Author contributions:Guarantor of integrity of entire study,V.M.G.M.; study concepts and design,V.M.G.M., J.S.B.; literature research,V.M.G.M., T.B.G.; clinical studies,V.M.G.M., T.B.G., M.H.P., B.J.D.; ex-perimental studies, V.M.G.M., B.J.D.;data acquisition, V.M.G.M., B.J.D.,J.S.B., T.L., A.E.O., N.D.J., M.H.P.; dataanalysis/interpretation, V.M.G.M., T.B.G.;statistical analysis, V.M.G.M., T.B.G.;manuscript preparation, definition of in-tellectual content, and editing, V.M.G.M.,T.B.G.; manuscript revision/review andfinal version approval, all authors.
Knee in Early JuvenileRheumatoid Arthritis: MRImaging Findings1
PURPOSE: To determine the magnetic resonance (MR) imaging findings in theknee in early juvenile rheumatoid arthritis.
MATERIALS AND METHODS: MR imaging (1.5 T) was performed in the moresymptomatic knee in 30 children with juvenile rheumatoid arthritis with a symptomduration 1 year or less. Conventional, fast spin-echo, three-dimensional gradient-echo, and gadolinium-enhanced T1-weighted images were assessed. Two radiolo-gists independently read the images, and a third resolved disagreements. Theseimages were compared with knee radiographs in 27 children.
RESULTS: Mean maximal synovial thickness was 4.8 mm 6 2.4 (SD). Mean synovialvolume was 15.4 mL 6 10.8. Suprapatellar joint effusions were seen in 26 (87%) of30 knees, meniscal hypoplasia in 11 (37%) of 30 knees, and abnormal epiphysealmarrow in eight (27%) of 30 knees. Three knees had articular cartilage contourirregularity, fissures, and/or thinning. One knee had a bone erosion. Knee radio-graphs showed suprapatellar fullness in 78% of the knees, joint space narrowing inone knee, and no bone abnormalities.
CONCLUSION: Synovial hypertrophy and joint effusions are the most frequent MRimaging findings of knees in early juvenile rheumatoid arthritis. Early in the disease,radiographically occult cartilage and bone erosions are uncommonly seen at MRimaging. The potential relationship of synovitis to cartilage abnormalities deservesfurther study.
Juvenile rheumatoid arthritis (JRA) is the most common rheumatic disorder of childhood,with 6 to 19.6 incident cases per 100,000 children yearly (1). The knee is the mostfrequently affected joint (2). With the increasing availability and use of disease-modifyingantirheumatic drugs, demonstration of early joint damage has become important (3–5).Physical examination may have low reliability in the assessment of disease activity becauseof subjectiveness among examiners (6,7). Conventional radiographs, the current standardof reference for initial radiologic evaluation and assessment of disease progression in JRA,are of limited use in early arthritis because they may not show an inflamed synovium,cartilage destruction, and early bone erosions, all of which are depicted with magneticresonance (MR) imaging (7–12). For this reason, MR imaging findings have been suggested(13–16) as an index of disease activity for arthritis.
To utilize MR imaging to monitor disease progression and response to therapy inchildren with JRA, imaging findings early in the course of disease must be known formeaningful interpretation of subsequent changes. Prior investigators (8,9,11) have de-scribed MR imaging findings in children with JRA of variable disease duration (meanduration, 4–5 years). The purpose of this study was to determine the MR imaging findingsin the knee in children with early JRA (symptom duration , 1 year) and clinical synovitis.
MATERIALS AND METHODS
Patients
Thirty children and adolescents (21 female, nine male; age range, 5–16 years; mean age,10.2 years) were enrolled in this prospective study during 31 months (June 20, 1996, toFebruary 16, 1999) after referral to the rheumatology clinic for evaluation of early JRA.
696
They met the American College of Rheu-matology criteria for JRA, which are thepresent standards for diagnosis. These stan-dards require the presence of objective ar-thritis (defined as either joint swelling orlimitation of motion associated with heat,tenderness, or pain) in at least one joint for6 consecutive weeks and the exclusion ofother diagnoses (17). Additional inclusioncriteria for our study were the following:clinically evident arthritis in at least oneknee (as determined by the attending pe-diatric rheumatologist [M.H.P.]), diseaseduration (defined as time since onset ofsymptoms) of less than 1 year, and no in-jection of intraarticular steroids into theaffected knee.
Patients were recruited as part of a pro-spective study of MR imaging of JRA inwhich patients are subsequently undergo-ing MR imaging and conventional radiog-raphy of the knee on a yearly basis. In this5-year prospective study, clinicians areblinded to MR imaging results, and radiol-ogists are blinded to details of the clinicaldata in individual patients until the end ofthe study. As a result, correlation of clinicaland imaging data cannot be accomplishedat this time. Disease duration of less than 1year was selected because, unlike the casewith rheumatoid arthritis, rapidly progres-sive erosions are uncommon in JRA withinthe 1st year of disease (18).
Institutional review board approvaland parental informed consent were ob-tained for MR imaging and radiography.Only children who could undergo imag-ing without sedation were enrolled. Stan-dard screening safety criteria for MR im-aging were observed.
Imaging
MR imaging was performed in themore symptomatic knee with a 1.5-Tmagnet (Signa; GE Medical Systems, Mil-waukee, Wis) by using a dedicated kneecoil and the imaging sequences detailedin Table 1. A fat-saturated three-dimen-sional T1-weighted gradient-echo tech-nique was used to evaluate the morphol-ogy of the articular cartilage (19). Fastspin-echo (SE) intermediate- and T2-weighted images were used to depictchanges in signal intensity within the ar-ticular cartilage (20). Immediately afterthe intravenous administration of con-trast material (gadopentetate dimeglu-mine [Magnevist]; Berlex Laboratories,Wayne, NJ; 0.1 mmol per kilogram ofbody weight), T1-weighted fat-saturated
images were obtained in the sagittal andcoronal planes. Sagittal image acquisi-tion was completed within 3–4 minutesof the administration of contrast mate-rial, which is well within the recom-mended 5 minutes allowed after contrastmaterial administration to capture peaksynovial enhancement and prevent vol-ume overestimation due to contrast ma-terial diffusion into the joint space (21).Total imaging time was 45–55 minutes.
