Expertise of French Laboratories in Detection, Genotyping, and Quantification of Hepatitis C Virus RNA in Serum

Abstract

Before initiating new large-scale therapeutic trials for hepatitis C virus (HCV)-infected patients, the French Health Authorities for HCV research decided to organize an evaluation of the expertise of laboratories that could be engaged to undertake molecular biology assays in such trials; 21 experienced laboratories participated in this national evaluation of laboratory expertise, which was performed in two successive rounds. The first round evaluated the laboratories for their abilities to detect HCV RNA in serum, determine genotypes, and quantify HCV RNA loads. The results observed by qualitative assays for HCV RNA detection were 100% sensitivity and 100% specificity for all laboratories. The genotyping results were 100% concordant for 9 laboratories and greater than 90% for 10 laboratories. By contrast, large coefficients of variation were observed for quantitative determination of HCV RNA loads, leading to a second round with standardized quantitative assays only. The dispersion of the results was larger by the AMPLICOR HCV Monitor assay than by the branched-DNA assay (mean coefficients of variation, 57.4 and 16.9%, respectively). In the majority of cases, discrepancies between the results of the two tests were found for samples with high viral loads. These results indicate the usefulness of validating, by controlling for expertise, both the reliabilities of laboratories involved in multicenter work and the standardized assays chosen for use in the evaluation of the biological impacts of new therapies.

Antiviral treatments are becoming more and more efficient, and monitoring of therapy requires reliable viral quantification tests. Hepatitis C virus (HCV) is an important cause of chronic liver disease, and recent studies have shown the efficiency of a combination of pegylated interferon and ribavirin in the treatment of patients with chronic hepatitis C (7, 9). Furthermore, several studies have shown that the HCV genotype and viral load prior to treatment have predictive values for the response to antiviral therapy and that measurement of serum HCV RNA levels is useful in the monitoring of treated patients (2, 5, 8). Theses studies have also emphasized the need for standardization of genotyping and HCV RNA quantification methods, especially in the framework of multicenter therapeutic trials.

In France, clinical research on HCV infection is conducted by a national organization, called “Agence Nationale de Recherches sur le Sida” (ANRS), which was initially designed to support research on human immunodeficiency virus infection. Before initiating new large-scale trials of therapies for HCV infection, the HCV ANRS group decided to organize an evaluation of the expertise of laboratories that could be engaged to undertake molecular biology assays (HCV genotyping and detection and quantification of HCV RNA) in such trials.

We present here the results of this evaluation of national laboratory expertise, in which 21 experienced French laboratories participated. The aim of this study was to verify the capacities and limits of the laboratories in the routine use of HCV molecular biology assays in order to select the most reliable laboratories for future multicenter work and to select the most reliable assays.

MATERIALS AND METHODS

Twenty-one French laboratories participated in the study; among them, 18 belonged to universities or hospitals and 3 belonged to blood transfusion centers. The study was performed through two successive rounds.

First round.

The first round of the study was based on a test panel consisting of 30 serum samples, including (i) 8 negative control samples (already used in previous French multicenter quality control studies [1, 3]) collected from HCV antibody-negative blood donors with no risk factors for HCV infection and (ii) 22 anti-HCV-positive (as determined by enzyme-linked immunosorbent assay and radioimmunoblot assay; Ortho Clinical Diagnostics, Issy-les-Moulineaux, France) and viremic samples from HCV-infected individuals detected through the systematic biological screening of blood donations. The viral loads of these samples, quantified by the AMPLICOR HCV Monitor test (version 2.0; Roche Diagnostic Systems, Basel, Switzerland), ranged from 7.9 × 10³ to 1.9 × 10⁶ viral copies (VC)/ml. The HCV genotypes were determined by a second-generation line probe assay (Inno-LiPA HCV II; Innogenetics, Ghent, Belgium). The following HCV genotypes were represented in the panel: 1a (four samples), 1b (seven samples), 2 (three samples), and 3 (seven samples). An additional serum sample was coinfected with genotypes 2 and 4. All samples in the panel were prepared and aliquoted by an external investigator, who randomly sent coded aliquots (0.5 ml) to each participating laboratory for blind testing. All laboratories simultaneously tested the samples in the panel in a limited period of time in order to simulate routine testing conditions.

The tasks performed in the first round included the detection of HCV RNA in serum, genotype determination, and quantification of the serum HCV RNA load. The techniques used by the 21 laboratories are described below.

(i) Detection of HCV RNA.

