Cost‐effectiveness of strategies to improve HCV screening, linkage‐to‐care and treatment in remand prison settings in England

A simplified cascade‐of‐care may improve screening and treatment uptake among incarcerated individuals. We assessed the cost‐effectiveness of traditional and simplified screening and treatment in a London remand prison.


| INTRODUC TI ON
The World Health Organisation (WHO) is committed to the elimination of viral hepatitis as a global health issue by 2030 as outlined in its 2016-2021 global health sector strategy. 1 It is estimated that 8% of the global burden of chronic hepatitis c virus (HCV) infection is concentrated among people who inject drugs (PWID). 2,3 A strong association exists between PWID and incarceration, with lifetime rates reported as high as 90% 4 and, furthermore, up to 15% of people living in detention globally are estimated to be infected with HCV. 5 Furthermore, recent modelling studies have estimated that PWID are accountable for 79% of HCV transmission and also underlined the value of sustained treatment as prevention with directacting antiviral (DAA) therapy in combination with scaling-up harm reduction (HR). 3,6 Thus, it is clear that addressing the HCV epidemic among PWID is a key priority, which will influence the success of global attempts to eliminate HCV. In practice, however, implementing a successful treatment initiative in this marginalised community can prove difficult. For example, a recent prospective evaluation reported less than 10% treatment uptake among newly diagnosed PWID in San Francisco. 7 Therefore, engaging infected individuals serving a custodial sentence has been identified as a potentially opportune circumstance.
Recognising the importance of HCV infection among prison populations, countries including: Australia, China, India, Iran and the USA, have started to include prison programmes for HCV screening and treatment as part of their national hepatitis plans. 8 In England, 'opt-out' dry blood spot (DBS) screening for HCV infection has been phased into the prison estates since 2014.
Mathematical models have estimated that scaling-up prison-based HCV screening, treatment and harm reduction in Scotland will reduce the incidence and prevalence of HCV among PWID by almost half. 9 The use of DBS-based prison HCV screening has been found to be cost-effective only when linkage to care is greater than 40% and is considered highly cost-effective when combined with incrementing levels of treatment using DAA therapy. 10,11 Furthermore, the scale-up of HCV interventions in prison have also been found to be cost-effective in Switzerland and the USA. 12,13 However, in England, there is clearly room for improvement, as despite having a national opt-out DBS screening policy, existing rates of screening are around 20%, while only a minority of individuals are accessing treatment. 14 We have recently demonstrated that a simplified screening and treatment algorithm, implementing point-of-care screening and streamlined treatment initiation of all viraemic individuals, resulted in screening uptake of 90% and treatment uptake of 85% in prisons. 15 However, no study to date has assessed the cost-effec-

LAY SUMMARY
Despite rolling-out a national screening programme for hepatitis C among people in prison in England, rates of screening and treatment remain suboptimal. In remand settings, point-of-care testing and fast-track treatment initiation can improve treatment uptake. In comparison to national average rates, optimising screening, linkage-to-care and treatment is highly cost-effective as both in universal and targeted testing among people who inject drugs in our remand prison settings. These data support the adoption of policies to simplify the testing and treatment practices in prison settings in England.  [15][16][17] Given the high-risk status of the population, it is assumed that the F I G U R E 1 Summary of universal opt-out general prison population DBS and targeted high-risk simplified pathways implemented at HMP Wormwood Scrubs. 15 * Patient given contact details to self-refer for treatment in community; ** direct acting antiviral treatment F I G U R E 2 Natural history model and transition probabilities for fibrosis progression, treatment outcomes and associated morbidity and mortality transmission of disease has occurred as a result of injecting drug use. All successfully treated non-cirrhotic patients experienced a halt in their liver disease progression. Meanwhile all untreated and cirrhotic individuals (irrespective of treatment status) continue to experience disease progression through the model lifespan. In addition to disease-specific morbidity and mortality, a standardised background prisoner mortality rate has been applied to our cohort. 18 The model also assumes a matched background chronic HCV prevalence and disease progression for those who are not retained along the cascade of care.

| Study setting
Where possible, empirical data from both opt-out universal HCV DBS general prison population screening and a pilot simplified screen-and-treat algorithm at HMP Wormwood Scrubs have been used to generate epidemiolocal, screening and costs parameters. 15 A review of the relevant literature was used to derive all other necessary parameters.

