The quantification of hepatitis C virus (HCV)

Transcription

1 Overestimation and Underestimation of Hepatitis C Virus RNA Levels in a Widely Used Real-Time Polymerase Chain Reaction Based Method Stéphane Chevaliez, 1,2 Magali Bouvier-Alias, 1,2 Rozenn Brillet, 1,2 and Jean-Michel Pawlotsky 1,2 The quantification of hepatitis C virus (HCV) RNA is essential for the everyday management of chronic hepatitis C therapy. Real-time polymerase chain reaction (PCR) techniques are potentially more sensitive than classical PCR techniques, are not prone to carryover contamination, and have a consistently wider dynamic range of quantification. Thus, they are rapidly replacing other technologies for the routine quantification of HCV RNA. We extensively evaluated the intrinsic characteristics and clinical performance of Cobas Ampliprep/ Cobas TaqMan (CAP/CTM), the most widely used real-time PCR assay for HCV RNA quantification. This study shows that CAP/CTM is sensitive, specific, precise, and reproducible and has a broad dynamic range of quantification well suited to HCV RNA monitoring in clinical practice. However, we identified 2 technical issues that will have an impact in clinical practice. First, the CAP/CTM assay overestimates HCV RNA levels in undiluted patient samples by approximately 0.6 log 10 international units per milliliter on average, and this overestimation increases with the viral load. Second, the CAP/CTM assay substantially underestimates HCV RNA levels in approximately 15% of genotype 2 samples and 30% of genotype 4 samples, probably because of mismatches with the target sequences due to the primer and/or probe design. Conclusion: As the CAP/CTM platform is widely available, easy to use, and suited to high-throughput screening for viral genomes, the manufacturer should improve the HCV RNA kit to resolve these 2 important technical issues that may affect everyday management of hepatitis C therapy. (HEPATOLOGY 2007;46:22-31.) The quantification of hepatitis C virus (HCV) RNA is essential for the management of chronic hepatitis C therapy based on the combination of pegylated interferon- and ribavirin. Indeed, the baseline Abbreviations: bdna, branched DNA; CAP/CTM, Cobas Ampliprep/Cobas TaqMan; cons geno X, consensus sequence of genotype X; HPC, high positive control; IU, international unit; LOD, limit of detection; LPC, low positive control; 5 NC, 5 noncoding; 5 NCR, 5 noncoding region; SD, standard deviation. From the 1 French National Reference Center for Viral Hepatitis B, C, and Delta, Department of Virology, Hôpital Henri Mondor, Créteil, France; and 2 Institut National de la Santé et de la Recherche Médicale U841, Créteil, France. Received December 13, 2006; accepted February 3, This work is part of the activity of the VIRGIL European Network of Excellence on Antiviral Drug Resistance supported by a grant (LSHM-CT ) from the Priority 1 Life Sciences, Genomics and Biotechnology for Health program in the 6th Framework Programme of the European Union. Address reprint requests to: Jean-Michel Pawlotsky, M.D., Ph.D., Department of Virology, Hôpital Henri Mondor, 51 Avenue du Maréchal de Lattre de Tassigny, Créteil, France. jean-michel.pawlotsky@hmn.aphp.fr; fax: Copyright 2007 by the American Association for the Study of Liver Diseases. Published online in Wiley InterScience ( DOI /hep Potential conflict of interest: Nothing to report. Supplementary material for this article can be found on the HEPATOLOGY website ( HCV RNA level is a significant predictor of the likelihood of sustained viral eradication. 1-3 In addition, early virologic response monitoring is now the standard of care for tailoring treatment to the individual patient s response. 4-8 Finally, the advent of a number of new anti-hcv drugs 9,10 will require close monitoring of HCV RNA replication in ongoing and future clinical trials. HCV RNA quantification assays should ideally be sensitive, specific, accurate, precise, and reproducible and should have a broad dynamic range of quantification that covers all possible clinical situations. No commercially available HCV RNA assay based on classical polymerase chain reaction (PCR) or the branched DNA (bdna) method has all these qualities. 11,12 Recently, real-time PCR techniques have been developed for use in this setting. These methods are potentially more sensitive than classical PCR techniques, are not prone to carryover contamination, and have a consistently wider dynamic range of quantification, and this makes them particularly useful for quantifying the full range of viral genome levels observed in treated and untreated patients Real-time PCR is rapidly replacing other 22

