We conducted this study to evaluate the analytical and clinical performance of LiCA^® anti-HCV across three clinical laboratory centers in Shanghai and Beijing, China. Comprehensive assay specifications including analytical characteristics and diagnostic performance were validated as described in Figure 1 . All samples were de-identified, separated into vials according to assay frequency and stored at −80°C before testing unless otherwise noted. Tests were performed with samples in freezing once only.

A total of 16,305 unselected clinical patient sera were enrolled for anti-HCV screening tests in parallel on the LiCA^® 500 system (Chemclin Diagnostics, Beijing, China) and the Architect^® i2000SR system (Abbott Laboratories, IL, USA). LiCA^® anti-HCV is a third-generation immunoassay and based on the random-access and fully automatic LiCA^® system (Wang et al., 2022). The assay uses two nano-scale beads (streptavidin-coated sensitizer beads and recombinant viral antigen-coated emission beads) and biotin-labeled antigens to bridge a one-step double-antigen sandwich reaction with the biotin-streptavidin amplification format. The antigens contain targeting epitopes from core, NS3, and NS4 domains of the viral protein. After incubation, the immunocomplex can be formed if anti-HCV exists in the sample. The sensitizer and emission beads are linked through the antibody–antigen binding chain, resulting in a short distance (<200 nm) between these two beads. Therefore, the singlet oxygens that are generated from the sensitizer beads can diffuse into the emission beads and initiate a chemiluminescence. When the sample does not contain anti-HCV, there is no binding reaction. A larger distance (>200 nm) between the sensitizer and emission beads blocks the singlet oxygen shifting; thus, no chemiluminescence occurs. With the unique methodology, a LiCA^® assay does not need any washing step to separate the immunocomplex from the free components, which may produce interfering signals if not washed out completely during a traditional immunoassay rather than LiCA^®. Architect^® anti-HCV is a two-step indirect immunoassay based on chemiluminescent microparticle technology. The particles contain antigens representing the regions of core, NS3, and NS4. Unlike the LiCA^® method, the Architect^® assay requires two steps of washing to elute the free components out of the reaction cuvette. Measurement with a ratio of S/Co ≥1.0 is regarded to be reactive and a negative result is considered as S/Co <1.0 for both assays.

Subjects with reactive S/Co ratios on any one assay were repeated in duplicate and identified by reflexing with the RIBA 3.0 assay (Mikrogen Diagnostics, Neuried, Germany). The RIBA results were classified as positive, negative, and indeterminate. The RIBA-indeterminate was further investigated using the real-time PCR test (Roche Diagnostics, Mannheim, Germany), which has the limit of detection (LoD) ≤15 IU/mL for HCV RNA. The RIBA-indeterminate but RNA-positive was considered to be true-positive while the RIBA-indeterminate and RNA-negative was further validated with the Roche Cobas^® anti-HCV assay. Cobas^® anti-HCV is a one-step double-antigen sandwich immunoassay that is based on the electrochemiluminescence method. The result is classified to be reactive as S/Co ≥1.0, non-reactive as S/Co <0.9, or borderline as S/Co within 0.9–1.0. The testing sequence is shown in Figure 2 .

Precision analysis in S/Co ratios was performed using patient sera and controls following the EP15-A3 protocol of the Clinical and Laboratory Standards Institute (CLSI) (CLSI, 2014). A maximum of 15% in coefficient of variation (CV) was acceptable. The C₅₀ test was guided by EP12-A2 using a serial of pooled sera away from both sides of the C₅₀ target (CLSI, 2008). Acceptance was considered as the C₅–C₉₅ interval was within C₅₀ ± 15%.

Assay on-board stability was evaluated using patient sera and controls refer to EP25-Ed2 (CLSI, 2023). Measurements were consecutively conducted over 20 days. Reloading a new reagent package and re-calibration were not allowed during study of this section. A difference of each point from the mean value of less than 3 times the standard deviation (SD) and 15% was considered to be acceptable.

Variability in lot-to-lot reagents and between-instruments was assessed refer to EP09c (CLSI, 2018) using 479 residual sera that were collected from randomly selected patients with a broad range of S/Co ratios. Comparative data were recorded in parallel. A satisfactory agreement was considered as correlation coefficient R > 0.975, slope within 0.85–1.15, intercept close to zero, and bias <15%.

To evaluate plasma equivalency to serum, we prepared 25 groups of anti-HCV-negative samples with 10 different types of matrix in serum and 9 commonly used plasma such as ethylenediaminetetraacetic acid (EDTA), heparin, and citrate. Each group was collected from the same individual free of HCV. Positive samples were prepared with the negative ones by spiking a high-positive anti-HCV specimen at 20:1 volume ratio. S/Co values were recorded for each group at the same time. A satisfactory equivalency was considered as percent difference was <15% in positive samples or S/Co difference <0.15 in negative ones.

