Hepatitis virus (HCV) diagnosis and access to treatment in a UK cohort

Setting and cohort

Our microbiology laboratory in the South East of the UK is located in a large tertiary referral teaching hospital (http://www.ouh.nhs.uk/) that provides one million patient contacts a year, and handles samples referred from the community as well as four in-patient sites.

We retrospectively interrogated electronic microbiology records for all HCV assays performed within two defined time-intervals, during which different diagnostic algorithms were operating, in each case starting with a screening test and then proceeding to confirm active infection. These are summarized in Fig. 1 and outlined as follows:

1. i.

   Group 1 (18 months; January 2013–June 2014). Samples were screened for HCV-Ab using an ADVIA Centaur automated immunoassay (Bayer). HCV-Ab-positive samples (excluding repeat samples from patients with a pre-existing HCV diagnosis) were sent for confirmatory testing by the regional reference laboratory (Public Health England, Colindale), using two further ELISA tests (Ortho and BioRad). Antibody positive samples (based on sample:cut-off ratio > 1) were tested for HCV RNA.

    
2. ii.

   Group 2 (15 months; January 2015–March 2016). HCV testing was undertaken using a combination of HCV-Ab and HCV-Ag, using Abbott Architect i2000SR, with Diasorin Liason XL for confirmation of HCV-Ab. The HCV-Ag assay provides a quantitative result up to an upper limit of 20,000 fmol/L. We classified values of ≥3 fmol/L as positive (as previously reported [19]). When we introduced the HCV-Ag assay, we tested the same sample for HCV RNA in parallel, for the puporses of validation. Following this validation period, we stopped routinely testing the same sample for HCV RNA, and changed the protocol to request a second sample for RNA testing (only in those that are HCV-Ag positive). This allows us to confirm the diagnosis using two separate samples, in keeping with good laboratory practice, and also provides material for genotyping.

    

For Groups 1 and 2, samples were tested for HCV RNA using the Abbott HCV M2000 assay.

Some individuals were tested for HCV infection on > 1 occasion; we removed duplicate tests from our overall positive cohort using unique identifiers (hospital number or NHS number). We recorded patient age, sex, and the location from which the sample was sent. Treatment data were captured and recorded from an electronic database within the Hepatology Department. Response to treatment was defined as sustained virologic response (undetectable HCV RNA using PCR) at ≥12 weeks following the end of therapy (SVR₁₂).

Ethnic origin is not routinely captured data in hospital electronic systems. Prior to anonymisation, we therefore used an analytical tool to estimate ethnicity, applying Onolytics software for all patients for whom a full name was part of the electronic record (https://onolytics.com [20–22]). This software was developed in 2006–7 funded by Economic and Social Research Council (ESRC) Knowledge Transfer Partnerships, and sets out to determine probable ethnic origin based on name.

Data analysis

Statistical analysis was performed using GraphPad Prism v.7.0b and Googlesheets (https://docs.google.com/spreadsheets). We compared binary values using Fisher’s Exact Test, Mann-Whitney U test for continuous non-parametric data, chi-square for analysis of categorical data, and linear regression for correlation between continuous variables.

Ethics approval

Ethics approval was not required, as this study was undertaken as a departmental quality improvement exercise within microbiology using anonymised patient data, and completed the audit cycle for previously approved audit projects [23, 24]. Data for Onolytics analysis were handled separately, subject to a confidential disclosure agreement drawn up by University of Oxford Research Services (February 2016).

HCV testing: Frequency and characteristics of infection

In total, we present data for 38,509 HCV tests done during the two intervals reviewed. On average we performed an average of 1068 tests / month during the earlier phase of the study (Group 1) and 1286 tests / month in the later time period (Group 2); Fig. 1.

We identified 353 active HCV infections across Group 1 and Group 2, using a combination of HCV-Ag and/or HCV RNA testing. We estimated the frequency of active HCV infection within this cohort at 0.9% based on a combined numerator (n = 353), and using the total number of samples tested as the denominator (n = 38,509); Fig. 1. Comparing the two time periods, there was no change in the frequency of positive HCV-Ab testing: HCV-Ab frequency 317 / 19,226 (1.6% in Group 1) vs. 325 / 19,283 (1.7% in Group 2); p = 0.8, chi-sqaure test. There was also no significant difference in the frequency of confirmed active HCV infection (based on HCV-Ag and/or HCV RNA PCR): 191 / 19,226 (1.0% in Group 1) vs. 162 /19,283 (0.8% in Group 2); p = 0.13, chi-square test.

