British Journal of Rheumatology 1996;35:50-59

Transcription

1 British Journal of Rheumatology 1996;35:50-59 QUANTITATIVE ASSESSMENT OF SYNOVIAL INFLAMMATION BY DYNAMIC GADOLINIUM-ENHANCED MAGNETIC RESONANCE IMAGING. A STUDY OF THE EFFECT OF INTRA-ARTICULAR METHYLPREDNISOLONE ON THE RATE OF EARLY SYNOVIAL ENHANCEMENT M. 0STERGAARD, M. STOLTENBERG, O. HENRIKSEN and I. LORENZEN Danish Research Centre of Magnetic Resonance and Department of Rheumatology, Hvidovre Hospital, University of Copenhagen, Denmark SUMMARY The effect of temporary inflammatory suppression on synovial membrane enhancement, as determined by dynamic and static gadolinium-dtpa enhanced magnetic resonance imaging (MRI), was studied. MRI of 18 arthritic knees was performed before and 1, 7, 30 and 180 days after intra-articular methylprednisolone injection until clinical relapse. MRI of another six Icnees was performed twice within 2-4 days in order to assess interobserver and inter-mri variation. The rate of early enhancement of the entire synovial membrane of a pre-selected central sagittal slice (REE^a), determined by dynamic Tl -weighted FLASH MRI, decreased in all knees within the first post-treatment week and remained low during remission, although gradually increasing. In cases of clinical relapse, REE^B4a increased to pre-treatment levels. The interobserver plus inter-mri variation was maximally 27%. The predictive values of a REE^^ below/above 1.0%/s were 0.94 and 0.91, respectively, with respect to the absence/presence of clinical synovitis. Evaluation of small synovial areas revealed marked regional heterogeneity. Static spin echo MRI was not informative. The study indicates that the rate of early synovial enhancement reflects synovial inflammatory activity. Subclinical changes may be revealed. Evaluation of large synovial areas increases reproducibility and reduces the effect of regional heterogeneity. Dynamic MRI may prove a clinically useful measure of synovial inflammation. KEY WORDS: MRI, Gadolinium, Gadopentetate dimeglumine, Rheumatoid arthritis, Arthritis, Synovitis, Knee. ESTIMATION of the severity of synovial inflammation is important in the assessment of patients with rheumatoid arthritis (RA) and other inflammatory joint diseases. Evaluation of the synovial membrane is generally based on physical examination of the joint and indirect laboratory parameters. Neither these methods nor conventional radiography, which only show the late results of preceding synovitis, allow detailed evaluation of the synovial inflammation [1-3]. Magnetic resonance imaging (MRI) allows detailed visualization of the tissues involved in arthritis, including the inflamed synovial membrane. After i.v. administration of the contrast agent gadolinium- DTPA (Gd-DTPA), the synovial membrane can be differentiated from the surrounding tissues, due to a marked increase in signal intensity (enhancement) on Tl-weighted images [3-10]. In addition to joint imaging, MRI perhaps provides a measure of synovial inflammatory activity. Following i.v. injection, Gd-DTPA is transported unlinked in the plasma and passes into the interstitial space, depending on the local capillary permeability and the tissue perfusion [11]. The blood flow, as well as the capillary permeability, is increased during inflammation. Consequently, the enhancement during the first minutes after Gd-DTPA administration, reflecting the initial distribution of Gd-DTPA, may indicate the Submitted 6 April 1995; revised version accepted 10 August Correspondence to: M. 0stergaard, Department of Rheumatology, HvidovTe Hospital, DK-2650 Hvidovre, Denmark. inflammatory activity of the joint. Registration of the time-dependent changes in the synovial signal intensity following Gd-DTPA ('dynamic imaging') by means of the gradient-echo technique FLASH (fast low angle shot) has been introduced as a potential marker of synovial inflammation [4,5,12-16]. Other studies have considered a marked synovial enhancement on Tl-weighted conventional 'static' spin echo MR images indicative of synovial inflammation [6, 7]. However, the reproducibility of the methods as well as their information value with respect to synovial inflammatory activity remain to be determined. Intra-articular corticosteroid injections cause a fast and pronounced suppression of the local joint inflammation. Usually, the effect is only temporary and joint inflammation recurs within a few months [17]. The predictable post-treatment course of the inflammatory activity