Prison‐based interventions are key to achieving HCV elimination among people who inject drugs in New South Wales, Australia: A modelling study

Abstract Background & Aims People who inject drugs (PWID) experience high incarceration rates which are associated with increased hepatitis C virus (HCV) transmission risk. We assess the importance of prison‐based interventions for achieving HCV elimination among PWID in New South Wales (NSW), Australia. Methods A model of incarceration and HCV transmission among PWID was calibrated in a Bayesian framework to epidemiological and incarceration data from NSW, incorporating elevated HCV acquisition risk among recently released PWID. We projected the contribution of differences in transmission risk during/following incarceration to HCV transmission over 2020–2029. We estimated the past and potential future impact of prison‐based opioid agonist therapy (OAT; ~33% coverage) and HCV treatment (1500 treatments in 2019 with 32.9%–83.3% among PWID) on HCV transmission. We estimated the time until HCV incidence reduces by 80% (WHO elimination target) compared to 2016 levels with or without prison‐based interventions. Results Over 2020–2029, incarceration will contribute 23.0% (17.9–30.5) of new HCV infections. If prison‐based interventions had not been implemented since 2010, HCV incidence in 2020 would have been 29.7% (95% credibility interval: 22.4–36.1) higher. If current prison and community HCV treatment rates continue, there is an 98.8% probability that elimination targets will be achieved by 2030, with this decreasing to 10.1% without current prison‐based interventions. Conclusions Existing prison‐based interventions in NSW are critical components of strategies to reduce HCV incidence among PWID. Prison‐based interventions are likely to be pivotal for achieving HCV elimination targets among PWID by 2030.


| INTRODUC TI ON
Over half of people who inject drugs (PWID) globally have been exposed to hepatitis C virus (HCV), 1 with injecting drug use (IDU) being the leading cause of HCV infection in developed countries. 2 Modelling studies suggest that disease elimination is achievable through the scale-up of highly effective direct-acting antiviral (DAA) therapies for chronic HCV infection. 3,4 In 2016, this led the World Health Organization (WHO) to set a goal of eliminating HCV as a public health threat by 2030. PWID need to be a high priority for elimination efforts. However, poor coverage of harm reduction measures, restricted access to DAAs, and criminalization of drug use are critical barriers to achieving HCV elimination among PWID. 5 Evidence suggests that incarceration can be an important driver of HCV transmission among PWID due to high incarceration rates (>50% ever 1 ) and heightened risk during incarceration and/or postrelease. [6][7][8] Our recent systematic review suggests that among community PWID recent incarceration is associated with a 62% increased risk of HCV acquisition, 9 increasing to 180% in Australia. 9 Prison-based interventions can have substantial impact on the overall HCV epidemic among PWID. 7,8 In NSW prisons, Opioid agonist therapy (OAT) has a high coverage (33% 10 ), comparable to the community coverage (40%). 11 Recognition of the importance of the prison sector for HCV transmission and an opportunity for DAA treatment scale-up, has also led to specific initiatives to ensure universal access to DAAs for prisoners in Australia. While prisoners only constituted 6% of individuals initiating DAA therapy in 2016, this has risen progressively to 29% in 2019. 12 We use a model of HCV transmission and incarceration among PWID in NSW to assess the contribution of incarceration to HCV transmission, evaluate the historical impact of prison-based OAT and HCV treatment, and determine their importance for achieving WHO elimination target of reducing HCV incidence by 80% by 2030.

