Reassessment of hepatitis B virus window periods for two transcription mediated amplification assays using screening data of South African blood donors

Abstract

Background

Transcription mediated amplification (TMA) assays for HBV DNA detection have transitioned from the Ultrio to the Ultrio Plus assay which features increased analytic sensitivity due to inclusion of a target enhancer reagent. The impact on HBV detection for different categories of HBV infection has not been fully evaluated.

Methods

HBV-DNA and HBsAg detection rates as well as viral load (VL) distributions in HBV-NAT yield samples were compared during one year of screening of South African blood donors with the Ultrio assay and the subsequent year by the Ultrio Plus version. HBV-DNA concentration at the HBsAg seroconversion point was established by regression analysis using a set of anti-HBc negative acute viremic samples.

Results

Ultrio Plus detected two-fold more window period (WP) NAT yield donations and 1.7-fold more occult HBV infections than Ultrio. The VL distribution data indicated that Ultrio not only missed samples below 100 copies/mL, but also a substantial number above this level. The VL at the HBsAg seroconversion point was estimated at 916 copies/mL whereas the VL at the NAT-conversion points was calculated at 63 and 4.1 copies/mL for Ultrio and Ultrio Plus. This reduced the infectious WP (compared to HBsAg testing) by 10.3 and 20.4 days, respectively.

Conclusion

The higher than expected rise in HBV-NAT yields after introduction of the Ultrio Plus assay is likely attributable to variable sensitivity of the former Ultrio assay for different HBV samples. Therefore, previously published HBV WP reduction and residual risk estimates based on analytical sensitivity of the Ultrio assay need to be revised.

INTRODUCTION

The South African National Blood service (SANBS) introduced individual donation nucleic acid amplification technology (ID-NAT) in October 2005 using the Procleix Ultrio assay. In April 2011 this transcription mediated amplification (TMA) method was replaced by the Procleix Ultrio Plus assay, which has a higher sensitivity for hepatitis B virus (HBV) DNA due to a target enhancer reagent^1,2. When comparing the last year of Ultrio screening with the first year of Ultrio Plus screening a higher than expected increase of HBV-NAT yield donations was observed in both window period (WP) and occult HBV infection (OBI) stages, while the HBsAg detection rate in first time and repeat donors did not change. The observed two-fold increase in the HBV WP NAT yield rate could not be explained by the previously reported 2.4-fold (1.4–4.8 fold) higher analytical sensitivity of the Ultrio Plus assay¹, which would have only marginally increased the pre-HBsAg NAT detection period by 3.2 days (1.3–5.9 days) from 17.2 to 20.4 days^3,4 and therefore would have predicted a 1.2-fold increase (20.4 days divided by 17.2 days) in the WP-NAT yield rate. To explore the reason for the unexpectedly high rise in HBV NAT yield rate we compared the viral load (VL) distributions in WP and OBI NAT yield samples in the Ultrio and Ultrio Plus screening periods, respectively. In addition, we analysed the VL distribution in the HBV ramp-up phase (HBV DNA positive/anti-HBc negative) to estimate the VL at the HBsAg seroconversion point which allowed for estimation of the differential pre-HBsAg detection period of the two NAT assays.

METHODS

HBV screening data

In April 2011 SANBS replaced the Ultrio assay by the Ultrio Plus assay without changing the previously described confirmatory testing algorithms for donations with discrepant HBV-DNA and HBsAg test results⁵. We compared the HBV screening data in the year before and after introduction of the Ultrio Plus assay. In the last year of screening with the Ultrio assay 775,444 donations were tested, whereas in the year thereafter 789,948 donations were screened by the Ultrio Plus assay.

