The calculation of sample size is based on two earlier studies regarding return-to-work in cancer patients. Based on the study by Spelten et al. [
18] with consecutive cancer patients, the expectation is that 18 months after diagnosis 36% of patients will not have returned to work [
18] after care as usual in the control group. With regard to the intervention group, the expectation is that 19% of the patients will not have returned to work 18 months after diagnosis [
23], based on the study by Nieuwenhuijsen et al. [
23]. Due to the inclusion and exclusion criteria, a percentage of patients with relatively less severe return-to-work problems will not be included in this study and this may lead to less favourable return-to-work rates. However, our intervention is more comprehensive than in the study by Nieuwenhuijsen et al. [
23] and thus the expectation is that the percentages of patients that will not have returned to work will be the same as in the study by Nieuwenhuijsen et al. [
23]. This indicates an Odds Ratio of 0.41 of the intervention versus usual care for higher percentages of patients that are not returned to work, which is the same as an Odds Ratio of 2.4 of the intervention versus care as usual for higher return-to-work rates. Since, the incidence of outcome is more than 10%, the Odds Ratio overestimates the magnitude of the association and therefore, we calculated the Relative Risk based on the Odds Ratio [
27]. This indicates a Relative Risk of not returning to work of 0.53 of the intervention versus usual care. Based on the PS Power and Sample size Program, with a power of 80% and two-sided significance level of p < 0.05, the sample size should be 109 patients in every arm, for a total of 216 patients [
28]. Assuming that 20% of the initial patients will be lost to follow-up during the course of the study, 270 patients must be recruited to gather 246 patients at 24 months. To account for at least 10% missing data at baseline, 300 patients are intended to be included. Furthermore, a sample size of 300 will have sufficient power to be able to control for the prognostic factors in a Cox regression analysis since we assume that 5 to 10 variables will be included in the final model [
17,
18] and a sample size of 10 per included factor in a Cox regression analysis is considered sufficient [
29].