From trials to programmes: Lessons learned from four trials of Universal Testing and Treatment (UTT) in Sub-Saharan Africa 24th July 2018 Estimates of coverage against the 90:90:90 UNAIDS targets: Comparison of methods and findings Kalpana Sabapathy London School of Hygiene and Tropical Medicine, HPTN 071 (PopART) trial On behalf of the Universal Test and Treat Trials Consortium (UT3C)

Overview of presentation Methods UNAIDS 90-90-90 targets Measuring impact of trial interventions on coverage over time Intervention communities of 4 UTT community randomised trials Baseline pre-intervention, and after ~2 years of intervention Findings Compare and contrast outcomes Limitations and strengths of coverage estimates Conclusions

UNAIDS 90-90-90 targets 73% VS among all PLWH 1st 2nd 3rd Aimed at achieving high coverage at population level Cross-sectional measure of coverage ‘Snapshot in time’ Population-wide Universal Test & Treat Trials Coverage has to be high to achieve the ultimate goal of population level HIV-incidence reduction 90-90-90 targets a way to measure the impact of UTT interventions on coverage over time

“…% of all people…” All people living in a community All people living with HIV (PLWH) 90% PLWH diagnosed and know status 90% diagnosed PLWH on ART 90% diagnosed PLWH on ART & virally suppressed

All adults living in the community Ideally… Identify all households and household members Reality of delivering intervention… Meet face to face Agree to participate in intervention Multiple time-points! Population flux Deaths Out-migration In-migration

Identifying the target population of adults BCPP 15 communities* PopART 4 communities** SEARCH 16 communities TasP 5 communities*** Population enumeration – baseline Community-wide enumeration of HHs Population enumeration – ~2y Baseline enumeration data which mapped all households updated using Google maps to estimate population growth Annual community-wide re-enumeration of HH and individuals during intervention delivery Updated data for in- and out-migration and deaths using community-wide re-enumeration at y3 Six monthly community-wide re-enumeration of HH and individuals during intervention delivery (Community wide enumeration was done at end of study in 3 of the intervention communities – data not shown today ) * BCPP – coverage estimates were restricted to Botswana citizens; primary by-arm coverage estimates will be derived from prospective cohort and end of study surveys in both arms ** PopART – coverage estimates shown are from 4 Arm A, full UTT intervention communities in Zambia pending data from SA ***TasP – 6 other intervention communities with ≤ 2y follow-up are not included here

Adults reached by intervention Total population BCPP PopART SEARCH TasP Proportion of adult residents reached by the intervention among all estimated eligible adult residents in community – baseline 66% (38,000/ 58,000) 80% (101,102/ 125,912)* 92% (73,310/ 79,818) 86% (4,024/ 4,681) Proportion of adult residents reached by the intervention among all estimated eligible adult residents in community – ~2y >100%! (61,645 observed/ 58,000 estimated at baseline 69% (88,098/ 127,438)* 77% (73,952/ 95,599) 95% (4,587/4,828) * PopART age eligibility changed from ≥18y at baseline to ≥15y by 2y

Extrapolating to total adult population BCPP PopART SEARCH TasP Key assumption for extrapolation to total population HIV status and steps of cascade among those not met during intervention delivery is the same as among those who were met prior to intervention using age and sex standardisation HIV status and steps of cascade* among those not met during intervention delivery is the same as among those who were met prior to intervention - after stratifying on a number of key characteristics - after controlling for a number of key characteristics Estimates are focused on those identified by the intervention as HIV-positive. Not extrapolated to total population in data shown here. Statistical methods Baseline HIV prevalence standardised according to age and gender distribution by community in Botswana Census 2011 Stratum specific estimates (based on key characteristics) summarised into summary estimate Targeted maximum likelihood estimation with Super Learner Sensitivity analysis (not presented here) include imputation of HIV status among those with no observed HIV-status information using mathematical modelling *data on viral suppression not currently available

First-90 All people living in a community All people living with HIV (PLWH) 90% PLWH diagnosed and know status

Approaches for identifying 1st 90: Knowledge of HIV-positive status and HIV-status BCPP PopART SEARCH TasP Blood sample (independent of uptake of HIV rapid testing) No Yes – dried blood spots for ELISA HIV rapid diagnostic testing Yes Public health record of receiving HIV care Self-report of HIV+ve status Yes – with documentation Self-report of HIV-ve status Yes, if documented test result in last 3 months Yes, if reported testing in last 3m Data from baseline intervention delivery for estimates at latter time-point

Second -90 All people living in a community All people living with HIV (PLWH) 90% PLWH diagnosed and know status 90% diagnosed PLWH on ART

Approaches for identifying 2nd 90 ART among those with knowledge of HIV-positive status BCPP PopART SEARCH TasP Public health record of receiving ART Yes No Yes – from both trial & government run clinics Self-reported ART Yes - with documentation Yes - (with documentation) Self-reported HIV+ve Yes, with documentation Data from baseline intervention delivery for estimates at latter time-points

Third -90 All people living in a community All people living with HIV (PLWH) 90% PLWH diagnosed and know status 90% diagnosed PLWH on ART 90% diagnosed PLWH on ART & virally suppressed

Approaches for identifying 3rd 90: Viral suppression among those on ART BCPP (<400c/ml) PopART SEARCH (<500c/ml) TasP (<400c/ml) Blood sample No Yes Public health record of viral load done for HIV care monitoring Yes – from both trial & government run clinics Public health record of receiving ART Self-reported ART Yes - with documentation *PopART data on viral suppression not currently available

Outcomes achieved after ~2 years of trial interventions

BCPP 68% VS % % n = 11207 N = 13705 n = 9610 N = 11207 n = 8796 N = 9050* n = 13705 N = 15093 n = 13051 N = 13705 n = 11803 N = 12087* *Known to have started ART and alive at end of second intervention round of delivery and VL result available

PopART 45% ART coverage n = 9648 N = 17686 n = 7805 N = 9648 n = 14929

SEARCH 42% VS % % n = 3484 N = 4149 n = 7988 N = 8408 n = 7287 N = 7988 n = 6392 N = 7287 n = 5136 N = 8305 n = 4149 N = 5316 These are interim results from intervention arm at baseline and y2. Full study results (y3, Intervention vs. Control) will be presented Wed 11 (WEAX01)

TasP 24% VS n =914 N = 1159 n=352 N = 914 n=275 N = 352 n=1268

Conclusions Challenging to get accurate representation of coverage and estimate 90-90-90 at population level to measure the impact of interventions Key unknown “behind the scenes” is population flux Estimates from all trials indicate high coverage is achievable 90-90-90 targets not originally part of the aims of the trial but useful way of summarising coverage

Thanks and Acknowledgements BCCP PopART SEARCH TasP Pam Bachanas Sian Floyd Laura Balzer Collins Iwuji Kara Bennett Richard Hayes Diane Havlir Joseph Larmarange Lisa Block Maya Petersen Joanna Orne- Gliemann Shahin Lockman Study participants and community members Funding organizations, sponsors, Ministries of Health, collaborating partners and all study staff Richard Hayes and Francois Dabis for leading the UT3C