RELIEVE-AD is an ongoing observational, prospective, longitudinal survey study in adult patients with AD who were enrolled in the Dupixent MyWay™ Patient Support Program and for whom dupilumab had been recently prescribed. Eligible patients completed a baseline survey before starting dupilumab and were followed at Months 1, 2, 3, 6, 9, and 12 post-initiation as they become eligible. Patients must have consented online to participate in the study prior to proceeding to completing the baseline questions through a secure online portal. The survey was performed in accordance with the Helsinki Declaration of 1964 and its later amendments and received a full review approval by the New England Independent Review Board in December 2017.
Patient enrollment into the RELIEVE-AD study began in January 2018 and the final data collection is expected to be completed in February 2020. The present study included patients in the RELIEVE-AD study who, on December 6, 2018, had completed the baseline and Months 1, 2, 3 and 6 surveys. Patients were eligible for inclusion in the RELIEVE-AD study if they met the following criteria at the time of the baseline survey:
- aged ≥18 years
- can speak and read English
- be willing to participate in the study and provide informed consent
- have not previously participated in a dupilumab clinical trial
- have not initiated treatment with dupilumab.
The surveys collected data on patient characteristics, including socio-demographics (age, sex, race/ethnicity, marital status, level of education, insurance, employment status, level of income, geographic region), medical history (self-reported age at AD diagnosis, comorbidities), and AD treatment and experience (treatment history prior to dupilumab initiation, concomitant therapy post dupilumab initiation, self-reported adherence to treatment and reasons for discontinuation), and treatment satisfaction.
In addition, PROM data were collected using the Patient Global Assessment of Disease (PGAD), Numerical Rating Scale (NRS) for patient self-reported symptoms (skin pain, burning, and sensitivity) (scores: 0–10; higher scores indicate worse symptom severity), disease control using ADCT (eczema-related symptoms, days with intense itching, overall bothersomeness, sleep problems, daily activities, mood/emotion; total score 0–24; higher scores indicate worse disease control), health-related quality of life (HRQL) using the Dermatology Life Quality Index (DLQI: 0–30, higher scores indicating worse HRQL), and the Work Productivity and Activity Impairment-Atopic Dermatitis questionnaire (WPAI-AD; percentages: 0–100, higher percentages indicate greater impairment) for patients in employment. (Table 1).
Analyses in this study were conducted using PROM data from multiple survey timepoints to ensure the robustness of the findings.
All data analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).
Assessments for reliability included internal consistency reliability and test–retest reliability. Internal consistency was assessed using Cronbach’s alpha (α ≥0.7) . ADCT item-to-total correlations were estimated at baseline and Months 1, 2, 3, and 6 using the Pearson correlation coefficient (PCC ≥0.5) . Test–retest reliability was evaluated based on the intra-class correlation (ICC) coefficient of the ADCT total score among patients with unchanged PGAD scores across month pairs (between Months 1 and 2, Months 2 and 3, and Months 3 and 6). An ICC ≥0.70 was expected for confirming test–retest reliability .
Convergent validity of the ADCT was assessed by computing Spearman’s rank-order correlation between the ADCT total score and the DLQI (total and item-level scores), skin pain, PGAD overall well-being, WPAI total work impairment (WPAI-TWI), and WPAI total activity impairment (WPAI-TAI) at baseline and Months 1, 2, 3, and 6. Given that the skin pain NRS directly measures AD-related symptoms and the DLQI includes questions on both symptoms and impacts due to skin problems, correlations between the ADCT and skin pain and DLQI were expected to be higher than the correlations between the ADCT and other measures, such as PGAD, WPAI-TWI, and WPAI-TAI. Cohen’s recommended guidelines for determining small, moderate, or large effects (0.1 to <0.3, 0.3 to <0.5, and ≥0.5, respectively) were applied, and a large effect (r ≥ 0.5) was used in this study as evidence of convergent validity . Divergent validity, established previously for the ADCT, was not assessed here owing to the lack of appropriate measures for use from the current study.
To confirm known-groups validity, mean ADCT total scores were compared across adjacent subgroups of patients based on PGAD responses (Excellent, Very good, Good, Fair, Poor) and categories of DLQI responses: no effect on patient life (score range: 0–1), a small effect (2–5), a moderate effect (6–10), a very large effect (11–20), an extremely large effect (21–30)  (Table 1). Patients in a worse PGAD or DLQI band subgroup were expected to display poorer AD control (i.e., higher mean ADCT total scores, indicating more severe symptoms/greater impact) than patients in a better PGAD or DLQI band subgroup. If the homogeneity of variance across the subgroups was rejected (p < 0.05) based on a Levene’s test of equality of variance, a Mann–Whitney U test was used to compare the mean ADCT total scores between the subgroups; otherwise, t-tests were applied. Cohen's d was calculated for the standardized differences in mean ADCT total scores between subgroups and was corrected for small sample sizes when the total sample size in the two groups was below 50 .
Ability to detect change (responsiveness)
Responsiveness was evaluated using correlations between the change from baseline (to Months 1, 2, 3, and 6) in ADCT total score and the change from baseline in DLQI total score (Pearson product-moment). The same analysis was conducted using (Spearman’s rank-order correlation) (r ≥ 0.5) for DLQI bands and PGAD scores.
Interpretation of change:
Anchor-based and distribution-based methods were used to establish a threshold characterizing meaningful within-person change in the ADCT total score.
Prior to applying the anchor-based method, the correlation coefficient between the change in the ADCT total score and the potential anchor measure was reviewed for the magnitude of association; in this study, a large effect (i.e., correlation at least 0.5) was required. Once established as appropriate, univariate regression analyses accounting for repeated measures were conducted; changes in ADCT total scores from baseline was the dependent variable and changes in the anchor measure from baseline was the independent variable. The change in PGAD and change in DLQI were considered as potential anchor measures and the following anchors were selected a priori: a 1-level improvement in the PGAD; a 4-point improvement on the DLQI total score ; or a 1-level improvement in the DLQI band. Patients who were not likely to change were excluded: e.g., reporting PGAD = “excellent” or DLQI = “no effect” (i.e., total score of 0 or 1) at baseline. Additional analysis was conducted using the subset of patients whose AD was considered not controlled at baseline based on the ADCT total score (i.e., score >7; Table 1), as established in previous research .
For the distribution-based approach, the half standard deviation (SD) method of the baseline ADCT scores, one-third SD, one unit of standard error of measurement (SEM), and two unit of SEMs were examined. Final recommendations for thresholds characterizing meaningful within-person change and considered as a clinical important responder were made considering the anchor- and distribution-based results.