Abstract
Background. An index is needed to assess the status of patients with rheumatoid arthritis (RA), as none of the existing measures are applicable to all individual patients. The 28-joint Disease Activity Score (DAS28) is the most specific and widely used index. Routine Assessment of Patient Index Data (RAPID3) is an index containing only the 3 patient self-report core dataset measures, without a laboratory test or formal joint count, and with simple scoring. RAPID3 is correlated significantly with DAS28, but calculated in 5–10 seconds on a Multidimensional Health Assessment Questionnaire (MDHAQ), compared to 114 seconds for DAS28.
Methods. DAS28 (0–10 scale) categories for high, moderate, and low activity, and remission (≤ 2.6, 2.6–3.2, 3.21–5.1, and > 5.1, respectively) and proposed RAPID3 (0–30 scale) categories for severity (0 ≤ 3, 3.1–6, 6.1–12, and > 12) were compared in patients taking abatacept and control-treated patients at the endpoint of the Abatacept in Inadequate Response to Methotrexate (AIM) and the Abatacept Trial in Treatment of Anti-TNF INadequate Responders (ATTAIN) clinical trials, using cross-tabulations and kappa statistics.
Results. Overall, 92%–99% of patients classified as having high DAS28 activity had high or moderate RAPID3 severity, while 64%–83% in DAS28 remission had RAPID3 low severity or remission; 50%–82% of patients with good or poor EULAR responses had good or poor RAPID3 responses. Kappa values ranged from 0.25 to 0.48, and weighted kappas from 0.32 to 0.52, indicating fair to moderate agreement for the 2 indices.
Conclusion. Proposed RAPID3 severity and response categories yield comparable results to DAS28 and EULAR criteria in AIM and ATTAIN. DAS28 is more specific for clinical trials. RAPID3 does not preclude also scoring DAS28, and may be informative in the infrastructure of routine care.
- RHEUMATOID ARTHRITIS
- ABATACEPT
- 28-JOINT DISEASE ACTIVITY SCORE
- ROUTINE ASSESSMENT OF PATIENT INDEX DATA
- MULTIDIMENSIONAL HEALTH ASSESSMENT QUESTIONNAIRE
Assessment of patients with rheumatoid arthritis (RA) requires a pooled index of various measures1, as no single measure is adequate to characterize clinical status in all individual patients. The most widely used index is the 28-joint Disease Activity Score (DAS28)2,3, which includes a tender joint count (TJC), a swollen joint count (SJC), an acute-phase reactant laboratory test [erythrocyte sedimentation rate (ESR) or C-reactive protein (CRP)], and patient global estimate of status.
The DAS28 is the most specific measure to assess patients with RA, but it has several limitations for usual care4,5: (1) a laboratory test (ESR or CRP) is needed, which often is not available at the time of the patient visit, and is normal in about 40% of patients6,7,8; (2) complex scoring is required, albeit with an excellent available Website (www.das-score.nl/ www.das-score.nl/index.html); (3) a formal 28-joint TJC and SJC must be scored, which is not performed at most visits in usual care9 as it involves about 90 seconds10; formal joint counts cannot be assessed in a general medical setting, home, or outside a specific rheumatology setting, and results vary considerably when performed by different rheumatologists11,12.
Routine Assessment of Patient Index Data (RAPID3) is an index of only 3 patient self-report core dataset measures (physical function, pain, and patient global estimate of status), with simple scoring, and no laboratory test or formal joint count. RAPID3 is calculated on a Multidimensional Health Assessment Questionnaire (MDHAQ) in 5–10 seconds13, compared to 114 seconds for DAS28 and 106 seconds for a Clinical Disease Activity Index (CDAI)14. RAPID3 is correlated significantly with DAS28 in studies in clinical care15.
