Comparison of two nucleic acid amplification technology (NAT) systems for detection of human immunodeficiency virus, hepatitis B and C virus

Abstract

Background

Human immunodeficiency virus (HIV) and hepatitis B virus (HBV) are endemic in South Africa (SA) while hepatitis C virus (HCV) infection is rare. Two NAT platforms, the Procleix Ultrio Elite assay on Panther (Elite) and the cobas MPX assay on cobas 6800/8800 Systems (MPX), are used worldwide. In 2015 these were evaluated in SA context.

Study design and methods

Sensitivity of HIV, HBV and HCV was evaluated using reference panels and two-fold dilutions of 51 positive plasmas tested in 12–24 replicates. The 95% and 50% lower limits of detection (LOD) were estimated by probit analysis and window period (WP) risk days by the Weusten model. Specificity was established by testing 3646 blood donations individually and instrument performance by evaluating all runs.

Results

Specificity was 99.94% for MPX and 99.97% for Elite. The following 95% LODs (95% confidence interval (CI)) were estimated for MPX and Elite respectively: HBV 17.8 (10.9–33.9) cp/mL and 47.9 (CI 29.1–92.4) cp/mL; HCV 21.9 (15.3–34.6) cp/mL and 13.8 (8.9–24.0) cp/mL; and HIV 8.3 (5.5–14.7) cp/mL and 10.4 (6.9–18.2) cp/mL. On SA HBV and HIV dilution panels relative sensitivity (range) of MPX was 3.20 (1.26–6.50) and 1.42 (0.26–2.72) fold higher than Elite. Downtime on cobas 6800 was 26 vs 6.6 hours on Panther (p<0.001). We estimated infectious WPs for HBV, HCV and HIV-1 at 13.8, 1.8 and 2.6 days for Elite and 10.3, 2.1 and 2.4 days for MPX.

Conclusion

Although MPX was significantly more sensitive for HBV, Elite was implemented due to instrument reliability during evaluation.

INTRODUCTION

Nucleic acid amplification technology (NAT) for blood and plasma screening was first introduced in minipool (MP) format in the 1990’s and commercial assays for hepatitis C virus (HCV) and human immunodeficiency virus (HIV) were implemented widely in 1999 ¹. From 2004 fully automated triplex NAT methods were introduced that included detection of hepatitis B virus (HBV) and allowed for donor screening in individual donation (ID) format ^{2, 3}. Rapid advances in terms of assay design and automation of testing processes were made during the past 15 years and nowadays more than 60 million donations are screened by NAT worldwide annually ⁴. NAT testing has since become the benchmark in blood screening and its ability to interdict transfusion transmissible infections is well described ^5–7.

Currently there are two NAT platforms used for blood screening worldwide, i.e. the Procleix Ultrio Elite assay (Elite) on the Panther instrument (Grifols Diagnostic Solutions, Inc., Emeryville, CA) and the cobas MPX assay (MPX) on the cobas 6800 (6800) or 8800 Systems (Roche Diagnostics GmbH, Mannheim, Germany) ^{8, 9}. Elite and MPX can both be used in Individual Donation (ID) or minipool (MP) format. The South African National Blood Service (SANBS) has used ID-NAT (Ultrio/Ultrio Plus) since 2005 to minimize the residual risk of HIV-1 and HBV transmission ^{2, 10, 11}. In contrast to HBV and HIV-1 infection which is endemic in South Africa the HCV prevalence is low.

In 2015 we compared the performance of the Elite on the Panther instrument with MPX on the in ID-NAT setting using South African donor samples.

MATERIALS AND METHODS

We compared the performance of Elite and MPX on South African donor samples and reference standards. The performance characteristics stated in the package inserts of the two NAT assays and systems are listed in supplemental Table 1 ^{12, 13}. The study was approved by the SANBS Human Research Ethics Committee and conducted in two parts. The first involved screening of donor samples on MPX and Elite in parallel with the then routinely used Ultrio Plus assay (Plus) (Grifols Diagnostic Solutions, Inc., Emeryville, CA). The second part involved testing of sensitivity panels. Instrument performance was evaluated using all runs.

Donor screening study (part 1)

Specimen selection and collection

A 9ml EDTA sample is routinely collected from all blood donors for NAT testing. For this study an additional 9ml EDTA sample was collected from 3846 donors in one geographical region of SANBS during February to March 2015. Test results for 3638 and 3815 donor samples on MPX and Elite respectively were obtained (difference due to failures). Specificity and possible NAT yields were reported.