Characteristics of the 353 individuals with active HCV infection are summarised in Table 2, and the complete metadata are available on-line (https://doi.org/10.6084/m9.figshare.5355097). The median age was 37 years (IQR 31–48). Men accounted for 55% of all individuals tested, but 84% of all new diagnoses (Table 2) [23]. Over one-third (36%) of new diagnoses were made in prison. Genotype was available in 186 cases (53% of new diagnoses), with genotype 1 (n = 84) and genotype 3 (n = 80) accounting for the majority (45% and 43%, respectively); Fig. 2.

	Group 1 (2014)	Group 2 (2016)	Group 1 + Group 2 (2014–2016)
Total number HCV-Ab positive	317	325	642
Total number confirmed positive for active HCV infection (HCV-Ag/HCV RNA positive)	191	162	353
Number male (% of active infections)a,b	165 (86.3%)	132 (81.5%)	297 (84.1%)
Age in years (median and IQR)a	39 (31–49)	36 (30–46)	37 (31–48)
Locationa, c
- Primary Care	64 (33.5%)	46 (28.4%)	110 (31.2%)
- Hospital out-patient	17 (8.9%)	17 (5.2%)	34 (9.6%)
- Prison	61 (31.9%)	66 (40.7%)	127 (35.9%)
- Hospital in-patientd	22 (11.5%)	9 (5.5%)	31 (8.8%)
- Emergency Dept.	5 (2.6%)	6 (3.7%)	11 (3.1%)
- Sexual health clinic	16 (8.4%)	13 (8.0%)	29 (8.2%)
- Occupational health	2 (1.0%)	1 (0.6%)	3 (0.8%)
- Other locationse	4 (2.1%)	4 (2.5%)	8 (2.3%)
Ethnic origina
- Black	5 (2.6%)	1 (0.6%)	6 (1.7%)
- Asian	17 (8.9%)	7 (4.3%)	24 (6.8%)
- European	149 (78.0%)	100 (61.7%)	249 (70.5%)
- Unknown	20 (10.5%)	54 (33.3%)	74 (21.0%)

^aBreakdown of characteristics is shown for the individuals who have confirmed active HCV infection, based on HCV-Ag and/or HCV RNA PCR

^bNumbers of positive tests are shown with percentage of positive tests in brackets

^cWe did not have access to data for PWID, but prison location may be an important surrogate marker for this group

^dHospital in-patient locations include general surgery, orthopaedics, transplant, renal, haematology, intensive care, unspecified ward locations in the hospital, and in-patients in community hospitals

^eOther locations include paediatrics, psychiatry, fertility and pathology

Outcomes and performance of HCV-ab and HCV-ag assays

In the earlier testing period (Group 1), 277/317 HCV-Ab positive samples were positive on confirmatory testing for HCV-Ab at the reference laboratory (Fig. 1), giving our in-house test a positive predictive value (PPV) of 87.4% compared to a regionally accepted standard. We used these results to investigate whether any host factors were associated with false positive antibody tests, and found that individuals identified as African have a higher chance of a false-positive HCV Ab test (Fig. 3). We confirmed this result by multivariate logistic regression analysis, in which African ethnicity was significantly associated with a false positive Ab test result (p = 0.0004), but age > 60 years and sex were not. Prison location was associated with a true positive Ab-test result (p = 0.01). We did not have access to data on representation of PWID in this cohort, but recognise that prison location is a likely surrogate marker for this risk factor for infection.

Individuals with a positive HCV-Ag test had a median HCV viral load of 5.9 × 10⁵ IU/ml (Fig. 4a), and there was a significant positive correlation between quantitative antigenaemia and viral load (r² = 0.3, p < 0.0001; Fig. 4b). However, in a small proportion of cases, the HCV-Ag test was falsely negative (Table 4).

Age group (years)	Sex	Patient location	Ethnicity	HIV status	HCV Ag (fmol/L)	HCV Ab (sample/cut-off ratio)	Genotype	HCV viral load (IU/ml)
30–39	F	Sexual health	Unknown	negative	0.0	11.8	N/A	25
40–49	M	Primary care	European	negative	0.0	12.2	N/A	226
50–59	F	Hospital in-patient (General Medicine)	European	positive	0.0	3.1	N/A	302
20–29	M	Prison	European	negative	0.62	14.2	2b	2916
20–29	M	Prison	European	N/A	0.00	12.6	N/A	8232
30–39	F	Hospital out-patient	European	N/A	1.98	15.8	1b	13,860
30–39	M	Prison	European	negative	0.00	12.2	3a	174,834

