The risk of HIV transmission during anal intercourse may be around 18 times greater than during vaginal intercourse, according to the results of a meta-analysis published online ahead of print in the International Journal of Epidemiology.
Moreover, as well as this empirical work, the researchers from Imperial College and the London School of Hygiene and Tropical Medicine carried out a modelling exercise to estimate the impact that HIV treatment has on infectiousness during anal intercourse. They estimate that the risk of transmission from a man with suppressed viral load may be reduced by as much as 99.9%.
Anal intercourse drives the HIV epidemic amongst gay and bisexual men. Moreover a substantial proportion of heterosexuals have anal sex but tend to use condoms less frequently than for vaginal sex, and this may contribute to heterosexual epidemics in sub-Saharan Africa and elsewhere.
Rebecca Baggaley and colleagues conducted a systematic review and meta-analysis (an analysis of all the medical research that meets predefined requirements) of the risk of HIV transmission during unprotected anal intercourse. The same authors have already conducted similar reviews of the transmission risk during vaginal sex and oral sex.
Despite the importance of the topic, only 16 studies were judged to be relevant enough to include in the review. While 12 were conducted with gay or bisexual men, others collected data on heterosexuals who frequently had anal intercourse. All studies were from Europe or North America.
Although the researchers looked for studies published up to September 2008, almost all the reports used data that were collected in the 1980s or early 1990s, which means that the findings do not reflect combination therapy’s impact on transmission. The researchers were not able to include a study with Australian gay men, published a few months ago.
Estimate of the per-act transmission risk
Four studies provided estimates of the transmission risk for a single act of unprotected receptive anal intercourse. Pooling their data, the summary estimate is 1.4% (95% CI, 0.3 to 3.2).
Two of these studies were conducted with gay men and two with heterosexuals, and the results did not vary by sexuality.
The estimate for receptive anal intercourse is almost identical to that in the recently published Australian study (1.43%, 95% CI, 0.48 to 2.85). This is despite the fact that the Australian data were collected after the widespread introduction of combination therapy.
The review did not identify any per-act estimates of the risk for the insertive partner. However, the recent Australian study did produce estimates of this: 0.62% for men who are not circumcised, and 0.11% for men who are circumcised.
Baggaley and colleagues note that their estimate for receptive intercourse is considerably higher than the estimates they produced in their previous reviews. In developed country studies, the risk of transmission during vaginal intercourse was estimated to be 0.08%, whereas the receptive anal intercourse estimate is 18 times greater. For oral sex a range of estimates exist, but none are higher than 0.04%.
Estimate of the per-partner transmission risk
Twelve studies provided estimates of the transmission risk during the whole time in which a person with HIV is in a relationship with an HIV-negative person. The authors note that most of these studies did not collect enough information on factors such as length of the relationship, frequency of unprotected sex and condom use to fully make sense of the data.
Ten of these studies were conducted with gay men only.
For partners having both unprotected receptive and insertive intercourse, the summary estimate of transmission risk is 39.9% (95% CI, 22.5 to 57.4).
For partners having only unprotected receptive intercourse, the summary estimate was almost the same, at 40.4% (95% CI, 6.0 to 74.9).
However, it was lower for people only having unprotected insertive intercourse: 21.7% (95% CI, 0.2 to 43.3). The authors comment that the data support the hypothesis that insertive intercourse is substantially less risky than receptive intercourse.
The individual studies that these estimates are based on often had very different results, in part due to different study designs and analytical methods. As a result, the confidence intervals for these pooled estimates are wide and the authors recommend that their figures should be interpreted with caution. (A 95% confidence interval gives a range of figures: it is thought that the ‘true’ result is likely to be within the range, but could be as high or as low as the extra figures given.)
Moreover, the researchers note that the per-act estimates do not appear to be consistent with the per-partner estimates. Their results would imply that there were relatively few instances of unprotected sex during the relationships studied.
The authors believe that some of this discrepancy could reflect variations in infectiousness and susceptibility to infection between individuals, and in infectiousness over the duration of an infection.
The impact of HIV treatment on transmission risk
As previously noted, almost all the studies come from the pre-HAART era. The investigators therefore carried out mathematical modelling work to estimate reductions in the transmission risk in individuals with a suppressed viral load.
To do this they used two different calculations for the relationship between viral load and transmission, derived from studies with heterosexuals in Uganda and Zambia.
The first calculation has been widely used by other researchers. In it, each log increase in viral load is assumed to increase transmission 2.45-fold. While this 2.45-fold relationship is thought to be accurate for viral loads between 400 and 10,000 copies/ml, Baggaley and colleagues believe that it overestimates transmission both at lower and higher viral loads.
The second, more complex, calculation reflects transmission being extremely rare at low viral loads and also transmission rates being pretty constant at higher viral loads.
Using the first method, the HIV transmission risk for unprotected receptive anal intercourse is 0.06%, which is 96% lower than without treatment. However using the second method, the predicted transmission risk would be 0.0011%, which is 99.9% lower than without treatment.
Extrapolating from these figures, the authors calculated the risk of HIV transmission in a relationship involving 1000 acts of unprotected receptive anal intercourse. Using the first method, the risk would be 45.6% and using the second method it would be 1.1%.
The authors note that very different predictions were obtained when two different sets of assumptions about viral load were used. In the debate on the use of HIV treatment for prevention they comment that “modelling cannot be a substitute for empirical evidence”.
Moreover, in a commentary on the article, Andrew Grulich and Iryna Zablotska of the University of New South Wales note the lack of data on viral load and transmission during anal sex (all the studies relate to heterosexual populations). They say that the fact that per-act estimates of transmission risks are so much higher during anal sex than during vaginal sex “is a strong argument for not simply extrapolating data from heterosexual populations.”
Baggaley and colleagues say that their findings suggest that the high infectiousness of anal intercourse means that even if treatment leads to a substantial reduction in infectiousness, “the residual infectiousness could still present a high risk to partners”. Given this, they say that prevention messages need to emphasise the high risk associated with anal sex and the importance of condoms.
Baggaley RF et al. HIV transmission risk through anal intercourse: systematic review, meta-analysis and implications for HIV prevention. Int J Epidemiol (online edition), doi:10.1093/ije/dyq057
Grulich AE and Zablotska I. Commentary: Probability of HIV transmission through anal intercourse. Int J Epidemiol (online edition), doi:10.1093/ije/dyq101