Data management
All information required by the study protocol is to be recorded in the case report forms and an explanation is to be provided for any missing data. Data is to be collected progressively, as obtained, and recorded explicitly in the case report forms.
An electronic case report form (e-CRF) will be made available and data capture will be carried out in the centers using a web interface (Clinsight software, Ennov, Paris, France). Only an Internet connection and a browser are required. A document on how to use this tool will be provided for the investigators. The interface between the clinical research associate and the investigator will thus be provided, also making remote data collection and control possible. Data consistency tests will be incorporated in electronic format. An audit function is incorporated in the e-CRF, thus making it possible to follow any change in study data. This function also makes it possible to clearly identify the person who has made a change, with the date; comments can be added.
If requested, a hard copy will be printed at the end of the study, authenticated (dated and signed) by the investigator, and copies will be sent to the sponsor and archived.
The procedures applied are those used Rennes CHU and comply with Good Clinical Practice and the legal and regulatory requirements in force.
Statistical methods
The statistical issue for the primary outcomes is the analysis of repeat measures (up to 12 measures for each 1-year period, on the PO-SCORAD, a self-administered questionnaire the results of which are collected through an application on computer or phone, at home) applied to a crossover design (two 1-year periods, two sequences). When repeat measures are analyzed, there are many possible hypotheses that could be of interest. We focused on a main effect hypothesis test for the effect of a single predictor variable (treatment: phototherapy) averaged across all other factors, and the sample size calculation was designed accordingly. We are interested in the usual factors, namely ‘Treatment’ (phototherapy/no phototherapy) and ‘Time’. However, the nature of the design (crossover trial) needs to be taken into account. This means that we need to consider the following factors: ‘Period’ – first winter/second winter, ‘Order’ – phototherapy–no phototherapy or no phototherapy–phototherapy, and ‘Carry-over’. It was anticipated that vitamin D could act as a quantitative effect-modifier on phototherapy. We therefore decided to include an interaction term in the analysis. We will use a single model with the following main effects: Treatment-1 (phototherapy/no phototherapy), and Treatment-2 (vitamin D/placebo), Period, Order, and Carry-over terms as well as an interaction term between Treatment-2 and Treatment-1. As an interaction of Time and Treatment could occur, we plan to add an interaction term to the model. The other primary outcome, the cumulate consumption of TAT (number of tubes) during the winter, is a quantitative variable. It will be analyzed using a mixed model with the following main effects: Treatment-1 (phototherapy/no phototherapy), and Treatment-2 (vitamin D/placebo), Period, Order, and Carry-over terms as well as an interaction term between Treatment-2 and Treatment-1. As this outcome is a summary of winter consumption, no interaction of Time and Treatment will be added to this model. Adjustment will be made on the variables used in the allocation process (region – north vs. south), age as a continuous variable, and disease severity (moderate vs. severe).
For descriptive purposes, it is instructive to carry out an analysis based on a summary of the statistics (a quantity calculated from each curve), which can reflect important aspects of the problem at hand – we have chosen the area under the curve (AUC) and the mean amplitude of PO-SCORAD score ranges. Descriptive statistics (mean and standard deviation) of the PO-SCORAD for each visit will also be provided. Normal distribution is assumed in the modeling analysis for the primary outcomes (PO-SCORAD severity score over two 1-year periods; cumulate consumption of TAT during winter) and will be checked at this stage.
The main analysis will be based on the adjusted model and will follow the ‘intention to treat’ principle, namely all participants, as randomized, will be analyzed; outcome data obtained from all participants, regardless of protocol adherence, will be used. Mixed model techniques (PROC MIXED) will be applied.
The multiplicity adjustment strategy used is a hierarchical closed test procedure, where the primary endpoints are ordered as follows: (E1) repeat measures of the PO-SCORAD severity score over 1 year and (E2) cumulate consumption of TAT over winter. We will test E1 and E2 sequentially at the same two-sided level of 0.05; we will test E1 first, if it proves significant, will we test E2. A Wald test with Kenward–Roger degrees of freedom will be used.
No subgroup analyses nor interim analyses are planned.
For secondary analyses, we will use up to four measures per 1-year period, collected at each clinical visit (every 3 months), including investigator-based severity scores (Eczema Area and Severity Index, Severity of Atopic Dermatitis Index, Investigator Global Assessment), patient-reported severity and quality of life scores (Patient-Oriented Eczema Measure, Dermatology Life Quality Index), serum vitamin D levels, total IgE serum levels, weeks of satisfactory control, inter-visit cumulate consumption of TAT, and synthetic patient-reported satisfaction at the end of each winter.
All analyses will use procedures available in SAS software 9.4 (SAS Institute, Carry, NC, USA).