| Interventions
In the following scenarios, we compare varied rates of screening, linkage-to-care and treatment uptake. The status quo scenario (strategy a) used is based on the national average rates of 20% screening and 40% linkage-to-care reported from the health and justice indicators of performance (HJIPs) for 2017-2018. 14 Strategy b reflects the existing HCV cascade of care at HMP Wormwood scrubs, while optimised rates of screening, linkage-to-care and treatment are varied based on data from a pilot 'simplified pathway' (incorporating point-of-care screening (Oraquick® anti-HCV screening followed by viraemia confirmation using Xpert® HCV FS VL assay), fast track linkage-to-care and unrestricted treatment initiation [irrespective of length of sentence]) in strategies c and d. 15 In addition, strategies e and f are comprised of varied intensity of universal general prison population screening and targeted screening of the high-risk prison population, defined as PWID initiated on OST (estimated to be 14% of the total prison population). 19 A summary of all scenarios is listed below, and the proportion of patients screened, linked-to-care and treated in each case can be found in Table 1. 1. National average reception-based universal general prison population DBS screening (20%), linkage-to-care (40%) and treatment (5%) (Status quo).

| Costing
Costs were considered from a healthcare provider perspective. The total assay cost for screening using DBS was £179.91 (£9.91 anti-HCV, 14 £100 DBS HCV RNA and genotype and £70 laboratory HCV RNA and genotype validation 15 ). While the cost of POC in the simplified algorithm was £117 (Oraquick® anti-HCV; £17, Xpert® HCV FS VL; £30 and laboratory HCV RNA and genotype validation; £70 15 ).
The cost of clinical assessment was £225 (Fibroscan®; £150 and healthcare professional time; £125 15 ) and the total cost of treatment and follow-up was £10 500 (DAA therapy; £10 000, healthcare professional time; £500). HCV disease-related costs were derived from existing literature. 20,21 The costs related to each scenario are presented per capita offered testing as opposed to per capita screened as this allowed us to take into account a health-related output costs among the estimated proportion of individuals who did not undergo screening.

| Outcome measures
Quality-adjusted life years (QALYs) were used to as an quantitative objective measure of HCV-related disease morbidity and each stage of liver disease (mild, moderate, cirrhosis, decompensation and HCC) and liver transplantation (at 1 year post-transplantation and long term) was assigned a value based on existing literature. 20,21 An incremental QALY was applied for non-cirrhotic prisoners who completed therapy successfully. 11

| Cost-effectiveness
Using the status quo scenario as the base case, an incremental costeffectiveness ratio (ICER) per QALY gained for each alternative strategy was generated using the following formula: A half-cycle correction was applied to all recurrent costs and health utilities. This methodology is commonly applied in Markov modelling. 22 In the model the transition from one state to another is discrete (occurring at the beginning or end of a cycle), however, in reality this process is likely to be continuous throughout the cycle. Half of the total accrued costs and effects are taken for the first and last cycle, effectively resulting in the estimate for all intermediate cycles being obtained from the middle time point, avoiding over-or underestimation of the incurred outputs associated with the transition between states. 23 In addition, a standardised discount was applied to all outputs at a rate of 3.5% per annum. In line with the UK National Institute of Health and Care Excellence (NICE), the willingness to pay threshold (WTP) used to assess whether a strategy was cost-effective or not was considered to be £20 000 to £30 000 per QALY gained. 24

| Sensitivity analysis
In order to evaluate the robustness of the results generated, both deterministic (univariate) analysis and probabilistic sensitivity analysis (PSA) have been performed for cost, health utility and probability parameters (  Table 3. and Figure S3A  The results of the cost-effectiveness analysis are presented in

| Sensitivity analysis
The outcomes for the five parameters with the largest impact on the ICER per QALY when compared to the status quo (strategy a) are presented in Figure 3. In all cases none of the parameters appear to independently raise the ICER value above the WTP of £30 000. However, the WTP threshold is breached across all strategies at a more conservative WTP (£20 000). In particular, three common parameters, cost of DDA therapy, the QALY associated with mild disease (F0/1) and the probability of completing treatment, appear to have the largest effect on the ICER value across all strategies.
Figures S1 and S2 represent outputs from the PSA Monte Carlo simulation. Figure S1 represents a cost-effectiveness acceptability

