Marinda, Edmore T MSc, MS*; Hargrove, John PhD†; Preiser, Wolfgang DTM&H, MRCPath‡; Slabbert, Hannes MSc‡; van Zyl, Gert MBChB, FCPath, MMed‡; Levin, Jonathan PhD§; Moulton, Lawrence H PhD‖; Welte, Alex PhD†¶; Humphrey, Jean ScD‖#
Numerous attempts have been made to measure HIV-1 incidence in a timely and inexpensive manner.1-8 Although some success has been achieved, obstacles still remain to widespread use of these methods, especially in the developing world, where such methods and technologies are most urgently required.1,9 The most reliable way of measuring HIV incidence is following up HIV-uninfected but susceptible cohorts, measuring new infections over specified time periods. This, however, is subject to a number of biases, is expensive, and involves following up large numbers of people over long periods of time.9-11
Accurate and timely measures of HIV incidence would provide ways of evaluating impacts of interventions, estimating epidemiological trends in HIV studies, and providing baseline measures at the start of clinical trials.4,10,12 Several laboratory assays have been proposed for estimating HIV incidence, with the BED capture enzyme immunoassay (henceforth BED; Calypte Biomedical Corporation, Rockville, MD) being one of the most promising. The assay measures the changes in anti-HIV-1 immunoglobulin G (IgG) activity relative to total IgG and is typically used to classify individuals as HIV-1 recently infected or long term infected depending on whether a normalized absorbance reading lies below or above a preset optical density cutoff point (C).4,13 The time taken post-seroconversion for the BED's optical density to exceed C is termed the window period. The assay is easy to perform, does not need specialized equipment, and has inbuilt quality control.4,13 Several analytical methods have been proposed to infer incidence from the distribution in a cross-sectional survey of HIV-negative, recently infected, and long-term infected individuals. These methods systematize, under different assumptions, the intuitive idea that a cross-sectional survey finding a large fraction of recently infected individuals is indicative of a high incidence. Studies in the United States, Asia, and Africa have however shown that estimates produced using this assay are 2 to 3 times higher than those obtained using prospective follow-up data.4,10,11,14 This is essentially a consequence of the proportion of individuals who test as recently infected, despite being HIV-1 positive for more than twice the window period. A proportion of these are patients who remain below the target threshold of the assay for all times. Others may revert to being misclassified as recently infected when they reach end-stage AIDS or on antiretroviral treatment (ART).15 We studied the latter problems by estimating the proportion testing as recent infections by BED among patients immediately before, and while they were undergoing, ART.
MATERIALS AND METHODS
We used cryopreserved plasma samples from HIV-infected (predominantly clade C) patients referred for ART to the South African national rollout program at the Tygerberg Hospital HIV clinic, Western Cape province, South Africa.
Most of the Tygerberg patients did not have evidence of previous positive HIV test results. To qualify for ART, HIV-seropositive patients had to have a CD4 count of less than 200 cells per microliter or had to suffer from an AIDS-defining condition, both of which almost invariably imply long-standing HIV infection. We therefore consider it safe to assume that patients qualifying for inclusion in the ART rollout program, that is, fulfilling the above-cited criteria, have been infected for several years, that is, more than twice the BED assay's window period.11,14
All patients were assumed to have been ART naive before enrollment at Tygerberg. Routine laboratory diagnosis at the time included measurement of HIV-1 viral load (i.e., quantitative HIV-1 RNA in plasma using the NucliSens EasyQ HIV-1 V1.1 system; bioMérieux bv, Boxtel, the Netherlands) and CD4 count at baseline, that is, when enrolling for ART. Once enrolled on ART, patients are normally monitored for HIV-1 viral load and CD4 count at 6-month intervals. Residual EDTA plasma specimens left over after routine HIV-1 viral load testing has been performed are routinely stored in the −20°C specimen cryobank at the Division of Medical Virology, National Health Laboratory Service Tygerberg.
Inclusion criteria for our study were being on ART and at least 1 residual specimen being available in the cryobank. All specimens were collected between January 14, 2004, and August 3, 2006. Wherever available, serial samples were selected; the first one taken at enrollment for ART and later ones from the same patient at subsequent clinic visits. For follow-up samples, the minimum duration of HIV infection is determined by the time since the baseline sample. Ideally, we tried to select samples at 12- and 24-month follow-up intervals. However, due to suboptimal adherence to follow-up schedules, many follow-up samples were obtained at intermittent intervals, that is, 6 and 18 months.
