Klaske van Sluis

Voice outcomes following total laryngectomy 51 3.2.4 Statistical analysis The data is analysed with help of IBM SPSS software to perform descriptive statistics [28] and R [29] for inferential statistics and modelling. No sample size calculation was performed since numbers of inclusion were based on the available patients admitted to TL. Study sample characteristics were tabulated and visualized. Primary out- comes were VHI-10, AVQI, Perc. Voice SLP, and Perc. Voice Pt. Paired t-tests between T0 and T2 were performed with statistical significance level set at p ≤ .05. To investigate treatment variables, three oncologic treatment variables were transformed to dichotomous variables, including a) primary surgical treat- ment vs. salvage surgical treatment, b) primary closure vs. major reconstruction of the neopharynx, and c) a History of CRT vs. RT. The variable c) History of CRT vs. RT proved to be redundant and was dropped. Definitions of the definite chosen variables are shown in Table 1 . Table 3.1: Transformed oncological treatment factors into dichotomous vari- ables Variable Includes participants with Treatment Primary surgical Total laryngectomy as a primary cancer treatment treatment Total laryngectomy as a treatment for a secondary primary tumour Salvage surgical treatment Total laryngectomy as a salvage treatment in case of recurrent disease Total laryngectomy as a treatment for a dysfunc- tional larynx Reconstruction Primary closure Primary closure of the neopharynx Major reconstruc- tion Major reconstruction of the neopharynx with the use of flaps including a Pectoralis Major-flap, free flap or gastric pull up Correlations between primary outcome measures are investigated using lin- ear mixed effect models with (pseudo) R2 andChi-square ANOVA on Y ∼ X +(1|Subject) + (1|T) against Y ∼ 1 +(1|Subject) +(1|T). Scatter plots are made for visualization (Appendix I, II). Because of multiple testing we used Bonferroni correction and adapted alpha to ≤ .01 To estimate the importance of the factors studied for outcomes in VHI- 10, AVQI, and perceptual rated voice quality overtime, linear mixed effect models were created (Appendix III, IV, V). The model analyses the relationship between AVQI,VHI-10, and Perc. Voice SLP on the one hand and the fixed effects Time (T0, T1, T2, T3), Treatment (primary surgery vs.salvage), and