TY - JOUR
T1 - Assessing the use of surveillance data to estimate the impact of prevention interventions on HIV incidence in cluster-randomized controlled trials
AU - Mitchell, Kate M.
AU - Dimitrov, Dobromir
AU - Hughes, James P.
AU - Moore, Mia
AU - Vittinghoff, Eric
AU - Liu, Albert
AU - Cohen, Myron S.
AU - Beyrer, Chris
AU - Donnell, Deborah
AU - Boily, Marie Claude
N1 - Funding Information:
This work was supported by the HPTN Modelling Centre, which is funded by the U.S. National Institutes of Health (NIH) [grant number UM1 AI068617 ] through HPTN. KMM and MCB acknowledge funding from the MRC Centre for Global Infectious Disease Analysis (MRC GIDA) at Imperial College London , [grant number MR/R015600/1 ]. This award is jointly funded by the UK Medical Research Council (MRC) and the UK Department for International Development (DFID) under the MRC/DFID Concordat agreement and is also part of the EDCTP2 programme supported by the European Union . MSC acknowledges funding from the HIV Prevention Trials Network , which is funded by the U.S. NIH [grant number U01-AI068619 ], and from the U.S. National Institute of Diabetes and Digestive and Kidney Diseases [grant number R37-DK049381 ]. The funders had no role in study design, data collection analysis or interpretation, decision to publish, or preparation of the manuscript.
PY - 2020/12
Y1 - 2020/12
N2 - Background: In cluster-randomized controlled trials (C-RCTs) of HIV prevention strategies, HIV incidence is expensive to measure directly. Surveillance data on HIV diagnoses or viral suppression could provide cheaper incidence estimates. We used mathematical modelling to evaluate whether these measures can replace HIV incidence measurement in C-RCTs. Methods: We used a US HIV transmission model to simulate C-RCTs of expanded antiretroviral therapy(ART), pre-exposure prophylaxis(PrEP) and HIV testing, together or alone. We tested whether modelled reductions in total new HIV diagnoses, diagnoses with acute infection, diagnoses with early infection(CD4 > 500 cells/μl), diagnoses adjusted for testing volume, or the proportion virally non-suppressed, reflected HIV incidence reductions. Results: Over a two-year trial expanding PrEP alone, modelled reductions in total diagnoses underestimated incidence reductions by a median six percentage points(pp), with acceptable variability(95 % credible interval -14,-2pp). For trials expanding HIV testing alone or alongside ART + PrEP, greater, highly variable bias was seen[-20pp(-128,-1) and -30pp(-134,-16), respectively]. Acceptable levels of bias were only seen over longer trial durations when levels of awareness of HIV-positive status were already high. Expanding ART alone, only acute and early diagnoses reductions reflected incidence reduction well, with some bias[-3pp(-6,-1) and -8pp(-16,-3), respectively]. Early and adjusted diagnoses also reliably reflected incidence when scaling up PrEP alone[bias -5pp(-11,1) and 10pp(3,18), respectively]. For trials expanding testing (alone or with ART + PrEP), bias for all measures explored was too variable for them to replace direct incidence measures, unless using diagnoses when HIV status awareness was already high. Conclusions: Surveillance measures based on HIV diagnoses may sometimes be adequate surrogates for HIV incidence reduction in C-RCTs expanding ART or PrEP only, if adjusted for bias. However, all surveillance measures explored failed to approximate HIV incidence reductions for C-RCTs expanding HIV testing, unless levels of awareness of HIV-positive status were already high.
AB - Background: In cluster-randomized controlled trials (C-RCTs) of HIV prevention strategies, HIV incidence is expensive to measure directly. Surveillance data on HIV diagnoses or viral suppression could provide cheaper incidence estimates. We used mathematical modelling to evaluate whether these measures can replace HIV incidence measurement in C-RCTs. Methods: We used a US HIV transmission model to simulate C-RCTs of expanded antiretroviral therapy(ART), pre-exposure prophylaxis(PrEP) and HIV testing, together or alone. We tested whether modelled reductions in total new HIV diagnoses, diagnoses with acute infection, diagnoses with early infection(CD4 > 500 cells/μl), diagnoses adjusted for testing volume, or the proportion virally non-suppressed, reflected HIV incidence reductions. Results: Over a two-year trial expanding PrEP alone, modelled reductions in total diagnoses underestimated incidence reductions by a median six percentage points(pp), with acceptable variability(95 % credible interval -14,-2pp). For trials expanding HIV testing alone or alongside ART + PrEP, greater, highly variable bias was seen[-20pp(-128,-1) and -30pp(-134,-16), respectively]. Acceptable levels of bias were only seen over longer trial durations when levels of awareness of HIV-positive status were already high. Expanding ART alone, only acute and early diagnoses reductions reflected incidence reduction well, with some bias[-3pp(-6,-1) and -8pp(-16,-3), respectively]. Early and adjusted diagnoses also reliably reflected incidence when scaling up PrEP alone[bias -5pp(-11,1) and 10pp(3,18), respectively]. For trials expanding testing (alone or with ART + PrEP), bias for all measures explored was too variable for them to replace direct incidence measures, unless using diagnoses when HIV status awareness was already high. Conclusions: Surveillance measures based on HIV diagnoses may sometimes be adequate surrogates for HIV incidence reduction in C-RCTs expanding ART or PrEP only, if adjusted for bias. However, all surveillance measures explored failed to approximate HIV incidence reductions for C-RCTs expanding HIV testing, unless levels of awareness of HIV-positive status were already high.
KW - HIV
KW - Incidence
KW - Marker
KW - Mathematical modelling
KW - Surveillance data
KW - Trials
U2 - 10.1016/j.epidem.2020.100423
DO - 10.1016/j.epidem.2020.100423
M3 - Article
C2 - 33285419
AN - SCOPUS:85097366207
SN - 1755-4365
VL - 33
JO - Epidemics
JF - Epidemics
M1 - 100423
ER -