Specificity based on one single screening test was calculated by the following formula:

$$\text{NR}/(\text{NR}+\text{NRR})$$

whereby $NR=NAT$ nonreactive and $NRR=NAT$ nonrepeat reactive

Algorithms for confirmation

The repeat testing algorithm is shown in Figure 1. An initial reactive followed by one more NAT reactive was considered a NAT repeat reactive (i.e. for MPX, if one of the three repeats were reactive; for Elite or Plus, if either of the repeats or the discriminatory test were reactive). For HBV test discrepancies between Plus, Elite and MPX, we performed replicate (x5) NAT (Plus) testing as well as hepatitis B serology testing (Roche Elecsys anti-HBc IgM, anti-HBc total and anti-HBs, Roche Diagnostics GmbH, Mannheim, Germany, Abbott Prism HBsAg, Abbott PRISM ChLIA system, Abbott, Delkenheim, Germany) on archive samples derived from the frozen plasma unit. Detection of HBV DNA on the plasma bag sample was considered a confirmed positive result.

Figure 1: Repeat testing algorithm for testing of donor samples.

Supplementary testing on plasma bag (Only HBV discrepancies were present during the study): HBV: 5x NAT (Plus), anti-HBc Total, anti-HBc IgM, anti-HBs, HBsAg; Interpreted as confirmed if DNA detectable on any system

Repeat testing after initial reactives was performed on the plasma bag samples if insufficient plasma was available in the original EDTA samples

Analytical sensitivity study (part 2)

External reference panels

Analytical sensitivity was estimated through the use of standard dilution panels (BioQControl, Heiloo, Netherlands). Eight members of the reference panels P0027 HIV-1 subtype C, P0007 HBV genotype A and P0019 HCV genotype 1 were tested in 24 replicates by each NAT system (Figure 2). The standards for these panels are calibrated in both copies/mL (Siemens bDNA 3.0 assay) and IU/mL (1^st and 2^nd WHO standards). Calibration of the HIV-1 subtype C against subtype B standard was based on the Abbott Realtime assay ¹⁴.

Figure 2:

Sensitivity panels run in parallel using MPX and Elite

In-house analytical sensitivity panels

Ten concordant positive donations (NAT repeat reactive, serology reactive) for HIV, HBV and HCV respectively were diluted with pooled plasma (tested NAT nonreactive in multiple replicates on MPX and Elite) to a viral load (VL) of 300 cps/mL followed by two-fold dilutions (1:1 to 1:64) (Figure 2). A similar two-fold dilution series were prepared from previously identified NAT yield (NAT repeat reactive, serology negative) donor plasmas (15 HIV, 15 HBV and 7 HCV). The VL in undiluted samples was measured at the time of collection in singular test using the Abbott RealTime assay and if below the quantification limit by probit analysis using MPX and Elite replicate test data and comparison against the proportion of reactive results on the BioQControl standard dilution panels as reference ¹⁵.

Each panel member of the SANBS plasma dilution panels were tested in 12 replicates on MPX and Elite.

Sensitivity analysis in serology yield samples

Ten or 11 serology yield samples (serology confirmed reactive but NAT negative on initial screen) for HIV, HCV and HBV respectively were tested undiluted in 12 replicates.

Comparative instrument performance (in both study parts)

Spreadsheets of all instrument runs were populated to assess invalid tests, instrument breakdowns and downtime. The mean-time-between-failure (MTBF), mean–time-to-repair (MTTR) and instrument availability were calculated based on all samples tested during the study (part 1&2) (Table 5)

Table 5:

Comparative instrument performance based on a total of 12773 tests on Panther and 13714 on cobas 6800

Measurement	cobas 6800	Panther	p value
	N (%) or Mean (SD)
Invalid tests N (%) (1)	540 (4.2%)	254 (1.9%)	<0.001
Breakdowns during evaluation	9	4	0.27
Mean time between failures (MTBF) (hours) = Total uptime/Number of breakdowns	171 h	400 h	N/A
† Mean time to repair (MTTR) (hours) = Total downtime/Number of breakdowns	2.9 h	1.6 h	N/A
Instrument availability % = MTBF/ (MTBF+MTTR)	98.34%	99.59%	N/A

^†Engineers were often on site as the Grifols Tigris system in use at the time was supported by the same company.

An additional 30 minutes added per call out to correct for this bias.

