SEARCH (Sustainable Research in Community Health) Thank you NIH and PEPFAR for supporting SEARCH We are grateful for all who work on SEARCH, our multilateral partners, and the communities we serve.

http://www.searchendaids.com NCT 01864603 Co-PIs Diane Havlir, Moses Kamya, Maya Petersen Economist: Harsha Thirumurthy Statistician: Laura Balzer, Mark van der Laan Social Scientist: Carol Camlin Vice-Chair: Edwin Charlebois Modeling: Britta Jewell, Anna Bershteyn Virologist: Teri Liegler KEMRI: Elizabeth Bukusi, Norton Sang, James Ayieko, Kevin Kadede UCSF: Tamara Clark, Gabe Chamie, Ted Ruel, Vivek Jain, Starley Shade, Doug Black, Cait Koss, Lillian Brown, Craig Cohen IDRC-UCSF: Dalsone Kwarisiima, Jane Kabami, Dathan Mirembe, Asiphas Owaraginise, Mucu Atukunda SEARCH Study

SEARCH Hypothesis: HIV “test and treat” with universal ART using a multi-disease, patient- 32 communities, of 10,000 persons each centered care model would reduce new HIV ~320,00 person study infections and improve community health compared to a country guideline approach East N=10 Study Design: Pair-matched, community Uganda West N=12 randomized study of 32 rural communities N=10

Study Population: Age > 15 years • Comprehensive baseline census with biometric identifier Study Interventions

INTERVENTION “CONTROL” COMMUNITIES COMMUNITIES N=16 N=16 HIV and NCD Diagnosis Health Fairs: Baseline + Annual * Health Fairs: Baseline only

ART eligibility All (Universal) Country guidelines adapted over time Patient-centered Care**3 years • “Chronic care” model: HIV and HTN/DM Country standard- of- care Care Delivery • Rapid ART start and VL counseling • Welcoming environment, flexible clinic hours for HIV and HTN/DM • Mobile phone triage and reminders *Multi-disease: HIV, HTN, DM, malaria and other Follow-up testing for non- participants Chamie, Lancet HIV 2016 **Kwarisiima, JIAS 2017 Study Endpoint Measurements Catategory Study Endpoint Measurement HIV cascade HIV testing Confirmed rapid antibody testing coverage HIV cascade ART start Ministry of Health record HIV cascade Viral suppression Plasma HIV RNA at health fair measured with Roche assay, Viral Suppression <500 c/mL Community Mortality Key informant interviews at Year 3; classified as due to illness, childbirth, suicide or Health accident Community Tuberculosis TB Registry at health dispensaries Health Community Hypertension Blood pressure < 140/90 mmHg Health control Community Diabetes control Blood glucose <11 mmol/L at the health fairs Health HIV *Cumulative 3 Rapid HIV antibody testing with Geenius and Western Blot confirmation at Year 3 year HIV incidence health fairs among 117, 114 persons HIV-negative at baseline HIV Annual HIV Confirmed rapid HIV antibody testing at annual health fairs in the intervention arm incidence *primary study endpoint Statistical Methods • Randomization: pair matched on region, population density, number of trading centers, occupation mix and migration • Powered to detect 25-40% reduction in 3 year cumulative HIV incidence • Comparison of outcomes between arms using community-level analysis 1. Calculate community-level outcomes 2. Compare between arms • Targeted maximum likelihood estimation (TMLE) and pre-specified adjustment to improve precision 3. Two-sided hypothesis testing, 5% significance level, pair as the independent unit • Pre-specified sensitivity analyses to look at robustness of findings

Balzer, Stat Med, 2016 Rapid Uptake of National ART Guidelines in Control clinics 2013 Kenya – Baseline to Year 3 2017

ART for all

Intervention Clinics

ART CD4 <350 ART CD4 <500 + ART for all (WHO 2010) (WHO 2013) (WHO 2015) Control Clinics Early implementation of expanded ART eligibility Results: SEARCH Adult Open Cohort

