Evaluating the utility of a two‐assay serological algorithm for hepatitis C screening in a low prevalence population

Abstract Introduction Screening for hepatitis C virus (HCV) is performed by testing for anti‐HCV antibodies, which may yield false‐positive results leading to additional testing and other downstream consequences for the patient. We report our experience in a low prevalence population (<0.05%) using a two‐assay algorithm aimed at testing specimens with borderline or weak positive anti‐HCV reactivity in the screening assay by a second anti‐HCV assay prior to confirming positive anti‐HCV results with RT‐PCR. Materials and Methods Retrospective analysis of 58,908 plasma samples was obtained over a 5‐year period. Samples were initially tested using the Elecsys Anti‐HCV II assay (Roche Diagnostics), with borderline or weakly positive results (defined in our algorithm as a Roche cutoff index of 0.9–19.99) reflexively analyzed using the Architect Anti‐HCV assay (Abbott Diagnostics). The Abbott anti‐HCV results dictated the final anti‐HCV interpretation for reflexed samples. Results Our testing algorithm resulted in 180 samples requiring second‐line testing, with final anti‐HCV results interpreted as 9% positive, 87% negative, and 4% indeterminate. The positive predictive value (PPV) of a weakly positive Roche result was 12%, which was significantly lower than the PPV using our two‐assay approach (65%). Conclusions The incorporation of a two‐assay serological testing algorithm in a low prevalence population provides a cost‐effective method of improving the PPV of HCV screening in specimens with borderline or weakly positive anti‐HCV results.

and afford treatment, and the prevalence of chronic HCV has slowly, yet consistently, declined. Unfortunately, given the financial barriers associated with these treatments, these therapies are not uniformly available, further illuminating the health disparities that plague even the most resource-rich countries. In contrast to the trends observed with chronic HCV, acute HCV infections have increased dramatically in the United States in the last 15 years, driven in large part by the ongoing opioid epidemic and associated intravenous drug use. 3 While trends regarding the incidence of acute HCV vary in different parts of the world, largely mirroring trends in intravenous drug use, it now predominantly occurs in younger adults, disproportionately impacting vulnerable communities that may lack access to high quality, consistent health care. These epidemiologic shifts highlight the importance of successful screening programs in identifying patients infected with HCV and facilitating early access to care. 4 Per guidelines from the United States Centers for Disease Control and Prevention (CDC), screening for HCV is based on the detection of circulating anti-HCV antibodies and should be performed once for all adults over age 18, for pregnant women in each pregnancy, and routinely in individuals at high risk of infection.
Routine, repeated HCV testing is also performed in specific clinical settings such as for patients undergoing hemodialysis. 5 The most common screening approach in the United States is testing with a single anti-HCV assay, with subsequent confirmation of positive results by detection of HCV RNA. One limitation to this approach is false-positive anti-HCV results, which are defined as a positive screening result in the absence of additional laboratory or clinical evidence of HCV infection. These false positives can lead to myriad adverse downstream consequences including erroneous diagnoses, additional testing, and unnecessary alerting of public health officials. Although commercially available anti-HCV assays demonstrate high specificity, widespread screening can result in a considerable number of false positives, particularly in low prevalence settings. 6 Published literature for a variety of anti-HCV tests demonstrates that false positives are more common in "weakly positive" specimens, such as those with low signal cutoff (S/CO) or cutoff index (COI), depending on the methodology employed. 7,8 Historically, the recombinant immunoblot assay (RIBA) served as a confirmatory step for samples with positive anti-HCV antibodies; however, the RIBA is no longer routinely available in the United States. Therefore, most laboratories now offer strategies for performing HCV RT-PCR as confirmatory testing for positive anti-HCV screening results. There are numerous approaches in use to accomplish this, many of which add complexity and additional expenses to specimen requirements (e.g., collection of a separate specimen tube and freezing of specimens). An alternative approach using a single-tube reflex to RT-PCR has been validated and utilized by some institutions to minimize workflow complexity and mitigate the need for additional appointments and sample collections. 9,10 Additionally, in a population with low HCV prevalence, such reflex testing may result in diversion of resources and increased manual effort and cost, despite the high specificity of contemporary anti-HCV serological assays. 11 An alternative screening approach, which has been adopted by several institutions including our own, is to reflexively perform a second anti-HCV assay for samples with weak positivity on the initial anti-HCV assay to mitigate false-positive anti-HCV results, prior to sending for confirmatory testing by RT-PCR. 8,[12][13][14] This approach can use the single specimen collected for the initial anti-HCV screening, without the more stringent specimen requirements needed for RT-PCR. An additional benefit is that the use of two automated assays can provide results with rapid turnaround time. Here, we report our experience using this two-step serological screening approach and evaluate its performance in a low prevalence population.

