Methodological constraints in interpreting serum paraoxonase-1 activity measurements: an example from a study in HIV-infected patients