of the MRI parameters todisease-modifying anti-rheumatic drug (DMARD) treat-ment in early RA patients. This was achieved by assessingsigns of synovitis in wrist and metacarpophalangeal(MCP) joints in healthy persons and early RA patientsby DCE-MRI analysed using Dynamika and by conven-tional MRI using the RAMRIS system.

Method

Patients

Weincluded14earlyRApatients [eightwomenand sixmen;median age 39 (range 26–63) years] and 26 healthy persons[16womenand10men;median age48 (range24–64) years].Patientswere all participants in theDanishCIMESTRA trial,a randomizedmulti-centre placebo-controlled clinical trial ofearly RA patients (17), and in an additional DCE-MRI pro-tocol performed at Copenhagen University Hospital atHvidovre. The RA patients received 12 months of treatmentwithmethotrexate, intra-articular betamethasone, and cyclos-porin or placebo–cyclosporin. Inclusion criteria were fulfill-ment of the American College of Rheumatology 1987criteria for RA, active disease for < 6 months, � 2 swollenjoints at baseline, and age between 18 and 75 years. Thehealthy persons were recruited among staff members, andtheir families, of the Departments of Rheumatology andRadiology at Copenhagen University Hospital at Hvidovre(18). They were without present or prior joint symptoms.Local ethics committee approval was obtained and thestudy complies with the Declaration of Helsinki of 1975 asrevised in 1983. All participants provided written informedconsent before participating. The trial was registered inClinicalTrials.Gov: reference NCT00209859.

Procedure

The MRI protocol consisted of conventional contrast-enhanced MRI and DCE-MRI of wrist and 2nd–5th

MCP joints of the non-dominant hand. MRI was per-formed at baseline and after 6 and 12 months of treat-ment for patients and once for healthy persons using a1.0-T Siemens Impact MRI unit (Erlangen, Germany)equipped with a circular polarized transmit–receiveextremity coil.

The MRI sequences included coronal short tau inversionrecovery (STIR) images, and coronal and axial T1-weightedspin echo (T1-SE) images before and after intravenous injec-tion of 0.1 mmol gadodiamide/kg bodyweight (Omniscan;GE Healthcare, Brøndby, Denmark). The DCE-MR imageswere obtained in one preselected coronal slice covering thewrist and 2nd–5thMCP joints and consisted of a series of 20T1-weighted fast low-angle shot MR images, each of 10 sduration. After four acquisitions, the contrast agent wasinjected intravenously as a bolus over approximately 10 s,and the following 16 acquisitions covered the initial 160post-contrast seconds. The static sequences were obtainedimmediately before and after the dynamic sequence. Thesame physician (BJE) performed all of the MRI scans.

The imaging parameters for the T1-SE images were asfollows: repetition time (TR) 600 ms, echo time (TE)15 ms, slice thickness 3 mm, field of view (FOV)109 � 145mm,matrix 192 � 256. For the STIR sequencethe following imaging parameters were used: TR 4500 ms,TE 30 ms, inversion time 150 ms, slice thickness 3 mm,FOV 109 � 145 mm, matrix 182 � 256. For the DCE-MRI the imaging parameters were as follows: TR 40 ms,TE 12 ms, flip angle 70˚, slice thickness 3 mm, FOV105 � 140 mm, matrix 192 � 256.

Conventional MRI

Conventional MR images were scored for synovitis (0–3per joint), BME (0–3 per bone), and bone erosions (0–10per bone) according to the OMERACT RAMRIS recom-mendations (19, 20). The sameRAMRIS-trained physician(BJE) performed scoring of all the conventional MRimages.

DCE-MRI

DCE-MR images were evaluated using Dynamika version4.2 (Image Analysis Ltd, London, UK). The software per-forms voxel-by-voxel analysis of enhancing voxels, auto-matically normalizes data to baseline images, and appliesmotion correction between the temporal slices to minimizeenhancement artefacts. We chose image frames 1–3 asbaseline to ensure that enhancement due to contrast wasnot present. Image frame 3 was used as reference formotion correction. ‘Rough ROIs’ were manually placeddelineating each individual 2nd–5th MCP joint and thewrist joint (Figure 1). The ROI placement was guided bycolour maps of enhancement parameters generated by thesoftware and by the corresponding STIR images from theparticipants (the latter only for patients). All DCE-MRimages were analysed once by one physician (MBA), and

2 MB Axelsen et al

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

RA patient baselineHealthy person RA patients 6 months

Persistent

Plateau

Washout

A

D E F

G H I

B C

5.0

4.0

3.0

2.0

1.0

0.10

0.08

0.06

0.04

0.02

Figure 1. Colour-coded maps of (A–C) enhancement, (D–F) maximum enhancement, and (G–I) initial rate of enhancement. Images from a healthyperson are shown in the first column (A, D, G), images from an early RA patient at baseline are shown in the second column (B, E, H), and images fromthe sameRA patient after 6 months of treatment are shown in the third column (C, F, I). The colour codes are explained in the vertical bars presented in theimages. In the first row the regions of interest (ROIs) for the metacarpophalangeal (MCP) joints and the wrist joint are shown (red lines).

