“…Considering the highly skewed distribution of redox markers and contaminants, the magnitude of the associations was analyzed by means of robust regression based on the MM-estimator . The covariates included were initially selected from among those most commonly used in the available literature. ,, Final covariate selection in the multivariable models was performed by using a combination of forward and backward stepwise methods. The model 1 was adjusted for age, sex, BMI (continuous), place of residence (urban/semirural), educational level (no studies/primary or higher), smoking (never, former, current), and alcohol consumption (regular consumer/nonconsumer) and type of work (nonmanual worker, manual worker, retired), while model 2 was adjusted by these covariates but also by the consumption of vegetable foods.…”