Article Text


Remission by imaging in rheumatoid arthritis: should this be the ultimate goal?
  1. Désirée van der Heijde
  1. Correspondence to Professor Désirée van der Heijde, Department of Rheumatology, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, The Netherlands; d.vanderheijde{at}


Remission is often selected as the ‘treat to target’. There is a plea to include imaging. Imaging remission can apply to structural damage and/or inflammation. For structural damage, radiographs are mostly used. A definition is needed which could be either strict, with no progression occurring, or which takes measurement error into account and uses the smallest detectable change. Mostly imaging remission refers to inflammation as assessed by ultrasound or MRI. The reason for arguing that imaging remission should be included for inflammation is that inflammation may still be present in patients who are in clinical remission. The level of inflammation depends on the clinical remission definition that is used. Bone marrow oedema is the feature that is most predictive of radiographic progression. However, before imaging remission can be implemented as a recommendation, a definition of remission by imaging needs to be established. A choice has to be made about the level of inflammation that can be tolerated and how this needs to be assessed (which imaging method, which feature, which joints, which cut-off point). Moreover, imaging remission should only be selected as a target if it can be proved that it can be treated and that the outcome of the patients will be improved by trying to achieve imaging remission in addition to clinical remission. This proof is not yet available, and too many unanswered questions remain to recommend including imaging remission of inflammation in a definition of remission.

Statistics from


Improvements in management of rheumatoid arthritis resulted in a better outcome and clinical remission as a feasible goal. Therefore, the first recommendation from the ‘treat to target’ recommendations is ‘The primary target for treatment of rheumatoid arthritis should be a state of clinical remission’.1 2 There are several definitions for clinical remission, some stricter than others.3 Especially in early disease up to 60% of the patients achieve clinical remission, depending on the remission definition used. In established disease it is more challenging to achieve remission, especially in patients for whom several drugs have already failed. Although not all patients achieve clinical remission, there is a plea to aim for an even stricter definition of remission: remission by imaging.4 In this manuscript, I will discuss the various aspects that should be taken into account before this recommendation can be widely implemented. Data and expert opinion will be presented to answer the following questions: Which imaging techniques should be used to define imaging remission? And which definition should be applied, what cut-off point should be used? What are the consequences of inflammation detected by imaging while in clinical remission on the long-term outcome? Does remission assessed by imaging lead to better (functional) outcome than clinical remission? Is aiming at remission by imaging feasible in clinical practice?

Definitions of remission

Clinical remission

There are several definitions for clinical remission based on composite scores, such as the Disease Activity Score (DAS), 28-joint count DAS (DAS28), Simple Disease Activity Index (SDAI), Clinical Disease Activity Index and a Boolean definition as recently proposed by American College of Rheumatology (ACR)/European League Against Rheumatism (EULAR).3 They all have in common that joint counts (tender and/or swollen) are included. During the validation process of the ACR/EULAR remission criteria one of the comparisons of the various definitions was the positive likelihood ratio (LR+) for a good outcome for radiographic progression, which was defined as progression ≤0. Both the SDAI and the Boolean definition (28 tender joint count ≤1, 28 swollen joint count ≤1, C-reactive protein ≤1, patient global ≤1) showed the highest LR+ of 3.0 and 2.9, although definitions of several other remissions had the same or a very similar LR+ (2.6–2.9). The performance of the DAS28 remission definition was clearly worse (LR+ 1.0). But, it should be realised that these data were based on very few patients in remission (depending on the definition ranging from 10 to 36) and a very strict definition of radiographic progression (≤0; see later). The overall conclusion is that the stricter the definition, the less radiographic progression.

Imaging remission structural damage/inflammation

For the definition of remission by imaging, we first have to define what we want to assess by imaging: structural damage or inflammation. Structural damage can be assessed by conventional radiographs (CRs), ultrasound (US) and MRI, while inflammation can be assessed by US and MRI. Assume we take CR for the definition of remission of structural damage by imaging. The remission definition should be applicable to an individual patient. There are again a few decisions to be taken. The ultimate is that we want to define ‘no progression’. Here we focus on progression and not on a status score: even patients who have established disease with structural damage in the joints should be able to achieve ‘remission by imaging’, which indicates that there is no further progression. In a clinical trial there are typically two observers and data from a comprehensive scoring method are available. But what cut-off point should be used to define no progression? Progression ≤0, ≤0.5 or less than the smallest detectable change (SDC)? The recommendation is to prefer ≤0.5 rather than ≤0 when the average score of two observers is being used (and ≤0 when there is one observer).5 But the selection of this cut-off point does not take measurement error into account. The SDC (the smallest change that can be detected beyond measurement error to determine if a patient shows change) takes measurement error into account but the SDCs obtained are often higher than the scores that rheumatologists consider clinically relevant progression and might therefore be too lenient.6 7

