Article Text

Download PDFPDF

Original research
Radiographic progression in clinical trials in rheumatoid arthritis: a systemic literature review of trials performed by industry
  1. Yune-Jung Park1,2,
  2. Ana Maria Gherghe3 and
  3. Desirée van der Heijde4
  1. 1Internal Medicine, The Catholic University of Korea, Seoul, Korea (the Republic of)
  2. 2Rheumatology, St. Vincent’s Hospital, The Catholic University of Korea, Suwon, Korea (the Republic of)
  3. 3Internal Medicine and Rheumatology, Cantacuzino Hospital, Bucharest, Romania
  4. 4Rheumatology, LUMC, Leiden, Netherlands
  1. Correspondence to Desirée van der Heijde, Rheumatology, LUMC, Leiden, Netherlands; mail{at}


Objectives To summarise radiographic data in randomised controlled trials (RCTs) as part of the radiographic inhibition claim of disease-modifying antirheumatic drugs (DMARDs) approved for patients with rheumatoid arthritis (RA).

Methods A systemic literature review was performed using the Medline database from 1994 to February 2020. The results were grouped based on the scoring methods (Sharp, Genant modification, van der Heijde modification) and RA patient populations.

Results One hundred sixty-eight publications were selected. After detailed assessment, 52 RCTs (7 methotrexate (MTX)-naive, 23 MTX inadequate response (IR), 9 DMARDs IR and 3 tumour necrosis factor-alpha inhibitors (TNFi) IR studies) were finally included. Information on patient population, scoring method used, reader reliability, statistical analyses and detailed radiographic data on baseline and change scores over multiple follow-up periods are presented.

Conclusion The data gathered in this review serve as a repository for the design of future trials with radiographic damage as an outcome.

  • Arthritis
  • Inflammation
  • Synovitis
  • Autoimmune Diseases
  • Rheumatoid Arthritis
  • Systemic Sclerosis
  • Outcomes research
  • Spondyloarthritis
  • Ankylosing Spondylitis

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

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.


Rheumatoid arthritis (RA) is a chronic inflammatory disorder characterised by synovitis and destruction of synovial joints, leading to severe disability and premature mortality.1 The introduction of disease-modifying antirheumatic drugs (DMARDs) in the treatment of patients with RA has led to improved management of RA, making not only (complete) symptom relief, but in addition the prevention of long-term structural damage the current goal of therapy.2 The prevention of structural damage is also recognised by the Food and Drug Administration (FDA)3 and the European Medicines Agency (EMA)4 as a separate claim for a drug and defines the disease-modifying capability of a drug.

To date, radiographs are still considered the most appropriate method to assess structural damage in RA. MRI is regarded as a supportive imaging method but is not yet accepted as an alternative for radiographs by the FDA and EMA.3 4 Validated radiographic scoring methods exist and are widely used for assessment and follow-up of joint damage in RA. Labelling for ‘inhibition of radiographic progression’ is granted to both synthetic and biological DMARDs (bDMARDs) based on randomised controlled trials (RCTs) in which retardation of structural progression is demonstrated using such validated scoring methods.

There is a general tendency for less radiographic progression in more recent RCTs.5 This may be due to: earlier, more effective treatment of patients included in RCTs, leading to less structural damage at baseline6; or to less exposure to placebo (control) therapy due to earlier rescue. These developments have made it challenging to demonstrate the superiority of new drugs in inhibiting radiographic progression in RCTs. For future RCTs, this will require even more careful selection of patients prone to radiographic progression and perhaps change in study design.5 In this context, an overview of data used to get a label for ‘inhibition of structural damage’ by pharmaceutical companies would be of interest. Existing reviews of radiographic data do not include trials of more recent bDMARDs and targeted synthetic DMARDs (tsDMARDs), such as certolizumab, golimumab, tocilizumab and janus kinase inhibitors,7 or do not consider methodological aspects of analysing radiographic data.6

The purpose of this overview was to summarise radiographic data in RCTs performed by pharmaceutical companies, usually to obtain the claim of radiographic inhibition, of all DMARDs approved for patients with RA. This can serve as a repository for the design including power calculations of future trials.


