Use of Hepatitis C Virus (HCV) Immunoglobulin G Antibody Avidity as a Biomarker to Estimate the Population-Level Incidence of HCV Infection

Abstract

Background. Sensitive methods are needed to estimate the population-level incidence of hepatitis C virus (HCV) infection.

Methods. We developed an HCV immunoglobulin G (IgG) antibody avidity assay by modifying the Ortho 3.0 HCV enzyme-linked immunoassay and tested 997 serum or plasma samples from 568 people who inject drugs enrolled in prospective cohort studies. Avidity-based testing algorithms were evaluated by their (1) mean duration of recent infection (MDRI), defined as the average time an individual is identified as having been recently infected, according to a given algorithm; (2) false-recent rate, defined as the proportion of samples collected >2 years after HCV seroconversion that were misclassified as recent; (3) sample sizes needed to estimate incidence; and (4) power to detect a reduction in incidence between serial cross-sectional surveys.

Results. A multiassay algorithm (defined as an avidity index of <30%, followed by HCV viremia detection) had an MDRI of 147 days (95% confidence interval [CI], 125–195 days), and the false-recent rates were 0.7% (95% CI, .2%–1.8%) and 7.6% (95% CI, 4.2%–12.3%) among human immunodeficiency virus (HIV)–negative and HIV-positive persons, respectively. In various simulated high-risk populations, this algorithm required <1000 individuals to estimate incidence (relative standard error, 30%) and had >80% power to detect a 50% reduction in incidence.

Conclusions. Avidity-based algorithms have the capacity to accurately estimate HCV infection incidence and rapidly assess the impact of public health efforts among high-risk populations. Efforts to optimize this method should be prioritized.

(See the editorial commentary by Mastro, Morrison, and Hamilton on pages 339–40.)

It is estimated that 185 million individuals are seropositive for hepatitis C virus (HCV) worldwide [1]. Although HCV prevalence and incidence have been stable or declining in many settings owing to human immunodeficiency virus (HIV) infection prevention efforts, particularly in people who inject drugs (PWID), emerging epidemics are occurring in some resource-limited settings [2], and HCV epidemiology has become dynamic in some high-income countries [3]. For example, in the United States, a surge of HCV infections has been noted among HIV-infected men who have sex with men [4]. There has also been a sharp increase in HCV infection incidence among PWID in some rural and suburban regions [5], with continued high incidence in some urban areas [6]. Prospective cohort studies of PWID in the United States collectively report an incidence rate of 27.7 cases per 100 person-years [7]. In parallel, screening guidelines have been updated [8, 9], and coverage of harm-reduction services has been expanded—especially in response to localized outbreaks [10, 11]. In addition, HCV treatment has been transformed by the development, regulatory approval, and availability of direct-acting antivirals that are now penetrating populations that could not previously be reached with interferon-based regimens [12–14]. Therefore, sensitive epidemiological tools are needed to monitor the impact of these interventions on this rapidly changing field [15].

Current methods to estimate HCV infection incidence are challenged, primarily because HCV infection is asymptomatic. The acute infection window period (during which individuals are HCV RNA positive and HCV immunoglobulin G [IgG] antibody negative) is 5–6 weeks [16–18]. Individual sample screening for HCV RNA–positive and HCV IgG antibody–negative samples to detect acute infections on a population-level requires large sample sizes, and approaches using pooled screening for acute infection may overestimate incidence [19]. The Centers for Disease Control and Prevention (CDC) surveillance system for the general US population is dependent on clinical report data of newly acquired HCV cases and uses a multiplicative strategy to overcome sample size barriers (an estimated 20 recent infections for every reported case is assumed) [20, 21]. Onofre et al have provided evidence that this surveillance system significantly underestimates HCV infection incidence [20]. Insights on the HCV epidemic have thus relied heavily on prospective follow-up of seronegative PWID, closely monitoring for acute infections and/or HCV IgG antibody seroconversion to estimate incidence [16, 17]. While this is the most appropriate population to study, prospective cohort studies are expensive, time-consuming, difficult to maintain, and prone to well-established biases [17, 22, 23].

