Interlaboratory Comparison of Epstein-Barr Virus Viral Load Assays

Authors


* Corresponding author: Jutta K. Preiksaitis, J.Preiksaitis@provlab.ab.ca

Abstract

To assess interlaboratory variability in qualitative and quantitative Epstein-Barr virus (EBV) viral load (VL) testing, we distributed a panel of samples to 28 laboratories in the USA, Canada and Europe who performed testing using commercially available reagents (n = 12) or laboratory-developed assays (n = 18). The panel included two negatives, seven constructed samples using Namalwa and Molt-3 cell lines diluted in plasma (1.30–5.30 log10 copies/mL) and three clinical plasma samples. Significant interlaboratory variation was observed for both actual (range 1.30–4.30 log10 copies/mL) and self-reported (range, 1.70–3.30 log10 copies/mL) lower limits of detection. The variation observed in reported results on individual samples ranged from 2.28 log10 (minimum) to 4.14 log10 (maximum). Variation was independent of dynamic range and use of commercial versus laboratory-developed assays. Overall, only 47.0% of all results fell within acceptable standards of variation: defined as the expected result ± 0.50 log10. Interlaboratory variability on replicate samples was significantly greater than intralaboratory variability (p < 0.0001). Kinetics of change in VL appears more relevant than absolute values and clinicians should understand the uncertainty associated with absolute VL values at their institutions. The creation of an international reference standard for EBV VL assay calibration would be an initial important step in quality improvement of this laboratory tool.

Abbreviations: 
SOT

solid organ transplant

HSCT

hematopoietic stem cell transplant

EBV

, Epstein–Barr virus

PTLD

posttransplant lymphoproliferative disorder

VL

viral load

GM

geometric mean

SD

standard deviation

LOD

lower limit of detection

CV

coefficient of variation

QNAT

quantitative nucleic acid testing

PCR

polymerase chain reaction

ASR

analyte-specific reagents.

Introduction

The use of more potent and targeted immunosuppression in solid organ transplantation (SOT) and T-cell depletion along with transplantation of higher risk, at times mismatched unrelated donors in hematopoietic stem cell transplantation (HSCT), has resulted in the emergence of Epstein–Barr virus (EBV) associated posttransplant lymphoproliferative disorders (PTLD) as a major complication of transplantation in these settings (1). The observation in the mid 1990s that patients who developed PTLD often had high EBV viral load (VL) in peripheral blood and that these high levels were found prior to the onset of clinical illness has led to the widespread use of this laboratory assay in preemptive programs for disease prevention, as a diagnostic assay in symptomatic patients and to monitor response to therapy (2). EBV VL monitoring has also been recommended as a means of monitoring safety in clinical trials of new immunosuppressive agents (3).

It has been argued that the utility of preemptive interventions that involve reduction in immunosuppression or more controversial approaches such as the use of anti-viral agents or anti-CD20 therapy in response to EBV quantitative nucleic acid testing (QNAT) has never been clearly demonstrated in randomized controlled trials (2,4). However, several centers have observed a reduction of PTLD incidence or mortality when cohorts of patients who underwent routine EBV VL surveillance and preemptive interventions were compared to historical cohorts who were not monitored (5,6).

Despite its widespread implementation in transplant settings, there are a number of barriers and limitations to the optimal use of EBV VL assessment in peripheral blood (6). The best sample type (plasma, whole blood or lymphocytes) and reporting units remains controversial and although frequent monitoring has been recommended, the optimal testing regimen in specific transplant populations is uncertain. Trigger points for preemptive intervention have often been defined at single centers for specific allograft types and have not been validated in multicenter studies; true natural history studies are rare (4,7). Our study addresses an additional limitation, testing variability that arises because EBV VL assays have not been standardized or extensively cross- referenced.

Advances in molecular laboratory diagnostic technology have resulted in the rapid evolution of EBV VL assessments in peripheral blood (8). Historical techniques that involved counting the number of EBV-transformed B-cell clones resulting from the spontaneous outgrowth of B cells isolated from peripheral blood and cultured under predetermined conditions were replaced initially by semiquantitative methods that directly detected EBV DNA by comparing Southern blot or agarose gel signals to calibrated controls. Assays moved from the research to the diagnostic setting and became more precise and quantitative when EBV DNA was coamplified with an internal calibration standard. Most recently, real-time QNAT has become the standard at most transplant centers and commercial laboratories.

