An alternative approach to confirming anti-HIV reactivity: a multi-country collaborative study.

An alternative approach to confirming anti-HIV reactivity: a multi-country collaborative study* J. Mortimer1 The confirmation of positive screening assay reactions for antibodies to human immunodeficiency virus type 1 (anti-H/V-1) by Western blot is expensive and often gives indeterminate results. We therefore carried out a collaborative study to investigate the confirmation of screening assay reactions using a second screening assay. For this purpose, seven laboratories prospectively tested sequential specimens, using at least one additional screening assay, until about 50 confirmed anti-HIV-1-positive specimens had been identified in each test centre. The reactions of 16 assays were analysed in pairs (assay A and assay B), using assay B on specimens reactive in assay A: A+/B+ reactions were considered positive and A-, negative anti-H/V results. These outcomes were compared with those obtained using confirma- tory Western blot. In all, 7950 specimens were tested, and 359 were reported as positive by the labora- tories. Within the test centres, eight screening assay pairings gave rise to no false-positive or false- negative results, and these combinations were at least as accurate as a single screening assay fol- lowed by Westem blot. From 6.3% to 8.3% of the Western blot results were indeterminate. The number of specimens examined was too small to justify recommending for general use named pairs of screening assays; the choice of these would, in any case, depend on local conditions. However, individual laboratory managers may wish to investigate the large potential savings to be made by confirming HIV infection using a second screening assay on initially reactive specimens. If the more sensitive screening assay is used first, the sensitivity of this approach may be improved by further investigation of specimens that react as A+B -. Introduction Serological tests for antibodies to human immuno- deficiency virus (anti-HIV) are now widely used to screen blood donors, to confirm clinical suspicion of infection, and for surveillance purposes. The sensi- tivity and specificity of commercially available enzyme-linked immunosorbent assay (ELISA) kits and rapid/simple, instrument-free assays have been well documented (J).a, b In order to compensate for The following individuals collaborated in the study: J. Parry & P. Mortimer (PHLS Virus Reference Laboratory, London, England); F. Brun-Vezinet & F. Simon (Claude-Bernard Hospital, Paris, France); J. Eberle (Max von Pettenkofer Institute, Munich, Germany); I. Gust & S. Nicholson (Fairfield Hospital, Melbourne, Australia); G. van der Groen & G. Vercautern (Institute of Tropi- cal Medicine, Antwerp, Belgium); G. Biberfeld & R. Thorstensson (National Bacteriological Laboratory, Stockholm, Sweden); and M. O'Shaughnessy (Federal Centre for AIDS, Ottawa, Canada). 1 Principal Scientist, PHLS Communicable Disease Surveillance Centre, 61 Colindale Avenue, London NW9 5EQ, England. Requests for reprints should be sent to Mrs Mortimer at this address. a Operational characteristics of commercially available assays to determine antibodies to HIV-1; report 1. Unpublished WHO document GPA/BMR/89.4. b Operational characteristics of commercially available assays to determine antibodies to HIV- 1 and/or HIV-2 in human sera; report 2. Unpublished WHO document GPA/BMR/90.1. Reprint No. 5340 any lack of specificity in the screening assays, and because of the serious prognostic implications of positive results, a two-tier system of testing has been developed in which sera that are reactive on initial testing are retested with a second (supplemental) test.c Although Western blot is the test commonly used for this purpose, it is expensive, is not easy to read or standardize, and often gives indeterminate results. Its cost, use, and interpretation pose difficul- ties for some laboratories, and there is therefore a need for alternative confirmatory procedures (2, 3).c The present study compared the accuracy of using a second screening assay (either ELISA or a simpler test) with that of using Westem blot to confirm a positive reaction in an initial screening assay. The study was carried out with the collabora- tion of seven laboratories in Europe, Australia, and Canada. Altogether, results from ordered pairs of 16 screening assays (68 combinations) were studied to determine whether a second screening assay would be an adequate substitute for a Western blot to confirm positive screening reactions. The outcomes of using either another screening assay or Western blot to confirm reactivity were compared. c Report of the WHO Meeting on Criteria for the Evaluation and Standardization of Diagnostic Tests for the Detection of H/V Antibody, Stockholm, 7-8 December 1987. Unpublished