MRI in RA patients and healthy persons 3

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

images from 10 healthy persons and from all patients atbaseline and after 6 months of treatment were analysed asecond time 6 weeks later by the same reader for thecalculation of intra-reader agreement, smallest detectabledifference (SDD), and smallest detectable change (SDC).For each ROI the following parameters were used for

further analysis: the number of enhancing voxels (Nvoxel),the average initial rate of enhancement (IRE) (%/s), andthe average maximum enhancement (ME), as describedpreviously (7, 14). A sum score of all five ROIs wascalculated for each parameter as follows:

Nvoxel ¼X5

J¼1NvoxelJð Þ

IRE ¼ 1Nvoxel

X5

J¼1IREJ �NvoxelJð Þ

ME ¼ 1Nvoxel

X5

J¼1MEJ �NvoxelJð Þ

where J denotes the number of joints assessed (1–5).IRE�Nvoxel and ME�Nvoxel were also calculated.Readers of conventional MRI and DCE-MRI were

blinded to clinical, biochemical, and other imaging data,and for patients also to the chronological order of the images.Clinical data regarding the C-reactive protein (CRP)

level and the Disease Activity Score using 28 joint counts(DAS28) were available for 12, 13, and 13 patients at 0, 6,and 12 months, respectively.All participants underwent the same MRI protocol.

DCE-MRI data were incomplete at five individual timepoints. Thus, DCE-MRI data were available for 13, 12,and 12 patients at 0, 6, and 12 months, respectively.Consequently, change scores included 11 patients.

Statistics

Non-parametric statistics were used. The median and rangeof values are presented. Healthy persons and patients werecompared using theMann–Whitney U-test. Data from visits0, 6, and 12 months were pooled as ‘patients’ and receiveroperating characteristic (ROC) curves were generated from‘patients’ vs. ‘healthy persons’, with a null hypothesis thatthe area under the curve (AUC) ¼ 0.5. AUC values > 0.5show that using the applied MRI method is better thanrandomly assigning the participant to a group (‘patients’or ‘healthy persons’). The best possible cut-off points forseparating the two groups were selected for each parameter.Changes between time points were tested using Wilcoxon’ssigned rank test. A p-value < 0.05 was considered statisti-cally significant.Intra-reader agreement was calculated as the intraclass

correlation coefficient (ICC, two-way mixed, absoluteagreement, single measure). Agreement was classifiedas follows: ICC � 0.8 (very good), 0.50 � ICC < 0.8(good), 0.20 � ICC < 0.5 (poor), and ICC < 0.2 (trivial)(7, 21). The variability is presented as 95% confidenceintervals (CIs) (22). Ta

ble1.

Values,cha

nges,and

thestan

dardize

drespon

semea

ns(SRM

s)form

agne

ticresona

nceimag

ing(M

RI)p

aram

eters.

Healthype

rson

s

Patientswith

rheu

matoidarthritis(RA) Ch

ange

fromba

selineto

6mon

ths

Baseline

6mon

ths

12mon

ths

Absolute

chan

ge%

chan

geSR

M

RAMRISsyno

vitis

0(0–3)[0/3]

13(3–18)***

5.5(3–19)***

5.0(3–21)***

�4.5(–14

to1)††

�45(–78

to33)

�1.00

Nvoxel

3(0–183)[0/172]

362(0–5526)***

20(0–989)

23(0–1251)*

�149

(–4748

to785)†

�86(–100to

325)

�0.52

ME(SI/b

aselineSI)

1.67

(0–2.16)[0/2.14]

1.91

(0–2.20)**

1.81

(0–2.13)

1.83

(0–2.16)

�0.05(�

1.83

to0.21)

�3(�

100to

11)

�0.54

IRE(%

/s)

0.73

(0–2.60)[0/2.20]

1.65

(0–2.82)***

1.05

(0–1.93)

1.10

(0–3.21)

�0.51(–1.80

to0.72)†

�32(�

100to

60)

�0.88

MExNvoxel(SI/b

aselineSI)

5(0–343)[0/314]

669(0–12047)***

38(0–2105)

42(0–2566)*

�273

(�10

696to

1710)

�89(�

100to

433)

�0.49

IRExNvoxel(%

/s)

3(0–267)[0/242]

607(0–15579)***

82(0–1904)

22(0–3849)*

�390

(�14

763to

1659)†

�93(�

100to

677)

�0.45

DAS28

–4.76

(2.76–6.85)

2.95

(1.77–6.65)

2.27

(1.67–6.65)

�1.05(�

3.68

to0.78)†

0(�

91to

276)

�0.95

RAMRIS,RA

MRIscoringsystem

;Nvoxel,n

umberofenhancing

voxels;M

E,maximum

enhancem

ent;SI,signalin

tensity;IRE

,initialrateofenhancem

ent;DA

S28,DiseaseAc

tivity

Scoreusing28

jointcounts.

