Development and Validation of Modified Disease Activity Scores in Rheumatoid Arthritis: Superior Correlation With Magnetic Resonance Imaging–Detected Synovitis and Radiographic Progression†
Joshua F. Baker,
University of Pennsylvania and Philadelphia VA Medical Center, Philadelphia
Division of Rheumatology, Department of Medicine, University of Pennsylvania, 8 Penn Tower Building, 34th Street and Civic Center Boulevard, Hospital of the University of Pennsylvania, Philadelphia, PA 19104. E-mail: firstname.lastname@example.org
University of Leeds, NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds Institute of Rheumatic and Musculoskeletal Medicine, Leeds Teaching Hospitals NHS Trust, and Chapel Allerton Hospital, Leeds, UK
Dr. Østergaard has received consulting fees and/or honoraria from Abbott/Abbvie, Bristol-Myers Squibb, Centocor, Janssen, GlaxoSmithKline, Merck, Mundipharma, Novo Nordisk, Pfizer, Schering-Plough, Roche, UCB, and Wyeth (less than $10,000 each) and research support from Abbott/Abbvie, Centocor, Merck, and Pfizer.
This study is a secondary analysis of the GO-BEFORE (ClinicalTrials.gov identifier NCT00361335) and GO-FORWARD (NCT00264550) randomized clinical trials. Funding for the original trials was provided by Janssen Biotech, Inc. There was no additional funding for the secondary study.
To develop and validate composite disease activity scores, based on widely available clinical measures, that would demonstrate improved correlation with detection of synovitis on magnetic resonance imaging (MRI) and radiographic progression, in comparison with conventional measures, in patients with rheumatoid arthritis (RA).
This study was conducted as a secondary study of 2 RA clinical trials, GO-BEFORE (development cohort) and GO-FORWARD (validation cohort). Generalized estimating equations were used to evaluate independent cross-sectional associations of component variables (from all time points) with concurrent MRI measures of synovitis and bone edema in the development cohort. Based on regression coefficients, modified versions of the Disease Activity Score in 28 joints (M-DAS28), Simplified Disease Activity Index (M-SDAI), and Clinical Disease Activity Index (M-CDAI) were generated for each subject in the validation cohort. The M-DAS28, M-SDAI, and M-CDAI scores were compared to conventional scores of disease activity with regard to associations with MRI measures of synovitis and radiographic progression, assessed using Pearson's and Spearman's correlations, linear/logistic regression, and receiver operating characteristic analysis.
Four variables were independently associated with MRI-detected synovitis and bone edema in the development cohort: C-reactive protein (CRP) level, erythrocyte sedimentation rate (ESR), swollen joint count in 28 joints (SJC28), and evaluator's global assessment of disease activity using a visual analog scale (EvGA score). Modified disease activity scores were generated using the regression coefficients obtained in the synovitis models for all subjects in the validation cohort; modified scores were calculated as M-DAS28 = 0.49 × ln(CRP) + 0.15 × SJC28 + 0.22 × EvGA + 1 and M-SDAI = CRP + SJC28 + EvGA. Both modified and conventional disease activity scores correlated significantly with MRI measures of synovitis. Modified scores showed superior correlation with synovitis, as compared to conventional scores, at all time points (P < 0.05). Furthermore, the M-DAS28 and M-SDAI had superior test characteristics for prediction of radiographic progression at 52 weeks (both P < 0.05).
Modified disease activity scores demonstrated superior correlation with MRI detection of synovitis at all time points, and more accurately predicted radiographic progression in patients with RA in a clinical trial setting.
A number of composite scores have been developed for rheumatoid arthritis (RA). Given that no gold standard measure of joint inflammation exists, clinical disease activity measures have hitherto been developed by correlating clinical variables with other surrogate measures, such as physician decision-making and radiographic progression ([1-3]). The available literature supports the use of a number of composite measures, comprising several variables for the assessment of disease activity in RA, with the goal of successfully treating to a target (). Widely used measures include the Disease Activity Score in 28 joints (DAS28) and the Simplified Disease Activity Index (SDAI) ([1, 5]).
