Article Text

Download PDFPDF

Erosions versus joint space narrowing in rheumatoid arthritis: what do we know?
  1. Désirée van der Heijde
  1. Correspondence to Professor Désirée van der Heijde, Department of Rheumatology, C1-43, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, The Netherlands; d.vanderheijde{at}


Both erosions and joint space narrowing (JSN) are aspects of structural damage in rheumatoid arthritis. Most information is available on structural damage as one concept. However, the differential aspect of the effects on bone and cartilage could yield interesting information. Comparative information of these aspects can be based only on radiographic data on erosions and JSN. Both erosions and JSN are the consequence of inflammation, and their progression is inhibited by drugs that inhibit inflammation. These two processes often occur in parallel but joints in which erosions are present show a preference for progression of erosions and, to a lesser extent, development of JSN. The reverse is true for joints with JSN present, where there is a preference for worsening of JSN over development of erosions. Repair is possible for erosions as well as JSN and this is related to the absence of inflammation and effective treatment (especially methotrexate in combination with a tumour necrosis factor blocker).

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.


Structural damage is an important consequence of inflammation in rheumatoid arthritis (RA). Although structural damage may lead to deformities, which can be observed clinically, the best way to visualise structural damage is by imaging techniques. Structural damage in RA typically affects both bone, resulting in erosions, and cartilage, assessed as joint space narrowing (JSN). These two features can be assessed separately and aggregated into a total score. This total score is most commonly used as the primary outcome in clinical trials and in clinical epidemiological research (eg, assessing the relationship between inflammation, damage and function). There has been less focus on the erosion score and JSN score separately. In this paper various aspects of erosions versus JSN will be highlighted, such as which joints can be assessed, which imaging techniques can be used and which assessment methods can be applied. Further, we show some examples of efficacy in clinical trials and effects in clinical research.

How to assess erosions and JSN

The first question is to decide which joints to assess. The first differentiation is small versus large joints. The advantage of the use of the small joints in hands and feet is that you may obtain a lot of information on many joints in a single film. Moreover, these small joints are affected in most patients and they are representative of what is happening in large joints.1 2 Erosions can be assessed more easily in small joints than in large joints, but the assessment of JSN is easier in large joints. Imaging of small joints is now available routinely, whereas films of large joints are not part of routine imaging. For each film you only obtain information on one joint. Perhaps most important is the fact that large joints are not affected in all patients, and if so, it is often different large joints. The main advantage of larger joints is the large cartilage volume, which can be quantified. On the other hand, it is more difficult to detect erosions at an early stage in large joints.

Radiographs of hands and feet are the current standard technique to assess structural damage in RA. Semiquantitative scoring methods are the standard to quantify damage. In this paper only the methods that give scores for erosions and JSN separately provide useful information. Most widely used is the Sharp method with several modifications.3,,5 In this method JSN is a surrogate for the loss of cartilage; however, JSN might be influenced by positioning of the joints on the film and possibly also by joint swelling. If the latter were the case, a reduction in joint swelling would lead to JSN with the same amount of cartilage. This is a theoretical option and so far no proof exists. Other imaging techniques such as ultrasound and MRI are used frequently to assess inflammation and erosions but have not often been used to assess cartilage in RA. Consequently, at present, all the clinical data we have are based on information from radiographs of hands and feet assessed by semiquantitative scoring methods with JSN as a surrogate for cartilage.

Relation between inflammation and damage

The general paradigm in RA is that inflammation leads to structural damage, both erosions and JSN. This has been shown in group evaluations and based on sum-scores of all joints and also in evaluations of inflammation in an individual joint leading to damage in the same joint.6 7 This has mainly been investigated as effect on the combined scores of erosions and JSN, but there are also data available for erosions and JSN separately. An indirect proof is that inhibition of inflammation results in a reduction of progression of both erosions and JSN. This effect has been established for conventional disease-modifying antirheumatic drugs such as methotrexate (MTX), the tumour necrosis factor (TNF) blockers and also other biological agents such as abatacept and rituximab.8,,12 On the other hand, when there is a specific effect on bone (by an anti-RANKL antibody, denosumab) bypassing the pathway of inflammation, an effect on erosions was found but not on JSN with the tested dosing schedule.13

Relation between erosive damage and JSN in individual joints

Another intriguing question is if progression of erosions and JSN is a parallel process or if these processes are independent in individual joints. Does one process occur more frequently than the other? This was evaluated in the ASPIRE trial comparing MTX plus placebo with MTX plus infliximab in patients with early disease, in the individual joints of about 870 patients.14 Overall, there were more joints with erosive damage at baseline than with JSN. Moreover, joints with damage showed more progression in the following year, especially the joints with erosions (either erosions only or in combination with JSN). This pattern was shown to be present in all treatment arms but much more pronounced in the MTX monotherapy arm. If we stratified for those with progression in erosions versus those with progression in JSN in the MTX monotherapy arm, the joints that had both erosions and JSN at baseline had the highest OR for developing worsening of erosions (OR=3.8 (95% CI 3.1 to 4.4)) and for worsening of JSN (OR=6.9 (95% CI 5.3 to 9.0)). For worsening of erosions the OR for joints with only erosions at baseline was 2.95 (95% CI 2.6 to 3.4) and for joints with only JSN at baseline 1.8 (95% CI 1.4 to 2.3). In comparison, for worsening of JSN the OR for joints with only erosions at baseline was 1.2 (95% CI 0.9 to 1.7) and for joints with only JSN at baseline 4.0 (95% CI 3.0 to 5.3). So there was a tendency to show worsening of the same feature that was already present in the joint as compared with development of the other feature. In other words, the presence of erosions in a particular joint predisposes to progression of erosions in that joint and similarly, the presence of JSN in a particular joint predisposes to progression of JSN in that joint.

