The past two decades have seen unprecedented advances in therapy for rheumatoid arthritis (RA), with nine effective biological agents introduced since 1998. Imaging played a central role in validating these novel agents, as none could have gained regulatory approval without definitive proof of structure-modifying efficacy through rigorously applied radiography in multicentre randomised controlled trials (RCTs). However, a number of factors have recently made the use of radiography in clinical trials increasingly impractical. Ironically, the most important of these has been the success of the biological agent revolution itself, as the availability of effective therapy has made extended placebo control unethical.

One alternative to placebo control is active-comparator study design. However, radiographic progression rates in such studies are low, necessitating prohibitively large numbers of patients and long study durations to demonstrate superiority of one treatment over another. Accordingly, there have been no such trials reported to date using radiography. Adding placebo or a test therapy to methotrexate (MTX) in patients with active disease despite prior use of MTX has been the most common alternative approach, but in these studies the placebo add-on is typically not maintained for longer than 6 months, and non-responders are offered rescue therapy or withdrawn by 12–16 weeks. Later time points are sometimes imputed, but the reliability of conclusions drawn from such data is questionable. Concern has also been raised about the ethics of blinding research patients and investigators, sometimes for years, until the last subject has completed follow-up.1

Another factor reducing radiography's utility in multicentre RCTs has been the conversion of most clinical facilities from screen-film radiography to digital radiography.2 While this has been a positive change for image archival, patient throughput and environmental impact, it has added complexity and variability to multicentre clinical trials. Not only is the spatial resolution of digitally acquired radiographs generally lower than that of images generated by digitising originally film-based radiographs, as was the standard in earlier RCTs, verifying the quality of digitally acquired radiographic images is harder for technologists. In the past, technologists could visually inspect film radiographs directly on backlit view boxes. However, digital radiography generates 12-bit electronic images comprising 4096 grayscale units. Since the human eye can discriminate only a small fraction of these, technologists must infer image quality indirectly from the exposure index, which is a parameter that is computed differently by different manufacturers. Not only has this added technical variability to multicentre RCTs, but, because some imperfections in digital image quality can be corrected through post-processing by the reading radiologist after the patient has left, the focus of radiology technologists at busy medical centres has shifted from meticulous image acquisition to rapid patient throughput. Unfortunately, reproducible patient positioning on serial radiographs, which is critical for reliably assessing change in RCTs, cannot be corrected by digital post-processing.

This combination of ethical, technical and practical problems has fuelled an urgent need for more sensitive and rapid ways of imaging joint damage and its inflammatory causes in RA. Of the medical imaging alternatives currently available, MRI provides the most viable solution for multicentre RCTs. The level of evidence supporting MRI's validity and utility in RA clinical trials was recently reviewed at the 4th Annual Meeting of the International Society for Musculoskeletal Imaging in Rheumatology3; that review forms the basis for this viewpoint.

Numerous methodological and observational studies have established the construct validity of MRI depiction of bone erosions and cartilage loss in RA.4–10 Not only has MRI been repeatedly shown to be more sensitive than radiography for detecting these classical features of joint damage in RA, it is uniquely able to visualise the upstream inflammatory drivers of bone erosion and cartilage loss, namely synovitis and osteitis,11–18 as well as other important features of the disease, such as tenosynovitis.19

Over the past few years, several RCTs have provided direct evidence of the ability of MRI to discriminate progression and treatment effect. All these studies used the validated semiquantitative RAMRIS (Rheumatoid Arthritis Magnetic Resonance Imaging Score) system,20 which scores erosions (from 0 to 10, in increments of 10% of articular bone loss), osteitis (from 0 to 3, in increments of 33% of articular bone) and synovitis (from 0 to 3, in increments of 33% of the synovial cavity) in the wrist and metacarpophalangeal joints of the hand. Figure 1 shows the rate of erosion progression in the placebo arms of all seven RCTs published or abstracted thus far that included more than 20 patients per arm. Five of these studies evaluating different therapeutic compounds (denosumab (DEN),21 fostamatinib (TASKi),22 adalimumab (OPTIMA),23 abatacept (ASSET),24 rituximab (SCORE)25) used the same MRI acquisition technique, image quality control process and centralised readers. The consistency in the progression-rate measurements (range=0.22–0.30 RAMRIS units/month; median=0.27; mean=0.27, SD=0.04) of these studies, despite having been conducted years apart and with the readers blinded to the time order of the visits in each study, attests to the accuracy and discriminative power attainable with MRI in multicentre RCTs. Two studies of golimumab (GO-BEFORE26 and GO-FORWARD27) that used slightly different MRI protocols, image quality control providers and central readers from the other five studies failed to show significant progression in the placebo arm, but one of the studies, GO-BEFORE, nevertheless demonstrated statistically significant reduction in bone erosion score with treatment in only 12 weeks.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Change in RAMRIS (Rheumatoid Arthritis Magnetic Resonance Imaging Studies) bone erosion score in the placebo arms of seven recent randomised controlled trials: ABA, ASSET study of abatacept24; ADA, OPTIMA study of adalimumab23; DEN, study of denosumab21; FosD, TASKi study of fostamatinib22; GLM-1, GO-BEFORE study of golimumab26; GLM-2, GO-FORWARD study of golimumab27; RTX, SCORE study of rituximab.25 Reproduced, with permission, from Peterfy.3

Table 1 shows the change in MRI bone erosion, osteitis and synovitis in the placebo and treatment arms of these seven RCTs as well as in one examining tocilizumab, in which both arms included active treatment (ACT-RAY28). Of the five studies that showed erosion progression in the placebo arm, only TASKi,22 a 12-week RCT of the oral spleen tyrosine kinase (Syk) inhibitor, fostamatinib, failed to show suppression of erosion in the treatment arm.22 In OPTIMA,23 adalimumab/MTX combination therapy showed significant suppression of erosion progression at 26 weeks, with only 27–32 patients with early RA per arm. Typically, several times this number of patients is needed to discriminate treatment effect with radiography. The ASSET Study24 showed suppression of bone erosion progression with abatacept+MTX in only 16 weeks, and both SCORE (rituximab)25 and GO-BEFORE (golimumab)26 showed suppression of bone erosion at 12 and 24 weeks, with statistical significance being reached in GO-BEFORE at 12 weeks. SCORE also showed suppression of cartilage loss using a nine-point MRI score8 at 24 weeks (cartilage loss was not evaluated at 12 weeks). In addition, all eight RCTs in table 1 showed suppression of synovitis and osteitis within 12 weeks. In ACT-RAY,28 significant suppression of synovitis was achieved in only 2 weeks. In all of these studies, MRI results correlated with clinical and laboratory findings. Moreover, MRI predicted subsequent radiographic findings in those studies that included radiography (DEN,21 SCORE,25 GO-BEFORE,26 GO-FORWARD27).

These RCTs demonstrate that it is possible to discriminate therapeutic efficacy of different structure-modifying therapies with MRI in less than 6 months—in some cases even less than 3 months—using small numbers of patients. In light of this evidence and the ethical imperative to limit the time that patients are exposed to ineffective treatment in RCTs, it is important that regulatory authorities now consider the use of MRI data as an alternative to radiographic data in support of claims of inhibition of progression of structural damage.