Assessment of Incidence of and Surveillance Burden for Hepatocellular Carcinoma Among Patients With Hepatitis C in the Era of Direct-Acting Antiviral Agents

Key Points Question What are the projected changes in incidence of and surveillance burden for hepatocellular carcinoma (HCC) among populations with hepatitis C virus (HCV) in the era of direct-acting antiviral agents (DAAs)? Findings In this decision-analytical model, the incidence of HCC associated with HCV and candidates for HCC surveillance were projected to increase before starting to decrease in the era of DAAs. The burden of HCC associated with HCV was projected to shift from patients with viremia to individuals with virologically cured HCV and to older individuals. Meaning Results of this study suggest that routine HCC surveillance is needed for early detection of HCC in individuals with virologically cured hepatitis C who may no longer be receiving specialty care of liver diseases.


Introduction
Hepatocellular carcinoma (HCC) is the fastest rising cause of cancer-related mortality in the US. 1,2 The incidence of HCC has increased over the last 2 decades owing to the hepatitis C virus (HCV) epidemic. 3 Most HCC cases are diagnosed in advanced stages, with a median survival less than 1 year. 4 Regular surveillance for HCC may help improve early cancer detection rates when curative treatment can be applied and is recommended in patients with HCV-associated cirrhosis. [5][6][7] The availability of direct-acting antiviral agents (DAAs) for HCV treatment has substantially altered the landscape of HCV. Though new DAA regimens can result in a virological cure (ie, sustained virological response [SVR]) in more than 90% of patients with HCV, 8 many patients remain at risk of developing HCC after virological cure. 9,10 Current clinical guidelines recommend that patients with virologically cured HCV with cirrhosis (or advanced fibrosis) to undergo routine HCC surveillance every 6 months. 11,12 Though the incidence rate of HCC in patients with virologically cured HCV is lower than that in patients with viremia, the absolute number of patients with virologically cured HCV requiring routine HCC surveillance is estimated to increase in the next decade. 13 Patients with virologically cured HCV may experience improvement in liver function and hence live longer and could undergo surveillance for several years. In contrast, most patients with virologically cured HCV, despite being at risk for HCC, may not undergo routine HCC surveillance because of the perceived lower risk. Previous studies have reported that HCC surveillance rates are much lower in patients receiving primary care vs patients receiving specialty care. 14,15 Thus, identifying HCC cases through surveillance in this cohort could be more challenging than that in other individuals at risk.
Estimating the burden of HCC surveillance in the cohorts of patients with virologically cured HCV may help guide primary and specialty care providers. The objective of this study was to evaluate the burden of HCC surveillance in the era of new HCV antiviral treatments by projecting the number and the characteristics of new HCC incident cases and candidates for routine HCC surveillance.

Methods
We used a previously developed mathematical model, Hepatitis C Disease Burden Simulation model (HEP-SIM), 8,16 to simulate the population with HCV who would be considered candidates for HCC surveillance in the era of DAAs in the US. The HEP-SIM is an individual-level state-transition model that simulates the changing landscape of HCV by replicating disease progression, screening for HCV, different waves of antiviral treatments for HCV, and insurance coverage from 2001 to 2040. The model was previously validated with results of the National Health and Nutrition Examination Survey (NHANES) 2003-2010, 17 reports from the Centers for Disease Control and Prevention, 18 and natural history results from a long-term follow-up study of patients with advanced fibrosis. 19 In this study, we projected temporal trends in HCC incidence and the changing characteristics of the number of patients with HCV (patients with viremia and patients with virologically cured HCV) who will be candidates for routine HCC surveillance in the era of DAAs. All data used in this analysis were publicly available and this study thus did not require approval from an institutional review board. This study followed the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline for decision analytical model studies. We describe the major model components; further model details can be found elsewhere. 16 the Centers for Disease Control and Prevention-reported estimates for the years 2006 to 2015 (eTable 3 in the Supplement), and we assumed that the annual incidence would change at the average rate observed during 2006 to 2015 until 2026 and then become constant.
We simulated the natural history of HCV, which was defined by different stages of liver disease, with chronic HCV infection defined using METAVIR fibrosis scores (F0, no fibrosis; F1, portal fibrosis without septa; F3, numerous septa without fibrosis; and F4, cirrhosis), decompensated cirrhosis, HCC, liver transplant, and liver-related death. Rates of fibrosis progression and decompensation were estimated from published metaregression analysis and observational studies [20][21][22] (eTable 4 in the Supplement). Patients with decompensated cirrhosis or with HCC had higher mortality rates 23 and were eligible to for liver transplant. [24][25][26]

