CROI 2016 Abstract eBook

Abstract Listing

Poster Abstracts

CEPHIA repository were tested. All assays were run in duplicate for each sample type. To examine the degree of variability within and between results for each sample type, both categorical and continuous results were analyzed. Bland Altman, R 2 values and Cohen’s kappa coefficient (ĸ) were used to assess correlation. Results: Intra-assay variability using the same sample type was high for all assays R 2 0.96 to 1.00. The R 2 values comparing DBS and plasma results for LAg-Avidity, BED-CEIA, and BioRad Avidity were 0.96, 0.93, and 0.84, respectively. The concordance and ĸ values between DBS and plasma for all three assays was >92% and >0.82, respectively. The Bland- Altman analysis showed significant differences between plasma and DBS samples, see figure. For all three assays, a higher number of samples were classified as recent using DBS samples, see figure. Seven DBS samples were excluded from the BioRad Avidity analysis due to protocol guidelines. Conclusions: Data generated using DBS and plasma samples was highly correlated. However, there is a scalar problem as the assays perform differently across their dynamic range. DBS samples were more likely to be classified as recent by all three assays, which may lead to over estimation of incidence in surveys using performance criteria derived for plasma samples. These results suggest that the performance characteristics of these incidence assays may be different than those calculated using plasma samples. LAg-Avidity BED-CEIA BioRad Avidity

Bland-AltmanPlot:PlasmaandDBS

Bland-AltmanPlot:PlasmaandDBS

Bland-AltmanPlot:PlasmaandDBS

-1 0 1 2 3 Difference (Plasma-DBS LAgAvidityResults) 0 1 2 3 4

-40 -20 0 20 40 60 Difference (Plasma-DBSAvidity Index)

-1 0 1 2 Difference (Plasma-DBSBED-CEIAResult) 0 0.5

7

6

5

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5

0 10 20 30 40 50 60 70 80 90 100 Average ofPlasmaandDBSAvidity Index

Average ofPlasmaandDBSBED-CEIAResult

Average ofPlasmaandDBSLAg-AvidityResult

Duration of Infection

Duration of Infection

Duration of Infection

<1Year >1Year Total

<1Year >1Year Total

<1Year >1Year Total

ODn<1.5 13 ODn>1.5 12

15 60 75

28 72

ODn<0.8 15 ODn>0.8 10

17 58 75

32 68

AI<50% 21 AI>50% 4

11 64 75

32 68

Total

25

100

Total

25

100

Total

25

100

PLASMA

Concordance= 73% kappa = 0.31 (p<0.01)

PLASMA

PLASMA

Concordance= 85% kappa = 0.63 (p<0.01)

Concordance= 73% kappa = 0.34 (p<0.01)

Duration of Infection

Duration of Infection

Duration of Infection

<1Year >1Year Total

<1Year >1Year Total

<1Year >1Year Total

ODn<0.8 16 ODn>0.8 9

21 54 75

37 63

ODn<1.5 14 ODn>1.5 11

19 56 75

33 67

AI<50% 18 AI>50% 4

14 57 71

32 61 93

DBS

DBS

DBS

Total

25

100

Total

25

100

Total

22

Concordance= 70% kappa = 0.31 (p<0.01)

Concordance= 70% kappa = 0.28 (p<0.01)

Concordance= 81% kappa = 0.54 (p<0.01)

512

Laboratory Evaluation of the GeneXpert® HIV-1 Viral Load Assay in Zimbabwe Zibusiso Ndlovu 1 ; Sekesai Mtapuri-Zinyowera 2 ; Carol Metcalf 3 ; Emmanuel Farjado 3 ; Kekeletso Kao 4 ; Maryam Rumaney 3 ; Daniel Orozco 4 ; Elton Mbofana 1 ; Helen Bygrave 3 1 Médecins Sans Frontières, Harare, Zimbabwe; 2 Natl Microbiology Reference Lab, Harare, Zimbabwe; 3 Médecins Sans Frontières, Southern Africa Med Unit, Cape Town, South Africa; 4 FIND, Geneva, Switzerland Background: In most resource limited countries with a high HIV burden, current Viral Load (VL) testing technologies are based at the centralized laboratory levels. New HIV VL point-of-care (POC) tests hold the potential to simplify ART monitoring and bring diagnostics closer to the patient. The objective of the study was to perform a laboratory validation of the diagnostic accuracy of the GeneXpert ® HIV-1 Quant assay compared to the predicate laboratory reference standard BioMérieux NucliSens Version 2.0 Easy Q/Easy Mag (NucliSens v2.0) for VL testing Methods: EDTA anticoagulated plasma samples from 392 consenting patients on ART were tested for HIV-1 VL at the National Microbiology Reference Laboratory (NMRL) in Harare, Zimbabwe from June-September 2015. Paired samples were tested with each reference and index test following the manufacturer’s instructions and laboratory SOPs. The technician performing the Xpert VL test was blinded to the result of the reference test. Statistical analysis was conducted in STATA 13.0 and included sensitivity, specificity, positive predictive value (PPV), linear regression and Bland-Altman (BA) analysis. Results: Of the 392 samples were tested in the Xpert ® HIV-1 Quant assay, 375 produced a VL result (3.4% true error rate). The sensitivity of the Xpert ® HIV-1 Quant assay in detecting HIV VL as at 1000copies/ml threshold compared to the NucliSens v2.0 assay was 98.2% [95% CI: 93.8-99.8], specificity 97.7% [95%CI: 95.1-99.2] and PPV 94.8%while NPV 99.7%. Bland-Altman mean bias between the two methods was -0.22 log 10 copies/ml [95% CI: -0.292 to -1.52] and the LOA for the bias were -1.11 to 0.66; p-value=0.125. Spearman’s correlation coefficient, R 2 =0.952. Concordance between the two assays at 1000copies/ml threshold was 97.8%. Only 1.6% of patients (6 of 375) and 0.5% (2 of 375) were misclassified above and below this threshold respectively by the index assay compared to predicate assay. The system proved easy-to-use with minimum sample preparation and handling time, with an average daily capacity of 16 samples per 8hour day Conclusions: Xpert® HIV-1 Quant gives HIV VL results comparable to NucliSens v2.0 assay. Findings from this study support the consideration of Xpert® HIV-1 Quant testing for the roll-out of HIV VL monitoring and may contribute towards reaching the ambitious UNAIDS target of 90% ART patients with undetectable VL

Poster Abstracts

198

CROI 2016