makes intra-articular corticosteroid treatment suitable for longitudinal studies of the relationship between synovial enhancement and inflammatory activity. The objective of the present 6 month longitudinal study was to analyse the course of the synovial membrane enhancement following intra-articular methylprednisolone injection (protocol A). Hereby we wanted to evaluate dynamic and static MRI as methods for the quantitative assessment of synovial membrane inflammation. The reproducibility of the methods was examined by repeated MR imaging before treatment (inter-mri variation) and by the determination of interobserver variation (protocol B) British Society for Rheumatology

2 0STERGAARD ET AL.: ASSESSMENT OF SYNOVITIS BY DYNAMIC MRI 51 PATIENTS AND METHODS Patients Twenty-four knees of 22 patients were included: 18 knees of 16 patients in protocol A and six knees of six patients in protocol B. Twenty-three knees had clinical signs of gonarthritis, i.e. joint swelling and/or joint tenderness, and a clinical indication for knee arthrocentesis and intra-articular corticosteroid injection. The underlying disease was RA in 15 knees, psoriatic arthritis in two knees and unspecified monoarthritis in three knees. The four remaining patients suffered from juvenile rheumatoid arthritis (JRA), reactive arthritis, enteropathic arthritis and clinical osteoarthritis of the knee, respectively. Two patients with RA had both knees treated within an interval of 1 month. One patient with RA included in protocol B had a history of knee joint synovitis, but at the time of the examination the knee was clinically inactive, i.e. had neither clinical joint swelling nor joint tenderness. The median age of the patients was 59 yr (range yr). The median duration of disease and knee symptoms was 5.5 yr (range 6 months-36 yr) and 2.5 yr (range 1 month-15yr), respectively. Three patients received no other treatment than the intraarticular methylprednisolone injection described below, while 14 patients received non-steroidal antiinflammatory drugs (NSAIDs). Seven patients received continuous therapy with low-dose oral prednisolone, alone (four patients) or in combination with a diseasemodifying drug (DMARD) (three patients). No intraarticular therapy had been administered within the last 4 weeks. Clinical procedure All patients, except the RA patients without clinical signs of synovitis, underwent knee joint arthrocentesis. During arthrocentesis, as much joint fluid as possible was aspirated, and 6 ml lidocaine 0.5% (lidocaine hydrochloride 5 mg/ml) and 80 mg methylprednisolone acetate (2 ml, 40 mg/ml) were injected. In 25 of the 27 treated knees, joint fluid could be aspirated (1-90 ml, median 25 ml). In the remaining two knees, correct intra-articular injection was verified by observation of an increased amount of fluid in the suprapatellar recess on sagittal spin echo MR images obtained immediately after arthrocentesis. In protocol A, MRI was performed just before arthrocentesis (day 0). Follow-up MRI was performed on days 1, 7, 30 and 180 until clinical relapse occurred. In cases of clinical relapse, MRI was performed and follow-up ended. Clinical relapse was diagnosed in the presence of joint swelling and/or joint tenderness. The erythrocyte sedimentation rate (ESR) and serum C-reactive protein (s-crp) were measured at each MRI session. In protocol B, MRI was performed 1-3 days before (day 1) and immediately before arthrocentesis (day 0) as well as at day 7 after arthrocentesis. In the RA patient without clinical signs of synovitis, MRI was performed twice, on two successive days. The variation in the enhancement values between the two pretreatment MRI sessions was determined (inter-mri variation). To evaluate the interobserver variation, two observers separately evaluated the MR images obtained on days 1 and 0. MRI The MR images were obtained using a 1.5 tesla Siemens MR unit (19 patients) or a 1.0 tesla Siemens Impact MR unit (five patients, all included in protocol B). The knees examined were positioned neutrally rotated in a dedicated knee coil. Continuous sagittal Tl-weighted spin echo images (TR/TE/slice thickness = ms/15-17 ms/5 mm) were obtained. A series of 30 Tl-weighted FLASH (fast low angle shot) images (TR/TE/flip angle/slice thickness = 40 ms/12ms/70 /5 mm), each with an imaging time of 10 s, was then performed in the same, pre-selected, sagittal slice (dynamic imaging). During the second period, 0.05 mmol Gd-DTPA/kg body weight was injected into a