2 HEPATOLOGY, Vol. 46, No. 1, 2007 CHEVALIEZ ET AL. 23 technologies for the routine quantification of HCV RNA. Currently, the most widely used real-time PCR assay for HCV RNA quantification is run on the Cobas Ampliprep/Cobas TaqMan (CAP/CTM) platform (Roche Molecular Systems, Pleasanton, CA). CAP/CTM combines the Cobas Ampliprep instrument, which ensures fully automated extraction of HCV RNA from 850 Lof a patient s serum or plasma, and the Cobas TaqMan 48 analyzer, which ensures fully automated real-time PCR amplification and detection, followed by the interpretation of HCV RNA quantitative results by means of Amplilink software. The manufacturer s stated limit of detection (LOD) for the CAP/CTM HCV RNA assay is 15 HCV RNA international units (IU)/ml (1.2 log 10 IU/ ml), and the dynamic range of HCV RNA quantification ranges from 43 to IU/ml ( log 10 IU/ml). In this study, we extensively evaluated the intrinsic characteristics and clinical performance of the CAP/ CTM assay for HCV RNA quantification. Our results illustrate not only the benefits of this technology in clinical practice but also the flaws of the current kit and their potential implications for therapeutic decision making. Materials and Methods Materials Standards. Two preparations of HCV genotype 1 viruses, NAP-HCV5E6 and NAP-HCV5E1 (AcroMetrix, Benicia, CA), that contained (6.7 log 10 ) and 50 (1.7 log 10 ) HCV RNA IU/ml, respectively, were used to study the analytical performance of the assay. Serial dilutions of each standard were used in the experiments. Clinical Specimens. Serum samples were obtained from 3 groups of patients. Group A comprised 205 HCVseronegative patients. Group B comprised 126 patients with chronic HCV infection. Genotype determination was performed by means of sequence determination in the nonstructural 5B region of the HCV genome followed by phylogenetic analysis, as recently described. 8 Group B comprised 29 patients with HCV genotype 1 infection, 27 with genotype 2 infection, 29 with genotype 3 infection, 30 with genotype 4 infection, 9 with genotype 5 infection, and 2 with genotype 6 infection. The sequence of the nearly full-length 5 noncoding region (5 NCR) of HCV, targeted by the CAP/CTM primers and probes, was systematically determined in the 126 HCV-positive clinical samples in group B. Group C was composed of another 20 patients with chronic HCV genotype 1 infection. The distribution of HCV genotype 1 subtypes was similar to that in group B. Methods Assessment of the CAP/CTM Performance. Analytical sensitivity. To determine the analytical sensitivity of the assay, the NAP-HCV5E1 standard was serially diluted from 50 to 7 IU/ml. Eight replicates of each HCV RNA concentration were tested. Specificity. The specificity of the assay was assessed by the testing of the 205 HCV-seronegative clinical specimens from group A. Linearity and accuracy: influence of the HCV genotype. The linearity of CAP/CTM quantification was assessed by the testing of the NAP-HCV5E6 standard, which contained IU/ml (6.7 log 10 IU/ml), both undiluted and after serial one-third dilutions to 30 IU/ml (1.2 log 10 IU/ml). Every dilution was tested 3 times in the same run, and the means and ranges were compared with the theoretical HCV RNA levels. In addition, among the 126 HCV-positive samples in group B, serial one-fifth dilutions down to signal extinction were tested for 9 genotype 1, 10 genotype 2, 7 genotype 3, 4 genotype 4, 2 genotype 5, and 2 genotype 6 samples. To assess the accuracy of CAP/CTM, we compared the results of CAP/CTM with those obtained for the same samples with the third-generation bdna-based assay, Versant HCV RNA 3.0 (Bayer HealthCare), in the 126 HCV-positive samples in group B. This assay has been shown to be precise and accurate and to equally quantify HCV genotypes 1-6 because of the use of a set of 6 capture and 17 extender oligonucleotide probes spanning the full-length 5 NCR and the 5 third of the core coding region for hybridization (Supplementary Fig. 1). 11,12,15-21 HCV RNA was quantified by both the bdna and CAP/CTM assays, and the results, expressed in log 10 international units per milliliter, were compared. Finally, samples from 20 group C patients with HCV genotype 1 infection were tested undiluted and at one-tenth dilution with both the CAP/CTM and bdna assays. Precision and reproducibility. For intra-assay precision, 3 replicates of one-third dilutions of NAP-HCV5E6 were tested down to 1.93 log 10 IU/ml, and 8 replicates of twothird dilutions of NAP-HCV5E1 were tested down to 0.85 log 10 IU/ml. To calculate interassay precision, the low positive control (LPC) and the high positive control (HPC) provided in the kits were tested 37 times in the corresponding run on different days. HCV RNA Quantification. CAP/CTM. HCV RNA was extracted from 850 L of serum in the automated Cobas Ampliprep extractor according to the manufacturer s instructions. The Cobas TaqMan 48 analyzer was used for automated real-time PCR amplification and the detection of PCR products. Generated data were analyzed with Amplilink software. HCV RNA levels were expressed as international units per milliliter.

