CROI 2017 Abstract e-Book

Abstract eBook

Poster and Themed Discussion Abstracts

Poster and Themed Discussion Abstracts

879 IDENTIFICATION AND VALIDATION OF AN INCIDENCE-TESTING ALGORITHM FOR HIV-1 SUBTYPE C Oliver Laeyendecker 1 , Jacob Konikoff 2 , Douglas Morrison 3 , Ronald Brookmeyer 3 , Jing Wang 4 , Connie L. Celum 5 , Charles S. Morrison 6 , Quarraisha Abdool Karim 7 , Audrey Pettifor 8 , Susan H. Eshleman 9 1 NIAID, NIH, Baltimore, MD, USA, 2 The Johns Hopkins Univ, Baltimore, MD, USA, 3 Univ of California Los Angeles, Los Angeles, CA, USA, 4 Fred Hutchinson Cancer Rsr Cntr, Seattle, WA, USA, 5 Univ of Washington, Seattle, WA, USA, 6 FHI 360, Durham, NC, USA, 7 CAPRISA, Durban, South Africa, 8 Univ of North Carolina at Chapel Hill, Chapel Hill, NC, USA, 9 The Johns Hopkins Univ, Baltimore, MD, USA Background: Accurate methods for estimating HIV incidence are needed for surveillance and to assess prevention efforts, particularly in HIV subtype C endemic areas where the burden of disease is the greatest. We evaluated assays and multi-assay algorithms (MAAs) for cross-sectional HIV incidence estimation in subtype C settings. Methods: We analyzed 2442 samples from 278 adults with known duration of infection (0.1 to 9.9 years after seroconversion). Samples were collected in 3 studies conducted in Zambia, Zimbabwe, and South Africa (CAPRISA; Hormonal Contraception and Risk of HIV Acquisition; HPTN 039). The following assays were evaluated: the Limiting Antigen Avidity assay (LAg, cut-offs: 0.5 to 3 normalized optical density [OD-n]); the Johns Hopkins modified BioRad Avidity assay (BioRad, cut-offs: 30 to 100% avidity index); CD4 cell count (cut-offs: 50 to 500 cells/mm3); and viral load (cut-offs: 400 to 10,000 copies/mL). We evaluated >25,000 MAAs, varying the assays used and cut-off for each assay. For each assay or MAA, we computed the mean window period (mean duration individuals are classified as ‘recent’); mean duration of recent infection (MDRI, mean duration individuals are classified as ‘recent’ in the first 2 years of infection); shadow (how far back in time incidence is being measured); and false recent rate (FRR, fraction of individuals infected >2 years misclassified as ‘recent’). We selected MAAs with the largest mean window period, where the upper 95% confidence interval (CI) of shadow was <1 year. Assays and MAAs were compared to the LAg standard algorithm (LAg <1.5 OD-n + VL > 1000) and were used to estimate HIV incidence in a longitudinal cohort from South Africa (HPTN 068). Results: The table shows data from the LAg standard algorithm (I: LAg <1.5 + VL >1000); two clade B algorithms (II: BioRad <40%+ LAg <2.9, III: BioRad <85%+ LAg <2.8 + VL >400 + CD4 >50); LAg alone (IV: LAg <0.7); BioRad alone (V: BioRad <40%); and the overall optimal MAA (VI: BioRad <95%+ LAg <2.8 + VL >400). The optimal MAA (VI) estimated incidence in the HPTN 068 cohort with an error that was 3-fold less than the LAg standard algorithm, with CIs for the incidence estimate that were half as wide. Conclusion: We identified an optimized MAA for cross-sectional HIV incidence in subtype C settings. This MAA, which includes the LAg assay, BioRad assay, and viral load, is more accurate than the LAg standard algorithm currently in use for global HIV surveillance. 