Abstract Body

Identifying risk factors and other epidemiological correlates of HIV transmission can inform the prioritization of health care services to specific groups. Typically transmission clusters are inferred based on genetic distance thresholds, then logistic regression is conducted to evaluate patient characteristics that are significantly associated with the probability of membership in an active transmission cluster. This method is limited in that all individuals within a cluster are treated as equally active transmitters, although in reality there is a range of transmission activity within a cluster. We introduce an alternative method to investigate risk factors associated with transmission in British Columbia (BC), Canada, based on the phylodynamically estimated viral diversification rate.

For 8,103 people living with HIV (PLHIV) in BC in March 2018, we recovered the oldest available HIV protease and RT sequences from the BC Centre for Excellence in HIV/AIDS database. Following alignment and removal of known drug resistance sites, we inferred 100 bootstrap approximate maximum likelihood phylogenetic trees in FastTree2.1 and then time-scaled the trees using Least Squares Dating. For each tip on each bootstrap tree, we calculated the lineage level phylogenetic diversification, which provides a proxy for transmission rates. The average diversification rate of all 100 trees for each tip was taken. We then built a generalized linear model (GLM) to evaluate patient attributes that were significantly associated with higher diversification rates.

Having a high HIV diversification rate was positively associated with being younger, reporting injection drug use, having co-infection with hepatitis C virus, having a high most recent viral load, and residing within the Northern BC Health authority or the Vancouver Coastal Health authority (Table 1). Interestingly, having ever had AIDS and identifying as black were both significantly associated with lower diversification rates (Table 1).

By identifying risk factors associated with HIV transmission using the viral diversification rate among PLHIV in BC, we can confidently recommend prioritized provision of treatment and prevention services for these key groups. Relative to logistic regression of phylogenetic cluster membership, this method has the added benefit of resolving finer differences in transmission activity between individuals, resulting in a more accurate assessment of risk.