Simplified and less expensive confirmatory HIV testing.

Simplified and less expensive confirmatory HIV testing* G. van der Groen,1 1. Van kerckhoven,2 G. Vercauteren,2 & P. Plot3 The conventional approach to human immunodeficiency virus (HIV) antibody testing, which relies on con- firmation of all initially positive screening results using a Western blot assay, is expensive. In an alterna- tive approach, we retested sera that were positive in an initial screening assay using a second screening assay, which differed from the first, and limited the use of Western blot to those sera that gave discrepant results in the two screening assays. This resulted in 100% sensitivity and specificity at a cost that was, on average, 6.1 times less than that of the conventional approach. This level of sensitivity and specificity was also achieved at a cost that was 9.0 times less than the conventional approach if the Western blot was replaced by a third screening assay that differed from the previous two. Retesting positive sera using the same assay did not increase the accuracy of the results obtained by testing the sera only once. Introduction Enzyme-linked immunosorbent assays (ELISAs) for the detection of antibodies to human immuno- deficiency virus (HIV) were first introduced in 1985. At present, most commercially available ELISAs for HIV have excellent sensitivity and specificity. If we assume a sensitivity and specificity of 99.9%, the pro- portion of false positives obtained with such ELISAs increases from less than 10% to over 70% as the prevalence of HIV in the population tested decreases from 1.0% to 0.04% (1). Therefore, initially positive ELISA results should always be confirmed. In the conventional testing strategy, a positive ELISA screening test is followed by confirmation using Western blot assay. However, this approach has several weaknesses (2-4): Western blot is costly and time-consuming, needs sophisticated equipment, is difficult to interpret and standardize, and can yield indeterminate results. To reduce the number of Western blots required in the conventional approach, we have evaluated the use of pairs of HIV-antibody screening assays, with an emphasis on the more recently developed simple, non-ELISA assays (5, 6) and second-generation assays that employ recombinant HIV proteins or synthetic peptides (7, 8). The use of this alternative approach could lead to accurate, cheaper, more rapid and/or less equipment-reliant testing for HIV infection. * From the Department of Microbiology, Institute of Tropical Medicine, Nationalestraat 155, 2000 Antwerp, Belgium. ' Deputy Director. Requests for reprints should be sent to Dr van der Groen. 2 Research worker. 3 Director. Reprint No. 5232 Materials and methods Sera. A panel of 361 human sera (220 of which were of African, 131 of European, and 10 of South Amer- ican origin) was used. A total of 164 (45%) of the sera were HIV-1 antibody positive. All samples were stored in aliquots and thawed at least once, but at most four times. Reference test. The results obtained using the various assays and combinations of assays were compared with those obtained using HIV-1 Western blot (Du Pont). A Western blot result was considered uositive when at least the bands representative of both HIV-specific core and env proteins were present. HiV antibody assays. The majority of the com- mercially available HIV assays shown in Table 1 were performed on all sera in the panel, according to the instructions provided by the manufacturers; some assays were, however, performed on a smaller number of sera, because the panel was incomplete when they were tested. For the HIV Chek assay, 'all sera were pretreated with Lipoclean (Behring) to facilitate penetration of the membrane. Data analysis. Initially all the assays were performed on all sera of the panel; the results that would have been obtained for combinations of pairs of assays using the alternative algorithms (Fig. 1) were then analysed retrospectively. The sensitivity and specificity of the different assays and assay combinations were calculated taking the Western blot results as the standard. The