Spatiotemporal Phylodynamics of Hepatitis C Among People Who Inject Drugs in India

Abstract

Background and aims: Implementing effective interventions for HCV requires a detailed understanding of local transmission dynamics and geospatial spread. Little is known about HCV phylodynamics, particularly among high-burden populations, such as people who inject drugs (PWID).

Approach and results: We used 483 HCV sequences and detailed individual-level data from PWID across four Indian cities. Bayesian phylogeographic analyses were used to evaluate transmission hotspots and geospatial diffusion of the virus. Phylogenetic cluster analysis was performed to infer epidemiologic links and factors associated with clustering. A total of 492 HIV sequences were used to draw comparisons within the same population and, in the case of coinfections, evaluate molecular evidence for shared transmission pathways. Overall, 139/483 (28.8%) of HCV sequences clustered with a median cluster size of 3 individuals. Genetically linked participants with HCV were significantly younger and more likely to be infected with HCV subtype 3b as well as to live and inject close to one another. Phylogenetic evidence suggests likely ongoing HCV infection/reinfection with limited support for shared HIV/HCV transmission pathways. Phylogeographic analyses trace historic HCV spread back to Northeastern India and show diffusion patterns consistent with drug trafficking routes.

Conclusions: This study characterizes HCV phylodynamics among PWID in a low and middle-income country setting. Heterogeneity and recent genetic linkage of HCV across geographically disparate Indian states suggest that targeted interventions could help prevent reimportation of virus through drug trafficking routes.

Fig. 1.. Transmission cluster structure of 139 genetically linked HCV sequences.

Circular nodes denote an individual sample and edges denote a putative link. Nodes are colored by participant study city.

Fig. 2.. Maximum clade credibility (MCC) tree of HCV sequences from Indian PWID (red) and references (blue).

MCC tree was selected from the posterior distribution of a 50 million state Markov Chain Monte Carlo (MCMC) simulation under a structured birth-death model and GTR+Γ₄+I substitution model.

Fig. 3.

Histogram of distance between centroid of self-reported zip code of residence of genetically linked HCV samples versus 5,000 permuted samples drawn randomly with respect to population density in New Delhi.

Fig. 4.. Geospatial diffusion of HCV among PWID across four Indian cities.

Circular polygons depict Markov reward counts such that the size of the polygons around a sampling location is proportional to the number of lineages that maintain that location. Colored lines depict movement into a particular city. Direction is indicated with arrowheads, with eastward movements further depicted by lines with an upward curvature and westward movements depicted by lines with a downward curvature. Only lines with well-supported rates (BF >3) that dominate the diffusion process by Bayesian stochastic search variable selection are displayed.

Fig. 5.. Transmission cluster structure of 109 genetically linked HIV sequences.

Circular nodes denote an individual sample and edges denote a putative link. Nodes are colored by participant study city.