Background
Remission in rheumatoid arthritis [RA] is arguably the ultimate goal of an anti-rheumatic therapy [
1,
2]. With modern therapeutic strategies, this goal can be achieved in the majority (up to 80%) of patients with early RA (ERA) [
3]. In this context, remission has been defined as a “state of absent disease activity”. In contrast, flares are defined as “a substantial increase of disease activity” associated with more radiological progression and worse functional outcome [
4]. Hence, continuous remission is the desired target state. A variety of response scores for RA patients based on clinical and serological data have been proposed and applied in clinical trials [
5]. Among these, the American College of Rheumatology (ACR) response criteria, which rely on a relative change of five core set variables [
6], and the European League Against Rheumatism (EULAR) response criteria, which are based on an absolute change of the composite Disease Activity Score in 28 joints (DAS28) including the ACR/EULAR remission criteria [
7‐
9], are most common.
In 2002, the OMERACT (Outcome Measures in RA Clinical Trials) magnetic resonance imaging (MR)I-group introduced a highly reliable sum-score (RA MRI Score (RAMRIS)) [
10] based on the semi-quantitative rating of the severity of synovitis, bone marrow edema and bone erosions in the joints of the hand and wrist [
10,
11]. The RAMRIS system has been shown to be a sensitive tool for the evaluation of therapy in patients receiving conventional synthetic and biologic DMARDs (Disease-modifying anti-rheumatic drugs) [
12,
13] similar to scores measuring clinical and serological parameters [
14]. However, Emery et al. reported a weak correlation between the individual change of the RAMRIS and the change of the DAS28 and C-reactive protein (CRP) levels, respectively. This was thought to be due to superior sensitivity of MRI compared to DAS28 and CRP [
15]. It is additionally known that the number and scope of erosions can increase instead of clinically low disease activity or remission (measured by DAS28). In particular, the existence and continuous presence of bone marrow edema as depicted by MRI is the strongest predictor for bony erosiveness in RA patients [
16,
17]: Imaging studies with ultrasound and MRI revealed signs of synovitis and/or bone marrow edema in patients with clinical remission (i.e. according to ACR or EULAR criteria). This phenomenon, often denominated “silent progression”, thus came into scientific focus [
18,
19]. Consequently, the question was raised whether extended remission criteria which incorporate modern imaging tools could be of superior value compared to clinical composite indices [
20].
We, therefore, investigated the value of MRI for the detection of erosive changes in patients with DAS28 improvement and/or remission of the German Remssion-plus cohort [
21].
Methods
Study design
Retrospective analysis was done on the Remission-plus cohort in which the data had been prospectively evaluated [
21].
Patients cohort
Datasets of 146 RA patients from the Remission-plus study cohort who fulfilled the ACR/EULAR 2010 Criteria for RA [
21] were retrospectively analysed in this study. Moreover, 80 patients who fulfilled advanced inclusion criteria consisting of (1) availability of two consecutive MRI scans (follow-up interval 1 year) of the clinically dominant hand and wrist, (2) the presence of DAS28 (CRP) scores at both time points and (3) had an DAS28 > 3,2 at T0 were investigated.
Clinical assessment
The following EULAR core set of variables was recorded: patient’s global assessment of overall disease activity, number of tender and swollen joints, erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP (<5 mg/l)).
The DAS28 [
22] was used to assess disease activity. Changes of disease activity were graded by the following classification criteria: DAS28 < 2.6 = clinical remission, ≤ 3.2 mild disease activity < 5.2 moderate disease activity and > 5.2 severe disease activity [
23,
24].
EULAR response assessment
Therapy response was graded by the following improvement criteria proposed by the EULAR committee [
7,
8]: DAS28 decrease >1.2 units and endpoint score <3.2 = good response, DAS28 decrease >1.2 units and endpoint score >3.2 or DAS28 decrease 0.6–1.2 units and endpoint score <5.1 = moderate response, DAS28 decrease <0.6 or DAS28 decrease 0.6–1.2 units and endpoint score >5.1 = poor response.
Imaging procedure [low-field MRI examination]
All examination were performed with the same low-field strength 0.2-T dedicated extremity MRI unit (Esoate, C-Scan, Esaote Biomedica Germany GmbH), and the same dedicated, dual phased-array coil. The clinically dominant hand was examined. Patients with renal dysfunction and known allergic reactions to gadolinium-diethylenetriaminepentaacetic acid (Gd-DTPA) were excluded from the study. The imaging protocol comprised pre- and post-contrast (i.v. gadolinium-based MRI contrast material, e.g. Magnevist, Schering AG, Berlin) T1-weighted images with a maximum slice thickness of 3 mm in at least two orthogonal planes and coronal fat-suppressed short tau inversion recovery (STIR) sequences [in detail coronar T1-weighted before contrast agent, coronar fat-suppressed STIR before contrast agent, 3-D GE T1-weighted after contrast agent with multiplanar reconstruction in three slide positions, coronar T1- weighted after contrast agent, axial T1-weighted after contrast agent].
