Article Text

Extended report
Associations between MRI-defined synovitis, bone marrow lesions and structural features and measures of pain and physical function in hand osteoarthritis
  1. Ida Kristin Haugen1,
  2. Pernille Bøyesen1,
  3. Barbara Slatkowsky-Christensen1,
  4. Sølve Sesseng2,
  5. Désirée van der Heijde1,3,
  6. Tore K Kvien1
  1. 1Department of Rheumatology, Diakonhjemmet Hospital, Oslo, Norway
  2. 2Department of Radiology, Diakonhjemmet Hospital, Oslo, Norway
  3. 3Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands
  1. Correspondence to Ida Kristin Haugen, Department of Rheumatology, Diakonhjemmet Hospital, PO Box 23, Vinderen, N-0319 Oslo, Norway; haugen_ida{at}


Objectives To explore associations between MRI features and measures of pain and physical function in hand osteoarthritis (OA).

Methods Eighty-five patients (77 women) with mean (SD) age of 68.8 (5.6) years underwent contrast-enhanced MRI of the interphalangeal joints (dominant hand) and clinical joint assessment. One investigator read the MRIs for presence/severity of osteophytes, joint space narrowing, erosions, bone attrition, cysts, malalignment, synovitis, flexor tenosynovitis, bone marrow lesions (BMLs) and ligament discontinuity according to the proposed Oslo hand OA MRI score. Pain and physical function were assessed by joint palpation (tenderness yes/no), self-reported questionnaires (Australian/Canadian (AUSCAN) hand index, Functional Index of hand osteoarthritis (FIHOA), Arthritis Impact Measurement Scale-2 (AIMS-2) hand/finger) and grip strength. Logistic regression with generalised estimating equations was used to explore associations between the presence of MRI features and joint tenderness, and linear regression for associations between the burden of MRI abnormalities and patient-reported outcomes and grip strength (adjusted for age and sex). MRI features with p<0.25 were introduced into a multivariate model. The final model included features with p≤0.10 (backward selection).

Results MRI-defined moderate/severe synovitis (OR=2.4; p<0.001), BMLs (OR=1.5; p=0.06), erosions (OR=1.4; p=0.05), attrition (OR=2.5; p<0.001) and osteophytes (OR=1.4; p=0.10) were associated with joint tenderness independently of each other (final model adjusted for age and sex). The sum score of MRI-defined attrition was associated with FIHOA (B=0.58; p=0.005), while the sum score of osteophytes was associated with grip strength (B=−0.39; p<0.001). No significant associations were found with AUSCAN pain/physical function or AIMS-2 hand/finger subscales.

Conclusion MRI-defined synovitis, BMLs, erosions and attrition were associated with joint tenderness. Synovitis and BMLs may be targets for therapeutic interventions in hand OA.

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.


Osteoarthritis (OA) is the most common musculoskeletal disorder,1 and the hands are often affected.2 Patients have pain and physical limitations, which may have great impact on health-related quality of life.3 Symptom-modifying treatment is mainly following general guidelines for pain management, and disease-modifying drugs are not yet available.4 However, more knowledge about the source of pain in hand OA may lead to better targeting of the treatment.

Hand OA is traditionally assessed by conventional radiography (CR). However, the disease is now recognised to involve the whole joint,5 and the discrepancy between radiographic severity and patient-reported outcomes may indicate that other structures than those seen by CR are important factors in pain and physical limitations.6 Indeed, nociceptive fibres are present in several joint components that are affected by the OA processes, including the capsule, ligaments, synovium, subchondral bone and periosteum.7

Modern imaging techniques such as ultrasonography and magnetic resonance imaging (MRI) provide information about inflammatory changes, which may represent potential therapeutic targets.8 MRI has also the benefit of providing a multiplanar image of the whole joint, and is the only imaging modality that can demonstrate bone marrow lesions (BMLs). Previous studies in knee OA have shown that both synovitis and BMLs are associated with pain.9 However, no studies have so far examined the associations between MRI features and tenderness/pain and physical limitations in hand OA.

Therefore, our primary aim was to explore whether MRI features of the proposed Oslo hand OA MRI score were associated with tenderness in the individual joints.10 Second, we explored the associations between the amount of MRI abnormalities and patient-reported hand pain, physical disability and grip strength.