RAPID3 is designed to be calculated on an MDHAQ16,17, but can also be calculated from the standard HAQ18, from which it was developed (8 items on the MDHAQ are identical16). However, RAPID3 from standard HAQ requires 42 seconds versus 5 seconds for RAPID3 on an MDHAQ13. RAPID3 incorporating the standard HAQ is correlated significantly with DAS28 in clinical trials of methotrexate19, leflunomide19, adalimumab20, abatacept21, and certolizumab22, and distinguishes active from control treatment similarly. While the specificity of DAS284 renders it optimal for clinical trials, studies to compare DAS28 and RAPID3 in clinical trials can provide useful information to clinicians in interpretation of RAPID3 scores in usual clinical care23,24,25.
We extended observations of correlations between DAS28 and RAPID3 with further post-hoc analyses of data from the Abatacept in Inadequate response to Methotrexate26 (AIM) trial and the Abatacept Trial in Treatment of Anti-TNF INadequate Responders27 (ATTAIN). We compared 4 categories of DAS28 versus 4 RAPID3 categories for activity severity: high, moderate, and low, and remission. We also compared European League Against Rheumatism (EULAR)-DAS28 improvement criteria to proposed RAPID3 improvement criteria in 3 categories: good, moderate, and poor.
MATERIALS AND METHODS
Clinical trials
Post-hoc analyses were performed on databases from the AIM26 and ATTAIN27 clinical trials. All 7 RA core dataset measures28,29,30 were analyzed at baseline and at endpoint, 12 months in AIM and 6 months in ATTAIN, using last observation carried forward. Further details concerning these clinical trials and post-hoc analyses are found in previous reports21,26,27.
Indices
DAS28 has a score range of 0–10 (Table 1)3,31. RAPID3 is designed to be scored in usual care, and includes only the 3 patient-reported RA core data set measures28 — physical function, pain, and patient global estimate of status. Each is scored 0–10, for a total range of 0–30 (Table 1)15,21. In earlier reports, 0–30 RAPID3 scores were recalculated to 0–10, dividing by 3. However, the raw 0–30 score requires about 5 seconds compared to 10 seconds for a recoded score, and is now recommended13. In our study, physical function was computed from the standard HAQ Disability Index (HAQ-DI) with standard 0–10 visual analog scale (VAS), although 0–3 HAQ-DI scores were converted to 0–10, as each of the 3 measures is weighted equally. DAS28 and RAPID3 scores were computed using the standard formulas rather than a Website or scoring templates.
Features of DAS28 and RAPID3 to assess patients with rheumatoid arthritis (RA).
The term “activity” is used to describe DAS28 categories, and the terms “severity” and “near-remission” to describe RAPID3 categories, as patient questionnaire scores are sensitive to both activity and damage32. Criteria for high, moderate, and low activity, and remission for DAS28 (0–10) are > 5.1, > 3.2–5.1, ≥ 2.6–3.2, and < 2.6, respectively (Table 1)33. Proposed criteria for high, moderate, and low severity and near-remission for RAPID3 (0–30) are > 12, > 6–12, > 3–6, and ≤ 3, respectively (Table 1)15,21.
EULAR-DAS28 response criteria categories are classified as good = decrease > 1.2 units AND endpoint score ≤ 3.2; moderate = decrease > 1.2 units AND endpoint score ≥ 3.2, OR decrease of 0.6–1.2 units AND endpoint score ≤ 5.1; poor = decrease < 0.6 units, OR endpoint score > 5.1 (Table 1)34. Proposed RAPID3 response criteria categories are classified as good = decrease > 3.6 units (on a 0–30 scale) AND endpoint score of < 6; moderate = decrease of > 1.8 and ≤ 3.6 units AND endpoint score ≤ 12, OR decrease of > 3.6 units AND endpoint score ≥ 6 to ≥ 12; poor = decrease of < 1.8 units OR endpoint score > 12 (Table 1).
Statistical analyses
DAS28 and RAPID3 were compared at baseline and endpoint (52 weeks in AIM, 26 weeks in ATTAIN) in patients who had been randomized to abatacept or control treatment, using cross-tabulations. The proportions of patients in categories of high, moderate, and low activity, and remission for DAS28 and RAPID3 severity were compared in each arm of each trial. Almost all patients had severe status at baseline, and only data at endpoint are presented. EULAR-DAS28 and proposed RAPID3 response criteria were compared with cross-tabulations for the proportions of good, moderate, and poor responses, with statistical significance analyzed according to kappa and weighted kappa statistics35.
