CROI 2015 Program and Abstracts

Abstract Listing

Poster Abstracts

625 False Recent Rates for Two Recent Infection Testing Algorithms, South Nyanza, Kenya Clement Zeh 1 ; David Maman 4 ; Harrison Omondi 2 ; Alex Morwabe 2 ; Collins Odhiambo 2 ; Beatrice Kirubi 4 ; Irene Mukui 3 ; MartinusW. Borgdorff 1 ; Jean-François Etard 4 ; Andrea A. Kim 1 1 US Centers for Disease Control and Prevention, Kisumu, Kenya; 2 Kenya Medical Research Institute, Kisumu, Kenya; 3 National AIDS and STI Control and Prevention, Nairobi, Kenya; 4 Médecins Sans Frontières, Paris, France Background: Evaluation of candidate tests for recent HIV infection (TRI), designed to distinguish recent from chronic HIV infection, is an essential step prior to estimating cross- sectional HIV incidence. The TRI’s false-recent rate (FRR), the probability that a chronic infection will misclassify as recent, is a required parameter for calculating HIV incidence and should not exceed 2% for accuracy. Because the FRR varies by TRI and sub-population, the FRR should be assessed in all settings in which HIV incidence will be estimated. We compare the FRR for the Limiting Antigen Avidity Enzyme Immunoassay (LAg) and Bio-Rad Avidity Enzyme Immunoassay (Bio-Rad), respectively, in a high HIV prevalence setting in South Nyanza, Kenya. Methods: We conducted a population-based household survey of persons aged 15-59 years in Ndhiwa District in South Nyanza, Kenya. HIV treatment naive participants with documented chronic HIV infection (defined as testing HIV+ in the survey and reporting the first HIV+ test result ≥ 12 months preceding the survey) were tested for recent infection using the LAg and Bio-Rad on serologic blood samples. Recent infection was defined based on two recent infection testing algorithms (RITA): 1) a multi-assay algorithm (MAA) which defined a recent case as: a) tested recent on the TRI; b) not virally suppressed defined as HIV-1 RNA concentration ≥ 400 copies/mL; and 2) a single-assay algorithm (SAA) which defined a recent case as tested recent on the TRI. The FRR was calculated by dividing the number of recent cases observed on the RITA by the number of chronic infections tested. Results: Of 1,465 HIV-positive samples, 835 (57.0%) were chronic infections. Based on the MAA, the FRR was 0.5% (95% CI 0.01 – 1.0) for LAg and 2.4% (95% CI 1.4 – 3.4) for Bio-Rad. Based on the SAA, the FRR was 4.6% (95% CI 3.2 – 6.0) for LAg and 7.2% (95% CI 5.5 – 9.0) for Biorad. The FRR did not differ by sex and RITA, but varied by age group for the two RITAs. In the MAA, the FRR was highest among youth aged 15-24 years (1.2%; 95% CI 0 – 3.5 for LAg; 3.5%; 95% CI 0 – 7.4 for Bio-Rad). In the SAA, the FRR was highest among persons aged 45-59 years at 5.7%; 95% CI 2.8 – 8.6 for LAg and 8.9%; 95% CI 5.4 – 12.5 for Bio-Rad. Conclusions: The recommended threshold for a FRR was met by LAg, but only in the MAA which excluded individuals with suppressed viral load. Performance of the TRIs using the SAA resulted in high FRRs that are inappropriate for estimating incidence. 626 Viral Load is Critical in Limiting False-Recent Results FromHIV Incidence Assays Reshma Kassanjee 1 ; Shelley Facente 2 ; Sheila Keating 3 ; Elaine McKinney 4 ; Kara Marson 2 ; Christopher D. Pilcher 2 ; Michael Busch 3 ; Gary Murphy 4 ; Alex Welte 1 The Consortium for the Evaluation and Performance of HIV Incidence Assays (CEPHIA) 1 South African DST/NRF Centre of Excellence in Epidemiological Modelling and Analysis (SACEMA), University of Stellenbosch, Stellenbosch, South Africa; 2 University of California San Francisco, San Francisco, CA, US; 3 Blood Systems Research Institute, San Francisco, CA, US; 4 Public Health England, London, United Kingdom Background: The cross-sectional use of (biomarker) tests for recent HIV infection in principle offers affordable, low-bias options for incidence estimation. For currently available assays, viral suppression (due to elite control or antiretroviral treatment) is predictive of long-term infections being (‘falsely’) classified as ‘recent’. Surveillance requires a not- too-transient ‘mean duration of recent infection’ (MDRI) – preferably at least 6 months. Assay readings below a chosen threshold are interpreted as indicating ‘recent’ infection, and any assay threshold sufficiently high to achieve a large MDRI inevitably incurs a substantial ‘false-recent rate’ (FRR), which should preferably be no higher than 1%. The performances of seven assays (BED, Limiting Antigen (LAg), Less-Sensitive (LS) Vitros, Vitros Avidity, BioRad Avidity, Architect Avidity, Geenius) were compared, in stand-alone form and in conjunction with a rule that low viral load is indicative of non-recent infection, allowing for varying assay and viral load thresholds. Methods: Specimens were used from a growing repository, previously described, of over 6000 specimens representing over 2000 subjects from studies in Africa, Brazil and the United States. Assay thresholds were adapted to produce the same MDRI, estimated by binomial regression. Within a model scenario inspired by the contemporary South African context, the net model population-level FRRs were estimated by combining FRR estimates for key subgroups (stratifying by time since infection and treatment status). Results: Table 1 shows the model population-level FRR for each assay, using an assay threshold that provides an MDRI of 200 days in each case, used either alone or with a viral load rule (using viral load thresholds of 75 and 1000 copies/ml).

Poster Abstracts

Table 1: False-recent rates for recent infection testing algorithms Conclusions: Adapted to provide a standard desirable MDRI of 200 days, none of the assays, used alone, provide an acceptably low FRR. With the use of any realistic viral load threshold, the FRR values drop dramatically, to between 0.4% and 3.3%, which is operationally feasible for population-level surveillance in high incidence contexts. Increasing the viral load threshold above 75 copies/ml offered little improvement in FRRs while decreasing MDRIs. Methods for optimally combining all information about predictors of ‘false-recent’ results into real-world context-specific FRR estimates require further development. Also, judicious combinations of these assays could potentially yield further improvements in performance.

394

CROI 2015