Sera from 483 patients at high (group 1, n = 313) and lower (group 2, n = 170) risk for exposure to hepatitis C were tested for antibodies to hepatitis C using first-generation (c100-3) and second-generation enzyme-linked immunosorbent assays and four-antigen recombinant immunoblot assay. The second-generation enzyme-linked immunosorbent assay and nitrocellulose-based immunoblot assay differ from c100-3-based systems in the addition of expression products from the NS3/NS4 (c33c, c200) and putative nucleocapsid (c22-3) region of the hepatitis C genome. In group 1, the sensitivity of detection of hepatitis C antibodies was 45%, 55% and 46% by the first- and second-generation enzyme-linked immunosorbent assays and recombinant immunoblot assay, respectively. In group 2, antibodies were detected by each test system in 26%, 32% and 7% of patients, respectively. Most sera (99%) reactive with the first-generation enzyme-linked immunosorbent assay were reactive with the second-generation enzyme-linked immunosorbent assay (in group 1, 89% of these specimens demonstrated reactivity to at least one antigen with the immunoblot assay, compared with only 31% in group 2). An additional 12% (group 1) and 6% (group 2) of specimens demonstrated reactivity with the second-generation enzyme-linked immunosorbent assay only (of these, 75% [group 1] and 9% [group 2] demonstrated reactivity to at least one antigen with the immunoblot assay). Ninety-eight percent of specimens not reactive with both enzyme-linked immunosorbent assay test systems were also nonreactive by recombinant immunoblot assay. Antibodies to c22-3 and c33c were more frequently detected in chronic hepatitis C virus infection and appeared earlier in acute hepatitis C virus infection, compared with 5-1-1 and c100-3. These results indicate that the addition of structural and nonstructural hepatitis C virus antigens in the second-generation enzyme-linked immunosorbent assay improves the sensitivity of detection of hepatitis C antibodies. (HEPATOLOGY 1992;15:19-25).