The risk of HIV reinfection, sometimes called superinfection, has not been clearly established. Given the risk of initial sexual HIV infection from a single exposure in population studies is commonly calculated as generally low (although these estimates are dependent on the background prevalence) this is difficult to assess in small studies.
Genetic analysis is specialised and expensive and the biological evidence is strongly supported by the many viral sub-clades and recombinant forms that can only have occurred in vivo. The extensive global viral diversification provides the most practical evidence for reinfection.
Case studies have highlighted cases where reinfection with drug resistant HIV has clinical implications leading to either more rapid disease progression or treatment failure and reduced future treatment options. Although initial reports were in early or acute initial infection, cases have also included people in chronic infection, indicating that this may not be restricted by immune responses to the initial infection and in people on suppressed ART, indicating that the pressure from PrEP/PEP can also be overcome.
Three studies at CROI focused on issues of reinfection.
Andrew Redd and colleagues from the US NIH and Johns Hopkins University used high-throughput deep sequencing to retrospectively test for HIV superinfection in two regions of the viral genome (p24 and gp41) in 203 people from the Rakai Community Cohort Study (RCCS) in Uganda, who seroconverted from 1997-2002 with later samples for a total 15,000 person years of follow up (PYFU). This was compared to the primary HIV incidence rate for the HIV negative general heterosexual population in Rakai (n = 20,220; > 100,000 PY).
They identified reinfection in 7/149 people with identifiable sequences in the seroconverter cohort, 1.44/100 PY (95%CI 0.37 to 2.51), all from significant changes in the gp41 region. These seven cases were initially infected with sub-type D. Four of the reinfections were with new sub-type D and three with sub-type A.
There were 1152 new infections in the general population over the same period giving an incidence rate of 1.15/100 PY (95%CI 1.08 to 1.21). This was not significantly different to from the primary HIV incidence rate (incidence rate ratio = 1.26, 95%CI 0.50 to 2.60; p = 0.26).
Differences between the risk factors for the people with reinfection (inherently at greater risk than the general population because they were clearly more susceptible to infection) were adjusted for using propensity score matching increased the background incidence rate to 3.28 /100 PYFU (95%CI 2.0-5.3) based on the baseline data but this reduced to 2.51 /100 PYFU (95%CI 1.5-4.3) using the follow up time point (when the difference between groups were more narrow).
The authors concluded: "Although other studies have examined superinfection in small groups of high-risk individuals, this is the first study to directly compare HIV superinfection rates to HIV incidence in a general heterosexual population. The finding that HIV superinfection occurs at approximately the same rate as primary HIV incidence has multiple public health ramifications, and could have significant implications for HIV vaccine research."
In a second oral presentation, Keshet Ronen from Fred Hutchinson Cancer Research Centre and colleagues looked at the incidence of reinfection in a high-risk cohort of female sex workers in Mombasa, Kenya who were enrolled within six months of initial infection and followed for two years. This is a cohort of almost 3000 HIV negative women who have been followed prospectively with monthly visits, 311 of whom have seroconverted since 1993, with median follow-up of five years.
This group amplified and analysed ~500 bp amplicons in gag, pol, and env from plasma RNA to identify cases of multiple infection. Between 100 and 2000 sequences were obtained per genomic region per time point for a total of ~380,000 sequences.
In earlier studies this group identified 12 cases of reinfection in 56 women. In this new analysis a further 94 women have been identified, with 7 cases of reinfection in the 63 women who have so far had data analysed. They presented combined result of 19 cases of superinfection among 117 women over 621 person-years of follow up (incidence of 3.06% PYFU for reinfection vs to 3.25 for initial infection) and ongoing screening and analysis continues to reach the 150 cases needed to be powered to compare these rates, adjusting for other risks. In this study, timing of reinfection was addressed and included cases plausibly occurring five or more years after initial infection.
However, some researchers suggest the possibility that cases attributed to reinfection could come from initial dual infections, with one infection outgrowing the other after several years. In the absence of being able to confirm a reinfection event by phylogenetic comparisons to the second donor an indirect way to rule out initial dual infection would be to look for closest ancestor for each dual strain to estimate whether one infection has been present for longer than the other.
A poster from the UCSF group that have previously presented this position included two cases where reinfection (superinfection) was initially described based on limited sample sequencing but that more sensitive analyses suggested were serially expressed dual infections (SEDI). 
The consensus after both studies seemed to be that initial HIV infection is not protective of subsequent infections. Researchers were divided over whether initial infection potentially increases the risk of second infection or whether longer duration of follow-up (>2 years) might uncover CTL responses.
Others suggested that cases of reinfection in these studies could easily have been underestimated by not looking early enough after initial infection and only reporting phylogenetically different infections. Further research is needed to determine risks for reinfection, currently assumed to generally be similar to those for initial infection (behaviour risk, viral load of the transmitting partner, STIs, genetics etc).
Ascribing reinfection to initial dual infection (SEDI), requires either one source partner (prompting the question of how this person became dually infected?) or exposures from multiple sources at a close time point, which becomes practically very close to dual infections as one infection must have predated the other, even if this involded a short window period.
Unless stated otherwise, all references are to the Programme and Abstracts for the 19th Conference of Retroviruses and Opportunistic Infections, 58 March 2012, Seattle.
Links to external websites are current at time of posting but not maintained.