CR provides a two-dimensional picture of bony changes, such as osteophytes, erosions, cysts, and sclerosis, and joint space narrowing (JSN) as an indirect measure of cartilage loss (Figure 1).

Osteophytes can be divided into 'true' intra-articular osteophytes and traction spurs. The 'true' intra-articular osteophytes are located at the joint margins 4 and can easily be seen on CR with a traditional posteroanterior view. The traction spurs, on the other hand, are located at the insertion of the extensor tendon or along the midshaft and are most easily seen on CR with an oblique or lateral view. Whether these enthesophytic changes are related to OA is not entirely clear, and previous studies have suggested that these changes are related mainly to age and local biomechanical factors and not a systemic enthesopathy 56.

Radiographic measurement of JSN is currently recommended as the imaging endpoint for clinical trials of disease-modifying OA drugs 7. The cartilage cannot be directly assessed by CR and therefore is indirectly judged by the inter-bone distance. The evaluation may be affected by positioning of the hands (for example, flexion deformities) and is further complicated by erosive development in the finger joints, which may lead to increased joint space width (JSW) (that is, pseudo-widening) despite worsening of disease.

Radiographic erosions in hand OA are seen as bone damage in the central part of the joints with a typical seagull wing configuration. They typically occur in the DIP and PIP joints 1 but have also been described in the thumb base joints 8. Longitudinal studies have shown that JSN precedes the erosive development, suggesting that local biomechanical factors are important for the erosive development 910. These findings may suggest that erosive hand OA represents severe hand OA rather than a separate disease entity.

Cysts are seen as loss of the trabecular structure, whereas sclerosis is seen as increased density on CR. Both features may be related to bone remodeling. Histological studies have shown that areas with sclerosis are characterized by increased thickness of the subchondral plate and the trabeculae, and this may indicate repair of bone trauma 11.

Here, we will briefly present the most widely used scoring systems for assessment of radiographic hand OA. Currently, there is no consensus on the preferred scale. The first proposed radiographic scoring system was the Kellgren and Lawrence (K&L) scale 12, which is still the most widely used 3. The K&L scale is a global scale that grades OA on a range from 0 to 4 points (a grade of at least 2 represents definite OA) on the basis of the presence/severity of osteophytes/ossicles, JSN, sclerosis, pseudocystic areas, and altered shape of bone ends. Different descriptions of the grades for various joint groups and across publications have caused confusion in how to interpret the different grades 1314. Further-more, the K&L scale has been criticized for too much emphasis on osteophytes 15 as narrowed/sclerotic joints cannot be classified as having OA unless osteophytes are present. Thus, several studies have used modified K&L scales to overcome these deficiencies.

Assessment of individual features instead of a global score may also optimize agreement, and the Osteoarthritis Research Society International (OARSI) atlas is most frequently used 1617. With this atlas as a reference, the presence and severity of individual features are assessed on semi-quantitative scales. However, scoring of individual features may be more time-consuming, and we have no agreed-upon definition of hand OA on the basis of individual radiographic features.

Verbruggen and colleagues 9 suggested an alternative approach and developed a numerical scoring system for the anatomic evolution of erosive and non-erosive hand OA. This system is based on an underlying assumption of hand OA as a disease that undergoes predictable phases. The scale is most suitable for the evaluation of erosive hand OA as the progression of JSN and osteophytes does not necessarily mean progression in terms of the proposed phases.

Reliability exercises have shown that hand OA can be assessed reliably by CR. However, despite the limitations of the K&L scale, studies have not been able to show better reliability with other scoring systems 1819. Hence, based on the different character of the scales, the choice of scale rather depends on the study objective.

Radiographic hand OA is usually a slowly progressive disease 12021. CR is most likely not a sensitive measure of hand OA progression or, as previously shown, of knee OA progression 22. Newly proposed computerized methods for quantification of JSW may provide a more sensitive set of tools for the assessment of OA progression over time 232425, but longitudinal studies are needed.