An alternative method to estimate the incidence of primary infection is to use measurable biomarkers that change with infection duration in a predictable manner and can distinguish between recent and nonrecent infections at a particular cutoff [24]. This cross-sectional, biomarker-based approach has been successfully used to measure HIV infection incidence but has received little attention for HCV [25, 26]. It has previously been demonstrated that HCV IgG antibody avidity, the binding capacity of maturing HCV-specific IgG antibody to antigen, correlates with infection duration [27–30]. However, previous HCV avidity assays were developed using samples with a limited distribution of infection duration and/or were evaluated by conventional statistical measures for discrimination of recent and long-standing infections—which are not reflective of the capacity to estimate population-level incidence. In this study, we explored the utility of HCV avidity-based testing algorithms that include HCV RNA status to estimate population-level HCV infection incidence via cross-sectional surveys.

METHODS

Study Specimens

The Baltimore Before and After Acute Study of Hepatitis (BBAASH) and AIDS Linked to the IntraVenous Experience (ALIVE) studies are ongoing, prospective, community-recruited, observational cohort studies of PWID in Baltimore, Maryland [16, 31]. Participants in each cohort provided written informed consent and received counseling to reduce drug use. Detailed study protocols were previously described [16, 31]. In brief, HCV-seronegative, HCV RNA–negative PWID were invited to enroll in the BBAASH cohort, providing blood samples at monthly visits to allow for the detection of HCV RNA and/or seroconversion. The ALIVE cohort enrolled HIV-negative and HIV-positive PWID, 89% of whom were HCV seropositive at baseline [32], and collected semiannual blood samples during follow-up. Subjects consented to long-term storage (at −80°C) of serum and plasma samples collected from them. This study conformed to all relevant ethical guidelines; the institutional review board of the Johns Hopkins University approved both cohort study protocols and the present analysis.

Samples were selected due to the availability of sufficient plasma or serum, HCV-positive serostatus, and known duration of infection. For samples obtained from BBAASH seroconverters, the date of seroconversion was estimated by 2 different methods, depending on if there was an acute time point identified during follow-up (HCV RNA positive and HCV IgG antibody negative). For individuals with such an acute time point, the estimated date of seroconversion was 22.5 days after the HCV RNA–positive and HCV antibody–negative visit date (16 of 56 individuals), since it was previously demonstrated in this cohort that the median time to HCV seroconversion is 45 days after infection [16]. For all other individuals, the estimated date of seroconversion was the midpoint between date of the last HCV antibody–negative visit and the date of the first HCV antibody–positive visit (40 of 56 individuals). The exact date of seroconversion for samples from the ALIVE cohort was unknown, but samples were only selected from individuals known to be HCV seropositive for >2 years.

Modified Ortho Avidity Assay

HCV IgG antibody avidity was determined by a modified protocol developed using the Ortho HCV Version 3.0 ELISA Test System (Ortho Clinical Diagnostics, Raritan, New Jersey). Samples were incubated in duplicate (intra-assay) on an Ortho 3.0 HCV EIA plate (10 µL of specimen in 200 µL of sample diluent) for 30 minutes at 4°C. The plate was washed with 1X wash buffer (300 µL) 5 times. For the dissociation step, the first well of each sample was manually treated with 200 µL of 0.025M diethylamine (diluted with dH₂O) to dissociate antigen antibody complexes, and the second well of the same sample was treated with 1X wash buffer as a control. The plate was then incubated for 30 minutes at 37°C, followed by washing with 1X wash buffer (300 µL) 5 times. The remainder of the assay (beginning with the addition of conjugate) was performed per manufacturer's protocol. Washes were automated using a Biotek ELx405TM Microplate washer. The absorbance of each well was read using a SpectraMax M5 microplate reader.

For each sample, HCV avidity was reported as an avidity index value (a percentage), which was calculated as the ratio of absorbance (optical density) of the well treated with diethylamine to that of the well treated with wash buffer only. An Ortho avidity index of 100% signified the strongest antibody-antigen binding. Inter-operator reproducibility was assessed among a subset of 82 samples enriched for recent infection.