All available cryopreserved plasma samples from each patient were tested by the BED assay. Basic demographic data and HIV-1 viral load and CD4 count results were retrieved from the laboratory database. Viral load results were log10 transformed.
Cryopreserved plasma samples were thawed and then immediately tested by the Calypte HIV-1 BED Incidence EIA (IgG-capture HIV-EIA; BED-CEIA) strictly following the manufacturer's instructions (cat. no. 98003; Calypte Biomedical Corporation, Lake Oswego, OR).
BED measures levels of HIV-1-specific IgG antibodies as a proportion of total serum IgG antibodies; it detects recent seroconversion through low levels of HIV-1-specific IgG. Specimens with normalized optical density (ODn = ODspecimen/ODcalibrator) of less than 0.8 are classified as recent infections and those with values of 0.8 or more as long-term infections. According to the test kit insert, all samples with ODn values of 1.2 or less must be retested in triplicate (confirmatory testing); the median ODn value of the 3 repeats is then used to classify the sample as recent (if ODn <0.8) or long term (if ODn ≥0.8).
We retested the first 81 samples with initial ODn values ≤1.2. Of these, 47 had ODn values of ≤0.8 and 34 had ODn values >0.8 and ≤1.2. Retesting did not change classification in 70 of the 81 samples (86.4%). Four samples were classified by retesting as long term despite an initial ODn ≤0.8 and 7 as recent despite an initial ODn >0.8. The observed high level of concordance between the initial and the confirmatory test results is in line with the results of a previous evaluation by Dobbs et al.13 We therefore decided to abandon retesting for the remainder of the study as it resulted in negligible difference on a population basis.
Data were entered into Microsoft Excel and exported to STATA (STATACORP version 10; College Station, TX) for analysis. Ninety-five percent confidence intervals (CIs) are reported for all estimates. We used 5% significance levels for all comparisons. The rate of false recent was estimated as a percentage with 95% binomial CI. We used 2-tailed t tests or nonparametric equivalents to compare BED recent versus BED long term infected for continuous variables and χ2 or Fisher exact tests for categorical variables.
To investigate factors associated with false-recent BED status before commencement of ART, multiple logistic regression models were used. To investigate trends in BED ODn values over the treatment period, we fitted random effect multiple logistic regression models to account for repeated readings for individuals over time.
Ethics and Confidentiality
Ethics approval was granted by the Committee for Human Research of the Faculty of Health Sciences, Stellenbosch University (project number N07/06/137) and the University of the Witwatersrand's Postgraduate Ethics Committee on Human research (clearance number M080224). Confidentiality was maintained as per standard laboratory protocol.
In total, 1061 samples collected from 505 patients were analyzed with the BED assay. Not all patients had BED readings at the first visit. There were 430 baseline BED ODn readings and 433, 127, 31, and 9 readings at 0.5, 1, 1.5, and 2 years of follow-up, respectively. Of the 505 patients in the study (median age 33 years), 132 (31%) were male (median age 37 years, range 22-79 years) and 298 (69%) were female (median age 32 years, range 18-58 years).
Proportion Testing as Recent HIV-1 Infections Just Before Commencing ART
Univariate analysis provided an overall BED false-recent rate of 11.2% (95% CI: 8.3 to 14.5%), independent of patient gender and age (P = 0.67 and 0.16, respectively). Among patients with CD4 counts below 50 cells per microliter, 18 of 89 (20.2%, 95% CI: 12.4 to 30.1) tested recent by BED compared with 30 of 341 (8.8%, 95% CI: 6.0 to 12.3) of those with higher CD4 counts (P = 0.002 for the difference).
In multivariate analysis with independent variables log10 viral load (stratified above and below 4) and CD4 count (stratified above and below 50 counts/μL), patients with low CD4 and low viral loads were 2.9 (95% CI: 1.5 to 5.6) (P = 0.01) and 3.8 (95% CI: 1.39 to 11.11) (P = 0.001) times more likely to test BED false recent than those with high CD4 and high viral loads, respectively.
Performance of the BED Assay When ART Commences
In univariate analysis, the percentage of patients testing recent by BED increased from 11.2% pre-ART to 17%, 25%, 38%, and 56% at 0.5, 1, 1.5, and 2 years, respectively, after treatment commenced (Fig. 1).