Screening and Treatment for Patients With HCV in the DAA Era
To accurately represent the HCV disease burden that would subsequently affect the burden of HCC, we simulated the clinical management of HCV, ie, screening and treatment, from 2001 to 2040. In each year, patients with HCV could be diagnosed through usual care, which included risk-based screening (eTable 5 in the Supplement). Since 2013, we implemented one-time birth-cohort screening for individuals born between 1945 and 1965 among the NHANES, nursing home, and homeless populations in addition to the diagnosis through usual care. From 2020 to projections to 2040, we included one-time universal screening of all adults between aged 18 to 79 years, following the updated screening recommendation by the US Preventive Services Task Force. 27 For universal screening, we assumed the screening rate of 9% per year, which is the rate observed in the one-time birth-cohort screening policy. 28,29 For HCV treatment, we modeled multiple waves of antiviral treatments available in the US during different periods. 16 The market share of different types of treatment regimens for each HCV genotype was estimated using commercial claims from research organizations (QuintilesIMS and Ipsos) (eTable 6 in the Supplement). Treatment efficacy measured as SVR rate was dependent on the regimen type, individual HCV genotype, presence of cirrhosis, and failure of previous treatment if any, which were estimated based on data from multiple clinical trials and real-world studies (eTable 7 in the Supplement). For all populations except individuals who were incarcerated or homeless, the number of patients receiving treatment in each year was estimated based on published data and drug sales until 2016 (eTable 8 in the Supplement). In the individuals who were incarcerated and homeless, treatment uptake was 1% per year from 2010 to 2016 and 5% per year starting in 2017. 30 Our model did not consider treatment for patients with decompensated cirrhosis. With limited treatment capacity, priorities were given to patients with bridging fibrosis (F3) or compensated cirrhosis. Patients whose previous treatment failed may be eligible for retreatment, depending on his previous regimen, new available regimen, and presence of cirrhosis (eTable 7 in the Supplement). 16

HCC Incidence
Patients with HCV having bridging fibrosis (F3), compensated, and decompensated cirrhosis were at risks of developing HCC. 9,10,22 The HCC risk in patients with virologically cured HCV was lower than in patients with viremia (eTable 4 in the Supplement). 9 Patients with compensated cirrhosis could progress to decompensated cirrhosis even after achieving virological cure of HCV and they remained at risk of developing HCC. 31 We assumed a lower risk of HCC in patients with decompensated cirrhosis with virologically cured HCV than in those with viremia based on a prospective cohort study. 32 or were uninsured (26.7%) based on NHANES data. 13,28,33 For the population of individuals who were homeless, the insurance distribution was estimated from a 2012 study (3.2% privately insured, 31.5% Medicaid, 65.3% uninsured). 34 We incorporated insurance expansion during the period from 2014 to 2017 owing to the implementation of the Affordable Care Act (eTable 10 in the Supplement). We also include patients who were incarcerated as a separate population because this population represents a substantial group of viremic infections that were not accounted for in NHANES data. 35