HBV infection categories

Classification into different infection categories took place on the basis of the quantitative (q)PCR (Roche TaqMan assay) and supplemental serologic test results in index donation and follow up samples as described elsewhere⁵. The following infection categories were distinguished:

1. Pre-HBsAg WP NAT yield infections. These were further subclassified as:

   1. Confirmed WP: If in a follow up sample, HBsAg or anti-HBc became reactive.
   2. Probable WP: If no follow up data were available but VL in the index donation was above the lower limit of quantification (LLQ) of 116 copies/mL in the qPCR assay, or when qPCR was negative (or below LLQ) and all five replicate discriminatory (d) HBV assays on archive samples taken from the Fresh Frozen Plasma (FFP) unit were reactive
   3. Possible WP: If qPCR result was negative (or below LLQ) on the index donation but when at least 2 of 5 replicate dHBV tests on the FFP unit showed HBV-DNA reactivity.
   4. Possible WP or false positive: as above but with only 1 of 5 replicate dHBV tests reactive on FFP sample
   5. Confirmed anti-HBs breakthrough/abortive infection: If in a follow up sample, a significant rise in pre-existing anti-HBs titer and/or anti-HBc seroconversion was established regardless of whether the donor had been vaccinated for HBV or had been immunized by natural exposure to the virus.
   6. Possible anti-HBs breakthrough/abortive infection: If no rise in anti-HBs titer was seen in a follow up sample or when no follow up samples were available. Note that this pattern can also be seen in chronic OBI with anti-HBs as the sole detectable serologic marker⁶.

   Note that some of the anti-HBc and HBsAg negative acute phase WP NAT yield cases with a high viral load (that would be expected to be HBsAg positive) could be acute occult HBV infections that would not have become HBsAg reactive at any (or only for a very short) time during the acute viremic phase.^7,8,9

2. Post HBsAg WP NAT yield infections: If HBV-DNA and IgM anti-HBc were found reactive.

3. HBsAg and HBV-DNA concordantly positive

4. HBV-DNA nonreactive HBsAg positive: If HBsAg and anti-HBc were reactive but ID-NAT was nonreactive in the initial screening test

5. OBI: If HBV-DNA was confirmed positive in conjunction with anti-HBc.

VL estimations in NAT yield samples

VL was measured by Roche TaqMan qPCR assay. If the qPCR result was below LLQ (<116 copies/mL) we determined the proportion of reactive ID-NAT results in a total of 9 replicate tests, i.e. three Ultrio (or Ultrio Plus) multiplex replicates on plasma aliquots from the primary NAT tube and six discriminatory (d)HBV replicates (using the Ultrio or Utrio Plus dHBV assay depending on the screening period), one on the primary tube and five on fresh frozen plasma (FFP)-derived samples. The proportion of reactive ID-NAT results was compared with probit curves established based on replicate testing of serial dilutions of the Eurohep standard with 50% and 95% LODs of 9.3 and 90.2 copies/mL for Ultrio and 4.5 and 43.1 copies/mL for Ultrio Plus, respectively¹⁰. In those cases where 9/9 ID-NAT replicates were reactive for HBV DNA, VL load was estimated from the reactivity detected on 1:4, 1:8 and 1:16 dilutions of the FFP sample tested in a single replicate. In this circumstance, the midpoint between the last positive and negative dilution was taken as the endpoint titer and was multiplied by the 50% LOD on the Eurohep standard to roughly estimate VL.

Estimation of VL at HBsAg seroconversion point in ramp up phase of viremia

Acute viremic donor samples collected by SANBS from October 2014 until October 2015 in the anti-HBc negative phase of infection with and without HBsAg reactivity in the PRISM chemiluminescence assay (CLIA) were used to estimate VL at the HBsAg seroconversion point by regression analysis as performed in other studies^11–14. Log values of VL in copies/mL measured in the Abbott real time HBV VL assay were plotted against log HBsAg S/CO ratios (Abbott PRISM) and the VL at the HBsAg cutoff crossing point and 95% confidence limits of the HBsAg LOD were calculated. The estimated value was compared with a PRISM HBsAg seroconversion point of 960 copies/mL that was previously established by regression analysis in five HBV seroconversion panels^13,14.