Statistical analysis

Analytical sensitivity

Parallel line probit analysis was performed to determine the 95% and 50% LODs of MPX and Elite on the BioQControl and SANBS HIV, HBV and HCV dilution panel results using SPSS v23.0 software (IBM Corporation, New York, USA). Of each dilution panel the relative sensitivity (with 95% CI) was calculated by comparing the difference in 50% LOD between the two assays. Where the 95% confidence intervals remain above or below 1.0 the difference in analytical sensitivity was significant (p<0.05).

Window period estimation

The infectious WP for HIV, HBV and HCV was calculated using the Weusten model ¹⁶ based on the 50% and 95% LODs found on the BioQControl reference panels. The HBV doubling time of 2.6 days was assumed as previously estimated for HBV ¹⁷, 0.85 days (20.5 hours) for HIV and 0.45 days (10.8 hours) for HCV ^{18, 19}. We calculated WPs for HBV, HCV and HIV-1 using a 50% minimum infectious dose (MID₅₀) of 1 and 3.16 virions per 20 mL plasma (the estimated residual plasma volume in a red blood cell (RBC) unit). For HIV-1, we also calculated the infectious WP using a MID₅₀ of 31.6 virions because infectivity of HIV-1 likely diminishes in stored RBCs ¹⁰.

Serology yields

The proportion of reactive results on serology yield samples cumulatively on MPX and Elite were compared using the chi-square test.

Instrument performance

The differences in invalid tests, were analyzed by using the chi-square test.

RESULTS

Test performance on donor samples

Specificity of 99.94% (CI 99.80%−99.99%), 99.97% (CI 99.86%−100%) and 99.89% (CI 99.73%−99.97%) was calculated for MPX, Elite and Plus respectively (p=0.99). No discordant results were found for HIV and HCV. Seven HBV NAT Yield (NAT+/HBsAg-) donor samples were detected by MPX and Elite combined which were not detected by Plus. Of these, five were HIV co-infections that tested both HIV RNA and anti-HIV positive but was not detected by Plus discriminatory HBV (dHBV) test. Three of the 5 HIV co-infected samples were detected by both MPX (HBV) and Elite dHBV, one by MPX (HBV) only and one by Elite dHBV only (Table 1). Both these samples had detectable HBV DNA on replicate (x5) Plus dHBV testing on the plasma bag sample, however anti-HBc and anti-HBs tested negative indicating either possible WP HBV infections or contamination of the plasma bag sample.

Table 1:

Test performance (MPX vs Elite) on routine donor samples in comparison to Procleix Ultrio Plus

Measureable	cobas MPX	Ultrio Elite	Ultrio Plus
Number of samples run per system	3638	3815	3790
Specificity	99.94% (CI 99.80%−99.99%)	99.97% (CI 99.86%−100%)	99.89% (CI 99.73%−99.97%)
Viral positive results detected in donors
HIV RNA reactive/anti-HIV reactive (HIV concordant)	12	12	12
HBV DNA reactive/HBsAg reactive (HBV concordant)	7	7	7
HCV RNA reactive/anti-HCV reactive (HCV concordant)	1	1	1
Total HBV DNA reactive/HBsAg negative (HBV Yields)	6 of 7	4 of 7	0 of 7
Donor 1	HIV/HBV	HIV/HBV	HIV
Donor 2	HIV/HBV	HIV/HBV	HIV
Donor 3	HIV/HBV	HIV/HBV	HIV
Donor 4	HIV/HBV	HIV only	HIV
Donor 5	HIV only	HIV/HBV	HIV
Donor 6	HBV	Not detected	Not detected
Donor 7	HBV	Not detected	Not detected

Two additional donor samples had detectable HBV DNA by MPX only. One was non-repeat reactive on Plus (1^st screen only reactive), had no detectable HBV DNA by replicate Plus dHBV testing (0/5) but tested anti-HBc and anti-HBs positive indicating a possible low viremic occult HBV infection (OBI). Replicate (x20) testing yielded 6/20 reactives on MPX vs 1/20 on Elite. This was estimated to correspond with a viral load of 0.9 (0.6–1.2) cp/mL by probit analysis based on the MPX reactivity. The second specimen was initially non-reactive on Plus, had detectable HBV DNA by replicate Plus dHBV testing but tested anti-HBc and anti-HBs negative indicating possible HBV WP infection. Replicate (x20) testing using new samples from the plasma bag showed 3/19 reactives on MPX vs 1/20 on Elite. This was equivalent to a viral load of 0.6 (0.4–0.8) cp/mL by probit analysis also based on MPX reactivity. This donor returned for retesting 46 months later. No HBV DNA could be detected by Elite (single screen only) and the donor remained anti-HBc and anti-HBs negative.