Additional Laboratory Testing

Samples were assessed for HCV RNA status or levels by the Cobas Amplicor HCV Monitor, Version 2.0; a Cobas Taqman HCV Test, Version 2.0 (Roche Applied Science); or the Abbott RealTime HCV Amplification Reagent Kit (Abbott, Des Plaines, Illinois) per the manufacturer's protocol. Owing to the use of multiple assays, the presence of HCV viremia was categorically defined as an HCV RNA level of ≥500 IU/mL for this analysis. HIV serostatus, HIV viral load, and CD4⁺ T-cell count were previously determined [31].

Statistical Analyses

Avidity-based testing algorithms were characterized by 2 incidence estimation parameters needed to use a general biomarker-based incidence estimator: the mean duration of recent infection (MDRI) and the false-recent rate [24, 33, 34]. These parameters are dependent on a time scale defined by a cutoff T, which truncates the range of time during which a biomarker can be considered indicative of recent infection. Since the humoral immune response to HCV infection is delayed [16], the time scale for this analysis was time after HCV seroconversion, and T was defined as 2 years [35].

The MDRI, defined as the average time an individual is identified as recently infected, was calculated by binomial regression without exclusion criteria, using samples collected ≤2 years after HCV seroconversion; the statistical method for calculating the MDRI and its 95% confidence intervals (CIs) has been described elsewhere [24, 33, 36]. The false-recent rate was calculated as the proportion of samples collected >2 years after HCV seroconversion that were misclassified as recent, and its 95% CIs were determined from a binomial distribution.

Among sera collected >2 years after HCV seroconversion, we determined factors associated with a reduced avidity index (ie, <80%) to explore indicators that may contribute to the potential misclassification of long-term infections as recent infections. The probability of an avidity index of <80% was expressed by univariate prevalence risk ratios (PRRs) calculated by modified Poisson regression models with a robust variance estimator—a preferred method to estimate risk when the prevalence of the outcome is >10% [37]. Models also incorporated generalized estimating equations to account for multiple samples (different visits) from the same individual. In a multivariate complete-case analysis controlling for demographic characteristics, adjusted PRRs (adjPRRs) were calculated for HCV viremia and HIV-induced immunosuppression as categorical variables.

The performance of 3 testing algorithms to estimate incidence were compared: method 1, HCV RNA detection among HCV IgG antibody negative samples (or individual acute HCV screening) [16]; method 2, Ortho avidity (index <30%) and HCV RNA detection; and method 3, a combination of methods 1 and 2. Precision of each testing algorithm was assessed in hypothetical contexts for demonstrative purposes. Precision was characterized by the sample size necessary to achieve a desired relative standard error (RSE; 30% and 15%) of the incidence estimate, as guided by the WHO/UNAIDS Technical Working Group on HIV Incidence Assays, which recommends the RSE of the incidence estimate to be ≤30% [38]. For all calculations, the RSE of the MDRI and false-recent rate was set to 10% and 20%, respectively. This analysis was conducted using simulated populations with varying HCV seroprevalence, HCV infection incidence, and HIV prevalence.

The power to detect a 50% reduction in HCV infection incidence between 2 serial cross-sectional surveys was examined for each testing method (1-tailed α = 0.05). This simulated analysis was conducted for populations with differential survey sample size, HCV seroprevalence, HIV prevalence, and HCV infection incidence in the baseline survey. Constant HIV prevalence was assumed between surveys.

The simulated analyses were conducted using the Assay-Based Incidence Estimation tool kit [24, 39]. Other statistical analyses were performed using R Statistical Software and Stata, version 14.

RESULTS

Study Specimens

Fifty-six HCV seroconverters from the BBAASH cohort contributed 233 samples from follow-up visits ≤2 years after HCV seroconversion (median days after HCV seroconversion, 241; interquartile range [IQR], 124–378) and 72 samples from follow-up visits >2 years after HCV seroconversion (median days after HCV seroconversion, 1152; IQR, 950–1672). The ALIVE cohort contributed 692 follow-up visit samples collected from 512 PWID who were known to be HCV seropositive for >2 years. The distribution of sex, race, age, and HIV status differed between samples collected ≤2 and >2 years after HCV seroconversion (Table 1). For the subjects with known genotype data, the majority were infected with genotype 1 (86.7% of BBAASH subjects [39 of 46] and 96.0% of ALIVE subjects [194 of 202]).