Estimation of HBV-DNA conversion point with Ultrio and Ultrio Plus assay

Studies based on performing multiple replicate NAT assays on serial samples from five HBV seroconversion panels^13,14 showed that the 50% HBV-DNA conversion point of the Ultrio Plus assay was comparable to the 50% LOD on HBV genotype A standard dilution series. The 50% Ultrio Plus conversion point (and 95% confidence interval (CI) in the five seroconversion panels was estimated at 4.6 (3.4–6.3) cp/mL comparable to the 50% lower limit of detection (LOD) estimate of 4.3 (2.9–6.1) cp/mL on the P0007 HBV-DNA genotype A standard dilution series (Biologicals Quality Control, Heiloo, the Netherlands). However, for the Ultrio assay the analytical sensitivity data were more variable which made it more difficult to establish a 50% HBV-DNA conversion point that is representative for this assay. In a previous publication¹⁰ we found a large difference in reactivity rates on (dilutions of) 107 HBV Ultrio NAT yield samples and on 32 HBsAg and anti-HBc positive samples that were initially HBV-DNA nonreactive in the Ultrio assay (serology yield samples). This is illustrated in supplemental figure 1 which shows the previously reported replicate-NAT testing data and the associated probit analysis results on (dilutions of) the HBsAg positive serology yield samples as compared to those on the Ultrio NAT yield samples¹⁰. The data indicated that the Ultrio assay had inconsistent reactivity rates on the two categories of low VL samples, and that the difference in analytical sensitivity disappeared when the Ultrio Plus assay was used¹⁰. For the Ultrio assay the 50% LOD (95%CI) was much lower in HBV-DNA reactive/HBsAg negative samples than in HBV-DNA nonreactive/HBsAg positive samples [10.6 (9.2–12.3) and 375 (156–1267) copies/mL, respectively]. In an attempt to estimate a 50% LOD that is representative for the Ultrio NAT seroconversion point in the South African population we decided to take the geometric mean values of the LODs on the two categories of yield samples as best estimates for calculation of the pre-HBsAg WP reductions and pre-NAT infectious WPs in the present study. The geometric mean value for the Ultrio assay was estimated at 63 (38–125) copies/mL, substantially higher than the 50% seroconversion point of 4.1 (3.2–5.3) copies/mL estimated for the Ultrio Plus assay (which is the geometric mean of 4.3 (3.7–5.0) and 4.0 (2.8–5.6) copies/mL found on (dilutions of) South African NAT yield samples and HBsAg positive/HBV-DNA nonreactive samples, respectively)¹⁰.

Estimations of lengths of pre-ID-NAT and pre-HBsAg WPs

The lengths of the infectious pre-ID-NAT and pre-HBsAg PRISM window periods were estimated by the risk day equivalent model³ using the estimated 50% HBV-DNA and HBsAg conversion points of the HBV screening assay, a 50% minimum infectious dose (MID₅₀) of 3.16 (1–10) virions in 20 mL plasma and a doubling time of 2.6 days^11,12. The pre-HBsAg WP reduction (in days) offered by the Ultrio and Ultrio Plus assays was calculated by subtracting the pre-NAT WPs from the pre-HBsAg WP. If the conversion point estimates in this study are correct, the pre-HBsAg WP reduction for Ultrio and Ultrio Plus would correspond with the pre-HBsAg WP-NAT yield rates for the respective NAT assays. Hence, we compared the ratio of the lengths of pre-HBsAg WP reduction for the Ultrio and Ultrio Plus assays to the ratio of WP NAT yields detected in the one year screening periods of the Ultrio and Ultrio Plus assays reported in this study.

