Abstract

Objectives. The 28-joint disease activity score (DAS-28) guides the use of biologics in RA. The aims of this study were to investigate agreement between the ESR- and CRP-based DAS-28 definitions, and to examine how this agreement may be improved.

Methods. Data were obtained from registers of early (n = 520) and established RA (n = 364) patients. Agreement over disease activity levels (remission, low, moderate and high) at baseline and 6 months, and EULAR responder status at 6 months, were assessed in the early cohort. Two alternative DAS-28(CRP) definitions, obtained through linear regression analyses at baseline in the early RA patients, were validated with 6-month data from both the cohorts.

Results. In early RA patients, despite a high percentage of exact agreement over DAS-28 categories (88.2%), 38 (30.4%) of 125 patients with ‘moderate’ DAS-28(CRP) at baseline had ‘high’ DAS-28(ESR). This agreement was improved by modifying the DAS-28(CRP) definition, and by incorporating age and gender: e.g. in early RA patients with moderate original DAS-28(CRP), 30.4% had ‘high’ DAS-28(ESR), whereas 3.2% had ‘low’ DAS-28(ESR); following DAS-28(CRP) transformation both proportions were 6.6%. Incorporating age and gender did not improve agreement over EULAR response states.

Conclusion. The DAS-28(ESR) and DAS-28(CRP) definitions differ substantially in classifying RA patients as having moderate or high disease activity, with the ESR definition resulting in a higher proportion of high DAS-28 especially in women. Our results suggest that modifying the DAS-28(CRP) definition may improve agreement with DAS-28(ESR). There are important implications for meta-analyses and for therapy driven by DAS scores.

Introduction

The 28-joint disease activity score (DAS-28) [1] for RA [2] is an important outcome for clinical practice and research, incorporating swollen joint counts (SJCs) and tender joint counts (TJCs), a self-determined assessment of patient general health and acute-phase response. Published thresholds define absolute DAS-28 scores representing remission (<2.6), mild (≤3.2), moderate or severe (>5.1) disease activity, and in combination with absolute score at end point, change in DAS-28 over time allows the extent of response to be categorized (none, moderate and good) [3]. While response categories and the sensitivity to change in absolute DAS-28 values are important outcomes when assessing treatment effect, in a number of European countries cross-sectional measurement of DAS-28 is crucial in determining which treatment a patient receives [4]; for example, in the UK to qualify for anti-TNF drugs their DAS-28 must be >5.1 at two time points.

The original DAS-28 was based on ESR, but an alternative formula incorporating CRP [DAS-28(CRP)] has been developed [5]. Reasons for using CRP include not only its potential to be a more direct measure of inflammation, but also the increasing use of B-cell therapies that may have differential effects on these inflammatory markers. However, recent data from two large observational studies [6, 7] suggested that DAS-28(CRP) scores tended to be lower than DAS-28(ESR) scores and Inoue et al. [6] suggested potential new thresholds for disease activity categories for DAS-28(CRP). Subsequently, another group validated DAS-28 and EULAR response criteria based on CRP and compared them with the DAS-28 based on ESR [8]. They concluded that although DAS-28(CRP) yielded a better EULAR response more often than the DAS-28(ESR), the validation profiles (in terms of construct and criterion validity and sensitivity to change) of the two definitions were similar. The authors suggested that amending the definition may be preferable to changing thresholds for disease activity categories, and that any amendments would need to be widely applicable and based on examination of diverse data sets. These studies [6–8] obtained data from patients with established RA, and it is possible that the level of agreement between the different DAS-28 definitions might also vary with disease duration.

In the current study, we aimed to compare the different DAS definitions in an early RA cohort using regression methods as suggested by Wells et al. [8]. We also sought to investigate methods for improving agreement between definitions, especially taking into account the known age- and gender-related variations in ESR [9]. We then validated our results with data from another patient cohort with established RA.

Methods

Subjects

Data for measuring agreement and amending the DAS-28(CRP) definition were taken from the Yorkshire Early Arthritis Register (YEAR) of patients in the early stages of RA (disease duration ≤12 months) recruited between 1998 and the present. A comparison data set was compiled from a cohort of patients with established RA, recruited from 1999 to 2007. Patients were selected if they had both DAS(ESR) and DAS(CRP) data available at baseline. Patients whose baseline CRP was <1 were excluded because their data severely skewed the distribution of scores. Most of the early RA patients received DMARD therapy during the first 6 months of observation (SSZ 51%; MTX 47%; HCQ 1%; and none 1%). Patients in the established disease cohort received biological therapy (infliximab 53%; etanercept 34%; adalimumab 9%; anakinra 4%; and rituximab <1%), having previously failed to respond to DMARDs. Ethical approval for both cohorts was granted by local research ethics committees [Airedale Local Research Ethics Committee; Bradford Local Research Ethics Committee; Calderdale Local Research Ethics Committee; Dewsbury District Hospital Local Research Ethics Committee; Huddersfield Local Research Ethics Committee; Hull and East Riding Local Research Ethics Committee; Leeds HA/United Leeds Teaching Hospitals Local Research Ethics Committee; Leeds HA/St James’ and Seacroft University Hospitals Clinical Research (Ethics) Committee; Harrogate Health Care NHS Trust Local Research Ethics Committee; Northallerton Local Research Ethics Committee; Scarborough and North East Yorkshire Local Research Ethics Committee; York Research Ethics Committee; Wakefield District Research Ethics Committee; and Barnsley Local Research Ethics Committee] and all patients gave written informed consent.