Studies have suggested a positive association between radiographic hand OA and hand pain, but the strength of the association varies across studies 26. The associations between radiographic hand OA and hand disability are more inconsequential, ranging from no to moderate associations 26. Patients with erosive hand OA experience more pain and physical disabilities compared with patients with non-erosive hand OA, but the difference seems to be due at least partly to a higher burden of disease in the erosive patients. Kortekaas and colleagues 27 just recently showed that hand osteophytes and JSN were associated with tenderness in the same joint independently of each other and synovitis. However, cartilage is aneural and cannot be a direct source of pain, and this means that the association is possibly mediated by bone damage (as a consequence of decreased bone unloading associated with cartilage loss).

Despite positive associations in cross-sectional studies, longitudinal studies have not been able to show any association between radiographic progression and clinical deterioration 192021. In general, studies focusing on the amount of structural features or the number of affected joints and the relation to hand pain and functioning are less likely show associations with measures of pain and functioning compared with analyses performed at the individual joint level. First of all, psychosocial factors affect the self-report of symptoms 28. Furthermore, radiographic studies are limited by the fact that pain in hand OA is related to not only structural abnormalities but also pain perception and inflammation.

Standardization of the definition of radiographic hand OA with respect to scoring methods, joints under evaluation, and the required number of affected joints could possibly reduce the variations across studies. Although the K&L scale has the benefit of being simple and is well known in the research community, the system has several limitations. However, comparative studies have not been able to conclude about the preferred instrument. Quantitative measurement of JSW may be a more sensitive measure of progression in hand OA, but the sensitivity to change must be explored in future longitudinal studies.

US allows visualization of a wide spectrum of hand OA features, including osteophytes, marginal erosions, and synovitis (Figure 2). US may, therefore, be a feasible tool for visualization of inflammation in patients with hand OA. One of the disadvantages of US is the inability of its beam to penetrate bony cortex. Thus, owing to the joint anatomy, visualization of the articular cartilage and bone damage is restricted mostly to the peripheral parts 29. Overlying osteophytes, which disturb the acoustic window, further complicate the evaluation. In severely damaged joints, it might be difficult to determine the point at which an erosion starts and an osteophyte ends.

Most US studies of patients with hand OA have reported a high prevalence of gray-scale synovitis 30313233, whereas power Doppler activity has been less frequent 303233. Kortekaas and colleagues 33 found that both gray-scale synovitis and power Doppler activity were present in the majority of patients with hand OA, but the number of joints with power Doppler activity was considerably lower than the number of joints with gray-scale synovitis. However, other studies have demonstrated greater similarity in the frequencies of power Doppler activity and gray-scale synovitis 3134. These variations across studies may be due to differences in study populations or US techniques.

Erosive OA is often called 'inflammatory' OA. Mancarella and colleagues 34 found a higher amount of power Doppler activity, synovial hypertrophy, and joint effusion in patients with radiographic erosive OA joints in comparison with patients with radiographic non-erosive OA joints. Synovitis seems to be most prevalent in joints with active erosions, whereas the prevalence is lower in joints that are remodeled 35. These findings may indicate a burnout of inflammation in late stages, but this hypothesis needs to be confirmed in longitudinal studies.

One preliminary US scoring system has been developed for hand OA. A group of experts in the fields of OA, US, and outcome measures agreed upon a scoring system for hand OA features, including assessment of synovitis (gray-scale hypertrophy/effusion and power Doppler) and osteophytes on semi-quantitative scales 36. Erosions, cartilage assessment, and JSN were not included in the scoring system, because of concerns about reliable definitions, acquisition, currently available US technology, and feasibility related to duration of scanning.

Dependency on the US operator represents one of the major limitations of US examination. Several studies report inter-reader reliability based on stored images, and this limits the variability related to the performance and technique of the US examination. A large reliability exercise was arranged in order to test the reliability of the preliminary US scoring system 36. Despite divergent results, the authors concluded that the results were satisfactory and that the proposed scoring system could provide a good basis for further development of a US outcome tool.

The validity of US against other imaging modalities (especially CR) has now been tested in several studies. The first report comparing US and CR found that CR was more sensitive than US in detection of erosions and indicated that the interposition of osteophytes could limit the acoustic window of US 37. However, later reports have shown that US is most sensitive in the detection of erosions 3138 as well as osteophytes and JSN 313839. The higher sensitivity is probably due to the multiplanar visualization of the joint by US. Erosions were found not only in the central part of the joints but also in the peripheral sections 31. Furthermore, some radiographic cysts appeared to be erosions as demonstrated by US 31. However, it must be noted that erosions may be difficult to assess in joints with severe OA with excessive bone formation and deformities, which limit the acoustic window.

