Declining incidence of hepatitis C related hepatocellular carcinoma in the era of interferon‐free therapies: A population‐based cohort study

The impact of interferon (IFN)‐free therapies on the epidemiology of hepatitis C virus (HCV) related hepatocellular carcinoma (HCC) is not well understood at a population level. Our goal was to bridge this evidence gap.


| INTRODUC TI ON
Hepatocellular carcinoma (HCC) is a major cause of excess mortality in patients with hepatitis C virus (HCV) infection. From an epidemiological perspective, the number of HCC deaths occurring in a given time period depends on four main factors: • Size of the 'at risk' population (For HCC, the major 'at risk' population are people with liver cirrhosis-ie the prevalence of cirrhosis in HCC patients exceeds 80%. 1 ) • Cumulative incidence of HCC among the 'at risk' population (Previous studies suggest the incidence of HCC among patients with HCV-related cirrhosis varies from 1-3 events per 100 person years of follow-up. [2][3][4][5][6] • Uptake of HCC curative therapy following an HCC diagnosis (Receiving curative HCC therapy is associated with improved survival, 7 but only patients with early stage HCC are eligible 1,8 ).
• Overall survival following HCC diagnosis (HCC is a leading cause of cancer mortality 9 ; Average survival after an HCC diagnosis is less than 1 year. 10 ) The arrival of interferon-free (IFN-free) direct acting antivirals (DAAs) from 2014 has not only revolutionised the treatment of chronic HCV infection but also offers new public health opportunities. 11 Whereas older treatments were long, arduous and ineffective, IFN-free therapies are short, tolerable and highly effective. 12,13 It is likely that IFN-free DAAs have already impacted the epidemiological landscape for HCV-related HCC. However, few studies have been able to quantify these changes at a population level, and so the picture remains uncertain. For example, the number of cirrhosis patients with cured HCV has probably increased dramatically in recent years, but precise numbers are not yet known. Similarly, although studies have shown that HCV cure is associated with reduced HCC incidence for cirrhosis patients, 6,14 this does not necessarily mean that HCC incidence is falling at a population level. The situation is complicated because HCV cure is also associated with a reduced risk of competing risk events (i.e. events that preclude the occurrence of HCC, such as death from decompensated cirrhosis), which may mitigate the population impact for HCC. By the same token, there is evidence that HCV cure is associated with improved post-HCC survival 15 The main obstacle to carrying out this analysis is that it requires comprehensive and linkable health registries in order to follow patients from cirrhosis diagnosis through to death. Historical data for patients diagnosed in the era of IFN-based therapies are also essential. Scotland is one of the few settings anywhere in the world with the requisite database and linkage infrastructure to carry out this type of analysis. Thus, the main aim of this study was to determine the trends over a twenty-year period (1999 to 2019) in the aforementioned domains for HCV patients in Scotland.

| Study design and primary objective
We conducted a population-based cohort study to assess trends over time in relation to; (a) the number of individuals living with HCV cirrhosis, (b) HCC incidence, (c) receipt of HCC curative treatment, and (d) post-HCC mortality risk. Secular trends were assessed using descriptive statistics and through competing risk multivariable regression models.
whereas the number of incident HCCs was four times higher. However, the cumulative incidence of HCC was significantly lower in the IFN-free versus PIR era (sdHR: 0.65; 95%CI:0.47-0.88; P = .006). Among HCC patients, diagnosis in the IFN-free era was not associated with improved uptake of curative treatment (aOR:1.18; 95%CI:0.69-2.01;

| Data Sources
We used data from four national health registries from Scotland to assess trends in HCV-related HCC over a period of 20 years (i.e. 1999 to 2019).

| Inclusion/exclusion criteria
All patients in Scotland with chronic HCV who received a diagnosis of cirrhosis between 1999 and 2019 were included in this study (N = 3350).
Cirrhosis was defined as a diagnosis of compensated or decompensated cirrhosis made during the course of routine clinical investigation and follow-up at a specialist liver clinic. Typically, cirrhosis is diagnosed through a combination of liver biopsy; transient elastography; abdominal ultrasound; clinical examination; and routine liver function tests. The earliest date of cirrhosis diagnosis was obtained from the Scottish HCV clinical database.
Patients were excluded if HCC occurred before cirrhosis diagnosis, which we defined as an HCC diagnosis more than one year prior to cirrhosis diagnosis (N = 3). Thus, there were 3347 cirrhosis patients in our study population after applying the aforementioned exclusion criteria (see Figure 1).

