CROI 2018 Abstract eBook

Abstract eBook

Poster Abstracts

range 11-13) was 8.4/100 person-years (IQR 7-9). Of the 1,260 putative links, 942 (74.8%) were between individuals residing in the same state and 116 (9.2%) were between individuals from states with a shared border. 43.3% (221/510) of the clusters spanned multiple states, and two-thirds (147/221,66.5%) included individuals from the Center-South region, which includes Mexico City (Figure 1). Sampled sequences from border states of Baja California and Quintana Roo, and the state of Puebla, with strong migrational links to New York City, were significantly more likely to cluster (OR: 1.57, 1,73 and 1.87 respectively (p<0.001). The median distance between linked individuals from different states (based on centroids) was 445 km [IQR:160-834], suggesting regional transmission patterns. Viral migration analysis revealed that the Center-South and East regions, which include the states of Veracruz and Puebla, were main hubs of the epidemic, as significantly more inferred transmissions originated from these regions to the rest of the country. Conclusion: Viral migration patterns highlight the regional nature of transmission in Mexico, and also demonstrate that the major metropolitan areas of Mexico City, Puebla and Veracruz were important hubs in interstate transmission. Together these results are consistent with a gravity model of transmission, and suggest that focusing prevention resources on major metropolitan hubs may have an enhanced effect on reducing new HIV transmissions.

from new transmission or diagnosis of existing infections. Nevertheless, these findings suggest that clusters in which not all persons are virally suppressed could be prioritized for further investigation and intervention. That a substantial portion of clusters with complete suppression continue to growmight represent transmission from unsuppressed persons whose disease is not yet diagnosed or reported. As we explore the use of this novel tool to guide prevention efforts, future analyses of factors associated with growth of clusters will help prioritize the most concerning clusters to maximize the primary and secondary prevention benefits of cluster identification and investigation. 956 HIGH VIRAL LOAD ACROSS STAGE OF INFECTION ASSOCIATED WITH CLUSTERING IN US NETWORK Joel O. Wertheim 1 , Alexandra M. Oster 2 , Nivedha Panneer 2 , Chenhua Zhang 2 , Neeraja Saduvala 3 , Ellsworth Campbell 2 , Jeffrey A. Johnson 2 , WilliamM. Switzer 2 , Walid Heneine 2 1 University of California San Diego, San Diego, CA, USA, 2 CDC, Atlanta, GA, USA, 3 ICF International, Atlanta, GA, USA Background: HIV spreads across sexual and injection drug-using partner networks, resulting in clusters of genetically similar viruses. The extent of this clustering is more often associated with risk behavior than viral traits (i.e., viral transmission fitness). Viral load (VL) is a viral trait associated with transmissibility and disease progression. We examined cases diagnosed at different stages of infection to test the hypothesis that VL is associated with transmission fitness (i.e., clustering) in a large U.S. transmission network. Methods: We analyzed HIV-1 polymerase sequences from 24,028 persons from the U.S. National HIV Surveillance Systemwho were genotyped < 3 months of HIV diagnosis during 2001–2016, received a VL measurement before or within 1 month of genotyping, were treatment-naïve at diagnosis, and had no drug resistance mutations. HIV-TRACE was used to construct a molecular transmission network. We used multiple linear regression analysis to assess the relationship between the log 10 earliest VL measurement and clustering in 5,914 cases diagnosed at Stage 1 infection (CD4 cell count ≥500/µL). Birth sex, transmission risk factor, race/ethnicity, age at diagnosis, year of diagnosis, subtype, and CD4 count were included as covariates. Similar analysis was performed on cases diagnosed at Stage 2 (CD4 200–499/µL) and Stage 3 (CD4 <200/µL). Results: The 2,787 cases diagnosed at Stage 1 that clustered in the network had a mean VL of 70,745 copies/ml, 27% higher than the 3,127 unclustered cases (p < 0.001). This finding was robust to the timing of VL measurement, demographic/risk covariates, CD4 count, network structure, genetic distance threshold for assigning partner clustering (0.005 to 0.015 substitutions/site), and inclusion/exclusion of people who inject drugs. Larger clusters (≥5 vs. <5 and ≥10 vs. <10 persons) had increasingly higher VL in cases diagnosed at Stage 1 (p = 0.01). Similar patterns of higher VL in clustered cases were observed for those cases diagnosed at Stage 2 (27%; p < 0.001) and Stage 3 (7.5%; p = 0.003). Conclusion: Circulating wild type viruses in a large transmission network differ in transmissibility. The robust association between VL and clustering reflects a heritable and durable viral trait maintained throughout infection stages. These findings heighten the importance of interrupting growing transmission clusters comprising cases with high VL through network-assisted targeting of public health interventions. 957 HIV TRANSMISSION CLUSTER DYNAMICS THAT INFORM PUBLIC HEALTH INTERVENTION IN ILLINOIS Manon Ragonnet-Cronin 1 , Joel O. Wertheim 1 , Christina S. Hayford 2 , Richard D’Aquila 2 , Fangchao Ma 3 , Cheryl Ward 3 , Nanette Benbow 2 1 University of California San Diego, San Diego, CA, USA, 2 Northwestern University, Chicago, IL, USA, 3 Illinois Department of Public Health, Springfield, IL, USA Background: Health departments across the US collect HIV sequence data from routine drug resistance tests. HIV sequences from different individuals that are closely related (i.e. clustered) can reveal high-risk groups warranting targeted public health intervention. However, clustering may be indicative of past transmissions, whereas recent cluster growth likely better reflects active transmission. Methods: We analyzed HIV sequences reported to the Illinois Department of Public Health between 2013-2017 using HIV-TRACE. Transmission clusters were identified at a pairwise genetic distance threshold of 0.015 substitutions/site. A cluster growth statistic was calculated: the number of new cluster members in 2016/2017 divided by the square root of cluster size. We sought epidemiological