Precision of scoring of imaging, in general, is a frequently debated topic. However, this discussion is rarely brought up for clinical assessments. For example, how reliably can a swollen joint of ≤1 (as being used in the remission criteria) be assessed? In a study comparing the reliability of clinical and radiographic measures, it has been shown that radiographic scores are more reliable than any of the clinical measures.8 So perhaps the requirements for radiographic progression are too rigorous and a cut-off point of ≤0.5 would be sufficient for the current aim. The situation in clinical practice is different: typically there is only one observer and infrequently a scoring method is applied. However, without a scoring method it is difficult to define a cut-off point. If, for example, the simple erosion narrowing score is used, a method that counts the number of affected joints, a newly affected joint could be used as the cut-off point to define progression.9 This omits progression in already damaged joints, but is sufficiently sensitive to change, especially in early disease.10 The implication is, however, that a scoring method in clinical practice should be implemented.

For remission of inflammation by imaging a decision needs to be taken on the imaging method used (US grey scale (GS), US power Doppler (PD) or MRI). Which feature (US-synovitis, MRI-synovitis, MRI-bone marrow oedema (BME))? Which joints should be assessed? Finally, which cut-off point should be used: based on the number of positive joints, and which cut-off point for each joint?

Døhn et al performed a nice comparison of the reliability of several scores: erosion score on CR, CT, MRI and US, and erosion volume on CT and MRI of metacarpophalangeal joint 2–5 (MCP2–5) of the dominant hand.11 They calculated for each method the SDC and presented the percentage of patients showing progression, no change and regression in two different ways: without taking measurement error into account (cut-off point 0) and after taking measurement error into account using the SDC as cut-off point. In the first instance the percentage of patients showing progression according to the various imaging methods ranged from 27% (MRI erosion score) to 58% (MRI erosion volume). After using the SDC the percentages of patients showing progression ranged from 12% (both MRI erosion score and erosion volume) to 24% (US erosion score). So which cut-off point is used to define progression has a major impact on the percentage of patients showing progression.

Inflammation on imaging while in clinical remission

Several studies have indicated that patients show progression in structural damage while they are in clinical remission.12 One of the explanations is that patients might be in clinical remission, but they still have subclinical inflammation.2 11 The level of progression is partly dependent on the definition of clinical remission (see above).3 But even in patients fulfilling a strict definition, progression can be seen. Subclinical inflammation can be detected by imaging. And indeed authors concluded that ‘Most patients in drug-induced remission continue to have synovitis, detectable using US and/or MRI’.2 13 Brown et al showed in 2006 that 59/107 (55%) patients who were in clinical remission according to their treating rheumatologist fulfilled the DAS28 remission criteria and 61 (57%) the (old) ACR remission criteria.2 Of the patients fulfilling the DAS28 remission criteria, 84.2% showed synovial hypertrophy with US and 50.9% showed an increased PD signal, 96.2% synovitis and 51.9% BME on MRI. This group of patients was mainly treated by conventional disease-modifying antirheumatic drugs (DMARDs).

Døhn et al describe in the above cited publication the results in a group of 52 patients after treatment of 12 months with a tumour necrosis factor (TNF) blocker.11 Thirty-eight per cent of patients reached a DAS28 remission. MCP2–5 of the dominant hand were assessed. At the 12-month follow-up all patients had at least one joint with synovitis on MRI (95% of joints); all patients had at least one joint with synovitis on the GS (45% of joints); BME on MRI was present in 63% of patients (20% of bones); and a positive PD signal was present in 87% of patients (26% of joints). Only a minority of the patients were in clinical remission, but the percentage of patients with positive imaging findings of inflammation was high, though no overall progression of structural damage assessed by MRI and US was seen over that same period.

In a case–control study, Saleem et al included patients in remission according to DAS28 and compared US findings of synovitis in 50 patients in remission while receiving DMARDs and 50 while taking a combination of DMARDs and TNF blockers.13 There were no differences between the groups in PD scores, and patients in the DMARD group had significantly lower GS scores, which was opposite to the result expected.

Is inflammation on imaging harmful?

As shown above, the few studies that have investigated inflammation on imaging showed clearly that even in patients in clinical remission, inflammation is present. The next question is: Is this harmful? Does inflammation on US/MRI lead to radiographic progression? Does inflammation on US/MRI lead to a worse outcome? Most predictive studies are performed in patients with various levels of disease activity, not only in patients in clinical remission. So it is hard to answer this question with data and we need to use the literature including patients with all levels of disease activity. Døhn et al checked the predictive value of inflammation at baseline to predict progression of erosions on CT.11 Only BME at baseline was a statistically significant predictor, and the PD signal just missed the significance level. But neither synovitis on MRI, nor on US at baseline were predictive. While the area under the curve of all four parameters (BME, PD, synovitis on MRI and US) was significantly related to erosive progression on CT.