| Study design
This is a retrospective study performed at an 860 bed tertiary/quaternary care academic medical center in a mostly rural state. The medical center serves as the major referral and specialty hospital in the geographic region. This study was approved by the University of Iowa Institutional Review Board (protocol #202008415) with waiver of informed consent. We queried our institutional EMR database for patients that had HCV screening performed between February 1, 2017, and February 4, 2022, and identified 59,055 samples from 47,707 individual patients. We removed duplicate samples, as well as those that were hemolyzed, icteric, or quantity not sufficient, resulting in a final dataset of 58,908 samples from 47,706 patients ( Figure 1). The subset of samples (180 samples from 139 unique patients) that were borderline/weakly positive on initial screening using the Roche assay (see testing methodology) were isolated for chart review including prior HCV testing and history of HCV infection. To evaluate the performance of our diagnostic algorithm, we grouped patients into one of the three groups according to their history of HCV using an approach similar to that employed by others. 15 Patients were determined to have a history of HCV if they had more than one consecutive positive serological test (any methodology), were found to be HCV positive by RT-PCR at any point, or had a clearly documented history of HCV in the medical record (multiple unique references). Negative HCV history was defined as uniformly negative serological testing, a lack of documented history (no references by patient or provider), or a single positive serological result with subsequent negative serological testing. Cases that did not fit clearly into either category were considered to have a possible HCV history and given the lack of confidence in their HCV status, were excluded from analyses relying on knowledge of HCV history (e.g., positive predictive value).

| Methods
We adopted a two-assay testing algorithm, like those described in other studies, 8,12 in which a second anti-HCV serological assay is  I G U R E 1 Analysis workflow. A graphical depiction of the data that were collected and analyzed for this study. Bolded terms indicate the results reported to our electronic medical record.

F I G U R E 2 Testing Algorithm.
A graphical depiction of the HCV testing algorithm utilized by our institution. Bolded terms indicate the results reported to our electronic medical record based on the interpretation of our two-assay algorithm.
Anti-HCV II assay are not reported by the manufacturer in the assay package insert, these test characteristics have been evaluated independently in a separate publication that determined a sensitivity of 99.3% and specificity of 99.86% in a population where the prevalence of HCV was 1%. 15 The Roche results are reported in the form of a COI, which is the signal detected in the sample divided by the cutoff measured by the instrument based on signal in the negative and positive calibrators. According to the manufacturer's guidelines, COI < 0.9 indicates a negative test, COI ≥ 1 is positive, and values of 0.9 ≤ COI < 1 are considered borderline.
In our two-assay algorithm, COI values <0.9 are reported to the electronic medical record (EMR, Epic, Inc., version May 2021) as negative and COI values ≥20 are reported as positive; no additional reflexive serological testing is performed in these instances. In contrast, specimens with Roche COI values ranging from 0.9 to 19.99 (i.e., borderline/weakly positive) have previously been shown in other publications to contain many false positives when HCV prevalence is low. 12 We thus perform reflex testing of specimens with weakly positive anti-HCV screens by the Roche method with the Abbott ARCHITECT Anti-HCV assay (ARCHITECT Anti-HCV assay package insert, 2012) run on an Architect i1000 analyzer (Abbott).
In this subset of samples, Abbott anti-HCV results (reported as S/CO ratio) then dictate the final interpretation in accordance with manufacturer guidelines: a S/CO ratio <0.8 is considered negative, S/CO ≥1 is considered positive, and S/CO ranging from 0.80 to 0.99 is con-

| Statistical analysis
Statistical analyses were performed in GraphPad Prism (version 9.3.1) using nonparametric methods, either the Kruskal-Wallis test for comparison of three or more groups or the Mann-Whitney test for comparison of two groups.