Estimation of residual risk by pre-NAT window period donations

The pre-HBsAg WP NAT yield rate (as described above in our HBV infection category 1) in all donations detected by Ultrio and Ultrio Plus in the one year sequential screening periods were used to calculate the residual risk (RR) by the NAT yield/WP ratio model⁴ using the formula: RR= WP_NAT/(WP_HBsAg-WP_NAT) x (N_{WP NAT yields}/N_donations)

RESULTS

Detection rates for different HBV infection categories in Ultrio and Ultrio Plus screening periods

Table 1 compares the HBV detection rates in the sequential Ultrio and Ultrio Plus screening periods for different categories of infection in all donations (table 1) as well as in first time, lapsed and repeat donors (supplemental table 1). There was no significant difference in HBV prevalence and incidence in the two one-year screening periods as can be deduced from the comparable rates of HBsAg positive infections among all donors as well as in first time (prevalent infections) and repeat donors (incident infections). In the Ultrio Plus screening period, pre-HBsAg/anti-HBc WP NAT yields were detected at a 1.70-fold higher rate (p=0.005) in all donors. However, if questionable anti-HBs breakthrough/abortive infections with a very low VL were excluded from the analysis (see supplemental table 2, category f cases) the rate after introduction of Ultrio Plus increased 2.1-fold, from 1:21,540 to 1:10,533. If also five low VL Ultrio Plus WP yields with only 1/5 replicates reactive on the FFP sample (supplemental table 2, categories d and f cases) were not counted as true WP infections, then the improvement factor would have been 1.9-fold. However, if all subcategories of WP infections that were classified as possible (rather than confirmed or probable) infections were excluded (categories c, d, and f in supplemental table 2), the WP-NAT yield increase was 1.7 fold.

Table 1.

Detection rates for different HBV infection categories in all donations in sequential Ultrio and Ultrio Plus screening periods of one year

	Ultrio	Ultrio Plus	factor$	p value
number donations	775 444	789 948
HBV-NAT yield	151 (1:5204)	259 (1:3 050)	1.72	<0.00001
pre-HBsAg WP^	47 (1:16 499)	80 (1:9 874)	1.70	0.005
post-HBsAg WP	10 (1:77 544)	17 (1:46 468)	1.67	0.19
OBI	94 (1:8249)	162 (1:46 468)	1.72	0.00004
HBsAg positive	820 (1:946)	841 (1:939)	1.03	0.89
HBsAg+/DNA+	744 (1:1042)	809 (1:976)	1.07	0.0598
HBsAg+/DNA−	38 (1:20 406)	16 (1: 49 372)	0.42	0.002
All HBV infections	971 (1:799)	1100 (1:718)	1.11	0.016

^{^}included all subcategories of table 2

The number of IgM-anti-HBc positive late acute phase ID-NAT yields increased 1.67-fold, comparable to the 1.72-fold increase in OBI detection rate (table 1). The higher sensitivity of HBV DNA detection by Ultrio Plus reduced the proportion of HBsAg positive donations that were missed in the initial ID-NAT screening test 2.4-fold. In first time donors, the OBI detection rate increased from 1:2065 to 1:1250 with Ultrio Plus, which is approximately 10% of the HBsAg detection rate (1:130). Due to more sensitive HBV-DNA detection, the overall rate of detectable HBV infections increased significantly, by an additional 11% in all donors (table 1); this increase was 2% in first time donors and 10% and 29% in lapsed and repeat donors respectively (supplemental table 1).

Viral load distribution in HBV-NAT yields detected by Ultrio and Ultrio Plus

Figure 1 compares the VL distribution in WP and OBI NAT yields detected by Ultrio and Ultrio Plus respectively. The density of NAT yields was higher in the Ultrio Plus screening period not only in the low VL range but also in the higher range (i.e., above the LLQ of the Roche TaqMan VL assay at >116 copies/mL). Table 2 and supplemental figure 2a–c compares the proportion of Ultrio and Ultrio Plus HBV-NAT yields in HBV-NAT yield donations as well as for WP and OBI categories separately in four VL ranges from 1–10, 10–100, 100–1000 and >1000 copies/mL. The p-values were calculated (Chi square) by using the total number of HBV infections in the Ultrio and Ultrio Plus screening years as the denominator (table 2). In the VL ranges of 1–10 and 100–1000 copies/mL the proportion of HBV-NAT yields was significantly higher in the Ultrio Plus screening year than in the Ultrio screening year. In table 2 and supplemental figure 2d, the data were also aggregated into two VL ranges - less than or greater than 100 copies/mL – and show that the proportion of HBV NAT yields with VL below 100 copies/mL increased 1.4-fold in the second year [from 129/971 (13%) to 198/1100 (18%) of HBV infections (p<0.005]), whereas the proportion of HBV-NAT yields with VL above 100 copies/mL increased 2.4-fold, from 22/971 (2.3%) to 61/1100 (5.6%) of infections (p<0.0002).