Patients and methods


The Oslo hand OA cohort was established in 2001 and included 209 patients (190 women) who underwent extensive examination in 2001–3.3 Of those patients, 128 (117 women) underwent a follow-up examination in 2008–9, and 107 (98 women) had MRI scans with T1-weighted (T1w) and short T1 inversion recovery (STIR) sequences of the dominant hand. In this study, we included 85 patients (77 women) with clinical finger joint examination and MRI performed on the same day (n=66) or within 22 days after the clinical examination (n=19). The excluded participants had MRI performed 1–3 months (n=10) or 4–13 months (n=12) after the clinical examination. Of the 85 patients, all except seven had post-gadolinium (Gd) T1w images.

The data collection was approved by the regional ethics committee and the data inspectorate. All patients signed informed consent.

Hand MRI

MRI was performed as part of the follow-up examination of the Oslo hand OA cohort. The 2nd–5th distal interphalangeal (DIP) and proximal interphalangeal (PIP) joints (n=680 joints) of the dominant hand (right: n=83, left: n=2) were examined using an extremity 1.0T MRI unit (ONI, GE Healthcare, Waukesha, Wisconsin, USA), as detailed previously.10 In short, coronal, sagittal and axial T1w fat-suppressed (fs) pre-and postintravenous Gd (0.1 mmol Gd/kg body weight; Magnevist, Bayer Schering Pharma AG, Leverkusen, Germany) images were acquired by a 3D dual-echo Dixon technique11 (repetition time (TR) 20 ms, echo time (TE) 5 ms, 1 mm slice thickness) for assessment of structural features and synovitis/flexor tenosynovitis (Gd enhancement). Coronal and axial STIR images (TR 2850 and 3150 ms, TE 16.3 and 21 ms, 2 and 3 mm slice thickness, respectively) were acquired for assessment of BMLs. Total acquisition time was approximately 30 min.

Two readers (IKH, PB) in consensus scored the 10 first MRIs, while one single reader (IKH) scored the remaining MRIs according to the proposed Oslo hand OA MRI score10: osteophytes (grade 0–3; distal/proximal part of the joint), joint space narrowing (JSN) (grade 0–3), cysts (absence/presence; distal/proximal), malalignment (absence/presence; frontal/sagittal plane), synovitis (grade 0–3), flexor tenosynovitis (grade 0–3), BMLs (grade 0–3; distal/proximal), collateral ligament absence/discontinuity (absence/presence; radial/ulnar), and BMLs at collateral ligament insertions (absence/presence; distal/proximal and ulnar/radial). Erosions (grade 0–3; distal/proximal) and bone attrition (absence/presence; distal/proximal) were scored separately, in contrast to the proposed combined definition and grading. Erosions were defined as more focal areas (often marginal) with bone loss, while attrition was seen as flattening/depression of the joint plate (often ‘sea-gull wing’ configuration). Cysts had no cortical break as opposed to erosions. IKH re-scored 10 randomly selected MRI scans after about 7 weeks, and the intra-reader reliability assessed by κ and intraclass correlation coefficients (two-way mixed effect model) was good/very good for all features except flexor tenosynovitis, which had fair/moderate reliability (online supplementary table S1).12 The scoring of MRI scans was performed blinded for clinical and radiographic status scores.

Hand radiographs

The patients underwent posteroanterior CR of both hands, and one investigator (IKH) scored the bilateral 2nd–5th DIP, 2nd–5th PIP, thumb interphalangeal, 1st–5th metacarpophalangeal and 1st carpometacarpal joints according to the Kellgren–Lawrence (KL) scale (grade 0–4).13 Presence of radiographic erosions was scored according to the OARSI atlas.14

Intra-reader and inter-reader reliability were good to very good for KL (weighted κ=0.86 and 0.79, respectively) and erosions (κ=0.88 and 0.81, respectively) as previously reported.12 15

Assessment of joint tenderness, hand pain, physical function and grip strength

All patients underwent clinical examination of 30 finger joints. One rheumatologist (BS-C) examined the bilateral 2nd–5th DIP, 2nd–5th PIP, thumb interphalangeal, 1st–5th metacarpophalangeal and thumb base joints for presence of tenderness upon palpation.16

The patients also completed the Australian/Canadian (AUSCAN) pain (n=83) and physical function subscales (n=85), Functional Index of hand osteoarthritis (FIHOA) (n=82), and Arthritis Impact Measurement Scale-2 (AIMS-2) hand/finger subscale (n=85) within 1 week before the clinical examination.17,,19 The AUSCAN hand index includes five items about pain and nine items about physical disabilities during the last 48 h (total range 0–20 and 0–36, respectively).17 20 FIHOA includes 10 questions about physical disabilities (total range 0–30).19 21 The questionnaire is originally investigator-based, but was in this study completed by the patients. The AIMS-2 hand/finger subscale includes five questions about physical disabilities (total range 0–20).18 22 High scores represent high levels of pain or physical disabilities for all questionnaires.