Model Outcomes
Our primary model outcome was the number of patients who developed HCC and the number of patients who were candidates for routine HCC surveillance from 2012 to 2040. Candidates for HCC surveillance were defined as patients with F3 fibrosis stage, or compensated cirrhosis, with or without virological cure. In addition, we assumed that 15% (range, 5%-25%) of patients with compensated cirrhosis (eligible for liver transplant) were candidates for HCC surveillance. 36,37 We further stratified the number of candidates for HCC surveillance by viremia status (ie, with viremia or virologically cured HCV), cirrhosis, and subpopulation category. We also projected the age distribution of HCC incident cases and surveillance candidates for each year. To account for model uncertainty in parameters, we conducted probabilistic sensitivity analysis by sampling model inputs from the recommended statistical distributions 38 (eTable 11 in the Supplement). We evaluated model outputs for sampled parameters by 2000 iterations and presented the 95% uncertainty intervals (UIs) for model outcomes. The simulation model was developed in C++ programming language and all statistical analysis was performed in R version 3.6.0 (R Project for Statistical Computing).

HCC Incidence
This decision analytical model study was conducted from January 2019 to February 2020. In 2012, the annual incidence of HCC among patients with HCV (with viremia and virologically cured HCV) was 18 000 (95% UI, 11 000-32 000). The annual incidence of HCV-associated HCC is projected to increase to 24 000 (95% UI, 18 000-31 000) cases in 2021 and decrease to 13 000 (95% UI, 11 000- We estimated that between 2012 and 2040, the cumulative incidence of HCV-associated HCC would be 583 000 cases (95% UI, 458 000 cases to 738 000 cases), and 27.1% (95% UI, 20.3%-34.7%) of these cases would develop among patients with virologically cured HCV. The 95% UIs for the annual HCC incidence by virologically cured status in each year are presented in eFigure 2 in the Supplement.    patients with virologically cured HCV is estimated to increase from 34.5% to 43.5% in the same time period (eFigure 3 in the Supplement).

JAMA Network Open | Gastroenterology and Hepatology
We further analyzed HCC surveillance candidates by subpopulation (Figure 3)

Aging of HCC Population
The

Discussion
The incidence of HCC has increased over the last 2 decades primarily because of the chronic HCV. 3 With the availability and wider use of new DAAs for HCV treatment, the burden of HCC and routine HCC surveillance attributable to HCV could change in the era of DAAs. In this modeling study, we found that the incidence of HCC is estimated to continue to increase until 2021.  Year surveillance were virologically cured, which is projected to increase to 65% by 2040. The average age of HCC incidence and surveillance candidates is also projected to shift from 55 years in 2012 to more than 70 years in 2040.
Our study provides insights on the trends in HCC in the DAA era. Some studies [39][40][41] conducted statistical analysis based on cancer registry data predominately in the pre-DAA era. The forecasted incidence was based on historical trends without factoring in the changes in the management of HCV. 39,42 Thus those findings may not be applicable to the DAA era. While other studies that used DAA-era data reported that the burden of HCC will decrease in the era of DAAs, 8,13,43 those studies excluded patients with HCV who were incarcerated, homeless, and residing in nursing homes, and were active-duty military personnel. Our study accounted for such populations with less access to screening and treatment for HCV than the general population, and we found that the overall HCC incidence in the population with HCV will continue to increase until 2021 despite the recent availability of highly effective DAAs.
Our study also provides information on the shifting burden of surveillance for HCC and the need to emphasize the clinical treatment for patients with virologically cured HCV. Despite virological cure, patients with advanced fibrosis or cirrhosis will remain at risk of developing HCC. Therefore, as the number of patients with virologically cured HCV increases, the new cases of HCC from this cohort will increase. In addition, the burden of surveillance for HCC will shift from patients with viremia to patients with virologically cured HCV. These changes may have substantial practical implications for the continuum of care. Most of the patients with virologically cured HCV will transition from receiving care from liver specialists to receiving care from primary care physicians. However, most people do not receive regular surveillance in primary care settings 44 and the knowledge of HCC surveillance in primary care settings remains low. 45 Therefore, it is vital to emphasize that biannual HCC surveillance is warranted for these patients in primary care settings.
The optimal clinical management of patients with virologically cured HCV remains unclear.
While the HCC management guidelines by the European Association for the Study of the Liver recommends routine HCC surveillance in patients with virologically cured HCV having advanced fibrosis and cirrhosis, 7 the guideline by the American Association for the Study of Liver Diseases recommends routine HCC surveillance only in individuals with cirrhosis without specifying the status of virological cure. 6 The optimal surveillance strategy in patients with virologically cured HCV will rely on more evidence and studies in the following 3 aspects. First, long-term data are needed for a better understanding of the risk of HCC after DAA-induced virological cure in patients with HCV. It is unclear how much of the current knowledge of the HCC risk among patients with virologically cured HCV from an interferon-based regimen could be used to inform surveillance policies in the era of DAAs.
Some studies suggest that patients with HCV cured by DAA regimens could have higher HCC risks than those with HCV cured by interferon-based regimens, 9,10,46 but the increase was not found to be significant after adjusting for confounders, 47 and the difference was not found to be conclusive. 48 In addition, compared with the available data in the interferon-based treatment era, the observations in DAA era are subject to shorter follow-up time and present an older at-risk population.
Second, health economic data are needed to identify cost-effective HCC surveillance policies for patients with virologically cured HCV. The primary reason for not recommending routine surveillance in patients with virologically cured HCV with advanced fibrosis may be that biannual surveillance using ultrasonography and α 1 -fetoprotein testing for the lifetime was not found to be cost-effective in this cohort. 49,50 Instead of a general surveillance, further research is warranted to identify riskbased surveillance in these patients that could provide a good use of limited resources. 50 In addition, the optimal age for stopping surveillance is unknown. It is plausible that patients with virologically cured HCV may not need HCC surveillance throughout their remaining lifetime.
Third, tailoring surveillance policies to individual-level factors may be a useful approach, especially as emerging molecular biomarkers show their value in the prediction of HCC risks in addition to current clinical risk factors. 51 Comprehensive risk prediction models can be developed by