Values

aregivenas

med

ian(ra

nge),and

forh

ealthype

rson

salso

[5th/95thpe

rcen

tiles].

Diffe

renc

esbe

twee

nhe

althype

rson

san

dRA

patients(M

ann–

Whitney):*p<0.05,**p<0.005,***p<0.0001.

Chan

gesfro

mba

selineto

after6

mon

thsof

treatmen

t(Wilcoxon

'ssign

edrank

test):†p<0.05,††p<0.005.

4 MB Axelsen et al

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

As measures of the variability between repeated mea-surements, the SDD (23) and the SDC (24, 25) werecalculated. The SDD%/(highest value measured) and theSDC%/(highest value measured) were calculated as

measures of the variation within the groups.Responsiveness was calculated as the standardizedresponse mean (SRM) (26). The SRM value was consid-ered small (< 0.5), medium (0.5–0.8), or large (> 0.8).

A25

20

15

10RA

MR

ISIR

E, %

/sM

E

ME

xNvo

xel

IRW

xNvo

xel,

%/s

Nvo

xel

5

0

4.00

3.00

2.00

1.00

2.50 12000

10000

8000

6000

4000

2000

0

Healthy RA 0 mo RA 6 mo RA 12 moHealthy RA 0 mo RA 6 mo RA 12 mo

Healthy RA 0 mo RA 6 mo RA 12 mo Healthy RA 0 mo RA 6 mo RA 12 mo

Healthy RA 0 mo RA 6 mo RA 12 moHealthy RA 0 mo RA 6 mo RA 12 mo

2.00

1.50

1.00

.50

.00

.00

20000

15000

10000

5000

0

6000

5000

4000

3000

2000

1000

0

C

E F

B

D

Figure 2. Box plots of (A) the RAMRIS synovitis score, (B) the number of enhancing voxels (Nvoxel), (C) initial rate of enhancement (IRE),(D) IRE�Nvoxel, (E) maximum enhancement (ME), and (F) ME�Nvoxel in healthy persons and RA patients. Boxes represent medians and 25th/75thpercentiles; bars represent outliers and open circles represent extreme values. The horizontal dotted reference lines represent the 95th percentiles forhealthy persons. Differences between healthy persons and RA patients at 0, 6, and 12 months (Mann–Whitney): * p< 0.05, ** p < 0.005, *** p < 0.001.

MRI in RA patients and healthy persons 5

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

SPSS version 19.0.0 (SPSS Inc, Chicago, IL, USA) wasused for data analysis.

Results

MRI parameters for healthy persons and RA patients arepresented in Table 1.

Differences between healthy persons and RA patients

The MRI parameters were statistically significantlylower for healthy persons than for RA patients at base-line (all p < 0.0001), except for ME (p ¼ 0.003). TheRAMRIS synovitis score was 0 for 18 of the healthypersons (69.2%), while none of the RA patients had aRAMRIS synovitis score of 0 at any time point. Thescore remained significantly higher for RA patientsafter 6 and 12 months of treatment compared to healthypersons (p < 0.0001 at both time points), despite amarked decrease from baseline. In 10 (38.5%) of the26 healthy persons there were no enhancing voxels inthe hand, while only one (8.3%) RA patient had noenhancing voxels at baseline. There were no significantdifferences in the dynamic parameters between healthypersons and patients treated for 6 months, but after12 months Nvoxel, ME�Nvoxel, and IRE�Nvoxel weresignificantly higher in patients (p ¼ 0.037, 0.037 and

0.047, respectively; Table 1). After 6 and 12 months oftreatment, four (30.8%) and two (15.4%) RA patients,respectively, were without any enhancing voxels. BMEwas not found in any of the patients at any time point, orin any of the healthy persons.

Differentiation between RA patients and healthy persons

Figure 2 shows box plots for all MRI parameters forhealthy persons and RA patients at baseline and after6 and 12 months of treatment. The 95th percentile forhealthy persons was calculated for all MRI parameters.Only one patient had a baseline RAMRIS synovitis scorebelow the 95th percentile of healthy persons. Threepatients had values below the 95th percentile for Nvoxel,ME�Nvoxel, and IRE�Nvoxel and 10 patients had valuesbelow the 95th percentile for both IRE and ME at base-line. ROC curves, AUCs, and selected cut-off points forall parameters are presented in Figure 3.