Analytical sensitivity using reference panels

Table 2 compares the 95% and 50% LODs on the BioQControl reference panels. The relative sensitivities (95% CI) of MPX compared to Elite (based on the shift in parallel probit curves) for HBV, HCV and HIV were 2.68 (1.50–5.40), 0.63 (0.38–0.98) and 1.24 (0.80–2.03) respectively.

Table 2:

Analytical Sensitivity of Ultrio Elite and cobas MPX as determined by parallel line probit analysis on 24 replicate tests per standard dilution

BioQ reference panel §	Limit of detection (LOD) in copies/mL
	cobas MPX 50% LOD (95% CI)	Ultrio Elite 50% LOD (95% CI)	cobas MPX 95% LOD (95% CI)	Ultrio Elite 95% LOD (95% CI)	Sensitivity of MPX relative to Elite
P0007 HBV Genotype A	1.9 (1.3–2.7)	5.0 (3.4–7.3)	17.8 (10.9–33.9)	47.9 (29.1–92.5)	2.68 (1.50–5.40)
P0019 HCV Genotype 1	3.0 (2.3–3.8)	1.9 (1.3–2.7)	21.9 (15.3–34.6)	13.8 (8.9–24.0)	0.63 (0.38–0.98)
P0027 HIV-1 Group M subtype C^	1.7 (1.2–2.4)	2.1 (1.5–2.9)	8.3 (5.5–14.7)	10.4 (6.9–18.2)	1.24 (0.80–2.03)

^§calibrated in cp/mL in bDNA 3.0 assay and in IU/mL against 1^st and 2^nd WHO standards with conversion factors of 5.33 cp/IU for S0011 HBV-DNA genotype A, 2.73 cp/IU for S0009 HCV genotype 1 and 0.58 for S0012 HIV-1 subtype B standard;

^{^}S0014 HIV-1 subtype C standard was calibrated in cp/mL against S0012 BioQ HIV-1 subtype B standard in Abbott RealTime assay.

Analytical sensitivity using SANBS NAT positive plasma dilution panels

Table 3 summarizes the analytical sensitivity data on the two-fold dilution series from SANBS HIV, HBV and HCV plasma samples as detailed in supplemental tables 2, 3 and 4 respectively. The geometric mean of the sensitivity factor of MPX relative to Elite obtained on 24 SANBS HIV-1 RNA positive samples was 1.42 (range 0.26–3.56). One HIV sample had significantly increased sensitivity by Elite (p<0.05) and eight were detected slightly more sensitive by MPX (p<0.05). For HBV DNA, the relative sensitivity (geomean) was 3.20 (range 1.26–6.50). MPX was significantly more sensitive for HBV DNA detection in 21 of the 23 SANBS in-house dilution panels. For HCV RNA, only data on four HCV-RNA positive samples could be analysed and the geomean relative sensitivity was 0.87 (range 0.75–0.98). The difference in analytical sensitivity was not significant on any of the four HCV-RNA positive samples.

Table 3.

Geometric mean values of 50% and 95% LODs (and 95% confidence intervals (CIs)) and geometric mean values of sensitivity factors of Cobas MPX relative to Ultrio Elite (and 95% CI) calculated from parallel line probit analysis results on 51 two-fold SANBS plasma dilution panels tested in 12 replicates per dilution (individual data presented in supplemental tables 2,3 and 4).

In-house plasma dilution panel §	Number of panels tested	MPX 50% LOD (95% CI)	MPX 95% LOD (95% CI)	Elite 50% LOD (95% CI)	Elite 95% LOD (95% CI)	Relative sensitivity MPX/Elite (CI)	Range of Relative sensitivity MPX/Elite
HIV-1 RNA	24	2.4 (1.0–4.1)	12.5 (7.0–36)	3.4 (1.7–3.4)	17.9 (10.4–56)	1.42 (0.68–3.87)	0.26–2.72
HBV DNA	23	1.1 (0.3–3.9)	6.1 (3.6–11.4)	3.4 (2.1–5.1)	19.5 (11.8–49)	3.20 (1.54–11.6)	1.26–6.50
HCV RNA	4	5.1 (2.4–9.0)	29.3 (15.4–133)	4.4 (2.0–8.0)	25.6 (13.5–112)	0.87 (0.34–2.10)	0.75–0.98

^§two-fold dilutions were prepared of 15 NAT yield samples with viral load quantified by Abbott RealTime assay or by probit analysis against BioQControl standards. In addition, 4–9 concordant serology and NAT positive samples were pre-diluted to 300 cp/mL for preparation 2-fold dilutions. Viral loads of NAT yield samples varied between 3.3 cp/mL and 295 cp/mL for HIV-1 and ranged from 18 to 267 cp/mL for HBV. (See supplemental tables 1, 2 and 3).

