Stefan Elbers

190 Chapter 7 correspondence table that contains both the identifier and patient personal information. This table will be stored in a separate folder that only the principal investigator (JP) and the head of the Lifestyle & Health Research Group (HW) can access. Any unforeseen data collection issues that may threaten to reveal an individual’s identity will be solved according to the recommendations of Tessier and Bonnemains (2019). Harm We do not expect any serious adverse events as a result of using the app. However, we will monitor any negative consequence that results from usage. During the training of the treatment team, we will instruct the health care providers to report negative experiences of the app (e.g. frustration due to low digital literacy skills). During biweekly contact with treatment teams, the researcher will also actively ask for any adverse event. In the unlikely event of harm, patients can appeal to the liability insurance of the sponsor that covers any damage to research subjects caused by the study within 4 years after the end of the study. Data Analysis During all analyses, the treatment condition will be masked to the researchers. We will analyse the data according to the intention-to-treat principle and include patients in the analysis regardless of their adherence to the treatment protocol.We will performno interim analysis. We will collect data at the following time points: –t 1 (baseline), t 1 (immediately postintervention), t 2 (3 months postintervention), t 3 (6 months postintervention), and t 4 (12 months postintervention). Prior to the main analysis, we will check the randomization by examining the distribution of baseline characteristics between both groups. Sample Size Calculation. The sample size calculation is based on a 2-factor repeated- measures analysis of variance with within-factor time (5 levels) and between-factor treatment (2 levels) conditions on the outcome variable PDI. We have set the α to .05, the power (1 – β ) to .95, and assumed a moderate treatment effect ( f = 0.2). To account for the expected dependencies of patients within each of the 5 participating treatment locations, we applied the Donner et al (2004) formula for the variance inflation factor, assuming an intracorrelation coefficient of 0.2 (Dickinson & Basu, 2005; Twisk et al., 2013). Furthermore, we corrected the analysis for an expected attrition of 20%, which is based on the average attrition of similar studies that used the PDI (Hållstam et al., 2017; Mangels et al., 2009; Volker et al., 2017). Based on these calculations, a minimum sample size of 157 participants, equally divided over 5 treatment locations, will be required. Primary Analysis: Pain Disability. In our primary analysis, we will test the difference in the development of pain disability over time between patients in the enhanced condition and patients in the treatment condition. To account for the assumed dependence of the