Impact of new direct‐acting antiviral therapy on the prevalence and undiagnosed proportion of chronic hepatitis C infection

Patients with chronic hepatitis C (CHC) can be cured with the new highly effective interferon‐free combination treatments (DAA) that were approved in 2014. However, CHC is a largely silent disease, and many individuals are unaware of their infections until the late stages of the disease. The impact of wider access to effective treatments and improved awareness of the disease on the number of infections and the number of patients who remain undiagnosed is not known in Canada. Such evidence can guide the development of strategies and interventions to reduce the burden of CHC and meet World Health Organization's (WHO) 2030 elimination targets. The purpose of this study is to use a back‐calculation framework informed by provincial population‐level health administrative data to estimate the prevalence of CHC and the proportion of cases that remain undiagnosed in the three most populated provinces in Canada: British Columbia (BC), Ontario and Quebec.


| INTRODUC TI ON
Chronic hepatitis C (CHC) is a progressive and infectious liver disease, which can remain asymptomatic until patients have progressed to the last stages of the disease, resulting in decompensated cirrhosis (DC), hepatocellular carcinoma (HCC) and death. 1 Treatment for this disease has rapidly evolved and diagnosed patients can now be cured with new interferon-free direct-acting antiviral (DAA) therapies available in Canada since the end of 2014. 1,2However, despite the availability of the highly effective DAA therapies, hepatitis C virus (HCV) infection continues to pose a major public health threat worldwide, with approximately 290 000 deaths resulting from HCVrelated complications in 2019. 3ny countries, like Canada, endorse the World Health Organization's (WHO) goal to eliminate viral hepatitis by 2030. 4 The development of effective and tailored strategies for hepatitis C care requires an understanding of trends in the geographic and demographic distribution of the disease burden, both in terms of the number of infections and, more importantly, the number of patients who remain undiagnosed.Thus, understanding the initial impact of DAAs on the burden of HCV across regions and populations is important in developing strategies that can effectively help meet the WHO 2030 targets.In this context, the estimation of prevalence and CHC undiagnosed proportion are important to inform the development of interventions and to design optimal policies, preventing disease transmission and improving access to the treatment.Such knowledge can also help jurisdictions in planning their resource allocation.
[7][8] However, the corresponding impact on CHC prevalence and the undiagnosed proportion are less known.With fewer than seven years left to achieve the WHO elimination targets, evaluating the impact of DAAs on CHC prevalence and the undiagnosed proportion is critical.
In this study, we estimated the initial impact of DAAs on the CHC prevalence and undiagnosed proportion in British Columbia (BC), Ontario and Quebec, the three most populated provinces in Canada, using a back-calculation modelling approach informed by highly accurate population-level health administrative data.The goal of this study is to provide vital evidence, including the most recent estimation of CHC prevalence and undiagnosed proportion, for each province to evaluate current strategies and to inform future HCV care policies.

| Overview
We have used a two-step process to estimate the CHC prevalence and undiagnosed proportion of chronic hepatitis C in BC, Ontario and Quebec.We first conducted population-based retrospective analyses of health administrative data in each province from 1999 to 2018 to generate population-level statistics on HCV-related health events cases for three birth cohorts: individuals born before 1945, individuals born between 1945 and 1965, and individuals born after 1965.Next, we used a validated natural history model 9 together with a back calculation modelling approach to back-calculate the historical distribution of fibrosis stages and clinical status of the CHC population for each province.This process was completed by calibrating the modelgenerated predictions of the annual number of CHC health events against the corresponding provincial health administrative data.

| Administrative databases and study populations
The BC analysis was performed based on the data collected from the British Columbia Hepatitis Tester's Cohort (BC-HTC), which is a provincial database of individuals tested for hepatitis B virus (HBV) and HCV. 10 The BC-HTC includes HCV confirmed cases reported to public health in BC from 1990 to 2015 as well as all individuals tested for HCV at the British Columbia Centre for Disease Control Public Health Laboratory (BCCDC-PHL) from 1990 to 2018.The record of physician visits, inpatient hospitalization, drug prescription, cancer cases and death are linked for each case. 10e Ontario CHC data have been collected from the provincial database of the Public Health Ontario Laboratory (PHOL), which includes individuals with HCV mono-infection aged 18 years and older in Ontario from 1 January 2000 to 31 December 2018. 11 This data were linked to health administrative data held at ICES (formerly known as Institute for Clinical Evaluative Sciences) using unique identifiers and analysed at ICES to construct the Ontario CHC study cohort from 2000 to 2018. 10 ICES is an independent, non-profit research institute whose legal status under Ontario's health information privacy law allows it to collect and analyse health care and demographic data, without consent, for health system evaluation and improvement.Similarly, a population-based cohort of all laboratory-confirmed hepatitis C diagnoses in Quebec, between 1990 and 2018, was obtained from the reportable disease database (MADO) or the Quebec Public Health Laboratory (LSPQ), linked to several provincial administrative data sets.The reported cases with HCV were linked to the provincial health insurance registry (FIPA) databases using a unique health care number. 12The sources of CHC case definition for retrospective analysis of BC, Ontario and Quebec are given in Hamadeh et al. 13

