Ires Ghielen

84 Chapter 5 Statistical analyses Linear mixed model analyses were used to assess the longitudinal associations between the outcome measures. Because repeated measurements, four time-points over 12 weeks, were nested within subjects, multilevel analyses is most appropiate [31]. Concerning missing values, multilevel analysis is quite good in handling missing data in the outcome variables because of this estimating method [32]. This longitudinal study consists of four repeated measurements of 153 patients, which means that the number of dummy variables would be large, compared to the total number of observations, and therefore it would be impossible to analyze the data in this way. In multilevel analysis the variance of the intercepts is estimated, not the separate intercepts for each patient. In this way, only one variance parameter (the ‘random’ intercept) is estimated. Equally, when it comes to the change of the outcome measures over the time points, the variance of the slopes is the one parameter that is estimated and further used in the analysis. Separate linear mixed model analyses were used to analyse the following longitudinal univariable associations: 1) between FoG and FoF; 2) between FoG and anxiety, in which the first-mentioned variable (FoG) of the associations is dependent, and the latter the independent variable. Subsequently, the longitudinal multivariate association 3) between FoF and anxiety together as independent variables, and FoG as dependent variable, was analysed to investigate which variable is contributing the most to the association with FoG. To investigate possible confounding factors, all associations were adjusted for the following groups of variables: * Demographics; including gender [33-35], age [36] and body mass index [37-40]. ** Disease characteristics; including UPDRS motor section 3 [41], disease duration [42], H&Y, HADSD [43], falls diary, the speed of the 10 meter walking test [44], standing balance (single leg stance test [45]), monitored physical activity (accelerometer), fatigue (multidimensional fatigue inventory (MFI)), cognitive executive function [46] (Mini Mental State Examination and Brixton Test Scaled score [47]). *** Medication use; including Levodopa, dopamine agonist, catechol-O-methyl transferase inhibitor (Entacapone), selegiline and other (the type of medication that a patient used was registered), and adverse effects (using the UPDRS part IV B).