Different signal-to-cut-off ratios from three automated anti-hepatitis C virus chemiluminescence immunoassays in relation to results of recombinant immunoblot assays and nucleic acid testing

Dear Sir,

The diagnosis of hepatitis C virus (HCV) infection in transfusion centres and clinical laboratories begins with serological tests for the detection of antibodies against HCV (anti-HCV). Currently, automated chemiluminescent immunoassay (CLIA) analysers are widely used and have been shown to have excellent sensitivity and specificity for analysing donated blood and routine clinical specimens^1–3. However, different false positive rates and seroconversion sensitivities have been reported for the various CLIA analysers².

The concentration of anti-HCV is expressed as a signal-to-cut-off (S/CO) ratio, and S/CO levels correspond to the antibody concentration¹. The Centers for Disease Control and Prevention recommended that HCV screening tests giving positive results be followed by high sensitivity and more specific supplemental assays such as a recombinant immunoblot assay (RIBA) or nucleic acid test, and that the clinical report include a statement that high S/CO ratios are usually (in ≥95% of cases) confirmed as being positive by supplemental tests². The aim of the present study was to investigate and analyse the S/CO ratios of three CLIA-reactive samples for predicting RIBA or HCV-RNA positive results. We also evaluated the diagnostic performances of three CLIA according to their optimal S/CO ratios.

All 87,675 specimens were consecutively screened for anti-HCV at Seoul St. Mary’s hospital (Seoul, Korea) between July 2008 and December 2010. Of these specimens, 24,896 were tested by the Architect anti-HCV assay (Architect, Abbott Laboratories, Abbott Park, Illinois, United States of America) on the Architect i2000, 12,178 by the ADVIA Centaur HCV assay (ADVIA) (Siemens Healthcare Diagnostics, Inc., Tarrytown, New York, United States of America) on the ADVIA Centaur system, and 50,601 by the Elecsys Anti-HCV assay (Elecsys, Roche Diagnostics, Mannheim, Germany) on the Modular Analytics E170. Out of all the samples that were screened, 1,994 (2.3%) were anti-HCV positive. There was no significant difference in positive rates among the three CLIA (2.2% on Architect, 2.1% on ADVIA, and 2.4% on Elecsys assay, P >0.05). The proportions of samples from patients with chronic liver diseases were similar among the CLIA-positive samples (P >0.05). The distribution of S/CO ratios of anti-HCV positive samples showed great variability among the CLIA (Architect: 1–17.8, ADVIA: 1–11, Elecsys: 1–1,350) (Figure 1). Comparisons were made using the χ² test for categorical data and the Mann-Whitney U test for non-normally distributed variables. All P values were two-tailed and P values <0.05 were considered statistically significant.

Figure 1.

Percentage distribution of the sera identified by RIBA (n =1,994) (a) and HCV-RNA testing (n =582) (b) as HCV-positive (black bars), HCV-indeterminate (grey bars), or HCV-negative (white bars) according to the S/CO ratios of the three CLIA. P: positive; ID: indeterminate; N: negative.

RIBA was additionally performed on all 1,994 CLIA-positive samples using LG HCD Confirm (LG Life Sciences, Seoul, South Korea). RIBA detected Core 14 (core), Core 518 (core/NS3), KHCV 897 (NS3), E1E2NS4 (E1, E2 membrane/NS4), and NS5 1.2 (NS5). Among the anti-HCV positive samples, 54.0% were RIBA-positive, 5.1% were RIBA - indeterminate, and 40.9% were RIBA-negative. Higher S/CO ratios corresponded to increasing RIBA-positive rates for each CLIA. In the receiver operating characteristic analysis of CLIA-positive samples, the optimal cut-off points of the S/CO ratios were chosen to identify RIBA-positive or RIBA-indeterminate samples. The estimated S/CO ratios of 7.5 (Architect), 10.0 (ADVIA Centaur), and 115 (Elecsys) showed a RIBA sensitivity of 92.0–96.3% and specificity of 88.3–96.6% (Table I). There was no difference in the positive detection rates of specific RIBA antigens for the three CLIA. For the RIBA-indeterminate samples, ADVIA showed a higher positive rate for anti-NS3 and Elecsys showed a higher positive rate for anti-core antigen (P <0.05). In a previous study in which blood donor and unselected routine serum samples were analysed, the specificities of the CLIA were reported to be above 99% for all subjects including the screen negative samples⁴. In the present study, we evaluated the diagnostic performance only for samples that screened positive, which may explain the lower specificity of our assays. The Architect assay showed a high sensitivity (94.9%) and positive predictive value (97.4%), and the Architect S/CO ratios were better correlated with RIBA reactivity. The ADVIA assay also showed excellent specificity and the highest RIBA-positive rate. Because HCV infection is often asymptomatic, highly sensitive serological screening tests play a fundamental role in the diagnostic process. In our study, the Elecsys assay had a high sensitivity (96.3%), and the Elecsys HCV-positive/RIBA-indeterminate samples showed higher positive rates for anti-core antigen, suggesting that these samples were truly positive.