After adjusting for CD4 count, stratified as above, the odds of testing recent by BED, compared with the baseline visit, were 2.1 (95% CI: 1.6 to 2.9), 3.5 (95% CI: 2.2 to 5.7), 6.8 (95% CI: 3.0 to 15.7), and 12.2 (95% CI: 3.2 to 46.1) at visits 0.5, 1, 1.5, and 2 years, respectively, after ART initiation. The wide CIs for estimated odds ratios are a consequence of sparse data for later time-points. Low CD4 count (below 50), but not viral load, remained predictive of testing recent in a random effect multiple logistic regression model.
Previously published analysis of data from the AIDSVAX B/B vaccine trial and Zimbabwe Vitamin-a for Mother and Babies Project (ZVITAMBO) studies showed that, if the BED were to be at all useful for estimating HIV incidence, it would be necessary to adjust for a proportion of patients infected with HIV-1 for long periods of time (much in excess of 1 year) who nonetheless continued to test as recent infections by BED.11,14 In these 2 studies, there were, however, few cases with CD4 counts below 50 cells per microliter and nobody was on ART.
The present study shows that the proportion testing recent increased significantly among pre-ART clients with CD4 counts below 50 cells per microliter, and even more sharply with time spent on ART, its value being significantly raised only after 6 months on ART. This poses serious problems because BED incidence estimates are very sensitive to changes in the proportion of long-term false-recent cases.14 Moreover, as previously observed, patients with a log10 viral load of 4 or less than 4 (i.e., <10,000 viral copies/μL) were more likely to test false recent than those with higher viral loads.16 If the BED method is to be used for the accurate determination of HIV incidence from cross-sectional surveys, we must be able to identify, and classify as long-term infections, cases who are on ART and/or have extremely low CD4 counts. Failure to do so will lead to unknown, but probably very large, errors in incidence estimates.
It is unclear why differences exist, between studies in the same region and among people infected with the same clade of HIV-1, in the proportion of long-term HIV-1 infections who continue to test recent by BED. Suggested factors influencing BED recency threshold include viral suppression, immune reconstitution on ART, concurrent infections, and late-stage HIV disease. Participants in the ZVITAMBO study were generally free of life-threatening conditions at recruitment (only 15/4495 HIV-1-positive individuals presented with CD4 count of less than 50 cells/μL), and a substantial number of patients with AIDS-defining conditions were thus excluded.14 In the Tygerberg study, conversely, patients either had to have a CD4 counts below 200 cells per microliter and/or an AIDS-defining condition to qualify for ART. Depending on the stage of the HIV epidemic in the population, patients at the AIDS stage are expected to be relatively fewer than those at other stages. The false-recent proportion of 11.2% for all patients, or 8.8% for patients with CD4 counts of 50 cells per microliter or above, in the Tygerberg study probably overestimates the true false-recent ratio among long-term HIV-1 infections in the population. (This problem would have been exacerbated if our unverifiable assumption that all patients recruited into this treatment center were ART naive was violated.) Conversely, because ZVITAMBO excluded AIDS patients, it is possible that the 5.2% figure from that study underestimated the true proportion in the female population.
By contrast, the Africa Center study reports a false-recent proportion of 1.7% among clients known to be infected with HIV-1 for more than 1 year, much lower than other similar clade C estimates.17 The study is one of the few that covers the whole spectrum of HIV-infected individuals: both sexes, different age-groups, and individuals at different HIV disease stage. Further studies are clearly indicated to discover the basis of the variability in the proportion of false-recent infections between populations. Such studies must be based on appropriate sample sizes and carried out in a variety of situations to assess variations with geography, HIV subtype, ART treatment, and other demographic factors. All factors that potentially influence BED ODn threshold levels, such as CD4 count, viral load, ART, and coinfections, should be investigated and appropriately used to aid in the estimation of robust HIV incidence estimates.