No previous studies have utilized magnetic resonance imaging (MRI) to aid in the development of a composite score. MRI detection of abnormalities appears to accurately depict synovitis, and findings of synovitis on MRI are correlated with histologic evidence of inflammation (). Therefore, a tool using MRI as a gold standard might more closely reflect the inflammatory disease burden, as measured by that imaging modality, than would indices developed in other ways. Correctly identifying and quantifying the burden of inflammation is important, since disease activity scores are primarily utilized to make treatment decisions with regard to immunomodulatory or antiinflammatory therapy. A score that correlates with the inflammatory disease burden in the joint would also be likely to correlate better with long-term outcomes, such as radiographic progression.
We therefore aimed to develop a novel disease activity score, based on widely available clinical measures, that would 1) show good correlation with the RA MRI Scoring (RAMRIS) system for synovitis, by assessing MRI features of synovitis in patients with RA in a clinical trial setting at all time points, and 2) improve prediction of radiographic progression of joint damage, by assessing patients with RA from baseline to 52 weeks of followup. With the background of these 2 widely used composite measures of disease activity (MRI-detected synovitis and radiographic progression), we developed modified versions of the composite disease activity scores using data from the GO-BEFORE clinical trial, and validated these scores using study data from the GO-FORWARD clinical trial ([7-9]).
PATIENTS AND METHODS
This study is a secondary analysis of the GO-BEFORE and GO-FORWARD randomized clinical trials. Both were multicenter, double-blind, placebo-controlled trials that evaluated the efficacy of golimumab, a fully human monoclonal antibody to tumor necrosis factor α, for the treatment of RA. The trial results have been previously published ([8-10]). Both studies evaluated the effect of golimumab in combination with methotrexate (MTX) as compared to MTX monotherapy and golimumab monotherapy. The GO-BEFORE study was performed in 637 MTX-treated patients who had not received treatment with biologic agents, of whom 297 had MRIs scored for synovitis, bone edema, and/or bone erosion. The GO-FORWARD study included 444 patients who had previously experienced an inadequate response to MTX, of whom 217 had MRIs that were scored for the same 3 measures as those in the GO-BEFORE study; 31 of these patients did not have adequate imaging for the quantification of synovitis at the initial visit, primarily due to a lack of gadolinium administration.
Disease activity was measured in a standard manner. Standard disease activity composite scores were determined for each subject at 0, 24, and 52 weeks. These included the DAS28 using the C-reactive protein (CRP) level (DAS28-CRP), the SDAI, and the Clinical Disease Activity Index (CDAI) ().
The trials were conducted according to the principles of the Declaration of Helsinki. As such, all patients provided written informed consent before participating in the study.
Magnetic resonance imaging
MRIs of the dominant hand at baseline and at weeks 24, 52, and 104 were obtained at participating trial centers for both clinical trials. MRIs of the patient's dominant wrist and the second through the fifth metacarpophalangeal (MCP) joints of the dominant hand were obtained using 1.5T MRI with contrast enhancement. The MR sequences were as follows: axial T1-weighted fast spin-echo (FSE) precontrast image, coronal T1-weighted FSE precontrast image, coronal short tau inversion recovery image (or T2 fat-suppressed precontrast image), and coronal T1-weighted fat-suppressed postcontrast image.
Images were scored by 2 independent readers who were blinded with regard to image time point or image sequence, patient identity, and treatment group. The 2 readers scored the MRIs for synovitis (score range 0–9 for the wrist joint, score range 0–21 for the wrist plus MCP joints), bone edema/osteitis (score range 0–69), and bone erosions (score range 0–230), using the RAMRIS system ().
Radiographs of the hands and feet
Radiographic assessments of the hands and feet were performed at baseline and at week 52. The radiographs were scored for damage progression by 2 blinded readers, using the Sharp/van der Heijde (SHS) scoring system (). Changes from baseline in the SHS scores at 52 weeks were determined by centralized readers using standardized methods, as previously described (). Progression was defined as a change in the SHS score of >0.5, as it was in the original trial.
