Screening for Hepatocellular Carcinoma in Patients with Hepatitis B

Chronic hepatitis B (CHB) infection is a significant risk factor for developing hepatocellular carcinoma (HCC). As HCC is associated with significant morbidity and mortality, screening patients with CHB at a high risk for HCC is recommended in an attempt to improve these outcomes. However, the screening recommendations on who to screen and how often are not uniform. Identifying patients at the highest risk of HCC would allow for the best use of health resources. In this review, we evaluate the literature on screening patients with CHB for HCC, strategies for optimizing adherence to screening, and potential risk stratification tools to identify patients with CHB at a high risk of developing HCC.


Introduction
Liver cancer is a leading cause of morbidity and mortality on a global scale, with approximately 30,000 deaths expected annually in the United States in 2020 [1], and significantly increased incidence rates modelled by 2030 world-wide [2]. The most common form of liver cancer is hepatocellular carcinoma (HCC), which accounts for approximately 75-85% of liver cancer cases [3]. HCC is within the seven most common cancers worldwide, and the third largest cause of cancer-induced mortality [4]. The leading cause of HCC worldwide is chronic hepatitis B virus (CHB) infection, which affects more than 250 million people world-wide [5]. Although every patient with CHB is at an elevated risk of developing HCC, this does not mean that each patient is at an equivalent risk. As demonstrated by the COVID-19 pandemic, surveillance programs can be easily disrupted. However, effective models of risk stratification can ensure that remaining resources are optimally allocated to the prioritization and prevention for those at the highest risk until the system returns to full capacity [6]. Several scoring systems have been created in an attempt to optimize the distribution of patients into different risk profiles, based on patient characteristics, along with viral characteristics [7]. The heterogeneity of the global CHB induced HCC population has made it difficult to develop a unified risk-scoring system, with differences among methodologies in terms of how these scoring systems identify, score, and evaluate relevant variables. This review aims to provide a comprehensive summary for the clinician, regarding screening for HCC in patients with CHB. Topics covered include reviewing the epidemiology of HCC in CHB, the state of research regarding screening patients with CHB for HCC, a summary of current guidelines for screening, highlighting the key literature The rationale for any screening strategy is to identify patients at earlier stages of disease where treatment has a higher likelihood of success, as well as to identify risk factors that increase the risk of developing the disease, as to allow for interventions [17]. Factors to consider in the evaluation of any screening strategy are whether it is a disease of concern, and whether the screening test is effective, reasonable in cost, and improves outcomes [17,18]. Screening for HCC clearly meets these criteria, given that HCC is a public health concern, leading to a significant loss of life years and quality of life [19].
Currently, an ultrasound with the possibility of adding alpha-fetoprotein are recommended as the screening strategies for HCC [20][21][22][23][24]. At this time, there are only three randomized clinical trials comparing HCC surveillance versus no surveillance, with conflicting results related to methodological challenges. The first trial studied patients from Shanghai that were employed in factories, schools or private enterprises, with CHB or chronic liver disease; they were allocated by cluster sampling into those who received AFP and an ultrasound every 6 months (n = 8109) and a control group (n = 9711) enrolled between 1992-1994 [25]. In screened patients, 38 patients with liver cancer were identified after 12,038 person-years, and in the control group 18 patients were diagnosed after 9573 person-years. Patients identified in the screening group had a 1-year survival of 88.1% and two-year survival of 77.5%, and HCC were found at an earlier stage than the control group; the control group had a 0% 1-year survival. Although this study estimated an average cost of $1500 USD per early-stage HCC diagnosis, the number of false positive screening results was not reported, and would likely have increased the procedural and administrative cost of executing this program.
A similar study was conducted through 1993-1995 in Shanghai, consisting of 18,816 patients aged 35-59 with chronic hepatitis B. Participants were randomly allocated to screening (9373) with alpha-fetoprotein and ultrasounds every 6 months, versus a control  [26]. The surveillance arm identified 86 cases with 32 deaths from HCC, and the control arm had 67 cases with 54 deaths, leading to a risk ratio of 0.63 (95% CI 0.41-0.98) for the screening arm. Notably, this trial has been criticized for a lack of information about each group's baseline characteristics, and how the outcome of death from liver cancer was defined [27]. Furthermore, the cohort likely has significant overlap with the last study by Yang et al. [25], given the population characteristics and time frame of recruitment.
The third randomized trial focused on men aged 30-69 between 1989-1995 in Jiangsu Province, China, using alpha-fetoprotein (AFP) levels every 6 months (n = 3712, mean follow-up 61.9 months) with a threshold of 20 µg/L, versus a control group (n = 1869, mean follow-up 62.8 months) [28]. Adherence rates for obtaining AFP were low at about 30%. A significantly larger number of cases were identified in the screening arm vs. the control arm at early stages (BCLC Stage 0-A [29]), with early detection rates of 29.6% and 6.0%, respectively. The 5-year survival between the groups did not differ at 4%, and the risk ratio in the screening arm was not statistically significant (RR 0.83; 95% CI 0.68-1.03), which may be related to the limited antiviral therapy available at the time of the study. The study has been criticized for the unclear randomization and allocation concealment techniques that were used [27].
A meta-analysis from 2014 which included the second and third trial, as well as 18 observational studies examining the role of screening in patients with chronic liver disease, concluded that it was uncertain whether systematic surveillance improved patient survival, and highlighted that more data were required given the relatively few studies on the CHB population [27]. A second meta-analysis published by Singal et al. [30] looking at patients with cirrhosis, suggested that the use of ultrasound was associated with a higher probability of diagnosing cancers at an earlier stage (OR 2.08, 95% CI 1.80-2.37) and improved survival (OR 1.90; 95% CI 1.67-2.17). These differences remained significant when accounting for lead time bias; however, only three publications looked at HBV patients in particular, with no subgroup analysis performed.
More recent epidemiological studies have shown potential survival benefits of HCC surveillance. One Canadian trial examined outcomes in patients with viral hepatitis diagnosed with liver cancer, and the impact of ultrasound screening prior to their diagnosis [31]. The 5-year survival in patients receiving routine surveillance (at least an ultrasound annually) was 31.93% (95% CI 25.77-38.24%) versus 20.67% (95% CI 16.86-24.74%) in patients who were not screened. Moreover, surveillance was associated with a mortality risk of 0.76 (95% CI 0.64-0.91), which indirectly supports the role of screening. Similarly, in patients with all-cause cirrhosis, screening for HCC has been suggested to have potential survival benefits [32][33][34]; although, a case-control study of the US Veterans Affairs cohort did not show a survival difference between those screened with ultrasounds and those who were not [35]. Current recommendations from the US National Cancer Institute suggest that screening patients at an increased risk does not lead to a decrease in mortality from HCC, as shown by fair quality evidence that is based on lead and length time bias [36].
Ideally, a randomized clinical trial in an attempt to best answer this question would be useful; however, this is unlikely to happen, given ethical concerns about the harm associated with no surveillance, as patients and practitioners prefer surveillance [37]. Further, answers to the true benefit of screening through cohort/case-control trials focusing on the role of screening patients with CHB will be important, given the limited high-quality literature to date. One potential barrier to a universal consensus regarding screening is the variation in efficacy of screening between regions. This is due to internal population factors, such as the prevalence of CHB and risk of tumor development based on prognostic factors of disease severity, in addition to health care resources available in the region. For example, although a male cohort between the ages of 70-79 may be at the highest risk of developing HCC, Japanese, African, and Chinese populations are skewed as significantly younger [21]. These differences in epidemiology are mirrored by regional differences in post-diagnosis outcomes [38], dependent on factors, such as availability of therapeutics and patient health status. Consequently, any analysis regarding the impact of surveillance may be limited to regional applicability, and if an RCT is conducted in the future it would be best to involve multiple regions, in an effort to create a generalizable scope of study.

