Objective　To evaluate the clinical value of sequential subtraction scintigraphy (SSS) with 99m　Tc red blood cell (RBC) for the early detection of gastrointestinal (GI) bleeding. Methods　Ninety patients referred with clinical evidence of GI bleeding underwent 99m　Tc-RBC scintigraphy with digital gamma camera after labeling of 99m　 Tc-RBC in vivo. Sequential 12 conventional non-subtraction (CNS) images of abdomen were taken at 5?min intervals for 60?min. Then 5-min images were subsequently subtracted from each other using a computer and 11 subtracted images were obtained. Delayed images up to 24 hr were obtained when early results were negative and/or recurrent bleeding was suspected. Results　62 of 90 patients with suspected GI bleeding were conformed to have active hemorrhage up to 24?hr. The scintigrams were divided into the following three sets: within 30?min, 60?min, and 24?hr. The sensitivity of SSS was 87% (54/62, 30?min) and 91.9% (57/62, 60?min) respectively, which were significantly higher (P<0.05) than that of CNS (56.4%, 35/62 and 62.9%, 39/62). 24?hr delayed image of CNS increased the sensitivity to 85.4%. No significant difference in specificity between the two methods was noted. Of the 62 patients with definite active hemorrhage, the bleeding sites were identified by surgical operation in 42. The concordant rate with surgical operation in SSS was 92.8% (39/42), higher than that of CNS (73.8%, 31/42). Conclusions　Sequential subtraction scintigraphy with 99m　 Tc-RBC, compared with conventional non-subtraction scintigraphy, is an effective technique for the early detection of GI bleeding. It increases the diagnostic sensitivity, detecting a small amount of bleeding earlier than CNS; SSS reduces background activity, more accurately localizing bleeding sites. It also shortens the examination time, making this approach more suitable for pediatric, elderly and critically ill patients.