cubital vein, while the patient remained in the same position in the MR unit. The FLASH sequence covered the enhancement during the first 285 post-contrast seconds. The dynamic FLASH sequence only allowed us to examine one slice. We chose to pre-select the sagittal slice through the insertion of the posterior cruciate ligament on the femoral bone. This slice was chosen because the synovium in the anterior (the suprapatellar recess) as well as the posterior (around the cruciate ligaments) part of the joint was well represented. Furthermore, the slice was easy to reproduce. Finally, the Tl-weighted spin echo sequence was repeated. In all sequences, the matrix size was x 256 and the field of view (FOV) mm. The early synovial enhancement, determined by dynamic Tl-weighted FLASH MRI The image-processing software package XPrime, installed on a Sun Sparc 10 computer (Unix), allowed subtraction of images and outlining of areas of interest. The image of t 0 was subtracted from the image of t ns. On this subtraction image the synovial membrane, that showed signal enhancement, was outlined (Fig. 1). The mean signal intensity of the outlined area, i.e. the entire synovial membrane of the slice, to each time was automatically calculated. The relative enhancement at t seconds C&I/^FLASH) ano< the relative synovial enhancement per second during the first 55 s (rate of early synovial enhancement, REE^^ was calculated by the following formulae: Relative enhancement at t seconds SI, - SI 0 SI, x 100% Rate of early enhancement (REE) = 100% SI0 x 55 where SI 0 and SI, are the signal intensities before and t seconds after contrast injection (Fig. 2). Evaluation of the enhancement after 55 s was chosen, because a

3 52 BRITISH JOURNAL OF RHEUMATOLOGY VOL. 35 NO. 1 The synovial enhancement on static Tl-weighted spin echo MR images The synovial membrane of the same slice as examined by FLASH imaging was outlined by means of image-processing software as described above. The outlining was done on post-gd-dtpa images, guided by subtraction images. The relative synovial enhancement on the spin echo images (ESE^J was calculated by means of the following formula: Relative synovial enhancement (ESE^J = I+G t7 ^ x 100% where S/.Q,, and SI+oa indicate the signal intensities of the synovium before and after Gd-DTPA administration. Statistical methods Non-parametric methods were used to analyse the data. Analysis of statistical correlation was performed by the Spearman test of rank correlation. The Mann-Whitney test (M-W, two-sample rank sum test) was used to analyse differences between groups of patients, while the Wilcoxon-Pratt test (W-P, one-sample rank sum test) was used to analyse changes in MRI parameters within the patient. RESULTS Protocol A Within the follow-up period of 6 months, clinical relapse occurred in 13 out of 18 knees included in protocol A, while five patients remained in clinical remission. The final MRI was not obtained in two remission knees and five relapse knees, respectively, due to technical problems and repeated glucocorticoid instillation without preceding MRI. The rate of early enhancement of the entire synovial membrane in the pre-selected slice (REE^^d ranged from 0.8 to 2.0%/s (median 1.2%/s) before arthrocentesis and glucocorticosteroid instillation (Table I). Seven days after treatment, REE^^ ranged from 0.1 to 1.0%/s (median 0.5%/s). All knees experienced a decrease in REE^xoi during the first 7 days after treatment. Thus, the W-P test revealed a highly statistically significant difference (P < 0.001). The decrease ranged from 0.1 to 1.5%/s (median 0.8%/s), corresponding to a relative decrease in REE^^ of 8-91% (median 64%). Fig. la, b Caption opposite. previous study showed maximal enhancement difference between knees with clinically active and clinically inactive arthritis in the interval from 50 to 90s after Gd-DTPA injection [12]. The rates of early enhancement REE^^ of the synovial membrane at four pre-selected locations in the pre-selected slice were also determined, as were REE values of joint effusion, muscle, bone marrow and subcutaneous fat. The four selected synovial membrane locations were: (1) the anterior wall of the suprapatellar recess, 1 cm above its insertion on the patella; (2) the top of the suprapatellar recess; (3) the posterior wall of the suprapatellar recess, just above its insertion on the femoral bone; (4) in front of the posterior cruciate ligament, as close as possible to its insertion on the femoral bone.

