Characterizing HIV transmission networks across the United States

Abstract

Background: Clinically, human immunodeficiency virus type 1 (HIV-1) pol sequences are used to evaluate for drug resistance. These data can also be used to evaluate transmission networks and help describe factors associated with transmission risk.

Methods: HIV-1 pol sequences from participants at 5 sites in the CFAR Network of Integrated Clinical Systems (CNICS) cohort from 2000-2009 were analyzed for genetic relatedness. Only the first available sequence per participant was included. Inferred transmission networks ("clusters") were defined as ≥2 sequences with ≤1.5% genetic distance. Clusters including ≥3 patients ("networks") were evaluated for clinical and demographic associations.

Results: Of 3697 sequences, 24% fell into inferred clusters: 155 clusters of 2 individuals ("dyads"), 54 clusters that included 3-14 individuals ("networks"), and 1 large cluster that included 336 individuals across all study sites. In multivariable analyses, factors associated with being in a cluster included not using antiretroviral (ARV) drugs at time of sampling (P < .001), sequence collected after 2004 (P < .001), CD4 cell count >350 cells/mL (P < .01), and viral load 10,000-100,000 copies/mL (P < .001) or >100,000 copies/mL (P < .001). In networks, women were more likely to cluster with other women (P < .001), and African Americans with other African Americans (P < .001).

Conclusions: Molecular epidemiology can be applied to study HIV transmission networks in geographically and demographically diverse cohorts. Clustering was associated with lack of ARV use and higher viral load, implying transmission may be interrupted by earlier diagnosis and treatment. Observed female and African American networks reinforce the importance of diagnosis and prevention efforts targeted by sex and race.

Figure 1.

Cluster overview with pictorial representation of all patients who clustered at ≤1.5%. Each cluster patient (n = 885) is represented as a colored dot with lines connecting phylogenetically linked sequences. Black line connections represent genetic distances of <0.05%; dark gray lines, 0.5%–1%; and light gray lines, 1%–1.5%. Sequences in cluster 3 can be seen in the center of the figure. Red dots represent patients from the University of California, San Francisco; green, the University of Washington; blue, Harvard/Fenway; purple, Case Western Reserve University; and orange, the University of North Carolina.

Figure 2.

Map of cohort population, percentage of clustering by site, and proportion of cluster patients in “dyads” (clusters of 2 patients), “networks” (clusters of >2 patients), or in “cluster 3″ (a unique cluster that spanned all sites and included 9% of cluster sequences).

Figure 3.

Analysis of networks. Clusters containing ≥3 patients were defined by the predominant demographic characteristic of patients in that network. For example, a network was defined as “female” if >50% of the patients in the network were female. A, Among women, 36 of 43 (84%) clustered with other women. B, Among blacks, 45 of 58 (78%) clustered with other blacks.