sensitivity of the individual assays was calculated using the initial assay results, and the specificity Bulletin of the World Health Organization, 69 (6): 747-752 (1991) fD World Health Organization 1991 747 G. van der Groen et al. Table 1: Characteristics of the HIV antibody assays studied Assay Type of Type of Cost per test (abbreviation) Manufacturer antigen assay (US$)' Du Pont HIV-1 Du Pont Recombinant Indirect ELISA 0.9 Rec ELISA proteins (EIA D) Recombinant Abbott Recombinant Indirect ELISA 3.3 HIV-1 ELISA proteins (EIA A) Vironostika Organon HIV lysate Indirect ELISA 2.0 anti-HTLV-I1 Teknika (EIA V) Enzygnost Behring HIV lysate Competitive ELISA 1.8 Anti-HIV Micro (EIA B) Wellcozyme Wellcome HIV lysate Competititve ELISA 1.0 anti-HTLV-III (EIA W) Recombigen Cambridge Recombinant Agglutination 3.0 HIV-LA BioScience proteins assay (LA CB) Serodia-HIV Fujirebio HIV lysate Agglutination 1.1 (AG FU) assay HIV Chek Du Pont Recombinant Immunodot 2.5 (CHEK) proteins assay Immunocomb PBS Orgenics Synthetic Immunodot 2.5 (ICOMB) peptides assay HIV-1 Western blot Du Pont HIV lysate Immunoblot 44.5 (WB) assay a Costs were provided by the distributor of the test in Belgium during the period of testing (July 1988-January 1989). Costs may vary according to country and the number of test kits ordered. Fig. 1. Schematic representation of algorithm 3: alterna- tive confirmatory strategy using Initially positive test results (* = different assay from the first). Algorithm 4 = algorithm 3, except that repeatedly positive sera in the first and second assays were used; algorithm 5 = algorithm 3, except that the third assay differed from the first and second; and algorithm 6 = algorithm 5, except that the repeatedly positive sera in the first, second and third assays were used. First assay + Second Report: assay* negative + Report: Western positive blot + Report: Report: positive negative using the results obtained after initially positive sera had been retested. Results The sensitivity and specificity of each of the HIV antibody assays are shown in Table 2. The results indicate that pretreatment of the sera used in the HIV Chek assay with Lipoclean seems to have reduced its sensitivity. Use of ELISA to screen sera for HIV and con- firmation of positive results using Western blQt is expensive: on average, US$ 19.7 per test if each serum is screened only once (algorithm 1) and US$ 21.5 per test if positive sera are retested using the screening assay before confirmatory testing (algorithm 2) (Table 3). In an alternative approach, sera found to be positive in the initial screening assay were retested using a different screening assay, followed by a Western blot only on those sera that gave discrepant screening assays (algorithms 3 and 4, Fig. 1); this resulted in 100% sensitivity and specificity at a cost that was, on average, 6.1 times and 4.7 times less, respectively, than that of the conventional algo- rithms 1 and 2 (Table 4). WHO Bulletin OMS. Vol 69 1991748 Simplified and less expensive confirmatory HIV testing Table 2: Sensitivity and specificity of the HIV antibody screening assays used in the study No. of Sensitivity Specificity Assay sera (%)- (%)b Du Pont HIV-1 361 100 91 Rec ELISA (98.0-100)c (86.0-94.5) Abbott Recom- 361 100 98 binant (97.8-100) (95.2-99.6) HIV-1 ELISA Enzygnost 361 100 100 Anti-HIV Micro (98.0-100) (98.0-100) Vironostika 361 100 99 anti-HTLV-11I (97.8-100) (96.9-100) Wellcozyme 361 100 100 anti-HTLV-111 (97.8-100) (98.0-100) Recombigen 314 95 96 HIV-LA (93.0-98.2) (93.0-98.2) Serodia-HIV 351 100 97 (98.0-100) (93.0-99.0) HIV Chek 361 95 99 (90.0-94.5) (96.0-100) Immunocomb 361 99 99 (96.0-99.9) (96.0-100) 8 The % sensitivity was determined by dividing the number of sera that were positive by Western blot, and which were initially positive in the HIV antibody test, by the total number of sera that were confirmed positive by Western blot, and multiplying this by 100. b The % specificity was determined by dividing the number of sera that were confirmed negative by Western blot, and which were negative in the HIV antibody test, by the total number of sera confirmed