To improve the detectability of gastrointestinal (GI) bleeding, we have developed a new method, 'sequential subtraction scintigraphy' (SSS) with 99Tcm-labelled red blood cells (RBC). Sequential subtraction scintigrams were obtained by subtracting 5 min interval images from each other using a digital gamma camera. With this method, we are able to detect the bleeding site and calculate the bleeding rate earlier because of the lower background activity. The results of the present study demonstrate that the minimum detectable bleeding rate was 0.05 ml min-1 in both a phantom and an animal model. The minimum detectable activity volumes within 10 min were 0.5 ml by SSS and more that 3 ml by conventional non-subtraction scintigraphy (CNS). The bleeding rates calculated with the subtraction method correlated well with the actual syringe pump rates. The detectability of GI bleeding by SSS and CNS in animal models was compared using the area under receiver operator characteristic (ROC) curves. The results showed that SSS significantly increased the area under the ROC curves from 0.55 +/- 0.05 to 0.79 +/- 0.04 for the 10 min image set and from 0.65 +/- 0.05 to 0.83 +/- 0.04 for the 30 min image set (Z = 5.45, P < 0.0001 and Z = 4.36, P < 0.0001, respectively). These results suggest that sequential subtraction scintigraphy with 99Tcm-RBC is an effective method for the early detection of GI bleeding when compared with the conventional non-subtraction method.