Objective: To evaluate a WHO testing strategy based on the use of two consecutive enzyme-linked immunosorbent assays (ELISA) as an alternative to ELISA followed by Western blotting (WB) for the serologic diagnosis of HIV infection.
Study design: The study was of 2069 consecutive serum specimens from patients suspected of HIV infection received for HIV diagnostic testing at the HIV laboratory, Muhimbili Medical Centre, Dar es Salaam. The strategy involved testing all sera with Behring indirect anti-HIV 1 + 2 peptide ELISA, followed by Wellcozyme anti-HIV-1 recombinant competitive ELISA on those sera reactive by the first ELISA. WB was done on a sample of the sera reactive on both ELISAs and on all those giving discordant results on the two ELISAs. Of the 2069 sera tested, 837 (40.5%) were negative on the first ELISA, 1172 (56.6%) were reactive on both ELISAs and 60 (2.9%) were initially reactive on the first test but not on the second assay.
Results: Of the 1172 sera reactive on both ELISAs, 329 (28.1%) were tested by WB. The diagnostic accuracy of the WHO alternative testing strategy using WB confirmation as the 'gold' standard was as follows: sensitivity 99.4% (326/328), specificity 99.7%, (893/896), positive predictive value 99.1% (328/331) and negative predictive value 99.8% (893/895). Repeated testing by ELISA of the sera which initially gave discordant results on the two ELISAs increased the sensitivity to 100%. Three sera giving false positive reactions on both ELISAs became negative on both ELISAs after retesting. In order to achieve a specificity and a positive predictive value of 100%, it would have been necessary to subject all sera reacting on both ELISAs to retesting on one ELISA.
Conclusions: A second ELISA based on different antigens and a different test principle compared with the first ELISA could be used as an alternative to the WB assay for confirmation of HIV antibodies. However, some modifications of the WHO strategy for diagnostic HIV antibody testing were required in order to maximize the diagnostic accuracy.