Endotoxin exposure has been implicated in the etiology of lung disease in cotton workers. We investigated this potential relationship in 443 cotton workers from 2 factories in Shanghai and 439 control subjects from a nearby silk mill. A respiratory questionnaire was administered and pre- and postshift forced expiratory volume (FVC) and flow in one second (FEV1) were determined for each worker. Multiple area air samples were analyzed for total elutriated dust concentration (range: 0.15 to 2.5 mg/m3) and endotoxin (range: 0.002 to 0.55 microgram U.S. Reference Endotoxin/m3). The cotton worker population was stratified by current and cumulative dust or endotoxin exposure. Groups were compared for FEV1, FVC, FEV1/FVC%, % change in FEV1 over the shift (delta FEV1%), and prevalences of chronic bronchitis and byssinosis, and linear and logistic regression models were constructed. No dose-response relationships were demonstrated comparing dust concentration to any pulmonary function or symptom variable. A dose-response trend was seen with the current endotoxin level and FEV1, delta FEV1%, and the prevalence of byssinosis and chronic bronchitis, except for the highest exposure level group in which a reversal of the trend was seen. The regression coefficients for current endotoxin exposure were significant (p less than 0.05) in the models for FEV1 and chronic bronchitis but not in the models for delta FEV1% (i.e., acute change in FEV1) or byssinosis prevalence. The coefficient for dust level was never significant in the models.(ABSTRACT TRUNCATED AT 250 WORDS)