Introduction
Since the introduction of the first enzyme immunoassay (EIA) tests to detect antibodies against HIV in 1985, the most widely used testing algorithm has been the screening of serum specimens with EIA and repeat testing of reactive specimens with the more specific Western blot assay [1]. This algorithm works well but has been difficult to implement in countries with limited resources because expensive, complex instruments and well-trained laboratory technicians are necessary. As the HIV pandemic has expanded, the demand has increased for accurate, less expensive alternative testing strategies. As early as 1990, alternative strategies for the detection of HIV antibodies were shown to be as sensitive and specific as the standard algorithm of EIA/Western blot [2-5]. In 1992, the World Health Organization (WHO) proposed three strategies [6] that used different numbers of HIV antibody screening tests depending on the reason for testing: seroprevalance studies, blood screening, or patient diagnosis. The choice of a particular strategy (number of tests run on each specimen) was also determined by the prevalence of HIV in the population and the presence or absence of HIV-associated symptoms. However, the WHO recommendation did not specify which combinations of tests should be used or provide data validating the sensitivity and specificity of the proposed algorithms under field conditions.
In 1991, the Honduras Ministry of Health (HMH) requested that the Centers for Disease Control and Prevention (CDC) assist them in developing and evaluating testing strategies based on their particular needs and using tests that were available in Honduras. This report describes the CDC/HMH field evaluation of testing strategies carried out during 1992-1993 at multiple sites in Honduras. This study evaluated different HIV serologic tests and test combinations under the field conditions existing in Honduras. The aim of this study was to identify strategies that would permit the completion of the HIV antibody screening as well as the diagnostic reporting in <1 h at all testing sites using simple, rapid immunoassays.
Materials and methods
WHO testing strategies
The WHO strategies (Table 1) being evaluated were those published in 1992 [6], and are outlined below.
Strategy I
All sera are tested with one EIA/rapid/simple assay. Reactive samples are considered positive for HIV antibody, and non-reactive samples are considered HIV antibody-negative.
Strategy II
All sera are tested as in strategy I. Any sample reactive in the first assay is retested with a second EIA/rapid/simple test based on a different antigen preparation, a different test principle, or both. Samples reactive in both tests are considered HIV antibody-positive. Any sample that is either negative in the first test, or positive in the first test but negative in the second, is considered HIV antibody-negative.
Strategy III
All sera are tested as in strategy I and reactive samples are retested with a different assay (strategy II). Strategy III, however, requires a third test if serum is reactive in the first and second assays. The third test should be based on a different antigen preparation or test principle than that used in the first two tests. Samples reactive in all three tests are considered HIV antibody-positive. Samples non-reactive in either of the first two tests are considered HIV antibody-negative. Samples that are positive in the first two tests but negative in the third are judged to have equivocal antibody status and require testing on follow-up samples.
Honduras study design
The study was divided into three phases.
Phase 1 was a retrospective blinded study designed to evaluate the sensitivity and specificity of seven selected rapid/simple tests against HIV antibody-positive (n = 306) and HIV antibody-negative (n = 294) specimens. Specimens were collected from pregnant women, male and female students (aged 15-19 years), and blood donors, and were submitted to the Honduras National Reference Laboratory for testing. The seven tests included in the evaluation were Retrocell (Abbott Laboratories, North Chicago, Illinois, USA), Serodia HIV-1 (Fujirebio, Tokyo, Japan), Testpack 1/2 (Abbott Laboratories), HIV-1/2 RTD (Cambridge Biotech, Galway, Ireland), Genie 1/2 (Multispot; Sanofi Pasteur Diagnostics, Marnes-la-Coquette, France), HIVCHEK 1 + 2 (Ortho Diagnostics, Roissy, France), and SUDS HIV-1 (Murex Corporation, Norcross, Georgia, USA). Aliquots of all specimens tested in each phase of the study were sent to CDC for EIA testing (Genetic Systems HIV-1 EIA, Seattle, Washington, USA) and Western blot (Cambridge Biotech, Worcester, Massachusetts, USA). The sensitivity and specificity of all tests were determined by comparison with CDC Western blot results.