Sensitivity in detecting serology yield samples

The cumulative reactivity rate on seven anti-HIV yield samples showed a significantly higher reactivity rate on MPX than Elite, i.e. 38/84 (45.2%) versus 12/84 (14.3%) (p<0.0005) (Supplementary table 5). The cumulative reactivity rate on seven HBsAg yield samples was also significantly higher in MPX than Elite assay, i.e. 69/84 (82.1%) versus 51/84 (60.7%) (p=0.002). There was no HCV RNA detected in any of eight anti-HCV serology yield samples by either MPX or Elite. The data on some samples could not be interpreted because of co-infection with another marker (4 for HIV, 1 for HBV and 8 for HCV) and two HBsAg yield samples were excluded from the analysis because all 12 replicate tests were reactive on both MPX and Elite.

Estimation of window periods

Table 4a gives the WP estimates for Elite and table 4b for MPX using ID-NAT or different minipool (MP)-NAT options based on the analytical sensitivity data obtained from the BioQControl reference panels (table 2). According to these results MPX reduces the infectious WP for HBV by 3.5 days as compared to Elite. The data show the impact of dilution of test samples in MP6, MP16 or MP24 format for Elite and MPX on lengthening of the infectious WPs.

Table 4.

Estimation of infectious WPs based on LODs on P0027 HBV genotype A, P0019 HCV genotype 1 and P0027 HIV-1 subtype C standard dilution panels based on the 50% LODs found on the BioQControl reference panels

Virus	MID50 (virions)	WP ID-NAT (days)	WP MP6-NAT (days)	WP MP16-NAT (days)	WP MP24-NAT (days)
HBV	1	18.1	25.5	28.7	30.1
HCV	1	2.5	3.7	4.3	4.6
HIV-1	1	4.0	6.3	7.4	7.9
HBV	3.16	13.8	21.2	24.4	25.8
HCV	3.16	1.8	3.0	3.6	3.8
HIV-1	3.16	2.6	4.9	6.0	6.5
HIV-1	31.6	0.66	2.1	3.2	3.7

Virus	MID50 (virions)	WP ID-NAT (days)	WP MP6-NAT (days)	WP MP16-NAT (days)	WP MP24 NAT (days)
HBV	1	14.5	21.8	25.1	26.5
HCV	1	2.8	4.0	4.6	4.9
HIV-1	1	3.7	6.0	7.1	7.6
HBV	3.16	10.3	17.5	20.8	22.2
HCV	3.16	2.1	3.3	3.9	4.1
HIV-1	3.16	2.4	4.6	5.7	6.2
HIV-1	31.6	0.56	1.9	2.9	3.4

Comparative instrument performance

Instrument performance indicators are presented in Table 5. The 6800 had 540 (4.2%) invalid tests compared to 254 (1.9%) with Panther (p<0.001). All invalid tests were due to instrument failure and none to detection failure of internal controls. Nine breakdowns (event leading to instrument not being able to continue testing due to hardware or software failure) occurred on the 6800 (26 hours downtime) compared to four (6.6 hours downtime) on Panther (p<0.001). The mean-time-between-failure for the 6800 was 171 hours compared to 400 hours on Panther.

DISCUSSION

According to the results obtained on dilutions of the HBV DNA genotype A2 reference panel (P0007, BioQControl) and 23 SANBS HBV (likely genotype A1) plasma samples in this study, the MPX assay is 2.7 (1.5–5.4) or 3.2 (1.3–6.5) fold more sensitive than the Elite assay. The 50% and 95% LODs on the reference panel were estimated at 1.9 (1.3–2.7) and 17.8 (10.9–33.9) cp/mL versus 5.0 (3.4–7.3) and 47.9 (29.1–92.5) cp/mL for MPX and Elite respectively. MPX in ID-NAT format would potentially reduce the infectious WP by 3.5 risk days from 13.8 to 10.3 days as compared to Elite (assuming a MID₅₀ of 3.16 virions/20 mL plasma in RBCs).We estimated that the infectious WPs estimated for HBV ID-NAT would increase by another 7.2 and 10.6 days in MP6 and MP16-NAT setting respectively.

