Elucidation of the relationship between synovitis and bone damage: A randomized magnetic resonance imaging study of individual joints in patients with early rheumatoid arthritis




To simultaneously image bone and synovium in the individual joints characteristically involved in early rheumatoid arthritis (RA).


Forty patients with early, untreated RA underwent gadolinium-enhanced magnetic resonance imaging (MRI) of the second through fifth metacarpophalangeal joints of the dominant hand at presentation, 3 months, and 12 months. In the first phase (0–3 months), patients were randomized to receive either methotrexate alone (MTX) or MTX and intraarticular corticosteroids (MTX + IAST) into all joints with clinically active RA. The MTX-alone group received no further corticosteroids until the second phase (3–12 months), when both groups received standard therapy.


In the first phase, MTX + IAST reduced synovitis scores more than MTX alone. There were significantly fewer joints with new erosions on MRI in the former group compared with the latter. During the second phase, the synovitis scores were equivalent and a similar number of joints in each group showed new erosions on MRI. In both phases, there was a close correlation between the degree of synovitis and the number of new erosions, with the area under the curve for MRI synovitis the only significant predictor of bone damage progression. In individual joints, there was a threshold effect on new bone damage related to the level of synovitis; no erosions occurred in joints without synovitis.


In early RA, synovitis appears to be the primary abnormality, and bone damage occurs in proportion to the level of synovitis but not in its absence. In the treatment of patients with RA, outcome measures and therapies should focus on synovitis.

Rheumatoid arthritis (RA) is a chronic inflammatory polyarthritis characterized by widespread synovitis and joint destruction (1). However, the relationship between synovitis and joint damage (conventionally described by radiographic erosions) remains controversial. Some studies have shown progression of joint destruction despite suppression of synovitis (2–6). Other studies have indicated that despite no change in clinical synovitis measures, certain therapies retard bone damage (7). More recently, it has been demonstrated that the suppression of disease activity slows or even halts progression of bone damage, although there was a poor correlation between clinical response and radiographic change (8, 9). Such clinical and radiographic observations, where synovitis and bone damage are seemingly independent processes, have been supported by experimental models of RA, where joint damage may be uncoupled from synovitis (10). All these studies used either indirect or insensitive measures or imaged complex joints such as the wrist, with consequent difficulty in interpretation. However, a recent analysis of hand joints in an early RA cohort (using tender and swollen joint counts as surrogate measures of synovitis) did demonstrate a close relationship between suppression of clinical synovitis and reduction of erosion progression (11).

An understanding of the interrelationship between synovitis and bone damage is critical for the optimal management of RA, especially in determining a logical approach for drug treatment, and provides the model for all diseases in which chronic inflammation is a dominant feature. Magnetic resonance imaging (MRI) offers advantages over clinical examination and radiographic techniques in its ability to simultaneously assess both synovitis and bone changes (12–19). We have previously used MRI of the metacarpophalangeal (MCP) joints as a model system for measuring damage (20) because these are the characteristically affected joints of early rheumatoid disease. Furthermore, their anatomy permits the most reproducible and interpretable documentation of the interrelationship between synovitis and bone damage at the individual joint level. Patients were studied at the onset of disease to minimize the confounding effects of drugs, disability, and disease duration.

Our previous cross-sectional study of early disease had suggested a close association between synovitis and bone damage (20). In the current prospective longitudinal study, undertaken with previously untreated patients, the same approach was used to assess individual joint disease and in particular the relationship between synovitis and bone damage over time. This was done in a model 2-stage system with 2 standard therapies known to produce differential rates of synovitis reduction. Patients were randomized to receive either methotrexate (MTX) alone, a relatively slow-acting agent, or MTX and an intraarticular corticosteroid (MTX + IAST), the latter drug being well known for its rapid reduction of synovitis.



Between December 1, 1997, and November 30, 1998, we recruited a total of 42 consecutive patients from an early arthritis clinic, all of whom fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) criteria for RA (21). Entry criteria included disease duration of less than 12 months and the presence of clinical MCP joint disease. Patients who had previously been treated with any disease-modifying antirheumatic drugs (DMARDs) or who were taking oral or intramuscular (IM) corticosteroids were excluded. All patients gave their consent to the study, and the protocol was approved by the local research ethics committee of Leeds Teaching Hospital Trust.

