Statistical Analysis

For this analysis, we will use the multivariate regression model in which motor improvement will be the dependent variable and changes in inhibitory activity (resulting from neurophysiological assessments performed before and after treatment) will be the independent variables.

To control the impact of different conditions on the multivariate model, we will create a “dummy variable” for each disease etiology. Moreover, demographic characteristics (age, education, sex, and ethnicity) and clinically relevant characteristics (duration of illness, comorbidities, and the use of medications) will be tested, as well as specific aspects for each disease, such as the stroke side, the level of spinal cord injury, the degree of bone deformity of knee OA, and the level of lower limb amputation. The neurophysiological biomarkers described earlier will be tested in the same model.

Although this study's aim is not to test interventions but to identify changes in biomarkers related to functional improvement (regardless of the therapy performed), the different therapies performed by patients will be quantified, including information such as the number of sessions, frequency, duration, among others, which can be used in future analyses.

For secondary analyzes, functional improvement can be assessed by the several, general and specific, scales used depending on the evaluated disease. In this situation, we will use the calculation of the functional modification's effect size and not the absolute values of the scale for the analysis in the multivariate model.

Also, the motor function of the upper limb will only be assessed for patients with stroke and spinal cord injury, the main scale used being the Fugl–Meyer Assessment. In this case, an analysis similar to the one described earlier will be performed but only including these two populations.

Some of the scales, such as those for mood, pain, cognition, and sleep, will be used to characterize the sample, in addition to possibly being used to control confounders in the multivariate model. Besides, mood, pain, and cognitive disorder are commonly present in these populations, so we will perform an exploratory analysis using a similar statistic method but with scales related to these aspects as dependent variables.

In the statistical analysis for the results of the obtained polymorphisms, classical methods of case–control studies' epidemiological analysis will be applied. Odds ratio and the respective 95% confidence intervals will be estimated by unconditional logistic regression to simultaneously control potential confounding variables. To assess the relationship between the dependent variable (stroke and its subtypes) and the independent variables (polymorphisms of the evaluated genes, smoking, lipid variables, etc.), the statistical technique used will be logistic regression analysis, which allows the evaluation of disease risk associated with a given variable considering all other independent variables in the model.