Most assays remain ‘laboratory developed’, calibrated with proprietary standards and assay parameters. Differences in diagnostic testing procedures have resulted in site-specific patient management algorithms. This makes inter institutional comparison of published and unpublished clinical practice difficult. Recently, a number of commercial kits and analyte-specific reagents (ASRs) have become available for EBV quantitative VL assessment. However, internationally accepted reference standards or reference panels for EBV QNAT are currently not available, making cross-referencing of both laboratory-developed and commercial assays difficult.

One of the best ways to assess the variability of EBV VL assessment in individual laboratories is to distribute proficiency panels and directly compare results. Since data regarding interlaboratory variability in EBV QNAT were limited, we designed and conducted this comparative study that included laboratories from Canada, the US and Europe.

Materials and Methods

Preparation of the reference standard and panel

Stock cell line preparation and cell count to define a reference standard:  Namalwa cells (ATCC, Catalog no. CRL-1432™), known to contain 1–3 EBV genome copies per cell, grown in RPMI 1640 with 7.5% fetal calf serum were used as a source of EBV DNA (9). To calculate EBV VL based on Namalwa cell count, a value of 2 genome copies/cell was used. Molt-3 cells (ATCC, Catalog no. CRL-1552™), known to be EBV negative, grown in RPMI 1640 with 10% fetal calf serum were used to prepare a background cellular matrix (10). The cell lines were passaged and grown in 75 cm2 flasks for 3 days in the appropriate media. Cells were counted using a Niebauer chamber and resuspended in EBV negative (DNA and antibody) plasma adjusted to a concentration of 1.0 × 106 cells/mL. These stock cell dilutions were used to prepare the constructed samples for the panels.

QNAT to define a reference standard:  In order to obtain a consensus result that would define a quantitative reference laboratory standard for the constructed samples in the panel, two aliquots of the Namalwa stock preparation were shipped to eight reference laboratories and tested using a range of EBV PCR assays. Laboratories were requested to perform two extractions and run the samples in triplicate and report results as genome copies/mL. In total, reference laboratories provided results obtained using Roche ASR using the LightCycler® platform, artus-Biotech ASR using both the LightCycler® and ABI platforms and five laboratory-developed assays. The consensus result was 5.11± 0.55 log10 copies/mL (geometric mean ± SD); range (4.33–6.16 log10 copies/mL). This consensus result was compared to the theoretical ‘expected value’ based on Namalwa cell count (2 EBV genome copies / Namalwa cell): 1.0 × 105 Namalwa cells / mL = 5.30 log10 copies/mL.

Preparation and constitution of the panel:  The panel consisted of 12 blindly coded samples. The dilution matrix for all samples in the panel was human plasma, seronegative for EBV IgG and EBV DNA. All nine constructed samples were prepared using dilutions of Namalwa and Molt-3 cells keeping the total cell count constant in all samples to mimic the total white cell number found in peripheral blood of normal subjects. Two samples of Molt-3 cells 1.0 × 106 cells/mL in EBV negative plasma (01 and 08) were used as negative controls. The seven EBV positive constructed samples had a dynamic range of 1.30 to 5.30 copies/mL based on Namalwa cell count, and were mixed with an appropriate number of negative Molt-3 cells to keep the total cell count constant in all samples. The panel included replicates of constructed samples at two dilutions 3.30 log10 copies/mL (05 and 10) and 4.30 log10 copies/mL (02 and 11). Three clinical plasma samples (04, 07 and 12) were also included. The clinical samples were EBV positive plasma samples obtained from three individual solid organ transplant patients, two of whom were diagnosed with B-cell PTLD. Clinical samples were diluted 1 in 10 to 1 in 20 with EBV negative plasma for use in the panel. The total volume of each panel sample was 500 μL and all sample sets were aliquoted from the same diluted pool on the same day by a single technician using one pipette to minimize any variation and possible errors in preparation. All samples were labeled, coded and stored at −70°C until shipment.