Table 1.

Characteristics of Samples Collected From Individuals ≤2 and >2 Years After Hepatitis C Virus (HCV) Seroconversion

Characteristic	≤2 Years After Seroconversion	>2 Years After Seroconversion
Sex
Male	54.9 (128/233)	70.3 (537/764)
Female	45.1 (105/233)	29.7 (227/764)
Race
Nonblack	93.6 (218/233)	19.5 (149/764)
Black	6.4 (15/233)	80.5 (615/764)
Age, y, median (range)a	24 (18–39)	45 (23–65)
HCV viremia level, IU/mL
< 500	30.5 (71/233)	21.6 (165/764)
≥ 500	69.5 (162/233)	78.4 (599/764)
HIV serostatus
Negative	97.4 (227/233)	75.8 (579/764)
Positive	2.6 (6/233)	24.2 (185/764)
Overall, no.
Samples	233	764
Individuals	56	547

Data are % of samples (no. of samples with the characteristic/no. of samples analyzed).

^a Age was not available for 2.5% of samples collected ≤2 years after HCV seroconversion and for 1.4% of samples collected >2 years after HCV seroconversion.

Development of Avidity-Based Algorithms

HCV avidity increased with time after HCV seroconversion (Figure 1). As time approached 2 years, the majority of samples had an avidity index >90% (Figure 1), thereby verifying 2 years after HCV seroconversion as an appropriate value for T. For avidity index cutoffs between 20% and 80%, the assay's MDRI ranged from 255 to 378 days (Supplementary Table 1). The avidity index of samples collected >2 years after HCV seroconversion was not affected by demographic characteristics such as sex, race, or age (Table 2). However, samples collected >2 years after HCV seroconversion were significantly more likely to have a lower avidity index (ie, <80%) if the individual did not have HCV viremia, compared with those with viremic infections (adjPRR, 2.94; 95% CI, 2.04–4.24; P < .001). In addition, samples from HIV-infected individuals with a CD4⁺ T-cell count of <300 cells/µL were more likely to have a lower avidity index, compared with HIV-negative individuals (adjPRR, 4.47; 95% CI, 3.07–6.49; P < .001; Supplementary Table 2). Inter-operator reproducibility of samples enriched for recent infection demonstrated significant agreement (93.9%) at an avidity index cutoff of 30% (Cohen κ = 0.84; P < .001; n = 82; data not shown).

Figure 1.

Ortho avidity index values as a function of time since hepatitis C virus (HCV) seroconversion (n = 305). Samples were collected from 56 seroconverters in the Baltimore Before and After Acute Study of Hepatitis (BBAASH) cohort. Box plots of the avidity index values depict the median (thick horizontal bar), and interquartile range (light gray box) in given intervals after HCV seroconversion. Open circles indicate samples collected from individuals without viremia (HCV RNA load, < 500 IU/mL), and closed circles indicate samples collected from individuals with viremia (HCV RNA load, ≥ 500 IU/mL).

Table 2.

Univariate Analysis of Factors Associated With an Avidity Index <80% Among 764 Samples Collected >2 Years After Hepatitis C Virus (HCV) Seroconversion

Characteristic	Ortho Avidity Index <80%
	Samples, Proportiona	PRR (95% CI)b	P Value
Sex
Male	80/537	Reference
Female	41/227	1.22 (.84–1.78)	.295
Race
Nonblack	27/149	Reference
Black	94/615	0.87 (.56–1.36)	.548
Age, y
18–34	18/117	Reference
35–44	41/251	1.00 (.62–1.61)	.995
45–54	43/293	0.98 (.60–1.60)	.923
≥55	13/92	0.93 (.51–1.70)	.820
HCV viremia level, IU/mL
≥ 500	69/599	Reference
< 500	52/165	2.67 (1.91–3.74)	<.001
HIV status
Negative	62/579	Reference
Positive	59/185	2.94 (2.09–4.14)	<.001
CD4+ T-cell count, cells/µL
≥300	16/85	Reference
<300	28/97	1.48 (1.01–2.15)	.042
Total	121/764	…

Abbreviation: HIV, human immunodeficiency virus.