Grip strength was measured by JAMAR hydraulic hand dynamometer with the patients sitting with the shoulder in a neutral position and 90 degrees flexed elbow.23 The best performance out of two attempts was recorded for each hand. In this study, we used measurements of the hand that had been imaged by MRI.


For joints that were scored for MRI pathology in both the proximal and distal part of the joint, we used the highest score for that joint (ie, a joint with osteophyte grade 3 distally and grade 1 proximally was interpreted as grade 3). We dichotomised the collateral ligament variables (ie, no pathology vs ≥1 absent ligament(s) or ≥1 BML(s) at the insertions). Flexor tenosynovitis grade 3 was only present in one joint, and we therefore collapsed grades 2 and 3.

A limited field of view caused a few missing values (ie, ≤7 of 680 missing for structural features, 11 of 624 missing for synovitis and 35 of 624 missing for flexor tenosynovitis). We calculated the numbers of affected joints and sum scores for each MRI feature by calculating the mean values for all joints and multiplying by eight (in order to correct for missing values).

We examined the associations between MRI features and tenderness in individual joints, while the associations with health status scores and grip strength were analysed at patient level.

For each MRI feature, we calculated the prevalence of tenderness upon palpation in joints with different levels of MRI pathology. Before the regression analyses, we dichotomised all MRI variables that were scored on 0–3 scales in order to limit the number of independent variables in the regression models. We used grade ≥1 as the cut-off point for osteophytes, JSN, erosions and BMLs, and grade ≥2 for synovitis due to similar prevalence of tenderness in joints with synovitis grades 0 and 1. We calculated the OR for tenderness in joints in which MRI pathology was present (joints without this MRI feature were used as reference) by generalised estimating equations in order to adjust for within-subject dependency using an unstructured covariance matrix. All analyses were adjusted for age and sex (ie, corrected univariate analyses). Those MRI features that were associated with tenderness (p<0.25) in the corrected univariate models were included in a multivariate model in order to assess the independent associations for each of the feature (adjusted for age and sex). One by one feature was excluded by backward selection. The final model was tested in the DIP and PIP joints separately.

We also tested whether synovitis and BMLs were associated with joint tenderness with additional adjustment for radiographic severity as assessed by KL. KL grades (KLG) 3 and 4 were collapsed in order to reduce the number of variables in the model.

We examined the associations between the amount of each MRI feature (ie, sum scores/number of affected joints) as independent variables and AUSCAN pain, AUSCAN physical function, FIHOA, AIMS-2 hand/finger and grip strength as dependent variables in separate models using linear regression adjusted for age and sex. The associations were reported as unstandardised B coefficients. Those MRI features that were associated with tenderness (p<0.25) in the corrected univariate models were included in a multivariate model (adjusted for age and sex). One by one feature was excluded by backward selection. The final models included features with p≤0.10. Owing to heteroscedasticity in the linear models, we also used logistic regression in order to confirm the validity of the linear regression models. We examined the associations between the MRI features as independent variables and symptoms above the ‘patient-acceptable symptom severity’ state for AUSCAN pain (>8.2) and AUSCAN physical function (>16.1) as binary dependent variables.24

The analyses were performed with SPSS version 17.0.


Table 1 shows demographic and clinical characteristics of the 85 patients. Tender finger joints were frequently present, and the median (interquartile range; IQR) number of tender joints was higher in DIP (2, 1–4) than in PIP joints (1, 0–2). Most patients had high burden of radiographic hand OA, and radiographic erosive disease was frequently present. Two patients had no joints with definite radiographic hand OA defined as KLG≥2.13 However, these two patients had one or more joint(s) with KLG=1 and fulfilled the clinical American College of Rheumatology criteria for hand OA.25

Table 1

Demographic and clinical characteristics for the 85 patients in the Oslo hand OA cohort

Structural MRI features such as osteophytes, JSN, erosions and attrition, and inflammatory features such as mild synovitis were frequently present (table 2). Moderate to severe synovitis (grade 2–3) was less common with a median (IQR) of 2 (1–3).