Study participants

Laboratories utilizing QNAT for VL determination in peripheral blood were invited to participate in this study by direct contact through the American Society of Transplantation and the Canadian Society of Transplantation. A total of 30 laboratories (17 in US, 11 in Canada and 2 in Europe) agreed to participate in the study. Panel samples were coded and shipped on dry ice by overnight courier. The recipient laboratories were asked to report the arrival and condition of the panel samples by e-mail and to return the results within 6 weeks. A questionnaire was also sent to obtain information on technical aspects of the procedures employed. To ensure confidentiality, all laboratories were requested to send their results and information to the central laboratory (Provincial Laboratory for Public Health, located in Edmonton, Alberta, Canada) for analysis.

Nucleic acid amplification tests (QNAT) utilized in the study

Analysis of the reference and subsequent testing of the EBV panels distributed to laboratories used a variety of kit/ASR combinations of laboratory-developed assays. The artus-Biotech EBV PCR kit and ASR (Qiagen Diagnostics, Hamburg, Germany) were utilized in two formats with amplification using the LightCycler® (version 1.0 or 2.0; Roche Molecular Diagnostics, Laval, Québec, Canada) or ABI SDS 7700/7000/7500 [Applied Biosystems (ABI), Foster City, CA]. The Roche ASR was utilized with the LightCycler® 1.0 or 2.0. Commercially available reagents and assays for EBV are at different stages of approval and regulation in Europe, USA and Canada. As a result the same reagents are defined as kits in some countries and ASRs in others. For the purposes of analyses, all laboratories utilizing the reagents noted above, were defined as being in the commercial group and were compared to all laboratory-developed assays.

Statistical analysis of panel results and determination of ‘expected result’:  For evaluation of the data sets, the value (genome copies/mL) was first converted to its logarithmic (log10) value. Negative results were not included in calculations of mean and other analyses of quantitative data. The geometric mean (GM) and SD were calculated for all positive samples. Any result reported as positive but not quantifiable was assigned the self-reported limit of detection (LOD) for the assay for calculating mean and variation. For constructed samples, the theoretical VL calculated based on the Namalwa cell count in each sample adjusted for dilution was utilized as the ‘expected result’ to which all results were compared. For each clinical sample, the GM in log10 copies/mL of positive results reported from all laboratories was defined as the ‘expected result’ to which all results were compared. The commercial assays and laboratory-developed assays were analyzed separately to assess differences. The intralaboratory and interlaboratory variability for the two sets of replicated samples in the panel was expressed as a coefficient of variation (%CV). Difference between the intralaboratory and interlaboratory variability, and the sensitivity between the commercial- and laboratory-developed assays were analyzed by the Fisher's exact test. A nonparametric statistical method (Mann–Whitney U-test) was used to analyze variation between commercial and laboratory-developed assays.

Results

Characteristics of participating laboratories and assays utilized

Thirty-two panels were distributed to 30 laboratories; two laboratories were sent two panels each to ensure all commercial assays were represented in the results. Twenty-nine (96.7%) of the 30 laboratories returned results: 16 from the USA, 11 from Canada and 2 from Europe. An additional laboratory's results were removed from analysis due to the semiquantitative nature of their assay. There were 30 data sets from 28 laboratories for final analysis. Laboratory-developed assays were used for 18 sample sets and commercial assays/reagents for 12 sample sets including 8 data sets for Roche ASR, and 5 for artus-Biotech ASR (3 using ABI SDS instrument and 2 using LightCycler 2.0). Assays targeted a range of genes and included a number of different nucleic acid extraction reagents with no particular target/extraction method predominating. There were not sufficient numbers in each of these separate methods/groups for more detailed analysis or conclusions to be drawn except for the broad separation into laboratory-developed methods versus approaches that used commercially available reagents (kits or ASRs).

