Abstract Body

Background: HIV transmission networks inferred from sequence data can be used to identify risk groups for intervention. Transmission clusters spread over larger geographic areas may require different methods of intervention than localized clusters. We investigated how individual transmission risk factor predicted the geographic spread of HIV within and outside NYC. Methodology: We analyzed 39,135 individuals’ earliest HIV pol sequence reported to NYC through routine laboratory surveillance during 2005-2012. . All sequences were aligned to the reference genotype HXB2. Potential transmission partners were defined as two individuals whose sequences had a pairwise genetic distance ≤2%; groups of linked partners formed transmission clusters. We applied multivariate logistic and ordinal logistic regression to identify characteristics associated with clustering and geographic spread. NYC geographical information was derived from zip code of residence at diagnosis as grouped into standard multi-zip NYC neighborhoods. We also included 86,216 non-US sequences from the Los Alamos National Laboratory (LANL) HIV sequence database to investigate global transmission links. Results: The network contained 8801 nodes (22.5% of all individuals) and 19,637 edges (connections between potential transmission partners). Propensity for clustering within the NYC network was not uniform: men who have sex with men (MSM) were more likely to cluster than heterosexuals (OR =2.4; p <0.001), and injecting drug users (IDU) were less likely to cluster than heterosexuals (OR = 0.52; p <0.001). Both MSM and IDU had transmission partners spread over a greater median distance than heterosexuals (10.3 and 9.0 vs. 7.8 km; p <0.01). Of all individuals clustered within the NYC network, 28.3% have a potential partner diagnosed within, but residing outside of, NYC. Among individuals connected in the NYC network, MSM were more likely to cluster with individuals residing outside of NYC than heterosexuals (36.6% vs. 8.2%; p <0.001), but not with individuals from other countries in the LANL reference dataset. IDU were unexpectedly more likely than heterosexuals to have an international connection in LANL (18.5% vs. 6.2%; p < 0.001). Conclusions: Risk factor plays an important role in predicting the geographic spread of HIV transmission clusters. MSM have the most geographically expansive clusters. However, IDU clusters are not as geographically constrained (locally or internationally) compared with heterosexual clusters. Interventions targeting specific risk groups should take the geographic spread of these clusters into account.