CROI 2015 Program and Abstracts

Abstract Listing

Poster Abstracts

and 2 active TB deaths averted per 1000 patients at an incremental cost of $2.01 per individual. Both Intervention 1 and Intervention 2 were found to be highly cost effective compared to no IPT with associated ICERs of $115.77 and $100.50/DALYs-averted respectively, at a willingness-to-pay threshold of Indian per capita GDP ($1500/DALYs-averted).

Conclusions: Implementation of 6 months of antepartum IPT for HIV-infected women (with or without CD4 cell count stratification) is a highly cost-effective strategy for prevention of TB compared to current practices in India. 1109 Epidemiologic Benefits and Cost-Effectiveness of Improving Rwanda’s HIV Care Cascade Eran Bendavid 1 ; Edward Mills 2 ; Steve Kanters 2 ; Sabin Nsanzimana 3 1 Stanford University, Stanford, CA, US; 2 University of Ottawa, Ottawa, Canada; 3 Rwanda Biomedical Center, Kigali, Rwanda Background: Connecting HIV+ individuals to treatment programs and sustaining effective antiretroviral therapy (ART) is a challenge in developing countries. The implications of addressing patient linkage and retention are not only partially characterized. We use the national coverage of data on HIV care in Rwanda to estimate the epidemiologic benefits and cost-effectiveness of improving the HIV care cascade. Methods: We parameterized an HIV disease and transmission simulation model to reflect the epidemic in Rwanda using age-structured prevalence and detailed information on ART program. The care cascade was modeled as a time-dependent process of patient loss between diagnosis and initial staging, staging and ART initiation, and loss after ART initiation. Costs were obtained from a top-down accounting of HIV testing, treatment, and laboratory monitoring in Rwanda. Alternative scenarios included improving HIV testing to 100% of the population, perfect linkage from testing to staging, and reduction of pre-ART and post-ART patient leakage to 0%. In addition, we simulated a strategy of immediate treatment among patients linked to care, as opposed to treatment initiation at CD4<350 cells/mm 3 . Model and parameter uncertainty were estimated using Monte Carlo and probabilistic simulations. Results: We calibrated the model to Rwanda’s adult HIV prevalence between 2004 and 2013, a drop from 3.6% to 2.8%. Incidence reduction was greatest under a scenario of combined perfect system performance, including immediate ART initiation (49% reduction in 10 years (uncertainty bounds 39-60%)), followed by immediate treatment initiation (36%, 28-43%) and reduction of post-ART patient loss (14%, 5-22%). Using a unit-cost approach to estimate the total costs of scaling up HIV treatment and services, we estimate that adopting immediate ART initiation will cost approximately $120 million additional to the status quo (discounted at 3% annually). The incremental cost-effectiveness of immediate ART initiation was estimated at $1,094 per life-year gained (556-1922), and that of a perfect system performance, including immediate ART initiation was $1,803 (1122- 2912) per life-year gained. No other scenario was cost-effective. Conclusions: Using national data on ART care and treatment in Rwanda, we demonstrate that the greatest epidemiologic benefits are likely to result from immediate ART initiation. While the budgetary implications of immediate initiation are large, this approach is cost-effective in Rwanda by traditional standards. 1110 The Cost-Effectiveness of Early ART Initiation in South Africa: A Quasi-Experiment Jacob Bor 1 ; Ellen Moscoe 2 ; Natsayi Chimbindi 3 ; Kobus Herbst 3 ; Kevindra K. Naidu 3 ; FrankTanser 3 ; Deenan Pillay 3 ;Till Barnighausen 3 1 Boston University School of Public Health, Boston, MA, US; 2 Harvard School of Public Health, Boston, MA, US; 3 Wellcome Trust Africa Centre for Health and Population Studies, Somkhele, South Africa Background: Clinical trials are not well suited to evaluate the effectiveness and cost-effectiveness of interventions in “real world” settings. Using a quasi-experimental regression-discontinuity design (Bor et al. 2014), we establish the causal effect of early (vs. deferred) ART initiation on patient survival in rural South Africa, and obtain empirical (as opposed to modeled) cost-effectiveness estimates. Methods: Demographic data from a large population surveillance in rural KwaZulu-Natal were linked to clinical records from South Africa’s public sector ART program. 4391 patients enrolled in HIV care between 2007 and 2011. CD4 counts were collected upon entry into care regardless of ART initiation. Subjects were eligible for ART if CD4 < 200 cells/ μ L, as per national guidelines during this period. Dates of death were obtained from the demographic surveillance; dates of initiation and follow-up CD4 counts were obtained from clinical records. Patients were followed for up to five years. We estimated the causal effect of immediate ART eligibility on survival, immune health, and time spent in pre-ART and on ART, which were used to estimate costs. Effects were estimated using a regression-discontinuity design, which exploits the quasi-random nature of treatment assignment for patients with first CD4 counts close to the eligibility threshold. Patients just above vs. just below the threshold are similar on all observed and unobserved factors; but they receive different treatment assignments. Results: Patients presenting with a CD4+ count just below 200 cells/ μ L were 4.3% points (95% CI 0.6, 8.0) more likely to be alive at two years compared to patients presenting with a CD4+ count just above the cut-off, an advantage that persisted at five years (Fig 1). These effects imply a 14.9% point two-year survival advantage for patients who actually initiated ART because they had an eligible CD4+ count. Large, persistent gains in clinical immune function were also observed among patients who were ART eligible. Over a five- year horizon, the additional medical care provided to ART-eligible patients implied a cost of $1967 per life year saved compared to treating patients with CD4+ counts close to 200 cells/ μ L.

Poster Abstracts

644

CROI 2015