For the detection of HCV RNA, 2 laboratories used AMPLICOR HCV (version 1; Roche Diagnostic Systems), 18 used Cobas AMPLICOR HCV (version 2; Roche Diagnostic Systems), and 1 used an in-house nested PCR.

(ii) Genotyping.

For genotyping, 17 laboratories used the Inno-LiPA HCV assay, 1 used sequencing of the 5′ noncoding region, 1 used a nested PCR followed by detection by the DEIA Sorin assay (DiaSorin, Antony, France), 1 used an in-house procedure, and 1 used a serotyping assay (Murex, Dartford, United Kingdom).

(iii) Quantitative assay.

For HCV RNA quantitation, 19 laboratories used one assay and 2 laboratories used two assays, giving a total of 23 determinations: 11 by the branched-DNA (b-DNA) HCV RNA (version 2.0) assay (Chiron-Bayer) (data are expressed in milliequivalents per milliliter); 6 by the AMPLICOR HCV Monitor assay, version 1 (Roche Diagnostic Systems) (data are expressed in VC per milliliter); and 6 by the Cobas AMPLICOR HCV Monitor assay, version 2 (Roche Diagnostic Systems) (data are expressed in VC per milliliter).

Second round.

Given the discrepancies and large coefficients of variation observed for quantitative determination of the HCV RNA loads (see below), a second round was organized (as stated in the study protocol) specifically for standardized quantitative assays. The panel used included 20 different anti-HCV-positive and viremic serum samples which were blindly tested by the 21 laboratories. The following HCV genotypes were represented in this panel: 1a (2 samples), 1b (10 samples), 2a/2c (3 samples), 3 (2 samples), and 4 or 5 (3 samples). Eleven laboratories used the AMPLICOR HCV Monitor assay (version 2.0), four used the b-DNA HCV assay (version 2.0), and six used both assays. Eleven of the 17 laboratories which used the AMPLICOR HCV Monitor assay performed sample dilutions when HCV RNA levels were above the highest limit of quantification (850,000 international units [IU]/ml), according to the instructions of the manufacturer. All results observed during the second round were expressed in IU per milliliter (8).

Statistical analysis.

For the first round, accuracy rates were calculated for the whole series of experiments and for each laboratory separately. The viral load was expressed as mean ± standard deviation (SD) and the coefficient of variation for each assay. For the second round, the results of each experiment were expressed as log₁₀ IU per milliliter, the mean ± SD, the range of values, and the coefficient of variation. For each sample, we pointed out results exceeding the interval of the mean ± 2 SDs. Linear regression analysis was performed to test the relationship between (i) the results of the two assays and (ii) the range of values and the viral loads obtained by both assays.

RESULTS

First round. (i) Qualitative assays.

The sensitivity and the specificity of the results of the qualitative assays were 100% for all laboratories. Neither false-positive nor false-negative results were observed for the 30 serum samples in the panel in any participating laboratory.

(ii) Genotyping.

Among the 17 laboratories using the Inno-LiPA assay, 9 had 100% accuracies (including the case of coinfection), 6 had 95% accuracies (the laboratories missed only the coinfection), 1 had 91% accuracy (the laboratory failed to detect the coinfection and did not correctly genotype one of the samples containing genotype 1b), and 1 had 86% accuracy (the laboratory failed to detect the coinfection and was unable to specify subtype 1a in two samples).

Among the laboratories using other genotyping assays, none had 100% accuracy; the laboratory that sequenced the HCV genomes in the 5′ noncoding region and the laboratory that used an in-house procedure had 95% accuracies (missing only the coinfection); the laboratory that used the Murex serotyping assay had 91% accuracy; and the laboratory that used a nested PCR followed by detection with the DEIA Sorin assay had 82% accuracy.

(iii) Quantitative assays.

Coefficients of variation ranged from 11 to 48% in the laboratories that used the b-DNA assay; from 9 to 55% in those that used the AMPLICOR HCV Monitor assay, version 1.0; and from 36 to 81% in those that used the AMPLICOR HCV Monitor assay, version 2.0.

Second round.

On the basis of the results of the HCV RNA load measurements obtained during the first round, the second round was undertaken, with quantification of HCV RNA performed exclusively by either of the two commercial assays available in France (the b-DNA assay and the AMPLICOR HCV Monitor assay). The correlation between the two tests was very strong (r = 0.975; P < 0.001).