^a Data are no. of samples with an Ortho avidity Index <80%/no. of samples analyzed among each characteristic.

^b Prevalence risk ratios (PRRs) and 95% confidence intervals (CIs) were calculated from univariate modified Poisson regression models, using generalized estimating equations, an exchangeable correlation structure, and a robust variance estimator to indicate associations with an Ortho avidity index of <80% among samples collected >2 years after HCV seroconversion. These data are indicative of misclassification of samples collected >2 years after HCV seroconversion as being from individuals with recent infection.

The MDRI and HIV-specific false-recent rates of the Ortho avidity assay at various index cutoffs in combination with viremic detection are presented in Supplementary Table 3. It was assumed that the MDRI and false-recent rate of method 1 (acute screening) was 45 days and 0%, respectively (Figure 2) [16]. The Ortho avidity assay at a cutoff of 30% in combination with HCV viremic detection had an MDRI of 147 days (95% CI, 125–195 days; method 2; Figure 2). Combining methods 1 and 2 assumed an MDRI of 192 days for method 3 (Figure 2). The false-recent rate of both avidity-based methods was 0.69% (95% CI, .19%–1.76%; n = 579) and 7.57% (95% CI, 4.20%–12.37%; n = 185) among HIV-negative and HIV-positive samples, respectively (Figure 2). Supplementary Table 4 presents the false-recent rates of methods 2 and 3 among subgroup populations, and their false-recent rates are shown by HIV prevalence in Supplementary Table 5.

Figure 2.

Three cross-sectional approaches to estimate hepatitis C incidence. Method 1 refers to individual acute-phase screening among hepatitis C virus (HCV)–seronegative individuals [18], and method 2 refers to Ortho avidity testing among HCV-seropositive individuals. Method 3 is a combination of method 1 and method 2. Abbreviation: IgG, immunoglobulin G.

Estimation of Incidence in Low-Risk Populations

In a simulated population where HCV prevalence was 2.5%, HCV infection incidence was low (<0.5%) and HIV prevalence was 0.0%, all methods required a sample size of >5000 individuals to estimate incidence with an RSE of 30% (Figure 3A). For an HCV epidemic with an incidence of 0.6%–1.0% nested within a population where HCV prevalence was 2.5%, method 2 required 3446–5907 individuals to estimate incidence (Figure 3A). Adding an HIV prevalence of 10% among HCV infections increased the necessary sample size for method 2 to 3739 and 6588 individuals for an incidence between 0.6% and 1.0% (Figure 3B). Method 3 required smaller sample sizes than Method 2, but this difference diminished as incidence increased (Figure 3). Methods 2 and 3 each required considerably lower sample sizes than method 1 (Figure 3).

Figure 3.

Minimum sample sizes required of a cross-sectional survey to estimate incidence in a general population with 2.5% hepatitis C virus (HCV) seroprevalence (relative standard error, 30%). The sample sizes represent the total number of HCV-seropositive and HCV-seronegative individuals required in a single cross-sectional to achieve an incidence estimate with a relative standard error (RSE) of 30%. The HCV seroprevalence was 2.5%, and the human immunodeficiency virus (HIV) prevalence among HCV-seropositive individuals was either 0% (A) or 10% (B). The RSE for the mean duration of recent infection and false-recent rate was 10% and 20%, respectively. Data are shown for 3 testing algorithms: method 1, HCV viremic detection among HCV-seronegative individuals only; method 2, Ortho avidity (index <30%) with viremic detection among HCV-seropositive individuals; and method 3, a combination of methods 1 and 2.

Estimation of Incidence in High-Risk Populations

In simulated populations with 25% HCV seroprevalence, HCV infection incidence between 10% and 40%, and HIV prevalence of 0.0%, method 2 only required a sample size of 125–469 individuals to estimate incidence with an RSE of 30% (Figure 4A). Increasing the HIV prevalence to 20% among the HCV infections increased the sample size to 137–564 individuals (Figure 4B). For surveys with 50% HCV seroprevalence, HCV infection incidence between 10% and 40%, and HIV prevalence of 0.0%, method 2 only required sample sizes of 192–788 individuals (RSE, 30%; Figure 4C), and adding an HIV prevalence of 20% to these parameters increased the requisite sample size for method 2, but the required sample size was still <1500 individuals (RSE, 30%; Figure 4D). In an HIV-negative population with an HCV seroprevalence of 75%, Methods 2 and 3 performed better than method 1, but this was not the case when the HIV prevalence was 20% (RSE, 30%; Supplementary Figure 1).

