Enzyme-linked immunosorbent assay (ELISA) was used to detect the presence of protein-acetaldehyde adducts (-AAs) in human serum samples. Two methods were compared: (1) direct ELISA: samples, rabbit anti-hemocyanin-AA IgG, and beta-galactosidase (beta-gal) conjugated goat anti-rabbit serum IgG added to a 96-well ELISA plate in a stepwise manner; and (2) two-site or sandwich ELISA: serum samples added to an ELISA plate that had been precoated with anti-hemocyanin-AA IgG (the capture antibody) and incubated stepwise with biotinated anti-hemocyanin-AA IgG (the signal antibody) and avidin-beta-gal conjugates. Serum protein-AA levels were then assayed by bound beta-gal activities at OD405. When human hemoglobin (Hgb)-AA was used as a model protein-AA for the sandwich ELISA, the EC50 (estimated concentration that corresponds to 50% of the OD405 response range) was 7 ng/ml. Direct ELISA was less sensitive (EC50 of 120 ng/ml). Adding control human serum to Hgb-AA increased the EC50 of the direct ELISA more than the sandwich ELISA. Intra- and interassay coefficients of variance for sandwich ELISA were both about 8%. Detection of Hgb-AA by sandwich ELISA was highly specific. The above results with anti-hemocyanin-AA IgG were also obtained when anti-myoglobin-AA IgG was used in sandwich ELISA. Using sandwich ELISA and anti-hemocyanin-AA IgG, OD405 for sera of control subjects and alcoholic patients were 0.036 +/- 0.033 (+/- SEM, n = 28) and 0.150 +/- 0.088 (n = 28), respectively. Serum protein-AAs reacted more strongly with anti-myoglobin-AA IgG than anti-hemocyanin-AA IgG.(ABSTRACT TRUNCATED AT 250 WORDS)