negative by Western blot, and multiplying this by 100. Initially false-positive sera that twice produced a negative result after repeat testing were included in the calculation. c Figures in parentheses are the 95% confidence limits. These levels of sensitivity and specificity for algorithms 3 and 4 were also achieved when instead of the Western blot a third screening assay that dif- fered from the previous two was used (algorithms 5 and 6, Fig. 1); in this case the cost was, on average, 9.0 times and 6.9 times lower, respectively, than that of the conventional algorithms 1 and 2 (Table 5). Not only combinations of three different ELISAs (Table 5, No. 53, 54), and of ELISAs with simple, non-ELISA tests (Table 5, No. 43-52), but also combinations of three different simple, non- ELISA tests (Table 5, No. 37-42) resulted in 100% sensitivity and specificity. These combinations cost, on average, 7.8 and 5.0 times less than the con- ventional algorithms 1 and 2. The cost per test was US$ 2.0-2.4 with algorithm 5 (Table 5, No. 38, 40, 42). Although a relatively small number of sera were tested, the test combinations in Tables 4 and 5 had 100% sensitivity and specificity, with the exception of the combination Immunocomb + HIV Chek (Table 4), whose sensitivity was only 98.8%. This arose because the sensitivity of Immunocomb, the first assay in the combination, was 98.8%. Discussion When a combination of two or three HIV screening assays is employed, and use of the expensive Western Table 3: Cost per test for HIV antibody screening using the conventional algorithms 1 and 2, with Western blot to confirm all positive screening results (see text)" Test combinationb No. of sera No. of to be confirmed Cost per test No. Algorithm 1st 2nd sera testedc by Western blot (US$)d 1 1 AG FU WB 189 61 15.5 2 2 AG FU WB 189 61 15.9 3 1 ICOMB WB 189 61 17.6 4 2 ICOMB WB 189 61 17.7 5 1 EIA B WB 361 164 22.0 6 2 EIA B WB 361 164 22.8 7 1 EIA D WB 361 183 23.5 8 2 EIA D WB 361 177 23.2 9 2 EIA W WB 361 164 21.7 10 2 EIA V WB 361 168 23.6 11 2 EIA A WB 361 168 25.5 aAlgorithm 1: Western blot confirmation of initially positive ELISA or non-ELISA screening results; algorithm 2: Western blot confirmation of repeatedly reactive ELISA or non-ELISA screening results. b See Table 1 for abbreviations. c The first assays were evaluated using a panel of 189 sera with a 320/0 HIV-1 antibody prevalence. The serum panel was later extended to 361 sera with a 45% HIV-1 antibody prevalence. d The cost per test was calculated for each combination, taking into account the cost of each assay (see Table 1) and the number of sera tested using each assay. WHO Bulletin OMS. Vol 69 1991 749 G. van der Groen et al. Table 4: Comparison of the sensitivity, specificity, and cost per test for HIV antibody screening, using the alternative algorithms 3 and 4 with pairs of ELISA or non-ELISA screening assays In combination with Western blot' Test combinationb No. of No. of sera to be Sensitivityd Specificity" Cost per test Price No. Algorithm 1st 2nd 3rd sera tested confirmedc (%) (%) (USS)' ratio" 12 3 AG FU ICOMB WB 189 2 100 100 2.4 6.5 13 4 AG FU ICOMB WB 189 2 100 100 3.6 4.4 14 3 AG FU LA CB WB 103 2 100 100 3.2 5.9 15 4 AG FU LA CB WB 103 2 100 100 4.9 3.9 16 3 AG FU CHEK WB 189 5 100 100 3.1 5.0 17 4 AG FU CHEK WB 189 5 100 100 5.7 2.8 18 3 ICOMB AG FU WB 189 5 100 100 4.1 4.3 19 4 ICOMB AG FU WB 189 2 100 100 4.5 3.9 20 3 ICOMB LA CB WB 126 1 100 100 3.9 4.6 21 4 ICOMB LA CB WB 126 0 100 100 5.5 3.4 22 3 ICOMB CHEK WB 361 9 98.8 100 4.8 4.8 23 4 ICOMB CHEK WB 361 8 98.8 100 6.9 3.5 24 3 AG FU EIA B WB 189 2 100 100 2.2 7.0 25 4 AG FU EIA B WB 189 2 100 100 3.1 5.1 26 4 AG FU EIA W WB 189 2 100 100 2.6 6.1 27 3 AG FU EIA D WB 189 2 100 100 1.9 8.2 28 4 AG FU EIA D WB 189 2 100 100 2.6 6.1 29 3 EIA B EIA D WB 361 0 100 100 2.5 8.8 30 4 EIA B EIA D WB 361 0 100 100 3.5 6.5 31 3 EIA D EIA B WB 361 19 100 100 4.2 5.6 32 4 EIA D EIA B WB 361 13 100 100 4.7 4.9 33 4 EIA D EIA W WB 361 13 100 100 3.9 6.0 34 4 EIA W EIA D WB 361 0 100 100 2.3 9.4 35 4 EIA V EIA A WB 361 4 100 100 6.5 3.6 36 4 EIA A EIA V WB 361 4 100 100 7.2 3.5 