Measures of disease activity

DAS-28 values were calculated according to the following equations: Patients were classified into DAS-28 categories (remission <2.6; low 2.6 to ≤3.2; moderate >3.2 to ≤5.1; and high >5.1) at baseline and 6 months. ‘Good’ EULAR responders showed an improvement of >1.2 and a score of ≤3.2 at 6 months; ‘moderate’ responders showed an improvement from >0.6 to ≤1.2 and had a score of ≤5.1, or an improvement of >1.2 and a score of >3.2; ‘non-responders’ showed improvement of ≤0.6, or an improvement from >0.6 to ≤1.2 and a score of >5.1 at 6 months [3]. Physical function was measured at baseline and 6 months using the HAQ disability index.

  • (i) DAS-28(CRP) = 0.56 × √(TJC-28) + 0.28 × √(SJC-28) + 0.014 × GH + 0.36 × ln(CRP + 1) + 0.96

  • (ii) DAS-28(ESR) = 0.56 × √(TJC-28) + 0.28 × √(SJC-28) + 0.014 × GH + 0.70 × ln(ESR)

Validation analyses for the DAS-28(CRP)

Criterion validity

The mean of the differences between baseline DAS-28 values obtained from patients according to the two DAS-28 definitions was used to construct limits of agreement between the two definitions. The classification of early RA patients into DAS-28 categories and EULAR response states according to DAS-28(ESR) and DAS-28(CRP) were cross-classified to identify any areas of disagreement.

Construct validity

Changes in physical function measured by HAQ were assessed across EULAR response categories and compared between the two DAS-28 definitions.

Effects of age and gender on the agreement between DAS-28(ESR) and DAS-28(CRP)

Linear regression modelling was used to investigate whether age and gender were related to lnESR and/or ln(CRP + 1) independently of other measures of disease activity and subsequently to create two new DAS-28(CRP) definitions: one that adjusted the constant in DAS-28(CRP) and the multiplier applied to CRP, and another that additionally incorporated both age and gender. Values obtained from the new DAS-28(CRP) definitions were compared with DAS-28(ESR) values at 6 months in the early RA cohort (to assess the performance of the alternative definitions across a wider range of DAS-28 values), and at both baseline and 6 months in the established RA data set.

Statistical methods

All statistical analyses were carried out in SPSS 16.0.2 (SPSS Inc., an IBM company, Chicago, Illinois, USA) Bland–Altman plots [10] were created by plotting the differences between DAS-28(ESR) and DAS-28(CRP) against the mean of the measurements. Limits of agreement and 95% CIs around these limits were calculated [10]. To assess agreement between the different DAS-28 definitions over the classification of patients into disease activity and response categories, binary κ [11] and quadratic-weighted κ (κw) [12] statistics were calculated, in addition to descriptive statistics regarding the proportion of cases over which the two exactly agreed [percentage of exact agreement (PEA)], or were in close (±1 category) agreement [percentage of close agreement (PCA)]. Category-specific agreement was also calculated: specific agreement for category i was defined as [2*(number of patients placed in category i by both definitions)]/{[total number placed in category i by DAS(ESR)] + [total number placed in category i by DAS-28(CRP)]}.

One-way analysis of covariance (ANCOVA) models, including baseline values as covariates, were used to identify significant differences in HAQ improvement after 6 months among patients who according to EULAR criteria were ‘good’ responders, ‘moderate’ responders or ‘non-responders’. To indicate whether age and gender affect ESR or CRP independently of other measures of disease activity, forced-entry linear regression models were constructed with either lnESR or ln(CRP + 1) as dependent variables and with age, gender, SJC-28, TJC-28 and patient assessment of disease activity visual analogue scale (VAS) included as independent variables. In order to construct a new DAS-28(CRP) definition, 0.7*lnESR [the ESR ‘component’ of DAS-28(ESR)] was entered as the dependent variable, with age, gender and ln(CRP + 1) included as independent variables. Regression coefficients obtained for each of the independent variables and the associated constant for the model were then added to the new DAS-28(CRP) definitions.

Results

A total of 520 early RA patients with full data for both DAS-28(ESR) and DAS-28(CRP) at baseline were included in the analysis; 284 (55%) had full data at 6 months. Eighty-six early RA patients whose CRP was <1 at baseline were excluded; median (range) ESR in these patients was 15.5 (10–100) mm/h. Three hundred and sixty-four patients with established RA were included in the baseline analysis; 189 had full data at 6 months. Baseline demographic and clinical data are summarized in Table 1. Baseline demographic and clinical results did not differ substantively between patients who had full data available at baseline and those who did not (data not shown). Baseline DAS-28(CRP) scores were generally lower than DAS-28(ESR) scores in both the cohorts (Table 1 and Fig. 1).

Fig. 1

Baseline DAS-28(CRP) plotted against DAS-28(ESR) in RA patients with early (○) and established (△) disease. The dotted black line indicates exact agreement between the two DAS-28 definitions.

Fig. 1

Baseline DAS-28(CRP) plotted against DAS-28(ESR) in RA patients with early (○) and established (△) disease. The dotted black line indicates exact agreement between the two DAS-28 definitions.

Table 1

Baseline demographics and clinical results for patients with recent-onset RA and both ESR and CRP measurements