Qualitative results for panel samples

A summary of qualitative results is given in Table 1. Two false-positive results were reported with use of laboratory-developed assays in two laboratories. The range of self-reported LOD for the QNAT assays was variable (1.70 to 3.30 log10 copies/mL) and the observed variation in LOD ranged from 1.30 to 4.30 log10 copies/mL. All laboratories were able to detect constructed samples with expected VL of 4.30 log10 copies/mL as positive, although a few laboratories were unable to produce quantitative results particularly when using laboratory-developed assays, even at VL levels as high as 5.30 log10 copies/mL. Complete concordance (100%) among all laboratories was observed in qualitative results for constructed samples with expected results ≥4.30 log10 copies/mL and in all clinical plasma samples. Greatest discordance in qualitative results was found in the constructed samples with low VL (1.30 to 3.30 log10 copies/mL). Although there was a trend toward lower sensitivity of laboratory-developed assays relative to commercial assays at VL of 1.30, 2.30 and 3.30 log10 copies/mL, these differences were not statistically significant (p = 0.35, 0.14 and 0.26, respectively, summarized in Table 1). The assays utilized in the study targeted a wide range of genes but most targeted either EBNA1 or LMP2. The study numbers were not large enough to clearly assess the impact of gene targets on qualitative detection, although no obvious qualitative false negatives could be attributed to the gene targets used.

Table 1.  Summary of qualitative results for 30 panels reported from 28 laboratories
Sample No.Expected results copies/mL (log10)Overall qualitative results (n = 30) No. of panel results (%)Qualitative results of laboratory developed assays (n = 18)No. of panel results (%)Qualitative results of commercial assays (n = 12) No. of panel results (%)
NegPos-NQPos-QNegPos-NQPos-QNegPos-NQPos-Q
  1. Expected results were calculated based on the Namalwa cell count, except for samples marked 1(clinical samples) where expected results were geometric mean (GM) of positive results reported from all laboratories.

  2. Pos-NQ = positive but not quantifiable.

  3. Pos-Q = positive with quantifiable results.

010.00 30 (100)0  0 18 (100)0012 (100)00
080.0028 (93)0 2 (7)  16 (89)0 2 (11) 12 (100)00
091.3027 (90)1 (3) 2 (7)  17 (94)0 1 (6)  10 (83) 1 (8)1 (8) 
032.3016 (53) 3 (10)11 (37) 12 (67)3 (17) 3 (17) 4 (33)08 (67)
053.30 3 (10)2 (7)25 (83)  2 (11)2 (11)14 (78) 1 (8) 011 (92) 
103.30 3 (10)1 (3)26 (87)  3 (17)1 (6) 14 (78) 0012 (100)
024.3001 (3)29 (97) 01 (6) 17 (94) 0012 (100)
114.3002 (7)28 (93) 02 (11)16 (89) 0012 (100)
065.3001 (3)29 (97) 01 (6) 17 (94) 0012 (100)
0714.080030 (100)0018 (100)0012 (100)
0413.950030 (100)0018 (100)0012 (100)
1214.210030 (100)0018 (100)0012 (100)

Quantitative results for panel samples

Quantitative results are summarized in Table 2. False- negative results were not included in the calculation of mean or SD but can be found in the analysis of qualitative results (Table 1). For all constructed samples with expected results ≥3.30 log10 copies/mL, the median and GM were approximately 0.30 log lower than the expected result. The range of reported quantitative results on individual constructed samples was large and was independent of dynamic range and use of commercial or laboratory-developed assays (Table 2 and Figure 1). The variability for reported quantitative positive results on individual constructed samples varied from a minimum of 2.28 log10 copies/mL to a maximum of 4.14 log10 copies/mL (Table 2 and Figure 1). The total range of variability for reported quantitative positive results on individual clinical samples was narrower with a minimum of 2.12 log10 copies/mL and a maximum of 2.40 log10 copies/mL (Table 2 and Figure 2), although the dynamic range covered by the available clinical samples was limited.

Table 2.  Summary of EBV quantitative results for positive samples included in the panel
Sample No.Expected results copies/mL (log10)Quantitative results for all data sets (maximum n = 30) copies/mL (log10)Quantitative results for laboratory-developed assays (maximum n = 18) copies/mL (log10)Quantitative results for commercial assays/reagents (maximum n = 12) copies/mL (log10)
GM ± SD2Median (range)GM ± SD2Median (range)GM ± SD2Median (range)
  1. Expected results were calculated based on the Namalwa cell count, except for samples marked 1(clinical samples) where expected results were geometric mean (GM) of positive results reported from all laboratories.

  2. 2GM ± SD = geometric mean plus or minus standard deviation; negative results were excluded.

