Statistical Analysis

To rule out any possible initial differences between the groups, a t-test was performed on age, the anthropometric variables (height and weight), and the cognitive evaluation (Mini-Mental State Examination) of the participants.

For the COP displacement variables, a mixed design of repeated measures analysis of variance (ANOVA) was performed to examine the potential differences between the two groups (PD participants and healthy controls) with respect to the effect of within-subject factors: vision (open or closed eyes) and task (single or dual-task), as well as possible interaction effects (three-way ANOVA). Data distribution normality was checked beforehand, as was the sphericity and homogeneity of variance. A measure of effect size for the statistically significant effects was obtained using partial Eta squared values (η_p²). When the post hoc analysis showed a significant difference in the pairwise comparisons, effect sizes were calculated using Cohen’s d for the within-subject differences (e.g., single vs. dual-task for healthy controls group).

In the introduction it was mentioned how aging is a factor that affects postural control. It has also been verified how it can affect dual-task management (Woollacott and Shumway-Cook, 2002; Lacour et al., 2008; Doumas et al., 2009; Olivier et al., 2010). Consequently, in the study it was important to be able to clearly separate the effect of Parkinson’s disease on the stability of the aging effect. Although there were no significant differences in the mean age of each group (Table 1), the age ranges were not the same (51 vs. 43), with greater variability in the control group. So, we decided to include age as a covariate to control for its influence on the results of the study participants. The use of age as a covariate was intended to reduce within-group error variance, assuming that some variability in the stability data of each group could be given by the age variable. None of the contrasts were made with age as a covariate in their interaction with the within-subject factors: vision (open or closed eyes) and task (single or dual-task) were significant. However, when removing this covariate from the analysis, the ANOVA did not produce the same results, showing significant effects of the vision and task factors on most of the variables of the COP displacement. In our statistical treatment, we opted for a more conservative approach that would allow for a more accurate determination of the effects of inter and intra-subject factors. Therefore, it was decided to keep age as a covariate in the statistical analysis.

To analyze the cognitive performance of the participants under dual-task conditions, a further repeated measures ANOVA was planned (two-way). Group (participant with PD or healthy control) was once again used as a between-groups factor. For the cognitive task, the three conditions of within-subject factor were sitting on a chair (baseline), during DTOE, and DTCE conditions. The variable that expressed cognitive performance, the ratio errors/phonemes, failed the data distribution normality and variance homogeneity assumptions. A non-parametric ANOVA-type test was conducted. This statistic allows the same analysis as a traditional ANOVA (i.e., the effect of each factor and the interaction between them) but is based on the use of ranks for calculating the so-called relative marginal effects (Noguchi et al., 2012). When a significant interaction was detected, paired comparison between groups was applied by using the Mann–Whitney U-test and paired comparison within the groups was determined with the Wilcoxon Signed-Rank test with Bonferroni’s adjustment. A measure of effect size was included by calculating Cohen’s d from the U and Z values reported by the non-parametric test used in pairwise comparisons (Fritz et al., 2012; Lenhard and Lenhard, 2016).

All statistical analyses were performed using SPSS software (IBM SPSS Statistics, release 20.0). A p-value of ≤0.05 was considered statistically significant.