Objective To determine if early MRI measures predict X-ray progression at 1 and 2 years in a large RA trial cohort.
Design This study included 256 methotrexate (MTX)-naïve RA patients from a randomised placebo-controlled trial of golimumab (GO-BEFORE). MRIs of wrist and 2nd–5th metacarpophalangeal joints at 0, 12, 24, 52 and 104 weeks were obtained and scored using the RAMRIS system. Multivariable logistic regression examined if baseline and early change (weeks 12/24) in RAMRIS scores independently predicted progression of the van der Heijde-Sharp (vdHS) score and MRI erosion score at 1 and 2 years of follow-up.
Results High baseline score and poor improvement over the first 24 weeks in synovitis (p=0.003 and p=0.003, respectively) and in bone oedema (p=0.02 and p=0.001, respectively) were independent predictors of X-ray progression at 1 year. Associations were significant or tended towards an association at 2 years. An increase in RAMRIS bone erosion >0.5 at weeks 12 and 24 also predicted X-ray progression (p<0.003). Poor 12-week improvement in bone oedema was associated with X-ray and MRI progression at 1 year (p<0.05). Regression models that incorporated baseline and 12-week and 24-week changes in MRI measures of synovitis (AUC=0.71) and bone oedema (AUC=0.70) improved the prediction of X-ray progression at 1 year above clinical disease activity alone (AUC=0.66, p<0.04).
Conclusions Baseline and early changes in MRI measures independently predicted X-ray and MRI progression at later time-points. The predictive validity established here supports potential use in shorter-duration studies to determine efficacy of RA therapies in preventing structural damage.
- Rheumatoid Arthritis
- Magnetic Resonance Imaging
- Disease Activity
Statistics from Altmetric.com
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.
In addition to improvement in the signs and symptoms of disease, the prevention of progression of structural joint damage is essential for the treatment of rheumatoid arthritis (RA). Clinical trials in RA generally require a study duration of at least 6–12 months to demonstrate differences in radiographic progression. Valid, objective and early predictors of progression in structural joint damage are currently lacking. Such predictors would have significant implications in the design and implementation of clinical trials, and could eventually play an important role in clinical care through risk stratification and early assessment of treatment efficacy.
MRI can detect and quantify synovitis, bone oedema and bone erosion. An RA MRI scoring system (RAMRIS) has been developed to quantify these measures.1 Previous studies have shown that MRI measures are reliable and that they correlate with clinical, laboratory and radiographic measures.2 MRI can identify bone erosions years before they are evident on standard radiographs.3 ,4 Additionally, baseline RAMRIS bone oedema scores independently identify subjects who are likely to have progression of erosions on X-ray and MRI.5–8 No previous studies have evaluated whether baseline and early changes in each component of the RAMRIS score (synovitis, bone oedema, and bone erosion) predict progression of structural damage on standard X-ray independently of other available clinical measures of disease activity in a clinical trial setting.
The aim of this study was to evaluate whether baseline and early (at 12 and 24 weeks) measures of change in MRI synovitis, bone oedema, and bone erosion predicted subsequent progression in structural joint damage as measured by conventional radiographic outcomes at 1 and 2 years of follow-up in a large cohort from a recent randomised controlled trial, independent of baseline and longitudinal measures of clinical disease activity. Secondary analyses evaluated the independent predictive validity of 12-week and 24-week MRI measures of synovitis and bone oedema for MRI erosion progression at weeks 52 and 104.
This prospective cohort study is ancillary to the GO-BEFORE trial (Clintrials.gov identifier NCT00361335) that included 637 patients with RA with 52 weeks of follow-up. Methods and results of the original trial have been previously reported.9 ,10 The original study compared the efficacy of methotrexate or golimumab alone to combination therapy with methotrexate and golimumab in methotrexate, and biologic therapy-naïve subjects. A total of 318 subjects at eligible study sites (based on technical capabilities) participated in the MRI substudy.
Patients 18 years of age or older who met American College of Rheumatology (ACR) 1987 criteria for RA for at least the past 3 months and had active disease were recruited into the MRI substudy at participating sites. Patient visits occurred at regular 4-week intervals as part of the original trial. Data collection at each visit included independent, blinded assessments of disease activity using the DAS28 with C-reactive protein (CRP) (DAS28(CRP)).
Of the 318 participants, 291 had MRIs of adequate quality to be scored for bone oedema/erosions, and 272 were adequate to be scored for synovitis at baseline. At 24 weeks, 280 were scored for bone oedema and erosions, and 268 were scored for synovitis.
The trial was conducted according to the principles of the Declaration of Helsinki. As such, all patients provided written informed consent before participating in the study. The ancillary study was approved by the internal review board at the University of Pennsylvania.