Twenty-seven patients had frontal andlateral radiographs of the non–weight-bearing knee. These were obtained at thetime of MR imaging in all children exceptone, in whom they were obtained 3 weekslater. Three patients and/or parents de-clined radiography.
Image Analysis
MR images were interpreted by readerswho were blinded to the specific clinicalhistory, including the duration, extent,and severity of symptoms; findings atphysical examination; and medications.The primary readers (V.M.G.M., J.S.B.) re-viewed the MR images of the 30 enrolledpatients. In the event of disagreementbetween the primary readers, a third reader(T.L.) provided independent interpretationof subjective variables that served as thetiebreaker. Each case was originally as-sessed for approximately 40 subjectivevariables (only a portion of which are re-ported in this article) at the time of MRimage interpretation. A third reading wastypically required to assess three to six sub-jective variables per case. Continuous vari-ables were measured once per patient.
The synovium was assessed for signal in-tensity characteristics, distribution, andmaximal thickness. Maximal synoviumthickness was measured in the suprapatel-
*ETL 5 echo train length, FSE 5 fast SE, Gd 5 gadolinium enhancement, IW 5 intermediate weighted, SPGR 5 spoiled gradient echo, T1W 5 T1weighted, T2W 5 T2 weighted, 3D 5 three dimensional.
† TR/TE 5 repetition time (msec)/echo time (msec).
Figure 1. Sagittal gadolinium-enhanced T1-weighted fat-saturated MR image (400/16)demonstrates the boundary of pixels (arrow-heads) used for volume calculation in the en-hancing synovium. (Fig 3c is an identical im-age, with boundaries not shown.)
Volume 220 z Number 3 Early Juvenile Rheumatoid Arthritis: Knee MR Imaging z 697
lar bursa and intercondylar region on sag-ittal T1-weighted gadolinium-enhancedimages. It was measured on digital imagesmagnified by a factor of two and wasgraded as follows: 1, 0–2 mm; 2, 3–5 mm;3, 6–9 mm; and 4, 10–14 mm.
Synovial volumes were also calculatedon the sagittal T1-weighted fat-saturatedimages by using semiautomatic segmenta-tion based on gadolinium enhancementand a dedicated software program (INTERAC-TIVE DATA LANGUAGE; Research Systems, Boul-der, Colo). A radiologist (V.M.G.M.) drewregions of interest around the synovium toexclude areas of high signal intensity dueto adjacent blood vessels, soft tissues, andphyses from the volume calculation.Boundaries were closely drawn around thebright areas within the user-defined re-gions by using a k-means-clustering algo-rithm (Fig 1) (22). Volume was calculatedfrom the number of brightest pixels thatwere automatically segmented by multi-plying the in-plane pixel dimensions bythe section thickness. Volumes (range,2–25 mL) of three volume-segmentationphantoms were calculated with varyingsection thicknesses and different x and yresolutions and were found to be accurateto within 13% of the actual volumes (Dard-zinski BJ, unpublished data, 1999); this ac-curacy is acceptable for this method to beused as a relative measure of joint inflam-mation (23).
The usefulness of maximal synovialthickness versus synovial volume in deter-mining synovitis was assessed by compar-ing knee MR images in patients with JRAand active synovitis (n 5 30) with those ofcontrol subjects without clinical synovitis
(12 knees in 12 children who underwentMR imaging for other reasons). Active sy-novitis is defined as swelling or limitationof motion with heat, pain, or tendernessand was required in the patients with JRAfor entry into the study (17). Although theinvestigators did not examine the controlsubjects, clinical synovitis was thought tobe absent on the basis of the history givenwhen the study was requested. Sensitivityand specificity in the diagnosis of synovitiswith the use of maximal synovial thicknessversus synovial volume were compared byusing the clinical diagnosis of synovitis asthe standard because this method remainsthe way JRA is defined (17).
Joint effusion size was subjectivelygraded within the suprapatellar bursa, in-tercondylar region, and tibiofibular joint asfollows: 1, none or trace; 2, small; 3, mod-erate; and 4, large or marked. Menisci weregraded as normal, hypoplastic (with short-ening and/or loss of triangular configura-tion), torn, or atrophic and/or absent. Ar-ticular cartilage was assessed for contour(smooth vs irregular) and focal destruction(intact, superficial loss and/or thinning, ordeep erosions to subchondral bone). Jointcartilage thickness of the middle of the me-dial and lateral femoral condyles and tibialplateaus was measured on digitized three-dimensional fat-saturated spoiled gradient-echo images that were electronically mag-nified by a factor of five. Bone was assessedfor marrow signal intensity abnormalitiesand focal erosions. The knee joint was eval-uated for abnormalities of the infrapatellarfat pad, tenosynovitis, synovial cysts, re-gional lymphadenopathy, and internal de-rangements.
Conventional knee radiographs ob-tained in 27 patients were randomly mixedwith 57 age- and sex-matched normal kneeradiographs (selected by V.M.G.M.). Radio-graphs were masked and reviewed bytwo pediatric musculoskeletal radiolo-gists (N.D.J., A.E.O.) who were blinded topatient history and MR imaging findings.In the event of disagreement, a third reader(T.L.), who served as tiebreaker, read themasked patient and normal radiographs asa batch 6 months after the reading of theMR images. Radiographs were scored byusing the Pettersson classification system(24). Independent of this scoring system,radiographs were also assessed for suprapa-tellar bursa fullness due to effusion and/orpannus (none, small, moderate, or large)and joint space narrowing.
Statistical Analysis
Statistical analysis was completed withSPSS (version 8.0; SPSS, Chicago, Ill) forWindows (Microsoft; Redmond, Wash).A Student t test was used to compare themeans of continuous variables. Nonpara-metric tests were used to compare ordinalvariables. When more than one controlwas used per subject, the mean value wasused for comparison purposes and wasrounded to the greater number. A Spear-man rank correlation was used to assessthe correlation between imaging vari-ables when at least one of the variableswas ordinal. A Pearson correlation coeffi-cient was calculated when two continu-ous variables were correlated. Wheneverpossible, the k statistic was calculated toassess interobserver reliability (25). A k
Figure 2. Summary of MR imaging findings in the knees of children with early (newly diagnosed) JRA.