091.302.06 ± 0.930.00 (0.00–2.74)2.450.00 (0.00–2.45)1.78 ± 1.230.00 (0.00–2.74)
032.302.54 ± 0.590.00 (0.00–3.78)2.11 ± 0.270.00 (0.00–2.51)2.86 ± 0.562.46 (0.00–3.78)
053.303.01 ± 0.522.92 (0.00–4.14)2.86 ± 0.432.89 (0.00–3.62)3.23 ± 0.573.15 (0.00–4.14)
103.303.08 ± 0.612.92 (0.00–4.12)3.02 ± 0.522.89 (0.00–4.12)3.15 ± 0.733.25 (1.56–4.05)
024.303.96 ± 0.594.03 (2.76–5.04)3.83 ± 0.523.95 (3.00–4.50)4.15 ± 0.664.26 (2.76–5.04)
114.303.93 ± 0.663.97 (2.18–5.00)3.84 ± 0.643.87 (3.00–4.82)4.07 ± 0.704.12 (2.99–5.00)
065.304.89 ± 0.814.96 (2.15–6.09)4.73 ± 0.864.93 (2.15–5.69)5.14 ± 0.695.03 (4.09–6.09)
0714.084.08 ± 0.604.09 (3.00–5.12)4.21 ± 0.614.21 (3.38–5.12)3.89 ± 0.573.82 (3.00–4.92)
0413.953.95 ± 0.563.96 (3.00–5.31)3.88 ± 0.483.87 (3.10–4.45)4.06 ± 0.674.08 (3.00–5.31)
1214.214.21 ± 0.614.36 (3.08–5.48)4.10 ± 0.574.22 (3.08–5.12)4.37 ± 0.654.46 (3.40–5.48)
Figure 1.

Figure 1.

Histogram of reported EBV quantitative results for constructed samples (30 panels, 28 laboratories) analyzed in 0.50 log10 increments. Each rectangle represents results from one laboratory for a single-panel sample. Expected results are 1.30 (A), 2.30 (B), 3.30 (C), 4.30 (D) and 5.30 (E) EBV genome copies/mL (log10) quantified by Namalwa cell count, respectively.

Figure 1.

Figure 1.

Histogram of reported EBV quantitative results for constructed samples (30 panels, 28 laboratories) analyzed in 0.50 log10 increments. Each rectangle represents results from one laboratory for a single-panel sample. Expected results are 1.30 (A), 2.30 (B), 3.30 (C), 4.30 (D) and 5.30 (E) EBV genome copies/mL (log10) quantified by Namalwa cell count, respectively.

Figure 2.

Histogram of reported EBV quantitative results for clinical plasma samples (30 panels, 28 laboratories) analyzed in 0.50 log10 increments. Each rectangle represents results from one laboratory for a single-panel sample. Expected results are 4.08 (A), 3.95 (B) and 4.21 (C) EBV genome copies/mL (log10), respectively, based on geometric mean of all quantitative results.

Figure 3 summarizes results reported by laboratories relative to the expected result for each sample. Assuming that ± 0.50 log10 variation from the expected result is an acceptable range for QNAT assays, our analysis showed that 47% of the overall results fell within this ‘acceptable’ variation. For samples with a sufficient number of reported positive results available for analysis, 47–57% of reported results on individual constructed samples and 53–67% of reported results on individual clinical samples fell within this acceptable range. Although there were notable exceptions, in general results reported within individual laboratories were relatively linear over the dynamic range represented by the samples tested (Figure 4).

Figure 3.

Log10 variation in reported results for individual positive samples in the panel relative to expected results (assigned 0 value) using laboratory-developed (LD) and commercial assays (Com). Blue and orange diamonds, constructed samples. Green and pink circles, clinical samples. The log10 numbers on the x-axis indicate the expected result for each sample.

Figure 4.

Result linearity over the dynamic range for 30 panels tested in 28 laboratories. Each line represents results from one panel. (A) Commercial assays (n = 12) and (B) laboratory-developed assays (n = 18). The x-axis shows expected results based on Namalwa cell count.

For the two sets of samples replicated in the panel, the reproducibility of quantitative results expressed as %CV was significantly better within a laboratory (intralaboratory variation) than between laboratories at different centers (interlaboratory variation) as shown in Table 3 (p < 0.0001).