Baseline Enumerated 355,848 persons 174,502 age <15 years 12 missing age

Baseline Age > 15 years 181,334 Non-resident 30,685 Died by start of CHC 254 Baseline Residents 150,395 Turned 15 years old Died 26,858 2,633 Enter cohort Leave cohort In-migrated Out-migrated 25,556 13,822 Year 3 Residents 186,354 Study Population Total N=150,395 Region (HIV prevalence) Kenya (19%) 53,872 Western Uganda (7%) 47,328 Eastern Uganda (4%) 49,195 Male 67,981 (45.2%) Age 15-20 years 36,655 (24%) 21-49 years 84,022 (56%) 50+ years 29,718 (20%) Hypertension (among >30 year olds) 22,599 (20%) Stable (≤6 mo. out of community) 143,870 (96%) Farmer 76,695 (51%) Male Circumcision 20,597 (34%) Proportion ever tested for HIV 100% 90% • Dramatic ramp-up in 80% HIV testing coverage in 70% both arms from 57% 60% 50% pre-SEARCH to 90% 90% 91% 92% 94% 40% after baseline testing 30% 57% 58% 20% • Continued increase in Proportion ever tested ever Proportion 10% cumulative testing over 0% time in intervention Pre-Baseline (self- Baseline Year 1 Year 2 arm report) Study Time

Intervention Control Among all residents ≧ 15 years, including in-migrants and aged-in; Pre-baseline: self-report at time of testing; Baseline-Ye a r 2 : Te sti n g status based on documented HIV test (SEARCH or Ministry record). Control Pre-BL: 65,074, N= BL: N=70,577; Intervention: Pre-BL: N= 72,978 ,BL N=79,818, Y1 N=89,994, Y2 N=93,008 ART start among baseline HIV-infected persons not on ART 90% 80% • More rapid ART start 70% in the intervention vs. 60% control arm 50% • More rapid ART in 40% 80% 73% intervention arm at 30% 60% all CD4+ strata, 20% 40% 30% including CD4<350: 10% Probability of Initiating ART Initiating of Probability 17% 80% intervention vs. 0% 45% control at 1 year 6 months 12 months 24 months (p<0.001) Study Time

Intervention Control Among baseline HIV+ residents not on ART at baseline (N=5,952, 44% of baseline HIV+); Community-level estimates of probability of initiating ART by 6,12, and 24 months based on Kaplan-Meier, censoring at death or outmigration Viral suppression among all HIV-infected persons over time

RR: 1.15 90% (95%CI: 1.11, 1.20 ) p<0.001 • Year 3 suppression 11% 80% 2020 UNAIDS population viral suppression 73% target higher (p<0.001) in 70% intervention vs. control 60% • Year 3 suppression 50% ~10% higher in 40% 76% 79% 71% 68% intervention vs. control 30% in: women, men, youth 20% 42% 42% 10% • Year 3 suppression in intervention: women 0% Proportion HIV+ with Viral Suppression Viral with HIV+ Proportion Baseline Year 1 Year 2 Year 3 (81%), men (74%), youth Study Time (55%)

Intervention Control Among all residents including in-migrants and aged-in, excluding out-migrants and deaths; adjusted for missing HIV serostatus and plasma HIV RNA measures; numerator for each suppression estimate using 4,192–6,800 RNA measures Impact of SEARCH on Community Health: Mortality was lower Among baseline HIV+ • Mortality among HIV-infected RR: 0.72 12% (95%CI: 0.57, 0.91) persons was 21% lower in the p<0.01 10% intervention vs. control arm (p=0.02) 8% RR: 0.79 (95%CI: 0.65, 0.96) • Mortality rate among all persons 6% p=0.02 11% lower in intervention arm RR: 0.97 4% 8% (95%CI: 0.75, 1.27) • RR 0.89 (95%CI 0.79, 1.02) 6% P=0.83 2% 3% 4% Probability of death by Year 3 Year by death of Probability 2% 2% 0% All CD4<350 CD4≥350 Intervention Control Death due to illness among baseline HIV+ stable residents (N=13,066) and all baseline stable residents (N=171,431); community-level estimates of risk by 2.7 years using Kaplan-Meier censored at outmigration or death due to other cause Impact of SEARCH on Community Health: Tuberculosis was lower