Baseline data Early RA (n = 520) Established RA (n = 364) 
Age, yearsa 57.7 (14.1) 52.8 (13.8) 
Female gender, % 64.2 76.6 
Caucasian race, % 96.7 (495/512) 92.8 (116/125) 
Disease durationb, months 6.0 (3.9–8.3) 108 (60–180) 
TJC-28a 13.3 (8.3) 15.3 (7.5) 
SJC-28a 11.1 (6.9) 9.4 (6.5) 
HAQa 1.43 (0.72) 2.05 (0.55) 
Patient global assessment 0–100 mm VASa 63.4 (24.8) 73.7 (18.9) 
ESRb, mm/h 39 (1–130) 36 (2–132) 
CRPb, mg/l 23 (1–339) 30 (2–244) 
DAS-28(ESR)a 6.15 (1.38) 6.37 (1.05) 
DAS-28(CRP)a 5.80 (1.31) 6.13 (0.94) 
Baseline data Early RA (n = 520) Established RA (n = 364) 
Age, yearsa 57.7 (14.1) 52.8 (13.8) 
Female gender, % 64.2 76.6 
Caucasian race, % 96.7 (495/512) 92.8 (116/125) 
Disease durationb, months 6.0 (3.9–8.3) 108 (60–180) 
TJC-28a 13.3 (8.3) 15.3 (7.5) 
SJC-28a 11.1 (6.9) 9.4 (6.5) 
HAQa 1.43 (0.72) 2.05 (0.55) 
Patient global assessment 0–100 mm VASa 63.4 (24.8) 73.7 (18.9) 
ESRb, mm/h 39 (1–130) 36 (2–132) 
CRPb, mg/l 23 (1–339) 30 (2–244) 
DAS-28(ESR)a 6.15 (1.38) 6.37 (1.05) 
DAS-28(CRP)a 5.80 (1.31) 6.13 (0.94) 

aMean (s.d.). bMedian (interquartile range).

Criterion validity

Limits of agreement

Examination of a Bland–Altman plot of the differences between DAS-28(ESR) and DAS-28 (CRP) in the early RA patients (Fig. 2) indicated that the magnitude of the differences was not related to the mean of the different DAS-28 definitions. The mean (95% CI) of the differences between the ESR- and CRP-based scores (ESR − CRP) was 0.34 (0.30, 0.38). The 95% limits of agreement (95% CI) were from −0.58 (−0.65, −0.51) to 1.26 (1.19, 1.33).

Fig. 2

The difference between baseline DAS-28(ESR) and DAS-28(CRP) plotted against the mean of the two values for patients with early RA (n = 520). The thick continuous line indicates the mean of the differences; dotted lines indicate 95% limits of agreement; and thin continuous lines indicate 95% CIs.

Fig. 2

The difference between baseline DAS-28(ESR) and DAS-28(CRP) plotted against the mean of the two values for patients with early RA (n = 520). The thick continuous line indicates the mean of the differences; dotted lines indicate 95% limits of agreement; and thin continuous lines indicate 95% CIs.

Cross-classification of DAS-28 categories

Comparing the classification of early RA patients as being in remission, or having mild, moderate or severe disease activity between DAS-28(ESR) and DAS-28(CRP), there were substantial levels of agreement when all categories were considered simultaneously (PEA = 88.5%; Tables 2 and 3). However, 38 (30.4%) of the 125 patients who showed ‘moderate disease activity’ according to their DAS-28(CRP) scores showed ‘high disease activity’ according to their DAS-28(ESR) scores. This disparity was more pronounced in females with moderate DAS-28(CRP) [31 (36.5%) of 85 patients] than for males [7 (17.5%) of 40 patients]. At 6 months, PEA between the two definitions went down to 70.1%; a greater proportion of patients were classified as having low disease activity, the category where agreement was poorest, and therefore the overall agreement was worse. DAS-28(ESR) remission (<2.6) was achieved in 56 (19.7%) of the 284 patients compared with 28.2% for DAS-28(CRP). The definitions agreed over remission status in 18.3% of patients and active disease status in 70.4% [κ = 0.69 (0.60–0.79)]. Of the 80 patients in DAS-28(CRP) remission, 28 (35.0%) had DAS-28(ESR) ≥2.6 [mean (range) 3.17 (2.63–4.14)].

Table 2

Agreement between DAS-28(ESR) and three different definitions of DAS-28(CRP) on the classification of patients into categories of disease activity (DA) at baseline and 6 months in early (n = 520) or established (n = 364) RA patients and EULAR responder states achieved at 6 months in a subset of patients in each group (early n = 284; established n = 189)

