CROI 2017 Abstract e-Book

Abstract eBook

Poster and Themed Discussion Abstracts

Conclusion: We have developed a predictive scoring systemwhich is based on combining LIA Algorithm 15.1 and the VL. This proof of concept analysis has demonstrated that data such as LIA and VL can be used to develop a multiassay algorithm to accurately predict an individual’s risk of recently acquired HIV infection.

514 THE HIV GENOMIC INCIDENCE ASSAY MEETS FRR AND MDRI PERFORMANCE STANDARDS

Sung Yong Park 1 , Tanzy Love 2 , Kathryn Anastos 3 , Jack DeHovitz 4 , Chenglong Liu 5 , Kord Kober 6 , Mardge Cohen 7 , Wendy Mack 1 , Elizabeth T. Golub 8 , Ha Youn Lee 1 1 Univ Southern California, Los Angeles, CA, USA, 2 Univ of Rochester, Rochester, NY, USA, 3 Albert Einstein Coll of Med, Bronx, NY, USA, 4 SUNY Downstate Med Cntr, Brooklyn, NY, USA, 5 Georgetown Univ, Washington, DC, USA, 6 Univ of California San Francisco, San Francisco, CA, USA, 7 Stroger Hosp, Chicago, IL, USA, 8 Johns Hopkins Univ, Baltimore, MD, USA Background: HIV incidence is a direct metric of HIV intervention and prevention trial efficacy. Incidence assay performance is evaluated by mean duration of recent infection (MDRI) and false-recent rate (FRR). A low FRR is required for accurate incidence determination and a higher MDRI allows incidence estimation from a smaller sample size. There is an immediate need for an assay meeting performance standards (MDRI > 1 year and FRR < 1%) to estimate incidence using a single measure. Methods: We conducted a meta-analysis of HIV envelope genes sequenced from 438 incident specimens and 305 chronic specimens representing a wide range of geographic locations, subtypes, risk behaviors, ART experiences, viral loads and CD4 T cell counts. The genomic assay’s FRR was measured from chronic specimens collected at least 2 years after documented HIV infection using the genome similarity index (GSI) as a biomarker of recency. The incident specimens included 186 serial HIV sequence samples likely collected within 6 months of infection (Fiebig stage IV). In order to estimate the MDRI, we statistically modeled the average GSI dynamics with a logistic regression assuming individual variabilities in a Beta distribution. We then tested our genomic assay by sequencing 407 HIV envelope gene segments from 15 Women’s Interagency HIV Study (WIHS) seroconverters who were followed from HIV negative status. To assess how closely the WIHS cohort GSI dynamics resemble the Beta distribution estimate, each WIHS specimen’s standardized residual was evaluated and the Anderson-Darling test was conducted. Results: All except one chronic specimen had GSIs below 0.67, yielding a FRR of 0.33 [0-1.0] %with a 2-year cutoff. The GSI probability density function estimated from 438 incident specimens peaked close to a GSI of 1 in early infection and a GSI of 0 around 2 years post infection. Around 1 year post infection the GSI probability density function peaked at both ends of the GSI spectrum. The resulting MDRI was estimated to be 420 [357, 469] days. Both standardized residuals and the Anderson-Darling tests suggest that our WIHS cohort sequence dataset was statistically consistent with the model GSI dynamics. Conclusion: This is the first incidence assay conforming to FRR and MDRI performance standards. Signatures of HIV gene diversification offer a foundation for a precise genomic assay with a desirable temporal range of incidence detection. Our finding suggests great promise for realizing accurate cross-sectional incidence surveys. 515 Background: Accurate measures of HCV incidence are needed for surveillance and prevention efforts. Acute HCV infection is a stage of recent infection (virologically+ but IgG antibody[Ab]-), but the duration of this state may vary depending on the assays used. We compared the duration of acute HCV infection before seroconversion as defined by various assay algorithms, and examined their impact on the precision of biomarker-based incidence estimation. Methods: Retrospective analysis of data from HCV seroconversion panels of initially uninfected individuals, sampled at a median of 5-day intervals (SeraCare and Zeptometrix), was used estimate the mean duration of acute HCV infection (the average duration between infection and Ab seroconversion). To estimate the date of infection, HCV uninfected individuals were assayed by nucleic acid testing (NAT) for HCV RNA detection and/or by the Abbott Murex HCV Ag/Ab Combo Assay. The estimated date of HCV IgG Ab seroconversion was monitored by various serologic assays. Using days before seroconversion as the unit of analysis, the mean duration of acute HCV infection was calculated using binomial regression with a logit cubic functional form and a maximum likelihood approach. Results: The range in the mean duration of acute infection was 20 to 35 days depending on the assay used to identify initial infection and IgG Ab seroconversion (Table 1). NAT testing resulted in longer estimated period of acute infection than using the Ag/Ab Combo Assay by approximately 6-9 days (Table 1). Although there is a loss in the mean duration of acute infection by using the Ag/Ab Combo Assay, this did not meaningfully affect sample size considerations for precise incidence estimation in many hypothetical contexts. To achieve a relative standard error of 20% for an expected incidence estimate of 20%, a survey of 1700 individuals would be required for the largest estimated interval of acute infection from these data (35 days). To achieve the same precision with a 20 day acute interval (smallest interval), the survey size would need to be increased to 4400. Conclusion: Efforts to optimize biomarker-based methods to estimate the population-level incidence of HCV infection may yield a sensitive and powerful tool to actively monitor the progress toward the WHO’s goal of HCV elimination by 2030. Employing acute screening in cross-sectional testing algorithms would extend the window period of identifying a ‘recent’ infection, but this will be dependent on the combination of assays used. DURATION OF ACUTE HEPATITIS C INFECTION AND IMPLICATIONS FOR INCIDENCE ESTIMATION Eshan Patel 1 , Anna Eisenberg 2 , Jeffrey Quinn 1 , Thomas C. Quinn 3 , Aaron Tobian 1 , Oliver Laeyendecker 2 1 Johns Hopkins Univ, Baltimore, MD, USA, 2 NIAID, Baltimore, MD, USA, 3 NIAID, Bethesda, MD, USA

Poster and Themed Discussion Abstracts

CROI 2017 214