Current Society Guidelines for HCC Screening
Currently, it is recommended by all major hepatology international societies to screen for liver cancer in patients with CHB deemed to be high risk with the use of abdominal ultrasound every 6 months, with variable recommendations regarding the inclusion of AFP. Although these guidelines are largely similar, there are some minor differences, i.e., the use of AFP and when to screen patients from an African background. We summarize the current society guidelines in Table 1. AFP is a glycoprotein produced primarily during the ontogenesis of the yolk sack and fetal liver, with a suppressed expression following birth. However, increased AFP production has been associated with reparative growth following liver damage [39], as well as oncogenesis, acting as a pro-proliferative protein involved in regulating apoptosis, growth, and angiogenesis [40]. Due to its association with abnormal hepatocyte proliferation, AFP elevations are well documented in both malignant liver disease [40] and nonmalignant liver diseases, such as acute hepatic failure [41] and CHB [42]. Unfortunately, the use of AFP as a screening tool remains unclear. A recent meta-analysis evaluated the effectiveness of screening with ultrasound versus using ultrasound and AFP in detecting early cancer [43]. Thirty-two studies were identified in this study (consisting of 13,367 patients). Ultrasound identified any stage of HCC with 84% sensitivity (95% CI 76-92%), while identifying early-stage HCC with 47% sensitivity (95% CI 33-61%). The addition of AFP to ultrasound increased the sensitivity to 63% (95% CI 48-75%) for early-stage HCC, and to 97% (95% CI 91-99%) for all stages of cancer. Subsequently, the addition of AFP reduced the specificity to 84% (95% CI 77-89%) from 92% (95% CI 85-96%). There were some limitations to this meta-analysis, as it had no studies that were looking at survival. Moreover, most studies only looked at HCC at any stage vs. early HCC, which may overestimate surveillance test performance. Furthermore, negative screening tests were not confirmed by another modality, which may lead to verification bias. Overall, it is critical to note that using AFP by itself is not recommended, as it has poor sensitivity and specificity [44].

