Modification of hemoglobin (Hb) by crosslinking and polymerization results in an improved oxygen release capacity and a prolonged vascular retention time. Modification improves the efficacy and prevents certain side effects. It eliminates leakage of Hb through the kidneys and accumulation in the tubuli. Another important issue is the degree of purification of Hb solutions. Traces of membrane fragments may cause immunogenic and thrombogenic side effects. To determine the contamination with erythrocyte membrane fragments, we developed assays for glycophorin-alpha and phospholipids. Special models were evaluated for testing the maximum allowable level of membrane contamination. As an in vitro model for thrombogenicity we used confluent monolayers of human umbilical vein endothelial cells. These cells were incubated with Hb solutions and subsequently tested on tissue factor (TF) procoagulant activity. TF was tested by the factor VII-catalyzed activation of factor X. The lower detection limit of this assay for endotoxin was 0.5 ng/ml. Hb did not cause any tissue factor expression even after prolonged incubation. No cooperation was found within endotoxin. As an in vivo test on thrombogenicity we developed a guinea pig model in which we can follow the generation of fibrinopeptide A (FPA). This is one of the most sensitive markers for thrombin activation in vivo. When slightly contaminated Hb solutions (phospholipid content 2 nmol/ml) were infused in the presence of factor Xa at a dose (9 micrograms/kg) which in itself did not induce FPA generation, we observed an increase in FPA levels in the plasma from 1.2 +/- 0.4 ng/ml to 5.2 +/- 0.7 ng/ml. Factor Xa is used to mimic a stressed clinical condition with activated coagulation.(ABSTRACT TRUNCATED AT 250 WORDS)