Figure 4.

Minimum sample sizes required of a cross-sectional survey to estimate incidence in simulated high-risk populations with 25% or 50% hepatitis C virus (HCV) seroprevalence (relative standard error [RSE], 30%). The sample sizes represent the total number of HCV-seropositive and HCV-seronegative individuals required in a single cross-sectional to achieve an incidence estimate with a RSE of 30%. HCV seroprevalence was 25% (A and B) or 50% (C and D), and HIV prevalence among HCV-seropositive individuals was either 0% (A and C) or 20% (B and D). The RSE for the mean duration of recent infection and false-recent rate was 10% and 20%, respectively. Data are shown for 3 testing algorithms: method 1, HCV viremic detection among HCV-seronegative individuals only; method 2, Ortho avidity (index <30%) with viremic detection among HCV-seropositive individuals; and method 3, a combination of methods 1 and 2.

Methods 2 and 3 were able to achieve greater precision surrounding the incidence estimate (RSE, 15%; Supplementary Figure 2). An RSE of 15% was attainable with sample sizes of <5000 individuals when the HCV seroprevalence was 25%– 50%, the HCV infection incidence was >15%, and the HIV prevalence was 0.0% (Supplementary Figure 2). An HIV prevalence of 20% increased the necessary sample size, but sample sizes remained <5000 individuals if the HCV infection incidence was >20% (Supplementary Figure 2). Overall, method 3 required smaller sample sizes than method 2, but as the HCV infection incidence increased, their performance was similar (Figure 4 and Supplementary Figure 2).

Statistical Power to Detect a Reduction in HCV Infection Incidence

The data demonstrate that avidity-based testing algorithms have the potential to have >80% statistical power to detect a 50% reduction in HCV infection incidence between 2 serial cross-sectional surveys among high-risk populations (Figure 5). Methods 2 and 3 had similar statistical power when the baseline incidence was ≥20%. Method 1 only had >80% power to detect a reduction in HCV infection incidence when the baseline HCV infection incidence was 40%, the HCV prevalence was 50%, and the survey sample size was 2000 (Figure 5).

Figure 5.

Capacity to detect a 50% reduction in hepatitis C virus (HCV) infection incidence between 2 serial cross-sectional surveys. The statistical power to detect a 50% reduction in incidence was compared for 3 testing algorithms: method 1, HCV viremic detection among HCV-seronegative individuals only; method 2, Ortho avidity (index <30%) with viremic detection among HCV-seropositive individuals; and method 3, a combination of methods 1 and 2. Statistical power is shown for simulated populations with differential HCV prevalence, human immunodeficiency virus (HIV) prevalence among HCV-seropositive individuals, and HCV infection incidence in the baseline survey. HIV prevalence was assumed to be constant between surveys. N refers to the total sample size of the study population of each independent survey, including both HCV-seronegative and HCV-seropositive individuals.

DISCUSSION

The global epidemiological profile of HCV infection is evolving [40], and accurate cross-sectional surveillance methods are needed to estimate population-level incidence [20]. Using a prospective cohort of frequently sampled PWID at high risk of newly acquired HCV infection [17], we demonstrated that HCV IgG antibody avidity is a biomarker of early HCV infection that can be used to precisely estimate incidence if HCV viremic status is considered. The Ortho avidity assay (index cutoff, <30%) in combination with viremic detection can be highly accurate in settings with a low HCV infection incidence, with a sufficiently large sample size. However, this algorithm had optimal performance in settings of high HCV infection incidence and a moderate HCV prevalence (25%–50%)—a burden that is reflective of HCV epidemics among PWID populations [7, 41]. In various simulated surveys of high-risk populations, avidity-based testing algorithms had the capacity to detect a 50% reduction in HCV infection incidence via serial cross-sectional surveys.

