Construct and Content Validity of Instruments Used in Clinical Trials of Psoriatic Arthritis

PI: Alexis Ogdie, MD MSCE

Supplemental Table: Variables to be requested. The following variables, when available, will be requested for the baseline visit and study completion.

Trial specific / Treated or placebo
Measures of Peripheral Arthritis Activity* / Tender joint count (0-28)
Swollen joint count (0-28)
Tender joint count (0-68)
Swollen joint count (0-66)
Tender joint count (0-78)
Swollen joint count (0-76)
Measures of Radiographic Damage / Modified Total Sharp-van der Heijde Score
Erosion subscore
Joint space narrowing subscore
Laboratory Measures of Inflammation / CRP (both mg/L and mg/dL)
ESR
Physician Assessment / Physician Arthritis Global VAS
Physician Disease Global VAS
Enthesitis / Leeds Enthesitis Index
MASES (0-13)
Modified MASES (0-15)
4-point enthesitis Index
SPARCC
Dactylitis / Dactylitis Count
Dactylitis Severity score
Leeds Dactylitis Index
Spondylitis / Bath ankylosing spondylitis disease activity index (BASDAI)
Bath ankylosing spondylitis functional index (BASFI)
Bath ankylosing spondylitis metrology index (BASMI)
Psoriasis / Psoriasis Area Severity Index (PASI)
Physician Global Assessment (PGA)
Body Surface Area (BSA)
Nails / Nail Global Assessment VAS
Nail Psoriasis Severity Index (NAPSI)
Modified Nail Psoriasis Severity Index (mNAPSI)
Patient reported outcomes
Global Assessment (arthritis and skin) (VAS) / Patient assessment of skin (VAS)
Patient assessment of arthritis (VAS)
Patient global assessment of disease activity (VAS)
Stiffness / Stiffness VAS
Duration
Depression / Beck Depression Inventory
CES-D Depression Scale
Columbia Suicide Severity Rating Scale (C-SSRS)
Hospital Anxiety and Depression Scale
Other depression or suicidality scales
Pain / Patient assessment of pain (VAS)
Fatigue / Fatigue Assessment Scale
Functional Assessment of Chronic Illness Therapy (FACIT) Fatigue score
SF-36 Vitality domain
Sleep / Pittsburgh Sleep Quality Index (PSQI)
MOS Sleep Scale
HRQOL / Psoriatic Arthritis Quality of Life Index (PsAQOL)
Short Form 36 (SF36) total score
SF36 Physical Component Summary
SF36 Mental Component Summary
Individual SF36 domains:
·  physical function, role physical, bodily pain, general health, vitality, social functioning, role emotional, mental health
EQ5D
SF-6D derived from mean changes across all 8 domains of SF-36
AsQoL
Skin / Dermatology Life Quality Index (DLQI)
Psoriasis Symptom Inventory (PSI)
Itch VAS
Function / Health Assessment Questionnaire-Disability Index (HAQ-DI)
SF-36 Physical Function Domain
Work/Productivity / Work Limitations Questionnaire
Work Productivity Survey (WPS)
Valuation of Lost Productivity (VOLP) questionnaire
Work Productivity and Activity Impairment (WPAI) questionnaire
We recognize that many of these variables will not be available for individual clinical trial but these instruments will be examined whenever available.
Joint counts may be provided as the joints recorded and then AO can create variables for the 28-, 66/68- and 76/78- joint counts or alternatively, the individual scores may be provided.

Supplemental Methods/Analysis: Data analysis will be performed by Dr. Ogdie (the study PI) in Stata 13.0 (College Station, TX).

Descriptive Reporting: We will first descriptively report the number and category of instruments in each clinical trial. Note that all clinical trials will be de-identified and labeled as “A”, “B”, and so on. This will be reported in Table 1 of the manuscript. Patient level data will not be reported (although patient-level data is required for analysis).