| Outcome event definitions
Incident cases of HCC were identified through the presence of an ICD code for HCC (i.e. ICD10: C22.0; ICD9:155.0) in either a cancer register, hospital admission or mortality record. Hospital admission and mortality records were only used to define HCC if they included an ICD code for HCC in the primary diagnostic/ cause of death position. The date of HCC incidence was defined as the earliest HCC diagnosis/presentation date from across these three registries. All HCC presentations through to 31 Dec 2019 were counted.  Table S1).
Deaths from HCC were identified by the presence of an HCC ICD code (ICD10: C22.0; ICD9:155.0) in any cause of death position.

| Definition of study covariates
The study outcome events were compared across three time periods, with each period corresponding to a distinct era of antivi-

HCC incidence
We calculated the number of cirrhosis patients presenting with HCC in each calendar year according to chronic HCV infection status.
Chronic HCV infection status at HCC diagnosis was inferred on the basis of whether the patient had achieved SVR prior to the HCC diagnosis date.

Curative treatment
We calculated the number of HCC patients who received curative HCC therapy. Again, we broke these numbers down accord-

HCC mortality
The number of deaths related to HCC was also counted according to the calendar year of death and SVR status. Calendar year corresponds to the year of HCC death, whilst HCV infection status relates to the infection status at the time of HCC diagnosis.

| Regression analyses
A range of multivariable regression models were fitted to assess time trends in relation to (a) HCC incidence; (b) receipt of curative therapy, and (c) post-HCC mortality risk. All models included statistical adjustment for relevant covariates to reduce bias from potential confounding.

HCC incidence
We performed time-to-event survival analysis to assess change in HCC incidence for cirrhosis patients over time, adjusting for relevant factors.
Time zero was defined as the date of first cirrhosis diagnosis.
Patients were then followed up through to either the date of incident HCC (if at all), date of mortality (if at all) or the study completion date (31st Dec 2019). However, in a small number of patients, the date of cirrhosis diagnosis predated first appointment at an HCV specialist liver clinic. In these cases, survival time was left truncated (i.e. delayed entry) to avoid the potential introduction of immortal time bias between diagnosis of cirrhosis and the eventual first clinic appointment. Delayed entry means that whilst susceptibility to HCC begins from the date of cirrhosis diagnosis, only follow-up time occurring after the date of first appointment was counted in this analysis.
Fine-Gray regression, accounting for non-HCC mortality as a competing risk, was used to assess the association between time period and time to first HCC presentation. We included adjustment for age group, gender, decompensated cirrhosis, history of heavy alcohol use, SVR achievement, HCV genotype, IDU history, and deprivation status. Time period, SVR achievement, age, and decompensated cirrhosis were modelled as time-dependent variables-meaning they were able to change/update during the course of follow-up.
Two multivariate models were fitted to assess if the association between time period and HCC incidence is mediated by SVR achievement: one model including SVR achievement as a covariate and one omitting SVR achievement as a covariate.
We also performed sensitivity analyses where standard Cox regression models were fitted, in which competing risks are treated as censored observations. 17 In this time-to-event analysis, we used the 'time window' method 18 to exclude all follow-up occurring in first six months since time zero. This was done to remove patients with 'prevalent' HCC (i.e. HCC already present at time zero) as opposed to incident HCC (ie HCC first emerging after time zero).
The cumulative incidence of HCC at specific time points was estimated using 'Stcomlist' command 19 in Stata version 17. Delayed entry patients were omitted from these calculations.

Curative treatment
Logistic regression was used to identify factors associated with receiving curative therapy among those with incident HCC. The primary exposure variable of interest was time period, based on the year of HCC diagnosis. Covariates considered were: age at HCC diagnosis; history of heavy alcohol use; gender; SVR achievement, decompensated cirrhosis and SIMD deprivation quintile.
As with the HCC incidence analysis, models including and omitting SVR as a covariate were fitted to assess if SVR mediates the association between time period and HCC treatment uptake.
We also performed a sensitivity analysis where individuals diagnosed with HCC within 18 months of the study completion date of 31st Dec 2019 were excluded. This was to assess if recently diagnosed patients -who may not have had sufficient time to access therapy before the study completion data -could be biasing time period trends.