Patients were evaluated at baseline for age and known severity factors, such as sex, disease duration, rheumatoid factor (RF) positivity, shared epitope positivity (previously known as HLA–DR1 or DR4, performed using molecular typing for HLA alleles DRB1*0101, *0102, *0401, *0404, *0405, *0408, and *1001), and the presence of radiographic erosions (based on standard hand and foot radiographs). The following were recorded at each visit (baseline and 1, 2, 3, 6, 9, and 12 months): early morning stiffness, patient assessment of disease activity on a visual analog scale, tender and swollen joint counts, modified Health Assessment Questionnaire (22) raw scores, and C-reactive protein levels. All patients had repeat radiographs performed at 12 months.

Treatment and MRI protocol.

All patients had MRI scans performed on the day of their entry into the study and were then randomized, using a computer-generated list kept by an independent research nurse, to receive either IAST (methylprednisolone) into all joints with clinically active RA (defined as both tender and swollen) or no corticosteroid. All patients were started on oral MTX, beginning at a dosage of 7.5 mg once weekly, and increasing over time by titration against disease activity. All patients also received 5 mg of oral folic acid twice weekly. Simple analgesics and nonsteroidal antiinflammatory drugs (NSAIDs) were allowed, but with no change in the NSAID dosage for 1 week prior to MRI scans. No more corticosteroids were given to the MTX-alone group until after the month-3 MRI scan. After the month-3 scan, patients in both groups received the same standard care, with continuing MTX dosage escalation and IA and IM corticosteroids as routinely indicated. All patients had repeat MRI scans after 12 months of therapy, although the time of the third scan could be moved up if disease control was considered unsatisfactory after attaining the maximum dosage of MTX for at least 6 weeks. No MRI scans were performed within 4 weeks after administration of corticosteroid therapy.

MRI evaluation.

Scans of the second through fifth MCP joints of the dominant hand were performed using a commercially available 1.5T Gyroscan ACS NT whole-body MRI system with image acquisition (Philips, Best, The Netherlands). A linear circular 11-cm surface coil (Philips) was placed on the dorsum of the hand when the patient was prone with an extended arm. All patients had 5 sequences performed: T1-weighted spin-echo (SE) coronal and axial pulse sequences (T1 SE), T2-weighted turbo spin-echo fat-suppressed (T2 TSE FS) coronal pulse sequences, T1 SE post–gadolinium–diethylenetriaminepentaacetic acid (Gd-DTPA) axial sequences, and T1 FS post–Gd-DTPA coronal sequences.

The imaging parameters for the T1-weighted axial images were repetition time (TR) 485 msec, echo time (TE) 20 msec, matrix 256 × 256, field of view (FOV) 10 × 5 cm, slice thickness 1.5 mm, slice gap 0.15 mm, number of signals averaged (NSA) 2, and total acquisition time 201 seconds. The T1 coronal SE parameters were similar, but the FOV was 10 × 10 cm, and the acquisition time was 250 seconds. The T1 coronal FS post–Gd-DTPA parameters were TR 450 msec, TE 20 msec, matrix 256 × 256, FOV 10 × 10 cm, and total acquisition time 524 seconds. The T2 TSE FS acquisition parameters were TR 2,000 msec, TE 100 msec, matrix 256 × 256, FOV 10 × 10 cm, slice thickness 2.0 mm, slice gap 0.2 mm, NSA 4, and acquisition time 224 seconds. The images were printed on radiographic acetate film and stored. Contrast-enhanced and non–contrast-enhanced MRI films were scored by 2 experienced readers (PGC, PO'C) who were blinded to clinical details and radiographic findings.

Synovitis was scored from those areas demonstrating increased enhancement on the T1 axial post–Gd-DTPA scans compared with the corresponding sequence on the pre–Gd-DTPA sequences, in accordance with recent expert recommendations (23). To avoid the recognized problems with Gd-DTPA leakage into the joint fluid, post–Gd-DTPA sequences were performed within 5 minutes of injection (24). The individual slice demonstrating maximum inflammation was chosen and then the maximum enhancing thickness of synovium per MCP joint was measured in millimeters (Figure 1A). Maximum thickness was chosen to represent the “ceiling” measure of synovitis, the measure that would be most responsive to change.

Figure 1.

Axial T1-weighted post-gadolinium sequences of the second through fifth metacarpophalangeal (MCP) joints. A, Before methotrexate and intraarticular corticosteroid (MTX + IAST) therapy. Marked MCP joint synovitis and tenosynovitis are visible, as demonstrated by high signal intensity. Arrow indicates the dimension measured for maximum synovial thickness. Arrowhead shows an erosion that becomes clearly visible in B after suppression of synovitis. B, After MTX + IAST therapy. (A defect in the cortex, fulfilling the definition of an erosion, was visible on the equivalent coronal film.)

