CROI 2017 Abstract e-Book

Abstract eBook

Poster and Themed Discussion Abstracts

would go undetected; (ii) if VL results can be available instantly (per POC) VL monitoring will capture 100% of viraemia but otherwise VL testing at any time point appears poorly predictive of raised VL occurring beyond 4 weeks in the future; (iii) the best-performing antenatal monitoring schedule (with 1 test at 36w gestation) predicts 77% of VL>1000 cps/mL at delivery but only 86% of women would be tested at this time due to preterm deliveries; (iv) the addition of pre-ART VL measures allows prediction of an additional 3-5% of all elevated VL. Conclusion: This simulation suggests that pregnant women warrant VL monitoring approaches different from non-pregnant adults. Simple VL monitoring guidelines can be used to predict approximately three-quarters of all elevated VL at delivery, but effective implementation would require rapid turnaround times. POC testing may be important to detect larger proportions of viraemic women on ART for intervention. 768 VIRAL SUPPRESSION AMONG HIV+ PREGNANT WOMEN ENTERING ANTENATAL CARE ON ART IN UGANDA Catherine A. Koss 1 , Paul Natureeba 2 , Deborah Cohan 1 , Teddy Ochieng 2 , Theodore Ruel 1 , Miriam Nakalembe 3 , Tamara D. Clark 1 , Edwin D. Charlebois 1 , Moses R. Kamya 3 , Diane V. Havlir 1 1 Univ of California San Francisco, San Francisco, CA, USA, 2 Makerere Univ–Univ of California San Francisco Rsr Collab, Kampala, Uganda, 3 Makerere Univ Coll of Hlth Scis, Kampala, Uganda Background: As access to testing and treatment expands throughout Africa, an increasing number of HIV+ pregnant women are presenting for antenatal care already on ART. However, little is known about rates of viral suppression (VS) in this population throughout pregnancy, as many settings still lack routine viral load (VL) monitoring. Methods: We evaluated VS among HIV+ pregnant women who enrolled in a study (NCT02282293) of HIV and malaria in rural Uganda between 12-28 weeks gestation. Participants had previously received care at area clinics per national guidelines without VL monitoring. At enrollment, women were ART-naïve or receiving NNRTIs; those on nevirapine (NVP) were switched to efavirenz (EFV). VL was tested at enrollment, 8 weeks after enrollment, delivery, and additionally as clinically indicated. Those with confirmed virologic failure (≥2 VL >1000 c/ml after >90 days on ART) were switched to protease inhibitors (PI). We assessed the prevalence of VS (HIV-1 RNA ≤400 c/ml) and used logistic regression to examine factors associated with VS at enrollment and delivery. Results: From 12/2014-10/2015, 200 pregnant women entered the study: median age 31 years (IQR 25-35); median gravidity 4 (IQR 3-6); median CD4 503 cells/mm³ (IQR 372-638); median years since HIV diagnosis 3.1 (IQR 0.7-6.4); 19.5% diagnosed with HIV in current pregnancy. At enrollment, among 161 (80.5%) participants already on ART (125 on EFV, 36 on NVP) for a median of 2 years (IQR 0.5-4.2), VS was 79.5%; among 135 who had been on ART for >90 days prior to enrollment, VS was 84.4%. In a multivariate model including ART regimen, VS at enrollment was associated with time on ART (aOR 1.41 per additional year, 95% CI 1.08-1.85, p=0.01) and older age (aOR 2.23 per 10 years, 95% CI 1.05-4.75, p=0.04). At delivery, 187/200 (93.5%) had VS and EFV vs. PI was associated with VS (aOR 16.3, 95% CI 3.9-67.9, p<0.01). Of 135 women with VS at enrollment, 98.5%maintained VS at delivery. Of 21 participants with unsuppressed VL at enrollment despite >90 days on ART, 17 (80.9%) achieved VL<1000 at delivery with switch to PI (N=11) and adherence counseling. Conclusion: The majority of HIV+ women already on ART at entry into antenatal care had previously achieved and were able to sustain viral suppression during pregnancy. However, 15% of women already on ART for >90 days were not suppressed, and VL monitoring in these asymptomatic women triggered enhanced adherence counseling or ART switch, leading to 81% percent achieving VL<1000 at delivery. 769 MEASURES OF ENGAGEMENT IN ROUTINE HIV CARE TO PREDICT ELEVATED VIRAL LOAD ON ART Tamsin Phillips 1 , Allison Zerbe 2 , Agnes Ronan 1 , Nei-Yuan M. Hsiao 3 , Greg Petro 1 , James A. McIntyre 4 , Andrew Boulle 1 , Elaine J. Abrams 2 , Landon Myer 1 , for the Maternal Child Health - AntiretroviralTherapy (MCH-ART) Study 1 Univ of Cape Town, Cape Town, South Africa, 2 ICAP at Columbia Univ, New York, NY, USA, 3 Natl Hlth Lab Service, Cape Town, South Africa, 4 Anova Hlth Inst, Johannesburg, South Africa Background: Engagement in HIV care is a precursor to effective antiretroviral therapy (ART) use and viral suppression (VS). Engagement in care is commonly used as an outcome in ART implementation science and health service evaluation, yet associations between measures of engagement and their ability to predict viral load (VL) and VS have not been examined. Methods: We followed a cohort of women initiating ART in pregnancy at a primary care clinic in Cape Town, South Africa, with regular study visits for up to 24 months (including VL measures [Abbott RealTime HIV-1] and self-reported engagement in HIV care [SR], taken separately from routine services). On completion of the study, we retrospectively assessed engagement in HIV care using different forms of routinely-collected medical records, including: (i) HIV-related laboratory testing, (ii) pharmacy dispensing of ART, and (iii) records of HIV/ART clinical visits. Analyses examined the agreement of different engagement measures expressed as kappa statistics. We also assessed the ability of each engagement measure (and an aggregate measure of engagement in any of laboratory, pharmacy and/or clinical service) to predict VS <50 and <1000 copies/mL at 12-18 months on ART, based on VL measured in the cohort study. Results: Of 471 women included, 441 women (94%) had any evidence of engagement in care at 12-18 months on ART, including 65%, 62%, 59% and 87%with evidence from laboratory testing, pharmacy dispensing, clinical visits and SR, respectively. Agreement between laboratory, pharmacy and clinical visit sources was good (kappa 0.58) but agreement of these with SR was very low (kappa -0.04). Among 411 women with study VL available at 12-18 months on ART, 66% and 74% of women had VS <50 and <1000 copies/mL, respectively. Laboratory, pharmacy and clinical data sources had >75% sensitivity in predicting VS<50 and <1000 copies/mL, and the aggregate measure was 95% and 90% sensitive, respectively (Table). Specificities were more variable, with varying engagement in HIV care among women with raised VLs. The sensitivity of SR was high but specificity very poor. Conclusion: This analysis provides novel evidence that routinely-collected medical records measuring engagement in HIV care are a strong predictor of VS on ART. While the utility of each measure of engagement requires careful consideration in different health systems, these results suggest that routine engagement measures can be robust outcomes for ART programme evaluations and implementation science.

Poster and Themed Discussion Abstracts

CROI 2017 333