Algorithms 3 and 4 are depicted in Fig. 1. b See Table 1 for abbreviations. c No. of sera with discrepant results in the 1st and 2nd assay to be confirmed by Western blot. d The % sensitivity was obtained by dividing the number of Western-blot-confirmed positive sera in the test combination by the total number of Western-blot-confirmed positive sera, and multiplying this by 100. ' The % specificity was obtained by dividing the number of Western-blot-confirmed negative sera obtained with the test combination by the total number of Western-blot-confirmed negative sera, and multiplying this by 100 ' See footnote d, Table 3. 9 The cost per serum to test n sera with the conventional algorithms 1 or 2, divided by the cost per serum to test the same number of sera with the alternative algorithms 3 or 4, respectively. blot is limited only to discrepant results or omitted entirely, it is critical that the first assay in the com- bination be the more sensitive, since this is the factor that limits the sensitivity of the combination. The specificity of the test combinations can be improved by combining assays that have a different format, i.e., a different type of assay and/or type of antigen. It is then less likely that a nonspecific effect in one assay will interfere also in the other (8). The data obtained using algorithms 2, 4, and 6 suggest that the accuracy of the test combinations does not improve by retesting initially positive sera using the same assay. Equally accurate results were obtained, at lower cost, when each serum was tested only once in each screening assay, as indicated by the findings obtained using algorithms 1, 3, and 5. It might be thought that the best test com- bination would be the one that can detect all true positive and negative sera at the lowest cost. However, the cost of a test or test combination is clearly not the only factor that plays a role in the selection; the operational characteristics of the assays, as well as the HIV antibody prevalence in the population studied, might also be decisive. Especially for small blood transfusion centres in developing countries, where only a few sera are screened, it is important that a combination of less sophisticated, non-ELISA assays can detect all true HIV positive and negative sera. Our findings on alternative less expensive con- firmatory strategies for HIV antibody screening should be followed up by further evaluations of WHO Bulletin OMS. Vol 69 1991750 Simplified and less expensive confirmatory HIV testing Table 5: Comparison of the sensitivity, specificity and cost per test for HIV antibody screening, using combinations of three different assays, without Western blot" Test combinationb No. of No. of sera to be Sensitivity Specificity Cost per test Price No. Algorithm 1st 2nd 3rd sera tested confirmedc (%)d (%)d (US$)- ratio' 37 6 AG FU ICOMB LA CB 189 2 100 100 3.1 5.1 38 5 AG FU ICOMB LA CB 189 2 100 100 2.0 7.8 39 6 AG FU ICOMB CHEK 189 2 100 100 3.1 5.1 40 5 AG FU ICOMB CHEK 189 2 100 100 2.0 7.8 41 6 AG FU LA CB ICOMB 103 2 100 100 4.1 4.7 42 5 AG FU LA CB ICOMB 103 2 100 100 2.4 7.9 43 6 AG FU EIA B ICOMB 189 2 100 100 2.7 5.9 44 5 AG FU EIA B ICOMB 189 2 100 100 1.8 8.6 45 6 AG FU EIA W ICOMB 189 2 100 100 2.2 7.2 46 6 AG FU EIA D ICOMB 189 2 100 100 2.1 7.6 47 5 AG FU EIA D ICOMB 189 2 100 100 1.4 11.1 48 6 AG FU EIA B LA CB 189 2 100 100 2.7 5.9 49 5 AG FU EIA B LA CB 189 2 100 100 1.8 8.6 50 6 AG FU EIA W LA CB 189 2 100 100 2.2 7.2 51 6 AG FU EIA D LA CB 189 2 100 100 2.1 7.6 52 5 AG FU EIA D LA CB 189 2 100 100 1.4 11.1 53 6 EIA W EIA D EIA B 361 0 100 100 2.3 9.4 54 6 EIA D EIA W EIA B 361 13 100 100 2.4 9.7 Algorithms 5 and 6 are depicted in Fig. 1. b See Table 1 for abbrevations. c No. of sera with discrepant results in the 1st and 2nd assay to be tested in a 3rd screening assay that differed from the 1st and 2nd. d. e See footnotes d, e, and f, Table 4. 