A main objective for patients with rheumatoid arthritis (RA) is to reach a health state they find acceptable or satisfactory. Therapeutic advances over recent years provide new opportunities,1 and more ambitious therapeutic goals can be achieved. One such goal is to target a level of remission with disease-modifying antirheumatic drugs (DMARDs).2 3 Thus, the current perception is that “for the patient it is good to feel better but it is better to feel good”.4

Measures for stringent evaluation of lower levels of disease activity that are coherent to the patient perspective are needed. The recently proposed concept of a “patient acceptable symptom state” (PASS) implies that results from clinical studies can be expressed as proportions—that is, the proportion of patients that reach or do not reach this state. Tubach et al have previously developed cut-off points for commonly used health status measures in knee and hip osteoarthritis5 and ankylosing spondylitis (AS)6 corresponding to PASS. Their methodology made use of a global “yes/no” question on perceived satisfactory health as an external anchor. Cut-off points for PASS have not been identified in RA, and exploratory analyses have previously not been done in the context of longitudinal observational studies of patients undergoing treatment with DMARDs.

Several composite indices and single measures are used to assess disease activity in RA.7^–10 Thresholds for remission, low and moderate disease activity have been proposed for the commonly used composite indices.10^–13 Agreement between some of these composite indices, patient-reported outcomes and investigators global assessment has been found to be unsatisfactory.14 Even though PASS has been suggested as an outcome criterion,15 the way in which this concept relates to already existing thresholds of low levels of disease activity has not been examined.

The objectives of this study were to identify new PASS cut-off points for composite indices and patient-reported outcomes and to examine the agreement between new PASS cut-off points and the established threshold values for remission, low and moderate disease activity in composite indices.

PATIENTS AND METHODS

Patients

Patients were recruited from NOR-DMARD, a Norwegian five-centre treatment register and longitudinal observational study that started in December 2000.16 All prescriptions of DMARDs in patients with inflammatory arthropathies are included, and patients are followed up longitudinally with assessments after 12, 24 and 52 weeks, and then yearly. As of August 2006, 5824 DMARD cases had been included. For the cross-sectional analyses we included patients with RA as well as patients who had responded to the global yes/no question on perceived satisfactory condition at baseline and at the 24-week follow-up visit (n = 1496). Before inclusion in the study, each patient signed an informed consent. The study had been evaluated by the regional ethical committee.

For examination of the longitudinal stability of the PASS cut-off points, we focused on a subgroup of these patients (n = 843 cases) who responded to the global yes/no question on perceived satisfactory health at baseline, 12, 24 and 52 weeks.

Anchoring question

The global question on perceived satisfactory health was formulated: “Is your current condition satisfactory, when you take your general functioning and your current pain into consideration?”.6 This question was used as an external indicator of satisfactory condition (PASS) with dichotomised yes or no response options.

Measures of disease activity

Composite indices

The patients were assessed with core measures of disease activity (28-swollen joint count (28-SJC), 28-tender joint count (28-TJC), erythrocyte sedimentation rate, C-reactive protein (CRP), patient and investigator global assessments of disease activity on a visual analogue scale (VAS), and intensity of joint pain on a VAS).

Commonly used composite indices were computed. The Disease Activity Score-28 (DAS28) includes 28-SJC and 28-TJC in addition to patient global assessments of disease activity on a VAS and erythrocyte sedimentation rate.7 Established thresholds for remission, low and moderate disease activity are 2.6, 3.2 and 5.1, respectively.11

The Simplified Disease Activity Index (SDAI)8 employs a linear sum of five untransformed, unweighted variables, including 28-SJC and 28-TJC, patient and investigator global assessments of disease activity on a VAS, and CRP. Predefined thresholds for remission, low and moderate levels of disease activity are 3.3, 11 and 26, respectively.12

The Clinical Disease Activity Index (CDAI)9 is a modification of the SDAI without laboratory evaluation (CRP) to allow immediate clinical assessment. Thresholds for separating remission, low and moderate levels of disease activity are at 2.8, 10 and 22, respectively.13

The Patient Activity Scale (PAS)10 is a validated composite patient-reported disease activity scale, comprising VAS scores of joint pain intensity and patient global and Health Assessment Questionnaire (HAQ) scores.17 The PAS score is computed as the average of each component score converted into a range of 0–10. A PAS score ⩽1.625 is defined as low disease activity.10

Global scales, pain and fatigue

VAS were used to measure investigator and patient global assessment of disease activity as well as intensity of joint pain and fatigue.