698 z Radiology z September 2001 Gylys-Morin et al
value of more than 0.75 denotes excel-lent reliability; 0.75–0.40, good reliabil-ity; and less than 0.40, marginal reliabil-ity (26). Because some cells were zero forsome variables, k could not be calculated.In these instances, the percentage ofagreement was calculated.
Receiver operating characteristic curveswere used to evaluate the ability to dis-tinguish patients from control subjectswith both conventional radiographs andMR images. These curves were constructedby using nonparametric methods withsoftware (Analyse-it Software, Leeds, En-gland) for EXCEL (Microsoft; Redmond,Wash). In each instance, analysis of thedata clearly revealed points that maxi-mized specificity while not substantially
compromising sensitivity. The Hanley andMcNeil method was used to compare re-ceiver operating characteristic curves.
RESULTS
Patients
All 30 patients presented with jointswelling, with or without pain and limi-
tation of motion. All had received non-steroidal antiinflammatory drugs (meanduration of therapy, 2.7 months) prior toMR imaging. All patients were thought tohave active knee synovitis when theywere examined on the day of MR imag-ing, but some patients had an improve-ment in objective arthritis after begin-ning therapy. Mean symptom durationwas 5.1 months (range, 1–12 months); 16(53%) of 30 patients had pauciarticular,10 (33%) had polyarticular, and four(13%) had systemic onsets of disease.
MR Imaging Findings
The two most frequent MR imagingfindings in our study were synovial pro-liferation and joint effusions (Fig 2).
Synovium.—The synovium was seen inthe suprapatellar bursa, lining the in-frapatellar fat pad, and adjacent to theposterior femoral condyles in all knees(Figs 3–5). The synovium was also adja-cent to or surrounding menisci in 29(97%) of 30 knees, in the intercondylarregion and posterior femoral recess in 28(93%) knees, in the posterior tibial orpopliteus recess in 20 (67%) knees, and inthe tibiofibular joint in 18 (60%) knees.Interobserver agreement was 78%–99%in the assessment of synovial distribu-tion.
Synovial signal intensity characteris-tics varied with the pulse sequences (Figs3–5), with 80%–85% interobserver agree-ment. On nonenhanced T1-weighted im-ages, the synovial signal intensity was in-termediate (similar to that of muscle) in23 (77%) of 30 knees, low to intermediatein two (7%), and not visible in five (17%).Compared with that of muscle, the signalintensity on fast SE intermediate-weightedimages was low to intermediate in 18(60%) knees, intermediate or isointense ineight (27%), and not distinguishable infour (13%). Compared with the high signalintensity of joint effusions on fast SE T2-weighted images, the signal intensity wasmixed low to intermediate in 13 (43%)knees, intermediate in eight (27%), mixedintermediate to high in five (17%), and notdistinguishable in four (13%).
Following the intravenous administra-tion of gadolinium-based contrast mate-rial, enhancing synovium was seen in allknees on T1-weighted fat-saturated im-ages in one of two enhancement pat-terns: (a) a hypervascular pattern (Fig 4)of homogeneously high signal intensityin 16 (53%) of 30 knees or (b) a mixedpattern (Figs 3, 5) of heterogeneous sig-nal intensity in 14 (47%) knees. The Co-hen k value for interobserver reliability
Figure 3. (a) Sagittal fast SE T2-weightedMR image (3,000/112) obtained in a 5-year-old girl shows bulky low-signal-intensitypannus (p) in the suprapatellar bursa, whichis outlined by high-signal-intensity joint ef-fusion (open arrow). Synovium lining thebursa (arrowheads) is of intermediate to highsignal intensity and is difficult to distinguishfrom adjacent high-signal-intensity effusion.The infrapatellar fat pad (curved arrow) hascontour irregularity and signal inhomogene-ity. Intermediate to low signal intensitypannus (solid straight arrows) is in the inter-condylar region. (b) Pannus has mixed en-hancement on this sagittal gadolinium-en-hanced T1-weighted fat-saturated MR image(400/16). The peripheral synovium has highsignal intensity (arrowheads), while the cen-tral bulky pannus (p) does not enhance and isindistinguishable from the surrounding low-signal-intensity joint fluid. (c) More medi-ally, hypervascular synovium is in the poste-rior femoral recess (white straight solid arrow)and encroaches onto the meniscal surfaces(black arrows) on this sagittal T1-weighted gad-olinium-enhanced MR image (400/16). Loss ofthe triangular configuration in the medial me-niscus (black arrows) suggests early meniscalhypoplasia. Prominent radially oriented vessels(open arrows) course through the femoral con-dyle epiphyseal cartilage. The distal femoralepiphysis has abnormally increased marrowsignal intensity (curved arrow).
Volume 220 z Number 3 Early Juvenile Rheumatoid Arthritis: Knee MR Imaging z 699
for the enhancement pattern was 0.543,which indicated good agreement. Kneeswith the mixed enhancement patternhad a greater mean maximal synovialthickness (6.3 mm 6 1.8 [SD] vs 3.4mm 6 2.2; P , .001) and had a trendtoward higher mean synovial volumes(20.9 mL 6 6.7 vs 7.4 mL 6 6.1; P 5 .08)compared with those with the hypervas-cular pattern.
The mean maximal thickness of thesynovium was 4.8 mm 6 2.4 (range, 1–10mm). In the suprapatellar bursa, maximalsynovium thickness was 1–2 mm inseven (23%) of 30 knees, 3–5 mm in 15(50%), 6–9 mm in seven (23%), and10–14 mm in one (3%). In the intercon-dylar region, maximal synovial thicknesswas 1–2 mm in nine (30%) knees, 3–5mm in 12 (40%), and 6–9 mm in nine(30%). The grade of maximal synovialthickness within the suprapatellar bursawas well correlated with that of the inter-condylar region (Spearman r 5 0.73; P ,.01). The mean maximal synovial thick-ness in knees with cartilage or bone ero-sions was greater than that of knees with-out destructive changes (7.3 mm 6 0.6 vs4.5 mm 6 2.4; P , .05).