There are clear advantages to a biomarker-based approach to estimate incidence as compared to prospective, longitudinal follow-up—it is more rapid, practical, and cost-efficient [36, 42]. Furthermore, it is more feasible to perform a cross-sectional incidence study among hard-to-reach populations that are difficult to follow. However, there are several nuances to this method for HCV that require attention. It is important to note that these caveats are expected from what we have previously learned from developing biomarker-based methods for HIV infection incidence estimation. First, HCV avidity cannot be used without HCV viremic status to estimate incidence, as this single biomarker would misclassify individuals with spontaneous or treatment-induced clearance of viremia as having recent infection [43]. Second, the sample size necessary to estimate the incidence of primary HCV infection is dependent on the desired level of precision and is context specific in terms of HCV prevalence and incidence. In addition, the performance of an HCV avidity-based testing algorithm will be dependent on the prevalence of HIV among HCV-infected individuals. Loss of high avidity associated with low CD4⁺ T-cell counts in this study population mirrors a previous study reporting decreased HCV antibody titer following HIV acquisition and progression to AIDS [32]. Given these factors, the false-recent rate of the selected avidity-based testing algorithm will need to be determined for a given study population prior to their application for incidence estimation. This recommendation is consistent for what is currently required of biomarker-based HIV infection incidence estimation, and guidelines on how to characterize the false-recent rate in a proposed study population have been previously described [34].

This study had several limitations. Samples collected from high-risk participants on a monthly basis in the BBAASH cohort enabled us to precisely estimate the date of seroconversion [17], but better precision of the MDRI estimate will be attainable through a larger sample size of such well-characterized HCV seroconverters. In addition, HCV antibody dynamics in early infection may differ among HIV-infected individuals [44]. However, this study was not powered to determine the effect of HIV infection on the MDRI estimate. The effect of genotype on the MDRI and false-recent rate estimates requires further investigation, as HIV subtype variation has previously been shown to be a challenge for HIV tests for recent infection [34]. Furthermore, the results of this study only reflect the ability to estimate the incidence of primary HCV infection, and the effect of reinfection and superinfection was not assessed. Additionally, testing a cross-sectional sample set taken from a prospective longitudinal cohort with known observed incidence is needed to further validate avidity-based algorithms for HCV incidence estimation.

Application of an avidity-based testing algorithm to estimate incidence from a cross-sectional survey can enhance current sentinel surveillance systems. Among other emerging epidemics, the CDC has reported that HCV infection incidence is markedly on the rise among PWID in the United States [5, 45]. However, in light of new information regarding the underreporting of newly acquired HCV cases to the CDC [20], these epidemics may be even more advanced than previously expected. Individual acute HCV screening practices in combination with an avidity-based algorithm (method 3) may collectively help to estimate incidence through practical sample sizes. A surveillance strategy using an avidity-based algorithm without HCV acute screening (method 2) may be a cost-effective alternative with comparable precision. Altogether, these avidity-based testing algorithms have the capacity to quickly identify and confirm emerging epidemics in many PWID populations.

With the advent of effective and tolerable direct-acting antivirals, clinical treatment is now reaching PWID [9]. On a population-level, targeting PWID, who account for 80% of new HCV infections, will be essential to the now feasible idea of disease elimination [6, 46]. Martin et al, among others, have mathematically modeled that treatment scale up can halve HCV prevalence in 15 years [46]. We now provide an empirical approach to assess these claims by measuring incidence before, during, and after treatment rollout is mobilized. It has been noted that treatment as prevention will be most effective in PWID populations with low-to-moderate baseline HCV prevalence (eg, 25%) [46, 47]. In parallel to this level of HCV burden, the avidity-based testing algorithms presented in this work had >80% power to detect a 50% reduction in HCV infection incidence in many settings (25%–50% HCV prevalence).

This study offers the field a sensitive tool to track the HCV epidemic. With sufficient power to assess the efficacy of community-level interventions, cross-sectional HCV infection incidence estimation can serve as a vital resource for HCV prevention research. Collaborative efforts to further optimize avidity-based or alternative antibody maturation testing algorithms to estimate HCV infection incidence should be prioritized.

Supplementary Data

Supplementary materials are available at http://jid.oxfordjournals.org. Consisting of data provided by the author to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the author, so questions or comments should be addressed to the author.