Article Text
Abstract
Objective Specific composite indices assessing disease activity in psoriatic arthritis (PsA) have not yet been sufficiently validated. The objective of this study was to identify instruments best reflecting disease activity in PsA.
Methods Measures for inclusion in clinical trials, as recommended by the OMERACT-7/8 PsA workshops, were assessed. A principal component analysis (PCA) was performed with cross-sectional data of 105 patients with PsA to identify a minimal set of important dimensions for a disease activity assessment tool for PsA. This was compared with components contained in existing composite indices.
Results The PCA revealed four principal components best reflecting disease activity. The first contained patient global and pain assessment; the second contained 66/68 swollen and tender joint counts as main variables; C-reactive protein (CRP) best loaded to the third component; and the fourth was loaded by skin assessment but did not reach significance. When comparing the three significant principal components with items of established composite measures, they were best covered by the Disease Activity Index for Reactive Arthritis (DAREA) which comprises patient pain and global assessments, 66/68 joint counts and CRP.
Conclusion Among the currently available indices used in arthritic conditions, the DAREA best reflects the domains found to be important in PsA.
Statistics from Altmetric.com
Introduction
Psoriatic arthritis (PsA) is an inflammatory joint disease associated with psoriasis, presenting as peripheral arthritis often accompanied by spinal involvement and enthesitis. Different instruments are employed for disease evaluation,1 most focusing on individual characteristics and therefore not spanning the disease spectrum. Even for arthritis, the instruments are insufficiently validated or “borrowed” from rheumatoid arthritis (RA) assessment. In contrast to RA, arthritis in PsA is often asymmetrical and oligoarticular, frequently involving distal interphalangeal joints (DIPs). Compared with ankylosing spondylitis, in PsA spinal involvement is infrequent, commonly asymmetrical and discontinuous.2
PsA shows variable signs and symptoms between and within individual patients. Composite disease activity indices compensate for such limitation.3 In ballots on important domains at PsA workshops (Outcome MEasures in Rheumatoid Arthritis Clinical Trials OMERACT-7/8),4 5 various items received majority votes, suggesting that they partly provided different information and should therefore be contained in potential composite scores. While several composite activity indices are used in PsA,6 they were mostly not developed for PsA and encompass only a few variables with majority votes.
We thus evaluated domains important for PsA assessment regarding commonalities and disparities in their ability to generate components for derivation of a disease activity score for PsA.
Patients and methods
Consenting outpatients (n=105) classified as having PsA7 were evaluated at two follow-up visits about 3 months apart. An independent biometrician performed assessments. The study, approved by the ethical committee, did not require treatment alterations.
In a cross-sectional primary analysis, we focused on variables having ≥50% of the votes at OMERACT-74 (for details see online supplement). In a sensitivity analysis we additionally performed assessments that included all tested variables (see online supplement).
Analyses
Principal component analysis (PCA) was chosen to generate a smaller number of combinations of variables accounting for most of the variability (see online supplement). Only loadings >0.3 are demonstrated. The “Eigenvalue” should be >1.0 for each component.
Sample size
In PCA, the ratio of subjects to items is important, ≥10 cases per item being ideal. With 105 patients and 11 items, we have accounted for this recommendation.
Results
Patient characteristics
The baseline characteristics of the patients are shown in the online supplement. Table 1 presents the data included in PCA on the first and second visits.
Principal component analysis
After obtaining significance supporting the overall approach (see online supplement), PCA revealed simple structures: four components showed strong loadings (table 2).
The first component contained patient global assessment (PtGA) and pain followed by the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) and represented “patient self-reported disease activity”. The second component (“joints”) contained swollen (SJC) and tender (TJC) joint counts as the main loading variables. The third was best loaded by C-reactive protein (CRP) and erythrocyte sedimentation rate, and the fourth by the “skin factor” Psoriasis Area and Severity Index (PASI), but did not reach significance (Eigenvalue <1.0).
Thus, assessment of PsA can be “reduced” to three significant principal components: patient self-reported disease activity, joint counts and acute phase reactants (APR).
Sensitivity analyses
Change over time
All patients were reassessed after about 3 months, allowing changes in activity to be evaluated. While changes were small (table 1), repeating PCA using the differences between visits gave similar results to the primary PCA. Again, the first component contained changes in pain and PtGA, the second contained change in APR, the third was best loaded by changes in SJC and TJC and the fourth by change in PASI.
Confirmatory PCA using more variables
Additional sensitivity analysis using 22 variables (see online supplement) showed five components with Eigenvalues >1.0. The first was again loaded most strongly by PtGA and pain. The second was new, showing self-reported function as an individual component best represented by the Dougados Functional Index (DFI) and Health Assessment Questionnaire (HAQ). The third component, representing “joints”, showed higher loadings for 66/68 than 28 joint counts. Interestingly, although low, enthesitis also loaded in this but no other component. The fourth and fifth components were loaded mainly by skin (PASI) and APR (CRP). The confirmatory PCA corroborated our primary PCA, showing the same major components and variables complemented by a component on functional status.