Five tests (Genie, HIVCHEK, Retrocell, SUDS, and Testpack) from phase 1 were selected for phase 2 on the basis of performance, cost, and availability. Phase 2 prospectively evaluated specimens at three large regional testing centers in Honduras: the National Hospital in Tegucigalpa, the National Hospital in San Pedro Sula, and the Regional Laboratory in San Pedro Sula. One day of training was provided by CDC for laboratory personnel at each site. Specimens identified as HIV antibody-positive by the routine screening test used at each site were prospectively tested in a blinded manner using the five tests. For each HIV antibody-positive specimen the next two HIV antibody-negative samples were also tested. Specimens were forwarded to the National Reference Laboratory for EIA (Abbott HIVAB HIV-1 EIA #3A11) and Western blot analysis (Cambridge Biotech). Serostatus of these samples was determined by EIA/Western blot testing performed at CDC. HIV antibody-positive (n = 100) and HIV antibody-negative (n = 200) specimens were collected at each site for a total of 300 HIV antibody-positive and 600 antibody-negative specimens.
Phase 3 was designed to evaluate the performance of rapid testing at the small, rural hospitals and clinics. Genie, HIVCHEK and Retrocell were selected for phase 3 and training was provided to the local laboratory personnel regarding each test. Specimens were collected prospectively for HIV testing in a blinded manner at six area laboratories chosen for the known high prevalence for HIV. These laboratories were small with limited resources and processed few specimens daily for HIV-antibody testing. Samples collected at these sites were also tested using EIA and Western blot at the National Reference Laboratory and at CDC. Phase 3 lasted 14 months with the aim of identifying at least 50 HIV-positive specimens at each site. All specimens found to be indeterminate by CDC Western blot were dropped from the data analyses since the true infection status of these samples could not be ascertained (i.e., follow-up by virologic methods or later blood samples was not performed). No personal identifiers were collected. Basic demographic data including age, sex, reason for testing, risk group, and place of residence for the preceding 12 months were collected on each specimen tested in phases 2 and 3. Laboratory results and demographic data were analysed using the EpiInfo 5.0 database (CDC, Atlanta, Georgia, USA). Confidence intervals were computed assuming that for each test the number of test errors had a Poisson distribution.
Results
Evaluation of rapid HIV-antibody detection assays
The evaluation was conducted in three phases. In phase 1, the sensitivity and specificity (Table 2) of seven rapid/simple tests were evaluated retrospectively against a panel of 600 sera previously submitted to the National Reference Laboratory for confirmatory testing. The sensitivity of five assays (Genie, Retrocell, RTD, Serodia and Testpack) was 100%; HIVCHEK missed one HIV-positive specimen (sensitivity, 99.7%) and SUDS did not detect two confirmed antibody-positive specimens (sensitivity, 99.3%). Both samples undetected by SUDS were from children <6 months old. The specificities of the seven assays ranged from 92.8% to 100%.
Phase 2 (Table 2) evaluated five rapid, simple assays using 900 prospectively collected specimens collected from three regional HIV testing laboratories in Honduras. The most frequent reasons given for seeking testing were as follows: symptoms compatible with AIDS (41%), blood donation (22%), prostitution (16%), and heterosexual contact with an HIV-positive person or a high-risk individual (15%). Specimens were classified as HIV-1-positive (n = 303), HIV-1-negative (n = 582), and indeterminate (n = 15) on the basis of the CDC EIA/Western blot results. Seven sera were antibody-negative by all tests performed in Honduras but the corresponding samples sent to CDC were antibody-positive by EIA and Western blot and by repeat testing with the five simple/rapid tests used in Honduras. The most obvious reasons for this discordance were labeling errors of the samples or a sample mix-up during aliquoting before shipping to CDC. The sensitivity of the five tests as performed in Honduras (field sensitivity) was between 96.4% (SUDS) and 98.0% (Testpack).