Statistical analysis was performed using Stata software version 11 (StataCorp). Skewed data were log-transformed to fit a normal distribution, before inclusion in linear regression models.
Development of a novel disease activity assessment tool using data from the GO-BEFORE trial
Within the GO-BEFORE clinical trial, we utilized robust generalized estimating equations (GEEs) in linear regression models to simultaneously evaluate independent cross-sectional associations of the explanatory variables from all time points (0, 24, 52, and 104 weeks) with the MRI synovitis score and MRI bone edema score from the same time points, in all subjects with available data (all treatment groups). Initially included in these models were all commonly utilized component variables from standard composite scores. Specifically, we included the linear log-transformed CRP (ln[CRP]) level (in mg/dl), linear log-transformed erythrocyte sedimentation rate (ln[ESR]) (in mm/hour), swollen joint count in 28 joints (SJC28), tender joint count in 28 joints (TJC28), patient's global assessment of disease activity using a visual analog scale (VAS; range 0–10) (represented as PtGA score in the models), and evaluator's global assessment of disease activity using a VAS (range 0–10) (represented as EvGA score in the models). Evaluator's global assessment was performed by the physician-investigator, who had full access to patients' clinical information and data on inflammation markers and formal joint counts.
By weighting the coefficients from the variables significantly associated with synovitis in the GEE model, we developed a formula for a novel, modified disease activity score (M-DAS28). To simplify this formula, we created a simplified summation formula (M-SDAI) by approximating the weighting of the significant untransformed coefficients. Finally, by approximating the weighting of the significant coefficients from the GEE regression model after excluding the CRP, we developed a modified CDAI score (M-CDAI).
Validation of the modified scores using data from the GO-FORWARD study
The M-DAS28, M-SDAI, and M-CDAI were calculated for each subject at all time points within the validation cohort (GO-FORWARD study), using the formulae developed in the GO-BEFORE trial. The formulae are presented below.
Pearson's correlation coefficients were used to compare the correlation of the M-DAS28 or DAS28-CRP, the M-SDAI or SDAI, and the M-CDAI or CDAI with the MRI synovitis score at 0, 24, and 52 weeks of followup in the trial. The “corcor” command in Stata statistical software was utilized to compare correlation coefficients. Spearman's correlation coefficients were used for assessing the correlation between composite disease activity scores and detection of bone edema on MRI at the same time points; Spearman's correlations were utilized for this because of the non-normal distribution of the bone edema scores. In univariate and multivariable logistic regression models, we compared the predictive value of 1) the baseline composite scores (M-DAS28, M-SDAI, DAS28-CRP, and SDAI), 2) 24-week composite scores, and 3) models incorporating both baseline and 24-week changes in composite scores.
Line graphs were generated to compare the predicted probability of radiographic progression at 52 weeks per quintile of baseline composite scores. Receiver operating characteristic analysis was used to evaluate the area under the curve (AUC) for regression models incorporating the M-DAS28 and M-SDAI, compared to the DAS28-CRP and SDAI, for prediction of subsequent radiographic progression at week 52.
The baseline disease characteristics of the patients in the development and validation cohorts (GO-BEFORE and GO-FORWARD) are shown in Table 1.
Table 1. Baseline disease characteristics of the subjects in the GO-BEFORE and GO-FORWARD cohorts*
GO-BEFORE development cohort (n = 633)
GO-FORWARD validation cohort (n = 444)
GO-BEFORE development cohort (n = 270)
GO-FORWARD validation cohort (n = 186)
Except where indicated otherwise, values are the mean ± SD. TJC28 = tender joint count in 28 joints; SJC28 = swollen joint count in 28 joints; VAS = visual analog scale; HAQ = Health Assessment Questionnaire; ESR = erythrocyte sedimentation rate; DAS28-CRP = Disease Activity Score in 28 joints using C-reactive protein (CRP) level; SDAI = Simplified Disease Activity Index; CDAI = Clinical Disease Activity Index; SHS = Sharp/van der Heijde; NA = not applicable.
aValues represent the scores for synovitis and bone edema as detected by magnetic resonance imaging (MRI) of the dominant hand, using the Rheumatoid Arthritis MRI scoring system.