RAMRIS performed better than the other parameters,with Nvoxel being the best performing DCE-MRI para-meter. For the RAMRIS synovitis score, it was possible toselect cut-off values that gave both high sensitivity andspecificity. With a cut-off value of 2.5, the sensitivity andspecificity were 1.00 and 0.88, respectively, and with acut-off value of 3.5 the corresponding values were 0.90and 1.00. For the Nvoxel, a high specificity of 0.92 was

0.00.0

0.2

0.4

0.6

0.8

1.0

0.2 0.4 0.6

1 - Specificity

Sens

itivi

ty

0.8 1.0

RAMRISNvoxelIREIRExNvoxelMEMExNvoxel

Figure 3. Receiver operating characteristic(ROC) curves for differentiation between RApatients and healthy persons by values of theRAMRIS synovitis score, number of enhan-cing voxels (Nvoxel), initial rate of enhance-ment (IRE) (%/s), maximum enhancement(ME), IRE�Nvoxel (%/s), and ME�Nvoxel.A p-value < 0.05 indicates that the areaunder the ROC curve (AUC) is significantlyhigher than 0.5. The values for the cut-offpoints are chosen to show the best perfor-mances with regard to sensitivity and/or spe-cificity. Two different cut-off values arepresented, with the corresponding sensitivityand specificity for differentiating patientswith RA from healthy controls inparentheses.

6 MB Axelsen et al

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

achieved using a cut-off value of 125, with a correspond-ing moderate sensitivity of 0.53. For the remaining para-meters, high specificities were only achieved at the cost ofvery low sensitivities and vice versa (Figure 3).

Responsiveness to treatment

RAMRIS synovitis score, Nvoxel, IRE�Nvoxel and IRE,but not ME and ME�Nvoxel, decreased from baseline to6 months after start of treatment (p ¼ 0.003, 0.038, 0.038and p ¼ 0.028, 0.11, and 0.051, respectively). In Table 1values, changes, and the SRM are presented. The SRMwas large for DAS28, the RAMRIS synovitis score, andIRE, medium for Nvoxel and ME, and small forME�Nvoxel and IRE�Nvoxel. Table 2 provides SDD andSDC values for the different DCE-MRI parameters. Thedecrease in DCE-MRI parameters after 6 months of treat-ment was larger than the SDC for 8, 7, 6, 6 for IRE, Nvoxel,ME,ME�Nvoxel, and IRE�Nvoxel, respectively. One patienthad no enhancing voxels at any time point, and for onepatient all DCE-MRI parameters increased (ME increasedin two patients) during the follow-up.

Intra-reader agreement

The intra-reader agreement for patients and healthy per-sons is presented in Table 2.

Clinical data

At baseline, eight patients had a tender and/or swollen wristjoint and seven patients had a tender and/or swollen MCPjoint, while three patients had no tender or swollen joints ofthe hand. All patients had active disease, with a meanDAS28 of 4.7 (95% CI 3.9–5.4), and 92% had moderatedisease activity (DAS28 > 3.2) (27). After 6 months oftreatment the mean DAS28 was 2.9 (95% CI 2.1–3.8) andsix patients had reached clinical remission (DAS28 < 2.6).At the 12-months visit the mean DAS28 was 2.8 (95% CI1.8–3.8) and a futher three patients had reached clinical

remission. There were no statistically significant correla-tions between DAS28 and MRI parameters, neither forstatus nor change scores (Spearman’s correlation coeffi-cient), data not shown.

Discussion

We assessed the conventional RAMRIS synovitis scoreand semi-automated analysis of DCE-MRI of the MCPand wrist joints in healthy persons and in patients withRA. We found that the RAMRIS synovitis score differ-entiated healthy persons and early RA patients, regardlessof the patients’ clinical disease activity and treatmentstatus. For DCE-MRI, Nvoxel differentiated early RApatients with active disease from healthy persons. Thesensitivity to change during therapy was high for IRE,and particularly for the RAMRIS synovitis score.

Differentiation between healthy persons and patients

Studies investigating the presence of synovitis andBME in healthy persons have been ambiguous, withsynovitis-like changes in healthy persons, usuallymild, having been found in some (18, 28), but notall (29, 30), studies. Olech et al used a modifiedRAMRIS scoring system with no contrast andreported synovitis-like changes in 17 of 40 healthypersons and in 32 of 40 RA patients, and could notdifferentiate between the two groups (31). We usedcontrast in both patients and healthy persons, and theRAMRIS synovitis score differentiated well betweenthe two groups. However, a quicker and less user-dependent method would be useful.

Quantification of the synovial volume is another poten-tial method for differentiating between groups and hasbeen assessed for both conventional MRI and DCE-MRI.Assessing conventional MRI, Klarlund et al found largersynovial volumes in theMCP joints of RA patients than inthose of healthy persons, but the variation within thegroups was large and the synovial volume could not

Table 2. Intrareader agreement: smallest detectable difference (SDD), smallest detectable change (SDC), and intraclass correlationcoefficient (ICC) for DCE-MRI parameters.