 Early RA
 
Established RA
 
DAS-28 (CRP) DAS-28 (CRP).1 DAS-28 (CRP).2 DAS-28 (CRP) DAS-28 (CRP).1 DAS-28 (CRP).2 
Baseline DAS-28 categories       
    Remission (CRP), % 1.5 1.0 1.0 0.0 0.0 0.0 
    Mild DA (CRP), % 1.9 1.3 1.3 0.0 0.0 0.0 
    Moderate DA (CRP), % 24.0 17.5 17.7 12.4 6.3 6.3 
    High DA (CRP), % 72.5 80.2 80.0 87.6 93.7 93.7 
    PEA with DAS-28(ESR) 88.5 91.9 93.3 91.2 94.0 94.5 
    PCA with DAS-28(ESR) 99.4 99.8 99.8 99.7 99.7 99.7 
 Specific agreement, %       
        Remission 71.4 72.7 72.7 – – – 
        Low 44.4 13.3 26.7 – – – 
        Moderate 73.8 79.2 82.7 56.8 57.7 61.5 
        High 94.0 96.6 97.2 91.2 97.2 97.5 
    ‘Moderate’ CRP with ‘high’ ESR, % 30.4 9.8 6.6 46.7 21.7 17.4 
    ‘Moderate’ CRP with ‘low’ or ‘remission’ ESR, % 3.2 7.6 6.6 6.6 13.0 13.0 
 κw       
    95% CI 0.79 0.84 0.87 0.57 0.61 0.64 
        Lower 0.73 0.79 0.82 0.46 0.49 0.52 
        Upper 0.86 0.90 0.92 0.69 0.73 0.76 
6-month DAS-28 categories       
    In remission, % 28.2 20.1 19.4 23.8 15.3 13.8 
    Mild DA, % 10.9 10.9 10.6 15.9 14.3 14.8 
    Moderate DA, % 34.2 36.3 36.3 42.3 48.1 49.2 
    High DA, % 26.8 32.7 33.8 18.0 22.2 22.2 
    PEA with DAS-28(ESR) 70.1 74.3 75.4 69.3 74.1 74.6 
    PCA with DAS-28(ESR) 96.4 98.6 98.9 97.9 97.4 98.4 
 Specific agreement, %       
        Remission 76.5 70.3 74.3 65.9 66.7 69.6 
        Low 19.0 16.1 22.2 25.5 41.5 38.5 
        Moderate 67.0 78.0 78.0 79.5 81.6 82.6 
        High 86.9 91.3 90.6 83.1 84.7 84.7 
    ‘Moderate’ CRP with ‘high’ ESR, % 23.7 9.7 11.7 13.8 7.5 7.7 
    ‘Moderate’ CRP with ‘low’ or ‘remission’ ESR, % 9.3 14.5 12.7 6.2 16.2 13.3 
 κw       
    95% CI 0.84 0.88 0.89 0.83 0.83 0.85 
        Lower 0.80 0.84 0.86 0.78 0.77 0.80 
        Upper 0.88 0.91 0.92 0.88 0.89 0.90 
6-month EULAR response       
    Good, % 35.6 28.2 26.8 39.7 29.6 28.6 
    Moderate, % 35.2 41.9 43.3 43.4 52.9 54.0 
    None, % 29.2 29.9 29.9 16.9 17.5 17.5 
    PEA with DAS-28(ESR) 79.9 83.1 83.1 84.1 84.7 85.7 
    PCA with DAS-28(ESR) 99.3 100.0 100.0 99.5 100.0 100.0 
 Specific agreement, %       
        Good 77.6 81.0 81.5 85.7 84.0 86.0 
        Moderate 73.9 79.5 79.1 82.8 84.7 85.6 
        None 89.7 89.8 89.8 84.1 85.7 85.7 
    ‘Good’ CRP with ‘mod’ or ‘none’ ESR, % 29.7 15.8 17.5 20.0 7.4 7.1 
    ‘Good’ ESR with ‘mod’ or ‘none’ CRP, % 13.4 22.0 19.5 7.7 23.1 20.0 
 κw       
    95% CI 0.82 0.86 0.86 0.83 0.84 0.85 
        Lower 0.77 0.82 0.82 0.77 0.78 0.80 
        Upper 0.87 0.90 0.90 0.90 0.90 0.91 
 Early RA
 
Established RA
 
DAS-28 (CRP) DAS-28 (CRP).1 DAS-28 (CRP).2 DAS-28 (CRP) DAS-28 (CRP).1 DAS-28 (CRP).2 
Baseline DAS-28 categories       
    Remission (CRP), % 1.5 1.0 1.0 0.0 0.0 0.0 
    Mild DA (CRP), % 1.9 1.3 1.3 0.0 0.0 0.0 
    Moderate DA (CRP), % 24.0 17.5 17.7 12.4 6.3 6.3 
    High DA (CRP), % 72.5 80.2 80.0 87.6 93.7 93.7 
    PEA with DAS-28(ESR) 88.5 91.9 93.3 91.2 94.0 94.5 
    PCA with DAS-28(ESR) 99.4 99.8 99.8 99.7 99.7 99.7 
 Specific agreement, %       
        Remission 71.4 72.7 72.7 – – – 
        Low 44.4 13.3 26.7 – – – 
        Moderate 73.8 79.2 82.7 56.8 57.7 61.5 
        High 94.0 96.6 97.2 91.2 97.2 97.5 
    ‘Moderate’ CRP with ‘high’ ESR, % 30.4 9.8 6.6 46.7 21.7 17.4 
    ‘Moderate’ CRP with ‘low’ or ‘remission’ ESR, % 3.2 7.6 6.6 6.6 13.0 13.0 
 κw       
    95% CI 0.79 0.84 0.87 0.57 0.61 0.64 
        Lower 0.73 0.79 0.82 0.46 0.49 0.52 
        Upper 0.86 0.90 0.92 0.69 0.73 0.76 
6-month DAS-28 categories       
    In remission, % 28.2 20.1 19.4 23.8 15.3 13.8 
    Mild DA, % 10.9 10.9 10.6 15.9 14.3 14.8 
    Moderate DA, % 34.2 36.3 36.3 42.3 48.1 49.2 
    High DA, % 26.8 32.7 33.8 18.0 22.2 22.2 
    PEA with DAS-28(ESR) 70.1 74.3 75.4 69.3 74.1 74.6 
    PCA with DAS-28(ESR) 96.4 98.6 98.9 97.9 97.4 98.4 
 Specific agreement, %       
        Remission 76.5 70.3 74.3 65.9 66.7 69.6 
        Low 19.0 16.1 22.2 25.5 41.5 38.5 
        Moderate 67.0 78.0 78.0 79.5 81.6 82.6 
        High 86.9 91.3 90.6 83.1 84.7 84.7 
    ‘Moderate’ CRP with ‘high’ ESR, % 23.7 9.7 11.7 13.8 7.5 7.7 
    ‘Moderate’ CRP with ‘low’ or ‘remission’ ESR, % 9.3 14.5 12.7 6.2 16.2 13.3 
 κw       
    95% CI 0.84 0.88 0.89 0.83 0.83 0.85 
        Lower 0.80 0.84 0.86 0.78 0.77 0.80 
        Upper 0.88 0.91 0.92 0.88 0.89 0.90 
6-month EULAR response       
    Good, % 35.6 28.2 26.8 39.7 29.6 28.6 
    Moderate, % 35.2 41.9 43.3 43.4 52.9 54.0 
    None, % 29.2 29.9 29.9 16.9 17.5 17.5 
    PEA with DAS-28(ESR) 79.9 83.1 83.1 84.1 84.7 85.7 
    PCA with DAS-28(ESR) 99.3 100.0 100.0 99.5 100.0 100.0 
 Specific agreement, %       
        Good 77.6 81.0 81.5 85.7 84.0 86.0 
        Moderate 73.9 79.5 79.1 82.8 84.7 85.6 
        None 89.7 89.8 89.8 84.1 85.7 85.7 
    ‘Good’ CRP with ‘mod’ or ‘none’ ESR, % 29.7 15.8 17.5 20.0 7.4 7.1 
    ‘Good’ ESR with ‘mod’ or ‘none’ CRP, % 13.4 22.0 19.5 7.7 23.1 20.0 
 κw       
    95% CI 0.82 0.86 0.86 0.83 0.84 0.85 
        Lower 0.77 0.82 0.82 0.77 0.78 0.80 
        Upper 0.87 0.90 0.90 0.90 0.90 0.91 
Table 3

