Modelagem GAMLSS da pressão arterial sanguínea em um estudo de hipotensão pós-exercício

Abstract

In Health Sciences, controlled studies play an important role in the development of techniques that promote prevention or cure of diseases. Frequently, statistical analyzes of repeated measures in time are used, since they allow to investigate the possible changes in the response variable over time. The more flexible methodology for the analysis of longitudinal data involves the use of mixed models, where the maximum likelihood (ML) or restricted maximum likelihood (REML) methods can be used in parameters estimation. Blood pressure (BP) is measured in terms of systolic (SBP) and diastolic (DBP) pressure and their high values lead to hypertension. This disease is a worldwide problem that, if not treated correctly, can lead the patient to suffer serious consequences. Currently, non-pharmacological methods, as the practice of physical activities, are prescribed for the control of BP. A longitudinal study was carried out in Lavras, MG, to evaluate the BP of 26 normotensive adults submitted to two training programs (resistance and hypertrophy). Semiparametric models with random effect were fitted to data from SBP and DBP obtained over 18 sessions. For this, we use the class of generalized additive models for location, scale and shape (GAMLSS). In all models (for SBP and DBP) the baseline was included as part of the response vector, assuming homogeneity between groups before training. Overall, the results indicated that both training programs lead to BP adaptations that are able to decrease pre-training blood pressure levels. Resistance training has been found to cause greater variability in SBP data than hypertrophy training and the latter showed a higher hypotensive effect on DBP. All analyzes were performed in R.