Materials and methods

Domains to be included

Ideally, any criteria need to contain the minimum of measurements (to allow feasibility) but must incorporate measures of all of the important aspects of the disease. The starting point was the OMERACT agreed core set of domains for PsA.7 However, HRQOL was excluded after a preliminary analysis of PsA cases with a low disease activity (defined as physician global disease activity visual analogue scale (VAS) of <3) showed no correlation between HRQOL and other disease activity measures (results not given). As recommended by OMERACT, peripheral joint activity was measured using the 68 tender/66 swollen joint count. Skin activity was initially planned to be represented by the Psoriasis Activity and Severity Index (PASI) score as this is the most commonly used outcome measure for psoriasis. However, concern was raised by dermatologists within the Group for Research and Assessment of Psoriasis and Psoriatic Arthritis (GRAPPA) that the PASI was not valid in patients with low psoriasis disease activity (personal communication, Dr G Krueger), and therefore both PASI and body surface area (BSA) were included. This allows physicians to use either PASI or BSA depending on their preference. Pain and patient’s global assessment of disease activity were measured using 100 mm VAS scales, and the health assessment questionnaire (HAQ) was used as a measure of physical function.

The core domains do not cover aspects of PsA such as dactylitis, enthesitis or spinal disease. The concern was that activity in any of these domains would not be recognised by the MDA criteria, allowing a person to be incorrectly classified as being in MDA. Dactylitis can be identified as part of a peripheral joint count where dactylitic digits will result in a quantifiable tender and/or swollen joint count. Therefore, further measurement of dactylitis was not felt necessary. Enthesitis was felt to be an important measure that may not be accurately identified using any of the other core measures. A decision was therefore made to include an enthesitis count. Multiple enthesitis counts have been proposed for clinical research, some “borrowed” from research in other spondyloarthropathies, but there is no clear consensus in PsA. As such, a raw enthesitis count was included with a maximum value of 13 representing the maximum score of the commonly used measures in PsA. This allows the criteria to be used with any available enthesitis outcome measures.

Spinal disease is particularly under-researched in PsA, and no validated disease activity measures are available. The Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) has been evaluated in PsA and was not shown to correlate with doctor’s perception of disease activity or with treatment decisions.8 Interestingly, it was found to correlate highly with the patient’s global disease activity VAS in both axial and peripheral joint disease, which is a core set measure in PsA and was already included in the profiles. Given the lack of outcome measures available, no specific measurement of axial disease activity was included. It was felt that active axial disease would be identified by the patient’s assessment of pain, global disease activity and physical function.

A questionnaire was compiled consisting of 40 PsA patient profiles using measures agreed above (tender joint count: 0–68, swollen joint count: 0–66, enthesitis count: 0–13, PASI: 0–72, BSA: 0–100, patient VAS pain: 0–100, patient VAS global disease activity: 0–100 and HAQ: 0–3). The profiles were created using data from real patients with PsA selected from a prospective longitudinal database. The profiles included a full range of disease activity (see table 1 and profiles available in table 6 in online supplement), but with additional patients with low disease activity (based on physicians’ assessments).

The questionnaire included the OMERACT definition of MDA and told the respondent: “This patient comes to see you in clinic. He/she has been on a stable dose of disease modifying anti-rheumatic drug (DMARD) therapy for over 6 months.” They were then asked: “Do you consider this patient to be in a ‘minimal disease activity’ state?” It was then circulated to rheumatologists and dermatologists with an interest in PsA, identified through membership of GRAPPA.9