JAMA Network Open | Gastroenterology and Hepatology
Surveillance Burden for Hepatocellular Carcinoma Among Patients With Hepatitis C integrating multilevel risk factors to further refine the risk predictions. 52 A tailored surveillance strategy could differentiate the patients at high risk who can benefit more from surveillance compared with the others, and thus further improve the cost-effectiveness and efficiency of medical resource use. 53,54 Like any modeling study, our model projection results are subject to inherent uncertainty that arises from the uncertainties in model parameter estimates and model assumptions. Some model parameters and assumptions cannot capture all possible disease variability and diverse real-world clinical practice. Therefore, instead of presenting the prediction of outcomes, our model projection results provide insights on the overall trends of the HCC incidence and surveillance burden, which has also been shown to be robust in the presence of model uncertainty from our sensitivity analysis.

Limitations
This study has limitations. First, we did not explicitly model HCC diagnosis process, which depends on incidence, underlying surveillance strategies, and testing accuracy. We considered HCC as an aggregate health state and used aggregate-level parameters for the population to characterize the overall burden of HCC. Further details would be necessary for analyzing the outcomes of particular HCC surveillance and treatment strategies. Second, our analysis was limited to HCV-associated HCC and did not account for other emerging risk factors such as metabolic disorder (eg, diabetes, nonalcoholic fatty liver disease), hepatitis B, and alcohol consumption. Management of HCC in these other populations at increased risk is becoming increasingly important but would require modeling of different disease factors, which is beyond the scope of this study. In addition, our model did not account for fibrosis regression following successful DAA treatment. Based on recent studies that suggested stable HCC risks following SVR in patients including those with improved noninvasive markers for fibrosis, 55,56 we assumed that patients would still be candidates for HCC surveillance even if they had fibrosis regression. However, it is possible that the risk of HCC could decrease with fibrosis regression as time from SVR accrues; therefore, our study may have overestimated the burden of HCC surveillance.

Conclusions
This decision analytical model study suggests that the new cases of HCC associated with HCV will continue to increase even in the era of DAAs. We also found that the risk of HCC as well as the burden of HCC surveillance is projected to shift from patients with active HCV to those who had achieved virological cure. Routine HCC surveillance in patients with virologically cured HCV who remain at risk of HCC may help reduce the burden of HCC in the era of new antiviral agents.