Figure 1.

Viral load distribution of HBV WP and OBI NAT yield samples detected by the Ultrio assay (figure 1a) and the Ultrio Plus assay (figure 1b) during one year of screening respectively. VLs are indicated by red diamonds (measured by Roche TaqMan assay), green triangles (estimated by NAT endpoint dilution titre), and blue squares (estimated by probit analysis on replicate NAT results).

Table 2.

Comparison of proportion of WP and OBI NAT yield cases in different VL ranges as compared to the total number of HBV infections in the Ultrio and Ultrio Plus screening year

	proportion all HBV NAT yields
range	Ultrio	Ultrio Plus
copies/mL	n=971	n=1100	p value
1–10	48 (4.94%)	88 (8.00%)	0.005
10–100	81 (8.34%)	110 (10.0%)	0.19
100–1000	14 (1.44%)	43 (3.91%)	0.0006
>1000	8 (0.82%)	18 (1.64%)	0.097
1–100	129 (13.3%)	198 (18.0%)	0.0033
>100	22 (2.27%)	61 (5.55%)	0.0001
	proportion WP HBV NAT yield^
range	Ultrio	Ultrio Plus
copies/mL	n=971	n=1100	p value
1–10	4 (0.41%)	20 (1.82%)	0.0028
10–100	25 (2.57%)	22 (2.00%)	0.38
100–1000	10 (1.03%)	23 (2.09%)	0.054
>1000	8 (0.82%)	15 (1.36%)	0.24
1–100	29 (2.99%)	42 (3.82%)	0.28
>100	18 (1.85%)	38 (3.45%)	0.025
	proportion OBI NAT yields#
range	Ultrio	Ultrio Plus
copies/mL	n=971	n=1100	p value
1–10	44 (4.53%)	68 (6.18%)	0.097
10–100	56 (5.77%)	88 (8.00%)	0.046
100–1000	4 (0.41%)	20 (1.82%)	0.0028
>1000	0 (0.00%)	3 (0.27%)	0.10
1–100	100 (10.3%)	156 (14.2%)	0.0074
>100	4 (0.41%)	23 (2.09%)	0.0008

^{^}included all WP subcategories of table 2

^#also included anti-HBc IgM positive, post HBsAg WP cases

Three of 80 (3.7%) anti-HBc negative HBV WP NAT yields detected by Ultrio Plus could potentially be classified as acute occult infections as they had VLs of 9,300, 20,300 and 21,700 copies/mL with PRISM HBsAg S/CO values of 0.76, 0.62 and 0.23 respectively. However, due to the lack of follow-up samples, the final classification of these cases as acute occult infections was not done and our data analysis did not break these out as a separate subcategory under WP-NAT yields.

Correlation of HBsAg and HBV-DNA levels in anti-HBc negative acute phase

Figure 2 shows the regression line and 95% confidence interval when comparing log HBV-DNA concentrations in copies/mL and log HBsAg S/CO ratios in anti-HBc negative, HBV-DNA positive samples, taking only samples into account with VL detected by the Abbott real time assay. Note that only a subset (14/55; 25%) of HBsAg positive, anti-HBc negative donations were tested for VL, whereas a higher proportion (109/192; 57%) of HBsAg negative WP NAT yield samples routinely confirmed by quantitative PCR had detectable VL in the assay. Using this data set, the HBsAg S/CO=1.0 crossing point (95% CI) was calculated at 919 (625–1352) copies/mL. Six of 36 (17%) HBV-NAT yields had VL above 2 times the upper confidence limit of the regression line and three of these (8%) were above 5 times the upper limit (figure 2). The latter could be considered potential acute occult HBV infections.