MRIs of the dominant hand at baseline and weeks 12, 24, 52 and 104 were obtained at participating trial centres. MRIs of the patient's dominant wrist and 2nd–5th metacarpophalangeal (MCP) joints were obtained using 1.5 T MRI with contrast enhancement. The MR sequences were as follows: axial T1 fast spin echo (FSE) precontrast, coronal T1 FSE precontrast, coronal short tau inversion recovery (STIR) (or T2 fat-suppressed precontrast) and coronal T1 fat-suppressed postcontrast.
Images were scored by two, independent readers who were blinded to the image time-point or sequence (visit number), patient identity and treatment group. The average score of two readers was determined for synovitis (0–9 for wrist joint, 0–21 for wrist plus MCP joints), bone oedema (osteitis, 0–69) and bone erosions (0–230), using the RAMRIS system.1 Intrarater reliability (R) ranged from 0.72 to 0.95, and inter-rater reliability (R) ranged from 0.69 to 0.89. RAMRIS synovitis scores were not determined for subjects who did not receive gadolinium enhancement.
Radiographs of hands and feet
Radiographs of hands and feet were performed at baseline, week 52 and week 104. They were scored by two blinded readers using the van der Heijde–Sharp (vdHS) system. Changes from baseline in vdHS scores at 52 weeks and 104 weeks were determined using centralised readers and standardised methods, as previously described.11 ,12 Progression was defined as a change in vdHS score of >0.5. This cut-off was chosen for the oritinal trial to reduce misclassification error. The magnitude of the coefficients of risk were similar when using a change cutoff of >3 in vdHS (not shown).
Statistical analysis was performed using STATA V.11 software (StataCorp, College Station, Texas). MRI synovitis and bone oedema scores were evaluated as continuous variables (per unit difference or difference in change). Change in RAMRIS bone erosion scores was highly skewed, and was therefore dichotomised at a cut-off of >0.5. This cut-off was chosen to be consistent with the definition of progression in vdHS. Again, these differences in means were assessed using Student t test or the Wilcoxon rank sum test, as appropriate. Group differences in categorical variables were assessed using the χ2 test. Subjects lost to follow-up were not included in longitudinal analyses. Since the change in MRI scores (synovitis and bone oedema) were dependent on baseline score, these scores were adjusted for baseline scores within group comparisons. Subjects with missing MRI data at early time-points and subjects with missing X-ray data at later time-points were excluded from analyses.
Parsimonious multivariable logistic regression models were built to evaluate if baseline and early measures of change in component RAMRIS scores (at 12 and 24 weeks) predicted subsequent X-ray progression independent of clinical disease activity (DAS28(CRP)), change in the DAS28 (CRP) over the same time period, age, sex, baseline vdHS score and treatment group. In regression models, baseline MRI erosion scores were log-adjusted given their skewed distribution.
To visualise associations, the relationship between early changes in synovitis/oedema scores with the predicted probability of X-ray progression was visualised while controlling baseline synovitis/oedema scores and other covariates in the logistic regression models. The Area Under the Curve (AUC) of the Receiver Operating Characteristics (ROC) curve for each regression model was determined as a test of the discrimination characteristics of the model. Multivariable logistic regression models evaluated independent associations between baseline and 12/24-week change in component MRI scores (synovitis and bone oedema) with progression in MRI erosion scores adjusting for the variables outlined above (table 1).
MRI findings in radiographic progressors and non-progressors
Baseline synovitis scores were higher among subjects who progressed on X-ray at 52 weeks and 104 weeks (table 2). After adjusting for baseline synovitis score, there was significantly less improvement in synovitis at 12 and 24 weeks in subjects who went on to progress on X-ray at 52 and 104 weeks. Baseline bone oedema scores were higher among subjects who went on to progress on X-ray at 52 weeks and 104 weeks (table 2). After adjusting for baseline bone oedema scores, there was significantly less improvement in bone oedema scores at 12 and 24 weeks in subjects who went on to progress on X-ray at 52 and 104 weeks.
Median bone erosion scores were higher among subjects who went on to progress at the later time points. Additionally, compared with those who did not progress on X-ray, a greater percentage of subjects who went on to progress on X-ray at 52 and 104 weeks had progressed on MRI at early time points (table 2). For example, 37.1% of subjects who progressed on X-ray at 52 weeks had progressed on MRI at 24 weeks, while only 16.5% of subjects who did not progress on X-ray at 52 weeks had progressed on MRI at 24 weeks. Similarly, among subjects who progressed on X-ray at 24 weeks, 31.6% had also progressed on MRI at 12 weeks (not shown). By contrast, among non-progressors on X-ray at 24 weeks, 16.7% progressed on MRI at 12 weeks (all p<0.03).