Table 2

The median (interquartile range) number of distal and proximal interphalangeal joints of the dominant hand with MRI pathology and MRI sum scores for the eight joints

Association between MRI features and joint tenderness in individual joints

The prevalence of joint tenderness increased with higher scores for most MRI features (table 3). Joints with mild synovitis had similar prevalence of tenderness as joints with no synovitis.

Table 3

The prevalence of joint tenderness across different scores of MRI pathology

We found significantly higher odds of tenderness in joints with all types of MRI pathology compared with joints without pathology, except for cysts (table 4). All features were included in a multivariate regression model using backward selection, and the final model contained osteophytes, erosions, bone attrition, synovitis and BMLs (table 4). Multivariate analyses with forward selection provided the same results (data not shown). We tested the final model in the DIP and PIP joints separately. Erosions were more strongly associated with tenderness in the DIP joints, while synovitis, attrition and osteophytes were more strongly associated with tenderness in the PIP joints (online supplementary table S2). Osteophytes and JSN appeared frequently, and we therefore used osteophytes ≥ grade2 and JSN ≥ grade 2 as cut-off points in confirmatory analyses with similar results (data not shown).

Table 4

Odds ratios for tenderness in joints with MRI pathology versus joints without the MRI feature (adjusted for age and sex)*

Synovitis (OR=2.4, 95% CI 1.5 to 3.6) and BMLs (OR=1.6, 95% CI 1.0 to 2.5) were also significantly associated with joint tenderness in multivariate analyses with both variables in the same model adjusted for radiographic severity as well as age and sex.

Association between MRI features and measures of pain and physical disabilities in individual subjects

MRI sum scores were not significantly associated with AUSCAN pain, AUSCAN physical function or AIMS-2 hand/finger subscales in univariate analyses corrected for age and sex (table 5). Several structural features were significantly associated with FIHOA and grip strength, and synovitis was associated with FIHOA and grip strength (the latter only borderline significant) (table 5). However, bone attrition and osteophytes were the only MRI features in the final multivariate models (backward selection) for FIHOA and grip strength, respectively.

Table 5

Associations between sum scores of MRI features and total scores of AUSCAN pain subscale, AUSCAN physical function subscale, FIHOA, AIMS-2 hand/finger subscale and grip strength (adjusted for age and sex)

Logistic regression analyses with AUSCAN pain and physical function above the patient-acceptable symptom severity state showed no significant associations confirming the validity of the linear regression models (online supplementary table S3).

Corrected univariate analyses with the number of affected joints as independent variables provided similar results as the analyses with MRI sum scores, and synovitis was associated with both FIHOA and grip strength (online supplementary table S4). However, synovitis did not remain in the final multivariate models (backward selection). The final model for FIHOA included osteophytes (B=0.88, 95% CI to 0.06,1.83) and bone attrition (B=0.63, 95% CI −0.01 to 1.27), while the final model for grip strength included JSN (B=−0.64, 95% CI −1.28 to 0.00) and erosions (B=−0.74, 95% CI −1.51 to 0.02)


This study shows for the first time that MRI-detected synovitis and BMLs are associated with joint tenderness independently of each other, other MRI features and also independently of radiographic structural abnormalities in patients with hand OA. Bone attrition and erosions were the only structural MRI features that were significantly associated with joint tenderness.

In line with our results in the hand, abnormalities in the subchondral bone and synovium seem to be the features that are most consistently associated with pain in knee OA.26,,35 A recent systematic review concluded that knee pain in OA was associated with BMLs and effusion/synovitis with moderate evidence, while the level of evidence for other MRI features was limited or conflicting.9 However, there are conflicting results especially for BMLs,36,,38 and several studies are limited by inadequate adjustment for demographic factors, other MRI features and/or radiographic severity, which may represent confounding factors.