| Outcomes definition for retrospective analysis
From the health administrative data for BC, Ontario and Quebec, we extracted the annual incidence of newly diagnosed hepatocellular carcinoma (HCC), decompensated cirrhosis (DC), confirmed CHC cases as well as HCV treatment initiations for the three birth cohorts from 2000 to 2018.DC and HCC diagnosis codes were defined based on previously published studies. 12-15

| Step 2: Model-based estimation of CHC population
We used diagnosed cases from the retrospective analyses in a backcalculation framework based on a validated natural history model of CHC to estimate a single birth cohort's CHC population through a calibration process. 16The following subsections describe the estimation process.

| CHC natural history model
[19] Accordingly, we employed three separate state-transition models for each of the cohorts of interest.In each model, simulated individuals move through the predefined health states.Figure A1 in Appendix A shows the state-transition model, describing the movement of patients infected with hepatitis C between the various health states.
Individuals who are infected with acute hepatitis C can develop chronic infection if the infection is not resolved.We considered five fibrosis stages for CHC based on disease progression (F0 to F4).In each fibrosis stage, individuals with CHC are stratified by undiagnosed and diagnosed status.Only diagnosed individuals have a chance to receive a treatment.If the treatment is successful, patients achieve sustained virologic response (SVR).Patients in the F4 stage may develop advanced liver diseases, which we considered as DC and HCC.

| CHC natural history model | Model parameters
The parameters of the CHC natural history model are obtained from the published literature. 9The parameter values, including viral genotype distribution (Table B1), disease progression data (Table B2), treatment efficacy data by viral genotype and fibrosis stage (Tables B3 and B4), the distribution of the fibrosis level for each birth cohort at diagnosis (Table B5) and annual probability of background death (Tables B6-B8), are shown in Appendix B.

| Back-calculation framework
We use a Bayesian Markov Chain Monte Carlo (MCMC) algorithm to back-calculate the historical prevalence of a single birth cohort's CHC population through a calibration process. 16The algorithm constructs probability distributions of the historical CHC prevalence and the undiagnosed proportion by comparing the model-generated predictions of the annual number of CHC health events through the CHC natural history model against the observed data obtained from the retrospective analyses of health administrative data extracted in step 1.The calibration targets include the annual number of newly diagnosed HCC, DC, CHC cases and new treatment initiations for each cohort from 2000 to 2018.The mathematical details of the calibration process were described previously. 16We also conducted a one-way sensitivity analysis to investigate the uncertainties associated with the model parameters on the CHC prevalence and the undiagnosed estimations.

| Trends in CHC diagnosis and treatment
The number of confirmed CHC diagnoses per 100 000 population is reported in Appendix Figures C1C-C3C

| Step 2: Results of model-based estimation of CHC population
For each birth cohort, the historical CHC population is estimated using a CHC natural history model and a back-calculation calibration process based on the data generated from the retrospective studies in step 1 (Appendix Figures C1-C3).The model estimation of prevalence and undiagnosed proportion are shown in Figures 1-6 and are reported in Tables 1 and 2. The final calibrated model demonstrates a good fit between the model-generated data and the observed data (Appendix D).  1F). Figure 2 shows the prevalence and undiagnosed CHC proportion estimates for BC averaged over all the birth cohorts for the years 2000-2018.The averaged CHC prevalence and averaged undiagnosed proportion was assessed to be 1.23% (95% CI: .96%-1.62%) and 35.44% (95% CI: 27.07%-45.83%) in 2018 respectively.

