AIDS 2016: Large Test-and-Treat Study Fails to Show Impact on New HIV Infections


The first major research study of "test and treat" as a public health intervention to report its final results -- ANRS 12249 -- has found that the strategy failed to reduce new HIV infections in the African communities where it was provided, according to a report at the 21st International AIDS Conference (AIDS 2016) last week in Durban.

[Produced in collaboration with]

François Dabis from the University of Bordeaux said data collection for the ANRS 12249 study was only completed a month ago, therefore his team has not yet been able to dig deep into the data to explain the findings. But it is already apparent that many individuals diagnosed with HIV did not link with medical care or took many months to do so. Only 49% of diagnosed individuals took antiretroviral therapy (ART).

The study was far more successful in terms of bringing HIV testing to people who needed it -- 92% of people with HIV knew their status. And treatment was highly effective among those who took it, with 93% having an undetectable viral load. These results were achieved in a rural, poor area of KwaZulu Natal, South Africa.

Therefore, in terms of the UNAIDS 90-90-90 targets, the study achieved 92% diagnosed, 49% on treatment, and 93% with viral suppression. The weakness at the crucial middle stage may explain the lack of impact on new HIV infections. But the reasons for the poor linkage to care still need to be unpacked.

Study Aim

The ANRS 12249 study aimed to find out what is the effectiveness of an HIV treatment as prevention (TasP) intervention at the population level in an African community profoundly burdened by HIV.

Whereas studies such as HPTN 052 and PARTNER have examined the impact of HIV treatment on individuals and couples, this is the first of 5 massive randomized studies to examine the populationimpact of scaling up TasP or "universal test and treat" in African countries.

The study aimed to see if the approach was feasible to deliver and acceptable to local communities, but its primary aim was to reduce the number of new infections. Therefore the primary outcome was HIV incidence, as measured in the general population.

The study was conducted in the Hlabisa sub-district of KwaZulu-Natal. Located a 3-hour drive from Durban, 3 in 10 people there are living with HIV -- the highest prevalence in South Africa and one of the highest in the world. Previous observational studies in the area showed that in neighborhoods in which more people with HIV were taking antiretrovirals, new infections were lower. But would a further scale-up of treatment be feasible and improve results further?

Study Design and Interventions

This was a cluster-randomized controlled trial in which the unit of randomization were geographic areas (clusters) of around 1000 residents each. There were 22 of these clusters in the area studied. They were divided into 2 groups -- 11 intervention clusters, and 11 control clusters.

Every 6 months, the entire population in both groups was offered home-based HIV testing and counseling. This involved trained counselors going door-to-door and offering HIV testing in people’s homes. Household members over the age of 16 were offered rapid testing and counseling individually, in a confidential space within the household.

This approach, already used successfully in the region, helps overcome some of the barriers to access to testing services. It can reach individuals who are harder to engage with other approaches, including adolescents, rural residents, and those with limited access to the formal healthcare system.

Subsequently, people diagnosed with HIV in the intervention clusters were offered immediate HIV treatment, regardless of symptoms or CD4 cell count. In the control clusters, diagnosed individuals received HIV treatment in line with national South African guidelines.

When the trial began in 2012, there was a clear difference between the intervention’s test-and-treat approach and the national guidelines, which recommended treatment for those with a CD4 cell count below 350 cells/mm3.  However, that level was changed in January 2015 to 500 cells/mm3. Guidelines are about to change again, to recommend treatment for all -- as in the intervention arm.

It’s worth stressing that the only difference between the intervention and control groups was in terms of eligibility for HIV treatment. Otherwise, there were few differences between the arms: the door-to-door testing happened in both; HIV treatment clinics (within a 45 minute walk of all homes) were provided in both.

So although the researchers describe the strategy tested as one of "universal test and treat," any difference that could have been observed between the intervention and control groups should only be due only to the change in treatment eligibility. Unlike some other test-and treat studies, the intervention arm did not include more intense provision of HIV testing or additional changes to make medical services easier to use.


There were 28,153 people in the study population at the beginning of the trial. Fewer men (37%) than women were enrolled. Most people reached were between 22 and 50 years old, with a median age of 30. As expected, 31% of people were already living with HIV. Just 34% of them were taking HIV treatment.

The door-to-door testing program was able to provide HIV testing, at least once, for 88% of people contacted. And in each round, over 70% of those contacted, accepted the offer of repeat testing.

But when individuals were diagnosed, linkage to care was poor:

Those figures were nearly identical for both the control and intervention groups. But as envisaged, far more people in the intervention group (91%) than in the control group (52%) began HIV treatment soon after linkage to care. Almost all people who began treatment achieved an undetectable viral load.

In terms of the 90-90-90 target, both the "first 90" and the "last 90" were surpassed in both intervention and control communities. But results for the "second 90," reflecting the numbers beginning treatment, were poor.

As a result of the problems at the second stage, only 42.4% of people with HIV in intervention communities and 40.2% of people in control communities had an undetectable viral load. (If each 90 had been achieved, 73% of all people with HIV would be undetectable.)

It is also worth noting how similar the results were between the intervention and control groups.

Overall, there were 495 new HIV infections recorded, over 22,434 person years of follow-up. This equates to an annual incidence of 2.21% (i.e. each year, 2 people in 100 newly acquired HIV).

Incidence did not differ between the intervention clusters (2.13%) and the control clusters (2.27%) -- this difference was not statistically significant.

"The universal test-and-treat strategy had no measurable effect on HIV incidence over the course of the trial," Dabis summed up.

However, he stressed two pieces of good news from the trial. First, there was good take-up of the offer of repeated HIV screening at home, with almost all people being tested at least once. Second, the virological response to HIV treatment was excellent in those people who took it.

Explaining the Results

The analysis presented today was preliminary, as data collection was only completed a month ago. Dabis said that further analyses will attempt to gain a better understanding of how results differed between men and women, and for people of different ages.

The researchers will try to clarify the reasons people did not link to care -- does the explanation lie in the way health services are provided, individual factors, or community stigma? They will seek to better understand the differences between the profile of individuals reached and not reached by interventions.

During the discussion of the presentation, HPTN 052 investigator Myron Cohen from the University of North Carolina suggested that delays in linking to care could have meant that individuals with recent HIV infection disproportionately contributed to onward transmission. Moreover, it will be important to understand the impact of migration and sexual networks which reach outside the study area, which may contribute to new HIV infections.

Dabis said that, although the study was hypothesized to show an effect of TasP after 4 years of follow-up, it may be that it will take longer to have an impact on incidence, in light of the slowness of linkage to care.

Other delegates suggested that the intervention clusters may not have received a package of interventions that was sufficiently intensive, in comparison with the control clusters. For example, a more intensive approach to help people link to care, such as home-based initiation of HIV treatment, could have had a greater impact.

Sheri Lippman from the University of California, chairing the session, commented that it may take more than technical solutions to deal with the structural barriers to engagement with care that exist. 



C Iwuji, J Orne-Gliemann, E Balestre, F Dabis, et al. The impact of universal test and treat on HIV incidence in a rural South African population: ANRS 12249 TasP trial, 2012-2016. 21st International AIDS Conference. Durban, July 18-22, 2016. Abstract FRAC0105LB.