The mean synovial volume was 15.4mL 6 10.8 (range, 2–44 mL). Synovialvolume was well correlated with synovialthickness grade in the suprapatellar bursaand intercondylar region (Spearman r 50.70 and 0.74, respectively; P , .01).Good correlation between synovial vol-ume and maximal synovial thickness inthe suprapatellar bursa was seen (Pearsoncorrelation coefficient, 0.73; P , .01).Trend analysis was performed (Fig 6); acubic regression line fit the data well(R2 5 0.588).
To determine the usefulness of maxi-mal synovial thickness versus synovialvolume in assessing synovitis at MR im-aging, receiver operating characteristiccurves were generated; these were used tofind the optimum values for maximal sy-novial thickness and synovial volume foruse in comparing patients with and thosewithout the clinical diagnosis of synovi-tis. A maximal synovial thickness of 3mm or more yielded 100% specificity and77% sensitivity for the diagnosis of syno-vitis (Table 2) (area under the curve 50.90; 95% CI: 0.81, 0.99). With a synovialvolume of 3 mL or more as the MR im-aging definition of synovitis, specificitywas 100% and sensitivity was 97% (Table2) (area under the curve 5 1.00; 95% CI:0.98, 1.00). Use of volume was superiorto use of maximal synovial thickness indistinguishing clinical synovitis (P ,.05).
Joint effusions.—Joint effusions hadhigh signal intensity on T2- or interme-diate-weighted images and low signal in-tensity on gadolinium-enhanced T1-weighted images (Figs 3–5, 7). They weremost common in the suprapatellar bursa,and moderate or large effusions wereseen in 17 (57%) of 30 knees (Table 3). kvalues for interobserver reliability in thedetection of moderate or large effusionswere 0.86 for the suprapatellar bursa,0.75 for the intercondylar region, and0.65 for the tibiofibular joint. Joint effu-sion size in the suprapatellar bursa wascorrelated with that in the intercondylarregion (Spearman r 5 0.63; P , .01) butnot with that in the tibiofibular joint(Spearman r 5 20.24). Size of the su-prapatellar joint effusion was correlatedwith maximal synovial thickness (Spear-man r 5 0.55; P , .01) and synovialvolume (Spearman r 5 0.66; P , .01).
Menisci.—Medial meniscal hypoplasiawas seen in 11 (37%) knees (Figs 3, 5).Three (10%) of all imaged knees also hadlateral meniscal hypoplasia. Interob-server agreement was good (Cohen k val-ues were 0.68 for medial meniscal hypo-plasia and 0.78 for lateral meniscal hy-poplasia). No meniscal tears or completeatrophy was seen. Meniscal hypoplasiawas positively correlated with larger su-prapatellar joint effusions, maximal sy-
novial thickness, synovial volume, andthe mixed synovial enhancement pattern(Spearman r 5 0.54, 0.56, 0.59, 0.59, re-spectively; P , .01). Although none ofthe knees with meniscal hypoplasia hadfocal cartilage or bone erosions, one hadabnormal subchondral enhancementand diffuse cartilage thinning.
Cartilage.—Total articular and epiphy-seal (growth) cartilage thickness de-creased with age in our series of childrenwith JRA. Mean cartilage thickness inskeletally immature knees (with openphyses) was significantly greater thancartilage thickness in mature knees (withclosed physes) (Table 4).
Prominent radial enhancement ofepiphyseal growth cartilage was seen inthree (10%) of 30 knees, with a meanpatient age of 6.3 years (Fig 3c). Cartilagedestruction, seen best on fast SE interme-diate-weighted and three-dimensionalspoiled gradient-echo fat-saturated im-ages, was identified in three (10%) of 30knees (93% interobserver agreement), allof which had fused or nearly fused phy-ses. One exhibited contour irregularityand signal intensity heterogeneity in thepatellar cartilage (Fig 7a). Another hadtwo articular cartilage fissures extendingto the subchondral bone of the patella(Fig 7b). The third knee had femoral andtibial cartilage thinning (Fig 5a, 5b),
Figure 4. (a) Sagittal fast SE T2-weighted fat-saturated MR image (4,000/126) obtained in a7-year-old boy shows a large knee joint effusion (p) that distends the suprapatellar bursa, posteriorfemoral recess, and posterior tibial recess. The effusion is lined by nearly imperceptible synovium(arrowheads). (b) The synovium (curved arrows) is readily seen on this gadolinium-enhancedsagittal T1-weighted fat-saturated MR image (400/17). Curvilinear high signal intensity at thecartilage-epiphysis junction (straight solid arrows) and enhancing vessels (open arrows) in thegrowth cartilage are normal findings.
700 z Radiology z September 2001 Gylys-Morin et al
which was verified at arthroscopy. Thesethree knees also exhibited abnormal sub-chondral linear enhancement on T1-weighted fat-saturated gadolinium-en-hanced images (Fig 5a, 5b). Additionalabnormalities included moderate or largejoint effusions, grade III synovial hyper-trophy, 19–34-mL synovial volumes, and
a radiographically occult bone erosion(Fig 8).
Bone changes.—Focal bone erosion wasseen in one knee (Fig 8), with 100% in-terobserver agreement. This knee alsohad patellar cartilage irregularity (Fig 7a),abnormal subchondral enhancement, 8-mm maximal synovial thickness, 22-mL
synovial volume, and an abnormal in-frapatellar fat pad.
Marrow signal intensity was abnor-mally increased in eight (27%) of 30knees on T2-weighted and gadolinium-enhanced T1-weighted fat-saturated im-ages. Marrow signal intensity was focallyincreased in one knee and diffusely in-creased in the femoral, tibial, and/or pa-tellar epiphyses of seven knees (Fig 9).Marrow signal hyperintensity was notcorrelated with cartilage or bone erosionsor the size of joint effusions. It was seenboth in knees with abundant synovialproliferation (Fig 3) and in knees withminimal or no synovial proliferation (Fig9).
No avascular necrosis, medullary in-farcts, marginal osseous defects and/or ero-sions, intraarticular fragments, or epiphy-seal overgrowths were seen.
