Screening for anti-HTLV-III/LAV antibody in high-risk subjects: sensitivity and specificity of commercial tests

1,834 serum samples from high-risk subjects were tested for anti-HTLV-III/LAV antibody by using the enzyme immuno assay (EIA) of Abbott (A), or the EIA of Abbott and Pasteur (P), or those two EIA plus Western blot analysis (WB). 983 samples were negative in A, 766 were strongly positive when tested in duplicate with both A and P, and 85 were slightly positive in A (23 samples) or discordant (62 samples) when tested further in P and WB. Of 62 discordant samples, 38 were reproducible. The pattern A+P-WB- was by far the most frequently encountered (26 samples) and interpreted as reflecting the higher incidence of positive results in the assay A. If one assumes that WB results are correct, the specificities of A and P are 93 and 100%, respectively (p less than 0.01), whereas the sensitivities are similar (98 and 99%).