HCC-related mortality
We performed time-to-event survival analysis to assess if time period was associated with risk of HCC mortality after HCC diagnosis.

| Study population
A total of 3347 cirrhosis patients were included in our final study population (see Figure 1). The average-mean age at cirrhosis diagnosis was 47.6 years, and three-quarters were male (74.8%). Most of the cohort were living in areas of high deprivation (i.e. 53.8% in SIMD quintile 1) and just under half were diagnosed with cirrhosis in the IFN-free era (i.e. 45%) ( Table 1).

| HCC incidence
The number of incident HCC cases also increased with time, from an average of 8 per year in the pegylated interferon and ribavirin era to 29 per year in the first-generation DAA era and to 34 per year in the IFN-free era. The proportion of HCC cases with SVR increased from 10.2% SVR in the pegylated interferon and ribavirin era; to 10.2% in the DAA first-generation era, to 53.0% in the IFN-free era ( Table 2 and Figure 2).

| Curative treatment
The number of HCC patients receiving curative therapy changed from an average of 3 per year in the pegylated interferon and ribavirin era to 13 per year in the first-generation DAA era and to 11 per year in the IFN-free era. The proportion of treated patients with an SVR was 12.7% in the pegylated interferon-ribavirin era, 10.8% in the first-generation DAA era, and 54.3% in the IFN-free era ( Table 2 and Figure 2).

| HCC-related mortality
The number of HCCs death increased from an average of 4 per year in the pegylated interferon-ribavirin era to 12 per year in the firstgeneration DAA era and to 22 per year in the IFN-free era. The proportion of patients dying with HCC who had achieved SVR was 6.5% in the pegylated interferon era, 1.3% in the first DAA generation era, and 34.7% in the IFN-free era (Table 2 and Figure 2).

| HCC incidence
A total of 3201 patients were included in the HCC incidence analysis after removing 146 individuals with less than six months follow-up (see previous explanation of 'time window' method). Of these 3201, survival time was left truncated (i.e. delayed entry) for 165 (5.2%) individuals (see Table S2).   were not materially altered in sensitivity analyses (Table S4).

| Post hoc analysis
As previously described, the reduction in the cumulative incidence of HCC observed in the IFN-free era was only partly attenuated when adjusting for SVR achievement (ie from sdHR:0.67 to sdHR:  (Table S5).

| DISCUSS ION
In this study, we show that the epidemiological landscape for HCVrelated HCC has changed markedly during the last twenty years. to be screened to identify a single HCCs) is higher now than ever before. These data, therefore, provide important context to the current debate around the need for a more targeted approach to HCC surveillance. 21,22 Our analysis also sheds new light on outcomes downstream of  HCC-related mortality risk in the IFN-free era (albeit this was not statistically significant; P = .09). This putative survival improvement may be a corollary of the higher SVR prevalence, given that SVR itself was associated with a lower HCC mortality risk in our analysis (as has also been noted elsewhere 15,16 ).
Finally, our results suggest a mixed picture with respect to the influence of deprivation on HCC outcomes. There was no indication that deprivation was associated with a higher HCC incidence among cirrhosis patients (indeed, if anything, the reverse was true).
However, both uptake of curative treatment and survival appeared to be adversely affected by higher levels of deprivation ( Figure S1).
Overall, our analysis is consistent with many strands of prior research. For example, like other studies 6,14 we show that an HCV cure is associated with decreased HCC incidence compared to untreated patients and treatment failures. We also show that an HCV cure may be associated with improved survival after HCC, which has also been reported elsewhere. 15,16 This study also chimes with previous population-based studies showing a reduction in all-cause mortality, 23 incident decompensated cirrhosis 24 and HCC mortality 25 in the IFN-free era. However, our finding of a reduced HCC incidence in the IFN-free era is novel and to the best of our knowledge has not been shown before. Therefore, this study will lend further confidence to patients, clinicians and policy makers regarding the utility of IFN-free therapies. At the same time, it also highlights the need now to focus on increasing curative treatment uptake for the full impact of IFN-free therapies to be realised. In conclusion, the epidemiological landscape for HCV-related HCC has changed considerably over the last two decades, particularly with respect to the number of diagnosed cirrhosis patients and the cumulative incidence of HCC. These changes have important implications for the implementation and sustainability of existing HCC screening guidelines.

ACK N OWLED G EM ENTS
None.

CO N FLI C T S O F I NTE R E S T
There are no conflicts of interest to disclose.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data used in this study are not publicly available, but can be acquired through a successful application to the Public Benefit and