| D ISCUSS I ON
Our study has shown that all of the evaluated strategies to increase coverage of HCV testing and treatment at HMP Wormwood Scrubs are cost-effective, compared to current national average outcomes for prison-based HCV initiatives in England. The ICERs ranged between £3402 (strategy b) and £10 300 (strategy c), while the net increase in annual cost ranged from £2794 (strategy b) to £35 800 (strategy d) and the proportion of inmates successfully being treated ranged between 7.4% (strategy b) and 53.6% strategy d.
One key message that our model emphasises is the importance  Table 4). This is a consequence of incurring the upfront costs of screening, while the impact of limiting disease progression has yet to be gained. Conversely, removing any discount to model outputs or increasing the length of the model further results in a more cost-effective ICER across strategies ( Figure 3). In terms of cost, the price of treatment has the most significant effect of the ICER values for all scenarios tested. This is in line with an economic evaluation of scaling-up HCV treatment in prisons in the USA, which forecasts that the most significant cost to the healthcare budget within the prison system would be the cost of treatment. 28 In our study we selected a relatively conservative estimate for DAA therapy of £10 000. However, in recent times, NHS England and the pharmaceutical industry have agreed to significant reductions on drug tariffs, and it can only be expected that these costs will continue to fall. 29 In addition to the cost of therapy, the successful completion of treatment is an important consideration in the sensitivity analysis.
As a result of the limited length of incarceration, in our experience only 46% of individuals complete treatment under supervision. We have applied a sub-optimal SVR rate (70%) to those who are released prior to completing therapy. Although, recent work to assess the sub-optimal adherence among PWID concluded that there was no observed difference compared to those with better compliance, this was defined as adherence of less than 90%, 30 whereas it is expected to be around 50% in our cohort. Currently there is very little continuity of HCV care with probation services and community harm reduction programmes on release. Thus, improved collaboration with community services to ensure adequate support for individuals to complete therapy would further enhance the cost-effectiveness of prison-based screening and treatment.
It is also interesting to note the impact of treatment in preventing the morbidity and mortality related to liver disease in this cohort. The most alarming finding is the high level of mortality observed despite the relatively mild degree of liver disease in the study cohort at baseline. Even after suspending the contribution of background mortality on the model, the proportion who die over a 30-year period is 3.7% higher among those who are untreated (8.0% vs 10.7%).
These findings are echoed in a recent analysis of mortality in an observational cohort of over 100 000 HCV infected non-cirrhotic individuals, reporting a hazard ratio of 0.32 among those successfully treated with DAA therapy compared with untreated individuals. 31 The gain in health-related quality of life from treatment also appears to be an important determinant of the ICER value across all strategies, particularly in the cohort with 'mild disease' (F0/1). This is because the averted progression of more advanced disease states in those undergoing treatment. A 20 per cent adjustment in the QALY value for treated and untreated individuals with mild disease resulted in an ICER per QALY gained between £20 000 and £30 000 in the more intensive screening and treatment strategies (Figure 3).
This suggests that even where there is no direct health benefit of treatment on an individual level, there is still a cohort benefit at a cost that is likely to be acceptable at the current threshold.
We do acknowledge that there are some limitations to our study. Firstly, our data are generated from a single remand prison site in London. Although this population typically serve short sentences and are therefore theoretically more difficult to treat, the make-up of the prison population, model of treatment delivery and allocation of staff and resources across England is not uniform and therefore our findings need to be generalised with caution.
Secondly, we have made the assumption that the fibrosis transition rates are the same for both PWID and non-PWID. Although the majority of cases can be attributed to PWID, other modes of transmission (prison tattoos, immigrants from high prevalence countries) exist in this setting and these individuals may have a slower rate of disease progression. 16,32 However, altering the rate of progression in our sensitivity analysis does not seem to significantly affect the cost-effectiveness. Thirdly, our study design also does not account for individuals who may be tested or treated subsequently in the community. However, we acknowledge that continuity of care for individuals being released without completing treatment remains inadequate by adjusting the SVR rate to 70%. Finally, our model is designed as a closed cohort, and as such does not account for any dynamic changes in disease incidence.
Thus, any beneficial effects of treatment as prevention is not captured and the role of re-infection is not considered. This concept maybe particularly pertinent to the prison setting in England, as needle-and-syringe services are not available. Therefore, scaling-up treatment en masse may prevent onward HCV transmission where high-risk injecting practices and tattooing may take place and therefore in practice be even more cost-effective than our study has found. Indeed, our data have demonstrated more than two-thirds of individuals are aviraemic 2 weeks into treatment. 15 Thus, it will be of interest to explore the effects treatment as prevention and re-infection in this setting with a dedicated dynamic transmission model.

ACK N OWLED G EM ENTS
The authors thank all patients for their willingness to participate and staff members working HMP Wormwood Scrubs for their support in facilitating the HCV clinic. We also thank Cepheid UK for donating Xpert® FS VL RUO cartridges, which were used as part of the alternative POC pathway.

CO N FLI C T O F I NTE R E S T S
All authors have no conflict of interest to declare in relation to this study.