Agreement between DAS-28(ESR) and DAS-28(CRP) on classification of patients with early RA (n = 520) by baseline DAS-28 categories

DAS-28(CRP) DAS-28(ESR)
 
Remission Low Moderate High 
Remission, n 5 
    Total, % 1.0 0.2 0.4 0.0 
    Remission (CRP), % 62.5 12.5 25.0 0.0 
Low, n 4 
    Total, % 0.0 0.8 1.2 0.0 
    Low (CRP), % 0.0 40.0 60.0 0.0 
Moderate, n 83 38 
    Total, % 0.2 0.6 16.0 7.3 
    Moderate (CRP), % 0.8 2.4 66.4 30.4 
High, n 368 
    Total, % 0.0 0.0 1.7 70.8 
    High (CRP), % 0.0 0.0 2.4 97.6 
DAS-28(CRP) DAS-28(ESR)
 
Remission Low Moderate High 
Remission, n 5 
    Total, % 1.0 0.2 0.4 0.0 
    Remission (CRP), % 62.5 12.5 25.0 0.0 
Low, n 4 
    Total, % 0.0 0.8 1.2 0.0 
    Low (CRP), % 0.0 40.0 60.0 0.0 
Moderate, n 83 38 
    Total, % 0.2 0.6 16.0 7.3 
    Moderate (CRP), % 0.8 2.4 66.4 30.4 
High, n 368 
    Total, % 0.0 0.0 1.7 70.8 
    High (CRP), % 0.0 0.0 2.4 97.6 

Bold values indicate exact agreement between the two DAS-28 definitions.

Cross-classification of EULAR response categories at 6 months

Levels of agreement between DAS-28(ESR) and DAS-28(CRP) over EULAR response rates at 6 months [PEA = 80.0%; κw (95% CI) = 0.82 (0.75, 0.89); Tables 2 and 4] were similar to those observed by Wells et al. [8] [DAS-28(CRP): PEA = 82.4%; κw = 0.80 (0.76, 0.83)]. However, 28 (27.7%) of the 101 patients who achieved a ‘good’ EULAR response according to their 6-month DAS-28(CRP) profile only showed ‘moderate’ response according to their DAS-28(ESR), and 2 (2.0%) showed no DAS-28(ESR) EULAR response at all. This level of disagreement is comparable to the findings of Wells et al. [8] where 108 patients showed ‘good’ improvement according to DAS-28(CRP), but 56 (34.2%) of these patients showed only moderate DAS-28(ESR) improvement.

Table 4

Agreement between DAS-28(ESR) and DAS-28(CRP) on classification of patients with early RA (n = 284) by EULAR response criteria at 6 months

DAS-28(CRP) DAS-28(ESR)
 
Good Moderate None 
Good 71 28 
    Total, % 25.0 9.9 0.7 
    Good (CRP), % 70.3 27.7 2.0 
Moderate 11 78 11 
    Total, % 3.9 27.5 3.9 
    Moderate (CRP), % 11.0 78.0 11.0 
None 78 
    Total, % 0.0 1.8 27.5 
    None (CRP), % 0.0 6.0 94.0 
DAS-28(CRP) DAS-28(ESR)
 
Good Moderate None 
Good 71 28 
    Total, % 25.0 9.9 0.7 
    Good (CRP), % 70.3 27.7 2.0 
Moderate 11 78 11 
    Total, % 3.9 27.5 3.9 
    Moderate (CRP), % 11.0 78.0 11.0 
None 78 
    Total, % 0.0 1.8 27.5 
    None (CRP), % 0.0 6.0 94.0 

Bold values indicate exact agreement between the two DAS-28 definitions.

Construct validity

Change in HAQ could be calculated for 268 patients. Mean improvements in HAQ differed significantly between EULAR responder states [DAS-28(ESR) ANCOVA F = 67.0, P < 0.001; DAS-28(CRP) F = 57.3, P < 0.001; all Bonferroni-corrected post hoc comparisons between pairs of responder states significant at P < 0.001, for either definition; Fig. 3] and the differences were of comparable magnitude across the different DAS-28 definitions [mean changes in HAQ were −0.86 and −0.80 for good responders; −0.46 and −0.41 for moderate responders; and 0.05 and 0.05 for non-responders, DAS-28(ESR) and DAS-28(CRP), respectively]. Non-responders showed a mean deterioration in HAQ, and in agreement with the findings of Wells et al. [8], function improved linearly across the EULAR responder states.

Fig. 3

Mean improvement from baseline in HAQ for patients with early RA who were good (light grey), moderate (dark grey) or non-responders (black) according to EULAR criteria based on DAS-28(ESR) or DAS-28(CRP). Error bars represent 95% CIs.

Fig. 3

Mean improvement from baseline in HAQ for patients with early RA who were good (light grey), moderate (dark grey) or non-responders (black) according to EULAR criteria based on DAS-28(ESR) or DAS-28(CRP). Error bars represent 95% CIs.

