Mariken Stegmann

Blinding  We cannot blind patients, GPs, or researchers because of the nature of the intervention  and because patients from the intervention group will report OPT scores. However, data  analysis will be performed by researchers blinded to patient allocation.  Sample size  In previous research in patients with cancer, a mean of 36 points (standard deviation 9  points) was found on the D‐SE scale, the primary outcome of the study. 23  Given that half  a standard deviation is considered clinically relevant in health‐related quality of life  studies, we aim to detect an effect of at least 4 points (effect size 0.44) on this scale. 29 Concerning two sided testing, with an alpha of 0.05 and a beta of 0.2 (power 0.8); we  would therefore require 80 persons in each group. Because comparison between groups  is to be done immediately after consultations, we expect a low rate of loss to follow up.  Accounting for a 5% loss, we aim to include 168 patients in total, or 84 per group. To  ensure adequate participation, we will recruit participants from ve hospitals (eight  locations). Both the doctors and participants will receive a newsletter regularly.  Statistical analysis  All data will be entered in a secured digital data management system. When data  collection will be completed, all data will be extracted into a statistical software package.  Descriptive statistics will be used to compare groups at baseline. The effect of using the  OPT on decision self‐efcacy will be primarily by comparison of the mean D‐SE values,  using an independent t‐test. The analysis of treatment effects will be on an intention‐ to‐ treat principle; that is, participants will be analysed by their randomisation, regardless of  the actual intervention received. Per‐ protocol analyses (based on completion of the  intervention) will    be conducted to investigate whether deviations  from  the  proto‐ col  inuence the effect  estimates.  If  baseline  differences  appear in the per‐protocol  analyses, they will be adjusted for by linear  regression modelling. In this model  treatment arm, the hospital site and the unequally distributed baseline variables will be  included as independent variables and self‐efcacy as the dependent variable. Linear  regression will additionally be used to explore the modifying effects of gender, age,  anxiety, depression, fatigue, tumour type, educational level, performance score, and  social network, by including and testing their respective interactions with the treatment  group.   Embedded observational study  An observational prospective cohort study of the patients in the OPT group will be  performed to explore the changes in preferences for treatment goals. The rst OPT  scores will be considered the baseline assessment, and patients will contact their GPs  again at 4 weeks, 3 months, and 6 months thereafter to reassess their OPT scores. As  secondary outcomes, the D‐SE scale, HADS, and MFI‐ 20 will be completed at each time  point (Fig. 1). OPT scores and changes over time will be described, and the association of  the OPT scores with gender, age, anxiety, depression, fatigue, tumour type, disease  course, educational level, performance score, and social network will be explored in a  hypothesis‐generating manner. Finally, the experiences with the OPT of both patients  and GPs will be assessed using a short questionnaire. Because of an expected survival  rate of around 60%, we expect the number of patients that will complete this  observational study to be around 50.  24 Chapter 2

RkJQdWJsaXNoZXIy ODAyMDc0