4 0STERGAARD ET AL.: ASSESSMENT OF SYNOVTTIS BY DYNAMIC MRI 53 From day 7 to clinical relapse, REE^^ had increased in all knees (W-P, P < 0.05). The increase ranged from 0.4 to 1.7%/s (median 0.9%/s), corresponding to a relative increase of % (median 60%). At clinical relapse, REE m ranged from 1.0 to 2.0%/s (median 1.4%/s), i.e. the early enhancement had increased to pre-treatment levels. No statistically significant difference between pre-treatment and relapse REE^^ was found (W-P, P = 0.74). In knees that were still in clinical remission at day 30, Fio. 1. (c-f) FIG. 1. Sagittal Tl-weighted FLASH images obtained before [(ahe)] and 7 days after ( 0 intra-articular methylprednisolone injection. Images are at times t3 (a), tu (b) and fjts (c)- The synovial signal intensity increases with time ( ), especially from /0 to t^. (d) Subtraction image (c a; 'as ~ 'o)- The enhancing synovial membrane appears white. This image was used for the outliningg (white lines) of the synovial membrane, (e) Subtraction image (c - a; ta - «). The rate of early enhancement (REE^ l l t d ffrom th h t oon this a g e REE was calculated the enhancement this iimage was 1.3%/*. ( 0 Subtraction image / - I* as (e), but obtained at day 7. was 0.4%/s. Thus, in this patient REE,,^, decreased from day 0 to day 7.

5 54 BRITISH JOURNAL OF RHEUMATOLOGY VOL. 35 NO. 1 (c) 140 E g 1 I a 60 j I 20 1 n D Q Time(Kx) ZC Mlo-Mo CD 2SV75% a Mcdnnhie DayO Day! Day7 Day 30 Day ISO Rdipse a o o 31 Min-Mtt CD25%-75% 0 Medunvihw o J I 1 a Day 0 Day 1 Day 7 Day 30 Day 180 Rdapte ' FIG. 2. (a) Example of the time-dependent changes in the relative enhancement of the entire synovial membrane of the pre-selected slice ( J E"' l tiaash^jj ont1 - wci 8l lted FLASH images before (day 0) and after (days 1,7, 30, 180 and clinical relapse) intra-articular methylprednisolone injection. The synovial enhancement during the first post-gd-dtpa minute is markedly higher before treatment and at clinical relapse than during clinical remission. As generally found in this study, the lowest enhancement is measured at day 7. At day 30, the enhancement has increased slightly, indicating an increased synovial inflammation, despite continued clinical remission, (b) and (c) Box-and-whiskers plots of the synovial enhancement of the pre-selected slice on (b) dynamic Tl-weighted FLASH MRI (REE^^J and (c) on static Tl-weighted spin echo MRI (ESE^J. All patients in protocol A are included.

6 No. of knees TABLE I Pre- and post-treatment values of REE and ESE (protocol A) Rate of early enhancement (REE) (%/s) Entire synovium Synovium 1 Synovium 2 Synovium 3 Synovium 4 DayO ( ) 1.4( ) 1.0( ) Dayl ( ) 1.1( ) 0.5( ) Day ( ) 0.5( ) 0.3( ) Day ( ) 0.3 (- 0.5 to + 1.9) 0.6( ) Day ( ) 0.4( ) 0.4( ) Relapse 8 1.4( ) 1.7( ) 1.3( ) 1.3( ) 0.8( ) 0.2 (-0.2 to +2.5) 0.6 (-0.1 to +2.1) 0.4( ) 1.9( ) 1.4( ) 1.0( ) 0.5( ) 0.8( ) 0.7( ) Joint fluid 0.0 ( ) 0.0 (-0.2 to +0.3) 0.0 ( ) 0.0 (-0.2 to +0.2) 1.4( ) 0.1( ) Muscle 0.2 ( ) 0.2( ) 0.1( ) 0.1 ( ) 0.1 ( ) 0.1 ( ) Median values are given, with the range in parentheses. Synovium 1-4 refer to four small synovial areas at four pre-selected locations (see the text). REErefersto the rate of early enhancement, while ESE indicates therelativeenhancement on spin echo images. Fatty tissue and bone marrow Entire synovium 71(28-116) 55(17-95) 45(23-81) 54 (4-93) 75(41-77) 75(49-79) 90 O 5 2 o D > O 2

7 56 BRITISH JOURNAL OF RHEUMATOLOGY VOL. 35 NO. 1 remained below pre-treatment values (W-P, P < 0.01). However, 10 out of 12 knees in clinical remission at day 30 had experienced a REE^^ increase compared to day 7. The increase ranged from 0.1 %/s to 0.8%/s (median 0.3%/s), corresponding to a relative increase of 4-400% (median 76%). This suggests an increased synovial inflammation from day 7 to day 30, despite continued clinical remission. The increase was statistically significant (W-P, P < 0.05), but also significantly smaller than the increase from day 7 to clinical relapse (M-W, P < 0.01). REE^^ at day 1 was somewhere in between values at day 0 and day 