Since treatment has prevention benefits at the individual level, it makes sense to think it may also have prevention benefits at the population level. In other words, increasing the number of HIV-positive people on treatment could lower the total amount of virus circulating in a community (also known as community viral load) and lead to a reduction in the number of new HIV infections.

This would appear to have occurred in San Francisco, where the epidemic is concentrated in men who have sex with men. Between 2004 and 2008, the number of men who had recently tested went up, rates of undiagnosed infection went down and the proportion of diagnosed men with an undetectable viral load rose from 45 to 78%. During the same period of time, annual new diagnoses fell by half.

Similarly, in British Columbia (Canada) between 1996 and 2012, the number of people taking treatment increased eight-fold, while the number of annual diagnoses was reduced by two thirds. It's estimated that for every 1% increase in the number of individuals taking treatment with a viral load below 500 copies/ml, new HIV infections (incidence) dropped by 1%.

The greatest change has been seen in people who inject drugs (92% fewer diagnoses) and this can be partly attributed to an expansion of harm-reduction services during the same period of time – consistent with this, infections with hepatitis C have also fallen. But benefit has also been seen in men who have sex with men (22% fewer diagnoses) and this does not appear to be due to increased condom use – rates of syphilis, gonorrhoea, and chlamydia have risen during the same period of time.

A sophisticated analysis, which took into account numerous factors that influence the risk of acquiring HIV, comes from KwaZulu-Natal, South Africa. During a seven-year period in which access to HIV treatment expanded rapidly but unevenly, researchers followed nearly 17,000 individuals who were initially HIV negative. The analysis took into account the prevalence of diagnosed HIV and access to HIV treatment in each person’s local area.

For every 1% increase in HIV treatment coverage among people with diagnosed HIV in the local community, the risk of HIV infection decreased by 1.4%. A person living in an area in which 30 to 40% of HIV-positive people were taking treatment was 38% less likely to acquire HIV than a person living in an area in which fewer than 10% of people were on treatment.

A number of community randomised trials are underway in African countries to assess the population impact of treatment as prevention policies.

However, infections and diagnoses have not fallen in gay men in the UK, or in heterosexuals infected in the UK, despite even higher levels of HIV treatment coverage and viral suppression than has been achieved in San Francisco. It is estimated that approximately 2400 gay men are newly infected each year, although around nine-in-ten men with diagnosed HIV and a CD4 count below 350 cells/mm³ are taking treatment and most of them have an undetectable viral load.

Likewise, despite the widespread use of HIV treatment in Australia and the Netherlands, these countries have not seen falls in the number of infections in men who have sex with men.

This does not mean that HIV treatment is not having any benefit at all. Mathematical modelling suggests that the rate of infections would be even higher if treatment had not been provided since 1996. But it is clear that HIV treatment is not – on its own – enough to curtail the epidemic in men who have sex with men in the UK.

This is partly because of a combination of low rates of HIV testing, high rates of partner change and inconsistent use of condoms – men who have undiagnosed HIV (especially undiagnosed acute infection) are the source of most new infections. It is also possible that sexually transmitted infections curtail the preventive benefits of treatment, or that treatment is not as effective in reducing transmission during anal intercourse as during vaginal intercourse.