docu- ment WHO/GPA/BMR/88.1. Bulletin of the World Health Organization, 70 (6): 751-756 (1992) ( World Health Organization 1992 751 J. Mortimer Method In January 1989 seven participating laboratories were invited to test prospectively consecutive serum specimens referred to them until 50 confirmed anti- HIV-positive specimens had been identified in each. Most of the sera had not previously been screened, but in some centres some of the sera had already been tested elsewhere. All the specimens were tested using the assays routinely employed in the centre plus additional assays suggested by the WHO Global Programme on AIDS. Specimens that gave a positive reaction in any assay were tested using Western blot. The centres were asked to record all the results on every serum and to report the antibody status according to their existing test procedure, i.e., ignor- ing the results of the additional assays. This pro- cedure generally included the use of one or more screening assays and Wesfern blot; the centre's result (CR) could be positive; negative, or undecided. The analysis of the results wag based on the following: The assumption that the CR recorded by the par- ticipating centre was correct. The application of two sets of criteria for scoring the Western blot reactions reported by each centre, either the WHO criteria for positivity (the presence of two of the envelope bands (gp41, gpl20 or gpl6O) or one of these envelope bands plus a band representing antibody to a gag or a pol gene product)d or those proposed by the Cen- ters for Disease Control (CDC) (presence of two of the three bands representing antibody against p24, gp4l or gpI20/160). In both schemes any Western blot reaction that did not meet the criteria was scored as indeterminate. If no bands (CDC criteria) or no virus-specific bands (WHO criteria) were reported, the specimen was scored negative. The simplification that an "indeterminate" Western blot result or an "undecided" CR was negative, and that equivocal reactions in rapid tests were positive. This was necessary to permit a manageable analysis and presentation of the data. Since the study centres were all in areas with relatively low incidences of HIV infection, the great majority of indeterminate Western blot results and undecided CR specimens were prob- ably anti-HIV negative. Equivocal reactions in rapid tests were regarded as positive in order to test the alternative approach to confirmatory testing as rigorously as possible. The biases that may have been introduced as a result of these simplifications should be borne in mind. d See footnote c, p. 751. The performance of pairs of assays was assessed on the basis of two simple algorithms. The first was as follows: if the initial screening test (A) was :unreactive, the final result was negative; if A was reactive, a second screening assay (B) was applied; if B was unreactive the final result was also negative; if A as well as B were reactive, the final result was positive. Thus: * A- = negative * A+B- = negative * A+B+ = positive The second algorithm differed from the first in that A+B- specimens were fully investigated, reduc- ing the possibility of a false-negative result. Results obtained upon repeat testing using the same assay were disregarded. Results Seven collaborating laboratories provided the results of anti-HIV testing using their current routine, additional assays, and Western blot (Table 1). A total of 7950 specimens were included in the analysis, of which 7516 were primary, while 434 (5.5%) were referred from another laboratory. Three other speci- mens were excluded: two anti-HIV-2 positive sera, and one sample of cerebrospinal fluid. For one of the collaborating laboratories, as an example, Table 2 shows the outcome of applying the two algorithms to pairs of anti-HIV assays, and to a single screening assay followed by Westem blot (scored by both criteria), as well as the results for single screening assays. The number of tests needed to arrive at an algorithm result varied with the assay pairings and with the order of a given pair. The accu- racy of the algorithm results was assessed using the CR as the "gold standard". The proportion of false- negatives is shown as incorrectly negative: total confirmed positive results and the false positives as incorrectly positive: total confirmed negative results. In some instances, the sensitivity of the second algorithm (further testing of A+B- specimens) was greater than that of the first, and the two false- negative proportions are shown separately. The second and fifth columns, taken together, show the number of specimens with outcome A+B-, while the second column shows the number of these which by algorithm 1 (i.e., without full investigation) would have given a false-negative result. There were some discrepancies between the CR and Western blot results (Table 3). Of the 557 sera that were examined by Western blot 503 gave concordant results with the CR using the WHO criteria, while for 495 there was agreement using 752 WHO Bulletin OMS. Vol 70 1992 An alternative approach to confirming anti-HIV reactivity Table 1: Assays used and specimens tested in the study No. of specimens: Test centrea Assays used Western blot Primary Referred Total CBV Existing: LAV Blot 413 0 413 Elavia HIV 1 Pasteur (Elavl) Wellcozyme HIV 1 recom. (Wellc1) Additional: Ortho HIV 1 Elisa (Ortho) Serodia (SEROD) HIV Chek (HIVCK) FCA Existing: In house 0 49 49 Genetic Systems (GenSys) Additional: Cambridge BioScience (CamBS) FH Existing: Biorad 5025 47 5072 Abbott HIV 1 recom. (Abblr) Additional: Welicozyme HIV 1 recom. (Welilc) ITM Existing: Du Pont 336 17 353 Abbott recomb HIV 1+2 (Abbl+2) Additional: HIV Chek (HIVCK) lmmunocomb (IMMCB) MVPI Existing: In house 231 77 308 Enzygnost HIV 1+2 (Enzl+2) Additional: Enzygnost HIV 1 (Enzl) Recombigen (RECOMB) NBL Existing: Du Pont 1202 130 1332 Enzygnost HIV 1+2 (Enzl+2) Wellcozyme HIV 1 recom. (Wellcl) Additional: Abbott HIV 1 recom. (Abblr) Serodia (SEROD) VRL Existing: Du Pont 309 114 423 Abbott HIV 1 recom. (Abblr) Additional: Du Pont HIV 1 recom. (Dupir) Vironostika HIV Uni-Form (Vuf) Wellcozyme HIV 1 (Wellcl) Serodia (SEROD) Total 7516 434 7950 a CBV = Claude-Bernard Virologie, Paris, France; FCA = Federal Centre for AIDS, Ottawa, Canada; FH = Fairfield Hospital, Mel- bourne, Australia; ITM = Institute of Tropical Medicine, Antwerp, Belgium; MVPI = Max von Pettenkofer Institute, Munich, Germany; NBL = National Bacteriological Laboratory, Stockholm, Sweden; VRL = PHLS Virus Reference Laboratory, London, England. the CDC criteria. No CR-positive sera were negative in the Western blot. On the other hand, four CR-negative sera were positive in the Western blot by WHO criteria and three by CDC criteria. There were 21 undecided CR results report- ed in the study, compared with 35 indeterminate Western blot results by WHO criteria and 46 by CDC criteria. Equivocal results in the screening assay were rare. Discussion In the study the value of using a second screening assay to confirm the presence of anti-HIV was compared with that of using Western blot, taking the test centres own final result (CR) as the "gold standard". The results obtained show that many pairs of screening assays performed at least as well as a single screening assay followed by Western blot. Moreover, in some cases where ordered pairs of screening assays did give rise to false-positive results, the ELISA absorbance readings were so close to the cut-off value that careful interpretation of the findings of the screening assay would have led to repeat testing or use of a third screening assay. The false-positive results were mostly based on weak reactions; the value of distinguishing between strong WHO Bulletin OMS. Vol 70 1992 753 J. Mortimer Table 2: Comparison between the centre's result (CR) and assay pairings, single assays + Western blot (WB) using WHO and CDC criteria, and single assays: an example using the data from the Institute of Tropical Medicine, Ant- werp, Beigiuma Centre's resultsb Positive Negative False: A+B+ A+B- A- A+B+ A+B- A- Negative (1)c Negative (2)c Positive Assay pairings Abb1+2 - HIVCK 47 2 0 0 11 293 2/49 0/49 0/304 HIVCK -+Abb1+2 47 0 2 0 1 303 2/49 2/49 0/304 Abb1+2-IMMCB 49 0 0 0 11 293 0/49 0/49 0/304 IMMCB - Abb1+2 49 0 0 0 0 304 0/49 0/49 0/304 HIVCK - IMMCB 47 0 2 0 1 303 2/49 2/49 0/304 IMMCB - HIVCK 47 2 0 0 0 304 2/49 0/49 0/304 Single assays + WB Abb1 +2 - WB (WHO) 49 0 0 0 11 293 0/49 0/49 0/304 HIVCK - WB (WHO) 47 0 2 0 1 303 2/49 2/49 0/304 IMMCB - WB (WHO) 49 0 0 0 0 304 0/49 0/49 0/304 Abb1 +2 - WB (CDC) 48 1 0 0 11 293 1/49 0/49 0/304 HIVCK - WB (CDC) 46 1 2 0 1 303 3/49 2/49 0/304 IMMCB - WB (WHO) 48 1 0 0 0 304 1/49 0/49 0/304 Single assays Abbl +2 0/49 11/304 HIVCK 2/49 1/304 IMMCB 0/49 0/304 a Data from the six other centres are available on request from the author. b One specimen that was WB positive by WHO criteria was indeterminate (and thus classed as negative for this table) by CDC criteria. c By algorithms 1 and 2, respectively. and weak ELISA reactions and modifying reports accordingly has been emphasized previously (5). In the algorithms used in the study, two interpre- tations of A+B- reactions were explored (either regarding them as negative or investigating them further). The first approach led to a few false- negative results with some pairs of screening assays. Thus, whether in practice it might be justifiable to adopt this approach would depend on the screening assays chosen and on the prevalence of anti-HIV in the population to be tested. However, the algorithm could be supplemented by follow-up testing after an interval of a few weeks to determine whether indi- viduals with