SDD SDC ICC

Healthy persons

RA patients RA patients

Healthy persons

RA patients

Baseline 6 months Baseline to 6 months Baseline 6 months

Nvoxel 14 (34) 145 (3) 119 (13) 98 (2) 0.87** 1.00** 0.98**ME (SI/baseline SI) 1.69 (78) 0.09 (4) 0.04 (2) 0.022 (1) 0.64* 1.00** 0.98**IRE (%/s) 2.68 (91) 0.01 (2) 0.05 (3) 0.06 (4) 0.09 NS 0.95** 1.00**MExNvoxel (SI/baseline SI) 26 (35) 291 (2) 255 (13) 204 (2) 0.87** 1.00** 1.00**IRExNvoxel (%/s) 43 (61) 406 (3) 210 (12) 228 (2) 0.28 NS 0.99** 0.94**

DCE-MRI, Dynamic contrast-enhancedmagnetic resonance imaging; Nvoxel, number of enhancing voxels; ME, maximum enhancement;SI, signal intensity; IRE, initial rate of enhancement; NS, not statistically significant.Numbers in parentheses indicate SDD and SDC in %/maximum measured value.* p < 0.05, ** p < 0.001.

MRI in RA patients and healthy persons 7

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

differentiate between them (32). Using DCE-MRI, Tanet al found a sixfold difference in mean number of enhan-cing voxels in the MCP joints in 28 healthy persons and33 RA patients (33), but an overlap between groups wasfound. Our results indicate that the number of Nvoxel, is apromising tool for differentiating between RA patients andhealthy persons also on an individual level.With the imaging technique used, only one dynamic

slice with a slice thickness of 3 mm was obtained. Thisraises the risk of under/overestimating the amount ofsynovitis within the joint (33, 34), and may at least partlyexplain why the RAMRIS synovitis score (generatedfrom the evaluation of 15 conventional MR images,each 3 mm thick) was the best parameter for separatingpatients and healthy persons, and the most responsivemeasure. It has since become technically possible toobtain and analyse several dynamic slices of the jointsof interest, giving a better coverage of the synovialvolume, and whether this, as expected, will increase thesensitivity and specificity of the DCE-MRI techniqueneeds to be investigated.Analysis of parameters reflecting dynamic properties of

the synovium, such as the early enhancement rate and themaximum enhancement, is a third possible approach. Wefound that both IRE and ME differentiated healthy personsand patients with active disease, but not healthy personsfrom patients with low disease activity. In a previous studyexamining the MCP joints, the early enhancement ratedifferentiated between arthritis patients and healthy personsfor the 2nd–4thMCP joints, but not between patient groups(35). In another study, both the relative enhancement andthe enhancement rate differentiated a group of healthypersons and inactive RA patients from a group of RApatients with intermediate and active disease, but notpatientswith lowdisease activity fromhealthy persons (36).Overall, Nvoxel may be useful in differentiating

between RA patients and healthy persons, whereasdynamic parameters such as the IRE, ME, and similarmeasures have not been able to do this.The selected cut-off points presented in Figure 3 show

that the RAMRIS synovitis score and to a lesser extentNvoxel are sensitive and specific parameters for differen-tiation between healthy persons and patients.

Responsiveness to treatment

The RAMRIS synovitis score is a validated and responsiveoutcome measure (10, 37–39) and the results of the presentstudy are in accordance with this. Dynamika parametershave previously been found to be responsive to treatmentwith intra-articular glucocorticoid injections in the kneejoint (7, 40). Other studies evaluating DCE-MRI para-meters have also shown that measures of the early enhance-ment rate are responsive to treatment with DMARDs andbiological treatments (6, 9, 10).We investigated the respon-siveness of the Dynamika parameters to intensive treatmentwith DMARDs and found accordingly that the IREwas themost responsive DCE-MRI parameter. The IRE correlates

well with histological synovitis and with subsequent ero-sive progression (14, 41, 42) and is a potentially reliableand feasible measure of current disease activity in clinicaltrials and in clinical practice. When analysing small joints,bias may be introduced through enhancement in the manyand relatively large adjacent blood vessels. As blood ves-sels have relatively high IRE andME (while the number ofenhancing voxels can be small), blood vessels unintention-ally included in a ROI can be part of the explanation for thelarger SDD for IRE and ME found in healthy personscompared to RA patients. Analysing the shape of the signalintensity vs. time curves for the individual voxels, as hasbeen done using different software (43, 44), may increasethe accuracy of this method. Signal intensity vs. time curveanalysis using Dynamika software currently requires sig-nificant user interaction.

It was unexpected that none of the participants had anyBME, as BME is common in RA patients with activedisease and has been found in approximately a quarterof patients, while it is rare in healthy persons (18, 45, 46).

A limitation in this study is the relatively small samplesize. In addition, although all patients had a high diseaseactivity at baseline, as measured by the DAS28, threepatients had no clinical involvement of the hand thatwas examined by MRI.