880 VALIDATION OF LIMITING ANTIGEN AVIDITY ASSAY TO ESTIMATE HIV INCIDENCE IN EAST AFRICA Oliver Laeyendecker 1 , Ronald H. Gray 2 , Kate Grabowski 2 , Steven Reynolds 1 , Anthony Ndyanabo 3 , Joseph Ssekasanvu 2 , Gilad Bismut 2 , Maria Wawer 2 , David Serwadda 4 , Thomas C. Quinn 1 1 NIAID, Baltimore, MD, USA, 2 The Johns Hopkins Univ, Baltimore, MD, USA, 3 Rakai Hlth Scis Prog, Kalisizo, Uganda, 4 Makerere Univ, Kampala, Uganda Background: Cross-sectional incidence testing will be used for Population-based HIV Impact Assessments in several East African countries. These will use a combination of the Limiting-Antigen (LAg) Avidity Assay with viral load (VL) >1000 copies/mL to estimate population level incidence. We aimed to validate the capacity of the LAg-Avidity + VL algorithm to estimate incidence in a subtype A and D epidemic compared to cohort observed incidence. Methods: We tested all samples from a single survey round of the Rakai Community Cohort (2008-2009) which had previously determined HIV subtype distribution and an observed incidence from individuals who were also surveyed in the prior round. The observed incidence was 1.05/ 100 person years (95% CI 0.90, 1.23). The sample set included 544 individuals infected >2years, for which a site-specific false recent rate (FRR) was determined. We compared incidence results per protocol (mean duration of recent infection [MDRI] of 130 days using a normalized optical density [OD-n] of 1.5 and 0% FRR after excluding those on ARV and VL<1000), to recent independent recommendations (MDRI 141 days and site specific FRR). Samples from HIV positive individuals who self-reported ARV use or with clinical records of ARV treatment were excluded from the analysis. Results: There were 9973 participants present at both surveys, with 1253 HIV+ subjects of whom 866 were not on ART. 822/866 HIV positive subjects not on ART had samples available for testing. 94/822 samples had LAg-Avidity values < 1.5 OD-n, and 49/94 had detectable VL. The site specific FRR was 1.1% (95% CI 0.4-2.4% [6/544]). Of the 161 individuals who seroconverted over the 18 months between surveys, the LAg-Avidity+ VL identified 27 as recently infected. The estimated incidence per protocol was 1.73% (95% CI 1.03, 2.22), 65% higher than the observed point estimate. Using the updated MDRI of 141 days and a site determined FRR of 1.1%, the incidence estimate was 1.38% (95% CI 0.83, 1.93), 33% higher than the observed cohort incidence. Conclusion: Revised MDRI and locally-estimated FRR improved the incidence estimate. The LAg-Avidity + VL estimate of incidence using revised performance characteristics was still 33% higher than the observed estimate, though confidence intervals overlapped. Nearly half of individuals with low LAg-Avidity values, with no self-reported or clinical record of ARV use had undetectable VL. 881 MISCLASSIFICATION RATE OF HIV ANTIBODY AVIDITY ASSAYS IN INDIVIDUALS FROM CAMEROON Briana Lynch 1 , Eshan Patel 2 , Colleen R. Courtney 3 , Aubin J. Nanfack 3 , Thomas Quinn 2 , Oliver Laeyendecker 4 , Phillipe N. Nyambi 3 , Ralf Duerr 3 , Andrew D. Redd 5 , for the Medical Diagnostic Center,Yaoundé, Cameroon 1 NIH, Bethesda, MD, USA, 2 The Johns Hopkins Univ, Baltimore, MD, USA, 3 New York Univ, New York, NY, USA, 4 NIAID, Baltimore, MD, USA, 6 NIAID, Bethesda, MD, USA Background: Accurate estimates of HIV incidence are critical to surveillance and prevention efforts. Current cross-sectional incidence testing strategies are dependent on antibody dynamics during recent and chronic infection, and many individual and population-level factors can affect their performance. Limited information exists on the

CROI 2017 381