We used 239 banked samples in different sources of plasma and serum, which has been identified with HCV genotypes by Roche real-time PCR in our labs, and 29 commercially available seroconversion panels in the matrix of sodium citrate for this study. Panels PHV913–926 were from SeraCare (n = 6, Milford, MA, USA), HCV6222–10235 were from ZeptoMetrix (n = 19, Buffalo, NY, USA), and SCP-HCV-004–012 were from BioMex (n = 4, Heidelberg, Germany). All samples were measured on LiCA^® and Architect^® anti-HCV in parallel.

We collected 108 samples free of HCV but positive of interfering substances, such as other viral antibodies, human anti-mouse antibody (HAMA), and auto-antibodies, to evaluate cross-reactivity from potential interferents. The specimens included different sources of plasma and serum samples. All samples were measured on LiCA^® and Architect^® in parallel. A significant cross-reactivity was considered as a reactive S/Co was observed in a negative sample unexpectedly.

To identify possible interferences from hemoglobin, triglycerides, bilirubin, and biotin, we used two patient sera containing anti-HCV of 0.71 and 4.75 S/Co, respectively. Patient samples were diluted with serial concentrations of each interferent. S/Co values were recorded on each dilution. A significant interference was considered as percent recovery change ≥15%.

Complement factors are mainly present in fresh blood samples and disappear rapidly during storage. To rule out the potential influence of complement factors on the assay, 25 fresh HCV-negative sera were spiked with a fresh serum sample in strong-positive to anti-HCV to obtain 5 low- and 20 high-positive fresh pools. All samples were stored at 2–8°C and replicate tests were performed for each dilution on days 0, 1, 2, 3, and 4. A significant influence was considered as assay difference from the baseline S/Co on day 0 was ≥15%.

All data were analyzed by the software program MedCalc 20.0.22 (MedCalc Software, Mariakerke, Belgium) and Excel 2019 (Microsoft, WA, USA). Anti-HCV screening tests were classified to be reactive as S/Co ≥1.0 or nonreactive as S/Co <1.0. Positive and negative results were confirmed upon RIBA and HCV RNA assays. A negative–positive reverse conversion was regarded to be primarily significant from the qualitative perspective. More valuable information was obtained with quantitative analysis in S/Co ratios. A statistical significance was considered as p < 0.05.

Based on the assay S/Co values, repeatability and within-lab imprecision of LiCA^® anti-HCV were determined to be 1.31%–3.27% across low and high levels of quality controls (QC) and patient sera ( Supplemental Table 1 ). From the perspective of qualitative analysis, the C₅–C₉₅ interval was identified to be −5.44%–5.03% away from the C₅₀ target ( Supplementary Figure 1 ). LiCA^® anti-HCV presented an excellent analytical precision.

Figure 3 shows daily waves of the assay S/Co on pooled sera and controls over a consecutive 20-day monitoring and demonstrates a good on-board stability. Most measurements were narrowed in a deviation within ±2 SD from the average. Only few points jumped slightly over the line of ±2 SD. In low and high S/Co levels of serum samples, total CV was 5.21% and 3.09%, respectively. The assay difference of each point from mean was within −11.18%–11.40%. In controls, total CV was 5.22% in QC-low and 5.36% in QC-high, respectively. The assay difference of each point from mean was within −7.08%–11.32%. No negative–positive reverse conversion was observed during 4 weeks of assays.

A good concordance was validated between different reagent lots and analyzers in serum pools with either full range (0.05–380) or low levels (<10) of S/Co ratios ( Supplementary Figure 2 ). The Spearman correlation coefficient was determined to be 0.976–0.998 and mean bias was 0.33%–4.94%. At the reactive cutoff value (S/Co =1. 0), assay difference from lot-to-lot reagents and between-instruments was estimated to be −1.32%–1.27%.

No significant deviation was observed in nine types of commonly used plasma from serum matrix ( Supplemental Table 2 ). In negative samples, assay S/Co values in any kind of plasma were measured to be almost equivalent to those in serum from the same individual. In positive groups, we recorded a minus but acceptable bias (−11.26 to −3.15%) in plasma from serum. No reverse conversion between negative and positive expectations was observed.

Early detection of HCV infection was evaluated with commercially available seroconversion panels in sodium citrate ( Table 1 ). Of 29 panels studied in total, LiCA^® presented 14 in earlier (mean −8.8 days, range −28 to −3 days), 3 in later (mean +5.3 days, range 3–11 days), and 12 in equal detection compared to Architect^®. Overall, LiCA^® detected seroconversion in an average of 28.9 (95% CI: confidence interval, 18.4–39.3) days and showed mean 3.7 (6.4–1.0) days earlier than Architect^®. The Wilcoxon test demonstrated that there was a significant difference in seroconversion detection between two assays (p = 0.0056).

A total of 239 banked samples were used for detection of HCV genotypes ( Supplemental Table 3 ). Both LiCA^® and Architect^® correctly detected all samples with reactive S/Co values.