Table I.

Diagnostic anti-HCV cut-off points (S/CO ratios) of the three CLIA for identifying RIBA-positive or -indeterminate samples.

	Architect	ADVIA	Elecsys
	Cut-off ratio
Performance	≥7.5	≥10.0	≥115
Area under the curve	0.984 (0.974–0.994)*	0.960 (0.938–0.982)	0.949 (0.937–0.962)
Sensitivity, %	94.9 (91.8–97.0)	92.0 (86.9–95.5)	96.3 (94.6–97.5)
Specificity, %	96.6 (93.3–98.5)	96.3 (89.4–99.2)	88.3 (85.2–91.0)
Negative predictive value, %	93.3 (89.4–96.1)	84.6 (75.5–91.3)	94.5 (92.0–96.4)
Positive predictive value, %	97.4 (94.9–98.9)	98.2 (94.7–99.6)	91.9 (89.6–93.8)

Legend Receiver operating characteristic plots were used to estimate the cut-off points of each CLIA by SPSS version 12.0 (SPSS, Chicago, Illinois, United States of America).

^*95% confidence intervals.

HCV-RNA was measured in 582 available samples using quantitative reverse transcriptase polymerase chain reaction assay (BioSewoom, Inc., Seoul, South Korea). The lower detection limit of HCV was 50 IU/mL. Viraemia was detected in 328 (56.4%) samples [97 (72.4%) of the Architect, 31 (63.3%) of the ADVIA, and 200 (50.1%) of the Elecsys HCV-positive samples]. The HCV-RNA-positive rates were increased in relation to the higher S/CO ratios, similar to previous reports showing that a high anti-HCV value is associated with viraemia⁵. Since anti-HCV levels tend to decline with time after the resolution of an infection, low S/CO ratios should be expected in a proportion of donors screened. In addition, anti-HCV screening can be persistently negative while HCV-RNA test results are positive for patients with chronic HCV infection, including those who are immunocompromised. In these patients, it is difficult to identify a cut-off S/CO ratio for detecting viraemia during routine clinical testing. The HCV-RNA test should, therefore, be performed for immunocompromised patients with undetermined HCV infection status regardless of the RIBA results.

When comparing the anti-HCV positive samples by HCV-RNA or RIBA, viraemia was detected in 76.1% of the RIBA-positive samples, 35.5% of the RIBA-indeterminate samples, and 6.2% of the RIBA-negative samples. Among the RIBA-positive samples, S/CO ratios showed similar distributions between the HCV-RNA-positive and -negative groups. All sera that were RIBA-indeterminate/HCV-RNA-positive were from immunocompromised patients.

The limitations of this study were the relatively small number of HCV-RNA tests, routine HCV screening samples from mixed populations including patients with chronic HCV infection, immunecompromised patients and healthy individuals, and lack of direct comparison of the performances of the different CLIA with the same samples. Despite these limitations, our study established different cut-off values for three widely used anti-HCV screening CLIA at a single centre.

In summary, our study analysed the S/CO ratios of HCV-positive samples using three different CLIA, comparing the ratios with the results of RIBA and HCV-RNA assays. While the three HCV CLIA showed comparable levels of diagnostic performance, the distributions and cut-off levels of S/CO ratios varied greatly among the three CLIA. Our data may be helpful for interpreting CLIA HCV-positive results from routine clinical laboratory testing and reducing the need for confirmatory tests.