Table 3.  Intra- and interlaboratory variation for samples replicated in the panel
 Mean coefficient of variation (%CV)
Duplicate samples (sample 05 and 10) 25 panels1Duplicate samples (sample 02 and 11) 30 panelsp-Value*
  1. *Fisher's exact test (2-tailed).

  2. 1Calculation based on quantifiable results only.

Intralaboratory39.01 30.480.234
Interlaboratory135.56 135.261.0  
p-Value*  <0.0001   <0.0001 

Discussion

Our study demonstrates significant interlaboratory variability in reported EBV QNAT results on identical specimens. We hypothesize that this variability is impacting the quality of programs associated with PTLD prevention, diagnosis and monitoring, preventing the meaningful interinstitutional comparison of results and making interpretation of results for the purpose of safety monitoring in clinical trials problematic. This report confirms and extends the observations of a recently published study that reported similar high variability when comparing EBV QNAT results for common sample panels tested by eight US laboratories (11). In our study, information regarding, the threshold values for intervention and site-specific rates of PTLD were not obtained. Collection of this type of data in future studies would allow a more clear determination of the impact of VL variability on clinical outcomes. Significant opportunities exist for quality improvement with respect to this laboratory tool.

EBV VL measured in peripheral blood is only a ‘surrogate’ marker for total body or tissue-specific lytic and latent EBV infection in both normal and aberrant/atypical cells (12). The EBV VL measured in the peripheral blood of transplant recipients is largely cell associated, with transcriptionally silent latently infected resting memory B cells (CD19+, IgD negative Ki67-, CD23-, CD80-, IgM+) containing a low genome copy/cell making the greatest contribution to VL in most patients. Some patients with very high VL have highly atypical B cells (Ig-null cellular phenotype) containing a high EBV genome copy number also contributing to VL in peripheral blood. Cell-associated lytic virus infection is rarely demonstrated in peripheral blood and when present is usually associated with plasma cells (CD38Hi, CD10−, CD19+, CD20lo) (13). Patients with infectious mononucleosis and perhaps all patients experiencing primary infection have evidence of both encapsidated virus and free DNA in the plasma fraction of peripheral blood, when EBV DNA is detected in this blood compartment. In contrast, preliminary evidence in patients with EBV-associated malignancies, suggests that the plasma fraction of peripheral blood, when positive, contains only free EBV DNA in the absence of encapsidated virions (14).

This complexity creates a significant challenge in the construction of a test panel that would mimic the clinical situation. Many laboratories in North America supporting organ transplant programs use whole blood as the preferred specimen type in the transplant setting. We chose to create samples that contained EBV DNA largely in the form of latently infected cells and a total DNA amount that replicated that found in most whole blood samples. It was felt that this, rather than the use of EBV virions or free EBV DNA, most closely replicated the characteristics of EBV VL in the peripheral blood of patients in the transplant setting. The lack of an internationally accepted standard for EBV QNAT makes it difficult to compare results to a well-defined quantitative ‘gold standard’. We elected to use the ‘theoretical gold standard’ based on Namalwa cell count rather than the consensus result provided by reference laboratories on the Namalwa stock preparation as the ‘expected value’ for our constructed samples. This decision was based on the fact that although the GM and median of the consensus result reported from the selected reference laboratories was close to the ‘theoretical gold standard’, the variation expressed as total range of results reported by these laboratories, spanning 1.90 log10 copies/mL, remained high. We acknowledge that this is an arbitrary decision that allowed us to analyze results but does not necessarily reflect true VL.