Bone erosions were defined as bone defects with sharp margins visible in 2 planes with a cortical break seen in at least 1 plane on T1 sequences (23). Each bone erosion was recorded according to its site within a given joint. Bone edema was defined as an area with ill-defined margins and high signal intensity on FS sequences (Figure 2A) and was only recorded when present without an associated erosion; bone edema was scored as present or absent in the metacarpal head or proximal end of the phalanx for each joint. To reduce artifactual count changes resulting from partial voluming (sectioning) effects, after the initial scoring each bone lesion was accounted for with random-time readings of a patient's films and with the readers still blinded to time sequence and treatment group. Interobserver disagreements were settled by consensus.

Figure 2.

Coronal T1 fat-suppressed sequences of the second through fifth metacarpophalangeal (MCP) joints. A, Before methotrexate and intraarticular corticosteroid (MTX + IAST) therapy. Bone edema is seen as areas of high signal intensity in both the proximal and distal subchondral bones of the third and fourth MCP joints. This edema has characteristic ill-defined margins (arrow). B, After MTX + IAST therapy. The bone edema has disappeared.

Statistical analysis.

Our primary analysis grouped all MCP joints from patients who received IAST at baseline, even though not all the MCP joints had been injected. We thought this was a cautious interpretation, given the possible effects of IAST on adjacent joints and the known difficulty with placement of small joint injections. Parametric and nonparametric tests were applied to baseline demographic and clinical parameters as appropriate. A repeated-measures general linear model approach assuming linearity was used to analyze the changes in MRI synovial thickness. Bone erosion data were analyzed using Fisher's exact test. The repeatability index was calculated to determine whether the “within-patient” changes over time were greater than the measurement error. Area under the curve (AUC) values for all MRI and clinical values were calculated using the trapezium rule. Statistical correlation tests were performed using sum scores for an individual patient (rather than individual joints) and using Spearman's rank correlation. Statistical analysis was performed using SPSS version 8.0 (SPSS, Chicago, IL).


Demographic baseline characteristics and medication use.

Forty-two patients fulfilled entry criteria, underwent baseline MRIs, and were randomized to either MTX alone (n = 20) or MTX + IAST (n = 22). One patient from each group did not have subsequent MRI scans (one due to malignancy and another due to claustrophobia) and were not included. The demographic data for age, sex, disease duration, serology, and radiographic erosions are presented in Table 1, with the baseline disease characteristics presented in Table 2. The only significant difference between the 2 groups was the presence of RF (P = 0.01).

Table 1. Patient characteristics*
VariableMTX groupMTX + IAST group
  • *

    MTX = methotrexate; IAST = intraarticular corticosteroid; RA = rheumatoid arthritis.

No. of patients (no. of women)19 (9)21 (13)
Mean age, years (range)55 (21–83)53 (28–76)
Median disease duration, months (range)6 (2–11)5 (1.5–11)
No. with family history of RA47
No. rheumatoid factor positive135
No. shared epitope positive1514
No. with radiographic erosions108
Table 2. Clinical outcomes*
OutcomeBaseline3 months12 months
  • *

    Except for modified Health Assessment Questionnaire (M-HAQ; see ref. 22) raw scores, which are presented as medians (with ranges), values are the mean ± SEM. MTX = methotrexate; IAST = intraarticular corticosteroid.

Early morning stiffness, minutes   
 MTX101.0 ± 21.064.2 ± 23.538.6 ± 14.2
 MTX + IAST94.0 ± 21.731.4 ± 10.436.4 ± 55.6
Disease activity, visual analog score   
 MTX52.6 ± 6.541.3 ± 6.927.3 ± 5.6
 MTX + IAST46.4 ± 4.430.8 ± 5.331.2 ± 5.7
Tender joint count   
 MTX18.0 ± 3.215.8 ± 3.313.3 ± 3.4
 MTX + IAST26.2 ± 3.117.9 ± 3.618.6 ± 3.8
Swollen joint count   
 MTX11.7 ± 1.67.4 ± 1.45.5 ± 1.4
 MTX + IAST12.2 ± 1.44.9 ± 1.37.6 ± 1.7
 MTX11.0 (0–22)6.5 (0–21)6.5 (0–21)
 MTX + IAST10.0 (0–24)7.0 (0–21)6.0 (0–20)
C-reactive protein, mg/liter   
 MTX35.8 ± 8.218.4 ± 5.015.8 ± 4.9
 MTX + IAST18.5 ± 4.217.9 ± 5.516.7 ± 5.3

