Modelo linear misto com interações e componentes principais para avaliar o efeito múltiplo de poluentes e variáveis climáticas na saúde respiratória

Abstract

The objective of this thesis is to investigate the effect of PM10, PM2.5 and SO2 atmospheric pollutants on the pulmonary function of children and adolescents living in an urbanized area, using linear mixed-effects regression models with interactions between some of the predictors and Principal Component Analysis (ACP). The sample consists of 82 children and adolescents aged 7 to 18 years, living in the surroundings (<2km) of an air quality monitoring station. Pulmonary function tests were performed monthly from July to December 2017. The main measure was the percentage of forced expiratory volume in the first second (FEV1%). The pollutant concentration data were obtained from the Automatic Air Quality Monitoring Network (RAMQAr) - Enseada do Suá Station) and the meteorological data from the Center for Weather Forecasting and Climate Studies - CEPETEC. To quantify the association between FEV1%, PM10, PM2.5, and SO2 pollutants, temperature, humidity and other covariates of interest, five mixed models (MM) models were considered. Models I and II explore the effects of pollutant and temperature on FEV1%. Model III explores the interactive effect between pollutants and climate variables on FEV1%. In model IV, in addition to the predictors of interest, pollution and climate variables and additional covariates related to the outcome are considered. In model V the PCA technique is considered. The analyses show a negative correlation between temperature and FEV1% and a positive correlation between pollutants and temperature. The consideration of the interaction effect between pollutant and temperature, as a predictor variable, contributes to mitigating the multicollinearity issue. When considering the maximum daily average concentrations of pollutants and the maximum average temperature for the period, reductions in FEV1% are observed. For models with PM10, FEV1% reduction percentages are 1% in the model with pollutant as the only predictor variable, 3.7% when the temperature is introduced as a predictor variable and 3% when considering the interaction between pollutant and temperature. For models with PM2.5, the FEV1% reduction percentages were 1.2%, 3.4% and 3.2%, respectively. In the models with SO2, the percentages of reduction are higher, 4.8% in the model with pollutant as the only predictor variable, 4% when the temperature is introduced and 4.6% when considering the interaction between pollutant and temperature. In the complete model, there is a significant inverse relationship between PM10 and SO2 pollutants and FEV1%. Temperature and passive smoking also have a significant inverse association with FEV1%. In LME with ACP, considering all pollutants, there is a 10% reduction in FEV1%, which is more significant compared to pollutants alone. The findings of this thesis suggest that PM10, PM2.5, and SO2 pollutant concentration and temperature level are associated with reduced lung function in children and adolescents.