Swaziland HIV Incidence Measurement Survey (SHIMS): A .

ArticlesSwaziland HIV Incidence Measurement Survey(SHIMS): a prospective national cohort studyJessica Justman, Jason B Reed, George Bicego, Deborah Donnell, Keala Li, Naomi Bock, Alison Koler, Neena M Philip, Charmaine K Mlambo,Bharat S Parekh, Yen T Duong, Dennis L Ellenberger, Wafaa M El-Sadr, Rejoice NkambuleSummaryBackground Swaziland has the highest national HIV prevalence worldwide. The Swaziland HIV IncidenceMeasurement Survey (SHIMS) provides the first national HIV incidence estimate based on prospectively observedHIV seroconversions.Methods A two-stage survey sampling design was used to select a nationally representative sample of men andwomen aged 18–49 years from 14 891 households in 575 enumeration areas in Swaziland, who underwenthousehold-based counselling and rapid HIV testing during 2011. All individuals aged 18–49 years who resided orhad slept in the household the night before and were willing to undergo home-based HIV testing, answerdemographic and behavioural questions in English or siSwati, and provide written informed consent were eligiblefor the study. We performed rapid HIV testing and assessed sociodemographic and behavioural characteristics withuse of a questionnaire at baseline and, for HIV-seronegative individuals, 6 months later. We calculated HIVincidence with Poisson regression modelling as events per person-years 100, and we assessed covariables aspredictors with Cox proportional hazards modelling. Survey weighting was applied and all models used surveysampling methods.Findings Between Dec 10, 2010, and June 25, 2011, 11 897 HIV-seronegative adults were enrolled in SHIMS and11 232 (94%) were re-tested. Of these, 145 HIV seroconversions were observed, resulting in a weighted HIV incidenceof 2 4% (95% CI 2 1–2 8). Incidence was nearly twice as high in women (3 1%; 95% CI 2 6–3 7) as in men (1 7%;1 3–2 1, p 0 0001). Among men, partner’s HIV-positive status (adjusted hazard ratio [aHR] 2 67, 1 06–6 82,p 0·040) or unknown serostatus (aHR 4 64, 2 32–9 27, p 0 0001) in the past 6 months predicted HIV seroconversion.Among women, significant predictors included not being married (aHR 2 90, 1 44–5 84, p 0 0030), having a spousewho lives elsewhere (aHR 2 66, 1 29–5 45, p 0 0078), and having a partner in the past 6 months with unknown HIVstatus (aHR 2 87, 1 44–5 84, p 0 0030).Interpretation Swaziland has the highest national HIV incidence in the world. In high-prevalence countries,population-based incidence measures and programmes that further expand HIV testing and support disclosure ofHIV status are needed.Funding President’s Emergency Plan for AIDS Relief (PEPFAR) by the Centers for Disease Control and Prevention.IntroductionSwaziland has the most severe HIV epidemic in theworld, with a measured HIV prevalence of 26% amongadults aged 15–49 years in 2006–07.1 To combat thisepidemic, in 2009, the Government of the Kingdom ofSwaziland (GKOS) initiated support for scale-up ofnational HIV prevention and treatment programmes,including a voluntary medical male circumcisioncampaign.2 The Swaziland HIV Incidence MeasurementSurvey (SHIMS) was designed to assess the effect ofthese programmes on HIV incidence by prospectivelymeasuring HIV seroconversions in a household-based,nationally representative sample of adults before andafter programme expansion.The approach of “know your epidemic, know yourresponse”3 is crucial to effective HIV programmes, andhaving accurate and detailed HIV incidence estimates isfundamental to this approach. Incidence estimates allowidentification of groups at the highest risk of new infectionsand, when repeated, determine the effect of programmesover time. Although incidence estimates of large populations are often modelled from trends in HIV prevalence,4such modelled estimates provide little demographic detail.HIV incidence laboratory assays5,6 are intended for use incross-sectional surveys but have yet to achieve optimumperformance.7 HIV incidence estimates based on the goldstandard of observed HIV seroconversions have not beenavailable at a national level although they have beenavailable from randomised clinical trials or observationalcohorts restricted to subpopulations at increased risk ofHIV.8,9 We report the national estimate, before programmeexpansion, of HIV incidence based on population-level,prospectively observed seroconversions.Lancet HIV 2016Published OnlineNovember 15, 4See (16)30187-4ICAP at Columbia University,Mailman School of PublicHealth, Columbia University,New York, NY, USA(J Justman MD, N M Philip MPH,Y T Duong PhD,Prof W M El-Sadr MD); Jhpiego,Washington, DC, USA(J B Reed MD); US Centers forDisease Control andPrevention, Center for GlobalHealth, Division of GlobalHIV/AIDS, Atlanta, GA, USA(G Bicego PhD, N Bock MD,B S Parekh PhD,D L Ellenberger PhD);Fred Hutchinson CancerResearch Center, Seattle, WA,USA (K Li MS, D Donnell PhD)Pact, Dar es Salaam, Tanzania(A Koler MPH); Clinton HealthAccess Initiative, Mbabane,Swaziland (C K Mlambo PhD);and Ministry of Health—Swaziland, Mbabane,Swaziland (R Nkambule MPH)Correspondence to:Dr Jessica Justman, ICAP atColumbia University,Departments of Epidemiologyand Medicine, Mailman Schoolof Public Health, ColumbiaUniversity, New York,NY 10032, USAjj2158@columbia.eduMethodsStudy design and participantsA two-stage sampling design was used to obtain across-sectional, nationally representative sample ofwww.thelancet.com/hiv Published online November 15, 2016 http://dx.doi.org/10.1016/S2352-3018(16)30190-41

