Reconciling HIV incidence results from two assays employed in the serological testing algorithm for recent HIV seroconversion (STARHS)

Abstract

The Abbott HIVAB and Vironostika HIV-1 Microelisa assays have both been validated for use in the serological testing algorithm for recent HIV seroconversion (STARHS). This ability to identify recently-acquired infection provides valuable insight into the epidemic. The availability of each assay during different periods led to longitudinal studies of annual HIV incidence being based on a mixture of results from each. We investigated whether results from both assays could be reconciled. Using statistical methods, the correlation of the two assays’ results and other performance characteristics were examined. Of 378 anti-HIV-1 positive specimens examined by both assays, the Abbott assay flagged 40 as from recent infections, whereas Vironostika flagged 50. The correlation coefficient between screening reactivities in each assay was 0.84, and 0.77 in confirmatory mode. Abbott screening results were significantly higher than its confirmatory results, and some specimens from recent infections may consequently have gone undetected by that assay. This problem was not found with the Vironostika assay. Observational data indicated that the estimated HIV incidence derived from HIVAB results increased as the assay threshold, with its pre-defined seroconversion window, was increased. For Vironostika, the estimated HIV incidence remained stable over a wide range of thresholds. Modelling of the observed relationship between the two assays allowed an estimate of the equivalent threshold in the alternative assay, thus providing a means of reconciling results. Our findings suggest that the Vironostika assay is more reliable than the HIVAB, is easier to use, and is able to allow processing of more specimens per run.

Introduction

Unlinked anonymous surveillance employing the residue of serum specimens remaining after routine laboratory testing is a proven means of obtaining minimally-biased measurements of HIV prevalence in well-defined clinic populations (Nicoll et al., 2000). However, HIV prevalence affords only an historical perspective on the HIV epidemic, principally providing information on the cumulative burden of infection. On the other hand, the ability to identify recently-acquired HIV infection provides additional valuable insight into the epidemic. First, it allows the HIV epidemic to be tracked in real time and, with appropriate epidemiological information, it permits the calculation of the rate of HIV incidence in a defined population. This can aid early recognition of local and national outbreaks with more timely interventions. Furthermore, applying a test of incidence to specimens from newly-diagnosed infections in a population to which interventions have already been applied will evaluate their efficacy. Also, when applied to individual patients at the time of first diagnosis, tests of HIV incidence can be helpful in clinical management and prognosis, and simplify contact tracing indicating recent infection.

A testing algorithm that is able to identify recent HIV-1 infections occurring during the 4–6 months following seroconversion has been described previously (Janssen et al., 1998). The serological testing algorithm for recent HIV seroconversion (STARHS) technique allows recent HIV infections to be determined using single anti-HIV-1 positive specimens. The method relies on serum anti-HIV titres rising gradually, and at a similar rate in each infected individual, for a period of several months following seroconversion. Specimens found to be reactive by a highly sensitive screening test, and that are confirmed to be anti-HIV-1 positive, are re-tested with an enzyme immunoassay (EIA) that has been made less sensitive (or ‘detuned’) by increasing the dilution at which the specimen is tested and reducing the assay incubation times. Recent seroconversion is inferred if a specimen confirmed to be anti-HIV-1 positive is negative in the detuned EIA. Employing specimens from patients with known dates of seroconversion, the duration of the interval between individuals becoming positive in the sensitive assay and subsequently becoming positive in the detuned assay was determined accurately (Janssen et al., 1998). Within limits, the duration of this incidence window can be adjusted by varying the threshold applied to results in the detuned assay. Strict adherence to the detuned assays’ protocols and the use of purpose-made controls and calibrators ensure reliable and reproducible results. Factors such as highly-active anti-retroviral therapy (HAART) and AIDS may lead to this assay falsely classifying a minority of infections as recent due to an associated fall in antibody titre.

Two assays, the Abbott HIVAB (catalogue number 3A11, Abbott Laboratories, Abbott Park, USA) and the Organon Teknika Vironostika HIV-1 Microelisa system (catalogue number 259605, bioMérieux, Cambridge, UK), have been licensed by the Food and Drug Administration (FDA) under an Investigational New Drug Application (IND 8193) co-ordinated by the Centres for Disease Control and Prevention (CDC), Atlanta. The Abbott HIVAB assay was the first to be validated. Its unreliable availability during 2000–2002, and subsequent world-wide withdrawal, led to the validation and adoption of the Vironostika assay. The duration of the interval between seroconversion in the sensitive and the less sensitive assay has been determined for both the Abbott and Vironostika assays. Using the cut-offs proposed by the CDC, the Abbott HIVAB assay detects recent infection occurring in the 129 days (95% CI: 109–149 days) following seroconversion, while the Vironostika assay detects recent infections that occur up to 170 days (95% CI: 162–183 days) after seroconversion (Dr. D. Withum, CDC, Atlanta, personnel communication). However, a lower cut-off of 0.75 has also been validated for the Vironostika assay, and this provides a shorter period of 133 days (95% CI: 127–140 days) during which recent infection can be attributed, similar to that of the Abbott HIVAB assay. Theoretically, this should permit direct comparisons in population-specific incidence rates determined by either manufacturer’s product.

Important products of STARHS are the determination of trends of HIV incidence over time and comparisons between different populations. However, as there has not been continuous availability of a single method, a comparison of the two methods on well-characterised specimens is essential to determine whether the findings are compatible and if a correction factor is needed to reconcile results. This report describes such an investigation.