1. Temporal trends of population viral suppression in the context of Universal Test and Treat: results from the ANRS TasP trial in rural South Africa Joseph Larmarange, Mamadou Hassimiou Diallo, Nuala McGrath, Collins Iwuji, Mélanie Plazy, Rodolphe Thiébaut, Frank Tanser, Till Bärninghausen, Joanna Orne-Gliemann, Deenan Pillay, François Dabis for the ANRS TasP Study Group
ANRS TasP trial
Good morning everyone. In this presentation, I will explore the temporal trends of population viral suppression in the context of Universal Test and Treat.
TUAC0103 • 24 July May 2018

2. Scientific Context  ART coverage (% of PLWHIV on ART)
Universal Test and Treat (UTT) aims to maximize PLWHIV on ART and virally suppressed in a community.
According to mathematical modelling, UTT would lead to reduction in HIV incidence.
 ART coverage
(% of PLWHIV on ART)
 Population Viral Suppression
(% being virally suppressed)
The universal test-and-treat strategy or UTT aims to maximize the proportion of all people living with HIV on antiretroviral treatment and virally suppressed in a community, i.e. to increase population viral suppression.
CLIC
According to mathematical modelling, UTT would lead to reduction in HIV incidence and potential elimination of the HIV epidemic in Southern Africa.
 HIV incidence
(new infections at population level)

3. The ANRS TasP trial
One of 5 international trials aiming at evaluating UTT approaches
Design: cluster-randomised trial
Timeline: March 2012-June 2016
Study setting: Hlabisa sub-district
~ individuals aged 16+
isiZulu speaking
HIV prevalence ~30%
frequent migration
low marital rates & late marriage
only 10% are employed
The ANRS Treatment as Prevention trial is one of 5 internationals trials aiming at evaluating UTT approaches.
It was a phased two-arm cluster-randomised trial implemented between March 2012 and June 2016 in Hlabisa sub-district, northeast KwaZulu-Natal, South Africa, in a rural area with approximately 28,000 isiZulu-speaking resident adults.
Adult HIV prevalence in the sub-district was around 30%. Hlabisa sub-district is characterized by frequent migration, low marital rates, and late marriage. On average only one adult in ten in the trial area is employed.
Hlabisa sub-district
South Africa
KwaZulu Natal

4. repeated every ~six months referred to trial clinic
TasP trial procedures
Homestead Identification
Homestead visit
Registration of resident adults
Update of resident members list
Exit forms
Homestead procedures
Individual questionnaires
DBS sample (lab tests)
Rapid HIV testing
In both trial arms, HIV counsellors visited all local households and enumerated all resident adult household members.
CLIC
Eligible individuals providing written informed consent responded to a socio-demographic and sexual behaviour questionnaire and gave a finger prick sample collected as a dried blood spot, used for HIV incidence estimation. HIV counsellors also offered individuals point-of-care rapid HIV counselling and testing.
Home-based survey rounds were repeated approximatively every six months. The list of resident household members was updated and exits (including deaths and out-migration from trial area) were documented.
All trial participants identified as HIV-positive (through rapid HIV test or self-report) were referred to a local trial clinic set up by the trial and situated in the trial cluster in which they lived, located at less than 45 minutes walking distance.
In the trial clinics of the control clusters, HIV-positive adults were offered ART according to national guidelines. In the trial clinics of the intervention clusters, all HIV-positive adults were offered the opportunity to begin ART immediately regardless of CD4 count or clinical staging.
The trial area was also served by three local governmental primary care clinics of the department of health providing HIV testing, care and treatment according to national guidelines only. HIV-positive participants of both arms could opt to receive HIV care in primary care clinics or transfer to a trial clinic. With the authorization of the ethical committee, we were able to link individual level data from the local governmental clinics with our trial data.
repeated every ~six months
Local governmental clinics
Matching between trial and governmental database at individual level
CD4 and viral load results / clinic visits
ART according to national guidelines
Trial clinics
Intervention arm: immediate ART
Control arm: ART according to national guidelines
if ascertained HIV+
(rapid test or self-report)
referred to trial clinic

5. TIMING of fieldwork 4 clusters (opened in 2012)
Light areas indicate the time required to complete the initial census of the population
The trial was implemented in 22 clusters in 3 steps: 4 clusters opened in 2012, 6 clusters in 2013 and 12 clusters in All clusters were followed until mid Therefore, the number of survey rounds and follow-up time differ per cluster.

