CROI 2017 Abstract e-Book

Abstract eBook

Poster and Themed Discussion Abstracts

1 VA Connecticut Hlthcare System, West Haven, CT, USA, 2 Yale Univ, New Haven, CT, USA, 3 Washington, DC, VA Med Cntr, Washington, DC, USA, 4 Baylor Coll of Med, Houston, TX, USA, 5 Univ of California San Francisco, San Francsico, CA, USA, 6 Columbia Univ, New York, NY, USA Background: As HIV infected (HIV+) individuals age, prevention of geriatric syndromes, particularly falls, is an important focus of research and clinical care. Using a machine learning algorithm to identify medically significant falls, we explored the relationship between HIV infection and falls. Methods: Using data from the Veterans Aging Cohort Study, we identified falls by external cause of injury codes and a machine learning algorithm that identified falls in radiology reports. As verified with chart review, reliability and accuracy of this algorithmwere excellent: sensitivity 94.5, and positive predictive value 92.6. In addition to HIV status, falls models included adjustment for race, sex, BMI, number of prescription medications taken in the past year (excluding antiretrovirals), specific medications associated with falls (antihypertensives, hypoglycemics, antipsychotics, benzodiazepines, hypnotics, muscle relaxers, and opiates), and comorbid conditions identified by ICD9 codes (HCV, anemia, coronary artery disease, heart failure, osteoarthritis, diabetes, COPD and pneumonia, end stage liver disease, hypertension, stroke, dementia, end stage renal disease, major depression, alcohol and drug use/abuse). With the exception of demographics, all covariates were time updated. We used general estimating equations (GEE) to assess the longitudinal association between HIV infection and falls in repeated six month intervals. Results: 130,107 Veterans were included (34% HIV+, 2%women, 48% Black, and 8% Hispanic, mean age 47±10 years). Falls incidence was 30/1000 person-years among HIV+ and 27/1000 person-years among uninfected persons (p<0.0001). HIV was associated with falls in unadjusted (odds ratio: 1.09; 95% confidence interval: 1.06, 1.12; p<0.0001) but not in fully adjusted models (Table).The strongest predictors included HCV infection, alcohol use/abuse, anemia, dementia, stroke, greater number of prescription medications taken, and the use of antipsychotics, benzodiazepines, and opiates. Conclusion: Falls are more common among HIV+ vs uninfected individuals but appear to be mediated by comorbid conditions known to be associated with falls (stroke and dementia) and by potentially modifiable factors: polypharmacy, opiate use, alcohol abuse, anemia and HCV infection. Efforts to reduce polypharmacy and hazardous medication use should be explored, especially in patients with high-risk comorbidities. Further research is needed to explore the role of HCV treatment in mitigating falls risk. 931 VIREMIA COPY-YEARS & MORTALITY IN HIV+ MEN STARTING ART: HOW MUCH HISTORY IS NEEDED? Ruibin Wang 1 , Sabina Haberlen 1 , Frank J. Palella 2 , Joseph B. Margolick 1 , Bernard Macatangay 3 , Otoniel Martínez-Maza 4 , Lisa Jacobson 1 , Alison Abraham 1 1 The Johns Hopkins Univ, Baltimore, MD, USA, 2 Northwestern Univ, Chicago, IL, USA, 3 Univ of Pittsburgh, Pittsburgh, PA, USA, 4 Univ of California, Los Angeles, Los Angeles, CA, USA Background: HIV replication while on combination antiretroviral therapy (cART) may reduce survival. Current methods to assess cumulative viral burden are limited by the need for extensive historical viral load (VL) data. We aimed to develop a prognostic metric of cumulative viral burden that requires less VL data. Methods: HIV+men in the Multicenter AIDS Cohort Study (MACS) who started cART between 1995 and 2014 were included. Viremia copy-years (VCY) was calculated as the area under the VL curve since cART initiation. This overall metric was compared to 20 different time-updated VCY metrics (10 calculated VCY over the most recent 1 to 10-year period; 10 calculated VCY over 1 to 10-year period immediately after cART initiation) and 3 cross-sectional VL metrics (VL at cART initiation, first post-cART VL and current VL). Men who did not have a VL measured in >1 year were censored. Lognormal survival models were used to assess the associations of VCY and cross-sectional VL metrics with all-cause mortality after cART initiation. Measure of association was the relative time (RT) to death, with a lower value indicating worse survival outcome. All models were adjusted for age, race, year of study entry, study center, baseline and current CD4 T cell count, and history of clinically-defined AIDS. A second model incorporated both VCY metrics and the 3 cross-sectional VLs. Model fit was evaluated using Akaike Information Criteria (AIC), a lower value indicating better fit. Results: Among 841 HIV+, cART-initiating men, 22%were black. At cART initiation, mean age was 43y; mean CD4 361 cells/µL; mean log 10 VL 4.5. Median follow-up was 5 years; 74 died. For the 10 recent VCY metrics, each log 10 increase in VCY was independently associated with 23% or more decrease in survival time (RT 0.72-0.77, Table). These RTs were similar to that of the overall VCY (RT 0.76). When further adjusted for cross-sectional VLs, only VCY over the recent 2 years remained significantly associated with death (RT 0.75, 95% CI 0.57-0.98). The model with VCY in the recent 2 years alone had the best fit based on AICs (Table). Conclusion: These results among cART-initiators suggest that the prognostic value of VCY over the most recent 2 years was similar to that of the overall VCY since cART initiation, while requiring much less viral load data. VCY over the most recent 2 years could potentially serve as a useful measure of mortality risk among cART-treated HIV+ persons.

Poster and Themed Discussion Abstracts

932 LOW-LEVEL VIREMIA IS ASSOCIATED WITH VIROLOGIC FAILURE IN A LARGE MILITARY COHORT Christie Joya 1 , Seung Hyun Won 2 , Karl Kronmann 3 , Jason Okulicz 4 , Timothy Whitman 1 , Robert Deiss 5 , Christina Schofield 6 , Julie Pavlin 5 , Brian K. Agan 5 , Anuradha Ganesan 5

CROI 2017 403