In conclusion, the RAMRIS synovitis score of conven-tional MRI and, to a lesser extent, the one-slice DCE-MRIparameter of synovial volume, Nvoxel, differentiated earlyRA patients from healthy persons. The RAMRIS synovitisscore and IRE, and to a lesser extentNvoxel, decreased duringtreatment and showed sensitivity to change. Currently, theRAMRIS synovitis score performs better than DCE-MRI.However, with further technical development DCE-MRIcould be a useful method for assessment of RA diseaseactivity.

Acknowledgements

The data analysis was supported by an unrestricted grant from AbbVie Inc,Denmark (formerly Abbott).

References1. Brown AK, Conaghan PG, Karim Z, Quinn MA, Ikeda K, Peterfy

CG, et al. An explanation for the apparent dissociation betweenclinical remission and continued structural deterioration inrheumatoid arthritis. Arthritis Rheum 2008;58:2958–67.

2. Haavardsholm EA, Østergaard M, Ejbjerg BJ, Kvan NP, Uhlig TA,Lilleas FG, et al. Reliability and sensitivity to change of theOMERACT rheumatoid arthritis magnetic resonance imaging score ina multireader, longitudinal setting. Arthritis Rheum 2005;52:3860–7.

3. Reiser MF, Bongartz GP, Erlemann R, Schneider M, Pauly T, SittekH, et al. Gadolinium-DTPA in rheumatoid arthritis and relateddiseases: first results with dynamic magnetic resonance imaging.Skeletal Radiol 1989;18:591–7.

4. Tamai K, Yamato M, Yamaguchi T, Ohno W. Dynamic magneticresonance imaging for the evaluation of synovitis in patients withrheumatoid arthritis. Arthritis Rheum 1994;37:1151–7.

5. König H, Sieper J, Wolf KJ. Rheumatoid arthritis: evaluation ofhypervascular and fibrous pannus with dynamic MR imagingenhanced with Gd-DTPA. Radiology 1990;176:473–7.

8 MB Axelsen et al

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

6. ReeceRJ,KraanMC,RadjenovicA,VealeDJ,O’Connor PJ, RidgwayJP, et al. Comparative assessment of leflunomide and methotrexate forthe treatment of rheumatoid arthritis, by dynamic enhanced magneticresonance imaging. Arthritis Rheum 2002;46:366–72.

7. Axelsen MB, Poggenborg RP, Stoltenberg M, Kubassova O, BoesenM, Hørslev-Petersen K, et al. Reliability and responsiveness ofdynamic contrast-enhanced magnetic resonance imaging inrheumatoid arthritis. Scand J Rheumatol 2013;42:115–22.

8. Buch MH, Boyle DL, Rosengren S, Saleem B, Reece RJ, RhodesLA, et al. Mode of action of abatacept in rheumatoid arthritispatients having failed tumour necrosis factor blockade: ahistological, gene expression and dynamic magnetic resonanceimaging pilot study. Ann Rheum Dis 2009;68:1220–7.

9. Huang J, Stewart N, Crabbe J, Robinson E, McLean L, Yeoman S,et al. A 1-year follow-up study of dynamic magnetic resonanceimaging in early rheumatoid arthritis reveals synovitis to beincreased in shared epitope-positive patients and predictive oferosions at 1 year. Rheumatology (Oxford) 2000;39:407–16.

10. Tam LS, Griffith JF, Yu AB, Li TK, Li EK. Rapid improvement inrheumatoid arthritis patients on combination of methotrexate andinfliximab: clinical and magnetic resonance imaging evaluation.Clin Rheumatol 2007;26:941–6.

11. Palosaari K, Vuotila J, Takalo R, Jartti A, Niemela R, Haapea M, et al.Contrast-enhanced dynamic and staticMRI correlates with quantitative99Tcm-labelled nanocolloid scintigraphy. Study of early rheumatoidarthritis patients. Rheumatology (Oxford) 2004;43:1364–73.

12. McQueen FM, Crabbe J, Stewart N. Dynamic gadolinium-enhancedmagnetic resonance imaging of the wrist in patients with rheumatoidarthritis: comment on the article by Cimmino et al. Arthritis Rheum2004;50:674–5; author reply 5–6.

13. Kubassova O, Boesen M, Cimmino MA, Bliddal H. A computer-aided detection system for rheumatoid arthritis MRI datainterpretation and quantification of synovial activity. Eur J Radiol2010;74:e67–72.

14. Axelsen MB, Stoltenberg M, Poggenborg RP, Kubassova O,Boesen M, Bliddal H, et al. Dynamic gadolinium-enhancedmagnetic resonance imaging allows accurate assessment of thesynovial inflammatory activity in rheumatoid arthritis knee joints:a comparison with synovial histology. Scand J Rheumatol2012;41:89–94.