RESULTS
Comparisons of RAPID3 and DAS28 activity/severity categories at the conclusion of the AIM trial
DAS28 remission and low, moderate, and high activity scores at the conclusion of AIM were seen respectively in 17%, 10%, 53%, and 20% of patients randomized to abatacept + methotrexate, versus 2%, 2%, 40%, and 56% randomized to placebo + methotrexate (Table 2). RAPID3 remission, low, moderate, and high severity scores at the conclusion of AIM were seen in 16%, 19%, 33%, and 33% of patients randomized to abatacept + methotrexate, versus 5%, 10%, 28%, and 56% of patients randomized to placebo + methotrexate (Table 2).
Number (%) of patients in the AIM trial in RAPID3 versus DAS28 disease activity/severity categories at 52 weeks.
Low or remission versus high or moderate severity was seen for abatacept patients in 27% versus 73% according to DAS28, and 35% versus 66% according to RAPID3, and for control patients in 4% versus 96% according to DAS28, and 15% versus 85% according to RAPID3.
Overall, 78%–80% of patients classified as having high disease activity according to DAS28 had high disease severity according to RAPID3, and 95%–99% high or moderate severity according to RAPID3 (Table 2). By contrast, 46%–50% classified as in remission according to DAS28 were in RAPID3 remission, and 64%–75% were in RAPID3 low severity or remission (Table 2). Kappa and weighted kappa statistics ranged from 0.28 to 0.42, indicating fair to moderate agreement between DAS28 and RAPID3 categories.
Comparisons of RAPID3 and DAS28 activity/severity categories at the conclusion of the ATTAIN trial
DAS28 remission and low, moderate, and high activity at the conclusion of the ATTAIN trial were seen respectively in 10%, 6%, 34%, and 50% of patients randomized to abatacept + methotrexate, versus 1%, 2%, 16%, and 81% of patients randomized to placebo + methotrexate (Table 3). RAPID3 remission and low, moderate, and high severity at the conclusion of ATTAIN were seen in 11%, 14%, 21%, and 55% of patients randomized to abatacept + methotrexate, versus 2%, 5%, 12%, and 81% of patients randomized to placebo + methotrexate (Table 3).
Number (%) of patients in the ATTAIN trial in RAPID3 versus DAS28 disease activity/severity categories at 26 weeks.
Overall, 76%–89% of patients classified as having high DAS28 activity had high RAPID3 severity, and 92%–97% high or moderate severity according to RAPID3 (Table 3), while 33% classified as in remission according to DAS28 were in RAPID3 remission, and 83% in RAPID3 remission or low activity. Kappa and weighted kappa statistics ranged from 0.25 to 0.40, indicating fair agreement between DAS28 and RAPID3 categories.
EULAR-DAS28 and proposed RAPID3 response criteria in the AIM trial
In the AIM trial, EULAR-DAS28 criteria indicated good, moderate, and poor responses in 27%, 62%, and 10%, respectively, of abatacept-treated patients compared to 4%, 59%, and 37% of control patients. Proposed RAPID3 response criteria indicated good, moderate, and poor responses in 35%, 45%, and 20% of abatacept-treated patients compared to 16%, 43%, and 42% for control patients (Table 4). Among all patients in the AIM trial, good, moderate, and poor responses were seen in 20%, 61%, and 19%, respectively, for EULAR-DAS28, compared to 29%, 44%, and 27% for RAPID3 (Table 4).
Number (%) of patients in the AIM trial in RAPID3 response categories as compared to EULAR-DAS28 response categories.
