A variety of albumin polymers were prepared and tested for binding with hepatitis B surface antigen (HBsAg): synthetic polymers cross-linked by either glutaraldehyde or carbodiimide; heat-aggregated polymers made by heating albumin solutions at 60 degrees C for 10 h with or without albumin stabilizer; and polymers isolated from fresh or long-stored commercial therapeutic albumin solutions. A sensitive solid-phase, competitive-inhibition radioimmunoassay, which can detect as little as 10 ng of glutaraldehyde-cross-linked human albumin polymer (PHALB-G), was developed and used to measure binding. The binding of PHALB-G with HBsAg was 150- to 1,000-fold greater than that of any other albumin polymer. Glutaraldehyde-cross-linked bovine albumin polymer showed no binding. Albumin monomer and dimer fractions produced by glutaraldehyde treatment exhibited some binding, albeit much weaker than PHALB-G. As measured by a direct-binding assay with solid-phase PHALB-G, the attachment of HBsAg particles from sera positive for antibody to the e antigen was less efficient than that from sera positive for e antigen, even when all sera were tested at equal HBsAg concentrations. In protein blot experiments with radiolabeled albumin preparations, PHALB-G bound almost exclusively to HBsAg polypeptide P31 and showed no binding with the major polypeptides P23 and P26. None of the other radiolabeled albumin polymers was reactive. These results indicate that the interaction between PHALB-G and HBsAg is not due to polymerization of albumin per se, but rather is unique and site specific.