15. Boesen M, Kubassova O, Bouert R, Axelsen MB, Ostergaard M,Cimmino MA, et al. Correlation between computer-aided dynamicgadolinium-enhanced MRI assessment of inflammation and semi-quantitative synovitis and bone marrow oedema scores of the wristin patients with rheumatoid arthritis – a cohort study. Rheumatology2012;51:134–43.

16. Boesen M, Kubassova O, Parodi M, Bliddal H, Innocenti S,Garlaschi G, et al. Comparison of the manual and computer-aidedtechniques for evaluation of wrist synovitis using dynamic contrast-enhancedMRI on a dedicated scanner. Eur J Radiol 2011;77:202–6.

17. Hetland ML, Stengaard-Pedersen K, Junker P, Lottenburger T,Ellingsen T, Andersen LS, et al. Combination treatment withmethotrexate, cyclosporine, and intraarticular betamethasonecompared with methotrexate and intraarticular betamethasone inearly active rheumatoid arthritis: an investigator-initiated,multicenter, randomized, double-blind, parallel-group, placebo-controlled study. Arthritis Rheum 2006;54:1401–9.

18. Ejbjerg B, Narvestad E, Rostrup E, Szkudlarek M, Jacobsen S,Thomsen HS, et al. Magnetic resonance imaging of wrist andfinger joints in healthy subjects occasionally shows changesresembling erosions and synovitis as seen in rheumatoid arthritis.Arthritis Rheum 2004;50:1097–106.

19. Ejbjerg B, McQueen F, Lassere M, Haavardsholm E, Conaghan P,O’Connor P, et al. The EULAR-OMERACT rheumatoid arthritis MRIreference image atlas: the wrist joint. Ann RheumDis 2005;64:i23–47.

20. Conaghan P, Bird P, Ejbjerg B, O’Connor P, Peterfy C,McQueen F,et al. The EULAR-OMERACT rheumatoid arthritis MRI referenceimage atlas: the metacarpophalangeal joints. Ann Rheum Dis2005;64:i11–21.

21. Bøyesen P, McQueen FM, Gandjbakhch F, Lillegraven S, Coates L,Wiell C, et al. TheOMERACTPsoriaticArthritisMagnetic ResonanceImaging Score (PsAMRIS) is reliable and sensitive to change: resultsfrom an OMERACT workshop. J Rheumatol 2011;38:2034–8.

22. Altman D. Practical statistics for medical research, 1st edn.Chapmann and Hall/CRC: London, 1990.

23. Bland JM, Altman DG. Statistical methods for assessingagreement between two methods of clinical measurement.Lancet 1986;1:307–10.

24. Beckerman H, Roebroeck ME, Lankhorst GJ, Becher JG, BezemerPD, Verbeek AL. Smallest real difference, a link betweenreproducibility and responsiveness. Qual Life Res 2001;10:571–8.

25. Bruynesteyn K, Boers M, Kostense P, van der Linden S, van derHeijde D. Deciding on progression of joint damage in paired filmsof individual patients: smallest detectable difference or change. AnnRheum Dis 2005;64:179–82.

26. Marra CA, Rashidi AA, Guh D, Kopec JA, AbrahamowiczM, EsdaileJM, et al. Are indirect utility measures reliable and responsive inrheumatoid arthritis patients? Qual Life Res 2005;14:1333–44.

27. Wells G, Becker JC, Teng J, DougadosM, Schiff M, Smolen J, et al.Validation of the 28-joint Disease Activity Score (DAS28) andEuropean League Against Rheumatism response criteria based onC-reactive protein against disease progression in patients withrheumatoid arthritis, and comparison with the DAS28 based onerythrocyte sedimentation rate. Ann Rheum Dis 2009;68:954–60.

28. Partik B, R and T, Pretterklieber ML, Voracek M, Hoermann M,Helbich TH. Patterns of gadopentetate-enhanced MR imaging ofradiocarpal joints of healthy subjects. AJR Am J Roentgenol2002;179:193–7.

29. Tonolli-Serabian I, Poet JL, Dufour M, Carasset S, Mattei JP, RouxH. Magnetic resonance imaging of the wrist in rheumatoid arthritis:comparison with other inflammatory joint diseases and controlsubjects. Clin Rheumatol 1996;15:137–42.

30. Østergaard M, Gideon P, Sorensen K, Hansen M, Stoltenberg M,Henriksen O, et al. Scoring of synovial membrane hypertrophy andbone erosions by MR imaging in clinically active and inactiverheumatoid arthritis of the wrist. Scand J Rheumatol1995;24:212–18.

31. Olech E, Crues JV 3rd, YocumDE,Merrill JT. Bonemarrow edemais the most specific finding for rheumatoid arthritis (RA) onnoncontrast magnetic resonance imaging of the hands and wrists:a comparison of patients with RA and healthy controls. J Rheumatol2010;37:265–74.