Figure 2.

Correlation between Log HBV-DNA concentration and Log HBsAg S/CO in acute viremic phase and estimation of HBsAg S/CO=1 cutoff crossing point (and 95% confidence limits) by regression analysis. This analysis included only acute phase anti-HBc negative, HBV-DNA positive samples that had detectable VL by qPCR during one year of Ultrio Plus screening, A limitation of the analysis is that only a subset of HBsAg positive donors had been tested for VL. Data points in ellipse had VL above 5 times upper 95% CI and could represent acute occult infections.

Estimation of lengths of Pre-ID-NAT and pre-HBsAg WPs

When using the geometric mean values of 63 and 4.1 copies/mL as the 50% LOD for Ultrio and Ultrio Plus, the infectious WPs were estimated at 23.2 and 13.1 risk day equivalents for Ultrio and Ultrio Plus respectively. Using a value of 960 copies/mL as the 50% HBsAg PRISM seroconversion point (previously established in HBV genotype A seroconversion panels and confirmed by regression analysis in this study) the infectious pre-HBsAg WP was estimated at 33.5 days. Based on the above mentioned average LODs, HBV-DNA screening by Ultrio reduced the 33.5 day pre-HBsAg WP by an average of 10.3 days (31%) and Ultrio Plus by 20.4 days (61%) for a theoretical WP NAT yield improvement factor of 2.0 over Ultrio. As a consequence of the higher sensitivity of the Ultrio Plus assay, the HBV WP NAT yield rate in the present screening study increased 1.7-, 1.9- or 2.1-fold depending on which subcategories of possible WP donations with low VL were included as true acute infections (supplemental table 2). The latter factors are in line with the predicted WP reduction from the estimated average conversion points of 960, 63 and 4.1 copies/mL for HBsAg PRISM, Ultrio and Ultrio Plus as illustrated in Figure 3.

Figure 3.

Illustration of the HBV ramp up phase with the calculated 50% HBV-DNA and HBsAg conversion points of 4.1, 63 and 960 copies/mL for Ultrio Plus, Ultrio, and PRISM HBsAg assays respectively as estimated by the WP risk day equivalent model¹⁰. The solid oblique black line represents the log linear increase of the average HBV-DNA concentration and the dotted oblique lines represent variation in this concentration based on different viral doubling times between individuals. The upper horizontal arrows and the dotted vertical lines represent the range of VL at the start of the infectious WP and around the HBV-DNA and HBsAg conversion points (represented by the solid vertical lines). The VL range around the average conversion points is wider for the Ultrio assay than for the Ultrio Plus assay because of the variable sensitivity of the former NAT method (explained in discussion); it is also wider for HBsAg, most likely due to differences in HBV to HBsAg particle ratios between donors. The light and dark blue arrows at the bottom represent the average lengths of the pre-HBsAg HBV-DNA detection periods for the Ultrio and Ultrio Plus assays. The red arrows represent the lengths of the infectious pre-NAT window periods for Ultrio and Ultrio Plus. The pink arrow represents the eclipse phase in which HBV replicates in the liver of the infected donor while the amount of HBV in the peripheral blood is below the minimum infectious dose that is required for HBV transmission.

Estimated residual risk of pre-NAT WP infections

Using the lengths of the pre-NAT window periods and detection periods shown in Figure 3 and the number of pre-HBsAg NAT yields in table 1, we estimated that residual risk decreased 2.1-fold from 136.5 per million RBC transfusions in the Ultrio screening period to 65.0 per million in the Ultrio Plus screening period.