Comparison of PCA data with disease activity and response criteria
Variables best representing the principal components should obviously be included in a possible composite disease activity index. Tables 3 and 4 show domains of various disease activity (table 3) and response measures (table 4) compared with PCA results.
The Disease Activity Score using 28 TJS/SJC (DAS28)8 comprises four variables, but not pain; it uses reduced joint counts. The Simplified and Clinical Disease Activity Indices (SDAI, CDAI) have the same limitation,9 10 CDAI also lacking APR. In contrast, the Disease Activity index for REactive Arthritis (DAREA)11 comprises variables contained in all PCA components, including 66/68 joint counts and CRP. The ACR criteria comprise five variables12: the EULAR criteria are limited by using DAS2813 and the Psoriasis Arthritis Response Criteria (PsARC)14 lack pain and APR.
Discussion
To date, there are no specific well-validated disease activity measures in PsA. Although measures “borrowed” from other rheumatic diseases are used in clinical trials, there are insufficient data showing which item composition best reflects PsA joint disease activity.
The data presented here using PCA seem to draw a clear picture of the areas important in PsA. Patient-reported disease activity represented the first component, loaded mostly by pain and PtGA and less by the BASDAI, SF-36 and HAQ. Spine involvement did not attain its own component by PCA in either our primary or confirmatory PCA. However, in this confirmatory PCA, functional assessment comprised a separate component with the DFI and HAQ achieving highest loadings. Since the DFI has not been validated in PsA while the HAQ has been used in recent clinical trials,15 16 we propose to use the HAQ in the follow-up of patients with PsA but not to include it in a composite measure as it is multifactorial and contains irreversible elements.17
Joint involvement constituted the second domain. Our confirmatory PCA showed 28 joint counts loading less strongly than 66/68 counts. Indeed, many patients with PsA have inflammation of DIPs and foot joints not captured by 28 joint counts.
The third component showed that APR loaded separately with no significant overlap with other components. Clinical trial data suggest a worse outcome in PsA with raised APR,18 and that APR are sensitive to change.16 Interestingly, in contrast to OMERACT-7, the second voting at OMERACT-8 rejected APR as an important domain;5 our data support OMERACT-7 and first OMERACT-8 votings.
The fourth component captured skin involvement with no overlap with any other domain, suggesting that severity of skin disease has no impact on arthritic activity. We certainly propose including dermatological evaluation and treatment in overall patient care, but it seems that skin involvement should be evaluated separately.
It is noteworthy that enthesitis did not load importantly to the components. This suggests that joint and entheseal pain may run different courses, do not occur together frequently, or that enthesitis may be less prevalent or less ailing than envisaged.
When looking at composite measures, the DAS28, EULAR response, SDAI and CDAI—all developed for RA—do not represent the principal components of PsA sufficiently since they use reduced joint counts and lack pain; indeed, the limitations of using DAS28 in PsA have recently been addressed.19 Also PsARC, developed for PsA, is not limited by either the inclusion of pain or APR. In contrast, ACR improvement criteria cover all components but do not reflect actual disease activity. Interestingly, DAREA, a score developed and validated for reactive arthritis and thus in a PsA-related disease,11 comprises all the major components: pain, PtGA, 66/68 SJC/TJC and CRP. Of all the available indices, DAREA therefore best reflects the domains important in PsA and may thus serve as a Disease Activity index for PSoriatic Arthritis (DAPSA). Indeed, in a preliminary analysis it showed validity and sensitivity to change in an observational cohort of patients with PsA.20
In a sensitivity analysis on changes over time, the highest loadings were exhibited by the same variables as in the original PCA. Thus, DAREA comprises the variables loading most strongly and significantly. Since mathematically it sums five variables, it is easy to perform in routine settings. Assessment of function by questionnaires such as the HAQ and assessment of skin, spine and entheseal involvement will expand information to disability and extra-articular features.
Since the joint is the primary organ involved in PsA, we feel that joint counts should be included in composite indices and not bypassed using surrogates which would not do justice to the patients or the caretaking physician. Moreover, in contrast to RA where reduced joint counts reflect the overall joint involvement well, comprehensive joint counts are needed for composite scores in PsA. Whether a composite measure also needs to comprise an APR will have to be determined separately. Clearly, the data suggest that skin, spine and entheseal involvement do not provide information on arthritic disease activity and should not be part of a composite score for PsA. However, as they capture important aspects of the disease, they should be assessed separately.
Our data suggest that developing a new score is not needed for the assessment of PsA since DAREA/DAPSA contains all the major components found to be important by PCA. Final validation in clinical trials of PsA is needed and planned.
Acknowledgments
This study was partly funded by a grant from GRAPPA (Group for Research and Assessments of Psoriasis and Psoriatic Arthritis).
References
Supplementary materials
Web Only Data ard.2009.117945
Files in this Data Supplement:
Footnotes
-
Ethics approval This study was conducted with the approval of the Medical University of Vienna.
-
Provenance and peer review Not commissioned; externally peer reviewed.