Inhaled corticosteroids in chronic obstructive pulmonary disease: results from two observational designs free of immortal time bias