Comparison of Composite Measures of Disease Activity in Psoriatic Arthritis Using Data From an Interventional Study With Golimumab

Authors


  • ClinicalTrials.gov identifier: NCT00265096.

  • Janssen provided the data for this post hoc analysis and was responsible for the original data collection in the GO-REVEAL Study.

Abstract

Objective

To compare the performance of the Psoriatic Arthritis Disease Activity Score (PASDAS), the Arithmetic Mean of the Desirability Function (AMDF), the Composite Psoriatic Disease Activity Index (CPDAI), and the Disease Activity Index for Psoriatic Arthritis (DAPSA) in the GO-REVEAL data set. The Disease Activity Score using 28 joints (DAS28) was used as a comparator.

Methods

The GO-REVEAL study did not allow full computation of all the composite scores (a modified version of CPDAI was used). The performance of the scores at baseline and followup (weeks 14 and 24) was compared using effect sizes.

Results

All indices could distinguish response to treatment at 14 and 24 weeks. Effect sizes at 24 weeks for the 50 mg (100 mg) golimumab doses were 2.18 (2.36), 2.08 (2.36), 1.09 (1.41), 1.80 (1.78), and 1.13 (1.18) for PASDAS, AMDF, modified CPDAI, DAS28, and DAPSA, respectively. Comparison of 24-week values across the 3 treatment groups (placebo, golimumab 50 mg, and golimumab 100 mg) by an analysis of covariance using the baseline values as covariates gave the following F statistics: PASDAS 18.3, AMDF 19.6, modified CPDAI 9.4, DAS28 13.6, and DAPSA 7.9; all of these are highly significant. When the analysis was confined to the 2 golimumab treatment groups, there were no significant between–treatment group differences with any of the composite measures.

Conclusion

PASDAS and AMDF were better able to distinguish treatment effect, having larger effect sizes at 24 weeks. PASDAS, AMDF, and modified CPDAI better reflected domains such as skin, enthesitis, and dactylitis.

INTRODUCTION

Psoriatic arthritis (PsA) is a heterogeneous disease, characterized by involvement of skin and nails, peripheral joints, entheses, and axial joints. Involvement in these different domains varies substantially among individuals, and can also vary over time in individual patients. All can impact patient quality of life (QOL). To assess disease activity in heterogeneous conditions such as PsA, and also to assess changes in disease activity with time, composite measures may be particularly useful. A composite measure is one way of assessing all relevant clinical outcomes in one single instrument. By definition, it incorporates several dimensions of disease status, often by combining these different domains into a single score. Composite measures can potentially provide a summary outcome for different groups of signs and symptoms at a specific time point. Composite measures focusing on various aspects of peripheral arthritis are well established in rheumatoid arthritis (RA). As PsA may be characterized by significant peripheral arthritis with semblance to RA, composite measures developed and validated for RA have been adopted for use in clinical trials involving patients with PsA ([1]). These include the American College of Rheumatology (ACR) responder index ([2]) and the Disease Activity Score for 28 joints (DAS28) ([3]). The ACR responder index measures improvement in tender and swollen joint counts plus improvement in at least 3 of the following 5 measures: acute-phase reactant, patient global assessment of disease activity by visual analog scale (VAS), physician global assessment of disease activity by VAS, pain by VAS, and physical function using the Health Assessment Questionnaire (HAQ). The ACR criteria for 20%, 50%, and 70% improvement scores refer to ≥20%/50%/70% improvements in these measures ([2]).