Algorithms that combine results of more than 1 test, either sequentially or in parallel, thus improving predictive values of the assay, could be used to improve incidence estimates.1 But such protocols and the use of clinical information to improve assay performance are unlikely to be feasible in resource-constrained setting. The essential problem in using the BED method to estimate HIV incidence thus continues to lie in the lack of unequivocal identification of persons with long-term HIV infections. If some alternative test can be developed to achieve this end, it could be used with or without other assays to render the statistical corrections unnecessary, and this should be a primary research aim.11
The authors wish to thank Stephen Korsman and Lynette Smit for their valuable contributions to sample selection and testing and Marina La Grange and colleagues of the Tygerberg ARV Clinic for contributing patient data from the President's Emergency Plan for AIDS Relief (PEPFAR)-funded database.
1. Gupta SB, Murphy G, Koenig E, et al. Comparison of methods to detect recent HIV type 1 infection in cross-sectionally collected specimens from a cohort of female sex workers in the Dominican Republic. AIDS Res Hum Retroviruses. 2007;23:1475-1480.
2. Parekh BS, Pau CP, Kennedy MS, et al. Assessment of antibody assays for identifying and distinguishing recent from long-term HIV type 1 infection. AIDS Res Hum Retroviruses. 2001;17:137-146.
3. Janssen RS, Satten GA, Stramer SL, et al. New testing strategy to detect early HIV-1 infection for use in incidence estimates and for clinical and prevention purposes. JAMA. 1998;280:42-48.
4. Parekh BS, Kennedy MS, Dobbs T, et al. Quantitative detection of increasing HIV type 1 antibodies after seroconversion: a simple assay for detecting recent HIV infection and estimating incidence. AIDS Res Hum Retroviruses. 2002;18:295-307.
5. Murphy G, Parry JV. Assays for the detection of recent infections with human immunodeficiency virus type 1. Euro Surveill. 2008;13:4-10.
6. Rawal BD, Degula A, Lebedeva L, et al. Development of a new less-sensitive enzyme immunoassay for detection of early HIV-1 infection. J Acquir Immune Defic Syndr. 2003;33:349-355.
7. Wilson KM, Johnson EI, Croom HA, et al. Incidence immunoassay for distinguishing recent from established HIV-1 infection in therapy-naive populations. AIDS. 2004;18:2253-2259.
8. Constantine NT, Sill AM, Jack N, et al. Improved classification of recent HIV-1 infection by employing a two-stage sensitive/less-sensitive test strategy. J Acquir Immune Defic Syndr. 2003;32:94-103.
9. Parekh BS, McDougal JS. Application of laboratory methods for estimation of HIV-1 incidence. Indian J Med Res. 2005;121:510-518.
10. Sakarovitch C, Rouet F, Murphy G, et al. Do tests devised to detect recent HIV-1 infection provide reliable estimates of incidence in Africa? J Acquir Immune Defic Syndr. 2007;45:115-122.
11. McDougal JS, Parekh BS, Peterson ML, et al. Comparison of HIV type 1 incidence observed during longitudinal follow-up with incidence estimated by cross-sectional analysis using the BED capture enzyme immunoassay. AIDS Res Hum Retroviruses. 2006;22:945-952.
12. Saphonn V, Parekh BS, Dobbs T, et al. Trends of HIV-1 seroincidence among HIV-1 sentinel surveillance groups in Cambodia, 1999-2002. J Acquir Immune Defic Syndr. 2005;39:587-592.
13. Dobbs T, Kennedy S, Pau CP, et al. Performance characteristics of the immunoglobulin G-capture BED-enzyme immunoassay, an assay to detect recent human immunodeficiency virus type 1 seroconversion. J Clin Microbiol. 2004;42:2623-2628.
14. Hargrove JW, Humphrey JH, Mutasa K, et al. Improved HIV-1 incidence estimates using the BED capture enzyme immunoassay. AIDS. 2008;22:511-518.
15. Corporation: CB. Calypte HIV-1 BED Incidence EIA. Enzyme Immunoassay for Population Estimation of HIV-1 Incidence. Portland, OR: Caplypte Biomedical Corporation; 2004.
16. Karita E, Price M, Hunter E, et al. Investigating the utility of the HIV-1 BED capture enzyme immunoassay using cross-sectional and longitudinal seroconverter specimens from Africa. AIDS. 2007;21:403-408.
17. Barninghausen T, Wallrauch C, Welte A, et al. HIV incidence in rural South Africa: comparison of estimates from longitudinal surveillance and cross-sectional cBED assay testing. PLoS One. 2008;3:e3640.
© 2010 Lippincott Williams & Wilkins, Inc.