DISCUSSION

In this study we compared the HBV-NAT yield rates detected by Ultrio and Ultrio Plus assays in two sequential years of screening during which incidence and prevalence of HBV infection was comparable as assessed by HBsAg detection rates in first time and repeat donations. The HBV-NAT yield rate in different stages of infection (pre-and post HBsAg WP and OBI) increased 1.7-fold and when subcategories of possible HBV-NAT yields with very low VL (including possible chronic OBIs with anti-HBs as the sole detectable serum marker) were excluded the improvement factor of WP NAT yields rose to 1.9 – 2.1-fold. This two-fold rise in HBV WP-NAT yield rate was in line with the estimated 50% NAT conversion points of 63.0 and 4.1 copies/mL for Ultrio and Ultrio Plus respectively because this translates to pre-HBsAg NAT detection periods of 10.3 and 20.4 days as compared to the HBsAg conversion point estimated at 960 copies/mL in seroconversion panels^13,14. Hence, our approach of taking the geometric mean values for the LODs in samples that were detected by Ultrio (NAT yields) and samples that were missed by Ultrio (serology yields) (methods section and supplemental figure¹⁰) provided representative NAT conversion point estimates for the Ultrio assay versions which are consistent with pre-HBsAg NAT detection periods that were calculated from the screening data.

An additional dataset was used to estimate the geometric mean HBV-DNA concentration at the HBsAg seroconversion point by regression analysis (figure 2). Not all acutely infected donors in the study period were systematically subjected to viral load testing which may have influenced the analysis (those that were HBsAg positive were underrepresented). The average HBV-DNA concentration at the HBsAg seroconversion point in anti-HBc negative ramp-up samples was estimated at 919 (625–1352) copies/mL, comparable to the value of 960 copies/mL that was found in five seroconversion panels^13,14.^.

The previously estimated ~43 -fold (11–350 fold) lower analytical sensitivity of Ultrio as compared to Ultrio Plus testing on HBsAg positive and Ultrio nonreactive samples¹⁰ was the first indication of under-detection of certain HBV samples by the Ultrio assay. This deficiency of the previous Ultrio assay (i.e. failure to detect some samples with moderately high VL) was also observed in the present study when comparing the VL distributions of HBV-NAT yield samples in the Ultrio and Ultrio Plus screening periods. Our data indicate that the original Ultrio assay not only missed NAT yields below 100 copies/mL, but also a significant number above this level. Of all HBV infections the proportion of HBV-NAT yields (WP and OBI combined) with VL below 100 copies/mL increased significantly (p<0.002) from 13% to 18% (69 additional infections) in the Ultrio Plus screening period. However we were surprised to also observe an additional 39 HBV-NAT yields with VL above 100 copies/mL in the Ultrio Plus screening year, which represented a significant rise of 3.3% (from 2.3% to 5.6%; p<0.0002) of all HBV infections. When considered separately, results from OBI-NAT yields and WP-NAT yields were not the same. In OBI-NAT yields, the proportion in all VL ranges was higher in the Ultrio Plus screening year, whereas among WP NAT yields, the proportion of NAT yields with VL between 10–100 copies/mL did not differ between years. A limitation of our study is that we were not able to test Ultrio Plus HBV NAT yields in the former Ultrio assay, however the HBsAg positive rate was similar in the one year Ultrio and Ultrio Plus screening periods confirming that there was no change in prevalence and incidence of HBV infections in the two sequential screening periods.

Although speculative, a possible explanation for the unexpected increase in HBV-NAT yields with moderately high VL is that there are some HBV infected donors in whom a relatively large proportion of HBV particles contain double stranded HBV-DNA in the Ultrio target region. It is known that the length of the double stranded portion of the HBV genomes varies in HBV particles in different donors and even in the same individual^15,16. Therefore the analytical sensitivity of the Ultrio assay for different HBV infected donor samples may be unpredictable and inconsistent. Hence the variable sensitivity of the previous Ultrio assay may be related to the extent of double stranded HBV-DNA in HBV particles circulating in the population (and this is likely unrelated to the genotype which is predominantly genotype A1 in South Africa). This under-detection problem of TMA for HBV has been overcome by inclusion of the target enhancer reagent in the Ultrio Plus assay which results in increased denaturation of double stranded DNA portions of the HBV-genome; i.e., due to the alkaline shock step in the Ultrio Plus assay, more single stranded DNA is available for the capture probes and the sensitivity for HBV detection is restored.