Independent predictive value of early MRI findings for subsequent radiographic progression
In multivariable regression models (adjusting for age, sex, treatment group, baseline vdHS score, DAS28(CRP) at baseline and 24-week change in DAS28(CRP)), baseline synovitis score and 24-week change in synovitis were independently associated with an increased risk of radiographic progression at 52 and 104 weeks of follow-up (Model 4, table 3).
Similarly, baseline bone oedema and 24-week change in bone oedema were independently associated with an increased risk of radiographic progression at 52 weeks (Model 5, table 3). Baseline bone oedema score was also associated with risk of progression at week 104, while the association with change in bone oedema at week 24 was marginal (p=0.07). Progression of RAMRIS erosion score (>0.5) at 24 weeks was also independently associated with an increased risk of X-ray progression at both later time-points (Model 6, table 3). The inclusion of CCP status in multivariable regression models did not significantly alter the results or conclusions (not shown).
The AUCs were similar between models 4–6 (no significant differences). Furthermore, in large models that included all MRI component scores, there was no single overarching independent predictor among the MRI component scores (not shown). This large model (inclusive of all component scores, including 24-week changes) had an AUC of 0.74 for 52-week progression, not different from models incorporating synovitis or bone oedema alone (p >0.05).
Table 3 (Models 1–3) also demonstrates the independent associations of 12-week changes in RAMRIS component scores and risk of radiographic progression at 52 and 104 weeks. Baseline bone oedema, and the change in bone oedema at 12 weeks, were significant predictors of X-ray progression at week 52. MRI progression >0.5 at 12 weeks was also significantly and independently associated with an increased risk of X-ray progression at 52 and 104 weeks. The baseline synovitis score and change in synovitis at 12 weeks tended towards an association. The AUCs and OR estimates for 12-week models tended to be similar to 24-week models suggesting that the use of 12-week MRI measures had comparable discriminative characteristics to the 24-week measures. Absolute synovitis and bone oedema scores at 12 weeks were also associated with X-ray progression independent of the 12-week absolute DAS28 (CRP) score (p=0.05 and 0.03, respectively).
Figure 1A and B illustrate the association between greater 24-week improvement in RAMRIS synovitis and bone oedema scores and a lower predicted probability of progression at 52 weeks as determined from the 24-week regression models. For example, for two similar subjects, both with a baseline synovitis score of 10, the subject with an improvement in synovitis of 5 points at 24 weeks would have a predicted probability of progression of 0.17. A subject with a change in synovitis of 0 would, in contrast, have a predicted probability of progression of 0.32. Similarly, for two similar subjects, both with a baseline bone oedema score of 6.5, the subject with an improvement in bone oedema of 5 points at 24 weeks would have a predicted probability of progression of 0.15. A subject with a change in bone oedema of 0 would, in contrast, have a predicted probability of progression of 0.24. The predicted probability of progression for two similar subjects was 0.20 for subjects without, and 0.36 for subjects with, progression in RAMRIS erosion score on MRI at week 24.
Figure 2 illustrates the improved test characteristics (sensitivity and specificity) for the logistic regression models that incorporated MRI measures of synovitis or bone oedema (including 24-week change). The AUCs were significantly greater for the models incorporating synovitis and bone oedema when compared with basic models without these measures (p<0.05). Incorporation of bone erosion scores tended to improve the AUC compared with the model incorporating clinical disease activity alone (p=0.06, not shown). A model incorporating 12-week bone oedema scores (Model 2, table 2) also had significantly greater AUC (for prediction of 52-week X-ray progression) than the model with disease activity alone (AUC 0.70 v. 0.66 (p=0.02)). The 12-week synovitis and bone erosion models (Models 1 and 3, table 2) had statistically similar AUCs to clinical disease activity alone (p=0.2 and 0.09, respectively).
Predictive value of early MRI findings for subsequent erosive progression on MRI
Table 4 demonstrates the independent associations between component MRI scores (synovitis and bone oedema) and progression of erosion score on MRI at 52 and 104 weeks. Baseline and 12/24-week change in bone oedema scores were independently associated with the risk of MRI progression at week 52. The change in bone oedema score at both 12 and 24 weeks was the only independent predictor of MRI progression at week 104. The AUC for a model (not shown) incorporating measures of bone oedema had a superior AUC to disease activity alone for prediction of MRI erosion progression (0.84 v. 0.73, p=0.003).