7, statistically significantly lower than at day 0 and higher than at day 7 (W-P; both P < 0.01). In Fig. 2b, the range, quartile and median values of REE^M at the different times of MRI are illustrated. Knees in clinical remission at day 7, 30 and 180 had a REE^n of maximally 1.05%/s, and only in 7 of 33 cases (21%) was REE^^ above 0.8%/s. REE^^ before treatment and at relapse was always above 0.8%/s and only in two cases (9%) below 1.0%/s. In other words, a REE^^ of > 1.1 %/s was only found in knees with clinically active arthritis, while a REE^^M of <0.8%/s was only found in knees in clinical remission. The predictive value of a REE,^^ < 1.0%/s was 0.94 with respect to clinical remission, while a REE tjn4ot > 1.0%/s had a predictive value of 0.91 with respect to clinical signs of synovitis. The rate of early synovial enhancement of the smaller areas of synovial membrane (REE^,^) followed the same pattern as REE^^^ i.e. a decrease from day 0 to day 7 followed by an increase in cases of clinical relapse (Table I). However, the overlap between the remission values (day 7, 30, 180) and pre-treatment/relapse values was far more pronounced (Table I). Absolute values of REE^,^ varied consider- Interobserver variation Observer 1 Observer 2 Obs 1-Obs2 Obil-Obs2 in/wi x 100% Obs 1 Inter-MRI variation MRI 1 (day -1) MRI 2 (day 0) MRI 1-MRI 2 TABLE II Interobferver and inter-mri variation (protocol B) 0.9 ( ) 0.9( ) 0.1 ( ) 6% (0-27%) 0.9 ( ) 1.0 ( ) 0.1 ( ) ably from REE^H,, suggesting a regional heterogeneity of the synovial inflammation. The median difference between REE^^ and the corresponding REE^^m was 0.27%/s (range %/s). The median relative difference was 38% (range 0-520%). The largest relative differences were found when REE^,^ was low. The enhancement of the synovial membrane on the 'static' spin echo images (ESE^J is given in Table I and illustrated in Fig. 2c. In most knees, ESE^ decreased from day 0 to day 7, and increased at clinical relapse. The change in ESE^ was statistically significant (W-P; day 0 vs day 7: P < 0.01; day 7 vs relapse: P < 0.05). The pattern was thus similar to what was found with respect to REE^^^. However, we found considerable overlap between the remission values (days 7, 30 and 180) and pre-treatment/relapse values. Correspondingly, no significant differences were found between ESE at relapse and ESE^, at day 30 and day 180 (M-W tests), while differences in REE^M were statistically significant (M-W; day 30 vs relapse: P<0.01; day 180 vs relapse: P < 0.05). REE rf^xox and ESE^, were not statistically significantly correlated to either s-crp or ESR. The Spearman correlation coefficients were as follows: REE^w vs s-crp: r = 0.24 (P = 0.08 for uncorrelated values); REE^tot vs ESR: r = 0.21 (P = 0.13); ESE^ vs s-crp: r = 0.21 (P = 0.12); ESE^ vs ESR: r = 0.18 (P = 0.21). REE values of joint fluid, muscle, fatty tissue and bone marrow were markedly lower than (Table I). Protocol B Six knees were examined by MRI twice within 3 days. No clinical changes were observed between the two MRI sessions. The maximal relative interobserver variation was 27% (median 6%), while the 1.1 ( ) 1.0 ( ) 0.1 ( ) 13% (0-107%) 1.1 ( ) 0.9 ( ) 0.3 ( ), (%/s) ESE-, (%) 84 (64-108) 83(58-115) 5(0-11) 6% (0-14%) 90(58-115) 81 (62-108) 7 (0-23) MRI I -MRI 2 x 100% MRI Interobserver plus inter-mri variation Highest value lowest value Highest value lowest value x 100% Lowest value 15% (8-23%) 0.2 ( ) 22% (18-27%) 27% (0-107%) 0.3 ( ) 35% (0-317%) Median values are given, with the range in parentheses. indicates numerical values. ^ lynjot and enhancement of the entire synovial membrane of the pre-selected slice and four small pre-selected synovial areas, respectively. the enhancement of the entire synovial membrane of the pre-selected slice on static spin echo images. 7% (0-23%) 23 (4-15) 19% (7-29%) refer to the rates of early indicates Interobserver plus inter-mri variations were calculated by comparing the highest and the lowest of the four enhancement values obtained by two observers evaluating two successive MRIs.