Functional measures

The panel of functional measures consisted of Modified Health Assessment Questionnaire (MHAQ)18 and the Medical Outcome Study Short Form-36 (SF-36).19 20

HAQ comprises 20 questions organised into eight categories (dressing, rising, eating, walking, hygiene, reach, grip and usual activities), with a score range of 0–3 (3 worst health). MHAQ is a modification of the HAQ, with only one item within each of the eight categories; the MHAQ ranges from 1 to 4. Our HAQ scores were derived from the MHAQ according to an algorithm that has been validated in 1000 patients with RA.21

SF-36 with 36 items and eight scales is the most commonly used generic health status measure with score range of 0–100 (0 worst health).19 20 Data on the role scales (role physical, role mental) are not presented since these scales were not adequate as continuous measures in the procedures used for identifying cut-off points.

Analyses

The demographic characteristics, values of composite indices and health status measures at baseline and 24 weeks are presented as mean (SD) for continuous variables, and as percentages for categorical values.

New PASS cut-off points for the composite indices and health status measures were identified by the method used by Tubach et al.5 The PASS cut-off points are defined according to the 75th centile of the values in patients with self-reported satisfactory condition.

Since we had a fairly large number of patients, we also took the opportunity to identify PASS cut-off points by using a receiver operating characteristic (ROC) analysis as a complementary approach, as also proposed by Maksymowych et al.22 New cut-off points using ROC analyses were identified with the maximum accuracy approach (best trade-off between sensitivity and specificity) for each variable.

To examine the robustness of the identified cut-off points, we repeated the analyses in subgroups, focusing on disease duration, age and gender. We also examined the stability of the cut-off points longitudinally after 12, 24 and 52 weeks’ treatment with DMARDs.

We cross-tabulated the proportions of cases with disease activity below existing thresholds for remission, low and moderate disease activity in composite indices against the proportions with values below and above the identified cut-off points for composite indices at the 24-week follow-up visit. Agreement between various disease activity categories and the PASS cut-off points was calculated by assessing accuracy (ie, the percentage of congruently classified patients divided by all patients) and quantified using κ statistics.

Analyses were performed with SPSS version 14 and SAS version 9.13.

RESULTS

Patients

Table 1 shows the demographic characteristics of the 1496 cases that were selected for this study (fig 1), the percentage of patients who responded yes to the PASS question, and also mean levels of disease activity and health status levels at baseline and after 24 weeks. The proportion of patients starting with anti-tumour necrosis factor regimens at baseline was 30.0%; initial methotrexate monotherapy, combination regimens, leflunomide, sulfasalazine and other DMARD treatments amounted to 41.4%, 12.0%, 6.7%, 7.0% and 2.9%, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Selection of cases included in the 24-week analyses. DMARDs, disease-modifying antirheumatic drugs; PASS, patient acceptable symptom state; RA, rheumatoid arthritis; ROC, receiver operating characteristic.

New PASS cut-off points

A comparison of cut-off points corresponding to PASS showed that those determined by employing the 75th centile approach, were generally higher than those identified by ROC curve analyses. The PASS cut-off points for PAS was 3.56 with the 75th centile approach and 3.28 with ROC in these analyses (table 2). Area under the curve was highest for PAS, patient global and joint pain (0.82, 0.83 and 0.80, respectively; table 2).

PASS cut-off points in subgroups

The cut-off points were robust across subgroups, but there was a tendency towards higher cut-off points for disease activity and health status (lower for SF-36) in female versus male patients, in patients with short versus long disease duration and in younger versus older patients (data not shown).

In the longitudinal analyses, focusing on patients responding to PASS after 12, 24 and 52 weeks, PASS cut-off points were relatively stable, with small fluctuations within the same moderate range of disease activity level (table 3). Interestingly, PASS cut-off points for patient-derived measures were very stable over time, as seen for example for patient global, joint pain, functional assessment and PAS. On the other hand, PASS cut-off points for the investigator global assessment, as well as DAS28, SDAI and CDAI fell slightly over time.

Agreement between PASS cut-off points and disease activity levels

Best agreement between PASS cut-off points and established thresholds for disease activity levels in composite indices was observed for low disease activity level for SDAI and CDAI (accuracy 73.8%, 74.6; κ = 0.70, 0.72, respectively). Weaker agreement was observed for remission than for moderate disease activity in DAS28, SDAI and CDAI (table 4).

DISCUSSION

Previous studies have identified PASS cut-off points for a variety of measures that are commonly used in clinical studies, both in AS,6 osteoarthritis5 and rotator cuff syndrome.23 Our approach was to identify new PASS cut-off points and to examine the agreement with existing thresholds of composite indices, in the setting of a longitudinal observational study of patients with RA starting treatment with DMARDs.16

In all analyses, after 24 weeks and in the longitudinal perspective, PASS cut-off points were in the range of the moderate disease activity level across all composite indices (tables 2 and 3). Agreement between PASS cut-off points and levels of disease activity in frequently used composite indices was best for the level of low disease activity and better for moderate disease activity than for remission (table 4).

