Estimation of HIV incidence in a large, community-based, randomized clinical trial: NIMH project accept (HIV Prevention Trials Network 043)

Abstract

Background: National Institute of Mental Health Project Accept (HIV Prevention Trials Network [HPTN] 043) is a large, Phase III, community-randomized, HIV prevention trial conducted in 48 matched communities in Africa and Thailand. The study intervention included enhanced community-based voluntary counseling and testing. The primary endpoint was HIV incidence, assessed in a single, cross-sectional, post-intervention survey of >50,000 participants.

Methods: HIV rapid tests were performed in-country. HIV status was confirmed at a central laboratory in the United States. HIV incidence was estimated using a multi-assay algorithm (MAA) that included the BED capture immunoassay, an avidity assay, CD4 cell count, and HIV viral load.

Results: Data from Thailand was not used in the endpoint analysis because HIV prevalence was low. Overall, 7,361 HIV infections were identified (4 acute, 3 early, and 7,354 established infections). Samples from established infections were analyzed using the MAA; 467 MAA positive samples were identified; 29 of those samples were excluded because they contained antiretroviral drugs. HIV prevalence was 16.5% (range at study sites: 5.93% to 30.8%). HIV incidence was 1.60% (range at study sites: 0.78% to 3.90%).

Conclusions: In this community-randomized trial, a MAA was used to estimate HIV incidence in a single, cross-sectional post-intervention survey. Results from this analysis were subsequently used to compare HIV incidence in the control and intervention communities.

Trial registration: ClinicalTrials.gov NCT00203749.

Figure 1. Algorithms used for quality assurance testing of study samples.

The figure illustrates the testing algorithms that were used to determine and/or confirm the HIV status of study samples. This quality assurance testing was performed at the HPTN Network Laboratory (see Methods). The algorithm used for quality assurance testing was determined by results obtained from HIV rapid testing performed at the study sites (for samples initially designated as HIV NEG, HIV DISC, and HIV POS, see Methods). Quality assurance testing was performed for HIV POS samples if results from the avidity assay suggested absent or very low levels of anti-HIV antibodies (weird avidity). In this case, the HIV DISC algorithm was used to determine HIV status. Neg indicates that a negative or non-reactive test result was obtained. Pos indicates that a positive or reactive test result was obtained. Arrows (non-bolded) indicate the next step in sample testing. The following abbreviations were used to describe assays and tests used in the analysis (see Methods): HIV Combo: ARCHITECT® HIV Ag/Ab Combo assay; EIA: Vitros EIA Human Immunodeficiency Virus Type 1 and/or 2 (HIV-1/2) Antibody Detection in Human Serum and Plasma; GA RNA: APTIMA® HIV-1 RNA Qualitative Assay; WB: Genetics System HIV-1 Western Blot.

Figure 2. Investigation of sample cross-contamination at a study site.

The figure shows two examples of results from two Western blot runs that were performed at the central laboratory as part of an investigation of discordant test results. Results from various laboratory tests are shown above the Western blot strips. HIV rapid tests were performed at a laboratory at the study site in Soweto, South Africa using whole blood; N indicates that both rapid tests were non-reactive, R indicates that both rapid tests were reactive. Samples were subsequently processed to produce plasma aliquots for storage which were later shipped to a central laboratory in the United States for analysis. Results from the ARCHITECT Combo HIV Ag/Ab test are shown (COMBO); N indicates that the Combo test was non-reactive, R indicates the Combo test was reactive. Samples were also tested using the Vitros EIA Human Immunodeficiency Virus Type 1 and/or 2 (HIV-1/2) Antibody Detection in Human Serum and Plasma (EIA); N indicates that the EIA test was non-reactive, R indicates the EIA test was reactive. Western blots were interpreted as negative (N) or positive (P) based on the pattern of bands observed. The banding pattern typically varies among different HIV-positive samples. The panel on the left shows that samples 11–15 were likely to have been cross-contaminated by transfer of plasma from sample 10 into those samples during aliquot preparation (sequential unintended transfer of plasma from tube to tube). Similar findings are shown in the panel on the right; samples 17–19 were likely to have been cross-contaminated by transfer of plasma from sample 16 into those samples. Further investigation at the study site confirmed that a technologist working at the study site prepared sample aliquots without changing pipette tips. All of the samples that may have been processed on the days that this technologist was working in the laboratory were excluded from the endpoint analysis.

Figure 3. Multi-assay algorithm (MAA) used for HIV incidence estimation.

Study samples were initially designated as HIV NEG, HIV DISC, and HIV POS based on HIV rapid testing performed at study sites (see Methods). HIV POS and HIV DISC samples (those that had at least one reactive HIV rapid test) were further evaluated at the HPTN Network Laboratory to determine the HIV status of each sample. The majority of the HIV POS samples and some of the HIV DISC samples were determined to be from individuals with established HIV infection (Table 3). Those samples were analyzed further using a multi-assay algorithm (MAA) developed for HIV incidence estimation. The figure shows the MAA testing schema. Samples were initially tested with the BED capture immunoassay (BED-CEIA) and an avidity assay. Samples that had a BED-CEIA result ≥1.2 normalized optical density units (OD-n) were considered to be MAA negative and were not evaluated further. The remaining samples were evaluated based on results of the avidity assay. Samples that had an avidity assay result (avidity index) ≥90% were considered to be MAA negative and were not evaluated further. The remaining samples were evaluated based on results of CD4 cell count testing that was performed at study sites around the time of sample collection (CD4). Samples that had CD4 cell count result <200 cells/mm³ were considered to be MAA negative and were not evaluated further; if a CD4 cell count result was not obtained at the time of sample collection, recency could not be assessed. The remaining samples were tested using an HIV viral load assay (VL). Samples that had a viral load result <400 copies/mL were considered to be MAA negative and were not evaluated further. Samples that met all of the criteria for the MAA (BED-CEIA <1.2 OD-n+avidity index <90%+CD4 cell count >200 cells/mm³+ HIV viral load >400 copies/mL) were classified as MAA positive.