Objective The advantages and disadvantages of three statistical models commonly used in food contaminant exposure assessment， namely the observed individual means （OIM） model， the beta-abinomial and normal （BBN） model and the non-parametric model were compared by an example of the estimation of long-term exposure to cadmium via rice.Methods Rice， cadmium and diet were used as Chinese and English keywords to search the literature on cadmium concentration in rice in China National Knowledge Infrastructure， Wanfang Data Knowledge Service Platform and PubMed database. Long-term levels of exposure to cadmium via rice for the total population and age groups in China were estimated by combining partial consumption frequency data obtained from the Chinese nutrition survey.Results In the total population， the OIM model showed that the 2.5^th to 97.5^th percentile （P2.5、P97.5） of the exposure to heavy metal Cd in rice was 0.081-0.576 μg/（kg·BW·d）， the non-parametric model result was 0.081-0.573 μg/（kg·BW·d）， and the BBN model result was 0.104-0.611 μg/（kg·BW·d）. The average results of the OIM model， non-parametric model and BBN model in different populations were close. The average values of the three models in the total population were 0.278， 0.277 and 0.278 μg/（kg·BW·d）， respectively.Conclusion With large sample data， non-parametric models have similar assessment results to the OIM model， while BBN models can allow for a more conservative assessment of exposure by subtracting differences in consumption frequency within individuals， and better evaluate the long-term exposure level of pollutants.