Generation of the M-DAS28, M-SDAI, and M-CDAI in the development study (GO-BEFORE).
GEE analysis was performed on the data from all subjects in the development cohort (GO-BEFORE study), and component measures that were independently associated with the MRI synovitis score and MRI bone edema score were identified (Table 2). For both outcomes, the independent predictors were the ln(CRP), ln(ESR), EvGA score, and SJC28. The TJC28 and the PtGA score were not independently associated with synovitis scores. A novel score, the M-DAS28, was thus developed based on the regression coefficients for MRI-detected synovitis (Table 2). Despite the fact that both CRP and ESR were significantly and independently associated with synovitis, only the CRP was included in the M-DAS28, in order to avoid overrepresentation of the acute-phase domain in the composite score (). A modified DAS28 using the ESR (M-DAS28-ESR) was also developed separately, in which the ESR was used to replace CRP as the acute-phase reactant. Adjustment for treatment group did not significantly affect the coefficients in these analyses.
Table 2. Regression analysis evaluating independent associations between component disease activity scores and MRI measures of synovitis and bone edema at all time points in the GO-BEFORE study*
β (95% CI)
β (95% CI)
Associations between component scores and magnetic resonance imaging (MRI) measures of synovitis and bone edema were assessed at 0, 24, 52, and 104 weeks in the GO-BEFORE study, utilizing generalized estimating equations. Values are beta coefficients (95% confidence interval [95% CI]) in models that included all component variables (inclusive) or models in which some component variables were omitted (reduced). ln(CRP) = linear log-transformed C-reactive protein; ln(ESR) = linear log-transformed erythrocyte sedimentation rate; SJC28 = swollen joint count in 28 joints; TJC28 = tender joint count in 28 joints; PtGA = patient's global assessment of disease activity; EvGA = evaluator's global assessment of disease activity.
Synovitis (n = 1,085)
0.50 (0.28, 0.72)
0.52 (0.29, 0.73)
0.20 (0.034, 0.36)
0.15 (0.096, 0.20)
0.15 (0.11, 0.20)
0.0061 (−0.023, 0.035)
0.031 (−0.040, 0.10)
0.20 (0.12, 0.28)
0.22 (0.15, 0.29)
Bone edema (n = 1,085)
0.49 (0.25, 0.73)
0.50 (0.26, 0.74)
0.36 (0.099, 0.62)
0.35 (0.087, 0.61)
0.069 (0.027, 0.11)
0.057 (0.017, 0.097)
−0.018 (−0.057, 0.021)
0.034 (−0.045, 0.11)
0.15 (0.058, 0.25)
0.16 (0.062, 0.26)
For generation of the M-SDAI, a simple summation formula was created on the basis of comparable coefficients for the component measures in a GEE model: for CRP (per 1 mg/dl), β = 0.22 (95% confidence interval [95% CI] 0.12, 0.33; P < 0.001); for SJC28 (per 1 joint), β = 0.17 (95% CI 0.12, 0.21; P < 0.001); and for EvGA (per 1-cm change in VAS score), β = 0.26 (95% CI 0.18, 0.33; P < 0.001). In addition, a modified version of the CDAI score (M-CDAI) was developed by weighting higher on the EvGA score (per 1-cm change in VAS score) (β = 0.31, 95% CI 0.23, 0.39; P < 0.001) than on the SJC28 (per 1 joint) (β = 0.12, 95% CI 0.089, 0.17; P < 0.001). All score calculations (modified and original/standard) were made using the following formulas: M-DAS28 = 0.49 × ln(CRP) + 0.15 × SJC28 + 0.22 × EvGA + 1 and DAS28-CRP = 0.36 × ln(CRP + 1) + 0.28 × SJC28 + 0.56 × TJC28 + 0.14 × PtGa; M-DAS28-ESR = 0.40 × ln(ESR) + 0.17 × SJC28 + 0.26 × EvGA and DAS28-ESR = 0.70 × ln(ESR) + 0.28 × SJC28 + 0.56 × TJC28 + 0.14 × PtGA; M-SDAI = CRP + SJC28 + EvGA and SDAI = CRP + SJC28 + TJC28 + EvGA + PtGA; M-CDAI = SJC28 + 2 × EvGA and CDAI = SJC28 + TJC28 + EvGA + PtGA.