'The cost per serum to test n sera with the conventional algorithms 1 or 2, divided by the cost per serum to test the same number of sera with the alternative algorithms 5 or 6, respectively. similar combinations of screening assays in resource- limited settings with different epidemiological pat- terns of HIV prevalence (9, 10). Acknowledgements We thank the manufacturers and distributors for kindly providing the screening assay kits. The contributions made by Dr P. Mortimer and Dr. H. Tamashiro are grate- fully acknowledged. Ms G. Beelaert is thanked for techni- cal support and Ms M. Mans and Ms C. Maeckelbergh are thanked for typing the manuscript. This study was supported by WHO (Technical Services Agreement, No. A20/181/68HQ/90/038231). Resume Approche simpliflee et moins couteuse de la confirmation de l'infection a VIH L'approche classique de recherche des anticorps anti-VIH, qui s'appuie sur la confirmation par Western blot de tous les cas positifs au premier depistage, est couteuse. Par exemple, pour exam- iner les 361 serums de 1'etude, pour lesquels la prevalence des anticorps anti-VIH est de 45%, le coOt par test serait de US$ 15,5-25,5. Lors de cette etude, nous avons utilise une autre approche dans laquelle les serums positifs au d6pistage initial ont et retestes par une epreuve de depistage differente, en limitant le Western blot aux serums ayant donne des resul- tats contradictoires lors des deux premiers tests. Cinq titrages immuno-enzymatiques (ELISA) et quatre titrages simples non ELISA a lecture visuelle ont ete effectues sur une serie de 361 serums, dont 220 d'origine africaine, 131 d'origine europeenne et 10 d'origine sud-americaine; la prevalence du VIH-1 etait de 45%. Les resultats obtenus avec differentes paires de titrages ont ete analyses retrospectivement. Le test Western blot HIV-1 de Du Pont a ete utilise comme test de refe- rence. Pour la plupart des associations, qui comp- taient egalement des associations de titrages non ELISA, la sensibilite et la specificite etaient de 100%, a un coOt en moyenne 6,1 fois plus faible que dans l'approche classique. De plus, on obte- nait la meme exactitude a un coOt 9,0 fois plus WHO Bulletin OMS. Vol 69 1991 751 G. van der Groen et al. faible que dans l'approche classique lorsque le Western blot etait remplace par un troisieme test de depistage different des deux premiers. Dans l'approche decrite ici, il est indispens- able que le premier test soit tres sensible, puisque cette sensibilite determine celle de I'association. Pour obtenir une bonne specificite, il est souhait- able de faire appel a deux titrages bases sur un principe different et/ou utilisant un type different d'antigenes. Le fait de retester les serums positifs par la meme methode ne semblait pas augmenter 1'exactitude obtenue pour chaque serum avec un seul test. References 1. Jenicek, M. & Cleroux, R. Theoreme de Bayes. In: Epidemiology, principles, techniques, applications. Quebec, Edisem, 1982, pp. 31-34. 2. Britz, J.A. et al. Interpreting HIV ELISA reactivity alternatives to Western blot. Journal of clinical labor- atory analysis, 2: 174-181 (1988). 3. Blomberg, J. & Klasse, P.J. Specificities and sensiti- vities of three systems for determination of anti- bodies to human immunodeficiency virus by electrophoretic immunoblotting. Journal of clinical microbiology, 26: 106-110 (1988). 4. Mortlmer, P. The fallibility of HIV Western blot. Lancet, 337: 286-287 (1991). 5. SpIelberg, F. et al. Field testing and comparative evaluation of rapid, visually read screening assays for antibody to human immunodeficiency virus. Lancet, 1: 580-584 (1989). 6. Riggin, C.H. & Thorn, R.M. Visually read HIV immu- noassays. Lancet, 1: 671-672 (1989). 7. Ferronl, P. et al. HIV antibody screening and confir- matory testing of Italian blood donors. Vox sanguinis, 55: 143-147 (1988). 8. Tosswill, J.H.C. et al. Sensitivity of the newer HIV assays. AIDS, 2: 230 (1988). 9. Spielberg, F. et al. Rapid instrument-free assays can be used in tandem for screening and confirmation of HIV infection. In: Program and Abstracts from Vth International Conference on AIDS, Montreal, 4-9 June 1989. p. 922. Abstract No. E.719. 10. Mitchell, S.W. et al. Field evaluation of alternative HIV testing strategy with a rapid immunobinding assay and an agglutination assay. Lancet, 337: 1328- 1331 (1991). 752 WHO Bulletin OMS. Vol 69 1991