In addition to the seven discordant samples described above, two other sera were discordant among the tests performed in Honduras. One specimen, antibody-positive by all tests at CDC and by EIA and Western blot in Honduras, was antibody-negative in Honduras by all simple/rapid tests except Testpack. A second specimen when tested at CDC was weakly positive by EIA and had only weak gp160, gp120, and gp41 by Western blot. This specimen was also missed by the rapid/simple tests in Honduras and at CDC.
Phase 3 evaluated the sensitivity and specificity of three tests (Genie, HIVCHEK, and Retrocell) against 1266 sera collected from consecutive patients at six test sites in Honduras. The most frequent reasons for being tested in this phase were blood donation (68%), symptoms compatible with AIDS (21%), and heterosexual contact with HIV-positive or high-risk persons (7%). Sera were also tested by the Abbott HIVAB HIV-1 EIA and Cambridge Western blot in Honduras and at CDC. Based on the CDC EIA/Western blot algorithm, the specimens were classified as HIV-1-positive (n = 37), HIV-1-negative (n = 1118), and indeterminate (n = 11). Two sera, which were positive by Western blot in Honduras and at CDC but were non-reactive in most of the simple/rapid tests performed in Honduras, were found to be reactive by the same simple/rapid assays when retested at CDC. The sensitivity of the three test methods as performed in Honduras (field sensitivity) was 99.3% (Table 2). The sensitivity as determined at CDC (corrected sensitivity) was 100% for all methods. Specificity (field and corrected) was >99.8% for all three assays.
WHO testing strategies
The WHO alternative strategies I and II were evaluated using the 1266 sera collected in phase 3. Strategy I was applied to the 857 blood donors. The performance of strategy I can be expressed as the sensitivity and specificity of the individual tests. All five tests (Abbott EIA, Genetic Systems EIA, Genie, HIVCHEK, and Retrocell) were used in phase 3 to test blood donors, and each detected all of the HIV-positive specimens (n = 13; sensitivity, 100%). The three rapid/simple tests (HIVCHEK, Genie, and Retrocell) were also 100% specific. Abbott EIA and Genetic Systems EIA were positive for one and two sera, respectively, that were indeterminate by Western blot. Western blot-indeterminate specimens were excluded from the analysis since their true infection status could not be determined. The specimen that was repeatedly reactive by the Abbott EIA had a gp160 band on the Western blot test and may represent early seroconversion. Two specimens were repeatedly reactive by Genetic Systems EIA but were Western blot-indeterminate with only gag-related bands (p17, p24).
In Honduras, test results are given to blood donors and therefore testing performed at the transfusion centers is diagnostic. WHO recommendations would require that strategy III (three tests) be applied to this population since all were asymptomatic. However, sensitivity, specificity, positive predictive value and negative predictive value (Table 3) all equalled 100% after testing by two tests. Further analysis using combinations of three assays was unnecessary but would also have been 100% sensitive and specific regardless of the third assay chosen.
The high-risk group (n = 406; prevalence, 30.5%) comprised all persons other than blood donors who were being tested for diagnostic purposes. WHO strategy II applies to this population because the HIV prevalence was >10% at all test sites. The sensitivity and specificity of these test combinations (Table 3) as performed in Honduras (field sensitivity) were less (98.4-100%) than when applied to the blood donor population. This decrease in sensitivity and specificity was due to the failure of all three simple/rapid tests to detect one or more samples. However, when testing was repeated at CDC the specimens that were HIV-negative in Honduras tested HIV-positive (corrected sensitivity, 100%) by all methods. The positive predictive values of the field results were 100% but the negative predictive value was between 99.3% and 100% due to one or two false-negative results.