Asian Pacific Association for the Study of the Liver (APASL)
The 2017 APASL guidelines [24] recommend abdominal ultrasounds with AFP every 6 months for HCC screening in patients with CHB cirrhosis in Asian females > 50 years, Asian males > 40 years, patients of African background > 20 years, and in patients with a Viruses 2021, 13, 1318 5 of 16 family history of HCC (no specific starting age recommended) ( Table 1). The AFP threshold recommended is 200 ng/mL to obtain the optimal positive likelihood ratio.

American Association for the Study of Liver Diseases (AASLD)
The 2018 AASLD guidelines [22,23] recommend abdominal ultrasounds every 6 months with or without AFP. The latter is a significant change from the 2010 guidelines [45], which explicitly recommended against AFP, due to concerns with false positives and subsequent testing. The AFP threshold recommended by the AASLD is 20 ng/mL. High risk groups that are recommended for HCC screening include patients with cirrhosis secondary to CHB with either Child-Pugh A or B; Child-Pugh C patients who are non-transplant candidates are excluded, as treatment would likely not be feasible. Other patients recommended for screening include Asian females > 50 years, Asian males > 40 years, patients of African background > 40 years, patients with a first-degree relative with a history of HCC (no specific starting age recommended), and patients co-infected with hepatitis delta (HDV) (no specific starting age recommended) ( Table 1).

European Association for the Study of the Liver (EASL)
The 2018 EASL guidelines [21] advocate screening high risk patients with abdominal ultrasounds every 6 months, and do not recommend using AFP routinely, due to concerns regarding false-positive results in the context of active liver inflammation. Patients recommended for screening include those with Child-Pugh A or B cirrhosis; Child-Pugh C cirrhosis awaiting liver transplantation; and patients with chronic hepatitis B determined to be at an intermediate or high risk of developing HCC, based on PAGE-B classes for those of Caucasian background (Table 1).

Canadian Association for the Study of the Liver (CASL)
The 2019 CASL guidelines [20] recommend performing abdominal ultrasounds every 6 months for patients deemed to be at a high risk for developing HCC. This includes patients with CHB cirrhosis, Asian females > 50 years, Asian males > 40 years, patients of African background > 20 years, and patients with a family history of hepatocellular carcinoma starting at age 40, as well as patients with HIV co-infection > 40 years (Table 1).

Guideline Based Approach to Co-Infection
All four guidelines discussed above highlight the potential impact of co-infection with hepatitis C virus (HCV), HDV or HIV as a risk factor for developing HCC, although each set of guidelines emphasizes a different viral co-infection. The data for the impact of co-infection on its development is strongest in HDV [46,47], with data for HCV [48] and HIV [49,50] being weaker. Although there may be an increased risk with co-infection, there is little evidence regarding how screening guidelines should be changed. This likely explains the lack of explicit comment throughout guidelines, regarding the impact of co-infection and HCC screening; this is an important area for further research. That being said, it is important to recognize that both patients with mono-infection and significant co-infection undergo the same HCC surveillance protocol under the previously mentioned guidelines, suggesting the primary driver for surveillance in this situation is the oncogenic potential of CHB.