Mapping of outcome measures to domains: All outcome measures will be mapped to one or more domains. For instruments with multiple domains, their individual domains and summary component scores will be analyzed separately.

Construct validity: The concept of construct validity is that an instrument measures what it is supposed to measure. We will establish construct validity by assessing convergent and divergent relationships of similar and dissimilar instruments pretreatment and post treatment separately. We will create a correlation matrix containing all of the outcome measures (change in measures over the course of the study). Pearson’s or Spearman’s correlation coefficients will be calculated for each cell of the matrix (depending on the parametric or non-parametric distribution of the measure respectively). Intra-domain and inter-domain correlations will then be compared. Intra-domain correlation should be higher than inter-domain correlation for an individual outcome measure. We will examine therapy categories to ensure that the correlation direction is the same for all measures but will report mean correlations among all trials included.

Responsiveness is also a part of construct validity. Responsiveness is the ability of an outcome measure to detect change over time. Correlation among change in the response measures will similarly be calculated as described above. In addition, we will calculate standardized response means over all of the treatment groups (not comparing among therapies but combining all therapies). Confidence intervals will be generated using a bootstrapping method (1000 samples with replacement).

Content validity: Content validity is the ability of an outcome measure (or set of outcome measures) to represent all facets of a condition. The PsA working group aims to measure as fully as possible patient response using available outcomes measures. We aim to determine the most complimentary group of outcome measures to best explain “response”. Thus, we will examine two different parameters for response: 1) achievement of minimal disease activity (MDA) and 2) patient global assessment of disease at study completion. MDA is defined as defined as meeting 5 of the 7 criteria: Patient Pain VAS(1-100)<15, Patient Disease Global (1-100)<20, Modified Health Assessment Questionnaire (0-3)<0.5, Tender joint count (1-68)<1, Swollen joint count (1-66)<1, Psoriasis Area Severity Index (0-72)<1 OR %BSA<3, Enthesitis sites (1-13, MASES) <1. The use of the patient global assessment as one of the outcomes addresses the patient perspective and the patient reported outcome domain but is not the sole measure for treatment response. The transition question of SF-36 “compared to one year ago, how are you feeling today?,” the EQ-5D and the Guyatt feeling thermometer will be examined as alternative model outcomes when available.

We will perform multivariable logistic and linear regression analyses with 1) the binary outcome of MDA and separately, 2) the continuous outcome of last visit patient global assessment as the outcomes. The predictor variables will be the individual outcome measures at the final visit (same visit during which MDA and patient global assessment are assessed). Outcome measures with the highest correlation coefficient with change in patient global will be selected from each domain for inclusion in the model. Variance inflation factors will be used to assess for the degree of collinearity among outcome measures included in the models. Clustering by trial and/or medication will be accounted for in the models. The R squared value will be used to identify the best model for overall improvement. The R squared value represents the goodness of fit (or total explained variation) in linear regression and ranges from 0-100%. If there is sufficient missing data (or non-overlap of outcome measures), modeling will be performed in each study separately and compared across studies.

Two examples of studies using the methods outlined for a similar purpose in fibromyalgia and rheumatoid arthritis are shown below. Dr. Mease, one of the co-investigators, was involved in both studies. Dr. Vibeke Strand, while she will not have access to the data, will be available to answer methodologic questions and will assist in interpretation of the data from this study.

·  Choy EH, Arnold LM, Clauw DJ, Crofford LJ, Glass JM, Simon LS, et al. Content and criterion validity of the preliminary core dataset for clinical trials in fibromyalgia syndrome. J Rheumatol 2009; 36(10): 2330-4

·  Strand V, Smolen JS, van Vollenhoven RF, Mease P, Burmester GR, Hiepe F, et al. Certolizumab pegol plus methotrexate provides broad relief from the burden of rheumatoid arthritis: analysis of patient-reported outcomes from the RAPID2 trial. Ann Rheum Dis 2011; 70(6): 996-1002