8 0STERGAARD ET AL:. ASSESSMENT OF SYNOVITIS BY DYNAMIC MRI 57 maximal relative inter-mri variation was 23% (median 15%). The maximal relative interobserver plus inter-mri variation was 27% (median 22%). Detailed information in given in Table EL The relative interobserver and inter-mri variations of ESE vn were similar (Table II). The maximal relative interobserver plus inter-mri variation was 29% (median 19%). The variations in REE of the four small synovial areas (REE^^) were considerably larger (Table IT). The maximal relative interobserver plus inter-mri variation was 317% (median 35%). The absolute interobserver variation of REE values of joint fluid, muscle, bone marrow and fatty tissue was <0.1%/s in all knees, as was the absolute inter-mri variation. MRI was repeated at day 7 after intra-articular glucocorticoid installation in order to ensure that knees included in protocol B were not essentially different from those included in protocol A. The synovial enhancement was determined by observer 1. REE^^ and ESE,^ had decreased in all knees compared to day 0, and REEL^^ had decreased in 19 out of 20 areas examined. The relative decrease from day 0 to day 7 ranged from 39 to 80% (median 56%) with respect to REE tybju>l, while the decrease in REE^,_4 and ESE^ ranged from -22 to +92% (median +56%) and from 33 to 77% (median 44%), respectively. These results are similar to findings in protocol A. DISCUSSION In the present study, temporary suppression of local knee joint synovitis was obtained by an intra-articular methylprednisolone injection. MRI was performed before treatment, several times during clinical remission and in cases of clinical relapse of synovitis. Hereby, we evaluated the relationship between clinical changes in synovial inflamation and two potential MRI-determined markers of synovial inflammation: the rate of early synovial enhancement (REE^), as determined by dynamic FLASH MR imaging, and the relative synovial enhancement on static spin echo MR images (ESE^). The changes in the rate of early enhancement of the entire synovial membrane of the selected slice (REE^ux) were in absolute accordance with the expectations of a marker of synovial inflammation, exposed to a temporary inflammatory suppression. REE^njux decreased in all patients within the first week after treatment. Already at day 1 after treatment, the decrease was statistically significant, and at day 7 all knees had experienced a marked (median 64%) decrease. In knees with clinical relapse, ^j increased to values of the same magnitude as before treatment. The changes in knees that were still in clinical remission at day 30 and day 180 continued to have a REE^^ markedly lower than before treatment. However, REE^^ increased in 10 of 12 knees from day 7 to day 30, despite no clinical changes. As the effect of intra-articular glucocorticoids generally subsides within the first post-treatment month [17], it is probable that the increase in REE,,,^ from day 7 to day 30reflectsa subclinical increase in synovial inflammation. At day 1, REE^^ showed a wide range of values, overlapping pre-treatment/relapse values as well as remission values, possibly reflecting various stages between acute inflammation and remission. The overlap between the remission values (days 7, 30 and 180) and the pre-treatment/relapse values was limited (Fig. 2b). In this study, a REE tyn, {oi < 1.0%/s had a predictive value of 0.94 with respect to clinical remission, and a REE^^ > 1.0%/s had a predictive value of 0.91 with respect to clinical signs of synovitis. -fce^iyimm < 0.8%/s was found in 79% of the knees in clinical remission, but not in any pre-treatment/relapse knees, while REE WfnMn > 1.1 %/s was found in 74% of the pre-treatment/relapse knees, but not in any knees in clinical remission. This suggests that REE^^ may be used to assess synovial inflammation, for instance in joints inaccessible to proper clinical examination (e.g. the hip or the sacroiliac joints) or in joints with suspected but not definite synovitis. However, evaluation of the predictive value of REE^^n with respect to histological synovial inflammation is needed to determine the value of dynamic MRI in arthritis. The reproducibility of the methods should also be considered (see below). The overall pattern of changes in synovial enhancement on spin echo images (ESE^J during the inflammatory suppression