This is the first study to evaluate MRI measures as predictors of X-ray progression in a large clinical trial. These data demonstrate the utility of MRI measures as early independent predictors of progression in structural joint damage on X-ray. Baseline synovitis and bone oedema scores, as well as early change scores, were significantly associated with radiographic progression even after adjusting for available clinical measures of disease activity.
The finding that baseline measures of synovitis and bone oedema are predictive of radiographic progression independent of clinical disease activity confirms previous reports by Hetland et al and extends them into the clinical trial setting.6 ,13 ,14 The observations from this study also suggest that baseline MRI scores could be used to identify subjects at high risk for structural joint damage progression. This has implications for clinical trial design since the enrichment of trials with high-risk subjects improves the ability to detect differences in treatments. In the future, MRI could potentially be used in a clinical setting to identify subjects at greatest risk and help stratify patients for appropriate use of more aggressive therapies.
Second, the finding that early changes in MRI scores independently associate with risk of radiographic progression on X-ray is novel and important. This observation suggests that these measures could be used early in the course of therapy to identify patients with meaningful improvements in inflammatory disease burden. This has powerful implications for clinical trials where early identification of effective therapies may save valuable time and resources. Indeed, we have previously shown that the use of MRI can identify differences in efficacy of therapy with approximately half the subjects in about half the follow-up time.10
The similar AUC for the regression models incorporating MRI component scores at 24 weeks suggests that the component scores have similar ability to discriminate between those who are likely and unlikely to progress on radiographs. In other words, synovitis, bone oedema and MRI erosion were similarly predictive. However, the change in bone oedema at early time-points was the strongest predictor of MRI erosion progression at both 52 and 104 weeks. The 12-week changes in bone oedema were also highly associated with 52-week X-ray progression, and improved the predictive validity of the model. These data suggest strongly that a therapeutic agent that reduces bone oedema to low levels by 12 weeks is very likely to prevent progression of structural damage on X-ray and MRI at later time points independent of the measureable effect of the drug on clinical disease activity alone.
Looking ahead, MRI could be potentially used to identify those patients who require escalation of therapy in a clinical setting as well. The probability of progression rises with greater baseline scores and less improvement over time, even after adjusting for clinical disease activity. While the elevated odds per unit change appear small in this study, we have shown that over the range of different MRI scores observed in this study, the probability of progression differs by twofold to fourfold. This provides evidence that this test could meaningfully affect a physician's pretest assessment of risk.
There are several limitations to this work. First, though highly applicable to clinical trial design, these results from a clinical trial may not be generalisable to routine clinical use, nor has the cost-efficacy been determined for this purpose. Second, there was some missing data as the result of dropout during the trial (approximately 10% at 52 weeks), as well as missing data for synovitis scores due to lack of contrast enhancement in some subjects (approximately 10%). Clinical characteristics were similar among those who did not have synovitis scores performed, and we consider it very likely that this data was completely missing at random, and did not introduce significant bias into the analyses. Third, it should be noted that synovitis was scored from coronal images, and more information would have been available from having additional axial images. Finally, statistical adjustment for multiple comparisons was not performed due to the inter-relatedness of the outcomes. We may also have underestimated strength of associations by intentionally comparing MRI of the dominant hand to X-ray progression in both hands and feet and by using the total vdHS score instead of the component erosion score.
In conclusion, the use of baseline and longitudinal MRI measures of synovitis, bone oedema, and bone erosion can enhance the prediction of subsequent X-ray and MRI progression at later time points. This has implications for the use of MRI in clinical trial design and could potentially impact future clinical practice.
The authors would like to acknowledge the GO-BEFORE study investigators.
Handling editor Tore K Kvien
Contributors All authors contributed to conception and design of the ancillary study or played a direct role in the analysis and interpretation of data. All authors also contributed to the drafting of the article or revising it for content. All authors approved the final version of the manuscript.
Funding Funding for the original GO-BEFORE trial was supported by Janssen Research and Development. Janssen did not provide additional funding for the ancillary study.
Competing interests Joshua Baker has no conflicts to disclose. Daniel Baker, Elizabeth Hsia and Jiandong Lu are employees of Janssen Research and Development. Dr Conaghan has done speakers bureaus or consultancies for BMS, Janssen, Merck, Pfizer and Roche. Dr Østergaard has received fees for consultancy or speaker fees and/or research support from Abbott, BMS, Centocor, GSK, Janssen, Merck, Mundipharma, Novo, Pfizer, Schering-Plough, Roche UCB and Wyeth. Dr Emery has done speakers bureaus or consultancies from Pfizer, MSD, Roche, Abbott, UCB and BMS.
Patient consent Obtained.
Ethics approval Approval was obtained at each participating centre for the original trial.
Provenance and peer review Not commissioned; externally peer reviewed.