Statistical analysis

A 70% agreement was selected as a reasonable consensus and therefore profiles were classified as MDA if there was ⩾70% agreement between respondents. Mann-Whitney U tests were used to test for a difference between values of each variable for those patients in MDA and those not. Regression analysis was used to investigate the strength of association of different variables with the state of MDA. Summary statistics for each outcome measure, using the profiles of patients agreed to have MDA, were calculated including mean, rounded mean, upper limit of the 95% confidence interval (CI), rounded upper limit of the 95% CI and maximum. These were investigated as potential cut-off points for each of the outcome measures. The number of outcome measures that could not exceed this cut-off point for the patient to remain in a MDA state was then investigated. There were seven outcome measures creating seven variations on the number of cut-off points that must be met (1/7, 2/7, 3/7, 4/7, 5/7, 6/7 or 7/7 where n/7 indicates that n or more of the outcome measures has a result at or below the cut-off point). This created 35 possible definitions for MDA when each of the five possible cut-off points was combined with the seven variations in the number of cut-off points met to reach the definition. For each definition, sensitivity, specificity and area under the receiver operating characteristic curve (ROC) for the prediction of MDA were calculated. From these the best definitions were selected. All statistical analysis was performed using SPSS Version 12.0 for Windows.

Results

A total of 60 respondents (82% rheumatology, 18% dermatology background) completed the questionnaire. There was substantial consensus between the experts on the profiles that represented MDA. There was agreement of ⩾90%, ⩾80% and ⩾70% that 7, 10 and 13 profiles, respectively, were in MDA. Summary statistics are shown for each of these levels of agreement and there is little variation in the cut-off points for these different levels (table 2).

Thirteen cases were therefore taken as being in MDA and the other 27 cases were considered not to be in MDA. When considering the different specialists separately, dermatologists designated only 10 cases as MDA compared with 13 designated by rheumatologists. As a result, the summary statistics derived from their MDA profiles were slightly lower than those agreed by the rheumatologists (table 3).

When considering the MDA cases compared with the remainder not classified as in MDA, there was a significant difference (p<0.05) between the values for all of the variables with the exception of enthesitis (p = 0.159). Individual regression analysis showed that all variables except enthesitis were predictors of MDA. There were few cases with active enthesitis, potentially explaining its lack of significance.

Using the 13 cases designated as MDA by the majority of all of the respondents, five cut-off points were calculated for each core measure (table 4).

For any final criteria, the cut-off points for each variable (eg, tender joint count) must be integers to allow application to patient data. The maximum value, rounded mean and rounded upper limit of the 95% confidence interval were therefore considered preferentially. The definitions based on maxima included outlying values (in particular, patient global disease activity, VAS 87 and enthesitis 6) that were felt to be clinically unrepresentative of MDA. The definitions based on maximum values were therefore not considered as ideal criteria. Application of all of the five cut-off points with seven variations depending on the number of cut-off points achieved yielded 35 definitions which were tested using ROC curves. The ROC curves generally showed high sensitivity and specificity and an area under the curve (AUC) ⩾0.895 for all definitions (table 5 and fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Receiver operator characteristic curve for candidate definitions of minimal disease activity.

The best criteria based on the rounded mean cut-off points was that with 5/7 criteria met (sensitivity 92%, specificity 89%, AUC 0.946). Two possible criteria using the rounded 95% upper limit (AUC 0.974) were identified with 5/7 criteria met (sensitivity 100%, specificity 82%) and 6/7 criteria met (sensitivity 85%, specificity 96%).

When considering these three candidate definitions, thought was given to the ideal characteristics of the criteria. Although both sensitivity and specificity were important, it was felt that a relatively low specificity would be more problematic. This is because it would be preferable to miss a few patients who were truly in MDA than to misclassify patients with significant disease activity into an MDA group where they may potentially be undertreated. The definition based on meeting 5/7 of the cut-off point criteria based on the 95% upper limit was therefore discarded as it had a lower specificity.

The final decision was between the two highest performing candidate definitions using the rounded mean with 5/7 cut-off point criteria met (sensitivity 92%, specificity 89%) or the rounded 95% CI with 6/7 criteria met (sensitivity 85%, specificity 96%). Again there was concern that one of the rounded 95% CI cut-off points may be too high to truly represent MDA (global disease activity VAS 40) and so the final definition selected used the rounded mean values.