This review is based on published articles reporting the results of RCTs for RA performed by pharmaceutical companies, in which the effects of new treatments on radiographic damage were evaluated. These trials were mostly used to obtain the registration as DMARDs for the respective treatment; however, some are pharmaceutical company-performed post-approval studies. A literature search on the topic was conducted in PubMed. The research question was translated into an epidemiological research question according to the PICO method (Patients, Intervention, Comparator and Outcome).8 Patients were defined as adults with RA according to the 1987 American College of Rheumatology (ACR) criteria9 or to the 2010 ACR criteria10; intervention was defined as any drug; comparator as placebo or another active drug; outcome was radiographic progression.

The literature search was carried out in PubMed. The database was searched using the following specific terms (synonyms and all possible combinations): rheumatoid arthritis, adalimumab, etanercept, infliximab, certolizumab, golimumab, anakinra, tocilizumab, rituximab, abatacept, tofacitinib, leflunomide, upadacitinib, baricitinib, peficitinib, ruxolitinib, filgotinib, ustekinumab, guselkumab, secukinumab, ixekizumab, canakinumab, brodalumab, sarilumab, secukinumab, sirukumab, radiographic, radiologic, structural or progression, Sharp, van der Heijde, Genant or Larsen. The search was limited to English language literature without a time limit. The last search was performed on February 6, 2020. The references of the selected articles were manually reviewed to identify additional relevant publications. Unpublished study enrolment dates were searched on, using, when available, the study identification number from publications. Pharmaceutical companies were also contacted to obtain unpublished data. Abstracts were not included as these contain insufficient detailed information.

The retrieved citations were managed using EndNote. One reviewer performed a selection based on titles and abstracts using predefined inclusion and exclusion criteria. The selected citations were discussed among two authors and included by consensus. To be included, articles had to contain data collected from any RCT performed by pharmaceutical companies for treatment registration (and their open-label extensions) or to further support the inhibition of radiographic progression, involving adult patients with RA (age >18 years). Articles with the following characteristics were excluded: investigator initiated and strategy studies, pediatric population, non-RA, languages other than English, no radiographic results reported, review articles, guidelines papers, case reports, commentary or letters. Based on this screening, full-text articles were obtained for more detailed reviewing.

Data extraction

An electronic form was used for the data extraction. The study characteristics including study design, patient enrolment dates, all relevant baseline demographics, clinical characteristics and all baseline and follow-up radiographic data were recorded. Trials were divided into methotrexate (MTX)-naïve, MTX inadequate responder (IR), DMARDs IR, or tumour necrosis factor-alpha inhibitors (TNFi) IR populations. The Larsen method was included in the literature search; however, this was used only in a limited number of older RCTs for which we have also results with the Sharp method. Therefore, we decided to exclude reporting data based on the Larsen method. A detailed data extraction flow chart is depicted in online supplementary figure 1.

Figure 1

Flow chart of the trial selection process.


A total of 1170 publications were identified in PubMed. Based on title and abstract review, 1002 publications were excluded because they did not include the population or intervention of interest, did not report radiographic results, were not randomised, controlled trials or were not performed by pharmaceutical companies. The remaining 168 publications were read full text. Of these, 104 manuscripts describing the results of 52 RA trials were included and were used for data extraction. A flow diagram summarising the screening and selection of articles is shown in figure 1.

