Background:  Molecular analysis of HIV transmission clusters can help elucidate risk factors for transmission at the network level and help locate at-risk individuals. Applying social network analysis approaches, we examined correlates of transmission clusters among men who have sex with men (MSM) with recent HIV infection in San Diego.

Methods:  Between June 1996 and June 2006, we recruited 285 MSM with acute (n = 83) and early (n = 202) subtype B HIV infection. Sequences of the 99 codons of protease and the first 305 codons of reverse transcriptase were compared phylogenetically using HyPhy to determine if MSM belonged to a transmission cluster, defined as <1% nucleotide divergence between 2 individual’s viruses. Exponential random graph models were fitted to the cluster matrix using the ergm function in the R statnet library, using likelihood ratio tests to determine factors associated with clustering between individuals.

Results:  The mean sequence divergence between individuals was 5.6%; 90 individuals (32%) belonged to a transmission cluster, 60 of whom were in a cluster of 3 or more individuals. There was no significant change in clustering over time (odds ratio [OR] 1.02, p = 0.15), although individuals sampled further apart in time were less likely to cluster (OR 0.6/year, p <0.01). After controlling for sampling time, individuals were more likely to cluster together if they were discordant by age (OR 0.96/year, p <0.01) or had drug resistant virus based on the Stanford algorithm (OR 3.0, p <0.01), and there was a trend for individuals with acute infection to cluster (OR 1.5, p = 0.08). There was no significant effect of age or number of sex partners in the previous 3 months (p >0.1).

Conclusions:  Previous studies have considered correlates of belonging to a cluster, rather than belonging to the same cluster, causing potential problems of ecological fallacy. We overcame this by treating a pair of individuals with similar HIV strains as the outcome; this also allowed us to address the role of discordance in clustering. Clusters of HIV sequences were significantly more heterogeneous in age, more likely to contain resistant virus, and consisted of more acute infections than expected. This study suggests that age discordance may increase the risk of HIV transmission; that there may be sub-epidemics with higher levels of drug resistant virus; and that people with acute infection are more likely to transmit HIV to others.