Brown et al investigated the predictive value of clinical, US and MRI findings in predicting radiographic progression in individual MCP joints in 90 patients in clinical remission according to the rheumatologist (about 55% DAS28<2.6).14 Of the 380 MCP joints, only 10 showed radiographic progression. In univariate analyses, the clinical joint findings of pain and swelling did not reach significance, while the US, PD, MRI BME and synovitis scores reached statistical significance. The highest prediction was reached by the number of joints with a positive PD signal, while this was not statistically significant for the number of joints positive for GS, MRI BME and synovitis. The major drawback of that study was the small number of joints showing progression.

McQueen and Naredo presented a good overview of the findings on US and MRI to predict radiographic progression.15 They found two additional studies showing that PD signals predict radiographic progression, but a much weaker correlation than with the study by Brown. MRI synovitis was investigated in many studies but it was a positive predictor only in the study by Brown, while MRI BME was consistently a positive predictor in many studies. This led the authors of the review to the conclusion ‘…. much evidence exists to suggest that osteitis (bone marrow oedema) is more strongly predictive of bone erosion than synovitis, supporting the notion that there is a more direct connection between bone inflammation and bone damage than between synovial inflammation and bone damage’.

Should imaging be part of the remission criteria?

Saleem et al applied several remission criteria to a group of 128 patients in clinical remission (about half treated with DMARDs and the other half with DMARDs and TNF blockers) and investigated the level of inflammation on imaging.4 For example, comparison of patients in SDAI remission with those not in SDAI remission showed that a similar percentage of patients had a positive PD signal (14.7% vs 14.6%); if the joints with a minimal PD signal were omitted the results were still similar in patients who did or did not fulfil the remission criteria, with the PD signal positive in 4.1% versus 4.5%, respectively. They also compared the GS and PD score across four subcategories of SDAI remission and again these were similar. Based on their data, they concluded that imaging should be part of the remission criteria.

So let us assume that we accept this conclusion. The next question is ‘how to monitor and treat subclinical inflammation’? How often should we assess subclinical inflammation? Which technique should be used? A PD signal predicts radiographic progression to some extent, but it is BME, in particular, that has the highest predictive validity. If we follow treat to target recommendations, imaging would need to be implemented when a patient has reached clinical remission. How often should the follow-up imaging be for these patients? Is there sufficient availability of personnel and equipment to perform these frequent assessments? Finally, how should we treat these patients? Should we change treatment while a patient is in clinical remission with the only aim being to abolish subclinical inflammation detected by imaging? What is the acceptability of this approach to the patient, the rheumatologist and the payer? In my opinion, this can be justified only if the long-term (patient-reported) outcome is better in patients in clinical and imaging remission than in patients in clinical remission only. This information is completely lacking. Only one case–control study shows that patients in imaging remission had similar Health Assessment Questionnaire and rheumatoid arthritis quality-of-life scores to those of patients in clinical remission (by DAS28) only.13 This was true both for the subgroup treated with conventional DMARDs and the subgroup treated with a combination of DMARDs and TNF blockers.


Only a few papers have investigated imaging in patients in clinical remission, and several are from the same institute, making it unclear how many patients are included in more than one paper. Overall, only relatively small numbers of patients are studied and information on only a few joints is taken into account. For structural damage CR seems to be the most feasible choice of imaging method, certainly in clinical trials. For clinical practice this would be mostly useful if a (simple) scoring method were applied. However, the definition of progression is unclear (taking measurement into account or not). For the imaging method used to detect inflammation, it is more practical to use US rather than MRI for frequent applications. But again a definition is needed of the joints on which this should be based and which definition for progression should be used. Patients in clinical remission show radiographic progression, but the stricter the remission criterion the less progression is seen.

Various studies showed that signs of inflammation are common in patients in clinical remission, but again the stricter the criteria the less synovitis is seen, although still common. But a strong PD signal is found only in about 4–5% of joints. In most US studies the PD signal is more predictive than the GS; however, synovitis on MRI does not predict damage progression and BME on MRI is the strongest predictor. So while US is the most feasible method to use, the highest prediction is obtained by BME on MRI. There is no proven positive effect on patient outcome of being in imaging remission in comparison with being in clinical remission only. Lastly, it is unknown if residual synovitis on imaging can be treated effectively. The final conclusion is that more data are needed before a recommendation that imaging, with far reaching consequences, is included in the definition of remission in addition to clinical remission.


View Abstract


  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.

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.