The liver injury induced by Polygonum multiflorum Thunb. (PM) was investigated based on idiosyncratic hepatotoxicity model co-treated with lipopolysaccharide (LPS) at a non-hepatotoxic dose. Sprague-Dawley (SD) rats were intragastrically administered with three doses (18.9, 37.8, 75.6 g crude drug per kg body weight) of 50% alcohol extracts of PM alone or co-treated with non-toxic dose of LPS (2.8 mg·kg(-1)) via tail vein injection. The plasma alanine aminotransferase (ALT) and aspartate aminotransferase (AST) activities were assayed and the isolated livers were evaluated for histopathological changes. The dose-toxicity relationships of single treatment of PM or co-treatment of LPS were investigated comparatively to elucidate the idiosyncratic hepatotoxicity of PM. The results showed that no significant alterations of plasma ALT and AST activities were observed in the groups of solo-administration of LPS (2.8 mg·kg(-1), i.v.) or different dosage (18.9, 37.8 and 75.6 g·kg(-1), i.g.) of PM, compared to normal control group (P > 0.05); while significant elevations were observed in the co-administration groups of PM and LPS. Treatment with LPS alone caused slight infiltration of inflammatory cells in portal area but no evident hepatocytes injury. Co-treatment with LPS and PM (75.6 g·kg(-1), i.g.) caused hepatocyte focal necrosis, loss of central vein intima and a large number of inflammatory cell infiltration in portal areas. When further reduce the dosage of PM, significant increases of plasma ALT and AST activities (P < 0.05) were still observed in co-administration groups of LPS and PM (1.08 or 2.16 g·kg(-1)), but not in LPS or PM solo-administration groups. Nevertheless, the co-treatment of low dosage of PM (0.54 g·kg(-1)) with LPS did not induce any alteration of plasma ALT and AST. In conclusion, intragastric administration with 75.6 g·kg(-1) of PM did not induce liver injury in normal rats model; while the 2 folds of clinical equivalent dose of PM (1.08 g·kg(-1)) could result in liver injury in the LPS-based idiosyncratic hepatotoxicity model, which could be used to evaluate the idiosyncratic hepatotoxicity of PM.