Estimation of JSN by US may be problematic as only the peripheral inter-bone distance can be documented and overlying osteophytes may further decrease the acoustic window 39. However, in two studies, the investigators measured the cartilage thickness quantitatively and found significant association between lower US-defined cartilage thickness and radiographic severity, JSN (semi-quantitative scale), and JSW (quantitative scale) 3440.

So far, few studies have compared the findings by US and MRI, but the current results support the use of US as a valid instrument. Wittoek and colleagues 38 compared US against MRI as reference and found good agreement for both structural features and inflammation. Iagnocco and colleagues 41 used US and MRI fusion imaging and found that that hyperechoic prominences seen by US corresponded to osteophytes seen by MRI. However, the optimum application of this technique has not yet been determined, and the extra cost and time limit its use.

Several studies have reported that US pathological features such as gray-scale synovitis, power Doppler signal, and osteophytes are significantly associated with pain at the individual joint level 273033. Kortekaas and colleagues 33 showed that gray-scale synovitis was significantly associated with pain. However, the analyses were not adjusted for structural features.

As discussed in the section on radiographic features and measures of pain and physical function (see previous 'Associations with pain and physical function' section), studies are less likely to show significant associations when the analyses are performed at the patient level instead of the individual joint level. Analyses of the association between US features (number of affected joints or sum scores) and measures of global hand pain, stiffness, and physical disability have revealed conflicting results. In line with analyses at the individual joint level, Kortekaas and colleagues 33 found significant associations between gray-scale synovitis and pain, stiffness, and physical disability. However, other studies have found no significant associations 3042. Koutroumpas and colleagues 42 found that clinically inflamed joints, but not US inflammation, were significantly correlated to physical limitations. These findings may indicate that US detects low-grade inflammation, which possibly does not reflect OA pathology.

The association between US features and pain should also be explored in longitudinal studies, but we are aware of only one observational study of patients with hand OA treated with intra-muscular metylprednisolone: Keen and colleagues 32 found a significant reduction in symptoms but no statistically significant reduction in US-detected synovial inflammation after 4 and 12 weeks. Interestingly, there was no association between reduction in symptoms and reduction in US inflammation. However, as this was an open-label study, one cannot rule out the possibility that the observed clinical response represented a placebo effect.

The reliability of US scoring could be greatly enhanced by the development of a scoring atlas. To the best of our knowledge, there are no hand OA randomized controlled trials that use US as an outcome measure. Longitudinal studies are also required in order to study the association between US features and pain. Ultimately, the role of US scans in daily clinical practice needs to be addressed.

MRI has the ability of providing a multiplanar image of all joint components, including structural features such as osteophytes, cartilage, erosions/cysts, misalignment, and collateral ligaments and inflammatory features such as synovitis and tenosynovitis (Figure 1). MRI is the only modality that is able to show bone marrow lesions (BMLs), which have been proven as an important feature of structural progression and as a source of pain in knee OA 43.

Tan and colleagues 44 imaged DIP or PIP joints with OA by using high-resolution MRI and showed that virtually all structures were affected in both chronic and early OA, confirming that OA is a failure of the whole joint. BMLs, erosions, and synovitis were common features in this small study. The authors highlighted that collateral ligament abnormalities were universal in both chronic and early disease and demonstrated a close anatomic relation between the ligaments and erosions, BMLs, and bone formation. However, it should be noted that collateral ligament pathology was also frequent in the older controls, and whether these changes are only age-related or play a role in the pathogenesis of the disease is currently not clear. Furthermore, collagenous structures, such as the collateral ligaments, may demonstrate increased signal intensity because of the 'magic angle' phenomenon, which may lead to overestimation of ligament pathology 45.

The prevalence of MRI pathology in patients with hand OA has been investigated in several cohorts, of which the Oslo hand OA cohort is the largest 46. In that study, the authors found a high prevalence of synovitis on the basis of enhancement of gadolinium 46. Synovitis was also highly prevalent in joints without radiographic OA, and this is in line with previous observations in knee OA 47. The high prevalence of synovitis has been confirmed in smaller cohorts 3848. However, minimal gadolinium enhancement may occur also in persons without OA, and therefore we propose that synovitis not be scored as present unless there is an accompanying thickness of the synovium.