The 52 included RCTs are presented in table 1. The MTX-naïve group included 17 RCTs (2 of a conventional synthetic DMARD (csDMARD) (leflunomide), 10 of a TNFi (adalimumab, certolizumab, etanercept, golimumab, infliximab) and 5 of a non-TNFi bDMARD or tsDMARD (abatacept, baricitinib, rituximab, tocilizumab, tofacitinib). MTX was mostly used as comparator. The MTX IR group included 23 RCTs (12 trials of TNFi (adalimumab, certolizumab, etanercept, golimumab, infliximab, biosimilar of etanercept, biosimilar of infliximab) and 11 of non-TNFi bDMARD or tsDMARD (anakinra, abatacept, baricitinib, denosumab, peficitinib, sarilumab, tocilizumab, tofacitinib, upadacitinib)) again with MTX as the most frequently used comparator. MTX IR trials enrolled 77–651 patients in the comparator group and 85–651 patients in the treatment group. The DMARD IR group included 9 RCTs (1 csDMARD (leflunomide), 4 TNFi (certolizumab, etanercept, golimumab) and 4 non-TNFi bDMARDs and tsDMARDs (baricitinib, sirukumab, tocilizumab)), which included 91–556 patients in the comparator group and 102–557 patients in the treatment group. There were three trials conducted in a TNFi IR population investigating adalimumab, rituximab and secukinumab, which studied 16–214 patients in the comparator group and 17–308 patients in the treatment group.

Table 1

Randomised controlled trials included for review

Main patient characteristics at baseline

The main baseline demographic and clinical characteristics of the patients included in the 52 trials are reported in table 2. MTX-naïve trials generally included patients with short disease duration (mean duration per treatment group was less than a year), while MTX IR trials had a longer mean disease duration per treatment arm (1.7–11 years). Rheumatoid factor (RF) positivity was reported in 48 out of 52 trials (92.3%), and anti-citrullinated protein antibody (ACPA) positivity was described in 20 out of 52 trials (38.5%). In recent trials, the proportion of RF or ACPA positive patients increased (online supplementary figure 2). The majority of patients has high level of disease activity (table 2).

Table 2

Baseline demographic and clinical characteristics of the patients*

Scoring methodology

The description of the radiographic methodology used in each trial is shown in table 3. Conventional radiography (CR) of hands and feet was performed in all trials, except for IL-1Ra, where only hands were included. CRs were usually evaluated by two readers. However, several trials, such as leflunomide trials,11 IL1-Ra,27 PRIZE22 and CAMEO37 trial, only one reader scored CRs. When CRs were scored by two readers, the average score of the two readers was reported. The inter-reader and intra-reader intraclass correlation coefficients of status scores were reported in several RCTs and showed a high reliability of the measurements between readers and within a reader. The readers employed the Sharp method, the van der Heijde modification of the Sharp (SvdH) method or the Genant modification of the Sharp (GS) method. All methods include separate scores for erosions (ES) and joint space narrowing (JSN) that add to a total score. The maximum total score is 398 for the Sharp method, 448 for the SvdH and 290 for the GS method. Results were reported for the total score, as well as for the separate scores, per treatment arm. The change (Δ) in radiographic scores, which represents the difference between the scores at the follow-up visit and the scores at baseline, was the main outcome. A variety of approaches were used to deal with missing data, including linear extrapolation (LE), last observation carried forward and multiple imputation methods. For the patients who withdrew early or who received rescue medication, CR scores were usually estimated by LE of the scores from the radiographs taken at an early visit.

Table 3

Radiographic methodology and statistical analysis in each trial

Radiographic results of the trials

Online supplementary table 1 presents the radiographic outcomes of all 52 trials until 1-year follow-up. The table is organised per scoring method (Genant, Sharp-van der Heijde and Sharp), and thereafter per patient population (MTX-naïve, MTX-IR, DMARD-IR, TNFi-IR). Per arm (intervention and control) the mean (SD), median (IQR) and range of the total score, erosion score and JSN score at baseline is presented. This is followed by the mean (SD) and median (IQR) change scores at 6 months and at 1 year. Finally, the percentage of non-progressors is presented. Non-progression in the RCTs is defined as: the number (%) of patients with ≤3 units of change in erosion scores at follow-up compared with baseline, the number (%) of patients with ≤0 units of change in total, erosion or JSN scores, the number (%) of patients with ≤0.5 units of change in scores, the number (%) of patients with ≤ smallest detectable difference (SDD), the number (%) of patients with ≤ smallest detectable change (SDC). The SDD is defined as the smallest difference between two independent measurements (ie, patients) that can be interpreted as a ‘real’ difference beyond measurement error, while the SDC represents the SDC beyond measurement between two successive scores of the same patient.62

Of the 52 studies, 37 studies were analysed using the SvdH scoring method. From these, 8 were conducted in early RA (EA) patient populations and 29 were conducted in established patient populations. The baseline total SvdH score were 5–25 in EA trials and 9–79 in established RA trials (online supplementary figure 3). In both patient populations, no clear change in baseline total SvdH score was observed over the years.

