Métodos de estadística espacial para evaluar la influencia de factores medioambientales sobre la incidencia y mortalidad por cáncer

Abstract

The objectives of the Thesis are: 1.- To study the relationship between incidence and cancer mortality and environmental factors, particularly atmospheric pollutio, using socioeconomic factors as control. 2.- To use those methods of spatial statistics suitable for each type of design. 3.- To distinguish different sources of spatial extra-variability between the models. 4.- To control the problem of excess zeros inherent in some of the environment-related neoplasias. 5.- To correctly introduce explanatory variables (according to the design oberved in them) of the geographical variability of the environment-related neoplasias in order to avoid possible biases. The conclusions are: -Both the incidence and mortality of environment-related neoplasias showed two wources of extra-variation: spatil extra-variation, in which neighbouring units tend to show similar incidence/mortality ratios, and non-spatial heterogeneity. -To smooth the SIR/SMR corresponding to variables with a precentage of zeros higher than 40-50% a model that captures this behaviour must be used. -The best model, in terms of fit, which recognises excess zeros in the variables of interest was the mixture relative risk model. -Smoothed SIR/SMR only showed a clear geographical pattern in some of the environment-related neoplasias. -Part of the remaining variability in the smoothed SIR/SMR could be explained by introducing explanatory variables, in particular atmospheric pollution and socieconomic variables. -As the atmospheric pollutants were observed in a geostatistical design and the environment-related neoplasias were studied in a lattice design, the exposure surface was modelled. -The effect of the pollutant in each municipality/electoral district as approached in the model by introducing the average value in each area and the intra-area variability. -Socioeconomic conditions were another of the variables that reduced the remaining variability in the smoothed SIR/SMR. -Explanatory variables observed with a lattice design, such as the privation index, were introduced in the model as fixed effects. -Greater privation, using atmospheric pollutans as control, implied a greater risk of suffering/dying from some of the environment-realted neoplasias in the three areas and periods studies. -The effect of privation on the incidence and/or mortality form throat, bronchial and lung cancer, using atmospheric pollutans as control, was greater in women than in men.