CROI 2020 Abstract eBook

Abstract eBook

Poster Abstracts

951 SIGNIFICANT UNDERQUANTIFICATION OF HIV RNA WITHIN ROUTINE SETTINGS IN SOUTH AFRICA Ahmad Haeri Mazanderani 1 , Tendesayi Kufa-Chakezha 1 , Tanya Murray 1 , Aurelie Mukendi 1 , Faith Moyo 1 , Kate Braithwaite 2 , Karl Technau 2 , Sergio Carmona 2 , Gayle G. Sherman 2 1 National Institute for Communicable Diseases, Johannesburg, South Africa, 2 University of the Witwatersrand, Johannesburg, South Africa Background: The extent to which pre-analytical variables impact HIV viral load (VL) accuracy is unknown. We describe VL log difference between paired point of care (POC) and centralized laboratory (lab) results within routine clinical settings, and determine whether result variability was associated with time to result, facility or season. Methods: Secondary analysis of data from a POC implementation study at four tertiary facilities in South Africa between March 2018 - August 2019. Two 4ml EDTA tubes were collected from HIV-positive women at time of delivery. One sample was centrifuged and plasma tested within 24 hours on an HIV VL assay at POC (Cepheid Xpert) and the other sent for routine lab testing (Cobas 8800 or Abbott RealTime) but only centrifuged at the testing lab. Results were transformed into log scale with those below limit of detection assigned a log value of 0.001 and results below limit of quantitation assigned value of lower limit of quantitation. Median log differences were calculated as POC - lab result. Ranksum and K-tests for equality of medians were used to determine if log differences varied significantly by time to lab result, facility, and season. Proportion of paired results with >0.5log difference and discrepancy at 1000 RNA cps/ml threshold were compared using X2 tests. Multivariable binomial regression was performed to identify variables associated with discrepant VL results at 1000 cps/ml threshold. Results: Among 1600 paired results, median POC VL was 2.1 (1.6–3.9) and lab VL was 2.0 (1.3–3.6) log (p<0.001), with median time to result of 1.6 (1.5–4.8) and 67.5 (45.5–99.1) hours, respectively. 475 (29.7%) specimen pairs had a VL difference >0.5log. Longer duration to lab result (p<0.001) and facility (p<0.001) were associated with >0.5log difference (Table 1). 59 (3.7%) paired samples had discrepant results at 1000 cps/ml threshold, of which 53 (3.3%) had a POC VL ≥1000 cps/ml but lab VL <1000 cps/ml. There was significant misclassification of lab VLs resulted ≥5 days compared to <5 days (aRR 1.82: 1.01–3.30). Conclusion: Longer time to lab result and facility, but not season, increased likelihood of underquantifying HIV VL by >0.5log and misclassifying results as <1000 cps/ml. Timely centrifugation and testing should be prioritized.

Methods: Analytical sensitivity was determined by probit analysis on dilution series of HIV-1 standard and 11 seroconversion panels. Specificity was determined by testing 500 HIV-1 negative WB samples and those containing hepatitis B, C and HTLV-I or II as well as potentially interfering substances such as bilirubin and ART drugs. Clinical sensitivity and specificity was determined using venous and finger-prick capillary blood collected from consecutive HIV-positive patients attending clinics for routine testing in four countries (Cameroon, UK, Ukraine, Zimbabwe). 1,490 venous and 816 finger-prick samples were tested. SAMBA HIV VL results using WB were compared with to Abbott Real-time PCR assay using plasma and analysed as a binary results (<1000 cp/ml or >1000 cp/ml ±0.3 Log 10 ). Samples with discrepant results were tested with a second gold standard HIV VL assay (Roche, Hologic or BioMerieux). Results: Limit of detection of the SAMBA II HIV-1 Semi-Q WB Test using Probit analysis is 1,037cp/ml (95% CI: 868-1,362cp/ml) with a 95% probability of detection at 1,000 cp/ml. Specificity was 100% in 500 HIV-1 negative samples (95% CI 99.4-100%). No cross-reaction or interference was observed with common ART drugs or samples containing other infections and potentially interfering substances. Concordance between SAMBA II WB POC assay and centralised gold standard assays in clinical samples at VL > or < 1,000 cp/ml ±0.3 Log 10 is given in the Table. Conclusion: The SAMBA II WB VL test is highly accurate using the WHO recommended > or < 1,000 cps/ml. Results can be obtained on site as soon as 95 minutes. Phlebotomy and centrifugation are not required, making SAMBA a simple ‘sample-in/results-out' true POC assay for remote settings

