       Document 2741
 DOCN  M94A2741
 TI    Selection of appropriate strategies for HIV testing.
 DT    9412
 AU    Vercauteren G; van der Groen G; Institute of Tropical Medicine, Antwerp,
       Belgium.
 SO    Int Conf AIDS. 1994 Aug 7-12;10(1):235 (abstract no. PB0369). Unique
       Identifier : AIDSLINE ICA10/94369834
 AB    OBJECTIVE: To evaluate the WHO recommended HIV testing strategies. To
       investigate the accuracy of a computer programme to predict the outcomes
       of certain test combinations in a defined test algorithm. METHODS: Five
       algorithms, including the three WHO algorithms (Lancet 1993;342:87-90),
       a fourth algorithm (simultaneous testing of all sera with two different
       screening assays A and B, considering A+B+ = positive; AB = negative;
       A+B/AB+ = indeterminate) and the conventional algorithm (A+ confirmed
       with Western blot) were investigated. The screening results of the 7
       different HIV screening tests as mentioned on the WHO bulk purchasement
       list (Global AIDS News 1992;1:13) obtained on the same well
       characterized serum panel (n = 450; HIV-1+: 42%; HIV-2+: 4.6%) were used
       for the retrospective analysis. The outcomes of the five algorithms
       using different test combinations were calculated retrospectively and
       compared with the outcome of computerized formula (introduced in a Lotus
       spreadsheet). The intrinsic sensitivity and specificity of each assay
       used in a particular algorithm, the sample size, and the HIV antibody
       prevalence, must be introduced in the computer to calculate the outcome
       of a particular algorithm. RESULTS: Several test combinations using the
       evaluated algorithms are capable to give equally good results as the
       conventional algorithm. The cost per sample to obtain these results was
       two to ten times cheaper as compared to the conventional algorithm
       depending on the alternative algorithm and/or test combination used.
       Only by using 2 assays simultaneously, it is possible to increase the
       sensitivity of the first assay. The data generated by the computerized
       formula were comparable to the data obtained retrospectively. The
       computer data predicted a decline in the number of false positives with
       an increasing HIV prevalence for each algorithm. We observed also a
       decline in the number of false positives from algorithms I towards
       algorithm II, III and IV respectively. The accuracy of all investigated
       test combinations was above 90% even at low prevalences. CONCLUSIONS:
       The computer programme can be used to predict accurately the outcome and
       the cost of a certain test combination in a particular algorithm at
       different HIV prevalences when the intrinsic sensitivity and specificity
       of the assays are known, without actually performing any tests. This can
       be a valuable tool for laboratory directors and policy makers in their
       choice of HIV assays and their decision of an HIV testing strategy. The
       data generated by the computerized formula for the evaluated algorithms
       at the different prevalences do support the proposed WHO strategies.
 DE    Algorithms  AIDS Serodiagnosis/ECONOMICS/*METHODS  Belgium  Comparative
       Study  Cost-Benefit Analysis  Human  HIV Infections/*DIAGNOSIS  HIV
       Seropositivity/DIAGNOSIS  HIV Seroprevalence  Predictive Value of Tests
       Software  World Health Organization  MEETING ABSTRACT

       SOURCE: National Library of Medicine.  NOTICE: This material may be
       protected by Copyright Law (Title 17, U.S.Code).