6. Previous results Research Question
Main results were presented in Durban in (Iwuji et al. Lancet HIV 2017)
No significant difference in HIV incidence between trial arms
Research Question
Did population viral suppression improve during the course of the trial?
Differentially by arm?
According to trial interventions or contextual changes?
Two years ago, we presented the main trial results in Durban.
We did not observe a significant difference in HIV incidence between trial arms.
CLIC
However, did Population Viral Suppression improve during the course of the trial? Differentially by arm? According to trial interventions or contextual changes, independent of the trial?

7. Approach: Computation of daily statuses
28,419 adult residents were registered
for each calendar day
initial census of the population,
16th birthday, in-migration events, out-migration events and deaths
RESIDENCY status
(resident / not resident)
among those residents
repeat DBS, repeat rapid tests, HIV-positive self-reports and HIV clinic visits seroconversion date imputed (random point approach)
HIV status
(HIV positive / negative)
28,419 adult residents were registered within the trial.
For each individual and each calendar day, we assessed their residency status considering the initial census of the population, 16th birthdays, in-migration events whose dates were imputed by comparing household members lists between survey rounds, out-migration events and deaths documented through specific exit forms.
CLIC
Daily HIV status was estimated using multiple sources: repeat blood samples collected at home, repeat rapid HIV tests, HIV-positive self-reports and HIV clinic visits in trial and local governmental clinics. We assumed a random seroconversion date between the last negative and the first positive observed HIV statuses. Probability of seroconversion by sex and cluster was also used to estimate possible seroconversion prior a positive status or after a negative one.
Clinic visits, ART prescriptions and CD4 counts collected in trial and governmental HIV clinics were used to estimate if HIV-infected individuals were in care and on antiretroviral treatment. Among those on treatment, all collected viral loads were taken into account to assess Viral Suppression, defined as less than 400.
among those HIV-positive
clinic visits, ART prescription, CD4 counts and viral loads
trial clinics and local governmental HIV clinics
HIV CARE position
Viral Suppression (<400)

8. cluster-level Population viral suppression
% being in care, on ART and virally suppressed
Computed at different time points (pre-intervention + daily)
Population Viral Suppression denominator
Resident adults living with HIV
16th birthday
HIV sero-conversion
Death
Out-migration
In-migration
The denominator used for computing Population Viral Suppression at cluster level, i.e. the local adult resident HIV-infected population, was changing over time due to in-migration, 16th birthday, HIV seroconversions, out-migrations and deaths.
CLIC
Population viral suppression corresponds to the proportion of residents adults living with HIV being in care, on ART and virally suppressed.
It was computed at different time points: a pre-intervention estimate, considering the situation of individuals at the initial census of the population performed during the first round of each cluster, plus an estimate per day once the initial census of the population was completed.

9. overall results
At baseline, population viral suppression slightly lower in intervention arm
Significant increase in both arms
A slightly higher increase in intervention arm
No significant difference between arms at the end of the trial
Intervention arm
Control arm
Pre-Intervention
23.5%
diff: -2.5
p=0.028
Pre-Intervention
26.0%
+22.8
p<0.001
diff in diff: +4.2
p=0.013
+18.6
p<0.001
At baseline, population viral suppression was similar between arms but slightly lower in intervention arm.
CLIC
It increased significantly in both arms between pre-intervention and January first 2016: +23 in intervention arm and +19 in control arm.
That increase remained similar between arms, although the increase was slightly better in intervention arm, difference in differences being of +4.
Therefore, at the end of the trial, population viral suppression was not significantly different between arms.
Jan. 1st 2016
46.2%
diff: +1.6
p=0.208
Jan. 1st 2016
44.6%

10. Modelling population viral suppression
Mixed linear model
One record per cluster and per day
Outcome: cluster-level population viral suppression
Factors:
calendar time
time since cluster opening
trial arm
interaction between trial and time since cluster opening
socio-demographic characteristics (cluster-level)
To disaggregate effects due to trial interventions implemented in both arms or intervention arm only and contextual changes, we used a mixed linear model to explore the relation between population viral suppression with calendar time, time since cluster opening, trial arm and interaction between arm and time since cluster opening, adjusting on sociodemographic changes at cluster level and including a random effect on cluster.