ArticlesResearch in contextFor the SHIMS protocol e/studyprotocol-case-review-formsEvidence before this studySwaziland’s Demographic and Health Survey done in 2006–07showed a severe generalised epidemic, with an HIV prevalenceof 26% in 15–49 year olds. UNAIDS’ modelled estimates ofprevalence and incidence indicated a similarly severeepidemic. With a goal of reducing new HIV infections,Swaziland planned to scale up national treatment andprevention programmes. The Swaziland HIV IncidenceMeasurement Survey (SHIMS) was designed to assess HIVincidence before and after scale-up of these interventionprogrammes by measuring prospectively observed HIVseroconversion in a nationally representative adult cohort.A search of PubMed for studies published in English throughto Nov 7, 2016, using the search terms “HIV incidence”,“longitudinal cohort”, and “nationally representative”confirmed no previous direct measurement of national HIVincidence using this method.HhohhoLubomboManziniAdded value of this studyThis study reports the baseline results, before scale up ofnational interventions, of the first national HIV incidencemeasurement on the basis of prospectively observedseroconversion, the “gold standard” measure of the spread ofinfection. High retention rates, a rigorous HIV testingalgorithm, and a large number of seroconverter casespermit an accurate and detailed description of HIV incidencein Swaziland, a country at the centre of the global HIVepidemic.Implications of all the available evidenceAs the global scale-up of antiretroviral treatment extends intoits second decade, accurate and detailed knowledge of eachcountry’s epidemic is increasingly crucial to implement aneffective local HIV response.adults aged 18–49 years in a survey of 14 891 householdsfrom 575 enumeration areas in Swaziland (figure 1), withhousehold sample size calculations and other details aspreviously reported.10 Each selected household wasapproached by study personnel trained in Good ClinicalPractice11 who asked responding heads of household toreport the sex and age of all household members. Allindividuals aged 18–49 years who resided or had slept inthe household the night before and were willing toundergo home-based HIV testing, answer demographicand behavioural questions in English or siSwati, andprovide written informed consent were eligible for thestudy. We enrolled in the prospective HIV incidencecohort those HIV-seronegative individuals whoconsented to have a 6 month follow-up home-based HIVtesting and counselling visit. The SHIMS protocol andconsent forms were reviewed and approved by the GKOSScientific and Ethics Committee and the institutionalreview boards at Columbia University Medical Centerand the US Centers for Disease Control and Prevention.ProceduresShiselweniSampled enumeration areasFigure 1: Distribution of the 575 enumeration areas sampled in the Swaziland HIV Incidence MeasurementSurvey across the four regions of Swaziland2Study teams comprising one nurse and one or twocounsellors did HIV counselling, venepuncture, andrapid HIV testing, provided condoms, and collecteddemographic, clinical, and behavioural informationwith questionnaires12 administered during face-to-faceinterviews in a private location in or just outside thehome. HIV test results were given to participants duringthe household visit. All HIV-seronegative individualsenrolled in the HIV incidence cohort had a 6 monthfollow-up visit, with similar procedures, includingverification of participant identity and repeat HIV testing.At the baseline and 6 month follow-up interviews,information was obtained about sexual behaviours in thepast 6 months and characteristics of the three mostwww.thelancet.com/hiv Published online November 15, 2016 http://dx.doi.org/10.1016/S2352-3018(16)30190-4

Articlesrecent sexual partners. Current pregnancy status inwomen was based on self-report. All individuals whotested HIV seropositive at either the baseline or follow-upvisit were counselled and referred to HIV care as pernational guidelines.Rapid HIV testing was done in the field on wholeblood samples obtained by venepuncture, as previouslydescribed.10,13 Samples were initially tested withDetermine HIV-1/2Ag/Ab Combo (Alere, Japan) andDetermine-reactive samples were confirmed withUni-Gold HIV Test (Trinity Biotech, Ireland), followingSwaziland’s serial testing algorithm. All HIVseronegative samples from the baseline visit, but notthe follow-up visit, had a nucleic acid amplification test(NAAT) with pools of ten samples to identify individualswith virological evidence of acute HIV infection.14,15Individuals with NAAT-positive results had follow-upvisits within 6 months to confirm seroconversionand were subsequently censored from the incidenceanalyses.1314 891 households selected for sample1556 households not occupied1207 head of household absent or not home349 household destroyed or vacant13 335 households occupied742 head of household refused22 occupied but no member listed12 571 participating households54 655 individuals residing in participating households13 582 women aged 18–49 years11 048 men aged 18–49 years1116 refused1422 not contacted1388 refused2527 not contacted11 044 participated in SHIMSbaseline survey6810 HIV uninfected7133 participated in SHIMSbaseline survey4232 HIV infected2 HIV status unknown5559 HIV uninfected45 refused175 other reasons1571 HIV infected3 HIV status unknown43 refused209 other reasons6590 enrolled in incidence5307 enrolled in incidence665 could not be relocated for re-testing and re-survey11 232 completed 6 month follow-up visit6230 were women5002 were menFigure 2: Study flow diagramSHIMS Swaziland HIV Incidence Measurement Survey.www.thelancet.com/hiv Published online November 15, 2016 http://dx.doi.org/10.1016/S2352-3018(16)30190-43