CROI 2017 Abstract e-Book

Abstract eBook

Poster and Themed Discussion Abstracts

Results: Between 7/1/15 and 6/30/16, 329 charts out of 4,741 HCV tested non-BC patients were reviewed; 80 (1.7%) HCV Ab+ or indeterminate patients were compared to 249 randomly selected HCV Ab negative (Ab-) controls (see table for demographics). In bivariate analysis, patients with at least one documented RF were more than twice as likely to have Medicaid (p = 0.005) and more than three times as likely to have Medicare than patients without RFs (p = 0.0034). Eighteen (23%) HCV Ab+ and 123 (49%) HCV Ab- had no identified RFs; 6 (33%) HCV Ab+ reported RFs only after a positive test result. In multivariate logistic regression, persons were more likely to be HCV Ab+ if they: reported drug use (OR adj 26, CI95 6.1-109.8), had Medicaid v. private insurance (3.4, 1.6-7.7), and were white v. other races (3.4, 1.5-7.9), adjusting for demographic factors and opiate prescriptions; sex behavior was no longer significant (ROC = 0.823). There was a significant interaction between age over 40 and opiate prescription use; these groups were 11x more likely to be HCV Ab+ (CI95 1.6-74.8). Conclusion: In non-BC patients, drug use remained a significant predictor of HCV positivity, as in the BC. However, white race was more significant than black race, which is reversed compared to the BC. The CDC has reported an increase of HCV in opiate abusers, and our data shows some signal for increased risk as well. RF testing in non-BC patients has the potential to miss a significant number of HCV Ab+ patients. Given patient- and provider-level barriers in elucidating RFs, universal HCV Ab testing may be warranted.

518 ESTABLISHING EPIDEMIOLOGICAL LINKAGE WITHIN HCV NETWORKS USING GENETIC DISTANCE Rebecca Rose 1 , Susanna Lamers 2 , Guido Massaccesi 3 , William Osburn 3 , Stuart C. Ray 3 , Andrea Cox 3 , Oliver Laeyendecker 4 1 BioInfoExperts, Norfolk, VA, USA, 2 BioInfoExperts, Thibodaux, LA, USA, 3 Johns Hopkins Univ, Baltimore, MD, USA, 4 NIAID, Baltimore, MD, USA

Background: Viral genetic data can be used for inferring epidemiological networks. We investigated genetic distance thresholds to identify clusters of related viruses among a highly networked population of people who inject drugs (PWID) from Baltimore. Methods: We used 908 core-E1 bulk sequences from 166 PWID in the Baltimore Before and After Acute Study of Hepatitis (BBAASH) cohort. All subjects had known dates of infection, including chronically infected individuals and those who experienced clearance and re- or co-infection. The number of sequences per subject ranged from 1 to 31, with follow up extending to 14 years post infection. We used HIV-TRACE to calculate genetic distance with the TN93 model and to determine clusters using pairwise genetic distance thresholds of 1-50%. The total number of clusters was calculated using the between- subject distances. Results: Sequences were predominantly subtype 1a(71%), as well as 1b(7%), 2b(10%) and 3a(11%). 28 subjects were infected with >1 sub/genotype during serial sampling. The largest peak in the distribution of genetic distances between subjects was at 10% (Fig 1A), corresponding to the within subtype comparisons (primarily subtype 1a). A gamma-shaped curve with a peak at 0 and nadir at 2% was present, corresponding to the within-subject comparisons. Four additional peaks were evident at 10%, 35%, 50%, and 65%, which corresponded to comparisons between within genotype, 1a and 1b, 1a/1b and 3a, and 1a/1b/3a and 2b, respectively. In contrast, the bulk (64%) of the distribution of distances within subjects was <3% (Fig 1B). This dataset also showed peaks at 10%, 35%, 50%, and 65%, although these larger distances were much less frequent than in the between-subject dataset. As expected, the number of clusters was highest at 1% (n=133) and decreased slightly at 2% (n=129) and 3% (n=120). A sharp decline was evident until 9% (n=9) and plateaued thereafter (Fig 1C, blue bars). The number of clusters with >1 subject at each threshold was highest at 3% (Fig 1C, red bars). However, the overall proportion of clusters containing >1 subject was lowest at 1% genetic distance (37/133) and increased with distance until all clusters contained >1 subject at 10%. Conclusion: A threshold of 3% appears to distinguish within versus between-subject distances, for the Core-E1 sequence of HCV. Even at the lowest estimate genetic distance (1%), 28% of clusters contained more than one subject, thus suggesting a high degree of connectivity in this cohort.

Poster and Themed Discussion Abstracts

CROI 2017 216