3 24 CHEVALIEZ ET AL. HEPATOLOGY, July 2007 bdna. In the Versant HCV RNA 3.0 assay, HCV RNA was recovered from 50 l of serum and quantified in the semiautomated Bayer System 340 bdna analyzer (Bayer HealthCare) according to the manufacturer s instructions. HCV RNA levels were expressed as international units per milliliter. Analysis of 5 NCR Sequences. Sequence analysis of the almost full-length 5 NCR targeted by CAP/CTM real-time PCR amplification was performed on the 126 HCV-positive samples in group B. Nested PCR was used to amplify a 306 base pair fragment. The first round used external sense and antisense primers 5 NCRS (5 - GGGCGACACTCCRCCATRAA-3, nt 5-24) and 5 NCE/AS (5 -CCCTGCGCGGCAACAAGTA-3, nt ). Two microliters cdna was added to the PCR mixture containing 10 mm Tris-HCl (ph 8.5), 50 mm KCl, 2 mm MgCl 2, 10 mm dntps, and 10 pmol of the primers, in the presence of Taq DNA Polymerase Advantage. Amplification included initial denaturation at 95 C for 1 minute, followed by 35 cycles of denaturation at 95 C for 30 seconds, annealing at 55 C for 30 seconds, and elongation at 70 C for 30 seconds, followed by a final elongation step at 70 C for 5 minutes. The second round used internal sense primer HCV28 22 and antisense primer 5 NCIAS (5 -CTTTTGGTTTTTCTTTGAGG-3, nt ). The second PCR round included an initial denaturation step at 95 C for 1 minute, followed by 30 cycles of denaturation at 95 C for 30 seconds, annealing at 52 C for 30 seconds, and elongation at 70 C for 30 seconds, followed by a final elongation step at 70 C for 5 minutes. The amplification products (5 l) were run on 1.5% agarose gels. The gels were stained with ethidium bromide, and DNA was viewed under ultraviolet light. The PCR products were purified and sequenced as described previously. The sequences were aligned with ClustalW. The Genbank sequence accession numbers are AM to AM Statistical Analysis. Descriptive statistics are shown as the mean SD or the median and interquartile range as appropriate. Groups were compared with the Kruskall- Wallis or Mann-Whitney test. Relationships between quantitative variables were studied by means of regression analysis. P values less than 0.05 were considered significant. Results Intrinsic Performance of the CAP/CTM Assay Analytical Sensitivity. The analytical sensitivity of the assay was assessed by the testing of two-third dilutions of a standard containing 50 (1.7 log 10 IU/ml) to 7 IU/ml (0.8 log 10 IU/ml) 8 times in different experiments. All 8 replicates containing 50 (1.7 log 10 IU/ml) and 33 IU/ml (1.5 log 10 IU/ml) tested HCV RNA positive. In contrast, only 4 of the 8 replicates with 22 IU/ml (1.3 log 10 IU/ml), 1 with 15 IU/ml (1.2 log 10 IU/ml), 2 with 10 IU/ml (1.0 log 10 IU/ml), and 3 with 7 IU/ml (0.8 log 10 IU/ml) were HCV RNA positive in the assay. Specificity. None of the 205 samples from anti HCV-seronegative patients from group A tested positive above the LOD of 15 IU/ml (specificity: 100%, 95% confidence interval: 98.1%-100%). However, the result was expressed as target not detected in 203 cases and as below 15 IU/ml in the other 2 cases. One specimen was truly negative on retesting ( target not detected ), whereas not enough serum was available for the second sample. Precision and Reproducibility. Precision (intra-assay reproducibility) was determined by the testing of undiluted and one-third serial dilutions of a standard containing IU/ml (6.7 log 10 IU/ml) and undiluted and two-third serial dilutions of a standard containing 50 IU/ml (1.7 log 10 IU/ml). As shown in Table 1, the coefficients of variation varied from 0.22%-2.15% between (5.7 log 10 IU/ml) and 10 2 IU/ml (2.0 log 10 IU/ml) and from 9.83%-14.57% between 10 2 (2.0 log 10 IU/ml) and 7 IU/ml (0.8 log 10 IU/ml). As shown in Table 1, interassay variability was characterized by a coefficient of variation of 2.85% for the LPC and by a coefficient of variation of 0.98% for the HPC. Accuracy and Linear Quantification: Influence of the HCV Genotype Linear Quantification of Standard Panel Dilutions. Serial