The higher sensitivity of MPX in ID-NAT format also became apparent in the donor screening study, which yielded two more HBV NAT yields (1 probable anti-HBs positive OBI and 1 possible WP case) with extreme low viral loads of 0.9 (0.6–1.2) and 0.6 (0.4–0.8) cp/mL according to reactivity rates of 6/20 (30%) and 3/19 (16%) by replicate testing with MPX on the plasma unit. The blood components could have been used for transfusion if not detected in this study. The RBCs of the OBI donation would contain 18 (12–24) HBV virions, which is well below the MID₅₀ of 316 (100–1000) virions estimated for anti-HBs negative OBIs in animal and human infectivity studies ²⁰. However, it was near the incidentally observed OBI minimum infectious dose of ~30 virions demonstrated in a SANBS RBC transfusion ²¹ and possibly another case reported by Candotti et al ²². The probability of infectivity of 12 (8–16) virions in the RBC transfusion from the presumed WP donation is expected to be higher since the worst case MID₅₀ is estimated 3.16 (1–10) virions ^{16, 20}. However, HBV marker testing in a follow up sample taken 46 months later did not confirm the infection in the donor. This indicates that either the infection in the donor was aborted or the source of the observed low MPX reactivity could have been contamination of the primary test tube and/or plasma sample. This stresses the importance of follow up samples in confirming WP infection in donors, especially where extreme low viral loads are detected. The higher sensitivity of MPX for HBV detection was also supported by replicate testing of HBsAg serology yield samples.

The geometric mean values of the 50% and 95% LODs of the 24 SANBS HIV plasma dilution panels were 2.4 (1.0–4.1) and 12.5 (7.0–36.2) cp/mL for MPX versus 3.4 (1.7–5.3) and 17.9 (10.4–56.1) cp/mL for Elite which were comparable to those found on the HIV-1 subtype C reference panel. Hence, the MPX assay is as sensitive as the Elite assay and reached somewhat higher analytical sensitivity on eight of 24 South African HIV-1 (likely subtype C) samples. A higher sensitivity of MPX was also observed on nine anti-HIV serology yield samples. In ID-NAT setting, there is no significant difference between the two assays in terms of the residual risk of HIV transmission with infectious WPs estimated at 2.4 versus 2.6 risk days for MPX and Elite respectively. Since HIV infectivity in stored RBCs is likely reduced 10-fold ¹⁰, we estimate WP risk day equivalents of 0.7 for Elite and 0.6 for MPX ID-NAT, which would increase to approximately 2 and 3 risk days for MP6 and MP16-NAT respectively.

For HCV we found Elite to be slightly more sensitive on the HCV reference panel by a factor of 1.6 (1.0–2.6), but on four local HCV (likely genotype 5) samples there was no difference in relative sensitivity. On eight anti-HCV positive serological yields samples tested in 12 replicates no HCV-RNA could be detected. Similar data were found in Egypt on 175 samples and in an efficacy study covering seven geographical regions ^{7, 23}. Hence, the clinical utility of anti-HCV testing in ID-NAT setting could be re-assessed.

Our study had some limitations. Firstly, only one standardized reference panel was tested per viral marker and analytical sensitivity therefore do not apply to all genotypes of the viruses. However, the subtypes tested in the study are those most prevalent in sub-Saharan Africa where HIV prevalence remains the highest in the world. Secondly, a potential bias exists in the in-house NAT yield samples used as these were originally identified by a Grifols assay. However, we believe we mitigated this by introducing a contrasting bias with the serology yield in-house panels, as these are all donations that were originally not detected using Grifols technology on initial screening. The large number of NAT yield and serology yield samples used in this study distinguishes this study from others and potentially adds insight on assay performance. A third limitation relates to operator inexperience on new instrumentation, especially on the 6800.

In conclusion, our study has given more insight in the performance characteristics of two widely used fully automated NAT systems. Despite significantly higher sensitivity of MPX for HBV detection (and marginally higher sensitivity on some of the samples for HIV detection), we decided to implement Elite on the Panther system. Chronic blood shortages and the concerns related to instrument reliability influenced this decision by SANBS in December 2015. No additional data on instrument performance on the cobas 6800 or 8800 instruments were available globally at that time as the system was only launched in August 2014. Significant hardware and software changes were made since 2015 and the cobas systems are now successfully being used by many blood screening centers.