HIV-TB/Death among baseline Annual TB incidence rate among HIV+ baseline HIV+ (N=13,066) • HIV-TB or death among 5% 600 RR: 0.41 RR: 0.80 (95%CI: 0.19, 0.86) 4% (95%CI: 0.70, 0.92) HIV+ 20% lower in the 500 p=0.02 4% p=0.004 intervention vs. control arm 400 3% (p=0.004) 3% 300

TB/Death by Year 3 Year by TB/Death • 59% lower TB incidence - 2% 200 4% rate in year 3 among 2% 3% 100 baseline HIV+ in 1% 0 1% intervention vs. control arm TB Incidence Rate (per 100,000 PY) TB Incidence Rate Year 1 Year 2 Year 3 (p=0.02) Probability of HIV Probability 0% Study Time Intervention Control Intervention Control HIV-TB/death from illness among baseline HIV+/unknown (N=26,096); TB incidence among HIV+ (N=13,430) stable residents; estimates of HIV-TB/death risk by 2.75 years using Kaplan-Meier censored at outmigration or death from other cause Impact of SEARCH on Community Health: Hypertension control was higher RR: 1.16 • Population-level HTN 70% (95%CI: 0.99, 1.36) RR: 1.23 RR: 1.26 p=0.07 (95%CI: 1.10, 1.40) control at year 3 was 60% (95%CI: 1.15, 1.39) p=0.002 26% higher in p<0.001 50% intervention vs. 40% control arm (p<0.001) 30% 55% 48% 52% • HTN-HIV dual 47% 42% 20% 37% control higher in 10% intervention vs. 0% control (p=0.002)

Proportion with HTN control at year 3 year at control HTN with Proportion Prevalent HTN* Prevalent HTN* & HIV+ Measured HTN & HIV+ • Similar findings for (dual control) HTN and HIV Disease Status at Year 3 combined HTN/DM Among baseline stable residents aged >30 years, N=59,218 with HTN measured at FUY3; adjusted for missing measures of HTN control; *also adjusted for unknown HTN status. Impact of SEARCH on Cumulative HIV incidence No difference detected RR: 1.01 1.6% (95%CI: 0.62 1.65) p=0.95 1.4% RR: 1.02 No difference in 3 year (95%CI: 0.66 1.56) 1.2% RR: 0.95 p=0.91 cumulative HIV incidence (95%CI: 0.77, 1.17) p=0.60 1.0% between arms

0.8% RR: 0.69 Intervention (95%CI: 0.23 2.03) 0.6% N=49,590 1.1% 1.1% p=0.39 360 seroconversions 0.9% 0.4% 0.8% 0.8% 0.8% Control 3 Year Cumulative HIV Incidence Cumulative 3 Year 0.2% 0.4% N= 45,493 0.3% 344 seroconversions 0.0% All Kenya Western Uganda Eastern Uganda Region Among incidence cohort of baseline HIV-negative stable residents; 91% intervention, 91% control alive and not out-migrated by year 3; of those, 89% intervention and 90% control with HIV status measured at year 3 Why no difference? 1. Active control: • 90% persons aware HIV status after baseline fairs in both arms • Greater health- seeking behaviors in the control after baseline • We implemented new guidelines; ART eligibility was “near universal” within one year • Mathematical model predicted 10% reduction in HIV incidence (0-19%) which we may not have detected*

2. New Infections from: • Outside the community • Acute infection outbreaks • Small subset of unsuppressed

*Jewell, IAC, 2018 In SEARCH study, did our annual measured HIV incidence decline over time? Yes 0.8 • 32% decline in 0.7 Kenya annual HIV 0.6 incidence from year 0.5 Uganda-West 1 to year 3 0.4 • 45% decline in 0.3 Overall Kenya 0.2 • 49% decline in men; 0.1 Uganda-East 19% decline in

HIV Incidence Rate (per 100 PY) 100 (per Rate Incidence HIV 0 women Year 1 Year 2 Year 3 Study Time Intervention arm - incidence rate calculated in 3 annual incidence cohorts of HIV-negative adult residents (inclusive of in- migrants and aging-in) with repeat HIV test one year later (Year 1 N= 52,468; Year 2 N=55,526; Year 3 N=57,858); Change over time using Poisson GEE adjusted for age, sex, mobility, w/ exchangeable covariance matrix What annual HIV incidence would we predict in the absence of SEARCH – a true control?

