Article Text

Download PDFPDF

Concise report
Bone marrow lesions and synovitis on MRI associate with radiographic progression after 2 years in hand osteoarthritis
  1. W Damman1,
  2. R Liu1,
  3. JL Bloem2,
  4. FR Rosendaal3,
  5. M Reijnierse2,
  6. M Kloppenburg1
  1. 1Department of Rheumatology, Leiden University Medical Center (LUMC), Leiden, The Netherlands
  2. 2Department of Radiology, Leiden University Medical Center (LUMC), Leiden, The Netherlands
  3. 3Department of Clinical Epidemiology, Leiden University Medical Center (LUMC), Leiden, The Netherlands
  1. Correspondence to W Damman, Department of Rheumatology, Leiden University Medical Center, C1-R, P.O. Box 9600, Leiden 2300 RC, The Netherlands; w.damman{at}


Objective To study the association of magnetic resonance (MR) features with radiographic progression of hand osteoarthritis over 2 years.

Methods Of 87 primary patients with hand osteoarthritis (82% women, mean age 59 years), baseline distal and proximal interphalangeal joint contrast-enhanced MR images were scored 0–3 for bone marrow lesions (BMLs) and synovitis following the Oslo score. Baseline and 2-year follow-up radiographs were scored following Kellgren-Lawrence (KL) (0–4) and OsteoArthritis Research Society International (OARSI) scoring methods (0–3 osteophytes, joint space narrowing (JSN)). Increase ≥1 defined progression. Associations between MR features and radiographic progression were explored on joint and on patient level, adjusting for age, sex, body mass index, synovitis and BML. Joints in end-stage were excluded.

Results Of 696 analysed joints, 324 had baseline KL=0, 28 KL=4 and after 2 years 78 joints progressed. BML grade 2/3 was associated with KL progression (2/3 vs 0: adjusted risk ratio (RR) (95% CI) 3.3 (2.1 to 5.3)) and with osteophyte or JSN progression, as was synovitis. Summated scores were associated with radiographic progression on patient level (RR crude BML 1.08 (1.01 to 1.2), synovitis 1.09 (1.04 to 1.1), adjusted synovitis 1.08 (1.03 to 1.1)).

Conclusions BMLs, next to synovitis, show, already after 2 years, graded associations with radiographic progression, suggesting that both joint tissues could be important targets for therapy.

  • Magnetic Resonance Imaging
  • Synovitis
  • Inflammation
  • Hand Osteoarthritis
  • Outcomes research

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


The hand osteoarthritis (OA) disease process leads to joint destruction, visualised as radiographic damage.1 With the need to develop effective therapies for hand OA, it is important to understand which processes are involved. By the time radiographic damage is visible, much of the disease process already took place.2 Visualisation of the disease process in an earlier stage will facilitate identification of treatment targets and performance of clinical trials.

From ultrasonography studies in hand OA we know that synovial inflammation plays a role in radiographic progression.3–5 Magnetic resonance (MR) has the advantage that subchondral bone can be visualised,6 where bone marrow lesions (BMLs) are seen as increased water content in the trabecular bone, compatible with possible inflammation or bone fibrosis and remodelling.7 ,8 In knee OA studies, BMLs were associated with structural progression.2 ,9 In hand OA, one MR study (1.0 Tesla (T)) showed that BMLs, next to synovitis, could predict radiographic progression after 5 years.10 However, clinical trials in hand OA measure outcome after 1 or 2 years follow-up, warranting more data on MR imaging.11

As it is unclear whether underlying processes play the same role in onset (incident) and progression of radiographic osteoarthritic damage,2 we studied both together and apart for their association with baseline MR features. Next to the joint level, with summated MR scores we investigated progression on patient level, the level most clinically relevant. This study used for the first time a midterm follow-up of 2 years.


Study design

We used data of Hand OSTeoArthritis in Secondary care, an observational cohort of consecutive patients from our outpatient clinic (a secondary and tertiary referral centre enabling inclusion of patients in all disease stages), who were included after the clinical diagnosis of primary hand OA was made by their treating rheumatologist. The present analysis concerns patients who received contrast-enhanced MR imaging, included March 2011 to October 2012.

Exclusion criteria were: any other pathological condition explaining the hand symptoms, secondary OA and routine MR contraindications. Written informed consent was obtained from all participants. The study was approved by the Leiden University Medical Center medical ethics committee.

For clinical assessment see online supplementary material.