Soft-tissue changes.—The infratentorialfat pad showed thickening, contour irreg-ularity, and/or signal intensity heteroge-neity on T2-weighted images in 21 (70%)of 30 knees (k 5 0.71). These knees had agreater maximal synovial thickness (5.7mm 6 2.1 vs 2.6 mm 6 1.6; P , .01) andsynovial volume (19.0 mL 6 10.6 vs 7.2mL 6 5.6; P , .01), compared with kneeswith a normal infrapatellar fat pad.
Popliteal synovial cysts were seen in six(20%) of 30 knees and measured 1.0–4.2cm3. One patient had two intact synovialcysts and a ruptured synovial cyst thatextended inferiorly into the calf. Thepresence of popliteal cysts was not corre-lated with the size of joint effusion(Spearman r 5 0.19), synovial prolifera-tion, or number and size of lymph nodes.
Popliteal lymph nodes were seen in 28(93%) of 30 knees. They numbered oneor two in five (18%) of 28 knees, three tofive in 21 (75%) knees, and more than sixin two (7%) knees. The mean maximallymph node length was 10.6 mm 6 4.2(range, 6–20 mm). There was no correla-tion between the number of lymph nodesand maximal lymph node size, joint effu-sion, and synovial volume (Spearman r 50.29, 0.1, and 0.32, respectively).
Conventional Radiographs
The mean Pettersson score was 0.63 61.18 (range, 0–4; median, 0; mode, 0),with 63% agreement. This score was sig-nificantly different from the mean Pet-tersson score of 0.05 6 0.29 (median, 0;mode, 0) for normal knee radiographs(P , .05, signed rank test). In patientswith JRA, Pettersson scores were 0 in 20(74%) of 27 knees, 1 in one (4%) knee, 2in three (11%) knees, 3 in two (7%)
Figure 5. (a) Coronal T1-weighted gadolinium-enhanced MR image (450/16) obtained in a15-year-old female adolescent shows abnormal subchondral enhancement (arrowheads). (b) Sag-ittal T1-weighted gadolinium-enhanced fat-saturated MR image (400/16) shows the abnormalsubchondral enhancement of the tibial plateau (arrowheads) and posterior femoral condyle(white arrows). At arthroscopy, diffuse articular cartilage thinning, pitting, and fissures of thetibial plateau were seen. Enhancing synovium lifts the inferior aspect of the medial meniscus(black arrow), with loss of the normal meniscal triangular configuration. At arthroscopy, themeniscus was compressed by the bulky pannus but was otherwise normal. (c) Correspondingsagittal fast SE T2-weighted MR image (3,500/120) shows normal subchondral signal intensity(arrowhead) and intermediate-signal-intensity synovium surrounding the meniscal tips (whitearrows). Large joint effusion in the suprapatellar bursa contains material (black arrows) withintermediate to low signal intensity that does not enhance with gadolinium-based contrastmaterial; these findings are compatible with the mixed enhancement pattern. At arthroscopy,this material was free-floating curdlike clumps of fibrin.
Volume 220 z Number 3 Early Juvenile Rheumatoid Arthritis: Knee MR Imaging z 701
knees, and 4 in one (4%) knee. The abil-ity to distinguish patients with JRA fromcontrol subjects by using abnormal (.0)Pettersson scores was not significant(area under the receiver operating char-acteristic curve, 0.60; 95% CI: 0.47, 0.73;P 5 .07). Although bone erosions wereinterpreted in four knee radiographs inpatients with JRA, corresponding MR im-ages showed none. The one focal boneerosion seen on MR images was not iden-tified on radiographs; the Petterssonscore was 0.
In patients with JRA, suprapatellar jointfullness was mild on 10 (37%) of 27 kneeradiographs, moderate on seven (26%),and large on four (15%), with 52% overallagreement. In these patients, moderate orlarge suprapatellar fullness on radiographswas highly correlated with moderate orlarge suprapatellar effusions in MR images(Spearman r 5 0.80; P , .01).
Joint space narrowing was seen on one(4%) of 27 JRA knee radiographs, withcomplete agreement. Corresponding kneeMR images showed meniscal hypoplasiaand diffuse cartilage thinning.
DISCUSSION
In early JRA, these MR imaging findingsreflect the underlying pathologic processin which synovial hypertrophy results inthe increased secretion of joint fluid.
At MR imaging, synovial abnormalityis best seen by using both T2-weightedand gadolinium-enhanced T1-weightedfat-saturated images (Figs 3–5). On non-enhanced images alone, the synoviummay be difficult to distinguish from jointfluid or adjacent soft tissues due to insuf-ficient signal intensity contrast (Fig 4).On the other hand, the pannus (seen onT2-weighted images as low-signal-inten-sity material outlined by high-signal-in-tensity joint effusion), which is mini-mally enhancing or nonenhancing, isnot visible on gadolinium-enhanced T1-weighted images (Figs 3, 5). In our series,this type of pannus had mixed signal in-tensity characteristics, which were simi-lar to those described in patients withadult rheumatoid arthritis (27,28). Thissynovium with mixed enhancement wasbulkier (greater maximal thickness andsynovial volume) than synovium withthe hypervascular enhancement pattern.Mixed synovial enhancement might in-dicate more long-standing or more severeinflammation wherein high-signal-inten-sity vascular proliferation and villous hy-pertrophy combine with the lower-signal-intensity fibrin and hemosiderin (29).
To assess disease activity with MR im-aging, investigators use intravenously ad-ministered gadolinium-based contrastmaterial to measure synovial inflamma-tion with time-activity curves of en-hancement (27,30,31), maximal synovialthickness (23), or synovial volumes (15,16,32,33). We found a maximal synovialthickness of 3 mm or more and synovialvolume of 3 mL or more to be usefulmeasures of synovitis. Although synovialvolume had higher sensitivity than max-imal synovial thickness in the diagnosisof synovitis, synovial thickness may bemore practical to use because it is easilymeasured, requires no special postpro-cessing, and is well correlated with syno-vial volume. Care should be taken toavoid overestimating synovitis with theuse of maximal synovial thickness whenthe synovium is irregular or nodular.Conversely, synovitis might be underes-timated by measuring thickness onlywhen the synovium is thin in the pres-
ence of a very large joint effusion. Syno-vial volume measurements are practicalto use in the clinical setting, requiringapproximately 5 minutes of postprocess-ing per case. Regions of interest weredrawn by a radiologist (V.M.G.M.) whowas cognizant of joint anatomy; then,measurements were completed by an-other author (B.J.D.), with the two work-ing together at the workstation.
