Recombinant human endostatin (rhEndostatin) has been shown to inhibit tumor growth, but the variable antitumor activity of different rhEndostatin preparations has necessitated the development of an accurate, reproducible in vivo bioassay for evaluating the rhEndostatin activity. To assess the in vivo antitumor efficacy of rhEndostatin, H22 tumor-bearing mice received three doses of rhEndostatin and the potency of rhEndostatin preparations in inhibiting tumor growth was determined by ED(50)-potency assay and validated by dose-response parallel-line assay. There was a consistent and highly reproducible linear regression relationship between rhEndostatin dosage and tumor growth inhibition rate. The ED(50) values were determined from dose-response regression lines for seven rhEndostatin preparations with high reproducibility. On the basis of the current study, the potency of rhEndostatin preparations was assigned a value of 6.09 x 10(5) U/ampoule and a 95% confidence limit of 5.96 x 10(5)-6.22 x 10(5). We consider that this procedure can be served as a potential candidate pharmacopoeial method for potency measurement of different rhEndostatin preparations.