Effects of age and gender on the agreement between DAS-28(ESR) and DAS-28(CRP)

Age and gender were significantly associated with both lnESR and lnCRP independently of the other DAS-28 variables in patients with early RA. In contrast to the findings of Radovits et al. [9], there were no significant interactions between age and gender (ESR β = −0.004, s.e. = 0.001, P = 0.218; CRP β = 0.003, s.e. = 0.006, P = 0.565). Females tended to have higher lnESR values (β = 0.194, s.e. = 0.067, P = 0.004), but lower lnCRP values (β = −0.152, s.e. = 0.079, P = 0.055). SJC-28 was significantly associated with both lnESR (β = 0.023, s.e. = 0.006, P < 0.001) and lnCRP (β = 0.018, s.e. = 0.007, P = 0.013), whereas TJC-28 was not (lnESR β = −0.009, s.e. = 0.005, P = 0.071; lnCRP β = −0.002, s.e. = 0.006, P = 0.695).

We felt it would be important to show that adding age and gender to the DAS-28(CRP) definition would lead to improvements in agreement with DAS-28(ESR) over and above those that might be achieved simply by transforming the original DAS-28(CRP). We therefore created two new definitions, one with the same structure as the existing definition, but with modified CRP multiplier and constant, and another that also incorporated age and gender. Regressing 0.7 × ln(ESR) onto ln(CRP + 1) produced a simple, modified DAS-28(CRP):

DAS-28(CRP).1 = (0.56 × √TJC-28) + (0.28 × √SJC-28) + (0.014 × GH) + [0.292 × ln(CRP + 1)] + 1.523

Regressing 0.7 × lnESR onto age, gender and ln(CRP + 1) yielded:

DAS-28(CRP).2 = (0.56 × √TJC-28) + (0.28 × √SJC-28) + (0.014 × GH) + [0.288 × ln(CRP + 1)] + (0.003 × age) + (0.159 if female) + 1.238

In terms of overall PEA, both of the amended DAS-28(CRP) definitions showed better agreement with DAS-28(ESR) over the classification of patients into disease activity categories, at both time points, in patients with early or established RA (Table 2). The definition that incorporated age and gender, DAS-28(CRP).2, generally performed slightly better than DAS-28(CRP).1, reducing and equalizing the amount by which DAS-28(CRP) over- and underestimated DAS-28(ESR).

Agreement over ‘low’ disease activity was generally poor, potentially because few patients were assigned to this category in either cohort, at either time point. Agreement over the designation of ‘low’ disease activity was not improved by either of the two amended DAS-28(CRP) definitions. Both definitions improved agreement over EULAR response states, but to a similar degree. In terms of EULAR response, the addition of age and gender did not result in improved agreement between DAS-28(ESR) and DAS-28(CRP) beyond that which the more simple modification of DAS-28(CRP) (first new definition) achieved.

Discussion

We examined agreement between DAS-28(ESR) and DAS-28(CRP) on categories of disease activity and EULAR response states and found disagreement, particularly in low to moderate disease activity and moderate to good response. We derived new definitions using an early RA cohort and demonstrated, by application to subsequent time points for this and an established RA cohort, that it is possible to derive a new DAS-28(CRP) definition that shows improved agreement with DAS-28(ESR), across a range of DAS-28 values and disease durations.

If ‘high’ disease activity is important for determining access to expensive biological therapies, and DAS-28(CRP) is substituted for DAS-28(ESR) when making treatment decisions, based on the results we present here as many as 30% of the patients categorized as ‘moderate’ (and therefore ineligible) with CRP would be categorized ‘high’ using ESR. Around 30% of the patients who are considered to have shown a ‘good’ DAS-28(CRP) response to an experimental therapy may actually have only shown a ‘moderate’ DAS-28(ESR) response. In light of these concerns, it seems prudent to be cautious when considering the use of the current DAS-28(CRP) definition. It was derived from one study of repeated measurements made in 334 patients, which was not fully published [5], but has nevertheless been adopted as a valid alternative to the DAS-28(ESR) despite mounting evidence that it produces lower scores.

Two large cohort studies from Japan [6, 7] have previously highlighted the tendency for DAS-28(CRP) to underestimate DAS-28(ESR) and recommended that changes to either the DAS-28(CRP) definition or the thresholds for disease activity and response categories should be made before DAS-28(CRP) can be considered equivalent to DAS-28(ESR). Inoue et al. [6] suggested new disease activity cut-offs of 2.3, 2.7 and 4.1. Applying these to our data from the early RA patients at 6 months, yielded exact agreement of 71.5%, which represents only a marginal improvement over the original cut-offs for DAS-28(CRP) (70.1%; Table 2); the gains in agreement were larger for our amended DAS-28(CRP) definitions, using the original cut-offs [DAS-28(CRP).1 74.3%; DAS-28(CRP).2 75.4%]. The cut-offs suggested by Inoue et al. [6] appear to be too low; the proportion of patients with moderate DAS-28(CRP) who had high DAS-28(ESR) was 2.7%, whereas 19.2% had low DAS-28(ESR) or were in DAS-28(ESR) remission. Matsui et al. [7] presented data for individual DAS-28 categories, from which we can determine that 488 (32.6%) of 1498 patients with moderate DAS-28(CRP) had high DAS-28(ESR), consistent with our findings. The authors also presented data from previous studies aimed at evaluating agreement between the two DAS-28 measurements, and pointed out that with the exception of the original abstract [5], mean DAS-28(CRP) was consistently lower than mean DAS-28(ESR).

Recent works [8, 13] aimed at assessing the relationship between DAS-28(ESR) and DAS-28(CRP) reported high levels of agreement between the definitions and concluded that DAS-28(CRP) was a valid alternative to DAS-28(ESR). Wells et al. [8] calculated outcomes such as the overall PEA without reference to distribution within categories. This approach can mask differences in agreement within specific categories; over 34% of patients in their study who showed a ‘good’ EULAR response according to their DAS-28(CRP) results only achieved a moderate response on DAS-28(ESR). Crowson et al. [13] used the strength of correlation between DAS-28(CRP) and DAS-28(ESR) to indicate that the two could be used ‘interchangeably’; however, this method may not be appropriate for assessing agreement because it does not detect consistent bias; indeed, examination of their published scatter plot suggests that in the majority of cases DAS-28(CRP) values were lower than the corresponding DAS-28(ESR). Our work highlights the need for rigorous investigation of agreement that takes into account the distribution of scores on each scale and measures category-specific agreement regarding classification into categories of status or response; overall measurements of agreement are not sufficient.

Wells et al. [8] cautioned that any transformation of the DAS-28(CRP) would need to be generalizable across different patient groups. Both of the novel definitions that we developed in this study improved agreement between the CRP- and ESR-based DAS-28 definitions, not only in the data collected at 6 months from the early RA patients whose baseline data had formed the basis for the transformations, but also at both time points in a separate cohort of patients with established disease. While we would not suggest that either of these novel definitions should be used in place of the existing DAS-28(CRP), we hope our results demonstrate that it should be possible to achieve a much better fit between DAS-28(CRP) and DAS-28(ESR) that is generalizable across different patient groups, at least for moderate to high scores. Unfortunately, our results show that agreement in the ‘low’ disease category was poor, even after transformation of DAS-28(CRP); Inoue et al. [6] also reported poor agreement over low DAS-28 scores. Given that ESR in patients with normal baseline CRP who were excluded from our analysis ranged up to 100 mm/h, it seems that there is quite some divergence between the two measurements of acute-phase response where there are low levels of inflammation. However, it is possible that the poor agreement is simply a function of the relatively small numbers of people in either cohort that had ‘low’ disease activity, either at baseline or following 6 months of therapy. Without actively selecting equal numbers of patients within each category of disease activity, this pattern is to be expected; the DAS-28 score interval representing low disease activity (2.6–3.2) only covers 6% of the 10-point scale, making it a very small category compared with those for remission (26%) and moderate (19%) or high (49%) disease activity. This fact alone may explain the relatively poor agreement observed for this category compared with others.

Although we have demonstrated that simply amending the CRP multiplier and constant in the current DAS-28(CRP) definition results in improved agreement with DAS-28(ESR), there was an additional (if comparatively small) gain to be made by including age and gender, at least in terms of classifying patients into the different cross-sectional DAS-28 categories, if not EULAR responder states. If the aim is to make DAS-28(CRP) as comparable as possible to DAS-28(ESR), it may be worth including these extra terms in any novel DAS-28(CRP) definition. Any additional effort in terms of calculation will be negligible, or non-existent, as the use of direct electronic entry methods becomes more widespread.

There were limitations to the current study, such as the lack of complete follow-up data. Since the data came from registers of patients who were treated at a number of different centres, not enrolled in controlled clinical trials, some attrition was inevitable. However, the primary focus of the analysis was to compare the different DAS-28 definitions within each cohort; hence while the follow-up data were incomplete, the main priority was to have complete data to obtain accurate estimates of the proportions of patients in each disease activity or EULAR response category. Larger sample sizes would have increased numbers within each DAS category, allowing for greater precision in assessing agreement and deriving the novel definitions. The relatively small number of patients with low disease activity was a particular problem that was exacerbated at baseline by the exclusion of patients with CRP = 0; however, even after 6 months of DMARD therapy the proportion of patients with low disease activity was low.

It should be noted that, like the developers of DAS(CRP), we are suggesting methods for amending the existing DAS formula, rather than constructing a new tool (with attendant requirement for a more detailed assessment of construct validity). CRP is recognized as a more sensitive measure of inflammation than ESR and can be measured centrally from stored serum samples. It could be argued that if a CRP-based measure of disease activity were to entirely replace DAS-28(ESR) in the future, then the process of identifying the combination of measures that best reflects disease activity should be carried out afresh, given that their relationships with CRP could differ. The lack of agreement at lower levels of inflammation supports this possibility. Our regression analyses found a slightly stronger relationship between TJC and ESR (standardized β = −0.103; P = 0.071) than CRP (standardized β = −0.022; P = 0.695).

This study indicates that if comparability with DAS-28(ESR) is to be maintained, the DAS-28(CRP) definition needs adjustment, and that although further work is needed before a definition can be provided that is shown to be valid across patients of different ages, sexes, weights and levels of disease activity, and at different stages of their disease, it should be feasible to obtain one. Our data suggest that a decision needs to be made as to whether the DAS-28(CRP) definition itself should be amended, or whether new thresholds for disease activity and EULAR response categories should be determined to account for the fact that the YEAR Consortium DAS-28(CRP) values are generally lower than the DAS-28(ESR) values on which the original thresholds were based. We feel that changes should be made to the definition itself, following similar methods to those presented in this article, but using a wide range of data sets; until such a definitive revision is published, caution should be used when interpreting DAS-28(CRP) results. Non-agreement between the DAS-28(ESR) and the current DAS-28(CRP) definition may have important implications for meta-analyses, for clinicians who apply therapies based on DAS(ESR)-based clinical trials but use DAS(CRP) and for therapy driven by DAS scores.

graphic

Acknowledgements

The YEAR consortium

The members of the YEAR Consortium are:

Management team: Professor Paul Emery1, Professor Philip Conaghan1, Professor Ann Morgan1, Dr Anne-Maree Keenan1 and Dr Elizabeth Hensor1.

Medical staff: Dr Mark Quinn2, Dr Andrew Gough3, Dr Michael Green2,3, Dr Richard Reece4, Dr Lesley Hordon5, Dr Philip Helliwell1,6, Dr Richard Melsom6, Dr Sheelagh Doherty7, Dr Ade Adebajo8, Dr Andrew Harvey9, Dr Steve Jarrett9, Dr Gareth Huson1, Dr Amanda Isdale2, Dr Mike Martin1, Dr Zunaid Karim9, Professor Dennis McGonagle1,10, Dr Colin Pease1, Dr Sally Cox1, Dr Victoria Bejarano1, Dr Jackie Nam1, Dr Edith Villeneuve1 and Dr Sarah Twigg1.