| Model description
We adapted a published 8  PWID initiate and leave OAT at constant rates, with excess overdose mortality at these points 14 and higher recruitment rates if they have been on OAT before. PWID are incarcerated or re-incarcerated at different rates (to capture greater rates of re-incarceration than primary incarceration), varying by injecting duration, and are released at a constant rate. PWID experience reduced rates of incarceration while on OAT. 15 Two-thirds of PWID are retained on OAT upon release. 16 PWID transmit HCV in their given setting (prison or community).
Transmission occurs at a rate proportional to the chronic prevalence in each setting and the infection rate. The infection rate differs by setting, and is elevated among PWID with <3 years of injecting 17,18 or who have been recently released from prison, 9 and is reduced if on OAT. 19 Most new infections lead to chronic infection. 20 A time-varying number of chronically infected PWID are treated annually. Community PWID on OAT are more likely to be treated than those not on OAT. 11,21 A proportion of treated PWID achieve sustained viral response (SVR). Those failing treatment are eligible for retreatment due to unrestricted access in Australia.

| Model parameterization and calibration
Data for parameterizing and calibrating the model came primarily from the ANSPS Survey, 11  We assume that the incarceration dynamics, HCV epidemic and demographics of PWID were stable in 2010 based on data from ANSPS. OAT initiation rates were calibrated to give stable coverages of 29%-37% among incarcerated PWID and 40%-52% among

Lay Summary
• The added risk associated with incarceration results in it contributing a fifth of all new HCV transmission among people who inject drugs (PWID) in New South Wales (NSW, Australia).
• Existing prison-based interventions (opioid agonist therapy and HCV treatment) have reduced HCV incidence among PWID by a quarter. The model assumes a factor increase in community HCV acquisition risk among PWID recently released from prison (2.78, 95% CI: 2.00-3.85), based on Australian data. 9 The HCV transmission risk in prison is calibrated to give the difference in HCV antibody prevalence among community PWID with and without a history of incarceration.
The model is calibrated using an approximate Bayesian computation sequential Monte Carlo algorithm (see Appendix S1). The algorithm begins with 1000 parameter sets sampled from prior distributions (Table 1), which are then resampled and perturbed in an iterative manner to better fit the data until the goodness of fit (measured using log-likelihood) no longer improves (<0.5% relative difference between successive iterations).

| Model analyses
Using the calibrated model, we projected the contribution (population attributable fraction, 'PAF') of incarceration to HCV transmission among PWID over 10 years from 2020 to 2029, calculated as Scale-up prison-based HCV treatment rates by 44% in 2020 as achieved by the SToP-C intervention study in NSW prisons. 10

| Scenario 2
Introduce NSP into prison at 50% or 100% coverage, assuming NSP reduces HCV transmission risk by 56% (95% CI 20%-76%). 19 This is modelled by reducing the force of infection in prison based on the coverage and effectiveness of NSP (see Appendix S1).
For each scenario, we projected the date by which an 80% reduction in HCV incidence (compared to 2016) could be reached among all PWID and the probability (estimated by proportion of runs) of achieving this target by 2030. In sensitivity analyses, we also considered how the contribution of prison-based interventions to HCV elimination would differ if community treatment rates reduced by 25% from 2020.

| Status quo model projections
The calibrated model agrees well with the available data ( Figure 3 and Appendix S1). Model projections on average fall within 74% of the 95% CI for the four HCV incidence validation points (not fit to) and project that 6328 (95% CrI: 4176-8109) PWID were treated for HCV between 2016 and 2018 ( Figure 2), similar to what was suggested by data (7560 35 ).
The model projects an HCV chronic prevalence of 20.5% (17.6-23.9) among PWID in 2020, and an HCV incidence of 1.9

| Historical impact of prison-based interventions
In 2020, the model projects OAT coverages of 34.3% (28.

| Role of prison-based interventions in HCV elimination
The status quo model projects ( Figure 4 and Figure 5) that an 80%

| DISCUSS ION
Prison exposure is common among PWID. Our modelling suggests 8% of PWID in NSW were incarcerated in 2020, approximately 50 times greater than the overall national average, with incarceration contributing one-fifth of ongoing HCV transmission among PWID in this setting. Prison-based OAT and HCV treatment have mitigated some of this increase in HCV transmission risk associated with incarceration, averting 16% of infections in 2019 and preventing the HCV incidence among PWID from being 30% higher in 2020. Over 2020-2030, prison-based interventions, particularly HCV treatment, will be pivotal for achieving HCV elimination among PWID, increasing the probability of achieving elimination by 2030 from 10% to 99%.