Validation study (GO-FORWARD).
Correlations with MRI-detected synovitis
The DAS28, SDAI, and CDAI scores all correlated significantly with the detection of synovitis on MRI, suggesting a relationship between clinical measures of disease activity and imaging measures of disease activity. However, compared to the original indices and scores, the correlation coefficients for the M-DAS28, M-SDAI, and M-CDAI composite scores demonstrated superior correlation with MRI measures of synovitis at all time points in the GO-FORWARD study (Table 3). For example, the baseline M-DAS28 correlated with the baseline synovitis score at a Pearson's R value of 0.32 (P < 0.001), while the baseline DAS28-CRP correlated at a Pearson's R value of 0.21 (P = 0.004) (P = 0.009 for the comparison). Similarly, the M-DAS28-ESR, M-SDAI, and M-CDAI all showed significantly higher correlations with the synovitis scores when compared to the original DAS28-ESR, SDAI, and CDAI, respectively (all P < 0.05) (Table 3).
Table 3. Associations between disease activity scores and MRI measures of synovitis and bone edema in the GO-FORWARD study*
Correlations between composite or component scores (per 1-unit difference) and magnetic resonance imaging (MRI) detection of synovitis or bone edema at 0, 24, and 52 weeks were assessed as Pearson's correlation coefficents (r values) for synovitis and Spearman's correlation coefficents (rho values) for bone edema. Sample sizes were as follows: for composite scores, n = 186 at baseline, n = 155 at 24 weeks, and n = 186 at 52 weeks; for component scores, n = 217 at baseline, n = 172 at 24 weeks, and n = 207 at 52 weeks. M-DAS28 = modified Disease Activity Score in 28 joints; DAS28-CRP = DAS28 using C-reactive protein level; M-DAS28-ESR = modified DAS28 using erythrocyte sedimentation rate; M-SDAI = modified Simplified Disease Activity Index; M-CDAI = modified Clinical Disease Activity Index; ln(CRP) = linear log-transformed CRP; ln(ESR) = linear log-transformed ESR; SJC28 = swollen joint count in 28 joints; TJC28 = tender joint count in 28 joints; EvGA = evaluator's global assessment of disease activity; PtGA = patient's global assessment of disease activity.
In addition, we assessed the correlations of individual component scores with MRI-detected synovitis (Table 3). In these analyses, the ln(CRP) and SJC28 demonstrated the closest correlations with synovitis among the component scores. The SJC28 showed a correlation that was comparable to that of the M-DAS28 at all time points.
With regard to bone edema, there were no significant associations between the standard clinical disease activity indices and MRI detection of bone edema at any time point in the GO-FORWARD study (Table 3). There was a trend toward a correlation between the M-DAS28 and bone edema score at baseline (Spearman's rho = 0.12, P = 0.07) and at 52 weeks (Spearman's rho = 0.13, P = 0.06). In addition, a low, but significant, association was observed between the M-SDAI and bone edema score at 52 weeks (Spearman's rho = 0.16, P = 0.02).
Prediction of radiographic progression
In the GO-FORWARD cohort, the baseline M-DAS28 score correlated with the risk of radiographic progression of joint damage at 52 weeks. The AUC was 0.62 for the baseline M-DAS28, compared to an AUC of 0.57 for the baseline DAS28-CRP (P = 0.04) (Table 4). The AUC for a model incorporating both the baseline score and the 24-week change in M-DAS28 was 0.67, compared to an AUC of 0.62 for models incorporating baseline and 24-week change in the DAS28-CRP (P = 0.02) (Figure 1, top). The baseline and 24- week M-DAS28-ESR did not improve prediction of radiographic progression (Table 4). Models incorporating baseline and 24-week change in the M-DAS28-ESR showed discriminative characteristics similar to those of the DAS28-ESR (AUC 0.65 versus 0.62; P = 0.1).