AMB- reactivity had changed to A+B+; if no such change occurred individual would be regarded as anti-HIV negative. The second ap- proach, in which A+B- results were investigated fur- ther, was sometimes more accurate, but consumed more laboratory resources; nevertheless, its use would still give considerable cost savings because all the A+B+ specimens would have been confirm- ed without having to resort to Western blot. Several observations should be made about local conditions in the centres which collaborated in this study and about the nature of the assays used. First, when specimens had been selected for referral by another laboratory on the basis of a reaction in a screening assay that was being investigated in a col- laborating centre, a false-positive outcome from an algorithm was more likely. Second, it was reported from one centre that a rapid assay had performed better than usual in the course of the study: whatever assays are used for HIV testing, users must be alert to batch-to-batch variations that affect sensitivity and/or specificity. Third, a few pairs of assays, e.g., Abbott and Wellcozyme, appeared to be susceptible to the same nonspecific effects, perhaps because they used a very similar recombinant antigen or had the same format. These considerations, and the fact that no assay pairings were used on enough specimens to establish the accuracy of any of the combinations, meant that there was insufficient information to allow any particular pairs to be recommended. Instead, we suggest that readers select from our WHO Bulletin OMS. Vol 70 1992754 An alternative approach to confirming antl-HIV reactivity Table 3: Concordance between centres' final anti-HIV and Western blot results, interpreted according to WHO and CDC criteria, for 557 sera Western blot Centres' final results result No. positive No. undecided No. negative Positive WHOa 359 3 4 CDCb 351 1 3 Indeterminate WHOa 0 6 29 CDCb 8 8 30 Negative WHO" 0 12 144 CDCb 0 12 144 a WHO criteria, see footnote c, p. 751. b CDC criteria, see ref. 4. Table 4: Screening assay pairings that In at least one study centre did not lead to false results (the order of their use did not alter the outcome) Pairing No. of specimens examined Abbir *- Enz1+2 1332 Abbir <- SEROD 1332 Abb1r " Wellcl 5072 Abbl+2 *- IMMCB 353 CamBS 4 GenSys 49 Elavl e SEROD 1332 Enzl+2 - WelIcl 1332 SEROD " Wellcl 1332 results those assay pairings that performed well in this study (Table 4) and use them with specimens of their own whose serostatus has previously been established. If these anti-HIV screening assays give reliable results on a sufficiently large sample of specimens of known status, Western blot confirma- tion can be omitted, making the testing cheaper and, with machine-read ELISA, more objective. Further savings might be achieved by considering the order in which the selected pairs of assays are used. With the first algorithm (see Table 2) the outcome of testing using a pair of assays was the same, irrespective of the order in which the assays were used. However, the total number of tests re- quired to achieve the final result depends on the spe- cificity of the first assay, which can vary considerably. The two simple algorithms used did not incorpo- rate repeat testing (which is often specified in more elaborate algorithms for investigating anti-HIV status) and such testing was not generally carried out. Furthermore, the analysis was concerned only with confirming positivity. In circumstances where the prevalence of HIV is high it may be just as important to confirm negative initial screening reac- tions by using a second screening assay. For almost all pairs of assays, a double negative outcome would have provided a correct negative result, at least for the specimens examined in the study (results not shown). A further potential advantage of using com- binations of screening assays to test and confirm is that fewer specimens will be indeterminate. In the study, 6.3% (by WHO criteria) and 8.3% (by CDC criteria) of specimens examined by Westem blot were indeterminate: 83% and 65%, respectively, of such indeterminate specimens were reported by the study centres as anti-HIV negative. It is a disadvan- tage of the Western blot technique that in many cases it fails to discriminate clearly between seropositivity and seronegativity. There should be broader investigation of cheap alternative confirmatory strategies for anti-HIV status. For example, very few of the specimens ex- amined in the study were collected in Africa; in other studies, African specimens have often been reported to give false-positive ELISA results. A fur- ther difficulty that now confronts confirmatory test- ing strategies occurs with specimens from popula- tions in which there might be HIV-2 infections. Although combined anti-HIV-1/2 screening assays have become widely available, only