| Initial impact of DAA treatment
The new DAA treatment was approved in 2014.Since then, the uptake of CHC treatment has increased rapidly across all the three birth cohorts in three provinces (Appendix Figures C1D-C3D).To assess the initial impact of DAA on the prevalence and the undiagnosed proportion, we compared the prevalence and the undiagnosed proportion estimations between 2014 and 2018.Based on our model estimation, the prevalence has declined from 1.39% to 1.23%, .97% to .91% and .65% to .57% in BC, Ontario and Quebec, respectively, from 2014 to 2018.Also, the undiagnosed proportion has declined from 38.78% to 35.44%, 38.70% to 34.28% and 47.54% to 46.32% in BC, Ontario and Quebec respectively.

| Sensitivity analysis results
The one-way sensitivity analysis results are shown in Figures E1-E3 in Appendix E. For individuals born before 1945, CHC prevalence was sensitive to the estimated number of new HCV infections occurring annually, SVR rate and CHC progression.For individuals born between 1945 and 1965, CHC prevalence was sensitive to treatment rate and CHC progression rates.For individuals born after 1965, CHC prevalence was sensitive to the estimated number of new HCV infections occurring annually.For all the three provinces, the CHC undiagnosed proportion was sensitive to the annual probability of CHC diagnoses for the three birth cohorts.

| DISCUSS ION
We have estimated CHC prevalence and the undiagnosed proportion using a validated CHC natural history model and a back-calculation framework informed by population-based health administrative data.We have estimated the CHC prevalence and undiagnosed proportion for the three largest provinces in Canada: Ontario, BC and Quebec, from 2000 to 2018, focusing on investigating the short-term impact of the new DAA treatment.Our results showed a continuous decreasing trend for both CHC prevalence and undiagnosed proportion in BC, Ontario and Quebec after the introduction of new DAA treatment.However, the impact of DAA treatment on prevalence is relatively small, and it may require more time to see the effect, as the new treatment became available in late 2014, but was initially restricted to patients with severe disease until early 2017. 20,21

| Comparison with previous studies
3][24] Recently, the Public Health Agency of Canada (PHAC) estimated the prevalence and undiagnosed proportion of CHC infection following the introduction of DAA therapy using a back calculation modelling approach and the workbook method. 22e model was calibrated using HCV surveillance data.However, in this study, only reported HCV diagnoses were considered.The study does not include CHC-related advanced liver diseases such as new HCC and DC diagnoses, which may generate estimations with substantive uncertainty.In Europe, a state-transition model was developed and used to back-calculate the prevalence of HCV for 28 European Union (EU) countries in 2015. 23However, this assessment was completed in the first year of the availability of DAA and did not capture the real-world impact of the therapy.Lastly, a study in the United States used data from the National Health and Nutrition Examination Survey together with mathematical modelling to estimate HCV prevalence. 24However, the study only reported an overall estimation during 2013-2016 and did not provide longitudinal estimations, which is needed for assessing the impact of the new treatment.6][27][28][29][30] The results demonstrated a reduction in CHC-associated HCC and CHC-related mortality due to DAA uptake. 26,27Similarly, DAA was also shown to be effective in reducing all-cause, liver-and drug-related mortality and HCC in British Columbia. 28,29In addition, DAA treatments were reported to be useful in lowering the economic burden of the disease. 30,31

| Limitations and strengths
We acknowledge some limitations of this work.First, it is possible that the health administrative data may have systematically under-sampled marginalized individuals, who may not have a permanent address from which to apply for publicly funded health insurance.Second, the model may underestimate the prevalence of CHC in the youngest cohort since the progression from CHC to DC or HCC may take 20-40 years, and our algorithm relies on these statistics to project the results.Similarly, the model may also underestimate the CHC undiagnosed proportion in the youngest cohort since the lower diagnoses cases of HCC would results in the lower estimation of undiagnosed proportion.Finally, the model has inherited the uncertainty that presents in the parameters of the previously published natural history model. 9However, our Bayesian back-calculation framework can reflect the impact of such uncertainties within the reported confidence intervals.

TA B L E 1
The estimation of CHC prevalence for the three birth cohorts as well as the averaged result across birth cohorts for each province (British Columbia, Ontario, and Quebec).

