CROI 2020 Abstract eBook

Abstract eBook

Poster Abstracts

Results: Across all cities, MAA-estimated incidence was generally highest, followed by the serial cross-sectional, with ICC estimates being substantially lower. MAA annual incidence ranged from 18.5% (New Delhi) to zero (Imphal), serial cross-sectional incidence from 16.1% (Kanpur) to 0.3% (Imphal), and ICC incidence from 7.3% (Aizawl) to 0.1% (Imphal). On average, the serial cross-sectional estimate was 19% lower than the MAA (range: -60% to +190%) and 20% higher than the ICC (range: -32% to +953%). While estimates were variable, rank order generally stayed the same across the estimates (Figure). Spearman rank correlation was 0.94 for the MAA-serial cross-sectional estimates, 0.83 for MAA-ICC, and 0.66 for serial cross-sectional-ICC estimates. Conclusion: While HIV incidence estimates within a given city were variable by method, the rank order by incidence was consistent. While use of facility-based data will generally underestimate population incidence, using cross-sectional population-based data to estimate HIV incidence can prioritize where resources may optimally be directed.

(Figure). Cost to identify one undiagnosed PWID was ~10 USD lower in the pRDS approach vs. the standard. Conclusion: A precision RDS approach identified nearly twice as many undiagnosed PWID significantly faster than the standard. While the NNR was not lower in pRDS, given the importance of timely identification and linkage to antiretroviral therapy, pRDS may be particularly useful in outbreaks when rapidly reaching undiagnosed people living with HIV is needed.

Poster Abstracts

891 LATENT CLASS ANALYSIS OF SUBSTANCE USE AND HIV VL TRAJECTORY PATTERNS AMONG PWH IN DC Morgan Byrne 1 , Anne K. Monroe 1 , Lindsey J. Powers Happ 1 , Rupali K. Doshi 1 , Michael A. Horberg 1 , Amanda D. Castel 1 , for the DC Cohort Executive Committee 1 George Washington University, Washington, DC, USA Background: People with HIV (PWH) with substance use disorders (SUD) have worse health outcomes than PWH without SUD. Our objective was to characterize substance use (SU) patterns and their impact on Viral Load (VL) trajectories among PWH. Methods: Data from PWH aged >18 years enrolled Jan 2011-Mar 2018 in the DC Cohort, a longitudinal observational study of PWH in care at 14 clinics in Washington, DC, were analyzed. Data were abstracted from participants’ electronic medical records. SU was defined as documented SU at DC Cohort enrollment and/or the presence of SU-related ICD9/10 codes during study follow-up. Treatments for alcohol and opioid use were also used to identify PWH receiving care for SU. Participants with least 3 VL were included in analysis. Latent class analysis (LCA) was used to determine classes with similar patterns of SU. HIV RNA values were examined using discrete mixture models to determine classes of group-based logVL trajectories and constructed using 3 VL measures. The number of classes for both SU patterns and VL trajectory were chosen using Bayesian Information Criterion, MLE, and maximized model fit. Differences in demographic and clinical characteristics between the SU classes were evaluated using a multivariable-adjusted multinomial model. The relationship between classes of SU patterns and classes of VL trajectories was examined using χ² test. Results: 6,301 participants were assigned to one of three LCA SU classes based on posterior probability: (1) No illicit SU;(2) limited SU and (3) polysubstance use. There were 4 VL trajectory classes: (a) always undetectable; (b) achieved undetectable VL; (c) always VS; and (d) high VL. In multivariable models, individuals in both the polysubstance or limited SU classes were less likely to have private insurance (P<0.05), more likely to be current smokers (P<0.001,) and homeless (P<0.01) compared to the no illicit SU class after adjusting for cohort demographics. Polysubstance use participants were most likely to be categorized in the trajectory that did not achieve VS, followed by participants in the limited SU class (28% and 24% respectively; p-value <0.001). Proportions of participants in each trajectory are shown given membership in SU classes (Fig). Conclusion: LCA identified distinct patterns of SU among PWH, with limited and polysubstance users having higher proportions of high VL trajectories.

890 OPTIMIZING SOCIAL-NETWORK SAMPLING TO FIND UNDIAGNOSED HIV- INFECTED PWID Allison M. McFall 1 , Bryan Lau 1 , Carl A. Latkin 1 , Aylur K. Srikrishnan 2 , Santhanam Anand 2 , Canjeevaram K. Vasudevan 2 , Shruti H. Mehta 1 , Sunil S. Solomon 3 1 Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA, 2 YR Gaitonde Center for AIDS Research and Education, Chennai, India, 3 Johns Hopkins University School of Medicine, Baltimore, MD, USA Background: People who inject drugs (PWID) experience high HIV burden and lag behind in UNAIDS 95-95-95 targets, particularly at diagnosis. We evaluated whether identification of undiagnosed HIV-infected PWID via respondent- driven sampling (RDS), a chain referral approach that leverages social networks, can be enhanced through a precision RDS (pRDS) approach. Methods: We identified characteristics that predicted recruitment of an undiagnosed HIV-infected PWID using previously collected RDS data from PWID in north India. We developed a multivariable prediction algorithm comprised of factors identified by the area under the receiver operator curve from logistic regression models and a random forest. pRDS was tested in Morinda, Punjab where participants were randomly assigned (1:1) to standard or pRDS. In the standard approach, all participants received 2 coupons. For pRDS, an individual’s probability of recruiting an undiagnosed PWID was determined by the algorithm and they received 2 (if low probability) or 5 (if high probability) coupons. The identification rate and number needed to recruit (NNR) - average number recruited in order to find one undiagnosed PWID - of each approach were compared. Results: Predictors of recruiting an undiagnosed HIV-infected PWID included HIV/HCV infection, network size, use of syringe services, and the injection environment. Among 1631 PWID recruited in Morinda, HIV prevalence was 10%, of whom 70%were undiagnosed. From the standard approach, 615 were recruited including 39 who were undiagnosed; from pRDS, 1012 were recruited including 77 who were undiagnosed. pRDS had a significantly higher identification rate of undiagnosed PWID (1.5/week) compared to the standard (0.8/week; difference: 0.7, 95% CI: 0.3, 1.1). However, the NNR for pRDS (13.1) was not significantly lower than the standard coupon system (15.8; difference=2.6, 95% CI: -2.6, 10.0). NNR differences were more substantial in the first four months but decreased over time (test for trend p-value=0.002)

These results may guide planning of SU treatment especially for newly diagnosed PWH to improve their ability to achieve and sustain VS.

CROI 2020 333