BACKGROUND: Paraquat has been widely used as a non-selective contact herbicide and it may induce damage to many organs. This study aimed to assess the factors that can predict the prognosis of paraquat poisoning and to determine the effect of hemoperfusion. METHODS: We retrospectively reviewed 132 patients who were poisoned with paraquat from January 2005, to December 2008. The patients were divided into two groups: The first groups included the death and survived groups, and the second groups included the hemoperfusion and non-hemoperfusion groups. We investigated the mortality, the factors that can predictive the prognosis and the effect of hemoperfusion. RESULTS: There were 79 males and 53 female (mean age: 56.1 +/- 15.1 years). The significant differences between the death and survival groups were the volume of paraquat ingested, the mental status, GCS (Glasgow coma scale), pH, base deficit, HCO3, serum Cr (creatinine), serum AST (aspartate transaminase), serum glucose, K (kalium), urine sodium dithionite test and hemoperfusion. The significant differences between the hemoperfusion and non-hemoperfusion groups were the mortality and the mean survival time. Multivariate regression analysis reveled four predictive factors and their's Odd ratio: 1) urine sodium dithionate test = strong 14.256, 2) hemoperfusion 0.493, 3) Cr > 0.95 mg/kg 31.603 and 4) an amount of ingested paraquat > 45 ml 16.945. CONCLUSIONS: The predictive factors for mortality were the amount of paraquat ingested > 45 ml, a urine sodium dithionite test = strong and a serum Cr > 0.95 mg/dl. Hemoperfusion couldn't be used a predictive factor for mortality, but it increased the mean survival time.