Haugen and colleagues 46 found a low prevalence of BMLs, in contrast to the high prevalence shown in the smaller studies 3848, and this could be due to lower field strength and poorer resolution. In the small finger joints, it is also important to be aware of partial volume artifacts that may mimic BMLs 49.

Patients with radiographic hand OA, in comparison with those with non-erosive hand OA, usually have a higher burden of disease. Wittoek and colleagues 38 confirmed that MRI-defined erosions, synovitis, and BMLs were more frequent in patients with radiographic erosive disease than in patients with radiographic non-erosive disease.

Haugen and colleagues 50 recently proposed a preliminary extensive MRI scoring system with an accompanying atlas for hand OA. The system includes assessment of osteophytes, JSN, erosions, cysts, misalignment, synovitis, flexor tenosynovitis, BMLs, and collateral ligament pathology such as absence/discontinuity and BMLs at insertion sites. The score was developed for the DIP and PIP joints, and future studies must confirm whether the score can be similarly applied to the metacarpophalangeal (MCP) and thumb base joints.

Haugen and colleagues 50 showed good intra- and inter-reader reliability for the assessment of most features in the proposed MRI scoring system. Good reliability of the scoring system has been confirmed in another cohort 48. These positive results suggest that MRI can be a reliable tool for assessment of OA pathology in the small finger joints.

The validity of MRI features in hand OA has been tested against histology and other imaging modalities 384648515253. Lewis and colleagues 51 compared MRI features and corresponding histological findings in three fingers from cadavers. Osteophytes and cartilage loss could be seen on histological sections, whereas only the largest structures were visualized by MRI. Tan and colleagues 52 recently combined high-resolution MRI and cadaveric histological studies in order to better understand the role of the collateral ligaments in the pathogenesis of OA. However, this study was limited by the fact that the MRI scans and histological sections were not from the same patients. The histological sections showed OA degenerative changes, fissuring, and cell clustering in the collateral ligaments and the enthesal organ, and the authors suggested that the MRI-defined ligament abnormalities were caused by degenerative changes.

Grainger and colleagues 53 were the first to report that high-resolution MRI was more sensitive than CR in detection of erosions and especially in marginal erosions in hand OA. These findings have since been confirmed by several studies using conventional MRI 384648. MRI was able to visualize more joints with erosions in patients with radiographic erosive hand OA but was also able to detect joints with erosions in patients with radiographic non-erosive disease. However, at this time, we do not know the prognostic value of these marginal erosions, and longitudinal studies are needed.

Two studies have shown that MRI is more sensitive than CR in detection of osteophytes 4648, and this may be due to the multiplanar demonstration of the joint by MRI. CR poorly visualizes bone formation located at the insertion of the extensor tendon unless there are oblique or lateral views. However, demonstration of osteophytes requires good contrast against adjacent structures because of the signal void of the cortical bone 38.

In line with the studies using US, Haugen and colleagues 54 found a significant association between synovitis and tenderness upon palpation, and the association was independent of other MRI features. This study is also the first to demonstrate an association between BMLs and pain in hand OA; together with synovitis, BML is the MRI feature that is most consistently associated with pain in knee OA 43. Significant associations with pain were also shown for bone damage. Kwok and colleagues 48 confirmed these findings but did not adjust for the co-occurrence of several MRI features.

Structural features such as bone damage and bone formation seemed to be associated with decreased physical functioning, but the results were not consistent and only weak associations were found. The MCP and thumb base joints were not imaged by MRI, and this may have affected these results 54.

An extensive scoring system with an accompanying atlas has been proposed 50. Validation studies have shown that MRI is more sensitive than CR in detection of erosions, suggesting that erosive hand OA is more common than previously indicated. Future studies should compare marginal erosions on MRI against histology or computer tomography or both, and longitudinal studies should evaluate the predictive value of these lesions. Synovitis and BMLs seem to be associated with pain in hand OA, and the associations should be confirmed in longitudinal studies. The sensitivity to change and the role of MRI as an outcome measure in clinical trials need to be determined. Optimally, further validation will lead to exclusion of less important features from the proposed scoring system, making it more feasible in practice.