The variable sensitivity of Ultrio also influences the HBV-DNA conversion point in the pre-HBsAg WP. We illustrate this in figure 3 which shows that the NAT conversion point in the HBV-DNA ramp up phase for Ultrio is less well defined (i.e. is spread over a wider VL range) than is the Ultrio Plus assay. There is also variability in VL at the HBsAg seroconversion point (3); this is likely due to another explanation which is that the ratio of HBV (DNA containing) particles to sub-viral HBsAg (non-DNA containing) particles can vary between 1:1000 in regular seroconversions to 1:10 in acute occult HBV infection cases⁹.

These observations indicate that previous residual risk estimations based on analytical sensitivity levels of the Ultrio assay on the Eurohep and WHO genotype A standards can no longer be considered valid¹. To correct this, we used the Ultrio and Ultrio Plus one-year screening datasets and previous analytical sensitivity data in NAT and HBsAg yield samples (methods section, supplemental figure)¹⁰ to reassess estimates for the lengths of the pre-HBsAg and pre-ID-NAT WPs and the residual risk of pre-NAT WP infections for the Ultrio and Ultrio plus screening periods. For RBC transfusions and an estimated MID₅₀ of 3.16 virions (between 1 and 10 virions), the estimated lengths were 13.1, 23.2 and 33.5 days for Ultrio Plus, Ultrio, and HBsAg PRISM respectively resulting in a 2.1 fold reduction in residual risk from 137 to 65 pre-NAT WP transmissions per million RBC transfusions after introduction of the Ultrio Plus assay. However these estimated WPs were calculated using a pre-formatted spreadsheet tool developed by Weusten et al³, and there is considerable uncertainty attached to the underlying parameters, i.e. viral doubling time, plasma transfusion volume, calibration in copies or virion numbers, LODs determined by probit analysis and finally infectivity of HBV as expressed by MID₅₀. Since confidence limits for some of these parameters are unknown, we did not engage in calculating confidence bounds for the lengths of WPs in this paper. Also these estimates of residual risk do not take into account that a considerable number of South African recipients are not susceptible to HBV infection and that HBV can be neutralized by anti-HBs in co-transfused blood components.

WP lengths for different HBV screening assays can be deduced from analytical sensitivity data provided that there is a uniform distribution of VL in the anti-HBc negative HBV ramp up phase. In the Ultrio Plus screening year we observed a similar number of WP NAT yields in the VL ranges of 1–10, 10–100 and 100–1000 copies/mL indicating random appearance of donors in the pre-HBsAg HBV ramp phase.

In the Ultrio Plus screening period we identified 20/80 (25%) of HBV WP-NAT yields that were classified as possible WP infections because of low VL (<10 copies/mL) and lack of follow up. Five of those were classified as possible anti-HBs breakthrough/abortive infections. The true status of these anti-HBc negative low VL NAT yields can only be established by follow up testing. The impact of these latter subcategories of possible HBV-NAT yields on the calculation of residual risk of WP infections are addressed in a separate paper [Lelie et al, manuscript in preparation].

The data in this study show that replacement of the Ultrio assay by the more sensitive Ultrio Plus assays has resulted in an important reduction of HBV residual risk. Moreover, the data in this report are essential to be able to use a multi-regional data base of 11 million donations and 9455 HBV infections (that were detected by HBsAg and Ultrio ID-NAT testing)⁶ for prediction of the residual HBV transmission risk with the current more sensitive HBV-NAT blood screening assays, as well as for estimation of the relative efficacy of different HBV screening strategies in epidemiologically and demographically diverse donor populations.