Methods: This prospective cohort study followed 447 cotton textile workers from 1981 to 2006.at approximately 5-year intervals. We used a generalized estimating equations approach to model FEV1 level and respiratory symptoms as a function of past exposure (cumulative exposure up to the start of the most recent 5-year survey interval) and cumulative exposure (within the most recent interval) to endotoxins, after adjusting for other covariates. Models were stratified by active versus retired work status and by years employed before the baseline survey (< 5 and ≥ 5 years).