Discordant Results of Two Automated Assays for Detection of Hepatitis C Virus-Specific Antibody

I am a member of a Veterans Administration facility routinely testing approximately 9,000 samples for hepatitis C virus (HCV) yearly, with about a 7% positive rate similar to the 8%-to-10% rate seen by Oethinger et al. (3) but lower than the 13%-to-15% rate reported by Dufour et al. for another population of veterans (2). Our current testing algorithm includes RNA testing of all positive samples in accordance with CDC guidelines for HCV testing (1).

For this evaluation I randomly selected 20 positive and 20 negative unlinked specimens from our routine HCV requests for comparison. These samples were evaluated for signal-to-cutoff ratio (S/CO) (with a correlation coefficient of 0.97) with two different AxSYM instruments at two different laboratories and with the Vitros Eci instrument. Positive S/COs for the assays were 1.21 and 1.0, with gray zones of 0.79 to 1.20 and 0.90 to 0.99 as stated in the package inserts for the AxSYM and Vitros instruments, respectively. There were no AxSYM-AxSYM discordant clinical results. I further evaluated an additional 10 “low-level positive” (as defined by the CDC) unlinked specimens that were known to have S/COs of less than 8 by the Vitros method (1). All specimens positive by either assay were tested for HCV RNA; discordant specimens were tested by Chiron RIBA 3.0 SIA.

In the first group of 20 positive specimens (S/CO ranges, 4.27 to 86.81 [AxSYM] and 1.10 to 36.1 [Vitros]), I found four samples (20%) that were positive by Vitros HCV and negative by AxSYM HCV (Table 1, row 1, and Table 2, samples 2, 7, 9, and 11 of the discordant specimens). Of these initial four discordant samples, all were negative for RNA, two were RIBA negative, and two were RIBA indeterminate. These four samples would be considered false positive by CDC guidelines. The first 20 negative specimens were concordant.

I then analyzed similar data from the 10 additional positive specimens (Table 1, row 3, and Table 2, samples 17, 18, and 19 and samples 1, 3, 4, 5, 6, 8, and 10 of the discordant specimens). Three samples were concordant positive, with one sample positive for RNA. There were seven samples that were Vitros HCV positive and AxSYM HCV negative. All seven samples were negative for RNA, two were RIBA indeterminate, and five were RIBA negative. All of the discordant specimens had S/COs of less than 5 in the Vitros HCV assay. Such false positivity has been reported in recent literature (3).

Oethinger et al. have used this fact to modify their Vitros HCV supplemental testing algorithm to exclude supplemental testing of all samples with an S/CO below 5 (reported as borderline) while continuing to perform supplemental testing on samples with S/COs of up to 20 (3). All of the Vitros discordant data shown in the tables would have been reported as “borderline” had this algorithm been used in our laboratory. I note that such algorithms are assay specific and that exact exclusions may not necessarily be applicable to other assays such as AxSYM (1).

In total, 13 of 30 (43%) positive specimens tested were found to be false positive for the Vitros anti-HCV assay, while 2 of 30 (7%) were found to be false positive for AxSYM HCV. This was a reduction of false positives with AxSYM HCV of 11 (36%).

Differences between the two assay formats alone could not account for this false positivity. The major difference in the capture phase of the two assays is the inclusion of NS5 in the Vitros anti-HCV assay. It has been noted in the literature that the addition of NS5 may be responsible for nonspecific reactivity in HCV assays (5), but I surmise that NS5 alone is not responsible for the results seen, as only 3 of the 11 discordant specimens gave a RIBA result of NS5 +/−.

At our facility, using only the initial 40 samples and applying a 20% false positivity rate, as confirmed by our validation with our annual test volume, we would reduce RNA testing by approximately 126 samples. At $65 per RNA test, which is at the low end of the cost spectrum, as our reference laboratory for hepatitis C RNA is another VA Medical Center, this is would be $8,190 in cost savings. As most of those results would have been negative RNA results, our algorithm would have mandated supplemental RIBA testing. At $105 per RIBA test (commercial laboratory), we would have incurred an additional cost of $13,230. We expect the new method to save over $20,000 annually in direct lab costs; this does not include the psychological and clinical care costs of 126 false-positive HCV tests.

• Copyright © 2006 American Society for Microbiology