Destructive changes in cartilage andbone were seen only in skeletally matureknees in our series. This finding, com-bined with the observation that joint car-tilage thickness decreases with age, sup-ports the speculation that thicker cartilage,along with intact growth cartilage bloodsupply and better repair processes in skele-tally immature children, may account forthe fewer erosive changes seen in JRA com-pared with those seen in adult rheumatoidarthritis (18,34). As children mature intoadults, their joints might also be at greater
Figure 6. Graph shows the relationship between maximal synovial thickness and synovialvolume; the cubic regression line indicates a strong positive correlation.
702 z Radiology z September 2001 Gylys-Morin et al
risk for joint destruction if synovitis per-sists.
The infrequent occurrence of destruc-tive changes in cartilage and bone inearly JRA is not entirely surprising. Dataobtained in the 1980s (18) demonstratedthat even in polyarticular and systemicdisease, the median time to the develop-
ment of destructive changes that are de-tectable at conventional radiography waslonger than 2 years. More frequent andaggressive use of methotrexate in the en-suing years is expected to decrease therate of development of erosions even fur-ther (3). Nevertheless, MR imaging ismore sensitive in the detection of carti-
laginous destructive changes in JRA (8–10,14). The prevalence of such changesin cross-sectional studies (8–10,14) in-volving patients with JRA has been esti-mated to be 50%–88%. The data pre-sented here are new because sensitive MRimaging techniques were applied to theassessment of early disease. Data aboutthe development of erosions will con-tinue to evolve as imaging techniquesand pharmacologic treatments improve.
Cartilage destruction in rheumatoidarthritis occurs either on the joint surfaceby degradative enzymes released by thesynovium or from subchondral resorp-tion resulting in diffuse cartilage thin-ning (35,36). The abnormal subchondrallinear enhancement pattern seen on ga-dolinium-enhanced T1-weighted imagesin three skeletally mature knees in ourseries may represent the increased vascu-larization of the basilar layers that leadsto cartilage resorption. This appearance isdistinct from the normal hyperintenserim seen around immature epiphyseal os-sification centers, which represents met-abolic activity at the edge of growingepiphyses (37). A patient with this abnor-mal subchondral linear enhancement didindeed have tibial articular cartilage thin-ning (verified at arthroscopy), which wasseemingly due to a combination of sur-face scalloping by the pannus and sub-chondral resorption. We postulate thatthis abnormal enhancement pattern is anearly MR imaging indicator of subse-quent cartilage loss. It is not seen inasymptomatic knees at skeletal maturity(Gylys-Morin VM, unpublished data,1999) or in adults (38).
Accentuated spoke-wheel enhance-ment of growth cartilage on gadolinium-en-hanced images may be a secondary MR im-aging sign of inflammation in JRA. A similarfinding in two knees in patients with juve-nile chronic arthritis was shown to be vesselsat Doppler ultrasonography (13). This prom-inent enhancement is distinct from the fineradial pattern of enhancement seen in nor-mal developing cartilaginous epiphyses,which represents vascular canals that supplynutrients to cartilage and induce ossification(37). With active inflammation, these vesselsmay enlarge and eventually contribute togrowth disturbances, such as epiphyseal en-largement, increased maturation, and pre-mature physeal closure.
The most common bone abnormalityseen at MR imaging in the knees of chil-dren with early JRA was marrow signal in-tensity abnormality. This finding may rep-resent bone marrow edema or hyperemiathat eventually contributes to epiphysealovergrowth.
Figure 7. (a) Cartilage destruction in knees of patients with earlyJRA. Adjacent transverse fast SE intermediate-weighted fat-saturatedMR images (4,000/30) show areas with heterogeneous high signalintensity (white arrows) and contour irregularity (black arrow) in thepatellar articular cartilage in a 15-year-old female adolescent. High-signal-intensity effusion (p) is seen in the adjacent joint space.(b) Transverse intermediate-weighted fat-saturated fast SE MR image(4,000/28) obtained in a 14-year-old female adolescent demonstratestwo fissures (arrows) that extend to the subchondral bone. These areaccentuated by high-signal-intensity joint effusion (p).
Volume 220 z Number 3 Early Juvenile Rheumatoid Arthritis: Knee MR Imaging z 703
Although bone erosions are infrequentin early JRA, MR imaging is inherently wellsuited to depict them due to its high signalintensity contrast and multiplanar tomo-graphic nature. The finding of one radio-logically occult bone erosion in our seriessuggests that erosions may occur earlierthan previously described (8–10,14). Con-ventional radiographs do not depict boneerosions or subchondral cysts smaller than8 mm in diameter due to bone densitysummation; centrally located cysts areeven more difficult to identify (39). Thus,MR imaging might be especially useful inthe detection of small bone lesions in ther-apeutic trials involving large joints.
We found meniscal hypoplasia early inJRA. Previous investigations (8,9) involvingpatients with JRA with a mean disease du-ration of 4–5 years reported a prevalence ofmeniscal hypoplasia of 65%–96%. Theearly occurrence of these changes was notappreciated, given the longer disease dura-tion. Since meniscal hypoplasia in our se-ries was well correlated with increased sy-novial volume, some meniscal changes inour study may have been caused by simplemechanical compression due to bulky pan-nus. Other described (35,40) mechanismsfor meniscal hypoplasia may play a moreimportant role in continued inflamma-tion.
Infrapatellar fat pad abnormalitieswere the only MR imaging soft-tissuefindings that we found to be useful as asecondary sign of inflammation in earlyJRA. This finding was well correlated withincreased synovial proliferation and jointeffusions. Popliteal cysts and lymphade-nopathy were not correlated with syno-vial proliferation or joint effusions wellenough to serve as useful secondary signsof synovitis.