A cost comparison was performed for different combinations of EIA, rapid tests, and simple tests in phase 3 versus the normal algorithm performed at the National Reference Laboratory (Table 4). This analysis only considered the cost of reagents which totalled US $7004 for the EIA/Western blot algorithm. The greatest savings (74.7%) were realized by using Retrocell followed by HIVCHEK. However, all of these methods produced considerable savings over the algorithms that included the expensive Western blot assay.
Discussion
In 1992, WHO introduced three HIV testing algorithms that were designed to provide rapid, accurate results equivalent to the widely accepted EIA/Western blot strategy and to reduce overall costs [6,7]. In this study, we have conducted a three-phase evaluation, which was designed to select rapid HIV antibody assay combinations that would provide economical and effective determination of HIV status and could be easily implemented at the small, rural laboratory level. The selected assay combinations were assessed according to WHO HIV testing strategies.
Evaluation of rapid assay algorithms for HIV testing have already demonstrated the effectiveness of these techniques (sensitivities >96%) with some combinations providing results comparable to the EIA/Western blot algorithm [5,8,9]. Spielberg et al. [2] have previously evaluated a combination of HIVCHEK and Serodia using 3878 sera collected from blood donors which correctly identified all HIV-positive and negative specimens. Laleman et al. [4] and Mortimer [10] have shown the application of these methods for excellent detection of HIV-positive specimens and a reduction of overall costs. However, most of these studies have relied on banked specimens which had equal numbers of HIV-positive and negative samples. These studies would not permit an evaluation of WHO testing strategies within the actual testing environment and the preselected high prevalence (50%) would be expected to yield higher positive and negative predictive values than would be observed in lower prevalence populations such as blood donors.
The sensitivities and specificities associated with the rapid tests in the first two phases of the study were comparable to those determined in other published reports [8,11-18]. All of the rapid tests had sensitivities >99% in phase 1 where the number of HIV-positive and negative specimens were approximately equal. Phase 2 incorporated the technical component of on-site testing at the regional laboratories into the algorithms, which resulted in lower field sensitivities. This decrease did not appear to be assay-related since repeat testing of the rapid tests at CDC were in agreement with the matched EIA/Western blot results and the corrected sensitivities were comparable to those determined in previous studies. The specificities for all tests in phase 2 were >99% which would make these tests useful for confirmation of initial test results. The selection of the tests for phase 3 was based on the overall performance in phases 1 and 2 as well as on assay availability, assay cost, and ease-of-use considerations.
Phase 3 of the study represents one of the first field trials of the proposed WHO testing strategies. Phase 3 was performed at the rural testing sites to assess the ability of the implemented algorithm to function under actual field conditions. This scenario presented logistical as well as technical difficulties. Since the field study was in place for more than 1 year, the laboratories had to be resupplied from the Honduran Central Laboratory in Tegucigalpa, and the records and specimen aliquots had to be transported back to the Central Laboratory. Many of the rural laboratories were not air-conditioned, suffered frequent power failures and had little equipment. The laboratory personnel who conducted the study were recent Honduran university graduates in microbiology who were beginning their first year of required public service. Despite the difficult working conditions, the sensitivities and specificities of all tests in phase 3 were excellent (>99.2 and >99.8%, respectively).
The appropriate WHO strategy was applied based on the population tested and the reason that testing was performed. In phase 3 the most common reason for testing in Honduras was to screen blood donors (68%). The remaining specimens were collected from individuals who came to the test sites because of symptoms associated with HIV infection or because they had engaged in high-risk activities and suspected that they might be HIV-infected. Only seven individuals came for testing who identified no HIV risk factor. The application of the WHO strategy II (combination of two assays) to the blood donor population (low prevalence) and high-risk groups (high prevalence) yielded excellent sensitivity and specificity (Table 3). In this study, the application of strategy III for diagnostic purposes in low prevalence populations would be unnecessary since 100% sensitivity and specificity was achieved after the application of only two tests (strategy II).