In PsA, the number of joints assessed should optimally include a 68 tender joint count and a 66 swollen joint count, which include the distal interphalangeal (DIP) joints of the fingers. The DAS28 in RA includes 28-joint tender and swollen counts, patient global, and either erythrocyte sedimentation rate (ESR) or C-reactive protein (CRP) level. The 28-joint count excludes the DIP joints of the fingers, as well as the ankles and feet. Although the DAS28 has been shown to be capable of distinguishing between patients with PsA treated with anti–tumor necrosis factor (anti-TNF) agents from those receiving placebo, it was noted that 25% of the patients would not have been included in this study because the primary joints involved were below the knees, which are not assessed as part of the DAS28 ([1]). Further, in cases of oligoarthritis, use of the DAS28 can misclassify 20% of cases, as shown in a cross-sectional data set ([4]).

A number of additional composite measures for assessing disease activity in PsA have been proposed. A composite measure for defining “minimal disease activity” (MDA) has been validated and includes assessments of joints, skin, and entheses, as well as assessment of physical function ([5]). The MDA criteria define a low disease state and can be used as a responder index in addition to a target for treatment interventions. Other disease-specific measures have been suggested. An adaptation of the Disease Activity Index for Reactive Arthritis, renamed the Disease Activity Index for Psoriatic Arthritis (DAPSA), was developed from a clinical cohort of PsA patients and validated using PsA clinical trial data ([6]). A domain-based approach has been proposed with the development of a composite measure known as the Composite Psoriatic Disease Activity Index (CPDAI) ([7]). In the CPDAI, disease involvement is assessed in up to 5 domains: peripheral joints, skin, entheses, dactylitis, and spinal manifestations. For each domain, instruments are used to assess both the extent of disease activity as well as the effect of involvement in that domain on patient function and health-related QOL. Domains were originally scored 0–3, with empirical cutoffs for disease severity/activity proposed in each domain, based on available data from the literature and consensus of expert opinion.

Recently, several specific analyses have been conducted to evaluate the appropriateness of composite disease activity measures. The GRACE (GRAPPA Composite Index Exercise) project has enabled the development of 2 new composite indices: the Psoriatic Arthritis Disease Activity Score (PASDAS) and the Arithmetic Mean of the Desirability Function (AMDF) ([8]). The PASDAS is a weighted index incorporating patient and physician global VAS scores, tender and swollen joint counts, dactylitis and enthesitis, health-related QOL, and CRP level. The AMDF is a composite index based on score transformations, incorporating measures of skin and joint involvement together with physical function. Data from the GRACE study have enabled an initial comparison of these measures.

Well-designed studies in PsA are scarce, so data from interventional clinical trials, such as the GO-REVEAL trial ([9]), offer an excellent opportunity to assess, validate, and further develop new composite outcomes measures in PsA. The main objective of this study was to compare the performance of the 2 new measures (PASDAS and AMDF) with the other proposed composite measures (CPDAI, DAPSA, and DAS28) and to examine their relationship to other measures of outcome, including ACR responses and MDA assessment, using data from the GO-REVEAL study.

Box 1. Significance & Innovations

  • New composite measures in psoriatic arthritis (PsA) assess all aspects of the disease in one index. This is in contrast to unidimensional articular indices such as the Disease Activity Score in 28 joints, which is borrowed from rheumatoid arthritis (RA).
  • An interventional study database was utilized to compare the performance of new composite indices for PsA and to assess their performance against RA indices.
  • The new indices that assess multiple facets of PsA, including joints, skin, enthesitis, dactylitis, and spine, perform better in statistical terms than traditional joint-only indices.
  • None of the instruments, either the new composites or the purely articular, can distinguish between different doses of golimumab in this database.