was similar to the changes in REE^w However, a considerable overlap between remission and pre-treatment/relapse values minimized the predictive value of ESE vn with respect to clinical signs of synovial inflammation (Fig. 2c). Neither REE^M nor ESE~ were correlated to the laboratory parameters of inflammation, probably because the laboratory parameters reflect inflammation of all joints, while REE reflects local joint inflammation. The reproducibility of the methods was evaluated. The median inter-mri variations were 15% (REE^^^) and 7% (ESE,^), while the median interobserver variation of both REE^ut and ESE^ was 7%. Taking inter-mri as well as interobserver variations into account, the enhancement of the synovial membrane of the pre-selected slice could be determined with a maximal relative variation of <30%, both on dynamic FLASH images and static spin echo images. Unavoidably, small amounts of joint fluid will sometimes be included in the outlined synovial membrane areas, due to partial volume phenomena. Furthermore, diffusion of Gd-DTPA into the joint fluid causes gradual enhancement of effusions. However, since this occurs at a much slower rate than in the synovium [8, 18,19], misinterpretations of the svnovium-effusion borderline are probably not of major importance on the early images used in the present study. Gd-DTPA-enhanced dynamic MRI of arthritic joints has been described in several earlier studies [4, 5, 12-16]. Two studies compare dynamic MRI findings with pathology [4,16]. In a study by Konig et al. [4], including eight patients, dynamic MRI could differentiate fibrous, slightly hypervascular and hypervascular pannus. This was not possible in a study by

9 58 BRITISH JOURNAL OF RHEUMATOLOGY VOL. 35 NO. 1 Tamai et al. [16] including 10 knees, although specimens with histological signs of severe inflammation showed a statistically significantly higher synovial enhancement than specimens with mild inflammatory changes. Reiser et al. [5] found no obvious correlation between the early synovial enhancement and clinical estimation of inflammatory activity, contrary to recent studies [12, 13, 15]. Konig et al. [13] found that the rate of signal increase was significantly reduced 2-3 weeks after intra-articular glucocorticoid installation. Systematic longitudinal studies and evaluation of the reproducibility of the methods have not been performed. In studies by other groups, the enhancement was measured in small circular areas of the synovium by means of the scanner software. In the present study, as well as in two previous studies by our group [12, 14], the signal intensity increase of the entire enhancing synovial tissue of a pre-selected slice was quantitatively analysed. Analysis of the enhancement of large volumes of synovial membrane, e.g. the entire synovial membrane of a slice, instead of small areas, reduces the variation caused by regional heterogeneity. In the present study, examination of the early enhancement of four small areas of synovial membrane revealed a large variation (median 38%, maximally 520%) from the enhancement of the entire synovium. This indicates a heterogeneity of the synovial inflammation, as described in studies evaluating the intra-articular variation of histological signs of synovitis [20, 21]. The considerable regional heterogeneity implies that evaluation of large synovial areas is necessary, if determination of the enhancement should be usable for determination of the inflammatory status of a joint, in clinical trials or in clinical routine. ^E^iyn.i-4 showed considerable overlap between remission and pre-treatment/relapse values and, furthermore, large maximal interobserver and inter-mri variations. This is probably partly caused by regional heterogeneities in prc-treatment synovial inflammation and in treatment-induced inflammatory suppression. Contributory causes may also be minor differences in patient positioning and/or selection of region of interest, resulting in unidentical areas of evaluation. Regional inflammatory heterogeneity, combined with a lack of exact spatial identity, may invalidate comparisons of histological and MRI-determined synovial inflammation. This problem may explain the lack of predictive value of the early synovial enhancement with respect to histological inflammation in the study by Tamai et al. [16]. The present study strongly supports the view