CROI 2016 Abstract eBook

Abstract Listing

Poster Abstracts

in 2015 among 40% of high-risk MSM, 10% of PWID and 10% of HRH and maintaining constant coverage through 2020. We applied a 96% reduction in HIV sexual transmission to PLWH who achieved VLS and assumed PrEP efficacy of 73% for MSM, 75% for HRH and 49% for PWID. Results: In the base case, PrEP reduced HIV incidence by 18% (48,221 cases) over 5 years. Compared with the base case, improving to 90% diagnosed, 85% LTC and 60% VLS among diagnosed reduced incidence by 34% (88,908 fewer cases); the marginal benefit of adding PrEP was an additional 12% (31,988 cases) reduction. Compared with the base case, the scenario with 80% VLS reduced incidence by 63% (168,132 cases); the marginal benefit of PrEP was an additional 7% (16,929 cases). Conclusions: Increasing diagnosis, care and treatment of PLWH resulted in large decreases in HIV incidence by 2020. The marginal benefit of PrEP decreased as diagnosis, care and treatment improved. However, even at high levels of viral load suppression, PrEP continued to achieve reductions in HIV incidence over 5 years.

1052 PrEP Is Only Cost-Effective Among MSM in the Netherlands When Used on Demand Brooke E. Nichols 1 ; Charles A. Boucher 1 ; Marc van derValk 2 ; Bart J. Rijnders 1 ; David A. van deVijver 1 1 Erasmus Univ Med Cntr, Rotterdam, Netherlands; 2 Academic Med Cntr, Amsterdam, Netherlands

Background: Pre-exposure prophylaxis (PrEP) with tenofovir and emtricitabine has been demonstrated to effectively prevent new HIV infections among men who have sex with men (MSM). It has recently been shown that on demand PrEP use, in which individuals take PrEP just before and after sexual contact, can be just as effective as daily PrEP in preventing HIV infections. The high cost of PrEP is still a primary limitation in its use for HIV prevention. The aim of this study was to compare the epidemiological impact and cost- effectiveness of both daily and on demand PrEP compared to no PrEP. Methods: A deterministic mathematical model was calibrated to the well-defined Dutch HIV epidemic among MSM. We aimed to predict the impact of PrEP targeted to 10% of the most sexually active individuals (median 4,500 MSM/year) over the coming 12 years, including two years of scale-up. Cost-effectiveness ratios were calculated to predict the cost-effectiveness of daily (€7,099/year) and on demand (€3,550/year) PrEP, and prevented infections were calculated to predict the epidemic impact. A decrease of 50% in the price of PrEP was considered in a sensitivity analysis, given that generic PrEP may soon become available. Cost-effectiveness ratios below €20,000 were considered to be cost- effective in this analysis. Results: Targeting PrEP to 10% of the most sexually active individuals was predicted to prevent 14.0% (interquartile range [IQR] 11.9%-16.3%) of new infections over the coming 12 years compared to no PrEP usage. Daily PrEP was predicted to cost €36,300 per quality adjusted life year (QALY) gained (IQR €34,000-€45,100). On demand dosing of PrEP has the potential to cut the cost per QALY gained by more than half to a cost-effective €17,700 (IQR €16,500-€22,200) per QALY gained over 12 years compared to no PrEP usage. This cost per QALY can be further reduced to €8,400 (IQR €7,700-€10,700) if the cost of on demand PrEP is reduced by 50%. If increasing numbers of patients are diagnosed and placed on treatment early, however, PrEP will have a diminished impact on the epidemic. This can result in the cost per QALY gained of on demand PrEP to increase to as much as €32,400 with full-price PrEP, and €16,200 with reduced-price PrEP. Con c lusions: The use of PrEP is only cost-effective when used on demand, and can become far more cost-effective when generic PrEP becomes available. The precise cost- effectiveness estimate is, however, dependent upon the impact of earlier treatment initiation on the future HIV epidemic. 1053 Linkage of UK HIV and Tuberculosis Data Using Probabilistic and Deterministic Methods Joanne R. Winter 1 ;Valerie Delpech 2 ; Peter Kirwan 2 ; Helen R. Stagg 1 ; SrividhyaVenugopalan 2 ; Andrew Skingsley 2 ; Maeve K. Lalor 2 ; H. L.Thomas 2 ; Ibrahim R. Abubakar 1 1 Univ Coll London, London, UK; 2 PH England, London, UK Background: Tuberculosis (TB)-HIV co-infection is a major cause of death worldwide. Understanding the impact of HIV and TB prevention and treatment programs requires high quality surveillance data of both diseases as well as co-infected persons. Internationally there is a lack of routine surveillance data on the proportion of cases with HIV who have had TB, and many TB surveillance systems do not consistently collect HIV data. Databases containing highly confidential information (e.g. HIV status) typically contain no unique identifiers and little identifying information, making record linkage difficult. The two key approaches to record linkage, probabilistic and deterministic, both have significant limitations, particularly when few common variables are available. We developed a novel approach to link national HIV and TB surveillance datasets by trialling probabilistic and deterministic linkage methods, plus combined approaches. Methods: A deterministic matching algorithm, using a hierarchy of matching criteria, was developed using Public Health England surveillance data; for 2001-2011 the TB database had 87,907 case reports and the HIV database 105,078. This was compared to an existing probabilistic method which utilised two thresholds of match weight for accepting/ rejecting matches and manual review of records between these thresholds. With no unique identifiers for validation, we evaluated deterministic criteria by calculating the mean probabilistic weight of the matches they identified. Criteria with a mean weight greater than the upper threshold were included in the algorithm. We compared algorithms based on probabilistic, deterministic, and combined approaches. Results: The novel, combined method accepts the above-threshold results from the probabilistic method, and the deterministic algorithm is then run in place of manual review to accept or reject intermediate-weight matches (Figure). Additional rules were added to accept/reject matches. The newmethod found 4308 (94.2%) and 4509 (85.5%) of the cases identified through probabilistic (n=4574) and deterministic matching (n=5277) respectively, and additional high-quality matches not identified by probabilistic (n=583) and deterministic (n=382) methods. Conclusions: Our novel approach combines probabilistic and deterministic methods and is quicker and more repeatable with less potential bias than a method requiring manual review. We propose that this approach is considered when linking large datasets with minimal common variables.

Poster Abstracts

451

CROI 2016