Cost Effectiveness of Screening Strategies for HCC
A key component of any screening strategy is ensuring it offers good value for money; that is, the strategy is cost effective. The majority of studies evaluating the cost-effectiveness of screening consider patients with cirrhosis, with a minority of studies looking at hepatitis B being mostly focused on Asian populations [51]. In studies comparing screening using abdominal ultrasound +/− AFP with no screening [52][53][54][55][56], screening programs were cost effective in all but one study [55]. There are two recent studies evaluating the cost effectiveness of imaging-based screening, as compared to no screening at all. A Taiwanese based study evaluated the cost effectiveness of screening patients with CHB with ultrasounds every 6 months versus no ultrasounds [56]. The base case was a 50-year-old individual who was followed over a 25year time horizon. Screening had a cost of $5912.37 USD for 13.78 years of life, as compared to $557.10 for 13.53 years of life in the unscreened arm, working out to an incremental cost effectiveness ratio of $20,856.25/year of life gained. Sangmala and colleagues expanded this analysis, and evaluated multiple screening strategies in a Thailand population of patients with CHB between the ages of 40-60 with a lifetime time horizon [52]. The strategies evaluated in this model included abdominal ultrasound, AFP and US, computed tomography (CT), and magnetic resonance imaging (MRI) at either a 6-monthly or 12-monthly frequency. Without any form of screening, only 6.24 quality adjusted life years (QALY) were gained. The use of ultrasound or ultrasound and AFP for HCC screening were found to be the most cost effective strategies. Notably, CT and MRI based strategies were not cost effective.
A recent analysis evaluating the cost effectiveness of screening compares abdominal ultrasound with or without AFP for HCC screening. This analysis is novel in its consideration of the potential harm of unnecessary tests, due to false positives [57], which is not considered in previous studies. The cohort studied were patients with compensated cirrhosis over a lifetime horizon. The most cost effective strategy for the cohort was using US with AFP, with a cost of $1,254,173.20 USD for 6.02 QALY; both using US and no surveillance in this cohort was more expensive and less effective than US with AFP. Modelling the impact of false positive testing in the CHB population is an important area to evaluate, until then, extrapolation from the available evidence is required.

Challenges with Screening Adherence
Adherence by patients to screening programs is a key indicator of success. Unfortunately, a major challenge with HCC surveillance programs is that of poor adherence rates. A study from Washington State looking at 1137 patients with cirrhosis highlights this challenge. Over a 2-year period, 33% of this group had at least one ultrasound (intermittent surveillance), and those with CHB cirrhosis were found to be more likely to undergo surveillance. Notably, only 2% of patients in this study underwent consistent surveillance, defined as having an ultrasound every 6 months [32]. A recent meta-analysis looking at screening for patients at a high risk of developing HCC identified 22 studies involving 19,511 patients, showing an overall adherence rate of 52% (95% CI 38-66%). Looking at the subset of studies of patients with CHB (4 trials, 2651 patients), the adherence rate was significantly lower (32%, 95% CI 13-51%). A subsequent meta-analysis focusing on patients with cirrhosis identified 29 studies (n = 118,779 patients) with a surveillance rate of 24% (95% CI 18.4-30.1); however, higher surveillance rates were observed when patients were seen by gastroenterologists and/or hepatologists [58].
Screening programs are complex, and there are many reasons why lower adherence rates to screening are observed. Simply, there are system, provider, and patient-based factors, as outlined in Figure 1. Physician failure to order HCC surveillance was the most common reason for patients not receiving surveillance in two studies [59,60], with a separate investigation of physician surveillance practices finding only 22% relied on biannual imaging for HCC surveillance [61]. However, patient-driven non-adherence remains a significant determinant of surveillance inconsistency, with Singal et al. estimating patient non-adherence accounted for <10% of adherence failures [62] in their study. Patient factors can be further divided into those that are related to feasibility (i.e., cost, insurance coverage [62], difficulty navigating system, transportation [63]) and patient perceptions (i.e., knowledge regarding need for surveillance, HCC presentation, management [63]). Both patient and provider factors are further exacerbated by systemic flaws, with issues related to supporting surveillance infrastructure and clinical care network [64], identified by existing studies.
can be further divided into those that are related to feasibility (i.e., cost, insurance coverage [62], difficulty navigating system, transportation [63]) and patient perceptions (i.e., knowledge regarding need for surveillance, HCC presentation, management [63]). Both patient and provider factors are further exacerbated by systemic flaws, with issues related to supporting surveillance infrastructure and clinical care network [64], identified by existing studies.

Methods to Improve Adherence
A number of strategies have been studied in an attempt to improve adherence to HCC screening, including the education of primary care providers, nurse led clinics, mailed outreach, and radiology led programs [58]. Although all of these techniques have merit, the impact of these strategies are variable in efficacy. There have been four studies utilizing a post-intervention outcome of the proportion of patients meeting the goldstandard of having ultrasounds every 6 months. A randomized control trial evaluating outreach letters by mail showed that the rate increased from a baseline of 7% to 21% in the intervention arm [67]. Moreover, systematic changes may be associated with higher