When we evaluated the samples for those with out-of-range results (defined as results out of the interval of the mean ± 2 SDs), we found that 3 samples whose loads were measured by the b-DNA assay and 12 samples whose loads were measured by the AMPLICOR HCV Monitor assay had out-of-range results. Figure 1 represents the intervals between the out-of-range measures and the mean values calculated without them for the 12 samples tested by the AMPLICOR HCV Monitor assay. Six laboratories each gave one extreme result, and one laboratory gave six extreme results. If we consider the six results that deviated from the mean by more than 0.5 log unit, we observed that two results were given by the latter laboratory and four results were given by four other laboratories.

FIG. 1.

Out-of-range values observed by the AMPLICOR HCV Monitor assay in the second round. The bars represent the differences between each out-of-range measure and the mean value of the corresponding HCV RNA load calculated without it. The letters under the bars correspond to the laboratories where the out-of-range measures were found. The numbers on the abscissa correspond to the numbers of serum samples.

The dispersion of the results was greater by the AMPLICOR HCV Monitor assay than by the b-DNA assay. Indeed, the range of values varied from 0.38 to 1.98 log₁₀ IU/ml for the 20 samples tested by the AMPLICOR HCV Monitor assay and from 0.12 to 0.38 log₁₀ IU/ml for the same samples tested by the b-DNA assay (mean coefficients of variation, 57.4 and 16.9%, respectively) (Fig. 2). The HCV genotype had no influence on the range of the viral RNA loads observed by either assay. Dilution protocols may have influenced the range of values observed by the AMPLICOR HCV Monitor assay for the samples with high viral loads (regardless of the dilution, however, the results were scattered [data not shown]). Discrepancies between the two tests were not laboratory dependent since in the six laboratories which tested the samples by both assays, the dispersion of the values was greater by the AMPLICOR HCV Monitor assay than by the b-DNA assay. Interestingly, in the majority of cases the greatest discrepancies between the two tests were found for samples with high viral loads. By the b-DNA assay, the range of values was not dependent on the viral load (Fig. 3a). By contrast, by the AMPLICOR HCV Monitor assay, the values were more scattered as the viral load went higher (Fig. 3b).

FIG. 2.

Ranges of HCV RNA loads measured in the 20 serum samples in the second round by the AMPLICOR HCV Monitor and b-DNA assays.

FIG. 3.

Relationship between the mean viral load and the range of values measured by the 21 laboratories by the b-DNA assay (a) and the AMPLICOR HCV Monitor assay (b).

DISCUSSION

The results presented in this multicenter study underscore the usefulness of validating, by controlling for expertise, both the reliabilities of laboratories involved in multicenter work and the standardized assays chosen for use in the evaluation of the biological impacts of new therapies.

The qualitative assays for HCV RNA detection showed 100% accuracy. This emphasizes the usefulness of the multicenter work performed within the “Groupe Français d'Etudes Moleculaires des Hepatites,” which previously organized national quality control studies of such assays (6). Thus, the 21 laboratories involved in the present study may be considered qualified for reliable HCV RNA detection in the context of multicenter trials.

Although genotyping methods were not evaluated and compared to a “gold standard,” most of the laboratories that used the same standardized genotyping assay (the Inno-LiPA assay) had concordant results (19 of the 21 laboratories had accuracies greater than 90%). These 19 laboratories may thus be considered qualified for participation in multicenter studies, including HCV genotype determination. The other laboratories will need to improve the reliabilities of their procedures, possibly by using another kind of genotyping assay.

By contrast, poor results were obtained during the first round of the quantitative assay, as reported in other independent studies (4, 6, 8). Our second round, based on the two standardized procedures available in France, gave far better results: most of the laboratories had concordant data, with a good correlation between the two quantitative assays. Indeed, interlaboratory comparisons showed few differences except for one laboratory (laboratory B in Fig. 1) and for one particular sample.

For blood samples with high viral loads, the number of samples with out-of-range results was notably higher by the AMPLICOR HCV Monitor assay (despite the use of the dilution recommended by the manufacturer) than by the b-DNA assay. The lack of linearity observed by the AMPLICOR HCV Monitor assay for the highest levels of viremia prompted us to alert the Roche Diagnostics Department. This manufacturer provided new recommendations and revised the method so that the linearity of the assay was improved: the previous highest limit of quantification (850,000 IU/ml) was estimated to be too high, and a new limit of linearity (500,000 IU/ml) was established. Indeed, Roche Diagnostics recommended that the results for all samples with viral loads between 500,000 and 850,000 IU/ml be verified by performing 1/100 dilutions in HCV-negative serum. For samples derived from patients receiving treatment, it was recommended that the previous serum sample be retested in parallel with the new one.

In conclusion, our results show that, even with the commercial assays that have currently been registered by health administrations, it is necessary to organize evaluations of the expertise of laboratories using such assays.