The main reason for discordance between the Honduran testing laboratories and the corresponding testing at CDC for all phases appeared to be pipetting and labeling errors during preparation of the specimens for transport to CDC. In general, specimens HIV-negative in the rapid tests in Honduras were found to be HIV-positive in the corresponding tests at CDC. Two sera in phase 3 were classified as HIV-negative by all rapid tests at the rural testing sites but were HIV-positive at the Honduran Central Laboratory and at CDC. However, since the purpose of this study was to measure performance of the tests and WHO algorithms under field conditions, the field results were reported without correction.
The use of combinations of HIV antibody assays provides choices for testing sites that have different levels of technical resources. In phase 3, the best results (100% sensitivity and specificity) were obtained using a combination of two EIA (Abbott EIA and Genetic Systems EIA). However, laboratories with fewer technical capabilities and resources can choose between combinations of simple or rapid tests, or both, with little loss of test accuracy. A combination of the simple Retrocell test followed by the rapid Genie test appears to be an excellent choice for a laboratory with few resources but a relatively high test volume. This strategy has been introduced in the rural Honduras screening laboratories, allowing testing and counseling to be completed with a single visit to a hospital or clinic. Since this study began, newer versions of the Genie (now called Multispot) and the HIVCHEK have been introduced. Evaluations of these new products indicate similar performance of these new versions compared to the tests used in phase 3 of our study [5,13].
As the need for HIV testing continues to increase, countries such as Honduras are recognizing the need to explore new testing strategies that more effectively serve their testing needs. The use of the simple, rapid tests can provide accurate results and reduce overall costs. Prior to this study, rural laboratories in Honduras tested samples using Retrocell. HIV-positive specimens were shipped to the regional laboratories for EIA testing with EIA antibody-positive specimens forwarded to the Central Laboratory in Tegucigalpa for Western blot testing. Results were returned through the same route and only became available to the patient several months after the sample was first collected. As a result many people never learned their HIV status and were never counseled. Using the algorithms of phase 3, testing can be completed during a single visit.
In addition to providing better service, WHO strategies are less expensive [3,9,10]. The savings described in Table 4 represent only the savings realized in reagent costs. Additional savings would be gained from reductions in equipment needs and personnel training and the elimination of the need to follow individuals with indeterminate Western blot results. The greatest savings (71.5%) were achieved where it was possible to use combinations of the less expensive EIA. However, substantial savings were realized even when using the relatively expensive rapid tests. These savings can be invested in expanding the amount of testing being done in these countries to offer testing to other groups at risk for HIV infection. Such testing combined with appropriate counseling can have a profound effect in reducing transmission.
The ability of rapid HIV antibody detection assays to identify sera from individuals infected with the various HIV-1 subtypes has not been extensively studied. One study indicates that these simple tests can identify sub-type sera [19] and our preliminary investigations affirm these results (unpublished data). We are currently performing a more extensive evaluation of commonly used rapid tests for their ability to detect HIV-1 sub-type sera.
In summary, advances in technology have produced highly accurate rapid and simple tests that equal EIA in sensitivity and specificity. When such tests are used in combination, as recommended by the WHO, the results are equivalent and in some ways better than the standard algorithm of EIA followed by Western blot. Developing countries should immediately consider adopting these tests and strategies to expand testing for their populations to include all blood for transfusion, and to expand counseling and testing for high-risk populations. When rapid testing algorithms are employed as one element of the HIV prevention program significant progress in slowing the spread of HIV will hopefully occur.
Acknowledgements
Our thanks to the clinical staff at the regional hospitals and rural sites throughout Honduras; J. Karon for statistical analysis; S. Phillips for technical assistance; B. Parekh and C. Pau for their manuscript review and contributions.
References
1. Centers for Disease Control: Update: serologic testing for antibody for human immunodeficiency virus. MMWR 1988, 36:833-840.