Among 102 abatacept-treated patients with good EULAR-DAS28 responses, 67 (66%), 29 (28%), and 6 (6%) had good, moderate, and poor responses according to RAPID3 (p < 0.001; kappa 0.35; weighted kappa 0.40; Table 4). Among patients randomized to control treatment, 68 had poor responses according to EULAR-DAS28 criteria, of whom 46 (68%) had poor responses according to RAPID3 (p < 0.001; kappa 0.29; weighted kappa 0.36; Table 4). Among all patients, 109 had good responses according to EULAR-DAS28, of whom 72 (66%) had good, 31 (28%) moderate, and 6 (6%) poor responses according to RAPID3 (Table 4). Among 341 patients who had moderate responses according to EULAR-DAS28 criteria, 186 (55%) had moderate, 85 (25%) good, and 70 (21%) poor responses according to RAPID3 (Table 4). Among 107 patients who had poor responses according to EULAR-DAS28 criteria, 75 (70%) had poor, 28 (11%) moderate, and 4 (4%) good responses according to RAPID3 (p < 0.001; kappa 0.35; weighted kappa 0.43; Table 4).
EULAR-DAS28 and proposed RAPID3 response criteria in the ATTAIN trial
In ATTAIN, good, moderate, and poor responses were seen in 17%, 56%, and 28%, respectively, of abatacept-treated patients for EULAR-DAS28 criteria, compared to 4%, 30%, and 66% in control patients. Proposed RAPID3 good, moderate, and poor responses were seen in 24%, 45%, and 31% of abatacept-treated patients compared to 6%, 32%, and 62% of control patients (Table 5). Among all patients in the ATTAIN trial, good, moderate and poor responses were seen in 12%, 46%, and 41%, respectively, for EULAR-DAS28 compared to 18%, 40%, and 42% for RAPID3 (Table 5).
Number (%) of patients in the ATTAIN trial in RAPID3 response categories as compared to EULAR-DAS28 response categories.
Of 30 abatacept-treated patients with good EULAR-DAS28 responses, 20 (67%) had good, 6 (20%) moderate, and 4 (13%) poor responses according to RAPID3 (p < 0.001; kappa 0.37; weighted kappa 0.44; Table 5). Among patients randomized to control treatment, 65 had poor EULAR-DAS28 responses, of whom 53 (82%) had poor, 11 (17%) moderate, and 1 (2%) good RAPID3 responses (p < 0.001; kappa 0.48; weighted kappa 0.52; Table 5). Among all patients in the ATTAIN trial, 34 had good responses according to EULAR-DAS28, of whom 22 (65%) had good, 8 (24%) moderate, and 4 (12%) poor responses according to RAPID3 (Table 5). Among 129 patients with moderate responses according to EULAR-DAS28, 75 (58%) had moderate, 26 (20%) good, and 28 (22%) poor RAPID3 responses (Table 5). Among 115 patients with poor responses according to EULAR-DAS28, 85 (74%) had poor, 29 (25%) moderate, and 1 (1%) good RAPID3 responses (p < 0.001; kappa 0.44; weighted kappa 0.51; Table 4).
DISCUSSION
The results indicate fair to moderate agreement between DAS28 and RAPID3, both for categories of high, moderate, and low activity/severity and remission, as well as for EULAR-DAS28 response criteria and proposed RAPID3 response criteria. Lower levels of both remission and good responses are seen according to RAPID3 compared to DAS28, in both abatacept and control patients. This finding may reflect, in part, that RAPID3 is more sensitive than DAS28 to fibromyalgia or joint damage, resulting in higher scores for pain and global status, in the absence of overt inflammation.
At the same time, the RAPID3 criteria may be more stringent, and patients who have a pain score or global estimate > 3 may not consider their disease in remission, even if they have no swollen and tender joints. A patient who has a TJC and SJC of 0, an ESR of 20, but a patient global estimate of 3 will have a DAS28 of 2.5, indicating remission. Conversely, a patient with no swollen joints, ESR of 18 (indicative of no inflammation), but patient global estimate of 10 and 28 tender joints (indicative of fibromyalgia) would have a DAS28 of 6.4 and CDAI of 38 (as well as RAPID3 of 20), all suggesting high disease activity.