N/A = not available. Total number of HCV-Ag tests carried out in this period n = 305. None of the patients with a false negative result underwent a repeat Ag test so laboratory error cannot be ruled out in this instance. Threshold for positive HCV-Ag defined as ≥3 fmol/L. Samples are ranked in ascending viral load order

Based on the strong correlation between HCV-Ag and viral load (Fig. 4b), and on a previous analysis that documents HCV-Ag down to a level equivalent to viral load 3000 IU/ml [13], we sought to determine whether the false negative HCV-Ag tests (n = 7) were associated with low viraemia. Indeed, in 5/7 cases, HCV RNA was < 10⁴ IU/ml, one of which had HCV RNA as low as 25 IU/ml. Among 128 samples for which we had both an HCV-Ag and HCV RNA assay result, the false-negative tests were significantly enriched in the group with HCV RNA < 10⁴ IU/ml (p < 0.0001, Fig. 4c).

Clinical management and outcomes

Of 353 patients with a new HCV diagnosis, 142 (40%) attended a hepatology clinic appointment, 79 were treated (22% of the cohort) and 66 met the SVR₁₂ criterion (19% of the whole cohort; 84% of all those treated). Among those treated with a DAA-based regimen (n = 59), a treatment endpoint was documented in 54, of which 50 were classified as SVR₁₂ (93%). For individual treatment regimens, see full clinical metadata available on-line (https://doi.org/10.6084/m9.figshare.5355097).

This study was not designed to examine or report on the outcomes of treatment. However, we examined existing treatment data to look for evidence of different outcomes between genotypes 1 and 3. Among treated genotype 1 infections with outcome data (n = 34), we recorded 32 cases of SVR₁₂, and two cases of relapse. For treated genotype 3 (n = 30), there were 25 cases of SVR₁₂ and five relapses, but this difference did not reach statistical significance (relapse rate 2/34 (6%) for geno-1 vs. 5/30 (17%) for geno-3; p = 0.2, Fisher’s Exact Test).

Summary comments

Careful scrutiny of HCV screening, diagnosis and treatment is important so that enhanced efforts can be made to identify individuals with active infection in order to provide access to DAA therapy, and to move towards international elimination goals [3, 9, 13].

In our setting in the UK, 0.9% of all samples submitted for HCV testing had evidence of active HCV infection. Following the implementation of HCV-Ag testing as part of the diagnostic algorithm, the PPV of a positive test increased to 100%, slightly exceeding that reported by other recently published studies [14]. The close correlation between HCV-Ag and HCV RNA viral load suggests that, in the absence of having access to a quantitative PCR result, HCV-Ag may be a useful surrogate marker of viraemia, particularly at higher viral loads (e.g. if HCV RNA > 10⁴ IU/ml). Genotypes 1 and 3 predominated (in keeping with other published studies in this setting [25, 26]).

Our data confirm the success of DAA treatment, with SVR₁₂ documented in 93%. However, this cohort also highlights the substantial loss of patients at each step of the clinical ‘cascade’ (Fig. 5; Table 5). This is due to a combination of factors that include poor linkage between services (community, prisons, sexual health and secondary care), itinerant populations, specific challenges in providing follow-up for those in prison and PWID, and deaths.

Treatment Status	Patient Classification	Number	Percentage of treatment subgroupa	Percentage of total cohort
Not yet treated (n = 130)	Offered appointment but did not attend	46	35.4	13.0
	Seen in another clinic (sexual health, prison, paediatrics)	32	24.6	9.1
	Seen by hepatology but not on treatment waiting list	23	17.7	6.5
	Seen by hepatology and on treatment waiting list for DAA	3	2.3	0.9
	Died	11	8.5	3.1
	Transferred out of area	10	7.7	2.8
	Seen in clinic but lost to follow-up	2	1.5	0.6
	Spontaneous clearer	3	2.3	0.8
Treated with DAA (n = 59)	SVR12	50	84.7	14.2
	Relapsed	4	6.8	1.1
	Outcome data pending	5	8.5	1.4
Treated with IFN/RBV (n = 20)	SVR12	16	80.0	4.5
	Relapsed	3	15.0	0.8
	No outcome data	1	5.0	0.3
Unknown (n = 150)	Not known to local services	144	100	40.8
TOTAL		353	400	100

^aTreatment subgroup is defined as the four categories listed in the first column of this table. The total of this column is 300%, as the total for each of three subgroups is 100%

DAA = direct acting antivirals, IFN = interferon, RBV = ribavirin; SVR₁₂ = sustained viraemic response at ≥12 weeks following therapy

Relevance to laboratory and clinical practice

Although HCV-Ag testing can potentially replace a nucleic acid test for HCV diagnosis or monitoring in some settings [14, 27, 28], guidelines from the UK [29], North America [30] and the WHO [12] still advocate use of PCR as a definitive test following HCV-Ab (± HCV-Ag) screening. RNA PCR also remains the gold-standard approach to monitoring progress during and after treatment and is currently still recommended for genotyping, to underpin optimum choice of DAA regimen [31]. However, as we move towards pan-genotype treatments for HCV, a single diagnostic test, with the potential to be applied at the point-of-care, is an appealing strategy [13].

