Iris Kanera

7 163 General Discussion did not report short- or long-term effects on lifestyle behaviors that were the scope in our study. We may conclude that multiple behavior web-based cancer aftercare is still in its infancy and needs further investigation (Post & Flanagan, 2016). Besides web-based interventions, also telephone-based interventions were developed that target lifestyle among cancer survivors (Goode et al., 2015). Particularly telephone-delivered programs including intensive in-person contact were able to achieve post-intervention effects on PA and diet behavior with effect sizes ranging from low to high. All PA and diet outcomes in long-term had small to very small effect sizes ( d ≤ .49) that mostly decreased between the end of intervention and long-term follow-ups. Strikingly, the KNW revealed comparable effect sizes on moderate PA with an increase over time (among survivors younger than 57 years) compared to the abovementioned interventions that included in- person contact and targeted fewer behaviors. Overall, to our best knowledge, the study in Chapter 6 was the first that showed significant long-term effects on PA of a multiple behavior web-based intervention for survivors of various types of cancer. METHODOLOGICAL CONSIDERATIONS For the interpretation of the results, some methodological issues should be taken into consideration. Most of the methodological issues of the separate studies in this thesis are discussed in the previous chapters. The following section explores the most important methodological issues with regard to generalizability and internal validity of the effect studies and the main outcome measurements. Study population Sample characteristics A strength of the main intervention study (Chapter 5 and Chapter 6) is the large study population ( N = 462) at baseline that resulted in adequate statistical power. The initially calculated required sample size of N = 376 was exceeded. The sample included mainly middle-aged, female survivors of breast cancer with a good prognosis, low numbers of comorbidities, relatively low levels of physical and psychological complaints, and engagement in PA. This sample composition does not represent the overall population of early cancer survivors, which hampers generalization to the overall cancer survivor population. The representativeness of our main study population may have been largely influenced by the recruitment strategy that took place in 21 hospitals. However, this number of hospitals did not result in a representative sample of each possible cancer type conform the prevalence in the Netherlands. Cancer survivors were eligible when having no sign of cancer recurrence and a good prognosis, thereby excluding the more fatal cancer types. It is not surprising that for instance survivors of breast cancer, having a high one-year