Individual-based network model (EMOD-HIV) of annual HIV incidence (blinded to all incidence results) in SEARCH communities under various scenarios

0.8 • Model: SEARCH 0.7 reduced incidence 40% 0.6 Modeled True Control • Model: 0.5 (no SEARCH interventions) Modeled SEARCH - No SEARCH: 0.7% 0.4 (active control) Modeled SEARCH - SEARCH: 0.3% 0.3 (intervention) • Measured: Measured SEARCH 0.2 annual incidence - SEARCH: 0.4% HIV Incidence (per 100 PY) 0.1

0 Year 1 Year 2 Year 3 Jewell, IAC, 2018 Summary

A community health approach with a patient- centered, multi-disease model rapidly increased population-level HIV suppression from 42% to 79% (intervention)-compared to control (68% ) at 3 years Improved Community Health Reduced HIV incidence 21% HIV mortality 32% Annual HIV incidence 59% HIV/TB year 3 annual incidence within arm Cumulative HIV incidence 26% HT control between arms* Explanation: SEARCH intervention *Explanation: Active control Hypothesis: Community health approach with patient-centered, multi-disease model would reduce HIV and improve community health compared to SOC with baseline HIV testing Study Design: 32 community RCT: N= 150,395 persons > 15 years rural Uganda/Kenya Intervention: Baseline + annual health fair, Universal ART, Streamlined care for HIV/NCD Control: Baseline health fair; ART by 2010,2013,2015 WHO guidelines Conclusions • SEARCH multi-disease, patient-centered approach is one model that can be adapted in rural Africa to accelerate reductions in mortality, TB and new HIV infections synergistically with improved NCD control-- in line with Sustainable Development Goals.

• To achieve HIV elimination at <0.1% incidence, we need to: increase viral suppression among youth, understand HIV transmission dynamics, and integrate new treatment and prevention( VAMC, PreP, vaginal ring, etc) interventions efficiently and effectively in a multi-disease and financed approach. Acknowledgements Univ. of California, SF: Ted Ruel Makerere University: Kenya Medical Research Institute: Diane Havlir Rachel Burger Moses Kamya Elizabeth Bukusi Edwin Charlebois Katie Snyman Infectious Disease James Ayieko Tamara Clark Monica Getahun Research Collaboration: Norton Sang Craig Cohen Carina Marquez Dalsone Kwarisiima Kevin Kadede Gabe Chamie Joshua Schwab Jane Kabami Winter Olilo Teri Liegler Univ. of California, Atukunda Mucunguzi Berkeley: Patrick Omanya Vivek Jain Geoff Lavoy Maya Petersen Bernard Awuonda Carol Camlin Emmanuel Ssemondo Mark van der Laan Jackson Achando Starley Shade Dathan Byonanebye Univ of Mass, Amherst: Erick Mugoma Wafula Doug Black Florence Mwangwa Laura Balzer Albert Plenty Asiphas Owaraganise Univ of Pennsylvania: & so many others in the Cait Koss Hellen Nakato SEARCH team! Harsha Thirumurthy Lillian Brown Joel Kironde A special thanks to our sponsors, partners and collaborators & advisory boards: - Sponsors: NIH, PEPFAR, Gilead - DSMB Board: Nicholas Jewell, Dorothy Mbori-Ngacha, Harriet Mayanja, Stephen Watiti, Carlos Del Rio - Scientific Advisory Board: Carl Dieffenbach, Haileyesus Getahun, Eric Goosby, Reuben Granich, Ade Fakoya, Nancy Padian, James Rooney, Doug Shaffer, David Wilson, - Uganda and Kenya Advisory Boards - Community members and local leaders in Uganda and Kenya