one-third dilutions of the NAP-HCV5E6 HCV genotype 1 standard were tested with CAP/CTM 3 times in 3 different experiments. As shown in Fig. 1A, a significant relationship was found between the average measured and expected HCV RNA levels (r , P ). The difference between the average measured and expected HCV RNA levels varied from log 10 IU/ml. These results were in keeping with accurate and linear HCV RNA quantification by CAP/ CTM over the range of tested values. The NAP-HCV5E1 standard, containing 50 IU/ml (1.7 log 10 IU/ml), was serially diluted to 7 IU/ml. As shown in Fig. 1B, although a significant relationship was found between the average measured and expected HCV RNA levels (r 0.880, P 0.03), the linearity of quantification was lost below 50 IU/ml, CAP/CTM overestimating HCV RNA levels within that range. Quantification of HCV RNA in Clinical Samples Containing HCV Genotypes 1-6. The 126 samples from patients with chronic hepatitis C infected by HCV

4 HEPATOLOGY, Vol. 46, No. 1, 2007 CHEVALIEZ ET AL. 25 Table 1. Intra-Assay (Precision) and Interassay Reproducibility of a CAP/CTM Real-Time HCV PCR Assay Dilution Control Target HCV RNA (log 10 IU/ml Number of Determinations Mean (SD) of Measured HCV RNA (log 10 IU/ml) Number of Replicates That Tested Positive (>15 IU/ml) Coefficient of Variation (%) Intra-assay reproducibility A (0.05) 3/ A (0.01) 3/ A (0.04) 3/ A (0.11) 3/ A (0.02) 3/ A (0.07) 3/ A (0.08) 3/ A (0.06) 3/ A (0.06) 3/ A (0.04) 3/ A (0.15) 3/ B (0.19) 8/ B (0.26) 8/ B (0.19) 4/ B /8 ND B (0.21) 2/ B (0.10) 3/ Interassay reproducibility LPC LPC* (0.08) 37/ HPC HPC (0.07) 37/ NOTE. For intra-assay reproducibility, one-third serial dilutions of a standard containing IU/ml, that is, 6.7 log 10 IU/ml (A1-A11), were tested 3 times, and two-third dilutions of a standard containing 50 IU/ml, that is, 1.7 log 10 IU/ml (B1-B6), were tested 8 times. For interassay reproducibility, the kit LPCs and HPCs were tested 37 times in different experiments. *Three kit lots were used, with an LPC value between 2.07 and 3.07 in two of them and with an LPC value between 2.20 and 3.20 in the remaining one. Three kit lots were used, with an HPC value between 6.20 and 7.20 in one, with an HPC value between 6.18 and 7.18 in another, and with an HPC value between 6.32 and 7.32 in the remaining one. Not determined because only 1 of 8 tests was HCV RNA positive. genotypes 1-6 (group B) were tested with both CAP/ CTM and bdna. All these samples fell within the dynamic range of quantification of both assays. As shown in Fig. 2, there was a significant relationship between the HCV RNA levels obtained in the same samples with CAP/CTM and bdna for genotypes 1-5 (only 2 genotype 6 samples were tested, and this ruled out regression analysis). Apart from HCV genotype 4, all the regression lines were substantially above the expected equality line as a result of generally higher values obtained with CAP/ CTM than with bdna for a given sample (Fig. 2). Figure 3A shows a Bland-Altman analysis of HCV RNA levels measured in the 126 group B samples by both CAP/CTM and bdna. The figure plots the difference between the 2 measured values (CAP/CTM minus bdna) as a function of the mean of the 2 measurements. HCV RNA levels were overestimated by CAP/CTM in comparison with bdna in 112 of the 126 samples (median CAP/CTM minus bdna: 0.57 log 10 IU/ml). This overestimation of HCV RNA levels by CAP/CTM was significantly larger with samples containing higher viral loads. Indeed, the median CAP/CTM minus bdna values were 0.38, 0.31, 0.60, and 0.88 log 10 IU/ml for mean HCV RNA levels between 3.0 and 5.0, 5.0 and 6.0, 6.0 and 7.0, and 7.0 and 8.0 log 10 IU/ml, respectively (Fig. 3A; P ). Fig. 1. HCV RNA quantification by CAP/CTM of (A) undiluted and serial one-third dilutions of a commercial standard containing HCV RNA IU/ml, that is, 6.7 log 10 IU/ml (NAP-HCV5E6, AcroMetrix, Benicia, CA), and (B) undiluted and serial two-third dilutions of a commercial standard containing 50 IU/ml, that is, 1.7 log 10 IU/ml (NAP-HCV5E1, AcroMetrix).