In our series, conventional radiogra-phy of the knee was insensitive and non-specific to early pathologic changes inJRA, as seen at MR imaging. To ourknowledge, the only radiographic scor-ing system currently available to gradedisease severity of the knee in JRA is thePettersson classification system (24). Thissystem, however, allows assessment ofonly bone abnormalities, which are lateand irreversible manifestations of the dis-ease process. Thus, it is impractical forthe grading of early disease.
Our study has several limitations. Thecurrent results are limited in clinical ap-plicability by the lack of correlation withclinical and follow-up imaging data.These issues are being addressed in anongoing trial (therapeutic control groupsare not being used). Measurement of sy-novial volume at MR imaging is relevant
to the outcome in adult rheumatoid ar-thritis and is expected to be similarly use-ful in JRA (16,30,33). Another limitationis the lack of a surgical or pathologicstandard for the verification of MR imag-ing findings in all patients except one,who underwent arthroscopy and syno-vectomy. Although biopsy proof of syno-vitis is desirable, it is not practical in theclinical setting with children. The clini-cal definition of synovitis, despite its lim-itations, remains the standard because itis the way in which JRA is defined (17).Especially in small joints, MR imagingmay be more sensitive than clinical ex-amination (41). Nevertheless, to date,MR imaging has not been accepted as adiagnostic criterion standard. The accu-racy in calculating phantom volumeswas acceptable (23). Even in patients whoundergo synovectomy, synovial volumecannot be accurately measured in an ex-cised specimen. The high sensitivity andspecificity of synovial volumes in the de-tection of clinically evident synovitis inthe knee joint validate both clinical as-sessment and MR imaging for the kneejoint.
A further limitation is that, with theexception of synovitis, imaging findingsin JRA were not compared with findingsin other diseases, because this compari-son was beyond the scope of this project.Furthermore, images in one joint may be
misleading representations of polyarticu-lar disease. Finally, it is difficult to drawmeaningful relationships between carti-lage destruction, the presence of osseouslesions, and severity of synovial inflam-mation because few patients exhibitedcartilage or bone destruction. Sample sizeis a limitation. The primary area in whichlack of power is a concern is in the factorsassociated with cartilage abnormalities,given the low number of such abnormali-ties. We expect that, with continuedenrollment and serial follow-up, thenecessary power to make inferences inthis regard can be achieved. These lim-itations, however, do not negate theMR imaging findings presented becauseeach knee can provide its own baselinefindings for future comparison in theassessment of disease progression or re-sponse to therapy.
In summary, MR imaging of the kneedepicts the synovium and its effects onjoint structures in children with early JRA.A maximal synovial thickness of 3 mmmore or a synovial volume of 3 mL or moreis a sensitive and specific MR imaging cri-terion for active synovitis in the knee. Al-though synovial hypertrophy and joint ef-fusions are the most frequent MR imagingfindings in knees in early JRA, radiograph-ically occult cartilage and bone erosionsmay also be identified. Thus, MR imagingmay potentially aid therapeutic decisions,
TABLE 3Frequency and Size of Joint Effusions at MR Imaging in Patientswith Clinical Synovitis
* Location of the middle or weight-bearing portion.† Data are the mean 6 SD.
704 z Radiology z September 2001 Gylys-Morin et al
particularly early in the disease process, byhelping in the quantification of synovitisand by depicting cartilage and bone de-struction not evident on conventional ra-diographs.
Acknowledgments: We thank Scott K. Hol-land, PhD, for assistance with volume segmen-
tation; Vincent J. Schmithorst, PhD, for writ-ing the dedicated INTERACTIVE DATA LANGUAGE
Figure 8. (a) Bone erosion in a 15-year-old female adolescent. Coro-nal fast SE intermediate-weighted fat-saturated MR image (4,000/14)depicts 13 3 6 3 11-mm focus (arrows) of high signal intensity withwell-defined margins at the tibial insertion site of the posterior cru-ciate ligament. (b) Compared with the nonenhancing joint effusion(p), this focus (arrow) is enhancing on this sagittal T1-weighted fat-saturated gadolinium-enhanced MR image (400/16) and likely repre-sents intraosseous pannus.
Figure 9. (a) Coronal fast SE T2-weighted fat-saturated MR image(4,000/102) shows abnormal marrow in a 9-year-old girl. Diffuse areas ofhigh signal intensity in the proximal tibial (arrows) and fibular (p)epiphyses have ill-defined borders compared with the well-defined bor-ders of bone erosions. (b) After the intravenous administration of gad-olinium-based contrast material, sagittal T1-weighted fat-saturated MRimage (400/12) shows diffuse enhancement in these areas, as well as inthe patella (p) and distal femoral epiphysis (arrow). Although this patienthad mild synovitis at clinical examination, no substantial joint effusionor synovial proliferation was seen at MR imaging.
Volume 220 z Number 3 Early Juvenile Rheumatoid Arthritis: Knee MR Imaging z 705
software to perform automatic volume seg-mentation; David Glass, MD, and Bill Ball,MD, for helpful suggestions; and Ed Giannini,PhD, for statistical advice.
arthritis. In: Textbook of pediatric rheu-matology. Philadelphia, Pa: Saunders,1995; 135.
2. Martel W, Holt JF, Cassidy JT. Roentgen-ologic manifestations of juvenile rheuma-toid arthritis. Am J Roentgenol 1962; 88:400–423.
3. Harel L, Wagner-Weiner L, Poznanski AK,et al. Effects of methotrexate on radio-logic progression in juvenile rheumatoidarthritis. Arthritis Rheum 1993; 36:1370–1374.
4. Giannini EH, Brewer EJ, Kuzmina N, et al.Methotrexate in resistant juvenile rheu-matoid arthritis: results of the USA-USSRdouble-blind, placebo-controlled trial.N Engl J Med 1992; 326:1043–1049.
5. Lovell DJ, Giannini EH, Reiff A, et al. Et-anercept in children with polyarticularjuvenile rheumatoid arthritis. N EnglJ Med 2000; 342:763–769.
6. Guzman J, Burgos-Vargas R, Duarte-Salazar C, et al. Reliability of the articularexamination in children with juvenilerheumatoid arthritis: interobserver agree-ment and sources of disagreement.J Rheumatol 1995; 22:2331–2336.