Methods to Improve Adherence
A number of strategies have been studied in an attempt to improve adherence to HCC screening, including the education of primary care providers, nurse led clinics, mailed outreach, and radiology led programs [58]. Although all of these techniques have merit, the impact of these strategies are variable in efficacy. There have been four studies utilizing a post-intervention outcome of the proportion of patients meeting the gold-standard of having ultrasounds every 6 months. A randomized control trial evaluating outreach letters by mail showed that the rate increased from a baseline of 7% to 21% in the intervention arm [67]. Moreover, systematic changes may be associated with higher rates of surveillance, although likely coming at a higher cost. For example, in Australia, the introduction of a nurse-led clinic led to 53% of patients seen in the clinic having appropriate ultrasounds [68], while a more intense system redesign and patient education program led to 63% of patients being screened [69]. From the radiology provider side, a radiology recall program in the United Kingdom [70] led to 46% of patients achieving appropriate screening. There is no one-size solution, but likely a structured intervention whereby the payer, patient, provider, and system will be important to increase the adherence rate of screening.

Risk Stratification Systems for HCC Risk in Patients with HBV
When comparing the various models of risk assessment, there appears to be some common areas of contention, which are addressed differently by different models. Risk stratification scores for HCC risk in HBV have been previously reviewed [71]. Here, in Tables 2 and 3, we present an updated critical summary of all published models of HCC risk assessment in patients with CHB. The majority of studies assessed in this review involve risk models, generated using homogenous populations. Although this is a result of the relatively homogenous populations in some of the countries in which these studies were generated (i.e., South Korea, Taiwan), this limits the utility of these models as international standards for HCC surveillance. In comparison, models generated in western countries, such as PAGE-B, have an advantage in terms of generalizability, as their study population for CHB often has a greater degree of ethnic diversity. However, as the number of models grows, newer studies such as the aMAP have been able to test their model in multiple populations of different ethnicities prior to publication, providing further support for the generalizability of their model. Although many studies report using gender as a risk factor, it is most likely that gender is incorrectly used, and this should be biological sex rather than the social concept of gender [72]; acknowledging this, we report the criteria as provided by the original publication. Ultimately, the majority of risk assessment models were reliant on follow-up studies, validating them in other population groups to strengthen their argument for generalizability.   Another area of difference between risk assessment models is the use of a three-tier vs. two-tier stratification of patient risk of developing HCC. While three-tier models are more likely to be statistically accurate than two-tier models, given they allow for a greater degree of objectivity over the data being analyzed, they are limited in their potential utility. Protocols in practice for patients with CHB [107] essentially culminate in a binary decision, and it must be determined whether the patient would be included in a screening protocol or not. Ultimately, three-tier models may have limited clinical applicability, as many of the systems in place for screening patients are not designed to have separate protocols for patients who are "high risk" vs. those categorized as "medium risk". Thus, having a paralleled stratification of outcome and model is beneficial for ease of use, and mitigates confusion on how patients should be stratified into surveillance vs. screening.
A limitation of most existing risk scores is a lack of discrimination between the stages of fibrosis beyond a diagnosis of overt cirrhosis. Studies have demonstrated a similar 5-year HCC probability for patients with a histopathological diagnosis of cirrhosis and those with a diagnosis of advanced fibrosis [108], elucidating the need for a more nuanced approach to fibrosis when developing a risk score. An improved assessment of the degree of fibrosis may help improve the cost effectiveness of screening strategies; there are a number of non-invasive techniques for fibrosis evaluation, which have been reviewed in recent guidelines [109]. Further, antiviral-treated patients require accurate staging of fibrosis for updating their risk of developing HCC, as patients with an appropriate treatment response may present with a regression of fibrosis, a positive prognostic indicator for HCC risk [110].
A recent study suggested the utility of a model combining AFP and the FIB-4 model, to predict HCC outcomes for patients with compensated cirrhosis, due to CHB being treated with antiviral therapy [111]. However, the novel approach was not compared to existing HCC-CHB risk stratification models, and the reliability of FIB-4 for non-Asian patients with CHB has also been called into question [112]. To appropriately utilize a FIB-4-like system for CHB, further adaptation may be necessary, with novel modifications being capable of identifying fibrosis, and stratifying HCC risk in the CHB population currently being investigated [112,113].

Conclusions
Although screening for HCC in high-risk patients with CHB is considered the standard of care and has proven benefits, there remains significant gaps in the knowledge base. Outstanding questions include whether alpha-fetoprotein should be a part of the screening program, as well as the risk stratification of patients with CHB in order to tailor screening programs to those with the highest risk of HCC. From a practical standpoint, investigations on how to develop structured interventions involving the payer, patient, and system to improve adherence to screening programs will be needed to maximize the effectiveness of any of these programs. Ultimately, surveillance programs are necessary for the effective identification of early-stage cancer in high-risk patients. This paper identifies potential areas of improvement to further improve their impact and accuracy in a clinical setting.