MATERIALS AND METHODS

These analyses used data from the GO-REVEAL study ([9]). Briefly, the GO-REVEAL study was a randomized placebo-controlled trial of golimumab in 405 patients with active, predominantly polyarticular PsA. The definition of active PsA included the presence of at least 3 swollen and 3 tender joints and the presence of plaque psoriasis with a qualifying lesion at least 2 cm in diameter. Patients were required to have active disease despite treatment with disease-modifying drugs, but prior treatment with biologic drugs was prohibited. Randomization was to placebo, golimumab 50 mg every 4 weeks, or golimumab 100 mg every 4 weeks. Patients not achieving a 10% reduction in swollen and tender joint count at week 16 were rerandomized (placebo to golimumab 50 mg, golimumab 50 mg to golimumab 100 mg, and golimumab 100 mg remained on the same treatment) to the final evaluation at week 24. Data from a random 80% sample of all GO-REVEAL patients (n = 324) at 3 study time points (baseline, 14 weeks, and 24 weeks) were used in this analysis. Individual patient data were analyzed. Only available data were used; there were no imputations for random missing data in this analysis. The data described below were used to calculate the composite measures.

DAS28.

Calculation of this composite requires a 28 tender and swollen joint count, patient global VAS, and CRP level/ESR. DAS28 (CRP) was calculated in this analysis (http://www.das-score.nl/das28/en/).

PASDAS

The PASDAS was calculated as previously described ([8]). The following variables were used: patient global VAS (rescaled from 0–10 to 0–100), physician global VAS (rescaled from 0–10 to 0–100), 66 swollen joint count, 68 tender joint count, CRP level (rescaled from mg/dl to mg/liter), enthesitis (measured in GO-REVEAL as modified Maastricht Ankylosing Spondylitis Enthesitis Score and rescaled to a 0–6 range by multiplying by a factor of 6/15 for this analysis), tender dactylitis count (the GO-REVEAL study scored each digit from 0–3 and these were recoded to 0–1, where any score >0 equaled 1), and, finally, the physical component summary (PCS) scale of the Short Form 36 (SF-36) health survey. The PASDAS is then given by the formula:

display math

AMDF

The AMDF was calculated by transforming the following variables, using predefined algorithms and expressing the total score as a mean with a score range of 0–1, where 1 indicates a better state than 0 ([8]): swollen joint count (66), tender joint count (68), patient joints VAS (using data for patient pain VAS and rescaled from 0–10 to 0–100), patient global VAS (rescaled from 0–10 to 0–100), Psoriasis Area and Severity Instrument (PASI; 0–72), and the HAQ (0–3). However, since the VAS for skin was not collected in GO-REVEAL this component of the AMDF was omitted (as the index is an arithmetic mean, this omission does not affect the score range of 0–1). For the PsAQOL a transformation algorithm was used:

display math

where MCS is the mental component summary scale. This equation was derived from the GRACE data set ([8]) using linear regression in which the R2 value was 0.804 and the Pearson's correlation between actual and predicted PsAQOL was 0.89.

ACR response scores

These were calculated as a percentage decrease from baseline in tender and swollen joint counts, together with a percentage decrease in at least 3 of 5 other outcomes (CRP level, HAQ score, physician global VAS, patient global VAS, and patient pain VAS scores).

MDA

A patient was deemed to be in MDA ([5]) if 5 of the 7 criteria were met: patient pain VAS ≤15 mm (0–100-mm scale), patient global VAS ≤20 mm (0–100-mm scale), HAQ score ≤0.5, tender joint count ≤1, swollen joint count ≤1, enthesitis ≤1, and PASI score ≤1, or body surface area ≤3%.

CPDAI

The CPDAI ([7]) assesses 5 domains (joints, skin, entheses, dactylitis, and spine). Within each domain a score (range 0–3) is assigned according to predefined cutoffs. The scores for each domain are then added together to give a final score range of 0–15. In the GO-REVEAL data set, data on axial involvement were not collected so the current (modified) version of CPDAI has a score range of 0–12.

CPDAI scores were calculated using the following assessments: joints (66 swollen/68 tender joint counts), HAQ score, PASI, dactylitis (number of tender digits), and enthesitis (using the rescaled 0–6 scale). The Dermatology Life Quality Index (DLQI) was not collected in the GO-REVEAL data set and the following equation was used to estimate the DLQI:

display math

In this equation MCS is the mental component summary scale of the SF-36 and PCS is the physical component summary scale of the SF-36. This equation was derived from the GRACE data set ([8]) using linear regression. Within the GRACE data set the correlation between the actual CPDAI and the CPDAI using the derived DLQI gave a Pearson's correlation coefficient of 0.99.