The small proportion of all diagnosed patients who access clinical care and receive successful treatment is in keeping with that reported in other centres [5, 32], reflecting many challenges for HCV elimination. Our data highlight the particular vulnerability of the prison population, in keeping with a worldwide estimate of 15% HCV prevalence in prisoners [33]. Offering treatment within the prison system has now become a realistic possibility, on the basis of oral DAA therapy, shortened treatment regimens, and a low rate of side-effects [34, 35].

In the longer term, bigger datasets are required to improve our insights into the HCV-infected population [32], and health economics analyses should be used to support optimum deployment of resources in different settings [36, 37]. On the global stage, optimization of laboratory testing and reduction of costs are essential to improve access to accurate diagnosis, advocacy for better testing and treatment for populations in resource-limited settings, and especially for targeting interventions at high-risk populations including MSM, PWID and prisons.

Performance of HCV-ag test

Our calculated PPV of 100% for the HCV-Ag test exceeds a previous estimation of the PPV of the Ag test (94.7%) calculated from assimilation of data from other comparable reports [14]. However, PPV is dependent on the overall prevalence of infection, and will therefore vary between settings.

A recent meta-analysis concludes that the HCV-Ag test performs similarly to HCV RNA PCR when viral load is > 3000 iu/ml [13]. Thus this diagnostic tool provides a high sensitivity without unduly influencing specificity. The HCV-Ag assay is likely to perform less reliably at low viral loads, and a low threshold should be applied for using NAT to confirm a diagnosis in the settting of an equivocal HCV-Ag result (3–10 fmol/L). Another potential explanation for false negative HCV-Ag test is mutations in the core region of the HCV genome which could account for a failure of antigen detection [38], or potentially cause lack of PCR amplification if mutations occur in primer binding regions.

Caveats and limitations

Our analysis must be set in the overall context of the low prevalence of HCV in our setting, and the retrospective approach to data collection. We cannot comment on population level epidemiology, as a large pool of individuals who are HCV-infected never have a screening or diagnostic test [9]. Due to gaps in the laboratory data-set, in which not every sample was tested for HCV RNA, we adopted a pragmatic definition of active HCV infection based on the best laboratory evidence available (in some cases this was an assay for HCV-Ag without an accompanying NAT). Our analysis of sensitivity and specificity of HCV-Ag testing could only be assessed in the subset of samples that had undergone testing with assays for both HCV-Ag and HCV RNA quantification.

The rate of false negative HCV tests is our population is likely to be low, but quantifying this was not possible within this study, as we relied on identifying samples that initially tested positive. In order to ascertain the PPV of the HCV-Ab test in-house, we referred to a Reference Laboratory test as ‘gold standard’. However, this repeat testing in a Reference Laboratory setting is itself subject to an error rate, and therefore may lead to a misrepresentation of our overall assay performance.

We found evidence that the HCV-Ab test performs poorly in individuals predicted to be of African origin. A similar high rate of false positive tests has previously been reported from Polynesia [39]. This illustrates how tests that have been validated in white European/Caucasian populations cannot necessarily be robustly applied in other settings. Although the tools used here have been validated [21, 22], use of name is an imperfect way to derive ethnic origin and is potentially confounded by a variety of factors, the most obvious of which is individuals who change their name through marriage.

In our setting, the sexual health clinic anonymises patient data, preventing robust linkage between services. We are therefore unable to trace outcomes for patients who were diagnosed via this route (8% of the total; Table 2). Likewise, consistent identification and tracing of individuals who use drugs and/or are in prison is challenging, and we cannot exclude the possibility of duplication of some of these individuals within our dataset.

In the context of this study we do not have prospective socio-demographic data that are required to investigate the reasons for the male excess we describe (e.g. MSM, PWID). Sex differences could also be accounted for by genuine biological discrepancies in susceptibility to, and outcomes of, infection between males and females [40, 41]. We have also not been able to characterise the prison population in detail; this group is difficult to treat and to follow-up, and so is at risk of worse outcomes, but we may have underestimated SVR₁₂ due to missing data.