Poster Abstracts

953 MODELING POINT-OF-CARE NUCLEIC ACID TESTS (POC NAT) TO MINIMIZE HIV MISDIAGNOSES Anne M. Neilan 1 , Jennifer Cohn 2 , Emma Sacks 2 , Aditya Gandhi 1 , Patricia Fassinou 2 , Kenneth Freedberg 1 , Marc N. Kouadio 2 , Rochelle P. Walensky 1 , Andrea L. Ciaranello 1 1 Massachusetts General Hospital, Boston, MA, USA, 2 Elizabeth Glaser Pediatric AIDS Foundation, Washington, DC, USA Background: The World Health Organization (WHO) adult HIV diagnostic testing strategy requires up to 4-7 rapid diagnostic tests (RDTs) prior to ART initiation. Although more expensive than RDTs, adding POC NAT to current testing strategies may minimize misdiagnoses and attrition, permitting ART initiation with fewer tests. Methods: Using the Cost-Effectiveness of Preventing AIDS Complications model, we simulated a one-time HIV test in addition to status quo (SQ) testing practices in a low HIV-undiagnosed prevalence setting (1.3%): Côte d’Ivoire (CI). Model inputs included mean age (37y), SQ HIV testing (74 tests/1,000PY), and costs of ART ($6-22/m), HIV care ($27-38/m), and assays (RDT $1.50; POC NAT $27.92). We assessed 3 testing strategies: RDT-based strategies recommended by the WHO (RDT-WHO) and CI (RDT-CI), and a novel strategy: POC NAT to resolve RDT discordancy (NAT-Resolve). We calculated the number of true/ false negative/positive (TN, TP, FN, FP) results for each strategy. We modeled 3 scenarios: A) sensitivity/specificity fromWHO prequalification reports and no attrition between tests, B) sensitivity/specificity fromWHO prequalification reports and reported attrition and result-delay rates, and C) field-based RDT sensitivity/specificity and reported attrition and result-delay rates. We reported life expectancy (LE) and costs per misdiagnosis and per person in the tested population, as well as incremental cost-effectiveness ratios (ICERs, in $/year-of- life saved [YLS]; threshold ≤$1,720 [CI per-capita GDP]). Results: Relative to the tested population, there were fewmisdiagnoses in Scenarios A and B (Table 1). A FN diagnosis led to a LE loss of 5y (vs. a TP); this LE loss was most sensitive to HIV detection rates after developing an opportunistic infection. A FP diagnosis increased costs by $6,500 (vs. a TN); this cost increase

952 EVALUATION OF A TRUE POC VIRAL LOAD TEST: SAMBA LEUCO-DEPLETED WHOLE BLOOD HIV ASSAY Gary Brook 1 , Tetiana Stepchenkova 2 , Innocent M. Ali 3 , Sandra Chipuka 4 , Helen Lee 5 1 London North West University Healthcare NHS Trust, London, United Kingdom, 2 Kyiv city AIDS centre, Kiev, Ukraine, 3 University of Dschang, Dschang, Cameroon, 4 Harare Central Hospital, Harare, Zimbabwe, 5 Cambridge University, Cambridge, UK Background: To meet the UNAIDS 90:90:90 targets, patients in developing countries require better access to HIV viral load (VL) testing to confirm anti- retroviral (ART) treatment success. Current laboratory-based assays are confined to cities with high infrastructure, limiting access to patients in remote settings. The SAMBA II HIV-1 Semi-Q Whole Blood (WB) Test is a portable, robust, heat- stable, point-of-care (POC) VL assay that automatically leuco-depletes 3 logs of cellular RNA and DNA, thus providing clinically meaningful results (VL 1000 cp/ml) using a finger-prick blood sample.

CROI 2020 357