MRI-scoring (RAMRIS)
MRI images were scored in each centre by MRI trained rheumatology specialists according to the RAMRIS based on OMERACT recommendations [
10]. MR images were read in consensus by two board-certified radiologists with special expertise in musculoskeletal MRI and trained for RAMRIS scoring.
Statistical analysis
Results of the analyses were reported as absolute numbers and percentages where appropriate. Data management and analysis was performed with SAS, version 9.3 (SAS Institute, Cary, North Carolina).
Discussion
Remission is the ultimate goal in RA-therapy. This has been underscored by successful applications of the Treat-to-Target (T2T)-strategies in studies and clinical practice in the last few years [
2]. Interestingly, MRI does not always reflect clinical improvement, but on the contrary, does show persisting or progressive joint pathologies in a considerable number of cases in most studies [
25,
26]. However, the presence of erosions is associated with a high risk of progression of the disease, while this was only shown for erosions in conventional x-rays, yet [
27,
28]. Up until now, therapy response criteria like the well-established EULAR response criteria are based on different constellations of clinical data, while matching MRI criteria are not available. In our study, a high number of 94% of patients showed erosions on MRI in at least one region. Importantly, roughly 40% of all patients who improved in DAS28 or who were in EULAR-defined remission, showed an increase of MR-detectable erosions after 12 months. Approximately 72% of these patients who worsened in ERO-subscore showed a subscore-deterioration of at least 2 points, so that an inaccuracy of the measurement is unlikely and a veritable increase of the MR-detectable erosivness must be assumed. Moreover, there was no relevant distinction between early and late RA, as even patients with a short disease-duration (less than 6 months) progressed.
The course of erosive changes depended on EULAR response in the current study: patients showing DAS28 improvement but EULAR non-response presented an increase of erosiveness in almost 72% of the patients, while only 31% of patients with good EULAR response had progressive erosions in MRI. Thus, our data is in accordance with a study by Van Gestel et al. who demonstrated that the improvement regarding the EULAR-response criteria is associated with less disease progression considering the clinical and conventional radiological course (highly sensitive imaging tools like MRI were not considered in this study) [
8]. In contrast, it has been demonstrated that up to 20–30% of patients reaching clinical remission showed progressive erosive joint damage (silent progression) [
9,
29]. Regarding the presented data, the proximal MCP-2 joints were most frequently affected by worsening in the ERO-subscore followed by the trapezoid bone and the proximal MCP-joints 3 and 4. The PIP joints were almost not affected (1/29). Regarding our additional analyses for possible predictive markers for silent progression (age, sex, antibody status, systemic inflammation (CRP) and RAMRIS-subscores), there were no statistical significant associations. However, we note that the study was not powered for specific subgroup analysis. It is known that erosive changes and BME detected by MRI lead to bone erosions which can be depict by conventional x-rays later on [
16]. There is a lot of evidence that erosive progression in conventional x-rays is related to functional loss in the course of disease [
30‐
32], while there is a lack of long term MRI data investigating the functional meaning of MR-detectable erosions, yet. Regarding that, long-term studies focused on this question are urgently needed.
Due to these issues, supplementary use of MRI scans could be of additional value to evaluate the therapy response, for example by using a smaller field of view to achieve a shorter examination time. In summary, MRI data in clinical routine confirm a high rate of silent progression despite DAS28 improvement or remission.
This study has some limitations. First, low-field MRI is used which is known to have a poorer local resolution in comparison to high-field MRI. Moreover, this multicentre study was a “real life” study without a static protocol, so that some patients were lost to follow-up or were excluded due to incomplete data. The study is not controlled for confounders such as RF, CCP-status, smoking or ethnicity. Moreover, we cannot completely exclude that progressive erosiveness detected by low-field MRI overestimates the risk of progression. In addition to that, it must be recognized that erosions were scored by MRI which is known as a very sensitive tool and could lead to occasionally over-interpretation. Hence, some sequences (for example STIR-sequences) are not fully comparable to high-field MR-scans due to the poorer resolution. To solve this last issue, both compound scores, such as the DAS28 and MRI based scores (e.g. RAMRIS), should be evaluated against gold standards such as functional or conventional radiological outcome measures in the long term in future studies.
Acknowledgements
We would like to thank Erika Rädisch for the assistance in receiving the MRI-scans.