Iris Kanera

140 Chapter 6 hematologic, kidney, liver, lung, prostate, stomach, testicular, and thyroid cancer). Type of treatment was categorized into (1) surgery and chemotherapy and radiotherapy , (2) surgery and chemotherapy , (3) surgery and radiotherapy , and (4) other types of treatment. Furthermore, aftercare ( yes/no ) and comorbidities ( yes/no ) were measured, and height and weight were assessed to determine BMI. The time since completion of primary treatment, measured in weeks, was based on registry data from the hospitals. Whether participants followed the modules PA and Diet was derived from program logging data. Module use was dichotomized ( yes/no ) and categorized into yes when at least the first three compulsory pages with important key information of the module were visited. Sample size Since the present study is part of a larger study project, sample size calculation was based on improvements in the main outcomes and revealed that each intervention condition needed to contain 144 participants (effect size = .30; one sided α = .05; β = 0.2; power = .80); intra-class correlation coefficient (ICC) = 0.005). With an expected dropout of some 20%- 23%, the required sample size was N = 376 (188 per condition) at baseline. Statistical analysis Baseline differences between IC and UC concerning lifestyle behaviors, and demographic and cancer-related characteristics were examined using independent t-tests and chi-square tests. Selective dropout after 12 months was assessed by applying logistic regression analysis with dropout as outcome variable (no = 0; yes = 1) and group assignment and baseline characteristics as predictive factors. In order to evaluate the main intervention effects on moderate PA and vegetable consumption, multilevel linear regression analysis (MLA) was conducted. A three-level longitudinal data structure was used, in order to account for interdependencies. Outcomes at two time points (6 and 12-month follow-up) were clustered within the participants and participants were clustered within hospitals. Time, individuals, and hospital were added to the MLA model with a random intercept, and intervention condition and baseline value of the dependent variable were added as random slopes. The models were adjusted by adding the baseline value of the outcome behavior, standard demographic and cancer- related characteristics, significant variables from dropout analysis, and significant baseline differences. In full, the added variables were gender, age, marital status, education level, income level, employment status, BMI, type of cancer, having had cancer before, type of treatment, time since completion of primary cancer treatment, aftercare, comorbidities, vegetable intake, fruit intake, whole grain bread intake and fish intake at baseline (added as fixed intercepts). These adjustments were in line with the MLA modelling of the prior study,