32. Klarlund M, Ostergaard M, Lorenzen I. Finger joint synovitis inrheumatoid arthritis: quantitative assessment by magnetic resonanceimaging. Rheumatology (Oxford) 1999;38:66–72.

33. Tan AL, Tanner SF, Conaghan PG, Radjenovic A, O’Connor P,Brown AK, et al. Role of metacarpophalangeal joint anatomicfactors in the distribution of synovitis and bone erosion in earlyrheumatoid arthritis. Arthritis Rheum 2003;48:1214–22.

34. Rhodes LA, Tan AL, Tanner SF, Radjenovic A, Hensor EM, ReeceR, et al. Regional variation and differential response to therapy forknee synovitis adjacent to the cartilage-pannus junction andsuprapatellar pouch in inflammatory arthritis: implications forpathogenesis and treatment. Arthritis Rheum 2004;50:2428–32.

35. Klarlund M, Østergaard M, Rostrup E, Skjødt H, Lorenzen I.Dynamic magnetic resonance imaging of the metacarpophalangealjoints in rheumatoid arthritis, early unclassified polyarthritis, andhealthy controls. Scand J Rheumatol 2000;29:108–15.

36. CimminoMA, Innocenti S, Livrone F, Magnaguagno F, Silvestri E,Garlaschi G. Dynamic gadolinium-enhanced magnetic resonanceimaging of the wrist in patients with rheumatoid arthritis candiscriminate active from inactive disease. Arthritis Rheum2003;48:1207–13.

37. Conaghan PG, Emery P, Østergaard M, Keystone EC, GenoveseMC, Hsia EC, et al. Assessment by MRI of inflammation anddamage in rheumatoid arthritis patients with methotrexateinadequate response receiving golimumab: results of the GO-FORWARD trial. Ann Rheum Dis 2011;70:1968–74.

MRI in RA patients and healthy persons 9

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.

38. Quinn MA, Conaghan PG, O’Connor PJ, Karim Z, Greenstein A,Brown A, et al. Very early treatment with infliximab in addition tomethotrexate in early, poor-prognosis rheumatoid arthritis reducesmagnetic resonance imaging evidence of synovitis and damage,with sustained benefit after infliximab withdrawal: results from atwelve-month randomized, double-blind, placebo-controlled trial.Arthritis Rheum 2005;52:27–35.

39. Østergaard M, Emery P, Conaghan PG, Fleischmann R, Hsia EC,XuW, et al. Significant improvement in synovitis, osteitis, and boneerosion following golimumab and methotrexate combinationtherapy as compared with methotrexate alone: a magneticresonance imaging study of 318 methotrexate-naive rheumatoidarthritis patients. Arthritis Rheum 2011;63:3712–22.

40. Boesen M, Kubassova O, Cimmino MA, Ostergaard M, Taylor P,Danneskiold-Samsoe B, et al. Dynamic contrast enhanced MRI canmonitor the very early inflammatory treatment response upon intra-articular steroid injection in the knee joint: a case report with reviewof the literature. Arthritis 2011;2011:578252.

41. Palosaari K, Vuotila J, Takalo R, Jartti A, Niemela RK, KarjalainenA, et al. Bone oedema predicts erosive progression on wrist MRI inearly RA – a 2-yr observational MRI and NC scintigraphy study.Rheumatology (Oxford) 2006;45:1542–8.

42. Gaffney K, Cookson J, Blake D, Coumbe A, Blades S.Quantification of rheumatoid synovitis by magnetic resonanceimaging. Arthritis Rheum 1995;38:1610–17.

43. van der Leij C, van de Sande MG, Lavini C, Tak PP,Maas M. Rheumatoid synovial inflammation: pixel-by-pixeldynamic contrast-enhanced MR imaging time-intensity curveshape analysis – a feasibility study. Radiology 2009;253:234–40.

44. van de Sande MG, van der Leij C, Lavini C, Wijbrandts CA, MaasM, Tak PP. Characteristics of synovial inflammation in earlyarthritis analysed by pixel-by-pixel time-intensity curve shapeanalysis. Rheumatology (Oxford) 2012;51:1240–5.

45. Gandjbakhch F, Conaghan PG, Ejbjerg B, Haavardsholm EA, FoltzV, Brown AK, et al. Synovitis and osteitis are very frequent inrheumatoid arthritis clinical remission: results from an MRI studyof 294 patients in clinical remission or low disease activity state. JRheumatol 2011;38:2039–44.

46. McGonagle D, Conaghan PG, O’Connor P, Gibbon W, Green M,Wakefield R, et al. The relationship between synovitis and bonechanges in early untreated rheumatoid arthritis: a controlledmagnetic resonance imaging study. Arthritis Rheum1999;42:1706–11.

10 MB Axelsen et al

www.scandjrheumatol.dk

Scan

d J

Rhe

umat

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y R

igsh

ospi

tale

t on

11/0

6/13

For

pers

onal

use

onl

y.