ABSTRACT
This study, involving 20 laboratories and using currently available assays for hepatitis C virus RNA quantification, demonstrated that differences in viral load values are due not to interlaboratory variations but rather to the nature of the assay itself. This underlines the importance of using the same assay in multicenter studies or when monitoring antiviral therapy.
In a previous multicenter study designed to assess the expertise of laboratories involved in large-scale therapeutic trials for providing reliable data on hepatitis C virus (HCV) viral loads (VLs) (11), we found a wide heterogeneity of results depending both on the laboratories and on the HCV RNA quantification assays. Since that study, new assays based on real-time (RT) PCR have been developed. These assays are very useful tools for antiviral therapeutic monitoring, because they offer several advantages, including lower limits of detection and larger dynamic ranges of quantification (1-3, 8). Moreover, the availability of automatic platforms makes them attractive for the monitoring of HCV infection, and hence an increasing number of laboratories have decided to introduce these assays for routine use. Thus, a variety of HCV RNA quantification assays are now in use in laboratories that may be involved in large-scale trials. This prompted us to conduct a multicenter study in order to evaluate the expertise of such laboratories for determining HCV VL in this new context.
The panel included 15 samples: 13 samples collected from HCV-infected blood donors (one was represented in duplicate, genotype 2k) and one negative sample. The characteristics of this panel are described in Table 1. For each positive sample, HCV genotype was determined with a method based on the analysis of the NS5b region sequence (9).
The 20 participating laboratories provided a total of 21 results: 9 with bDNA Versant HCV 3.0 (Bayer Healthcare, Eragny, France), 6 with HCV Monitor 2.0 (Roche Diagnostics, Meylan, France), 4 with Cobas Ampliprep/Cobas TaqMan 48 HCV (Roche Diagnostics), and 2 with Abbott RealTime HCV PCR (Abbott Diagnostics, Rungis, France). The two latter assays are based on RT PCR. Table 1 shows the mean VLs obtained with each assay and the related coefficients of variation (CV). No false-positive result was observed for the HCV RNA-negative sample. Although not designed to evaluate specificity, this study confirms that unlike first-generation assays (4), standardized commercial assays, including those based on RNA extraction platforms, guarantee an optimal specificity score.
In contrast to the observation recently made by our working group on hepatitis B virus DNA quantification assays (10), we observed no significant intra-assay differences. Indeed, all HCV RNA quantification assays showed a high reproducibility rate, as evidenced by the low variations observed, irrespective of the VL level and the genotype. As shown in Table 1, the interlaboratory CV ranged from 0.8 to 6.2% (mean, 2.6%) for Bayer bDNA assay, from 0.8 to 10.9% (mean, 3.52%) for the Roche Monitor assay, from 0.5 to 5.3% (mean, 2.25%) for the Cobas TaqMan 48, and from 0.1 to 15.2% (mean, 5.12%) for the Abbott RealTime HCV PCR. The relatively high CV for results from laboratories using the latter assay may suggest interlaboratory variability. As the calculation relied on only two independent values, more extensive studies are needed to assess more accurately the performance of this new method of HCV RNA quantification.
The low interlaboratory CV rate observed for each assay contrasts with the interassay CV, which reached values above 10% (Table 1) for two samples (sample 4, genotype 1b, and sample 12, genotype 4d). Interestingly, as shown in Fig. 1 and in Table 1, the difference between the highest and the lowest VLs for each sample ranged from 0.33 (genotype 4a sample) to 1.95 log10 (genotype 1a). For 12 of the 14 samples, the lowest value was given by the bDNA assay, which was performed by four laboratories, eliminating the hypothesis that differences in laboratory procedure were responsible. Moreover, as depicted in Table 1, the bDNA assay provided significantly lower VLs than Monitor (13 samples) and Cobas TaqMan (11 samples) assays. The two remaining samples were underestimated by Cobas TaqMan 48 (one genotype 4a sample) and Abbott RealTime (one genotype 4d sample). This supports the concern of previous studies about the ability of the new RT PCR assays to correctly quantify samples of genotype 4 (2, 12). However, more studies of genotype 4 are required before a conclusion on the ability of the new RT PCR assays to correctly quantify such samples can be reached. The highest values were obtained in eight cases with the Monitor assay and in six cases with the Cobas TaqMan HCV 48.
Many recent comparative studies of commercially available HCV RNA quantification assays have shown contradictory results, in particular concerning the ability of assays to accurately quantify certain HCV genotypes. A significant underestimation of genotypes 2, 3, 4, and 5 was observed with the Cobas TaqMan 48 when the High Pure system was used for RNA extraction (5, 12). The same RT PCR method used in conjunction with the Cobas Ampliprep sample preparation device was described as having an efficiency equal to that of the Abbott RealTime HCV mostly for genotype 1 samples (7), while it was reported to underestimate genotype 4 VL (2, 12). In another study (8), the Cobas TaqMan provided a poor correlation with Amplicor Roche and Bayer bDNA assays, leading to a 1.5-log10 IU/ml difference in the quantification of the same sample in 5% of the cases. However, this study was based on the use of an extraction device different from those recommended by the manufacturer of this assay. Conflicting results have also been published concerning the performance of the Bayer bDNA assay: some studies have shown a good sensitivity irrespective of genotype (5, 6, 8), while another study demonstrated a constantly lower quantification of genotypes 1 to 3 (12). In comparison with the majority of previous studies, a major asset of our study is the involvement of several laboratories using the same assay; thus, the impact of potential interlaboratory variations on the comparison of the assays was minimized. In this way, we observed a global underestimation of HCV VL with the bDNA assay, since this method gave a mean VL lower than the total mean for all the samples, with a statistically significant difference for six of them (Table 1).
As a whole, these results demonstrate that any laboratory with good expertise in HCV RNA quantification techniques is able to provide reliable VLs when using a given assay. Therefore, the centralization of samples in one single laboratory is not necessary when a multicenter study is carried out. However, the interassay differences, in particular in the extreme VLs, lead us to recommend that the same assay be used in multicenter clinical trials or when the response to therapy in a single patient is being monitored.
Plot of the mean VLs for each assay. For each sample, means of the VLs (as log10 IU/ml) obtained by all laboratories using the same assay are shown. The upper and lower black squares indicate the extreme individual values, the box shows the 25th and 75th percentiles, and the line within the box indicates the median value. x axis, genotype; y axis, VL.
Mean VLs for each assay and comparison with the total meana
ACKNOWLEDGMENTS
This work was supported by a grant from the Agence Nationale de Recherches sur le SIDA et les Virus des Hépatites (ANRS).
We thank Véronique Descamps, Christine Portal, Annie Razer, and Patricia Zawadzki for their technical assistance and Christopher Payan for his helpful advice in the reading of the manuscript.
FOOTNOTES
- Received 18 April 2007.
- Returned for modification 17 June 2007.
- Accepted 15 September 2007.
- Copyright © 2007 American Society for Microbiology