As this version of the CPDAI did not include the axial domain, and since one component of the skin domain was derived using data from another data set, we have described this as a modified CPDAI (mCPDAI).

DAPSA

DAPSA was calculated as the sum of the following components: tender joint count (0–68), swollen joint count (0–66), CRP level (mg/dl), patient VAS for pain (0–10), and patient VAS for global disease activity (0–10).

Statistics

Patients were analyzed according to their original group allocation. The relative performance of each index was compared using within-group paired t-tests and the calculation of effect size (ES; the difference between the first and second score divided by the SD of the first score) and standardized response mean (SRM; the mean difference of the 2 scores divided by the SD of the difference) at weeks 14 and 24. Although the magnitude of the ES and SRM is used purely for comparative purposes in this work, the following values are generally accepted: “trivial” (<0.20), “small” (≥0.20 to <0.50), “moderate” (≥50 to <0.80), and “large” (≥0.80). Composite change scores were also compared according to the achievement of ACR responses and the achievement of MDA at week 24. ACR responses were categorized as follows: ACR0: <20% improvement in ACR response measure; ACR20: from 20–49% improvement in ACR response measure; ACR50: from 50–69% improvement in ACR response measure; and ACR70: ≥70% improvement in ACR response measure. The magnitude of change between ACR categories was compared using an independent samples T statistic and across all categories with a one-way analysis of variance. Between-treatment comparisons of change in all indices at weeks 14 and 24 were evaluated with analysis of covariance (ANCOVA) models with baseline as covariate; these comparisons are evaluated with the F statistic, which is the ratio of variance between and within contrasted groups. Correlations between the indices and various continuous disease measures used Spearman's correlation coefficient assuming non-normality. Exploratory stepwise regression models were used to evaluate predictors of change in composite scores using P values of 0.15 to enter the model and 0.10 to stay in the final model.

RESULTS

The GO-REVEAL study treatment allocation of our sample at baseline was as follows: n = 90 for placebo, n = 125 for golimumab 50 mg, and n = 109 for golimumab 100 mg. At week 14 patients in the placebo group with inadequate response could “escape” to active treatment, which meant final per-treatment allocation group numbers were n = 41 for placebo, n = 97 for golimumab 50 mg, and n = 87 for golimumab 100 mg. This analysis was undertaken with data at baseline, 14 weeks, and 24 weeks using subjects who had completed the followup as per treatment allocation. Baseline, week 14, and week 16 values for individual clinical and laboratory variable(s) are shown in Supplementary Table 1 (available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.22204/abstract).

Table 1 gives the baseline, 14-week, and 24-week values for each of the composite indices for both doses of golimumab, together with the ES and SRMs at each time point. The mean changes in each index are also depicted in Figure 1. ES ranged from 1.47 to 2.36 according to composite index and dose of golimumab. On the whole, ES and SRMs were higher for golimumab 100 mg subjects, as might be expected. Additionally, ES and SRMs were higher for the newer composite measures such as PASDAS and AMDF compared to the articular-based measures DAPSA and DAS28.