7. Forslind K, Larsson EM, Johansson A, etal. Detection of joint pathology by mag-netic resonance imaging in patients withearly rheumatoid arthritis. Br J Rheuma-tol 1997; 36:683–688.
8. Herve-Somma CMP, Sebag GH, PrieurAM, et al. Juvenile rheumatoid arthritis ofthe knee: MR evaluation with GD-DOTA.Radiology 1992; 182:93–98.
9. Senac MO, Deutsch D, Bernstein BH, et al.MR imaging in juvenile rheumatoid ar-thritis. AJR Am J Roentgenol 1988; 150:873–878.
14. Brown DG, Edwards NL, Greer JM, et al.Magnetic resonance imaging in patientswith inflammatory arthritis of the knee.Clin Rheum 1990; 9:73–83.
15. Østergaard M, Stoltenberg M, Gideon P,et al. Changes in synovial membrane andjoint effusion volumes after intraarticularmethylprednisolone: quantitative assess-ment of inflammatory and destructivechanges in arthritis by MRI. J Rheumatol1996; 23:1151–1161.
16. Østergaard M, Hansen M, Stoltenberg M,et al. Magnetic resonance imaging-deter-mined synovial membrane volume as amarker of disease activity and a predictorof progressive joint destruction in thewrists of patients with rheumatoid arthri-tis. Arthritis Rheum 1999; 42:918–929.
17. Brewer EJ, Bass J, Baum J, et al. Currentproposed revision of JRA criteria. ArthritisRheum 1997; 20(2 suppl):195–199.
18. Wallace CA, Levinson JE. Juvenile rheu-matoid arthritis: outcome and treatmentfor the 1990s. Rheum Dis Clin North Am1991; 17:891–905.
20. Potter HG, Linklater JM, Allen AA, et al.Magnetic resonance imaging of articularcartilage in the knee: an evaluation withuse of fast spin-echo imaging. J BoneJoint Surg Am 1998; 80:1276–1284.
21. Yamato M, Tamai K, Yamaguchi T, et al.MRI of the knee in rheumatoid arthritis:Gd-DTPA perfusion dynamics. J ComputAssist Tomogr 1993; 17:781–785.
22. Bezdek JC, Hall LO, Clarke LP. Review ofMR image segmentation techniques us-ing pattern recognition. Med Phys20:1033–1048.
23. Østergaard M, Stoltenberg M, HenriksenO, Lorenzen I. The accuracy of MRI-deter-mined synovial membrane and joint ef-fusion volumes in arthritis. Scand J Rheu-matol 1995; 24:305–311.
24. Pettersson H, Rydholm U. Radiologicclassification of knee joint destruction injuvenile chronic arthritis. Pediatr Radiol1984; 14:419–421.
25. Fleiss JL. Statistical methods for rates andproportions. 2nd ed. New York, NY:Wiley, 1981.
26. Rosner B. Fundamentals of biostatistics.Belmont, Calif: Duxbury, 1995; 426.
27. Konig H, Sieper J, Wolf KJ. Rheumatoidarthritis: evaluation of hypervascular andfibrous pannus with dynamic MR imag-ing enhanced with Gd-DTPA. Radiology1990; 176:473–477.
28. Tamai KA, Yamato M, Yamaguchi T, et al.Dynamic magnetic resonance imagingfor the evaluation of synovitis in patientswith rheumatoid arthritis. Arthritis Rheum1994; 37:1151–1157.
29. Cohen AS, Brandt KD, Krey PR. Synovialfluid. In: Cohen AS, ed. Laboratory diag-nostic procedures in the rheumatic dis-eases. 2nd ed. Boston, Mass: Little,Brown, 1975; 1–64.
30. Gaffney K, Cookson J, Blake D, et al.Quantification of rheumatoid synovitisby magnetic resonance imaging. ArthritisRheum 1995; 38:1610–1617.
31. Østergaard M, Stoltenberg M, HenriksenO. Quantitative assessment of synovialinflammation by dynamic gadolinium-enhanced magnetic resonance imaging: astudy of the effect of intra-articular meth-ylprednisolone on the rate of early syno-vial enhancement. Br J Rheumatol 1996;35:50–59.
32. Østergaard M, Gideon P, Henriksen O, etal. Synovial volume: a marker of diseaseseverity in rheumatoid arthritis? Quanti-fication by MRI. Scand J Rheumatol 1994;23:197–202.
33. Sugimoto H, Takeda A, Kano S. Assess-ment of disease activity in rheumatoidarthritis using magnetic resonance imag-ing: quantification of pannus volume inthe hands. Br J Rheumatol 1998; 37:854–861.
35. Harris ED. Role of collagenase in jointdestruction. In: Sokoloff L, ed. The jointsand synovial fluid. New York, NY: Aca-demic Press, 1978; 243–266.
36. Milgram JW. Juvenile rheumatoid arthri-tis. In: Radiologic and histologic pathol-ogy of nontumorous diseases of bonesand joints. Vol 1. Northbrook, Ill: North-brook Publishing, 1990; 675–682.
37. Barnewolt CE, Shapiro F, Jaramillo D.Normal gadolinium-enhanced MR im-ages of the developing appendicular skel-eton. I. Cartilaginous epiphysis and phy-sis. AJR Am J Roentgenol 1997; 169:183–189.
38. Boegård T, Johansson Å, Rudling O, et al.Gadolinium-DTPA-enhanced MR imag-ing in asymptomatic knees. Acta Radiol1996; 37:877–882.
39. Poleksic L, Zdravkovic D, Jablanovic D, etal. Magnetic resonance imaging of bonedestruction in rheumatoid arthritis: com-parison with radiography. Skeletal Radiol1993; 22:577–580.
40. Salisbury RB, Nottage WM. A new evalu-ation of gross pathologic changes andconcepts of rheumatoid articular carti-lage degeneration. Clin Orthop Rel Res1985; 199:242–247.
41. Klarlund M, Ostergaard M, Rostrup E, etal. Dynamic magnetic resonance imagingof the metacarpophalangeal joints inrheumatoid arthritis, early unclassifiedpolyarthritis, and healthy controls. ScandJ Rheumatol 2000; 29:108–115.
706 z Radiology z September 2001 Gylys-Morin et al