5 26 CHEVALIEZ ET AL. HEPATOLOGY, July 2007 Fig. 2. Correlation between HCV RNA levels measured by CAP/CTM and bdna in the same samples from group B, which consisted of 126 clinical samples containing HCV genotypes (A) 1, (B) 2, (C) 3, (D) 4, (E) 5, and (F) 6. Despite the overall overestimation of HCV RNA levels by CAP/CTM in comparison with bdna, 9 samples containing HCV genotype 4 and 4 samples containing HCV genotype 2 had lower HCV RNA levels in CAP/ CTM than in bdna. Of these, 7 genotype 4 samples and 3 genotype 2 samples had a CAP/CTM minus bdna difference below the mean difference minus 1.96 SD (Fig. 3A). Box plots of individual differences between the 2 methods are shown for each genotype in Fig. 3B. They confirm the significantly lower overestimation of HCV RNA levels by CAP/CTM for HCV genotype 4 (median difference: 0.20 log 10 IU/ml) than for genotypes 1, 2, 3, and 5 (median differences: 0.71, 0.53, 0.66, and 0.39 log 10 IU/ml, respectively; P ). The largest overestimation of HCV RNA levels was seen with HCV genotypes 1 and 3 (Fig. 3B). Figure 4 shows the distribution of individual CAP/ CTM minus bdna differences within each genotype. The distribution was heterogeneous, and this made it impossible to establish a conversion factor between the 2 methods. Figure 4 also shows the extent to which HCV RNA levels were underestimated in 4 genotype 2 samples and in 9 genotype 4 samples. Target Sequence Analysis in Clinical Samples Containing HCV Genotypes 1-6. The nearly full-length 5 NCR of the HCV genome was sequenced in the 126 clinical samples in group B to determine the role of sequence polymorphisms in the assay differences among HCV genotypes and in the underestimation of HCV RNA levels in several genotype 2 and 4 samples. Figure 5A shows the alignment of the consensus sequences established for each genotype group in our study. Figure 5B,C shows the individual sequence alignments of genotype 2 and 4 samples, respectively, with respect to the corresponding consensus sequences. The sequences are ranked in order of increasing CAP/CTM minus bdna differences; that is, the samples for which the HCV RNA level was underestimated by CAP/CTM with respect to bdna are shown at the top. This presentation and the comparison with genotype-related variations shown in Fig. 5A allowed us to identify sequence polymorphisms associated with the underestimation of HCV RNA levels in these samples. Three polymorphisms were shared by underquantified samples in both the genotype 2 and genotype 4 groups, namely, an A-to-G substitution at position 107 of the 5 NCR, an A-to-T substitution at position 165, and the insertion of an A after position 206 (Fig. 5B,C). As shown in Fig. 5A, the genotype 1 and 3 consensus sequences bore a G at position 107, which did not appear to alter the overall efficiency of quantification for this genotype. In contrast, the A at position 165 was conserved across the genotypes, whereas the inserted A was almost exclusively seen in the samples with underestimated HCV RNA levels. The underquantified genotype 2 samples also often exhibited an A-to-T substitution at position 204, whereas the underquantified genotype 4 samples often had T-to-A and A-to-T substitutions at positions 203 and 205, respectively (Fig. 5B,C). As shown in Fig. 5A, the consensus sequence of genotype 3 bore a G at position 203, whereas an A was present at position 204 in the consensus sequence of genotypes 2 and 5, and a T was present in genotype 4. In contrast, position 205 was conserved across the genotypes. Supplementary Fig. 2 shows the alignments of the 5 NCR sequences of the 126 samples with respect to the genotype 1 consensus sequence. Linear Quantification of HCV-Infected Serum Dilutions. Serial one-fifth dilutions down to signal extinction were tested for 9 genotype 1, 10 genotype 2, 7 genotype 3, 4 genotype 4, 2 genotype 5, and 2 genotype 6 samples randomly selected among the 126 infected blood samples from group B. As shown in Fig. 6, the difference between the measured values in the undiluted serum ver-

6 HEPATOLOGY, Vol. 46, No. 1, 2007 CHEVALIEZ ET AL. 27 To confirm this observation, blood samples from 20 additional patients with chronic HCV genotype 1 infection (group C) were tested undiluted and after one-tenth dilution in both CAP/CTM and bdna. The average HCV RNA levels did not differ between the samples from group C and the genotype 1 samples from group B ( versus in bdna, respectively). Figure 8 shows the average HCV RNA levels obtained in CAP/CTM and bdna with the undiluted and diluted samples. The mean undiluted to one-tenth dilution difference was log 10 IU/ml with CAP/CTM and log 10 IU/ml with the bdna method (P ). Fig. 3. (A) Bland-Altman plot analysis of HCV RNA levels measured by both CAP/CTM and bdna in the 126 group B samples. The difference between HCV RNA levels measured by CAP/CTM and bdna is represented as a function of the mean of the 2 values. Different genotypes are represented by different colors. The gray area corresponds to the mean difference 1.96 SD. (B) Distribution of the differences between HCV RNA levels measured by CAP/CTM and bdna in the same samples according to the HCV genotype (genotype 6 is not shown because there were too few samples). The difference was significant (P ). Discussion This analysis of the intrinsic performance of CAP/ CTM HCV RNA quantification shows that the assay has an LOD of the order of the value stated by the manufacturer (15 IU/ml). However, the dilution containing 33 IU/ml was found to be HCV RNA positive in the 8 experiments, whereas further dilutions were not positive in all experiments. We also found that the specificity of the CAP/CTM assay was excellent, although results falling below the stated LOD of 15 IU/ml were reported either as below 15 IU/ml or as target not detected, potentially creating some confusion. We recommend that both results be interpreted as negative HCV RNA detection. The CAP/CTM assay was also very precise and reproducible, as already reported We identified 2 technical issues that may have an impact on CAP/CTM performance in clinical practice. The sus the value obtained with the same serum one-fifth diluted was significantly larger than any other difference between 2 subsequent one-fifth dilutions (P ). The expected difference for a one-fifth dilution being 0.70, the undiluted to one-fifth dilution difference was substantially larger than expected (on average, log 10 IU/ml), whereas the differences between subsequent dilutions were compatible with the expected value ( , , , , , and , respectively). The same phenomenon was noted with all 6 genotypes studied. Figure 7 shows individual examples of values obtained with serial one-fifth dilutions of samples containing HCV genotypes 1-4. Fig. 4. Individual differences between HCV RNA levels measured by CAP/CTM and bdna, ranked in increasing order of bdna values, according to the HCV genotype. Different HCV genotypes are represented by different colors. Bars above the 0 line correspond to samples for which the CAP/CTM value was higher than the bdna value, whereas bars below the 0 line correspond to samples for which the CAP/CTM value was lower than the bdna value in the same sample.

7 Fig. 5. Nearly full-length 5 noncoding (5 NC) sequence alignments of the 126 group B patients samples containing HCV genotypes 1-6. (A) Alignment of the consensus sequences of genotypes 2-5 with respect to genotype 1. The consensus sequences were established by the determination of the most frequently observed nucleotide at each position. Cons geno X is the consensus sequence of genotype X. No consensus sequence could be established for genotype 6, of which only 2 samples were analyzed. The recently proposed numbering of 5 NCR nucleotides was used. 27 (B) Alignment of individual 5 NC sequences in the samples containing HCV genotype 2. The sequences are presented with respect to the consensus sequence (cons geno 2). They are ranked in increasing order of CAP/CTM minus bdna differences (indicated in the second column). The average HCV RNA level, calculated from the results of the 2 methods, is shown in the first column on the left. (C) Alignment of the individual 5 NC sequences in the samples containing HCV genotype 4. The presentation and legend are the same as those for genotype 2.

8 HEPATOLOGY, Vol. 46, No. 1, 2007 CHEVALIEZ ET AL. 29 first was a global overestimation of HCV RNA levels, and the second was occasional substantial underestimation of HCV RNA levels in samples containing HCV genotype 4 and, to a lesser extent, genotype 2. The comparator was the third-generation bdna assay, which is well calibrated to the World Health Organization HCV RNA standard and is also precise, reproducible, accurate, independent of the HCV genotype, and robust to sequence polymorphisms Mismatches with 1 or a few of the probes in the bdna assay are unlikely to affect the final result. Although not a gold standard for HCV RNA quantification, the third-generation bdna assay is thus a satisfactory comparator for assessing the accuracy of other HCV RNA quantification technologies. Our results showed substantial overestimation of HCV RNA levels in the CAP/CTM method with respect to the bdna assay. This phenomenon affected all HCV genotypes. In addition, we found that this overestimation increased with the viral load of the sample. This overestimation was not related to inappropriate calibration of the CAP/CTM assay. Indeed, we found that HCV RNA levels were overestimated by the CAP/CTM assay exclusively in undiluted clinical samples but not when the same samples were tested after dilution; this phenomenon was observed with all 6 HCV genotypes tested. The molecular basis for this HCV RNA overestimation by CAP/ CTM in undiluted samples is unknown. Our results point to a biochemical interaction with a blood component during one of the reaction steps, an interaction that vanishes when the concentration of this putative component is reduced in diluted samples. This would explain why the problem was not observed during assay calibration, which was done with dilutions of secondary standard panels. For the same reason, the problem was not seen when we tested the assay calibration with dilutions of the Acrometrix standard, itself a secondary standard derived from an infected patient s plasma. This overall overestimation of HCV RNA levels by CAP/CTM may have implications for patient management. A switch to CAP/CTM will result in a global increase in observed HCV RNA levels. Patients who have been tested before the switch should thus be retested by Fig. 6. Distribution of the differences in the HCV RNA levels between a given sample and the next one-fifth dilution in a series of 34 clinical samples randomly selected among the 126 group B samples, containing HCV genotype 1 (n 9), genotype 2 (n 10), genotype 3 (n 7), genotype 4 (n 4), genotype 5 (n 2), or genotype 6 (n 2). The difference was significant (P ). Fig. 7. HCV RNA levels measured by CAP/CTM in serial one-fifth dilutions of clinical samples containing HCV genotype (A) 1, (B) 2, (C) 3, or (D) 4. Two examples are shown for each genotype.