Nursing staff: Claire Brown1, Christine Thomas1, David Pickles1, Alison Hammond1, Beverley Neville3, Alan Fairclough4, Caroline Nunns4, Anne Gill2, Julie Green2, Belinda Rhys-Evans1, Barbara Padwell1, Julie Madden10, Lynda Taylor10, Sally Smith1, Heather King1, Jill Firth6, Jayne Heard7 and Linda Sigsworth6.

Support Staff: Diane Corscadden1, Karen Henshaw1, Lubna-Haroon Rashid1, Stephen G Martin1, James I Robinson1, Dr Lukasz Kozera1, Sarah Fahy1 and Andrea Paterson1.

1Section of Musculoskeletal Disease, LIMM, Leeds, 2York District Hospital, York, 3Harrogate District Hospital, Harrogate, 4Huddersfield Royal Infirmary, Huddersfield, 5Dewsbury District and General Hospital, Dewsbury, 6St Luke's Hospital, Bradford, 7Hull Royal Infirmary, Hull, 8Barnsley District General Hospital, Barnsley, 9Pinderfields General Hospital, Wakefield and 10Calderdale Royal Hospital, Halifax, UK

Funding: The Yorkshire Early Arthritis Register is in part supported by a programme grant from Arthritis Research UK. E.M.A.H. and P.G.C. are funded in part by Arthritis Research UK.

Disclosure statement: P.G.C. has done advisory boards, presentations or holds research grants from Abbott, BMS, Centocor, Novartis, Schering Plough, UCB and Wyeth. P.E. has done advisory boards, presentations or holds research grants from Abbott, BMS, Centocor, Novartis, Roche, Schering Plough, UCB and Wyeth. All other authors have declared no conflicts of interest.

References

1
van der Heijde
DM
van’ t Hof
MA
van Riel
PL
, et al.  . 
Judging disease activity in clinical practice in rheumatoid arthritis: first step in the development of a disease activity score
Ann Rheum Dis
 , 
1990
, vol. 
49
 (pg. 
916
-
20
)
2
Prevoo
ML
van’ t Hof
MA
Kuper
HH
van Leeuwen
MA
van de Putte
LB
van Riel
PL
Modified disease activity scores that include twenty-eight-joint counts. Development and validation in a prospective longitudinal study of patients with rheumatoid arthritis
Arthritis Rheum
 , 
1995
, vol. 
38
 (pg. 
44
-
8
)
3
van Gestel
AM
Haagsma
CJ
van Riel
PL
Validation of rheumatoid arthritis improvement criteria that include simplified joint counts
Arthritis Rheum
 , 
1998
, vol. 
41
 (pg. 
1845
-
50
)
4
Emery
P
van Vollenhoven
R
Ostergaard
M
, et al.  . 
Guidelines for initiation of anti-tumour necrosis factor therapy in rheumatoid arthritis: similarities and differences across Europe
Ann Rheum Dis
 , 
2009
, vol. 
68
 (pg. 
456
-
9
)
5
Fransen
J
Welsing
PMJ
de Keijzer
RMH
van Riel
PLCM
Disease activity scores using C-reactive protein: CRP may replace ESR in the assessment of RA disease activity
Ann Rheum Dis
 , 
2003
, vol. 
62(Suppl. 1)
 pg. 
151
 
6
Inoue
E
Yamanaka
H
Hara
M
Tomatsu
T
Kamatani
N
Comparison of Disease Activity Score (DAS)28-erythrocyte sedimentation rate and DAS28-C-reactive protein threshold values
Ann Rheum Dis
 , 
2007
, vol. 
66
 (pg. 
407
-
9
)
7
Matsui
T
Yoshiaki
K
Kaneko
A
, et al.  . 
Disease Activity Score 28 (DAS28) using C-reactive protein underestimates disease activity and overestimates EULAR response criteria compared with DAS28 using erythrocyte sedimentation rate in a large observational cohort of rheumatoid arthritis patients in Japan
Ann Rheum Dis
 , 
2007
, vol. 
66
 (pg. 
1221
-
6
)
8
Wells
GA
Becker
J-CP
Tend
J
, et al.  . 
Validation of the disease activity score 28 (DAS28) and EULAR response criteria based on CRP against disease progression in patients with rheumatoid arthritis, and comparison with the DAS28 based on ESR
Ann Rheum Dis
 , 
2009
, vol. 
68
 (pg. 
954
-
60
)
9
Radovits
BJ
Fransen
J
van Riel
PLCM
Laan
RFJM
Influence of age and gender on the 28-joint disease activity score (DAS28) in rheumatoid arthritis
Ann Rheum Dis
 , 
2008
, vol. 
67
 (pg. 
1127
-
31
)
10
Bland
JM
Altman
DG
Statistical methods for assessing agreement between two methods of clinical measurement
Lancet
 , 
1996
, vol. 
i
 (pg. 
307
-
10
)
11
Cohen
J
A coefficient of agreement for nominal scales
Educ Psychol Meas
 , vol. 
20
 (pg. 
37
-
46
)
12
Cohen
J
Weighted kappa; nominal scale agreement with provision for scaled disagreement or partial credit
J Psychol Bull
 , 
1968
, vol. 
70
 (pg. 
213
-
20
)
13
Crowson
CS
Rahman
MU
Matteson
EL
Which measure of inflammation to use? A comparison of erythrocyte sedimentation rate and C-reactive protein measurements from randomised clinical trials of golimumab in rheumatoid arthritis
J Rheumatol
 , 
2009
, vol. 
36
 (pg. 
1606
-
10
)

Author notes

*See the acknowledgements for the members of the YEAR Consortium.

Comments

0 Comments