Poster Abstracts

955 ASSOCIATION BETWEEN VIRAL SUPPRESSION AND MOLECULAR CLUSTER GROWTH, UNITED STATES Nivedha Panneer 1 , Alexandra M. Oster 1 , Cheryl B. Ocfemia 1 , Sheryl Lyss 1 , Joel O. Wertheim 2 , Anne Marie France 1 1 CDC, Atlanta, GA, USA, 2 University of California San Diego, San Diego, CA, USA Background: Molecular clusters identified through analysis of HIV sequences can identify groups of persons among whom HIV is rapidly spreading; these clusters can be prioritized for prevention interventions. Since not all clusters are equally likely to contribute to ongoing transmission, identifying factors predictive of cluster growth is critical for prioritization. As part of an early effort to identify such factors, we assessed whether lack of viral suppression within a small cluster was associated with cluster growth. Methods: We analyzed HIV-1 pol sequences reported to the National HIV Surveillance System through December 2016 for 51,750 persons with HIV diagnosed during 2013–2016. We identified potential transmission pairs at a genetic distance threshold of ≤0.5%. We restricted analysis to clusters of 3 persons with HIV diagnoses during 2013–2015. We categorized as virally suppressed those persons with a reported viral load result of <200 copies/ml during 2015 and, for those with >1 viral load result in 2015, if the most recent was <200 copies/ml. We considered clusters to be incompletely suppressed if ≥1 person in the cluster was not suppressed or completely suppressed if all persons in the cluster were suppressed. We then determined which clusters grew by ≥1 person in 2016. Results: Of 494 clusters of size three, 84 (17%) grew by ≥1 person in 2016. Cluster growth was identified for 68 (20%) of 347 clusters with incomplete suppression compared with 16 (11%) of 147 clusters with complete suppression. The relative risk of growth among clusters with incomplete versus complete suppression was 1.8 (95% CI: 1.08, 3.00). Conclusion: Incomplete viral suppression was associated with cluster growth among clusters of 3 persons. We cannot determine whether growth resulted

CROI 2018 365