Table 4. Independent associations between disease activity scores at baseline and 24 weeks and radiographic progression of joint damage at 52 weeks in the GO-FORWARD study*
Radiographic progression at 52 weeks
Odds ratio (95% CI)
P for AUC
Associations between composite or component scores (per 1-unit difference) and radiographic progression in univariate models were assessed as the odds ratio (95% confidence interval [95% CI]) and area under the curve (AUC). M-DAS28 = modified Disease Activity Score in 28 joints; DAS28-CRP = DAS28 using C-reactive protein level; M-DAS28-ESR = modified DAS28 using erythrocyte sedimentation rate; M-SDAI = modified Simplified Disease Activity Index; M-CDAI = modified Clinical Disease Activity Index; SJC28 = swollen joint count in 28 joints; TJC28 = tender joint count in 28 joints; EvGA = evaluator's global assessment of disease activity; PtGA = patient's global assessment of disease activity.
Baseline (n = 385)
1.41 (1.17, 1.68)
1.36 (1.07, 1.72)
1.32 (1.09, 1.58)
1.27 (1.01, 1.59)
1.05 (1.02, 1.09)
1.02 (1.00, 1.04)
1.04 (1.01, 1.07)
1.01 (1.00, 1.03)
1.23 (1.11, 1.37)
1.04 (1.00, 1.08)
1.00 (0.97, 1.04)
1.16 (1.01, 1.34)
1.10 (0.99, 1.21)
24 weeks (n = 384)
1.65 (1.36, 2.00)
1.39 (1.16, 1.65)
1.49 (1.25, 1.77)
1.34 (1.14, 1.57)
1.08 (1.04, 1.11)
1.03 (1.01, 1.05)
1.06 (1.03, 1.09)
1.03 (1.01, 1.04)
The AUC values at baseline were 0.60 for the M-SDAI and 0.56 for the original SDAI (P = 0.01) (Table 4). The AUC for a model incorporating baseline and 24-week change in the M-SDAI was 0.65, compared to an AUC of 0.62 for models incorporating baseline and 24-week change in the SDAI (P = 0.04) (Figure 1, bottom). The greater AUC for the modified scores demonstrates that models based on baseline and 24-week changes in the M-DAS28 and M-SDAI have significantly better ability to discriminate between subsequent radiographic progressors and radiographic nonprogressors. Of the scores' individual components, only the CRP showed discriminative characteristics comparable to those of the composite scores, including the modified scores. Adjustment for treatment group did not affect these associations.
The line graphs in Figure 2 illustrate the predicted probability of radiographic progression of joint damage in each quintile of baseline disease activity, as measured by the M-DAS28 and original DAS28-CRP (top), and the M-SDAI and original SDAI (bottom). The findings demonstrate that, over the range of possible disease activity scores, the modified scores more accurately identified those with highest and lowest probability of progression.
In this study, we developed novel disease activity scores, based on readily available and widely used component scores, by determining associations with synovitis scores as measured by MRI. We validated these scores in a second study, and showed that these scores demonstrate 1) superior correlation with MRI measures of synovitis, and 2) superior prediction of radiographic progression. Interestingly, while swollen joint counts correlated well with MRI synovitis scores, they were not predictive of radiographic progression. In contrast, while the CRP level correlated relatively poorly with synovitis over the 3 time points, it was predictive of radiographic progression. This helps to explain why the M-DAS28 and M-SDAI (which are composed of both the SJC28 and CRP) were predictive of both synovitis and radiographic progression.
The DAS28-CRP and SDAI correlated significantly with the MRI synovitis score and with radiographic progression, demonstrating the relationship between clinical activity and imaging activity and progression. The correlation and predictive value were greater when the M-DAS28 or M-SDAI was used. Thus, the association between clinical and imaging activity and between clinical activity and subsequent radiographic progression can be augmented by the proposed modifications to the scores.
