Does antiretroviral treatment reduce HIV transmission among MSM?

A simulated risk model has estimated the effect of antiretroviral treatment on HIV transmission rates among men who have sex with men (MSM). The study, published in the April 15^th edition of Clinical Infectious Diseases, estimated that antiretroviral therapy (ART), if used in all MSM with CD4 cell counts below 350 cells/mm³, might reduce secondary infections by 26% over one decade, assuming no other changes in risk behaviour.

Antiretroviral therapy (ART) may have two opposing effects on HIV transmission. Since ART reduces HIV viral loads in both blood and semen, it could arguably reduce secondary infection rates (infections resulting from people who are on therapy). Conversely, ART might lead to more secondary infections as HIV-positive people on ART remain alive and sexually active longer. A further complication is whether and how much ART affects sexual risk behaviours (i.e., changes in higher-risk activity because people perceive themselves to be less or non-infectious).

To study this issue, researchers from Boston University, Harvard, Yale, and several north-eastern US hospitals statistically modelled the impact of ART on infections resulting from unprotected anal intercourse (UAI) in MSM. The analysis compared two hypothetical cohorts of MSM: one not on treatment, and one in which all men with CD4 cell counts below 350 cells/mm³ were treated.

In the analysis, a single untreated individual was calculated to cause a total of 1.9 new infections after ten years, 2.5 after 20 years, and a total of 2.5 new infections after 30 years (estimated mean numbers). A treated individual caused only slightly fewer cumulative secondary infections: 1.4 after 10 years, 1.8 after 20 years, and 2.3 after 30 years. After 33 years, the treated individuals actually began to cause more secondary infections than the untreated (due to treatment failure and longer survival); the treated group caused 23% more infections than the untreated group over an entire simulated lifetime.

Obviously, this type of model depends greatly on its built-in parameters and assumptions. The main parameters were: viral load, disease progression, frequency and level of risk of sexual behaviour, and the effect of ART.

Estimates: Disease stage, viral load, and infectiousness

The risk of infection from unprotected insertive anal intercourse (UIAI) was estimated based on previous studies in both MSM and heterosexuals. Risk levels for a single act of intercourse were taken to be eight-fold higher than risks for heterosexual vaginal intercourse at comparable viral loads and points during the course of disease, an assumption which would be disputed by other experts, who have argued that the transmission risks for anal and vaginal intercourse are similar.

During chronic infection (i.e., excluding primary infection and late-stage disease), the infection risk from a single act of UIAI was estimated to be:

• 0.08% (for viral loads below 500 copies/mL)
• 0.88% (between 501 and 3000 copies/mL)
• 0.96% (between 3001 and 10000)
• 1.1% (between 10001 and 30000)
• 1.8% (for viral loads over 30000 copies/mL)

The infection risk per act was taken to be 6.5% during primary infection (the first three months after infection, with viral load over 100000 copies/mL), and 3.44% during late-stage infection.

Estimates: Sexual behaviour

Based on findings from the Boston Partners Study, each person in the model was assumed to have between 0.8 and 2.7 acts of UIAI per month: e.g., an average of 1.87 acts was assumed per person per month during primary infection. Other frequencies were assigned based on the Boston data, according to CD4 count. Infection rates were calculated as though each sexual act took place with a different partner.

What could affect the results?

In a “sensitivity analysis” component of the study, the various parameters were adjusted to gauge how much they affected the results generated. Rates of treatment failure did not have a large effect. Neither did it make much difference whether risky sexual behaviour in fact varied with CD4 cell count.

There is preliminary evidence that ART may reduce viral load to lower levels in semen than in plasma. If widely true, this would make treatment more effective at preventing transmission.

Estimates of transmission risk affected both groups roughly equally: increasing the per-sexual-act risk estimates by 25% increased the ten-year infection rates by 21% in both groups.

The model’s greatest sensitivity was the assumption that treatment did not affect risky behaviour. Results assumed that UIAI rates were the same in treated and untreated groups. However, if risky behaviour doubled in men for whom treatment was available (regardless of whether they actually started), this led to an average of 2.8 new infections per treated person over the first ten years – twice as many as if no change in behaviour was assumed, and 1.5 times more than the untreated men.

The report also acknowledges that “averaged” sexual behaviour, as modelled here, underestimates real-world infection rates, “because highly sexually active individuals are more likely to become infected and to transmit infection.”

Previous published models have suggested that “widespread ART use could eradicate the HIV epidemic in … MSM … in San Francisco if the incidence of unsafe sexual behaviour remained stable.” (Emphasis added.) However, the authors of the current study state that their results, “although still predicting that treatment would reduce infectiousness over the first ten years, are less optimistic … regarding the likelihood of HIV eradication.” The findings of this study indicate that “ART alone will not eradicate the HIV epidemic.”

Estimates: Effects of antiretroviral treatment

Two main scenarios were compared: one in which everyone with CD4 cell counts less than 350 cells/mm³ was treated, versus one in which no-one received ART. Since therapy changes over time, analysis focused on the first ten years after infection, but was also extrapolated out to “lifetime” time scales. Untreated individuals were taken to survive a mean of 145.4 months (just over twelve years); treated individuals 366.7 months (30.6 years).