Analysis of Moderators and Mediators

Individual patient data from randomized controlled trials from the POLARIS database are pooled to conduct the statistical analyses, ensuring sufficient power. In different research questions, we will evaluate which of the following socio-demographic, clinical, and psychosocial characteristics moderate or mediate the psychosocial- or exercise intervention effect on quality of life.

  • Socio-demographic characteristics: age at baseline, gender, marital status at baseline, level of education at baseline, employment status at baseline, ethnicity, and performance status at baseline
  • Clinical characteristics: cancer diagnosis, cancer stage, time since diagnosis when starting the intervention, treatment type and presence of co morbidities
  • Psychosocial characteristics: depression, anxiety, mood, emotional distress, self-esteem, anger , sleep quality, social support, optimism, illness/caregiving (appraisal), uncertainty (appraisal), hopelessness (appraisal), happiness (appraisal), and sexual (dys)function

Other covariates that will be used in the statistical analysis, when not analyzed as moderator, are baseline QoL, cancer diagnosis,  cancer stage, time since diagnosis when starting the intervention, gender, age at baseline, education level at baseline, treatment (e.g. radiotherapy), length of treatment, treatment confidence, performance status (e.g. Karnofsky Performance Scale), and presence of co morbidities.

Other covariates (depending on which intervention type was used) that will be used in the statistical analysis are intervention type (e.g. resistance, endurance exercise, cogntive behavior therapy), intervention duration, exercise intensity, exercise frequency, exercise session duration, exercise supervision (i.e. yes/no), intervention format (e.g. group, individual, couples), number of care providers involved in the intervention, total number of sessions of the intervention.

To test for the presence of a moderator variable, we will use moderated multiple regression analyses (MMR), see Figure 1

To examine potential mediators of the intervention effect on quality of life, a series of regression analyses are conducted, according to the product-of-coefficients test described by MacKinnon, see Figure 2.

 

 Figure 1: Moderation analysis                                                                                                                     

 

 

 

 

Figure 2: Mediation analysis