| Strengths and limitations
This study presents the first impact analysis of an ongoing HCV elimination initiative among PWID in the community and prison. Main practices and drug and alcohol clinics. It is likely that the COVID-19 pandemic will have further impacted on the 2020 treatment rates.
To account for these possible decreases, we considered scenarios in which community treatment rates are reduced by 25% from 2020, with these sensitivity analyses showing an increased important for existing and new prison-based interventions for achieving HCV elimination.
Secondly, we did not model HCV testing and diagnosis; instead we assumed that all chronically infected PWID could be treated with greater treatment rates. We were therefore unable to consider the contribution of prison-based HCV screening. While most people diagnosed with HCV in prison will begin treatment in prison, this may also lead to increased HCV treatment in the community because the time in custody is typically short (weeks-months). We therefore may have underestimated the importance of prison-based interventions for achieving HCV elimination.
Thirdly, our findings may not be generalizable to other settingseither in Australia or internationally, as our model was parameterized and calibrated using detailed epidemiological data from NSW.
However, given that NSW has high coverage of harm reduction interventions and high levels of ongoing HCV treatment in the community, it is likely that prison-based interventions will also be important for achieving HCV elimination among PWID in many other global settings, particularly those with similarly high levels of incarceration or in settings with longer prison sentences.
Finally, although there is uncertainty around the effectiveness of prison-based OAT, with the HITS-p study in NSW finding no protective effect on HCV incidence, 23

| Comparisons with existing studies
Our analysis is consistent with previous modelling that showed incarceration to be an important contributor to HCV transmission in Scotland and Kentucky. 7,8 However, our estimates are lower than F I G U R E 5 Date of achieving an 80% reduction in overall HCV incidence among PWID compared to 2016 if (A) community treatment rates continue or (B) community treatment rates reduce by 25%. The box plots signify the uncertainty (middle line is median, limits of boxes are the 25% and 75% percentiles, and whiskers are 2.5% and 97.5% percentile range) in the time estimates for achieving elimination due to uncertainty in the model parameters. Labels represent the probability of achieving elimination before 2030.
previous illustrative modelling for Australia, which found a PAF of incarceration of 49%, 6 although that analysis assumed a much higher HCV incidence within prisons and was not calibrated to incarceration or epidemiological data from Australia.
Our results are also consistent with findings from previous modelling studies that show prison-based prevention and treatment interventions can reduce HCV transmission in the community. 7,8 However, ours is the first study to directly quantify the importance of ongoing prison-based interventions for achieving HCV elimination among all PWID. Previous modelling of HCV transmission in NSW prisons found that HCV elimination targets for incidence in the prison setting could be achieved by 2030 via scaling-up HCV treatments in the community, but that a combination approach of scaled-up in-prison HCV treatment and improved harm reduction (OAT and/or NSP) would be required to substantially reduce incidence in prisons by 2030 if community HCV treatment rates reduced. 27 This study was limited by not mechanistically modelling the interface between community and prison and so was unable to fully capture the effects of interventions in either setting. Other modelling at the national level suggests that existing HCV treatment rates are probably sufficient for Australia achieving HCV elimination targets for incidence, 3,37 but that testing may need scaling-up to maintain these treatment rates. 38 These studies only modelled community PWID and so did not consider the effect of differences in transmission risk or intervention uptake between prisons and the community. Our study therefore extends previous analyses by dynamically modelling HCV transmission in prisons and the community, allowing us to demonstrate the importance of prison-based interventions for reaching the HCV elimination targets for reducing incidence.

| Implications
Australia has committed to eliminating HCV as a public health threat, with the ongoing universal access DAA program achieving high levels of treatment uptake. Our modelling demonstrates that NSW is on track to achieve HCV elimination targets for incidence