We observed significant variability in the sensitivity of EBV VL assays being used, with observed LODs varying over 3.00 log10 copies/mL. However, a key unresolved question is what level of assay sensitivity is clinically relevant to manage EBV infection and disease in the transplant setting? This question involves two set points in the assays, the LOD for the assay and the trigger point for clinical intervention. It would be useful if the LOD of assays being used were above the threshold for detection of normal EBV latency in immunocompetent adults, estimated to be approximately 0.01–0.10 copies per 105 peripheral blood mononuclear cells (PBMCs), in order to differentiate a ‘normal’ state from clinically relevant reactivation events. Although our panel did not include samples this low in the dynamic range, approximately 10% of laboratories in our study reported the ability to detect the positive sample with the lowest VL in our panel, 1.00 log10 copies above values that would be associated with normal EBV latency. Whether assays in these laboratories are ‘too’ sensitive from a clinical management perspective is uncertain. A group of patients with early EBV-positive B-cell PTLD in the presence of low or nondetectable EBV VL in peripheral blood have been described (15). Although this may represent a true clinical subgroup of PTLD patients, our data would suggest that laboratory artifact due to variability in LOD of assays being used at various centers should be ruled out first in defining this subgroup of patients. In monitoring response to therapy in PTLD using EBV VL in blood, it is not clear that treatment to VL ‘negativity’ is required to prevent short- or long-term disease relapse. If collation of data from multiple centers using variable assays is used to answer this question, our data suggest it is important to realize that EBV VL ‘negativity’ may be a highly institutional and assay-dependent term. Trigger points for intervention in preemptive programs aimed at PTLD prevention have not been validated in interinstitutional studies. Single-center recommendations were seldom based on natural history studies and were determined using a variety of specimen types, reporting units and assays. However, many of these recommended ‘trigger’ points are in the lower limit of the dynamic range where variability in LOD of assays in addition to quantitative precision may play a significant role. For example, given the trigger point for intervention recommended by the University Medical Center Utrecht for HSCT recipients of 1000 (3.00 log10) copies/mL of plasma (6), or the University of Pittsburgh recommendation in solid organ transplant recipients of approximately 4000 (3.60 log10) copies /mL of whole blood (16), our study, suggests that 10% of laboratories would identify samples with VL in this dynamic range as negative and a further 3–7% would correctly identify these samples as positive but be unable to quantify the VL. Among laboratories who correctly report samples with EBV VL in the range of 1000–4000 copies/mL as positive, our data suggest the actual reported values would range from approximately 40 to 14 000 copies/mL. We, like Hayden et al. (11), observed that intralaboratory result variability is significantly less than interlaboratory variability. Our study did not evaluate intraassay precision in detail but results were generally close within an individual laboratory for the two duplicate samples provided in the panel showing the reproducibility of results. This combined with the general linearity of results over the dynamic range within most centers, provides some reassurance to physicians managing patients being tested using a single assay at one center. Our study illustrates that kinetics of change in VL is likely more important that absolute VL values and the clinician must understand the uncertainty associated with absolute quantitative values reported in their specific center. In general, for all QNAT assays differences in absolute values of <0.50 log10 copies/mL should not be considered significant.

The large interlaboratory variability we observed in quantitative results reported suggests that the absence of an internationally recognized reference standard for EBV QNAT calibration is a significant problem. The introduction of QNAT for a number of pathogens such as human immunodeficiency virus (HIV), hepatitis C virus (HCV), hepatitis B virus (HBV), hepatitis A virus (HAV) and parvovirus B19 in blood donors internationally has resulted in the establishment of internationally recognized reference standards that have become regulatory requirements for use as calibrators for commercial and other assays being used to screen blood donors. This approach, combined with proficiency testing, commercial products and automation has had a significant impact on the reproducibility of QNAT assays, not only in blood donor testing but also in the diagnostic sector that also perform QNAT for these pathogens in support of patient management (17–21). Hayden et al. (11) observed that the use of common calibrators reduced interlaboratory variability in EBV VL testing. We believe that the establishment of an international reference standard for EBV VL is the most critical initial step in improving EBV VL assessment and it is realistic to set up a goal of reducing both inter and intralaboratory variability in EBV VL assessment to ≤± 0.50 log10 copies relative to expected values, consistent with standards for QNAT for other pathogens. Our data suggest that, at present, just under half of all reported results achieve that target.