A core set definition therefore places patients with PsA in MDA when they meet 5/7 of the following criteria:

• Tender joint count ⩽1

• Swollen joint count ⩽1

• PASI ⩽1 or BSA ⩽3

• Patient pain VAS ⩽15

• Patient global activity VAS ⩽20

• HAQ ⩽0.5

• Tender entheseal points ⩽1

Discussion

The aim of this study was to produce clinical criteria for MDA in PsA, and it provides the first measure of clinical “state” in PsA. Such measures defining remission, low and high disease activity have been available in RA for some time and have been shown to be important for use in research trials and in daily clinical practice.

Observational data in PsA suggest a link between inflammation and damage. It has been shown that clinical and radiological joint damage is related to the number of actively inflamed joints as a time-varying indicator.10 11 Therapeutic trials of highly effective biological therapies such as tumour necrosis factor (TNF) blockers in PsA have shown a reduction in disease activity and a corresponding reduction in radiological progression.12 13 However, the lack of criteria or cut-off points for remission or low disease activity in PsA has meant that the concept of “tight control” of disease activity using objective measures has never been evaluated.

This study is based on a questionnaire using profiles of patients sent to experts in PsA including both rheumatologists and dermatologists. This raises some limitations in the research. It is not known in PsA whether judgements about these profiles or “paper patients” are truly reflective of clinical assessments. However, in RA, research has shown that there is a high correlation between judgements made on “paper patients” and those based on real patients seen in clinical practice.14 Our profiles also contained more information than that given in the validation study in RA14 to counteract the problem of a more complex and heterogeneous disease.

A link to the questionnaire was sent to all members of GRAPPA with a request for their participation and a follow-up email was sent 2 weeks later. This recruited 60 respondents and the response rate is similar to that of previous GRAPPA membership surveys. The online tool used to complete the questionnaires is familiar to many GRAPPA members and allowed respondents to link to the questionnaire on repeated occasions if they did not have time to complete it in one sitting. It also ensured that respondents could not answer the questionnaire more than once. A lower proportion of dermatologists completed the questionnaire, probably because dermatologists constitute a minority of the experts who were contacted. However, analysis has not shown a marked difference in the responses. The possible cut-off points for skin disease (rounded mean, rounded 95% upper limit and maximum) were similar in the assessments of both groups.

The number of profiles included is limited in these questionnaire-based exercises by the number of profiles that physicians can be expected to rate reliably, and a compromise at 40 was reached. Obviously the results of this study are based on a particular sample of 40 patients and 60 physicians and the results might be different, to some extent, when the study would be replicated in larger and distinct samples of patients and physicians. The relatively small number of cases also means that multivariate logistic regression could not be performed to compare the contribution of each core measure towards the MDA status. This could be performed in future in large distinct cohorts to evaluate the contribution of these different core measures. The proposed criteria need external validation in distinct cohorts before they can be recommended as an outcome measure in future trials. In particular, the performance of the proposed criteria must be analysed in a variety of clinical presentations including oligoarticular, polyarticular and axial disease. Work is underway with further validation following the principles of the OMERACT filter.15

Importantly, this analysis of expert opinion represents current views and these may change in the future. The OMERACT definition of MDA is “that state of disease activity deemed a useful target of treatment by both the patient and physician, given current treatment possibilities and limitations”. This relationship to current treatment possibilities means that acceptable targets of treatment are likely to have changed rapidly in the last few decades as new therapies have become available. Studies in RA have shown a lowering target for remission and low disease activity based on expert opinion over time, and have suggested that this is related to the increasing potential to achieve these states with newer therapies. Similarly, in future the acceptable levels of disease activity may fall further as the quest for remission becomes increasingly important.

In summary, this study provides the first definition of a “state” of disease activity in PsA and defines a target for treatment. It is based on current expert opinion and uses a composite of key outcome measures in PsA to encompass all of the domains of the disease. Given the new therapeutic options in PsA, patients are likely to benefit if disease activity states are targeted rather than simply looking for a subjective clinical response.