two (Abbott and Enzygnost) were included in the present study. The use of a combined screening assay creates the need and HIV-2 but can avoid using two Western blots, which would be inordinately expensive. It may not be easy to devise a simple test algorithm that will identify HIV-1- and HIV-2- positive specimens, dis- tinguishing them from falsely reactive specimens, as well as from each other. Though it has not been possible to recommend specific alternatives to the use of Western blot for confirming positive anti-HIV reactions, our results show that pairs of screening assays can be just as accurate and give rise to fewer indeterminate (equivocal) results. Laboratory managers who are interested in devoting to other needs the funds which they spend on Western blot strips may there- fore want to investigate the savings to be gained by confirming HIV infection through the use of further screening assays on initially reactive specimens. Acknowledgements The WHO Global Programme on AIDS is thanked for providing the kits used in the study, and Dr H. Tamashiro's help and encouragement in preparing this article are gratefully acknowledged. WHO Bulletin OMS. Vol 70 1992 755 J. Mortimer Resume Une autre faqon de confirmer la s6ropositivite pour le VIH: resultats d'une 6tude multicentrique Sept centres collaborateurs de l'OMS ont appliqu6 de facon prospective leur test habituel de depista- ge des anticorps dirig6s contre le virus de l'immu- nod6ficience humaine (VIH), et un ou plusieurs tests supplementaires, a des 6chantillons de s6rum, jusqu'a ce qu'ils aient obtenu environ 50 resultats positifs selon leur protocole de confirma- tion habituel. Au nombre des tests supplemen- taires figuraient diverses epreuves immuno- enzymatiques (ELISA) et des tests rapides. Au total, 16 tests de depistage ont 6te 6valu6s sur 7950 prelevements dont 359 se sont rev6els posi- tifs apres confirmation. Tous les echantillons qui ont donne une r6action positive dans un ou plusieurs tests de d6pistage ont 6t6 soumis a un Western blot dont la r6action a 6te interpr6t6e selon les criteres proposes par les Centers for Disease Control d'une part et par l'OMS d'autre part. Dans chaque centre, on a appliqu6 soit un deuxieme test de depistage, soit la m6thode du Western blot pour confirmer la r6action au premier test de d6pistage. Les r6sultats ont ensuite et6 compar6s au r6sultat final de la procedure habi- tuelle de confirmation, consid6ree comme 6talon de r6f6rence. Les r6sultats obtenus en utilisant successivement deux tests de d6pistage (A et B) ont 6te analys6s comme suit: A- a et consid6r6 comme n6gatif; A+B+ comme positif et A+B- soit comme un resultat n6gatif (algorithme 1), soit comme un r6sultat n6cessitant un nouveau test (algorithme 2). L'application de ces algorithmes a montre que l'utilisation de deux tests de d6pistage donnait dans bien des cas des r6sultats au moins aussi pr6cis que ceux de la confirmation classique d'un test de d6pistage unique par Western blot. Huit paires de tests de d6pistage n'ont donn6 aucun faux r6sultat et les r6sultats douteux ont 6t6 rares. Par contre, la confirmation par Western blot a donn6 des resultats indetermines dans 6,3% a 8,3% des cas. Depuis la fin de cette etude, d'autres tests de d6pistage, dont plusieurs permettent de d6tecter a la fois les anticorps anti-VIH 2 et anti-VIH 1, sont apparus sur le marche. L'application de la tech- nique du Western blot est tres coOteuse, surtout lorsqu'elle est utilisee pour confirmer les deux types d'infection, sans etre n6cessairement plus pr6cise qu'une combinaison de deux ou plusieurs tests de depistage. Nous pensons que les labora- toires de diagnostic devraient 6valuer des algo- rithmes moins coOteux de confirmation de la s6ro- positivit6 pour le VIH et les adopter s'ils se r6velent pratiques. References 1. 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). 2. Spielberg, F. et al. Performance and cost- effectiveness of a dual rapid assay system for screening and confirmation of human immuno- deficiency virus type 1 seropositivity. Journal of clinical microbiology, 28: 303-306 (1990). 3. Lepine, D.G. et al. Evaluation of a human immuno- deficiency virus test algorithm utilizing a recombi- nant protein enzyme immunoassay. Journal of clini- cal microbiology, 28: 1169-1171 (1990). 4. Centers for Disease Control. Interpretation and use of the Western blot assay for serodiagnosis of human immunodeficiency virus type 1 infections. Morbidity and mortality weekly report, 37 (suppl.): 1-7 (1989). 5. Crofts, N. et al. Evaluation of enzyme-linked immunosorbent assays: a method of data analysis. Journal of virological methods, 22: 51-59 (1988). 6. van der Groen, G. et al. Simplified and less expen- sive confirmatory HIV testing. Bulletin of the World Health Organization, 69: 747-752 (1991). 756 WHO Bulletin OMS. Vol 70 1992