Baseline and 2-year follow-up radiographs of distal interphalangeal joints (DIPJs), proximal interphalangeal joints (PIPJs), interphalangeal joints (IPJs), metacarpophalangeal joints (MCPJs) and first carpometacarpal joints of both hands (30 joints per patient) were scored 0–4 following Kellgren-Lawrence (KL) scoring and 0–3 (IPJs 0–1) for osteophytes and joint space narrowing (JSN) following the OARSI atlas (MCPJs following the PIPJs atlas).12 ,13 Joints with the highest score or with arthroplasty were in end-stage. Reader WD scored paired in known order, blinded for demographic and clinical data. Intraobserver reliability (based on 10% of pairs) was high: cross-sectional intraclass correlation coefficients (ICCs) were 0.89–0.91 and longitudinal percentages exact agreement for progression 92%–96% for the different methods.

Radiographic progression was defined as an increase in score above the smallest detectable change (SDC):14 SDCs on joint level were 0.28–0.39, so ≥1 grade defined progression. For subanalysis, joints were classified as incident OA when they changed from no OA at baseline (KL score 0) to radiographic osteoarthritic damage (KL score 1–4). Joints progressed when they had signs of OA at baseline (KL score ≥1) and increased in score.

Scores of KL (range 0–120), osteophytes or JSN (both 0–86) of all 30 hand joints were summated to study progression on patient level. SDCs were 2.2, 1.4 and 1.8, respectively. Therefore, increase ≥3 grades in KL or ≥2 grades in osteophyte or JSN summated scores defined progression.

MR imaging

MR imaging of the right PIPJs and DIPJs (n=8 joints per patient) was performed at baseline, using an ONI-MSK-Extreme 1.5 T extremity MR imaging scanner (GE, Wisconsin, USA), acquiring coronal and axial T1-weighted precontrast and postcontrast injection and coronal and axial T2-weighted images (protocol in online supplementary material).

MR imaging scoring was performed blinded for demographic and clinical data by RL, using a modified version of the Oslo hand OA MR imaging scoring.15 Cross-sectional intrareader reliability was high: ICC 0.84–1.00 (based on 11 patients).

Synovitis and BMLs were scored 0–3, while effusion, flexor tenosynovitis (PIPJs) or flexor tendon involvement (DIPJs), extensor tendon involvement and cysts were scored present/absent (detailed scoring in online supplementary material).

BML and synovitis scores were summated (range 0–24) for patient level analysis.

Statistical analysis

Risk ratios (RRs) with 95% CIs were estimated to study the association of MR features (determinant) with radiographic progression (outcome) on joint level using generalised estimating equations to account for the patient effect (joints within a patient as within-subject variable), while adjusting for age, sex and body mass index. An exchangeable working correlation matrix, a log link function and the Poisson distribution with robust standard errors (SEs) were used.16 Joints without the MR feature served as reference. BML or synovitis grades 2 and 3 were merged. Joints in radiographic end-stage at baseline were excluded, as they have no potential for progression.

The association between summated scores of MR features (eight joints) and the presence of radiographic progression on patient level (both hands) was studied using the modified Poisson approach for binary data (ie, a Poisson regression model with robust SEs).

Statistical software from SPSS for Windows, V.23.0 (IBM SPSS statistics, New York, USA) was used.


Study population and prevalence of imaging features

Baseline MR imaging was performed in 107 patients, whereof 87 (83%) (82% women, mean age 59 years, follow-up time 2.1 years, online supplementary material) had follow-up available. Reasons for no follow-up: 11 patients stopped, two were excluded, five skipped visit and two radiographs missed. Patients with and without follow-up did not differ (not shown).

At baseline, 28 (4%) joints were in end-stage for KL, 38 (6%) for osteophytes and 42 (6%) for JSN. Progression was seen in 12%, 9% and 10% of joints not in end-stage, respectively (see online supplementary material). At follow-up, one PIPJ had an arthroplasty and 25 patients showed no progression.

BMLs were present in 14.7% (102/693) of joints, while 41.4% (286/691) had synovitis, with missing data in three and five joints, respectively. Effusion, flexor or extensor tendon involvement or cysts were present in 8% (57/693), 3% (20/692), 7% (48/692) and 3% (23/696) of joints, respectively.

MR features and radiographic progression

BMLs grade 2/3 were associated with KL progression (vs 0 RR (95% CI) 3.3 (2.1 to 5.3), figure 1, see online supplementary material), while BML grade 1 was not. Synovitis showed graded associations with KL progression. Similar results were found for associations with osteophyte and JSN progression (see online supplementary material). Adjustment for BMLs decreased the strength of the association between synovitis and progression and vice versa. Neither effusion (present vs absent RR 0.8 (0.3 to 2.0)), nor flexor (1.1 (0.2 to 5.9)), nor extensor tendon involvement (0.9 (0.3 to 2.5)) nor cysts (1.3 (0.5 to 3.3)) were associated with KL progression.

