Previous work from this laboratory has shown that injection of thrombin into rat basal ganglia causes brain edema. This study investigates the effect on rat brain of thrombin-soaked gelatin sponge (used for intraoperative hemostasis in clinical situations) at a concentration similar to that used in humans. Three models were developed to evaluate this effect. In the first model, a gelatin sponge soaked with vehicle or thrombin (100 U/cm3) was placed on the intact pia of the right frontal lobe in rats without cortical lesions. In the second model, frontal cortex was excised (3 mm3) and the exposed brain was cauterized with electrocoagulation. Gelatin sponge was soaked with vehicle or thrombin (1000, 100, 10, or 1 U/cm3) and placed in the lesion site. In the third model, hirudin, a specific thrombin antagonist, was added to the thrombin-soaked gelatin sponge and placed in a similar cortical lesion to determine if the observed effects were specific to thrombin. The dose-response range for thrombin was determined qualitatively by magnetic resonance (MR) imaging and quantitatively by brain edema formation 24 hours after exposure. We found no edema in the cortically intact rats. The rats given cortical lesions developed significant edema when subjected to 1000, 100, and 10 U/cm3 thrombin as seen on MR imaging and at 100 and 10 U/cm3 thrombin as revealed by wet/dry weight and ion studies of brain tissue. Topical hirudin prevented thrombin-induced edema. It is concluded that thrombin-soaked gelatin sponges cause or enhance significant brain edema in rats at concentrations typically used for human neurosurgery.