Any clinical measure or index must be regarded as one component of the data needed for clinical decisions, and should not be the only variable that triggers a response from a clinician. Consider ESR, which, for example, might be elevated to 75 mm/h from a normal value 3 months earlier. This finding could signify a flare of RA, but also an infection or neoplasm, which must be excluded prior to initiating new RA therapy. Similarly, DAS28, RAPID3, or any index may provide valuable information to help guide management of RA, but should not be a sole determinant of therapeutic decisions. A complete medical history, physical examination, appropriate laboratory assessment, and ancillary studies should be performed at every clinical encounter to provide an optimal database for clinical decisions.
Discordance of DAS28 and RAPID3 is seen in certain patients, reflected in kappa levels in the range of 0.2–0.4, indicating fair to moderate agreement. This is common with most measures in clinical medicine. For example, both ESR and CRP are valid measures of inflammation, but these measures are not invariably congruent, being correlated with one another at about rho = 0.515. About 15% of patients with a high ESR have a normal CRP, while about 15% have a high CRP and normal ESR8. RAPID3 is correlated with DAS28 and CDAI at about rho = 0.6–0.715, higher levels than the correlation of ESR with CRP, although nonetheless reflecting incomplete concordance.
RAPID3 certainly is not regarded as replacing a careful joint examination. On the contrary, such an examination is necessary for a diagnosis of RA and should be included in every visit to a rheumatologist. Nonetheless, a formal quantitative joint count may not be necessary to guide clinical decisions. Involvement of, say, 12–14 versus 1–2 swollen joints may be very important in treatment decisions, but whether a patient has 12 versus 14, or 1 versus 2 swollen joints generally will not affect clinical decisions. A careful examination without a formal, quantitative joint count is generally adequate to recognize 12–14 versus 1–2 swollen joints.
We describe clinical trial data concerning RAPID3 based on a HAQ, rather than on an MDHAQ13,15,36. The MDHAQ was adapted from the HAQ for usual clinical care over 25 years16,17,37,38, and has evolved to differ in many respects: (1) 10 (vs 20) items for physical function, adding complex activities that better reflect status of patients at this time than 3 decades earlier; (2) scoring templates for physical function; (3) 21-circle visual analog scale for pain and patient global estimate of status, as well as fatigue (rather than 10-cm lines, so that a ruler is not needed); (4) self-report RA disease activity index (RADAI)39 joint count; (5) review of systems; (6) medical history self-report; (7) queries regarding depression, anxiety, sleep quality, morning stiffness, exercise status, and smoking status; and (8) demographic measures. Nonetheless, the HAQ and MDHAQ physical function scores are highly correlated, as 8 of the 10 physical function scale items on the MDHAQ are identical to 8 of the 20 HAQ items16,17. The similarity of results to distinguish active from control treatment in clinical trials according to RAPID3 or DAS28 is presented here to reassure clinicians that RAPID3 data reflect the more traditional DAS2813,15.
RAPID3 appears to be a useful, feasible index for busy clinical settings. It approximates information found in DAS28 for categories of activity/severity and response criteria. The MDHAQ and RAPID3 do not replace a standard medical history and physical examination, including a careful joint examination. The MDHAQ enhances capacity to focus on concerns of the patient more directly, and can save time for the rheumatologist. A review of systems and recent medical history within 2 pages on the MDHAQ help improve doctor-patient communication, saving time and focusing the patient on a more productive visit40. We suggest that all rheumatologists consider distribution of an MDHAQ at every visit of every patient with any rheumatic disease, and calculating RAPID3 as an adjunct to clinical decisions in the infrastructure of usual rheumatology care.
Footnotes
-
Supported by a research grant from Bristol-Myers Squibb to T. Pincus. Dr. Pincus has received research support from Amgen, Bristol-Myers Squibb, and UCB; and has acted as a consultant to Bristol-Myers Squibb, Centocor, and UCB. Y. Yazici and M. Bergman have acted as consultants to Bristol-Myers Squibb.
- Accepted for publication August 3, 2011.