Figure 1

Radiographic progression of a second distal interphalangeal joint. Radiograph at baseline (A) and after 2 years (B) with corresponding MR features (C and D) at baseline. (A) Dorsovolar conventional radiograph at baseline shows discrete joint space narrowing and subchondral cyst formation on the medial side. (B) Dorsovolar conventional radiograph after 2 years shows progression of joint space narrowing, subchondral cyst and osteophyte formation. (C) Axial T1-weighted fast spin echo (FSE) image with frequency-selective fat-suppression (FSFS) post-Gadolinium at baseline shows synovial enhancement (synovitis grade 2) at the dorsal side (arrow). (D) Coronal T2-weighted FSE image with FSFS at baseline, shows high signal in the trabecular bone (bone marrow lesion grade 2) (arrow).

MR features and onset or progression of radiographic osteoarthritic damage

Of joints that increased in KL score, 33 had baseline KL=0 (incident OA), while the other 45 joints had baseline KL≥1 (prevalent OA). Both BML and synovitis were associated with onset and progression and these associations were similar in strength (table 1).

Table 1

Baseline MR imaging features associated with radiographic progression in the same joint* after 2 years of follow-up in 87 patients with hand osteoarthritis in the HOSTAS cohort, stratified to the presence radiographic osteoarthritic damage at baseline

Summated MR features and progression on patient level

Median (range) summated BML score was 1 (0; 10) and synovitis score was 4 (0; 13). Both BML and synovitis summated scores were crudely associated with progression. However, after adjustment, only the associations for synovitis remained statistically significant (table 2).

Table 2

Associations between summated scores of MR imaging features and progression of radiographic osteoarthritis on patient level in 87 hand osteoarthritis patients*


MR imaging-defined BMLs, like synovitis, showed dose–response associations with radiographic progression in hand OA already after 2 years, confirming earlier studies on ultrasound-detected synovitis,3 but indicating that BMLs in hand OA, like in knee OA,2 ,9 are an important additional factor in the disease process. Also, because the presence of BMLs decreases the strength of the association between synovitis and progression, and vice versa.

A strength of our study is inclusion of patients in all disease stages from early to severe. Other cohorts have more severely affected hand OA patients with more joints in end-stage at baseline;3 ,10 these joints have no potential for onset of OA or progression and are thus excluded from the analysis.

Another strength is the distinction in onset and progression of radiographic damage in individual hand joints. Of note, this distinction resulted in few joints in some groups, and therefore, results should be interpreted with caution. We used a cut-off at doubtful to definite OA (KL 1), since lesions can already be present at KL=1. Like in knees, where KL=1 at baseline was a strong predictor for progression and considered as early OA.17 We showed that both BML and synovitis were associated with onset and progression and that these associations were similar in strength. This is in line with results for ultrasound-detected synovitis,4 but was not described before in MR-detected BMLs and synovitis.

Novel is our approach to investigate progression on patient level, which is most relevant from a clinical perspective. Summated BML or synovitis score showed crude associations with progression, although only for synovitis this remained statistically significant after adjustment. This means that the more severe the inflammatory state is, the higher the risk of progression in both hands. We hypothesise that inflammatory MR imaging features could be modified by anti-inflammatory medication like steroids. Future proof-of-concept randomised controlled trials could explore this hypothesis.

This is the first study using 1.5 T MR scanner in hands, enabling more precise identification of lesions with a higher signal-to-noise ratio compared with 1.0 T. Consequences are indicated by our results: we found an association between JSN progression and synovitis grade 1, where another hand OA MR study using 1.0 T did not.10

Our study also had some limitations and restrictions in interpretation of results. First, we did not have information whether MR imaging features are persistent or fluctuating. Especially persistent or progressing lesions have shown to be associated with progression and onset of OA.3 ,18 ,19 Nevertheless, we already found the association with only one time measurement. Another limitation is the number of patients in our study. However, the circumstance that in every patient eight joints can be studied provided enough power to study associations with progression.

Our study indicates that all joint tissues, including BMLs, are important in the disease course of hand OA and it illustrates the use of MR imaging, visualising BMLs, in detecting early OA and detection of joints and patients prone to progress. Future studies should focus on the persistent or fluctuant nature of BMLs in hands and on hand MR imaging in the short term.


We thank the patients of the HOSTAS cohort for participation in this study, research nurses, Mrs B van Schie-Geyer and Mrs A Wongsodihardjo and data manager, Mr C Kromme, in the rheumatology department of the LUMC and technicians in the radiology department of the LUMC for support in data collection.



  • Handling editor Tore K Kvien

  • Contributors WD, FR and MK designed the study. WD and RL included patients in the HOSTAS cohort. MK supervised the HOSTAS cohort. WD scored the radiographs. RL performed the MR imaging and scored the MR images. JLB and MR supervised the MR imaging protocol and scoring. WD, FR and MK were involved in data analysis. WD drafted the manuscript. All authors reviewed the manuscript and approved the final version.

  • Funding Dutch Arthritis Foundation (Reumafonds): 10-1-405.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval The study was approved by the Leiden University Medical Center medical ethics committee.

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