11. Model results: Population viral suppression, Tasp ANRS 12249
Contextual changes
Effect of Universal Testing
This graph represents the coefficients estimated by the model.
CLIC
Population viral suppression increase was mainly driven by time since cluster opening, measuring the impact of repeat home-based HIV testing and implementation of local trial clinics, both having been implemented in all clusters.
As already seen, Population Viral Suppression at baseline was lower in the intervention arm.
However, the increase was more important in the intervention arm, measuring the effect of initiating treatment regardless of CD4 count compared to national guidelines. The effect of Universal Treatment was smaller than the effect of Universal Testing, due to low level of linkage to care.
Finally, they were also some effect due to contextual changes, measured by calendar time. In 2015, South Africa changed its treatment initiation guidelines, from 350 to 500 CD4 count.
Baseline difference between trial arms
Effect of Universal ART
The coefficients corresponds to PVS annual change.

12. Limitations
Care received in governmental clinics probably underestimated due to participants not matched between governmental and trial datasets
Care received in private sector or outside the trial area not captured
9.5% of trial population with no observed HIV status and excluded from the analysis
Sensitivity analysis: results unchanged
Our analysis presents some limitations. In particular, population viral suppression is probably underestimated due to the fact that some trial participants receiving care in local governmental clinics were probably not successfully matched between governmental and trial database and therefore wrongly classified as not in care.
In addition, we did not capture HIV care received in private sector or outside the trial area.
Overall, 9.5% of the trial population had no observed HIV status and was excluded from the analysis. It could induce some overestimation as we can hypothesise that these individuals are less likely to receive care. However, we did perform some sensitivity analysis not presented here and results remained unchanged.

13. discussion
Although suboptimal, the TasP strategy significantly improved population viral suppression over time.
Mainly due to universal testing rather than universal treatment
Increase similar between arms  explain the null effect of HIV incidence
Changes in treatment guidelines not enough to increase population viral suppression
Although suboptimal, the UTT strategy implemented in TasP trial improved significantly population viral suppression over time.
As it was mainly due to universal testing rather than universal treatment, it did not induce difference between arms, explaining the null effect observed on cumulative incidence. Unfortunately, the trial was not powered to measure if incidence decreased over time within each arm, but to compare if overall incidence was different between arms.
CLIC
Changes in treatment initiation guidelines alone are not enough to significantly increase population viral suppression if no additional intervention is implemented to improve linkage to care in this rural setting.

14. Acknowledgments ANRS 12249 TasP trial
Trial participants
Africa Centre staff
Traditional Authorities
ANRS Study Group (by alphabetical order):
Kathy Baisley, Eric Balestre, Till Bärnighausen, Sylvie Boyer, Alexandra Calmy, Vincent Calvez, François Dabis (co- PI), Anne Derache, Adama Diallo, Hermann Donfouet, Rosemary Dray-Spira, Jaco Dreyer, Ken Freedberg, Andréa Gosset, Kobus Herbst, John Imrie, Collins Iwuji (Coordinator South), Sophie Karcher, Joseph Larmarange, France Lert, Richard Lessells, Thembisa Makowa, Anne-Geniève Marcelin, Laura March, Kevi Naidu, Colin Newell, Marie- Louise Newell (co-PI), Nuala McGrath, Nonhlanhla Okesola, Tulio de Oliveira, Joanna Orne-Gliemann (Coordinator North), Delphine Perriat, Deenan Pillay (co-PI), Mélanie Plazy, Camélia Protopescu, Bruno Spire, Frank Tanser, Rodolphe Thiébaut, Thierry Tiendrebeogo, Johannes Viljoen, Thembelile Zuma.
Department of Health, South Africa
Merck/Gilead
ANRS TasP trial
Thank you very much