Province
British Columbia Ontario Quebec

TA B L E 2
The estimation of CHC undiagnosed proportion for the three birth cohorts as well as the averaged result across birth cohorts for each province for British Columbia, Ontario and Quebec.
providers (eg healthcare organizations and government) prohibit ICES from making the data set publicly available, access may be granted to those who meet pre-specified criteria for confidential access, available at www. ices.on.ca/ DAS (email: das@ices.on.ca).
The full data set creation plan and underlying analytic code are available from the authors upon request, understanding that the computer programs may rely upon coding templates or macros that are unique to ICES and are therefore either inaccessible or may require modification.

E TH I C S S TATEM ENT
This project has been approved by the Research Ethics Board at the

1 |
Step 1: Results of retrospective analyses of health administrative data Appendix Figures C1-C3 show the confirmed incident rate of CHC diagnoses, CHC-related HCC diagnoses, CHC-related DC diagnoses and CHC-related treatment for BC, Ontario and Quebec from 2000 to 2018 for individuals born before 1945, individuals born between 1945 and 1965, and individuals born after 1965 respectively.The source data are given in Appendix C.

3. 1 . 1 |
Figure C2A,B).As shown in Appendix FigureC3A, a much smaller number of cases with DC per 100 000 population were reported from 2000 to 2018 for the three cohorts respectively.All three provinces display a decreasing trend in the cohort of individuals born before 1945 and the cohort of individuals born between 1945 and 1965.However, an increasing trend is observed for the cohort of individuals born after 1965 in BC and Ontario since 2010.In addition, a lower confirmed CHC diagnoses rate has been observed for Quebec in comparison to Ontario and BC for the three cohorts.In terms of treatment, Appendix Figures C1D-C3D show the number of new CHC treatment initiations per 100 000 population for the three provinces from 2000 to 2018.The CHC treatment rates rapidly increased after 2014 when the new DAA treatments were introduced.

Figure 1
Figure 1 illustrates the back-calculated estimation of the BC CHC prevalence and undiagnosed proportions over 2000-2018 for the three cohorts.The prevalence in 2018 was .75% (95% CI: .63%-.97%) among the individuals born before 1945 (Figure 1A), 2.23% (95% CI: 1.72%-3.51%)among the individuals born between 1945 and 1965 (Figure 1C) and .89%(95% CI: .62%-1.12%) among the individuals born after 1965 (Figure 1E).The CHC prevalence has decreased over the years for individuals born between 1945 and 1965 and individuals born after 1965 while it remained steady for individuals born before 1945.However, a decreasing trend of the undiagnosed CHC proportion can be observed over the years for all the three cohorts (Figure 1B,D,F).In 2018, the undiagnosed CHC

F I G U R E 2
British Columbia CHC estimates averaged over all birth cohorts for 2000-2018.(A) CHC prevalence.(B) Undiagnosed CHC proportion.F I G U R E 3 Ontario CHC prevalence and undiagnosed CHC proportion estimates by cohort over 2000-2018.(A) CHC prevalence for pre-1945 birth cohort based on ON data in Appendix Figure C1.(B) Undiagnosed CHC proportion for pre-1945 birth cohort based on ON data in Appendix Figure C1.(C) CHC prevalence for 1945-1965 birth cohort based on ON data in Appendix Figure C2.(D) Undiagnosed CHC proportion for 1945-1965 birth cohort based on ON data in Appendix Figure C2.(E) CHC prevalence for post-1965 birth cohort based on ON data in Appendix Figure C3.(F) Undiagnosed CHC proportion for post-1965 birth cohort based on ON data in Appendix Figure C3.

F I G U R E 4
Ontario CHC estimates averaged over all birth cohorts for 2000-2018.(A) CHC prevalence.(B) Undiagnosed CHC proportion.F I G U R E 5 Quebec CHC prevalence and undiagnosed CHC proportion estimates by cohort over 2000-2018.(A) CHC prevalence for pre-1945 birth cohort based on QC data in Appendix Figure C1.(B) Undiagnosed CHC proportion for pre-1945 birth cohort based on QC data in Appendix Figure C1.(C) CHC prevalence for 1945-1965 birth cohort based on QC data in Appendix Figure C2.(D) Undiagnosed CHC proportion for 1945-1965 birth cohort based on QC data in Appendix Figure C2.(E) CHC prevalence for post-1965 birth cohort based on QC data in Appendix Figure C3.(F) Undiagnosed CHC proportion for post-1965 birth cohort based on QC data in Appendix Figure C3.

F I G U R E 6
Quebec CHC estimates averaged over all birth cohorts for 2000-2018.(A) CHC prevalence.(B) Undiagnosed CHC proportion.