Please use this identifier to cite or link to this item:

http://hdl.handle.net/10609/91486
Title: Análisis exploratorio multivariante para el estudio de la distribución anormal de grasa, masa magra y masa ósea en individuos infectados por VIH
Author: Jácome Pérez, Longino
Director: Pérez Álvarez, Nuria
Tutor: Ventura Royo, Carles  
Keywords: multivariate analysis
osteoporosis
lipodistrofia
sarcopenia
HIV
Issue Date: Jan-2019
Publisher: Universitat Oberta de Catalunya (UOC)
Abstract: The objective of this study is to evaluate the associated variables to the bone mass, fat mass and lean mass, and its possible relation with osteoporosis, lipodistrophy, and sarcopenia in patients infected with HIV. Supervised learning techniques are used to obtain the relationship between the variables of interest in the study: regression and classification techniques. A longitudinal analysis of unbalanced data is done through mixed linear models in order to model the relationship between individuals and the time or infection. The database includes 1,475 individuals some of which have more than one dexa. The database was divided in men and women in order to do the analysis separately. The variables density of total bone mass, total fat mass and total lean mass were considered as response variables. The criterion of the World Health Organization was taken to characterize the presence of osteoporosis through the T brand. The indexes of fat mass and lean mass were considered to evaluate the lipodistrophy and the sarcopenia. The 65% of men population presents osteoporosis. For women, 60% presents such condition. The model of bone mass for men explains the 33% of the data and for women the 37 %. The 96.5% of men has the lipodistrophy condition and 99.7% in women. The regression model explains the 97.5% of the data for both men and women. The 11.8% of men has sarcopenia according to the defined criteria. The 0.6% of women presents this condition. The regression model for the total lean mass has an r2 of 0.98 for men and 0.99 for women. The bone classification model, based on the knn method, has an accuracy of 60% for men and 62% for women. The corresponding model for lipodistrophy shows an accuracy of 96% for men and 100% for women. In the case of lean mass, we got an accuracy of 0.92 for men and 1 for women. Regarding the longitudinal model, the decrease in bone mass with time is observed as a general trend, specially in those observations with more than two dexas, for both men and women. The results obtained in this study in terms of the indices, do not differ greatly from those obtained for other populations with HIV. The percentages of individuals with osteoporosis are consistent with those reported by the literature. The lipodistrophy values obtained are very high, perhaps due to the cutoff points used. The sarcopenia values are very different from those reported, so they should be reviewed. Considering the result obtained it is necessary to perform analysis by age, review cutoff points and association of variables to eliminate collinearity problems.
Language: Spanish
URI: http://hdl.handle.net/10609/91486
Appears in Collections:Bachelor thesis, research projects, etc.

Share:
Export:
Files in This Item:
File Description SizeFormat 
longinoTFM0119memoria.pdfMemoria del TFM4.05 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons