Using a physiologically based pharmacokinetic (PBPK) model, the effects of variation of exposure concentration of acetone on three biological indicators--acetone concentrations in blood, urine, and exhaled air--were investigated. The effect of the difference in work load was also examined. It was confirmed that the model could be used to estimate acetone concentrations during fluctuating exposure by comparing simulated acetone concentrations with the corresponding values observed in field surveys. By inputting the exposure situations into the PBPK model, the variabilities of the biological indicators were simulated. The variation of acetone exposure was expressed by seven 1-hour time-weighted averages (CEXPs). The arithmetic means of the CEXPS were 200 and 750 ppm. The geometric standard deviations (GSDs) were 1.5, 2.0, and 3.0, representing low, moderate, and high variations, respectively. Work loads were set at 15 and 50 W. Consequently, there were 12 exposure situations. The acetone concentrations in venous blood (CB) and exhaled alveolar air (CA) at 1 minute after the end of the work shift were selected as biological indicators of exposure because they were predicted to decrease rapidly at the end of exposure and become relatively stable after 1 minute. The acetone concentration in urine excreted during the last 2 hours of the work shift (CU) was also used as a biological indicator. Simulation was repeated 100 times with randomly permuting CEXPs for each situation. The mean values of CB, CU, and CA showed almost no variation regardless of the difference in the GSD of CEXPs. The coefficients of variation increased with the GSD of CEXPs but were less than 0.2. Consequently, these variables were acceptable as biological indicators of daily average exposure for the same work load. However, the difference in work load greatly changed the mean values of CB, CU, and CA, thus making it difficult to use these variables as indicators of daily average exposure for different work loads.