Long-term extension (LTE) trials

There were 22 LTE trials as shown in online supplementary table 2. All trials have a follow-up of 2 years and several an additional follow-up up to 10 years in one trial.

Withdrawal or tapering trials

Finally, the data of the 7 trials that investigated radiographic progression after tapering or withdrawal are summarised in online supplementary table 3.


This is the first overview of radiographic data from all RCTs performed by pharmaceutical companies to obtain registration for new drugs that inhibit radiographic progression in RA or to further support their efficacy. As such this provides a rich source of information for planning future trials with radiographic damage as an outcome.

Fifty-two trials (7 trials used the GS method, 7 the Sharp method and 38 the SvdH method) conducted over 26 years have included a wide variety of RA patient populations. Over time, there has been no significant decline in the mean baseline radiographic score in the RCTs (online supplementary table 1 and figure 3). This result is different from the previous study by Rahman et al.6 They described a dramatic decrease in severity of RA patients who participated in the TNFi trials. There are several possible explanations on this discrepancy. First, the previous study included only 5 trials in MTX-experienced population. It used to ATTRACT trial28 conducted in 1999 as an anchor study that had the highest baseline radiographic score out of all the trials so far. If ATTRACT trial28 is used as the reference point, the scatter plot is likely to show a negative slope. Second, the actual severity may have decreased, but the clinical trials have adapted the inclusion criteria to select patients with a high propensity for progression. For example, there is a trend that recent trials included more RF or ACPA positive patients. Some trials even required the presence of bone erosions as an inclusion criterion: among 17 trials conducted since 2010, 64.7% of the trials had the mandatory presence of erosions, as compared with 28.6% of the studies prior to 2009. However, overall it is difficult to compare the true trend as data are obtained by different scoring methods and within the same scoring method by different readers. This may all result in variation of the scores, which may challenge the interpretation over time.

In clinical trials, missing values are inevitable. Because missing values can be a potential source of bias, various methods have been proposed to deal with this issue. LE has been the most widely used method in RA clinical trials. In this overview, 37 trials (71.2%), especially the older trials, employed LE methods. However, currently, the use of all available data in mixed models are the preferred method of analysis. For more detail, we refer to the literature.63

In conclusion, we summarised radiographic data from clinical trials used for the registration of drugs for the treatment of RA. This may serve as a repository for designing future clinical trials in RA with structural damage as an endpoint.

Key messages

What is already known about this subject?

  • Radiographic progression has been an important outcome assessment in rheumatoid arthritis randomised controlled trials (RCTs).

What does this study add?

  • This is a systematic literature review of the available published information on demographic features, radiographic scoring methods, statistical analyses and detailed radiographic data.

How might this impact on clinical practice?

  • This systematic literature review will help the design of RCTs with the radiographic inhibition claim of new drugs in the future.


The authors thank Jan Schoonese, librarian at the Leiden University Medical Center, for his advice.



  • Contributors All authors discussed and formulated the clinical questions and interpreted the results. YJP, AMG and DvdH collected the data, performed the analysis and wrote the manuscript. All authors read and critically reviewed the manuscript prior to submission.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests DvdH received consulting fees from AbbVie, Amgen, Astellas, AstraZeneca, BMS, Boehringer Ingelheim, Celgene, Cyxone, Daiichi, Eisai, Eli-Lilly, Galapagos, Gilead, Glaxo-Smith-Kline, Janssen, Merck, Novartis, Pfizer, Regeneron, Roche, Sanofi, Takeda, UCB Pharma and is Director of Imaging Rheumatology bv.

  • Patient consent Not required.

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

  • Data sharing statement Data sharing not applicable as no datasets generated and/or analysed for this study.