9 30 CHEVALIEZ ET AL. HEPATOLOGY, July 2007 Fig. 8. Mean SD HCV RNA levels measured by CAP/CTM (left) and bdna (right) in undiluted samples (black) and one-tenth dilutions (gray) of the same samples from 20 patients infected by HCV genotype 1. The mean SD difference between the undiluted sample and the one-tenth dilutions is shown. These differences were significantly different in the 2 methods. CAP/CTM if they are on treatment and their virologic response is being evaluated. More generally, serial HCV RNA quantification should always take place in the same laboratory, with the same assay, to ensure appropriate therapeutic decisions. Finally, the average reported viral load will increase in clinical trials, and recommendations based on specific HCV thresholds will have to take this into account. It will also be mandatory to determine whether the same phenomenon occurs with other realtime PCR assays and with other viral markers. The second issue raised by this study is the underestimation of HCV RNA values in genotype 2 and 4 samples (15% and 30% of tested samples, respectively). Underestimation of this type is generally observed when mismatches occur between the primer pair and/or the TaqMan probe and the target viral sequences. As shown in Fig. 5A, we confirmed the sequence differences in the 5 NCR among the 6 genotypes. These well-known differences were likely taken into consideration by the manufacturer during the design of the primers and probe to make the assay suitable for all HCV genotypes. Our analysis of individual 5 NCR sequences provided interesting clues as to why HCV RNA levels were underestimated in some samples. Indeed, several polymorphisms were found exclusively in underestimated samples. Interestingly, 3 of them were common to both genotype 2 and 4 samples, namely, an A-to-G substitution at position 107 (but this position bearsagingenotypes 1 and 3), an A-to-T substitution at position 165 (the A being highly conserved among all HCV genotypes), and an insertion of an A after position 206 (a change almost exclusively seen in underestimated samples). One polymorphism was seen only in underestimated genotype 2 samples (A-to-T substitution at position 204), whereas 2 were seen exclusively in the underestimated genotype 4 samples (T-to-A and A-to-T substitutions at positions 203 and 205, respectively). These results strongly suggest that one or more of these nucleotide substitutions were responsible for mismatches with the primers and/or probe, resulting in the underestimation of the HCV RNA level in the sample. Unfortunately, this could not be confirmed as Roche Molecular Systems refused to provide us with the sequences or positions of their proprietary primers and probe. Our data show that HCV RNA is accurately quantified by the CAP/CTM assay in most clinical samples containing HCV genotypes 2 and 4. However, a Genbank search showed that the polymorphisms that we found to be associated with the underestimation of HCV RNA levels in the CAP/CTM assay were present in approximately 5% of displayed HCV genotype 2 sequences and in 7%-25% of displayed HCV genotype 4 sequences. This occasional underestimation of HCV RNA levels in patients infected by these genotypes will be an issue in clinical practice because it cannot be predicted at the individual level. The underestimation of baseline HCV RNA levels may result in erroneous appreciation of the early virologic response. It is also unknown whether the magnitude of this underestimation is affected by the decline in HCV replication during therapy. The interpretation of log HCV RNA declines in these patients may well be erroneous and lead to inappropriate therapeutic decisions in the context of future guidelines based on the virologic response at week 4 for all HCV genotypes. In conclusion, this study shows that the real-time PCR CAP/CTM method of HCV RNA quantification is sensitive, specific, precise, and reproducible and that it has a broad dynamic range of quantification well suited to HCV RNA monitoring in clinical practice. However, we identified 2 important technical issues with this widely used assay that have implications for clinical practice: (1) the CAP/CTM assay overestimates HCV RNA levels in undiluted patient samples by approximately 0.6 log 10 IU/ml on average, and this overestimation increases with the viral load, and (2) the CAP/CTM assay substantially underestimates HCV RNA levels in approximately 15% of genotype 2 samples and 30% of genotype 4 samples, probably because of mismatches with the target sequences due to the primer and/or probe design. As the CAP/CTM platform is widely available, very easy to use, and particularly suited to high-throughput screening for viral genomes, the manufacturer should improve the HCV RNA kit to resolve these 2 important technical issues that may affect everyday management of hepatitis C therapy.