Table 1. Scores for composite indices at baseline, 14 weeks, and 24 weeks for each dose of golimumab*
 Mean ± SDEffect sizeSRM
BaselineWeek 14Week 24Week 14Week 24Week 14Week 24
  1. Only patients with complete data and in original allocated randomization group are included. SRM = standardized response mean; PASDAS = Psoriatic Arthritis Disease Activity Score; AMDF = Arithmetic Mean of the Desirability Function; mCPDAI = modified Composite Psoriatic Disease Activity Index; DAS28 = Disease Activity Score in 28 joints; DAPSA = Disease Activity Index for Psoriatic Arthritis.
Golimumab 50 mg, no.12512197    
PASDAS6.01 ± 1.293.54 ± 1.553.20 ± 1.571.922.181.862.04
AMDF0.47 ± 0.130.71 ± 0.170.74 ± 0.181.852.081.771.93
mCPDAI10.98 ± 5.205.86 ± 4.635.31 ± 4.700.981.091.051.11
DAS285.00 ± 1.103.13 ± 1.223.02 ± 1.271.701.801.691.69
DAPSA50.54 ± 28.3821.10 ± 20.8318.54 ± 19.061.041.131.341.41
Golimumab 100 mg, no.10910887    
PASDAS5.96 ± 1.083.72 ± 1.693.41 ± 1.482.072.361.522.13
AMDF0.46 ± 0.110.68 ± 0.180.72 ± 0.172.002.361.642.00
mCPDAI11.62 ± 4.496.75 ± 4.855.29 ± 4.351.081.411.101.42
DAS284.91 ± 1.053.37 ± 1.233.04 ± 1.141.471.781.391.94
DAPSA47.37 ± 24.2524.04 ± 20.0418.68 ± 16.120.961.181.251.56
Figure 1.

Mean values for each composite measure at baseline, 14 weeks, and 24 weeks for golimumab 100 mg (solid line), golimumab 50 mg (patterned line), and placebo (broken line). PASDAS = Psoriatic Arthritis Disease Activity Score; AMDF = Arithmetic Mean of the Desirability Function; CPDAI = Composite Psoriatic Disease Index; DAPSA = Disease Activity Index for Psoriatic Arthritis; DAS28 = Disease Activity Score in 28 joints.

ACR and MDA responses

Differences in composite scores according to ACR category and MDA category at 24 weeks are given in Tables 2 and 3. Significant differences between ACR groups are seen for all comparisons for PASDAS; but only 2 comparisons are seen for AMDF and DAS28, only 1 comparison for mCPDAI, and none of the comparisons for DAPSA are seen. These comparisons are more noticeable for the higher golimumab dose. For MDA categories only the PASDAS, AMDF, and DAS28 achieve significant differences for the 50-mg dose of golimumab, whereas only DAPSA fails to achieve significance for the 100-mg dose of golimumab.

Table 2. Differences between baseline and week 24 scores for each composite index according to ACR response*
 ACR0ACR20aACR50bACR70cFd
  1. ACR = American College of Rheumatology; ACR0 = <20% improvement in ACR response measure; ACR20 = 20–49% improvement in ACR response measure; ACR50 = 50–69% improvement in ACR response measure; ACR70 = ≥70% improvement in ACR response measure; PASDAS = Psoriatic Arthritis Disease Activity Score; AMDF = Arithmetic Mean of the Desirability Function; mCPDAI = modified Composite Psoriatic Disease Activity Index; DAS28 = Disease Activity Score in 28 joints; DAPSA = Disease Activity Index for Psoriatic Arthritis.
  2. aIndependent sample t-test compares scores with ACR0.
  3. bIndependent sample t-test compares scores with ACR20.
  4. cIndependent sample t-test compares scores with ACR50.
  5. dStatistic for one-way analysis of variance across ACR groups for each composite index.
Golimumab 50 mg, no.27241625 
PASDAS    22.18
Mean ± SD1.70 ± 1.142.43 ± 0.973.29 ± 0.664.01 ± 1.19 
t 2.093.292.12 
P 0.0430.0020.041 
AMDF    26.93
Mean ± SD−0.15 ± 0.11−0.25 ± 0.09−0.28 ± 0.06−0.41 ± 0.11 
t 3.330.914.11 
P 0.0020.370.0001 
mCPDAI    8.83
Mean ± SD2.93 ± 4.775.21 ± 4.105.63 ± 4.068.64 ± 5.09 
t 1.800.321.99 
P 0.080.750.05 
DAS28    27.53
Mean ± SD0.92 ± 0.891.92 ± 0.922.24 ± 0.533.10 ± 0.96 
t 3.471.733.20 
P 0.0010.0920.003 
DAPSA    5.10
Mean ± SD21.22 ± 19.4230.82 ± 18.0936.19 ± 18.3344.06 ± 28.30 
t 1.810.900.96 
P 0.080.380.34 
Golimumab 100 mg, no.24201821 
PASDAS    29.88
Mean ± SD1.65 ± 0.842.14 ± 0.732.84 ± 0.914.04 ± 0.85 
t 1.982.544.15 
P 0.0550.0160.0001 
AMDF    35.87
Mean ± SD−0.15 ± 0.08−0.20 ± 0.09−0.30 ± 0.08−0.041 ± 0.09 
t 1.963.643.72 
P 0.570.0010.001 
mCPDAI    2.81
Mean ± SD4.46 ± 4.275.85 ± 4.378.17 ± 4.067.86 ± 4.70 
t 1.071.690.22 
P 0.290.100.83 
DAS28    16.98
Mean ± SD1.18 ± 0.731.56 ± 0.552.27 ± 0.862.76 ± 0.90 
t 1.843.061.69 
P 0.070.0040.10 
DAPSA    2.55
Mean ± SD21.78 ± 18.6926.18 ± 11.0433.16 ± 20.5235.65 ± 19.57 
t 1.011.350.38 
P 0.320.190.71 
Table 3. Differences between baseline and week 24 scores for each composite index according to MDA response*
 −MDAa+MDAbtcP
  1. Values are the mean ± SD unless indicated otherwise. MDA = minimal disease activity; PASDAS = Psoriatic Arthritis Disease Activity Score; AMDF = Arithmetic Mean of the Desirability Function; mCPDAI = modified Composite Psoriatic Disease Activity Index; DAS28 = Disease Activity Score in 28 joints; DAPSA = Disease Activity Index for Psoriatic Arthritis.
  2. aNot in MDA.
  3. bIn MDA.
  4. cIndependent samples t-test.
Golimumab 50 mg, no.4346  
PASDAS2.46 ± 1.483.14 ± 1.20−2.360.02
AMDF−0.22 ± 0.15−0.31 ± 0.11+3.320.001
mCPDAI4.66 ± 5.556.47 ± 4.22−1.750.08
DAS281.70 ± 1.252.30 ± 1.07−2.460.02
DAPSA33.75 ± 25.7731.58 ± 20.75−0.440.66
Golimumab 100 mg, no.4044  
PASDAS1.88 ± 0.843.26 ± 1.14−6.190.0001
AMDF−0.18 ± 0.09−0.33 ± 0.11+6.530.0001
mCPDAI4.36 ± 3.918.32 ± 4.31−4.360.0001
DAS281.36 ± 0.822.37 ± 0.84−5.500.0001
DAPSA26.18 ± 18.4330.77 ± 18.15−1.140.26

ANCOVA, correlation, and regression models

Mean values for each composite index for each treatment group are shown in Figure 1. Comparison of 24-week values across the 3 treatment groups by ANCOVA using the baseline values as covariates gave the following F statistics for the PASDAS (F = 18.3), AMDF (F = 19.6), mCPDAI (F = 9.4), DAS28 (F = 13.6), and DAPSA (F = 7.9); all F values were highly significant. When the analysis was confined to the 2 golimumab treatment groups there were no significant between–treatment group differences for any of the composite measures at 14 and 24 weeks.

Correlation between composite indices and other variables are given in Table 4. Correlation coefficients >0.5 are highlighted, although values >0.11 are significant at the 5% level. Correlations between indices are high, particularly between PASDAS and AMDF, and between DAS28 and DAPSA. Reflecting this, correlations between articular scores such as tender and swollen joint count are higher for the DAS28 and DAPSA. On the other hand, for PASDAS and AMDF correlations are high across a wider spectrum of variables, including enthesitis, dactylitis, function, and QOL. For mCPDAI correlations are moderate across all the variables.