Scoring versus measuring

All available data on the evaluation of treatment efficacy are based on semiquantitative scoring. For example for JSN the observer is judging the joint space width but is not measuring the distance between the joint surfaces. Several efforts have been undertaken to develop automated measurement of erosions (eg, volume) and joint space width. Most data are available on JSN. Several semiautomated and fully automated joint space measurement methods are under evaluation.

An OMERACT study group under leadership of the late John Sharp performed a major validation study.15 Five methods were compared. They differed in the number of joints they evaluated (ranging from eight metacarpophalangeal joints to all the joints included in the Sharp scoring method); the way they measured the distance (the shortest distance, average distance, etc); the way they located the bone surfaces and in whether the method was operator dependent or fully automated. Images from the COBRA trial for 107 patients were available for the analyses at baseline and various follow-up points. For all these patients, Sharp–van der Heijde scores were available.

The first aspect that was tested was feasibility: how many films can be measured. One method had sufficient data for only one patient (for two time points) on all joint pairs, a second method had data for nine patients and the other methods had between 61 and 81 (of the 107 patients) sets of data. By reducing the requirement to a minimum of 50% of the joint pairs should be measurable for each patient, four of the methods were able to measure 97–100% of the patients, and one method still only 23%. Testing the reproducibility of results in 30 patients, the ICC of one method (able to measure 97% of the patients) was only 0.41 but for the other four methods it was good with a value of 0.80–0.98. The most important step was the testing of the discrimination between the trial arms. The modified Sharp score showed a statistically significant difference after 6 months in both the total score and the erosion, but not the JSN score. Three of five methods measured a significant difference based on the joint space width. Three methods scored the large majority of the patients (if a minimum of 50% of joint pairs was accepted) and showed good reproducibility. The total score followed by the erosion score obtained the highest discrimination, but it was promising that three measuring methods of JSN could discriminate between treatment arms whereas conventional scoring of JSN could not. The success rate of these methods clearly needs to improve. These data were obtained in patients with early RA and so with relatively little damage, and it is expected that in patients with more established disease the success rate will be even less.

The development of the measurement of erosions lags further behind. Several groups are developing methods but no validation studies are available yet. For the future a combination of manual scoring of erosions combined with measuring joint space width might be an option to increase the discriminatory power.

Repair of damage

Consideration of the repair of damage usually focuses on erosions. There have been various steps in the validation process of repair of erosions.16 Evidence of repair was shown at a group level in a trial arm, but also at an individual joint level. It was shown that negative scores in the modified Sharp score were in agreement with what experts judged as less damage.17 Finally, it was shown that negative scores occur almost exclusively either in joints without inflammation or in joints with improving inflammation, and also preferentially in joints in the combination of MTX and a TNF-blocker arm.18 All these pieces of information show that negative scores (obtained in blind scoring for time order) are an indication of repair of the erosive process. The majority of this information was obtained in the TEMPO trial. The same analyses have been performed with JSN as the outcome measure and exactly the same results were obtained (unpublished data). This indicates that repair of bone (erosions) and of cartilage (JSN) is an option. This is in line with animal data showing that cartilage repair can occur. To determine whether repair has consequences for the patient, a relationship with clinical outcomes should be established. The first indication that it might be clinically meaningful is that there is a tendency for patients with negative changes to have better function than patients without change, with mild and severe progression (in progressive order of worse function).19 This is based on the total score; information on erosions and JSN separately is not available.

Relevance of erosions versus JSN in the diagnostic process of RA

Both in the previous American College of Rheumatology (ACR) criteria for RA and in the recently developed ACR/EULAR criteria for RA only erosions are considered as one of the features that classify patients as having RA.20 21 In the 1987 criteria erosions in the hands was one of the features while in the 2010 criteria those patients with typical erosions are classified as having RA without the need to fulfil the criteria. This applies mostly to patients with advanced disease. JSN has not been included in any of the criteria sets as this is a much less typical feature for RA than erosions. JSN occurs frequently in patients with osteoarthritis and inflammatory arthritides of a different origin.

Overall conclusions

Comparison of the effects on bone and cartilage in RA can be based only on radiographic data on erosions and JSN. Inflammation leads to development of both erosions and JSN, which is largely a coupled process. Nevertheless, there is a preference for progression of erosive disease in joints with erosive damage and of worsening of JSN in those joints with JSN present. Drugs reducing inflammation have an effect on both erosions and JSN. Repair is possible for erosions as well as JSN and this is related to the absence of inflammation and effective treatment (especially MTX in combination with a TNF blocker). There are many unanswered questions, which are of interest for further evaluation.



  • Competing interests None.

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