As the protocol had dictated, no patient in the MTX-alone group had any corticosteroid before the month-3 scan and no patient in either group received a DMARD other than MTX. The mean (range) weekly dosages of MTX were as follows: 3 months, MTX alone 12.5 mg (10.0–12.5), MTX + IAST 10.0 mg (10.0–15.0); 12 months, MTX alone 17.5 mg (12.5–20.0), MTX + IAST 16.25 mg (10.0–20.0). All patients continued to take MTX. According to protocol, all patients in the MTX + IAST group received IAST at baseline. The mean (range) number of injections per patient was 9 (3–16). The mean (range) number of the second through fifth MCP joints injected of the dominant hand was 1.8 (1–4). In the first phase of the study, the mean (range) total methylprednisolone dose was 181.9 mg (30–300), while in the second phase (3–12 months) the mean (range) total corticosteroid doses were MTX alone 117.8 mg (0–800) and MTX + IAST 141.6 mg (0–540). The clinical and laboratory outcomes related to the imaging time points are presented in Table 2. The imaging findings are presented in Table 3.

Table 3. Imaging outcomes for the second through fifth metacarpophalangeal joints
Outcome*Baseline3 months12 months
  • *

    MTX = methotrexate; IAST = intraarticular corticosteroid; MRI = magnetic resonance imaging.

  • Total numbers of metacarpophalangeal joints imaged for erosions = 76 in the MTX group and 84 in the MTX + IAST group. Bone edema was imaged at 2 sites per joint.

Synovitis, mean ± SEM mm per joint   
 MTX2.5 ± 0.172.7 ± 0.191.9 ± 0.13
 MTX + IAST2.7 ± 0.171.6 ± 0.161.6 ± 0.12
Bone edema sites, total no.   
 MTX + IAST1863
New MRI erosions, no. of joints   
New radiographic erosions, no. of joints   

MRI: synovitis and erosions.

Baseline. Since there was unequal distribution of seropositive patients, with a greater number in the MTX-alone group, the outcome was analyzed according to RF status. There was no difference in baseline synovitis scores between seropositive and seronegative patients. When the presence of the shared epitope was analyzed, there was a trend toward higher synovitis scores in the patients who possessed the shared epitope compared with those who were negative for this epitope (mean ± SEM 2.8 ± 0.2 mm versus 2.2 ± 0.3 mm; P < 0.07).

For MRI synovitis scores, the repeatability index (the smallest detectable difference, twice the SD of the difference in scores) was calculated using 20 scans assessed twice (with a 12-week interval) in a blinded manner. The repeatability index was 1.6 and this was incorporated on a Bland-Altman plot (data not shown), together with the change in synovitis scores seen from baseline to 3 months in the MTX + IAST group. The plot showed that many of the changes seen were greater than the repeatability index, indicating that the observed differences were beyond measurement error and therefore most likely real.

Phase 1. During the first 3 months, there was a significant difference in synovitis reduction between the 2 groups. The synovitis scores (mean ± SEM) per joint for MTX alone were baseline 2.5 ± 0.17 mm, 3 months 2.7 ± 0.19 mm; for MTX + IAST, baseline 2.7 ± 0.17 mm, 3 months 1.6 ± 0.16 mm (difference in change between groups, P < 0.001). During the same period, in the MTX-alone group 7 joints developed new erosions compared with 1 joint in the MTX + IAST group (P < 0.03). Furthermore, for the patient group as a whole, there was a close correlation between the level of synovitis and the likelihood of developing erosions during this period (Figure 3A), with no new bony erosions in the joints without synovitis.

Figure 3.

Percentage of all joints with new erosions versus mean level of synovitis, during phase 1 (A) and phase 2 (B). Level of synovitis is rounded to the nearest whole number. Data are based on individual joint measurements.

Phase 2. For the period between 3 and 12 months, there was no significant difference between the treatment groups in either the synovitis scores (at 12 months, MTX alone 1.9 ± 0.13 mm versus MTX + IAST 1.6 ± 0.12 mm) or the progression of erosion scores (MTX alone 5 versus MTX + IAST 7). Thus, at the end of 12 months the total erosion rate was 12 for MTX versus 8 for MTX + IAST (P not significant). Again, there were no new erosions seen in any joints without synovitis, and the likelihood of bony erosions was directly related to the level of synovitis (Figure 3B).

Correlation analysis. The increase in MRI bone erosion score was significantly correlated with the AUC for MRI synovitis (ρ = 0.420, P < 0.007). No other variables (including baseline synovitis, RF, HLA status, and all clinical parameters) were correlated with increase in bone erosions.

MRI: synovitis and bone edema.

In phase 1, there was a significant reduction in the number of sites with bone edema, similar to the pattern for reduction in synovitis (number of bone edema sites for MTX alone, baseline 10, 3 months 14; for MTX + IAST, baseline 18, 3 months 6; P < 0.01). By 12 months, there was no difference between the groups for number of sites involved (MTX 5 versus MTX + IAST 3).