Table 4. Spearman's correlations between composite indices and other measured variables at baseline*
 PASDASAMDFmCPDAIDAS28DAPSA
  1. PASDAS = Psoriatic Arthritis Disease Activity Score; AMDF = Arithmetic Mean of the Desirability Function; mCPDAI = modified Composite Psoriatic Disease Activity Index; DAS28 = Disease Activity Score in 28 joints; DAPSA = Disease Activity Index for Psoriatic Arthritis; VAS = visual analog scale; HAQ = Health Assessment Questionnaire; PASI = Psoriasis Area and Severity Instrument; CRP = C-reactive protein; SF-36 = Short Form 36 health survey; MCS = mental component summary scale; PCS = physical component summary scale.
  2. aIndicates value >0.5.
PASDAS −0.82a0.69a0.62a0.62a
AMDF−0.82a 0.69a0.64a0.65a
mCPDAI0.69a0.69a 0.470.49
DAS280.62a0.64a0.47 0.87a
DAPSA0.62a0.65a0.490.87a 
Tender joint count0.440.470.430.77a0.92a
Swollen joint count0.400.350.280.62a0.77a
VAS pain0.64a0.76a0.370.440.43
VAS global0.72a0.81a0.410.490.42
Physician VAS global0.69a0.54a0.370.440.45
Dactylitis score0.360.060.340.030.10
Enthesitis score0.51a0.450.55a0.460.55a
HAQ score0.64a0.76a0.64a0.480.42
PASI0.020.170.21−0.06−0.04
CRP0.340.170.240.320.11
SF-36 MCS−0.260.480.300.180.19
SF-36 PCS−0.71a0.64a0.50a0.360.33

Stepwise multivariate regression also reflected the above findings (see Supplementary Table 2, available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.22204/abstract). For PASDAS score change was largely predicted by patient and physician global VAS, although significant contributions were also made by the PCS scale of the SF-36, the dactylitis and enthesitis scores, the tender joint count, and the PASI. For the AMDF the most significant contributors to the final model were the patient global VAS, the HAQ score, the tender joint count, the PASI, the VAS for pain, and both subscales of the SF-36. For mCPDAI the most significant contributors were the HAQ score, the PASI, and the enthesitis and dactylitis scores. For DAS28 virtually all the variance was provided by 3 variables: tender joint count, patient global VAS, and CRP level. The DAPSA score had a similar result with tender and swollen joint counts, patient pain VAS, and CRP level as the major contributors.

DISCUSSION

PsA is a complex condition affecting both articular and nonarticular structures. For example, skin and nail involvement is a significant contributor to functional loss and poor QOL. Ideally, outcome measures used for assessing PsA in both clinical trials as well as in the clinic would capture disease involvement and activity across these heterogeneous domains. To date, clinical trials have focused largely on peripheral joint activity, using measures developed for RA. In PsA, even within the musculoskeletal system, important features such as spondylitis, enthesitis, and dactylitis clearly add to the burden of disease and are not captured in simplified assessments of peripheral joints. Therefore, a composite measure assessing varied domains may be of value in PsA.

