Using HIV networks to inform real time prevention interventions

Abstract

Objective: To reconstruct the local HIV-1 transmission network from 1996 to 2011 and use network data to evaluate and guide efforts to interrupt transmission.

Design: HIV-1 pol sequence data were analyzed to infer the local transmission network.

Methods: We analyzed HIV-1 pol sequence data to infer a partial local transmission network among 478 recently HIV-1 infected persons and 170 of their sexual and social contacts in San Diego, California. A transmission network score (TNS) was developed to estimate the risk of HIV transmission from a newly diagnosed individual to a new partner and target prevention interventions.

Results: HIV-1 pol sequences from 339 individuals (52.3%) were highly similar to sequences from at least one other participant (i.e., clustered). A high TNS (top 25%) was significantly correlated with baseline risk behaviors (number of unique sexual partners and insertive unprotected anal intercourse (p = 0.014 and p = 0.0455, respectively) and predicted risk of transmission (p<0.0001). Retrospective analysis of antiretroviral therapy (ART) use, and simulations of ART targeted to individuals with the highest TNS, showed significantly reduced network level HIV transmission (p<0.05).

Conclusions: Sequence data from an HIV-1 screening program focused on recently infected persons and their social and sexual contacts enabled the characterization of a highly connected transmission network. The network-based risk score (TNS) was highly correlated with transmission risk behaviors and outcomes, and can be used identify and target effective prevention interventions, like ART, to those at a greater risk for HIV-1 transmission.

Figure 1. The inferred transmission network (excluding unconnected individuals) in the SDPIC.

Only clustered individuals (nodes) within the network are shown (52.3%). Despite the likely presence of unsampled (i.e., missing) nodes, a partial HIV-1 transmission network is color coded; the intensity of coloring of nodes determined by their TNS score, while that for directed edges corresponds to the viral load of the putative initial partner at the timepoint closest to the transmission event. Absence of blue shading indicates that no VL was available for the sampled individual at any timepoint or that the direction of the edge could not be ascertained using EDI (see text). Absence of red shading indicates a TNS  = 0 (i.e. nodes that were unconnected at the time of enrollment). Nodes are connected with an edge (i.e., a line to indicate potential transmission) if the minimum distance between the respective pol sequences (i.e., possible transmission pairs) is less than 1.5%. A direction is assigned to an edge if the EDI for the secondary partner (i.e., putative “recipient”) is at least 30 days after the sampling date of the putative transmitting partner (i.e., putative “source”). The direction of transmission was resolved for in 332 of the 540 individuals (61.5%).

Figure 2. Simulations of ART provided either to those with the highest TNS or to a random subset of clustered individuals.

Black nodes are those that are being treated (the assumption is that ART is 100% effective at stopping all secondary transmissions). Other nodes are colored according to how likely they are to be prevented from becoming infected assuming that we have removed the treated nodes from the infectious pool; they are also labeled by the rate at which they are expected to be effectively protected (dark red = very high probability of preventing transmission). These values are derived from simulating treatment where the randomness comes from the fact that should a node have N possible infectious connections, K of which are treated/removed due to treatment of other nodes, the node itself will NOT become infected with probability K/N. Targeting high TNS in panel A shows (i) The removal of an entire large cluster, where many nodes have high TNS (ii) prevented chains of transmission (i.e. even nodes that are not directly connected to the treatment subset have a high probability from being protected). Targeting the same number of random nodes in panel B shows: largely a very local effect and almost no chains being disrupted (with the exception of a cluster that is randomly chosen). Both panel A and panel B networks have the same topology, though the appearance is slightly different to allow labeling of specific nodes.

Figure 3. Schematic of TNS Clinical Application and Outcomes.

The schematic illustrates in a step-by-step fashion (numbers 1-6), the application of TNS to clinical care and potential outcomes. The standard of clinical care for newly HIV diagnosed persons (1) includes baseline HIV pol sequence evaluation (2) to screen for ART drug resistance. With development of appropriate privacy preserving methods, these same data could be evaluated to determine a TNS (3). Feedback of TNS with drug resistance results (4), including an interpretation and description of potential limitations, could inform clinical care decisions (5). The opportunity to focus prevention intervention resources to those at greatest risk of subsequent HIV transmission could result in more efficient and effective use of these limited resources. Generalized use of these data within a transmission network is expected to reduce HIV transmission (6) to a greater degree than delivery of these same interventions provided at random (i.e., guided by traditional metrics of risk for disease progression and behavioral risk).