No significant cross-reactivity was detected on LiCA^® to 19 sources of potential interferents such as various viral antibodies, HAMA, and auto-antibodies ( Supplemental Table 4 ). Only background signals (S/Co < 0.2) were recorded in all 108 samples free of HCV but positive of interfering substances.

A percent recovery change <15% was observed in both low and high S/Co levels of specimens (0.71 and 4.75) by spiking interferents up to 169.5 mmol/L triglycerides, 4,276.0 µmol/L bilirubin, 5.0 g/L hemoglobin, or 409.0 nmol/L biotin, respectively.

The assay difference on day 1–4 storage from the baseline S/Co on day 0 was determined to be −1.37%–8.85% and −2.65%–10.34% in 5 low-positive (mean S/Co = 5.80) and 20 high-positive (mean S/Co = 84.10) fresh serum pools, respectively. No significant influence was observed on the assay to complement factors.

Of 16,305 clinical patient serum samples, 287 were confirmed to be anti-HCV true-positive and 16,018 were true-negative ( Figure 2 ). The diagnostic performance (95% CI) for anti-HCV detection is summarized in Table 2 . Compared to Architect^®, LiCA^® presented a significantly better performance in specificity (99.91% vs. 99.14%, n = 16,018, p < 0.0001), positive predictive value (PPV, 95.29% vs. 67.06%, n = 419, p < 0.0001), and overall accuracy (99.89% vs. 99.12%, n = 16,305, p < 0.0001), while no significant difference in sensitivity (98.61% vs. 97.91%, n = 287, p = 0.5217) and negative predictive value (NPV, 99.98% vs. 99.96%, n = 15,886, p = 0.3021). Among 16,305 patient sera, LiCA^® detected 4 false-negative (0.25‰) and 14 false-positive (0.86‰) anti-HCV cases, while Architect^® recorded 6 false-negative (0.37‰) and 138 false-positive (8.46‰) subjects, respectively ( Table 3 ). We listed all of 10 false-negative assays in terms of S/Co, positive RIBA bands, and RNA results in Table 4 . LiCA^® presented 100% of agreement with Cobas^®. There was one case that was confirmed to be positive (857 IU/mL) by HCV RNA. LiCA^® detected with a good-reactive S/Co of 12.11 but Architect^® misdiagnosed with an entirely nonreactive S/Co of 0.05.

With stratified S/Co ratios in a range of 1.00–2.99, 3.00–4.99, and ≥5.00, and overall tested-reactive results (≥1.00), the proportion of confirmatory positive results was determined to be 33.33% (n = 21), 100% (n = 4), and 100% (n = 272), and 95.29% (n = 297) on LiCA^® in contrast to 16.79% (n = 137), 50.00% (n = 28), and 96.06% (n = 254), and 67.06% (n = 419) on Architect^®, respectively ( Table 3 ). The distribution of S/Co ratios in more detailed stratification clearly revealed that Architect^® had much more false-positive tests especially in a range of 1.0–5.0 S/Co ( Figure 4 ). LiCA^® reported the results with much wider distribution of S/Co ratios (0–475.87 vs. 0–24.94) and less count in the “borderline” range of 0.9–5.0 S/Co (26 vs. 174) than Architect^® did.

Probit regression estimated that the predicting S/Co value (95% CI) with a positivity ≥95% was 3.28 (2.74–4.12) on LiCA^® ( Supplementary Figure 3 ). Using the predicting value as a cutoff, the data set of this study presented 100% of PPV and 99.93% of NPV.

LiCA^® showed a weak correlation with Architect^® based on the assays with S/Co ≥1.0 ( Supplementary Figure 4A , R = 0.296, p < 0.0001). An improved correlation was observed with logarithmic transformation of S/Co ratios ( Supplementary Figure 4B , R = 0.650, p < 0.0001). Overall agreement (95% CI) between both methods was 99.07% (98.91%–99.21%, n = 16,305) with Cohen’s kappa 0.78 (0.75–0.82). Positive and negative agreements were 67.30% (62.58%–71.78%, n = 419) and 99.91% (99.84%–99.95%, n = 15,886), respectively ( Table 3 ).

A total of 152 (0.93%, n = 16,305) discordant measurements were observed between these two assays ( Supplemental Table 5 ). Most discrepancies (133/152, 87.50%) resulted from the false-positive assays on Architect^®, especially from those with a lower reactive S/Co between 1.0 and 5.0 (123/133, 92.48%). For these discordant assays between LiCA^® and Architect^®, LiCA^® presented a high agreement with Cobas^® (143/152, 94.08%). There were 47 (0.28%, n = 16,305) discordances with RIBA-indeterminate but RNA-negative results. The Cobas^® anti-HCV assay was used for further validation and confirmed that 46/47 were false-positive and 1/47 was false-negative from Architect^®. A perfect consistency (47/47, 100%) was observed between LiCA^® and Cobas^® assays. The detailed records were summarized in Supplemental Table 6 .