Laboratory technique and staff competence may also be a factor in variability, particularly intralaboratory variability. The false-positive results reported from two laboratories highlights the vulnerability of this type of technology to cross-contamination events. Although, we did not examine the relationship between the number of VL assays done annually and laboratory performance in our study, this may be an important parameter to consider in future studies. In addition to assay calibration, the large diversity of assay design observed in our study including parameters such as extraction methods, gene targets and platforms/instruments used may be contributing to the variability observed, and may offer opportunities for quality improvement through further standardization. Theoretically, the more extensive use of well-defined commercial reagents/kits and automated systems for specimen handling and processing should reduce variability. However, we did not observe any significant difference in variability in interlaboratory quantitative precision when commercial reagents/assays rather than laboratory-developed assays were used. This may not be a reflection of the potential of commercial assays but simply reflect the variety of assays used and the lack of common calibrators that would allow standardization of quantitative results among manufacturers. Hayden et al. (11) observed large interlaboratory EBV VL result variability even when the extraction method was removed as a variable. Although our data are limited, we observed a trend toward less variability when plasma rather than cellular samples were used, illustrating the need to validate extraction methods and efficiency for specific sample types. The development and implementation of international reference standards for assay calibration will remove a currently important variable in EBV VL and facilitate the identification and standardization of other assay parameters that might be contributing to result variability.

In summary, it is important that transplant clinicians understand the current limitations of EBV VL as a laboratory tool and exercise caution in direct interinstitutional comparison of data dependent on the use of EBV VL. Recently, an international group [Standardization of Genome Amplification Techniques (SoGAT)] made up of regulators, representatives of industry producing laboratory standards and academics involved in laboratory science and practice, who historically have worked at the development of international World Health Organization (WHO) approved reference standards for NAT testing of blood donors, have taken on the responsibility of developing reference standards for both EBV and cytomegalovirus CMV VL assays. Working with the National Institutes of Biologic Standards and Controls (United Kingdom) and using the WHO framework for standards development and approval, this initiative will require laboratories internationally to participate in testing of potential reference material that will result in a consensus value assignment. It is important that the transplant community support this initiative, an important initial step in the standardization of this laboratory tool. In the interim, interinstitutional comparison of results could be improved by cross calibration through specimen exchange among laboratories or use of common commercial calibrators.

Acknowledgments

We would also like to acknowledge Kimberly Martin and Sandy Shokoples for their technical support at ProvLab, Dr. Alan Health and Dr. John Saldanha for statistical assistance as well as all laboratory participants in this study. This study was supported by the American Society of Transplantation and an arms-length educational grant to the Canadian Society of Transplantation by Roche Canada.

Dr. Angela M. Caliendo was supported in part by the Emory Center for AIDS Research (P30 AI050409).

The authors would like to acknowledge the participating laboratories:

BC Children's Hospital, Vancouver, BC: E. Thomas; Children's Hospital, Birmingham, AL: F. Lakeman; Children's Hospital of Eastern Ontario, Ottawa, ON: T. Karnauchow; Children's Hospital of Pittsburgh, Pittsburgh, PA: R. Wadowsky; Cleveland Clinic, Cleveland, OH: B. Yen-Lieberman; Emory University Hospital, Atlanta, GA: A. Caliendo; Erasmus Medical Center, Rotterdam, The Netherlands: H. Niesters; Institute for Medical Microbiology, Basel, Switzerland: H. Hirsch; Johns Hopkins Hospital, Baltimore, MD: M. Forman; Laboratoire regional de virologie du CHUQ-CHUL, Sainte-Foy, QC: G. Boivin; London Laboratory Services Group, London, ON: R. Wheeler; Mayo Clinic, Rochester, MN: M. Espy; Mount Sinai Hospital, New York, NY: S. Jenkins; Sainte Justine Hospital, Montreal, QC: C. Alfieri; St. Joseph's Healthcare, Hamilton, ON: J. Mahony; St. Louis Children's Hospital, St. Louis, MO; W. Mulford; Newfoundland Public Health Laboratory, St. John's, NL: S. Ratnam; Provincial Laboratory for Public Health, Calgary, AB: J. Fox; Provincial Laboratory for Public Health, Edmonton, AB: X. Pang; Stanford University Hospital, Stanford, CA: E. Thomas; Toronto Medical Laboratories/Mt. Sinai, Hospital, Toronto, ON: T. Mazzulli; UCLA Clinical Laboratories, Los Angeles, CA: D. Bruckner; University of Michigan, Ann Arbor, MI: D. Newton; University of North Carolina Hospital, Chapel Hill, NC: M. Gulley; University of Washington Medical Center, Seattle, WA: L. Cook; Vanderbilt University Medical Center, Nashville, TN: Y. Tang; ViraCor, Lee's Summit, MO: S. Arnoldi; Yale-New Haven Hospital, New Haven, CT: M. Landry.

Ancillary