10 HEPATOLOGY, Vol. 46, No. 1, 2007 CHEVALIEZ ET AL. 31 Acknowledgment: We are grateful to Françoise Darthuy, Jocelyne Rémiré, and Guillaume Dameron for their helpful technical assistance. The CAP/CTM kits used in this study were kindly provided by Roche Diagnostics (Meylan, France). References 1. Fried MW, Shiffman ML, Reddy KR, Smith C, Marinos G, Goncales FL Jr, et al. Peginterferon alfa-2a plus ribavirin for chronic hepatitis C virus infection. N Engl J Med 2002;347: Hadziyannis SJ, Sette H Jr, Morgan TR, Balan V, Diago M, Marcellin P, et al. Peginterferon-alpha2a and ribavirin combination therapy in chronic hepatitis C: a randomized study of treatment duration and ribavirin dose. Ann Intern Med 2004;140: Manns MP, McHutchison JG, Gordon SC, Rustgi VK, Shiffman M, Reindollar R, et al. Peginterferon alfa-2b plus ribavirin compared with interferon alfa-2b plus ribavirin for initial treatment of chronic hepatitis C: a randomised trial. Lancet 2001;358: National Institutes of Health Consensus Development Conference Statement: management of hepatitis C: 2002-June 10-12, HEPATOLOGY 2002;36(Suppl):S3-S Davis GL, Wong JB, McHutchison JG, Manns MP, Harvey J, Albrecht J. Early virologic response to treatment with peginterferon alfa-2b plus ribavirin in patients with chronic hepatitis C. HEPATOLOGY 2003;38: Ferenci P, Fried MW, Shiffman ML, Smith CI, Marinos G, Goncales FL Jr, et al. Predicting sustained virological responses in chronic hepatitis C patients treated with peginterferon alfa-2a (40 KD)/ribavirin. J Hepatol 2005;43: Jensen DM, Morgan TR, Marcellin P, Pockros PJ, Reddy KR, Hadziyannis SJ, et al. Early identification of HCV genotype 1 patients responding to 24 weeks peginterferon alpha-2a (40 kd)/ribavirin therapy. HEPATOLOGY 2006;43: Bronowicki JP, Ouzan D, Asselah T, Desmorat H, Zarski JP, Foucher J, et al. Effect of ribavirin in genotype 1 patients with hepatitis C responding to pegylated interferon alfa-2a plus ribavirin. Gastroenterology 2006;131: Pawlotsky JM. Therapy of hepatitis C: from empiricism to eradication. HEPATOLOGY 2006;43(Suppl):S207-S Pawlotsky JM, Gish RG. Future therapies for hepatitis C. Antivir Ther 2006;11: Pawlotsky JM. Use and interpretation of virological tests for hepatitis C. HEPATOLOGY 2002;36(Suppl):S65-S Pawlotsky JM. Molecular diagnosis of viral hepatitis. Gastroenterology 2002;122: Chevaliez S, Pawlotsky JM. Use of virologic assays in the diagnosis and management of hepatitis C virus infection. Clin Liver Dis 2005;9: Chevaliez S, Pawlotsky JM. Hepatitis C virus serologic and virologic tests and clinical diagnosis of HCV-related liver disease. Int J Med Sci 2006;3: Beld M, Sentjens R, Rebers S, Weegink C, Weel J, Sol C, et al. Performance of the new Bayer Versant HCV RNA 3.0 assay for quantitation of hepatitis C virus RNA in plasma and serum: conversion to international units and comparison with the Roche Cobas Amplicor HCV Monitor, version 2.0, assay. J Clin Microbiol 2002;40: Germer JJ, Heimgartner PJ, Ilstrup DM, Harmsen WS, Jenkins GD, Patel R. Comparative evaluation of the Versant HCV RNA 3.0, Quantiplex HCV RNA 2.0, and Cobas Amplicor HCV Monitor version 2.0 assays for quantification of hepatitis C virus RNA in serum. J Clin Microbiol 2002; 40: Morishima C, Chung M, Ng KW, Brambilla DJ, Gretch DR. Strengths and limitations of commercial tests for hepatitis C virus RNA quantification. J Clin Microbiol 2004;42: Elbeik T, Markowitz N, Nassos P, Kumar U, Beringer S, Haller B, et al. Simultaneous runs of the Bayer Versant HIV-1 version 3.0 and HCV bdna version 3.0 quantitative assays on the system 340 platform provide reliable quantitation and improved work flow. J Clin Microbiol 2004;42: Elbeik T, Surtihadi J, Destree M, Gorlin J, Holodniy M, Jortani SA, et al. Multicenter evaluation of the performance characteristics of the Bayer Versant HCV RNA 3.0 assay (bdna). J Clin Microbiol 2004;42: Hendricks DA, Friesenhahn M, Tanimoto L, Goergen B, Dodge D, Comanor L. Multicenter evaluation of the Versant HCV RNA qualitative assay for detection of hepatitis C virus RNA. J Clin Microbiol 2003;41: Trimoulet P, Halfon P, Pohier E, Khiri H, Chene G, Fleury H. Evaluation of the Versant HCV RNA 3.0 assay for quantification of hepatitis C virus RNA in serum. J Clin Microbiol 2002;40: Soler M, Pellerin M, Malnou CE, Dhumeaux D, Kean KM, Pawlotsky JM. Quasispecies heterogeneity and constraints on the evolution of the 5 noncoding region of hepatitis C virus (HCV): relationship with HCV resistance to interferon-alpha therapy. Virology 2002;298: Caliendo AM, Valsamakis A, Zhou Y, Yen-Lieberman B, Andersen J, Young S, et al. Multilaboratory comparison of hepatitis C virus viral load assays. J Clin Microbiol 2006;44: Germer JJ, Harmsen WS, Mandrekar JN, Mitchell PS, Yao JD. Evaluation of the Cobas TaqMan HCV test with automated sample processing using the MagNA pure LC instrument. J Clin Microbiol 2005; 43: Konnick EQ, Williams SM, Ashwood ER, Hillyard DR. Evaluation of the Cobas Hepatitis C Virus (HCV) TaqMan analyte-specific reagent assay and comparison to the Cobas Amplicor HCV Monitor V2.0 and Versant HCV bdna 3.0 assays. J Clin Microbiol 2005;43: Sarrazin C, Gartner BC, Sizmann D, Babiel R, Mihm U, Hofmann WP, et al. Comparison of conventional PCR with real-time PCR and bdna-based assays for hepatitis C virus RNA quantification and clinical significance for genotypes 1 to 5. J Clin Microbiol 2006;44: Kuiken C, Combet C, Bukh J, Shin IT, Deleage G, Mizokami M, et al. A comprehensive system for consistent numbering of HCV sequences, proteins and epitopes. HEPATOLOGY 2006;44: