Objective Based on the reference dose （RfD） of deoxynivalenol （DON）， the applicability of probabilistic RfD was explored.Methods Based on the uncertainty analysis guidelines established by the World Health Organization， an experimental study of chronic oral exposure to DON in mice was performed to derive RfD. The probability distribution of the reference dose of DON was calculated by the approximate probabilistic method and the complete probabilistic method， respectively. Then， sensitivity analysis was conducted to study the ability of the two probabilistic methods to analyze uncertainty factors.Results The complete probabilistic RfD of DON was 0.42 μg/kg·BW·d， which was higher than the approximate probabilistic reference dose of 0.30 μg/kg·BW·d， and the complete probabilistic method had an uncertainty spanning factor of 42， which was slightly lower than factor of 46 of the approximate probabilistic method， suggested that complete probabilistic RfD had lower conservatism. Sensitivity analysis showed that the complete probabilistic method greatly reduced the uncertainty caused by point of departure （POD）.Conclusion The complete probability method has advantages over the approximate probabilistic method in analyzing uncertainty， especially in POD aspects. If the data can satisfy the Bayesian analysis of dose-response model， the use of the complete probabilistic method is recommended.