For the estimation of PASS cut-off points, we used the validated 75th centile method as the primary tool,5 but also performed ROC analyses to study the robustness of the cut-off points across approaches. The 75th centile approach reflects the 75% level of sensitivity in a group of patients reporting satisfactory condition, whereas the ROC analyses reflect levels giving the highest accuracy in individual response to the global question of satisfactory condition. The identified cut-off point differed in the direction of lower cut-off points with the ROC analyses, but remained within the range of moderate disease activity level. Lower cut-off points with ROC analyses versus the 75th centile approach has recently also been described in a study focusing on AS.22 ROC analyses may represent the most precise estimate because of the individual approach in these analyses, and because cut-off points defined by ROC analyses are based on maximum values for sensitivity and specificity. Nevertheless, in studies of smaller patient numbers or low area under the curve, the 75th centile approach may give the better estimates.

We observed that the PASS cut-off points of investigators’ global VAS assessments of disease activity were consistently lower than those reported on VAS by patients. Thus, while patient global assessments of disease activity of acceptable symptom state amounted to 35–37 mm, it ranged between 19 and 26 mm for the evaluators’ global assessment (tables 2 and 3). Patient-reported outcomes (PROs) were more stable over time than investigator-based measures (table 3). These observations may reflect the different perspectives and expectations of doctors and patients about disease activity and health status.24

Identified cut-off points for patient-reported measures on VAS were comparable to previous findings. Tubach et al found that PASS cut-off points for pain and patient global VAS levels after 12 weeks of treatment in AS were 34.5 and 35.7 mm, respectively.6 The cut-off points in knee and hip osteoarthritis after 4 weeks’ treatment were between 32 and 35 mm.5 In our study the values were 35.0 mm for joint pain VAS and 37.0 mm for patient global VAS (table 2).

Since most composite indices include joint counts, laboratory values and patient and/or investigator global assessment, PROs do not constitute a major part. PAS is a composite score based purely on PROs. The high level of area under the curve for PAS (0.82) in the ROC analyses supports the high correspondence of this measure with PASS. However, the identified PAS value corresponding to PASS was as high as 3.56 with the 75th centile approach and 3.28 in the ROC analyses, which is far above the PAS score of 1.625 that has been suggested as a threshold for low disease activity.10 Importantly, we also found a poor agreement between PAS low disease activity level and the new PASS cut-off points corresponding to satisfactory condition (table 4). PAS is the only composite measure that includes pain. Pain also has significant impact on patient global assessment VAS, one of the two other components of PAS.25 The strong association between pain outcomes and PASS (table 2) as well as our previous observations,26 highlight that pain is an important measure when the patient perspective on outcome assessment is taken into account. For the HAQ the identified PASS cut-off point value was 1.04 (with 75th centile), which is similar to the average HAQ score found in large patient populations.27

Good remission criteria and refined instruments to enable discrimination at lower levels of disease activity are needed to monitor patients according to new ambitious treatment goals.28 We found better agreement between the new PASS cut-off points and SDAI and CDAI low disease activity state than DAS28 low disease activity (table 4). This result may support the observation that the SDAI and CDAI low disease activity levels are closer than DAS28 low disease activity to the patient perception of a satisfactory condition and is in line with data suggesting that the currently used low disease activity threshold of the DAS28 may be too low.12

Although one of the limitations of our study is its observational, uncontrolled nature, observational studies do allow for analysis of a “real-world” situation and have been regarded as providing important insights and as being complementary to the randomised controlled trials.29 Moreover, the data provided here may not be definitive, but need to be validated cross-culturally in other open as well as controlled clinical trials. Importantly, data were collected prospectively in a large number of patients using many DMARDs, allowing for generalisability and extrapolation of the results into clinical practice after further validation. Further, subgroup analyses according to gender, age and disease duration were performed as were longitudinal analyses, all of which showed a general robustness of the cut-off points. The cut-off points were also similar in the subgroup of patients who reported that they were in an unsatisfactory condition at baseline (data not shown).

PASS has been referred to as the value beyond which patients consider themselves well5 or feeling good.23 However, the external anchoring PASS question is about whether or not the patients consider their condition as satisfactory. The PASS cut-off points of disease activity levels identified in the current study (table 2) exceeded the remission and low disease activity state. In line with this observation, in another study, the PASS threshold for the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) has been identified as 4.8,22 although 4.0 has been used as the primary inclusion criterion for disease activity warranting institution of anti-tumour necrosis factor therapy after standard treatment has failed.30 The residual disease activity for PASS is located within a range of disease activity that will lead to progression of joint damage and irreversible disability, if not treated.31 32 PASS accentuates the importance of taking patient-reported measures into consideration in treatment decisions, but it should not be used as the ultimate therapeutic goal, which has to remain low disease activity or remission if joint damage progression is to be minimised or halted. However, an intermediate state as identified through the current PASS question may also be useful to report in trials and to communicate with patients.