Análise da associação entre interleucina-6 e doença cardiovascular e a busca de um modelo em redes neurais artificiais para identificação de risco cardiovascular em pacientes com síndrome metabólica

• Ano de defesa: 2014
• Autor(a) principal: Helegda, Lara Colognese
• Orientador(a): Bodanese, Luiz Carlos
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul
• Programa de Pós-Graduação: Programa de Pós-Graduação em Medicina e Ciências da Saúde
• Departamento: Faculdade de Medicina
• País: BR
• Palavras-chave em Português: MEDICINA; SÍNDROME METABÓLICA; DOENÇAS CARDIOVASCULARES; HIPERTENSÃO ARTERIAL; REDES NEURAIS
• Área do conhecimento CNPq: CNPQ::CIENCIAS DA SAUDE::MEDICINA
• Link de acesso: http://tede2.pucrs.br/tede2/handle/tede/1795

Introduction: The metabolic syndrome (MetS) is characterized by a complex disorder represented by a cluster of cardiovascular risk factors associated with central fat distribution and insulin resistance. Interleukin-6 (IL-6) is a adipokine produced in the visceral adipose tissue and have been proposed as an inflammatory marker of MetS and may also be related to the development of cardiovascular disease (CVD). The Artificial Neural Networks (ANN) comprise a processing structure composed of interconnected units, known as artificial neurons, seeking solutions to complex problems through learning. Objective: To evaluate the serum levels of the inflammatory cytokine interleukin-6 in patients with Metabolic Syndrome, with and without Cardiovascular Disease through clinical, biochemical and anthropometric parameters of these patients, to develop and to facilitate a pilot study to aid the propensity of Cardiovascular Disease using models in Artificial Neural Networks. Methods: Serum IL-6 levels were assessed in 80 patients with metabolic syndrome, 40 with CVD and 40 without CVD. Adding to this, clinical, biochemical and anthropometric parameters were selected from the registry database of Cardiometabolic Risk Clinic in a controlled cross-sectional study consisting of a historical sample. From this, parameters were selected relevant to CVD and presented two proposals for training models in ANN. ANN templates were used type Multilayer Perceptrom (MLP) and the setup process, training and validation was performed with the aid of MATLAB computational tool version 6.5, the Mathworks, with Neural Networks Toolbox packages specific to ANN and implemented with Backpropagation algorithm. Results: Among the variables we found that patients without CVD had baseline DBP (p = 0.008) and LDL- cholesterol (p = 0.026) higher than patients with CVD. It was also in the group of patients with CVD, statin use was significantly higher (82.8% vs. 27.5%; p = 0.001) compared to those without CVD group. Serum IL-6 levels were higher in patients with established CVD, 23.52 + 10.39 + 59.78 x 3,50pg/mL; p = 0.036, compared to patients without CVD. ANN was tested two topologies: with IL-6 and without IL-6. The network topology MLP with best result of classification, was considering all parameters relevant to CVD. This topology presented a mean absolute error of 2.41% and a configuration with one hidden layer and 80 internal neurons. Conclusion: Patients with MetS and with established CVD, showed serum levels of inflammatory cytokine IL-6 higher, which are associated with persistent inflammation. The ANN proved to be a complementary instrument at diagnosis, potentially useful for situations that demonstrate the complexity of characterizing the risk of cardiovascular events in patients with MetS