Blood Tests and Questionnaires in Studying Adherence to Preventative Swallowing Exercises in Participants With Metastatic Head and Neck Cancer

Adherence to swallowing exercises

The primary analytic approach is to use linear mixed models to determine whether illness perceptions significantly predict adherence 6 months. The study will estimate and test the illness perception subscale scores' relationship to adherence using a linear mixed model. In the model, the dependent variable is the patient's self-reported adherence score and independent variables include illness perception, follow-up assessment time point and other covariates such as demographic and medical variables, depression, pain, fatigue, emotional coping, objective coping, appraisal of coping, stress, and lack of social well-being.

Illness perceptions

The study's statistical tool is mediation analysis: the independent variables are the sickness behaviors (depression, fatigue, pain); the mediators are the cytokine levels, and the outcome variable is illness perception. The study will test the mediator effects based on the product of coefficients method, rather than the classical Baron \& Kenny causal steps tests, since the former approach has better statistical power. The study will conduct both single-mediator analysis for each of the mediators, and multiple-mediator analysis with all mediators in the mediating process

Effect of social support on coping

The moderating effect of social support on coping and appraisal of coping is the statistical equivalent of testing the two-way interaction between these variables and the social support variable. Multivariate linear regression will be used, in which the dependent variable is appraisal of coping at 6 months follow-up, and independent variables include the emotion-focused coping and action-focused sub scales, social support, and their two-way interactions. In the regression, the study will control for baseline variables. In addition, we will also apply the linear mixed model to investigate the moderating effect of these variables longitudinally. The mixed model naturally accounts for the correlation of repeated measures at baseline and 6 months follow-up.