This analysis is a validation of several candidate composite measures for PsA and has allowed a comparison of their performance in a trial of a highly effective intervention. All indices were able to distinguish treatment from placebo, with better statistical discrimination for the PASDAS and AMDF. Of note, none of the measures could successfully distinguish between the 2 doses of golimumab at 14 and 24 weeks. That the PASDAS and AMDF appeared to perform somewhat better than the other composite indices is probably a reflection of the wider spectrum of domains included in each of these indices. The PASDAS, a weighted composite in which the greatest weights are placed on both physician and patient global scores, also has contributions from joints, skin, dactylitis, enthesitis, acute-phase response, and health-related QOL, all important aspects of disease for patients and physicians. The AMDF does not contain weighting but incorporates equal contributions from domains that include the joints, skin and global assessments, function, and QOL. In contrast, the only other composite that includes domains other than the joints is the CPDAI. The slightly lesser performance of the CPDAI compared to the PASDAS and AMDF might possibly relate to the way the CPDAI uses predefined cutoffs for severity in each domain, a strategy that may result in a “blunting” of responses to treatment, although the CPDAI performed well in another trial data set comparing 2 doses of etanercept in PsA ([10]). In contrast, largely articular composites such as DAS28 and DAPSA work very well and are clearly discriminatory, but appear to be less powerful than some of the other composite indices in statistical terms, reflected by lower scores for ES and SRM. This is relevant as these statistical considerations affect sample sizes necessary for future trial design, an area of great importance to patients, companies, and regulators.

This study has several potential limitations. The data were derived from an interventional trial of golimumab in PsA. Because TNF inhibitors are highly efficacious across the manifestations of PsA, it is possible that examination of the performance of these measures in studies of less effective agents, or with agents that had differential effects on distinct domains of disease, may yield different results. This will hopefully be addressed by further analyses similar to this one being performed in other studies, particularly future trials. Also in this analysis, slight modifications were made to some of the instruments used to accommodate available collected data. These very minor changes are unlikely to have had substantial impact on the performance, but it does merit consideration as one considers assessment of measures from various sources, including different clinical trials or registries. Another potential limitation is that enrollment of patients into this clinical trial was predicated upon a level of active peripheral arthritis, among the other inclusion and exclusion criteria. In other populations of PsA patients, some of whom may not have met the entry criteria for this study (e.g., with less active peripheral arthritis), the measures may have performed differently. This concern is common when analyzing clinical trial data, and is of importance in the consideration of adoption of outcome measures from clinical trials into the clinic. In that regard, issues such as feasibility of outcome measures are relevant and will need to be further assessed. It is also relevant when considering alternative populations such as PsA patients with early disease. Hopefully these types of analyses will be extended to other populations in the future.

At what point have we reached in the development of these new composite indices for PsA? In terms of the Outcomes Measures in Rheumatology, filter aspects of truth, discrimination, and feasibility have been tested in this analysis and the new composite indices have demonstrated both face, content, and convergent validity and, of course, response to change. Although individual components of the new indices have been subjected to reliability studies ([11, 12]) further work is needed using the indices themselves. As indicated above, more data are needed in people with less severe disease, predominant skin disease, and in early arthritis. It would also be helpful to develop cutoffs for disease activity states and responses that can be used to compare outcomes in the current and future data sets. At this stage it is probably premature to decide which of the composite indices should be chosen as the definitive measure for future studies.

If it is worth collecting the additional data for the composite indices, then we need to be clear about the situation in which the additional data are to be collected. Should the new composite indices only be used in clinical trials? At the moment the answer to this is probably in the affirmative, but with further use and evolution it is possible that a shorthand version of the composites will be developed for use in the clinic situation.

In summary, although all composite measures tested in this analysis were able to clearly discern a treatment response in patients treated with golimumab for active psoriasis and PsA, the PASDAS and AMDF were better able to distinguish between the treatment groups and had much bigger ES at 24 weeks. The articular component of the disease was equally reflected by all composite scores, but the PASDAS, AMDF, and CPDAI better reflected other domains such as skin, enthesitis, and dactylitis.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Helliwell had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Helliwell, Kavanaugh.

Acquisition of data. Helliwell, Kavanaugh.

Analysis and interpretation of data. Helliwell, Kavanaugh.

ROLE OF THE STUDY SPONSOR

Janssen had no role in the study design or in the analysis or interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication. Janssen reviewed the submitted manuscript; publication of this article was not contingent upon approval by Janssen.

Ancillary