| RE SULTS
Our institution utilizes a two-assay testing algorithm for detecting the presence of anti-HCV antibodies in patient plasma (Figure 2).
We first reviewed the efficacy of this approach by assessing the results of 58,908 anti-HCV tests from 47,706 patients performed over a 5-year span. A total of 97.7% (57,555 samples) of anti-HCV tests were negative (Roche COI < 0.9) and 2.3% (1338 samples) were positive (Roche COI ≥ 20). A total of 0.3% (180 samples) had borderline/weakly positive Roche COI values ranging from 0.9 to 19.99 and were thus tested by the second anti-HCV assay (Abbott) in our twoassay algorithm (Figure 1). The baseline characteristics of the patients whose samples are represented in this study (i.e., borderline/ weakly positive Roche anti-HCV) are included in Table 1.
The low prevalence of HCV infection in our patient population (~0.026%, Iowa Department of Public Health, 2020) led us to ask whether our two-step testing algorithm was beneficial. To address this, we closely examined the subset of samples that required both Roche and Abbott serological assays. These consisted of 180 samples (0.3%) from 139 unique patients that had borderline/weakly positive results using the Roche anti-HCV assay (0.9 ≤ COI < 20) and Finally, as the primary goal of our current algorithm is to enhance the specificity of HCV screening in samples with borderline or weakly positive anti-HCV results, we sought to directly determine the ability of our two-assay approach (Roche + Abbott) to predict positive HCV history relative to a single-assay approach using Roche alone (positive results defined by the Roche recommend cutoff of COI ≥ 1). Our two-assay approach significantly improved PPV from 12% using Roche only to 65% using Roche and Abbott. When restricting this analysis to the subset of samples from patients with confirmatory RT-PCR, our two-assay approach again outperformed Roche alone, improving PPV from 16% to 50%.

| DISCUSS ION
Acute HCV infection often causes a mild or moderate self-limiting illness with nonspecific symptoms that can be easily missed. While many acute infections are cleared spontaneously, over 50% of infections persist in the form of chronic hepatitis, which is frequently asymptomatic until irreversible complications develop. 16 Effective screening programs are therefore vital to facilitating diagnosis and treatment. Here, we explored the benefit of a two-step serological screening algorithm for weakly positive anti-HCV samples in a population with low HCV prevalence (<0.05%).  In our study, the major difference between a single-assay and two-assay approach is in the number of false positives. If we had followed the Roche assay package insert for COI interpretation, eight samples would have been resulted as "borderline" Our study has limitations meriting discussion. First, this study was performed at a single center in Iowa, and thus the number of samples are smaller and from a less diverse population than could be obtained in a multicenter study. Another limitation is that it is functionally impossible to be certain as to the true HCV status of most patients undergoing HCV screening, especially as current guidelines suggest no further testing following negative serology. A similar approach to determine HCV status to the one used here has been employed previously and served as a model for our study. 15 It is certainly encouraging that patients with a presumed history of HCV have statistically higher COI by the Roche assay than those without, but false negatives are still possible with our approach (Figure 4). Third, our borderline/weakly positive Roche COI range warranting additional anti-HCV testing by a secondary assay (i.e., 0.9 ≤ COI < 20) may not be optimal. Based on our anal- HCV II assay to be 99.3% and 99.86%, respectively, in a population with a high prevalence of HCV infection. 14 Using our two-assay approach, we would predict that specificity would be enhanced, a critical advantage when testing in low prevalence areas.
Taken together, we demonstrate that a two-assay serological testing algorithm for weakly positive anti-HCV samples significantly increases the PPV of HCV screening when compared to a singleassay approach using manufacturer's guidelines. We find that the extra testing required by such an approach comes at a minimal cost in our low prevalence population, affecting only 180 samples in

ACK N OWLED G M ENTS
We thank Alexandra Ehlers for her thoughtful review of the article.

CO N FLI C T O F I NTE R E S T S TATE M E NT
None declared.

F I G U R E 4
Anti-HCV screening results grouped by HCV history.
Results from Elecsys Anti-HCV II assay for all samples requiring second-line testing grouped by history of HCV infection. Data are graphed as COI with each result displayed as a single point, and error bars representing standard deviation. Statistics were performed using Mann-Whitney test, with **** indicating a p-value <0.0001.