2. Spielberg F, Kabeya CM, Quinn TC, et al.: Performance and cost-effectiveness of a dual rapid assay system for screening and confirmation of human immunodeficiency virus type 1 seropositivity. J Clin Microbiol 1990, 28:303-306.
3. Van der Groen G, Van Kerckhoven I, Vercauteren G, et al.: Simplified and less expensive confirmatory HIV testing. Bull WHO 1991, 69:747-752.
4. Laleman G, Kambale M, Van Kerckhoven I, et al.: A simplified and less expensive strategy for confirming anti-HIV screening results in a diagnostic laboratory in Lumbashi, Zaire. Ann Soc Belge Med Trop 1991, 71:287-294.
5. Behets F, Bishagara K, Disasi A, et al.: Diagnosis of HIV infection with instrument-free assays as an alternative to the ELISA and Western blot testing strategy: an evaluation in Central Africa. J Acquir Immune Defic Syndr 1992, 5:878-882.
6. World Health Organization: Recommendations for the selection and use of HIV antibody tests. Wkly Epidemiol Rec 1992, 67:145-152.
7. Tamashiro H, Maskill W, Emmanuel J, Sato P, Heymann D: Reducing the cost of antibody testing. Lancet 1993, 342:87-91.
8. Brattegaard K, Kouadio J, Adom ML, Doorly R, George JR, De Cock KM: Rapid and simple screening and supplemental testing for HIV-1 and HIV-2 infections in West Africa. AIDS 1993, 7:883-885.
9. Urassa W, Bredberg-Raden U, Mbena E, et al.: Alternative confirmatory strategies in HIV-1 antibody testing. J Acquir Immune Defic Syndr 1992, 5:170-176.
10. Mortimer J: An alternative approach to confirming anti-HIV reactivity: a multi-country collaborative study. Bull WHO 1992, 70:751-756.
11. Asihene P, Kline R, Moss M, Carella A, Quinn T: Evaluation of rapid test for detection of antibody to human immunodeficiency virus type 1 and type 2. J Clin Microbiol 1994, 32:1341-1342.
12. Healy D, Dax E: An evaluation of two simple synthetic peptide based anti-HIV assays. J Virol Methods 1992, 38:305-312.
13. Kelen G, Bennecoff T, Kline R, Green G, Quinn T: Evaluation of two rapid screening assays for the detection of human immunodeficiency virus-1 infection in emergency department patients. Am J Emerg Med 1991, 9:416-420.
14. Malone J, Smith E, Sheffield J, et al.: Comparative evaluation of six rapid serological tests for HIV-1 antibody. J Acquir Immune Defic Syndr 1993, 6:115-119.
15. Moss GB, Maitha GM, Wamola IA, Ndinya-Achola JO, Plummer FA, Kreiss JK: Evaluation of a rapid membrane-based assay (HIV-CHEK) for detection of antibodies to HIV in serum samples from Nairobi. AIDS 1990, 4:351-353.
16. Myrmel H, Holm-Hansen C, Haukenes G: Evaluation of two rapid tests for the detection of HIV-1/HIV-2 antibodies. AIDS 1990, 4:1164-1166.
17. Scheffel J, Wiesner D, Kapsalis A, Traylor D, Suarez A: Retrocell HIV-1 passive hemagglutination assay for HIV-1 antibody screening. J Acquir Immune Defic Syndr 1990, 3:540-545.
18. Van Kerckhoven I, Vercauteren G, Piot P, van der Groen G: Comparative evaluation of 36 commercial assays for detecting antibodies to HIV. Bull WHO 1991, 69:753-760.
19. Engelbrecht S, de Jager G, van Rensburg E: Evaluation of commercially available assays for antibodies to HIV-1 in serum obtained from South African patients infected with HIV-1 sub-types B, C, and D. J Med Virol 1994, 44:223-228.
© Lippincott-Raven Publishers.