that the early synovial enhancement reflects synovial inflnmmntory activity. Subclinical changes may be revealed. Evaluation of large synovial areas increases reproducibility and reduces the effect of regional heterogeneity. Dynamic MRI may prove a clinically useful measure of synovial inflammation. ACKNOWLEDGEMENTS We acknowledge the Foundation of , the University of Copenhagen, the Thomas & Elisabeth Fralund Nielsen Foundation, the Danish Rheumatism Association and the Danish Medical Research Council for financial support. Schering Diagnostika, Denmark, is acknowledged for providing the contrast agent. We thank Steen Ahlmann for skilful technical assistance, and Jens Arnth Jensen and Poul Ring for developing the image-processing software package XPrime. REFERENCES 1. Scott DL. European preferences in assessing rheumatoid arthritis. / Rheumatol 1993^0: Felson DT. Choosing a core set of disease activity measures for rheumatoid arthritis clinical trials. J Rheumatol 1993^0: Kaye JJ. Arthritis: Roles of radiography and other imaging techniques in evaluation. Radiology 1990; 177: Konig H, Sieper J, Wolf KJ. Rheumatoid arthritis: Evaluation of hypervascular and fibrous pannus with dynamic MR imaging enhanced with gd-dtpa. Radiology 1990;176: Reiser MF, Bongartz GP, Erlemann R et al. Gadolinium-DTPA in rheumatoid arthritis and related diseases: First results with dynamic magnetic resonance imaging. Skel Radiol 1989;18: Adam G, Dammer M, Bohndorf K, Christoph R, Fenke F, Giinther RW. Rheumatoid arthritis of the knee: Value of gadopentetate dimeglumine-enhanced MR imaging. Am J Roentgenol 1991;15& Bjorkengren AG, Geborek P, Rydholm U, Holtas S, Petterson H. MR imaging of the knee in acute rheumatoid arthritis: Synovial uptake of gadolinium-dota. Am J Roentgenol 1990;155: Herve-Somma CMP, Sebag GH, Prieur A, Bonnerot V, Lallemand DP. Juvenile rheumatoid arthritis of the knee: MR evaluation with Gd-DOTA. Radiology 1992; stergaard M, Gideon P, Henriksen O, Lorenzen I. Synovial volume a marker of disease severity in rheumatoid arthritis? Quantification by MRI. Scand J Rheumatol 1994^3: Corvetta A, Giovagnoni A, Baldclli S et al. MR imaging of rheumatoid hand lesions: comparison with conventional radiology in 31 patients. Clin Exp Rheumatol 1992; Weinmann H-J, Brasch RC, Press W-R, Wesbye GE. Characteristics of gadolinium-dtpa complex: A potential NMR contrast agent. Am J Roentgenol 1984;142: stergaard M, Lorenzen I, Henriksen O. Dynamic gadolinium-enhanced MR imaging in active and inactive immunoinflammatory gonarthritis. Ada Radiol 1994; 35: Konig H, Bolze X, Sieper J, Wolf KJ. Quantitativ evaluierte dynamische magnetresonanztomographie bei chronischer polyarthritis des kniegelenkes. Therapiekontrolle nach intraartikularer kortisonapplikation. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1992; 157: stergaard M, Stoltenberg M, Gideon P et al. Effect of intraarticular osmic acid on synovial membrane volume and inflammation, determined by magnetic resonance imaging. Scand J Rheumatol 1995^4: Braun J, Bollow M, Eggens U, Konig H, Distler A, Sieper J. Use of dynamic magnetic resonance imaging

10 0STERGAARD ET AL.\ ASSESSMENT OF SYNOVTTIS BY DYNAMIC MRI 59 with fast imaging in the detection of early and advanced sacroiliitis in spondylarthropathy patients. Arthritis Rheum 1994;7: Tamai K, Yamato M, Yamaguchi T, Ohno W. Dynamic magnetic resonance imaging for the evaluation of synovitis in patients with rheumatoid arthritis. Arthritis Rheum 1994;& Bird HA. Pharmacological approaches. Intra-articular and intralesional therapy. In: Klippel JH, Dieppe PA, eds. Rheumatology. London: Mosby-Year Book Europe Limited, 1994: Yamato M, Tamai K, Yamaguchi T, Ohno W. MRI/)f the knee in rheumatoid arthritis: Gd-DTPA perfusion dynamics. / Comput Assist Tomogr 1993; 17: Drape J, Thelen P, Gay-Depassier P, Silbennann O, Benacerraf R. Intraarticular diffusion of Gd- DOTA after intravenous injection in the knee: MR imaging evaluation. Radiology 1993; 18& Lindblad S, Hedfors E. Intraarticular variation in synovitis. Local macroscopic and microscopic signs of inflammatory activity are significantly correlated. Arthritis Rheum 1985O& Hutton CW, Hinton C, Dieppe PA. Intra-articular variation of synovial changes in knee arthritis: Biopsy study comparing changes in patellofemoral synovium and the medial tibiofemoral synovium. Br J Rheumatol 1987;26: 5-8.