The synovial thickness in all joints with bone edema was much greater than in those without edema (mean ± SEM 3.6 ± 0.3 versus 2.4 ± 0.1 mm; P < 0.01). The level of synovitis reduction in joints without persistent edema was significantly greater than in those joints with persistent edema (phase 1 1.4 ± 0.5 versus 0.5 ± 0.1 mm, phase 2 1.9 ± 0.4 versus 0.3 ± 0.1; P < 0.001). Bone edema preceded a subsequent MRI erosion in 9 of 22 new erosions (41%).


In the MTX-alone group, only 2 erosions were seen radiographically in the study MCP joints at baseline, with no increase over time. In the MTX + IAST group, only 6 erosions were seen at baseline and again this number did not increase over time. These radiographic erosions correlated by site with MRI erosions. Based on hand and foot radiographs, in the MTX-alone group 10 patients had erosions at baseline and 12 had erosions at 12 months, while in the MTX + IAST group, 8 patients had erosions at baseline and 9 patients had erosions at 12 months.


The results of this study strongly suggest the primacy of synovitis in patients with RA and demonstrate that effective suppression of synovitis prevents bone damage. There was no evidence for a dichotomy between synovitis and bone damage in those patients with early RA. In fact, there was a close parallel between the level of synovitis and the rate of appearance of new erosions in patients independent of whether they received slow- or fast-acting therapies. An important finding was that without synovitis, no erosions occurred and there was a threshold level of synovitis below which there was no association with new bony damage. The findings have major implications for pathogenic concepts and the development of therapies for RA.

In the last decade, the view has emerged that there are 2 independent processes in RA, a phase characterized by synovial inflammation and a phase characterized by autonomous joint destruction (25). These concepts have culminated in the idea that strategies specifically aimed at inhibiting bone damage are necessary in addition to the control of synovitis (26). In contrast, previous MRI studies in RA have demonstrated that persistence of synovitis is associated with ongoing joint damage (5, 27, 28). However, these studies often focused on the wrist (a complex structure with multiple separate joints), used a variety of DMARDs, and included patients with RA of differing durations. We used a sensitive validated imaging system to study individual joints over time, and the patient group was selected to be homogeneous in terms of being newly presenting patients with RA of short disease duration with no prior exposure to DMARDs. Two DMARDs selected for their differing times until onset of action were used. Our data suggest that any proposed dichotomy between synovitis and bone damage has arisen from previous inappropriate analyses of the clinical and radiologic outcome measures used in those studies. Also, the addition of biologic agents to the treatment regimen in patients whose disease has been only partially responsive to MTX is more effective in preventing damage, because additional therapy with biologic agents results in more effective synovitis suppression.

This study, using a novel synovial outcome measure, confirms previous findings obtained by traditional methods (11). Synovitis, as detected by Gd-DTPA–enhanced MRI, has been correlated with microscopic features of inflammation (29). Also, the MRI sequences used fall within the recommendations of the Outcome Measures in Rheumatology Clinical Trials MRI working party for synovitis measurement (23). There is strong face value for the synovial measure used, in that corticosteroids produced a reduction in MRI synovitis within the expected time period, as previously demonstrated (28, 30–32). Similarly, in the second phase, after 12 months of similar DMARD therapy, both groups had comparable synovial thickness scores.

Unfortunately, the MRI sequences used and the current scoring system preclude the measurement of any effects on cartilage. With respect to bone abnormalities, MRI erosions have been shown to be equivalent to radiographic erosions in a validation study (33). The insensitivity of radiography is due to its inability to detect small cortical defects that can be imaged by other modalities (33). They are specific for RA (20), correlate with radiographic erosions, and are predictive of long-term joint destruction (5, 27). Although the difference in MRI erosions between treatment groups at 12 months was not significant, this study was not designed to detect such a change. Care was taken to distinguish these structural lesions from areas of bone marrow edema seen on fat-suppressed MRI sequences only.

This study provides the strongest evidence to date that bone edema is related to the degree of synovitis and is the forerunner of erosions on MRI. Furthermore, effective suppression of synovitis can reverse these pre-erosive changes and subsequent structural damage. Finally, this study demonstrates that synovitis is primary for the pathogenesis of bone disease in early RA and does not support the concept of independent destruction of bone. Future management of RA is likely to focus on controlling synovitis and, by doing so, may obviate the need for separate monitoring of bone damage progression. This may be particularly relevant with regard to expensive therapies not limited by toxicity. Therefore, titration of therapy to optimize the reduction of synovitis may become the preferred method of monitoring disease activity.