While there was a trend toward better correlation with bone edema when the modified scores were used, there was relatively poor correlation overall between clinical activity measures and this outcome. This is likely attributable to the low bone edema scores overall in this cohort, which is, in part, due to the relatively high burden of erosive structural damage in the GO-FORWARD cohort. Greater erosive damage may interfere with the accurate assessment of inflammatory osteitis, since eroded joints are more difficult to evaluate for concomitant osteitis.
The major characteristic of the M-DAS28, M-SDAI, and M-CDAI that distinguishes these measures from the other available composite measures of disease activity is the exclusion of the TJC28 and the patient's global assessment. These findings speak to the subjectivity of the TJC28 and the patient's global assessment, each of which may often be high in subjects with relatively low levels of inflammatory disease. Results from previous studies have suggested that subjective measures may affect the accuracy of the DAS28-CRP in identifying inflammatory disease ().
The observations presented reveal that the M-DAS28 and M-SDAI (when compared to use of the DAS28-CRP and SDAI) would be more likely to correctly identify subjects who would be most likely to benefit significantly from escalation of disease-modifying therapy. Thus, use of this score in practice might decrease the risk of disease progression in patients with RA, and thereby reduce the cost to the healthcare system. The optimal cutoff level to define “low disease activity” or “remission” in these novel measures still needs to be defined.
Importantly, while MRI scoring was used to derive the modified indices, the calculation of these indices was solely based on routine clinical and laboratory measures, and the good performance observed during their derivation in the development study was validated in a second trial setting. Moreover, the data suggest that these modified scores can be used in early as well as established RA.
Limitations to this study include the lack of an agreed gold standard measure of joint inflammation. Whether synovitis on MRI is the best reflection of inflammatory disease burden remains unknown. Moreover, since MRI is performed only on the dominant hand, it may not adequately characterize the full burden of disease. However, we have previously shown that the synovitis score and bone edema score, as detected on MRI of the dominant hand, are highly and independently predictive of radiographic progression in the joints of the hands and feet (), suggesting that these scores are likely to be an important surrogate measure of overall joint inflammation. Furthermore, the validity of the M-DAS28 was confirmed in the demonstration of its ability to predict radiographic progression of joint damage.
A second limitation is that correlations with radiographic progression may not be generalizable to higher-risk cohorts, given that few of the subjects in the GO-FORWARD study experienced significant radiographic progression (median score change 0). Nevertheless, the M-DAS28 might be more predictive in other settings, within cohorts that may have greater subject heterogeneity and greater overall progression. Evaluation and validation in other cohorts with different patient populations is critical before the modified scores can be used more widely in other populations. Although the modified scores were found to be statistically superior to the original scores, there was overall a modest correlation between all of the clinical indices and both MRI-detected synovitis and radiographic progression. This suggests that there is still likely to be room for improvement in terms of accurately quantifying the objective disease burden.
Additionally, inclusion of other known predictors of progression (such as seropositivity, presence of erosions at baseline) in statistical models may improve the discriminative power. Finally, although these modified scores may correlate better with synovitis and radiographic progression, they are expected to be more discordant with patient's reported outcomes.
Further study in real-world settings may help to establish the utility of these scores in other populations. It should be emphasized that the previously developed disease activity scores have been validated and remain appropriate for following up disease activity over time. These indices correlate quite well with MRI scores, although not as well as the modified scores presented herein.
In conclusion, these findings have shown that, in comparison with the original DAS28-CRP and SDAI, the modified disease activity scores correlated better with synovitis on MRI, and also demonstrated greater predictive value for radiographic progression of joint damage at later time points, in patients with RA in a clinical trial setting.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. J. F. Baker had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. J. F. Baker, Conaghan, Smolen, Emery, D. G. Baker, Østergaard.
Acquisition of data. J. F. Baker, Conaghan, Emery, D. G. Baker, Østergaard.
Analysis and interpretation of data. J. F. Baker, Conaghan, Smolen, Aletaha, Shults, Emery, D. G. Baker, Østergaard.
Funding for the original GO-BEFORE and GO-FORWARD randomized clinical trials was provided by Janssen Biotech. There was no additional funding for this secondary study. Dr. D. G. Baker is an employee of Janssen Biotech.