Our results are also in concordance with previous studies showing an association between US-detected synovitis and pain in patients with hand OA.39 40 However, the analyses in these studies were not adjusted for structural features. Several features of OA are often co-occurring and may also be associated to each other. Hence, the relations between MRI features and pain are complex and involve both direct and indirect effects. In order to limit the effect of confounding, we performed multivariate analyses with adjustment for other MRI features and also radiographic severity in a second model. We found significant associations between synovitis and joint tenderness which were independent of other MRI features and radiographic severity, which strengthens the hypothesis of inflammation as a potential therapeutic target in hand OA. A corticosteroid drug, CrX-102, has been shown to improve health status and clinical joint findings in hand OA.8 41 Keen et al performed a 4-week observational study on parenteral corticosteroids in hand OA, but found no statistically significant reduction in US-detected synovitis.42 However, whether MRI synovitis is more sensitive for change and corresponds to clinical improvement needs to be adressed in future studies.

Few studies have explored the role of BMLs in hand OA. Tan et al described BMLs at the entheseal sites with use of high-resolution MRI.43 The high prevalence of BMLs in hand OA was confirmed by Wittoek et al, especially in erosive DIP joints.44 However, this study is the first to show an association between BMLs and tenderness in hand OA (although of borderline significance in the multivariate model). Increased trabecular bone pressure, ischaemia and inflammation in the BMLs are possible stimuli for pain,,45 46 but the pathophysiology behind the observed association between BMLs and pain is still elusive.47

A strong association with joint tenderness was found for bone attrition, but also erosions were significantly associated with tenderness independently of other MRI features. A previous study on radiographic erosive hand OA showed that erosive OA was associated with higher levels of pain and physical disabilities compared with non-erosive OA. However, the association seemed to be confounded by higher burden of disease in the patients with erosive disease.48 Hernández-Molina et al found an association between bone attrition of the knee and pain, but in contrast to our results, the association was lost in OA knees where other pathological features that may cause pain also coexisted.49

The association between erosions and joint tenderness was strongest in the DIP joints, while the associations for synovitis, attrition and osteophytes were strongest in the PIP joints (online supplementary table S2). These observed differences may be due to a higher prevalence of joint tenderness in the DIP joints (also in joints without MRI pathology), making it more difficult to detect significant associations.

In support of previous US studies,39 we found significant associations between MRI features and tenderness in individual joints, while the associations between the level of MRI abnormalities and patient-reported measures and grip strength were much weaker. First, person-related psychosocial factors influence the report of pain and physical disabilities,47 while this influence is smaller on the analyses in individual joints. Second, we had information about MRI pathology in the interphalangeal joints of the dominant hand only. Hence, MRI pathology in the metacarpophalangeal and thumb base joints of the dominant hand may also contribute to the level of symptoms experienced and the observed grip strength. Affection of the non-dominant hand may also influence the level of self-reported symptoms. Third, non-unidimensionality and overlapping questions within the subscales may make us less likely to detect significant associations.50

There are several limitations to this study. First, the study was cross-sectional, and the association between MRI features and joint tenderness should also be explored in longitudinal studies to explore a causal relationship. Second, our patients were recruited from an outpatient rheumatology clinic and the majority of patients had advanced radiographic disease. Hence, it remains uncertain whether our findings can be extrapolated to the general population of patients with hand OA.

In conclusion, we found that synovitis and BMLs were associated with tenderness in the same joint in patients with hand OA, independently of each other and also other MRI features (BMLs only borderline significant) and radiographic severity. Also bone damage, reflected as attrition and erosions, was significantly associated with joint tenderness. The link between MRI findings and clinical examination also supports the validity of MRI as an imaging tool in hand OA. However, we were unable to demonstrate any convincing association between the amount of MRI pathology and patient-reported measures of hand pain, physical disabilities and grip strength. Longitudinal studies are needed to assess whether changes in MRI pathology are associated with changes in pain in hand OA.


The authors thank the patients of the Oslo hand OA cohort for participating in this study, study nurses and technicians in the department of radiology for helping us to collect the data, Jessica Bijsterbosch for reading hand x-rays for calculation of inter-reader reliability and Inge C Olsen for help with the statistical analyses.



  • Handling editor Dr. Johannes WJ Bijlsma

  • Funding The study is supported by grants from the South-Eastern Norway Regional Health Authority.

  • Competing interests None.

  • Ethics approval Regional ethics committee and the data inspectorate.

  • Provenance and peer review Not commissioned; externally peer reviewed