Comparing validated hepatocellular carcinoma risk scores for chronic hepatitis B against current Australian surveillance guidelines

Abstract

BACKGROUND

Chronic hepatitis B (CHB) patients are at increased risk of developing hepatocellular carcinoma (HCC), however risk varies greatly between individuals. The Gastroenterological Society of Australia (GESA) consensus guidelines were developed to ensure patients receive appropriate surveillance; however, these guidelines do not account for other established risk factors. Various HCC risk scoring systems have been developed, including the Risk Estimation for HCC in CHB (REACH-B), modified REACH-B (mREACH-B) and Platelet Age Gender-Hepatitis B (PAGE-B) scores which have been validated in specific ethnic populations.

AIM

To evaluate differences in HCC surveillance patterns between current GESA recommendations and risk scoring tools for non-cirrhotic CHB patients.

METHODS

A prospectively collected health registry identified all patients with CHB from a regional tertiary Australian hospital in South East Queensland from 2019 to 2022. REACH-B, mREACH-B and PAGE-B scores were calculated for non-cirrhotic CHB patients. Surveillance patterns between scoring systems and current guidelines were compared.

RESULTS

A total of 314 patients were included. 35.0% (n = 110) were male, with a median age of 50.6 years. 69.7% of patients originated from Asia, 9.6% had a positive hepatitis B e-antigen status, and 37.9% were on antiviral therapy. Only 2 (0.6%) patients had progression to HCC. Using the GESA guidelines, 57% of patients qualified for HCC surveillance. However, when applying REACH-B, mREACH-B and PAGE-B scores, surveillance was recommended for 23.2%, 12.4% and 46.8% of patients respectively. This represented marked discordance between guideline-based and score-based surveillance eligibility. Notably, none of the scores identified the two patients who developed HCC as candidates for surveillance. All scores demonstrated poor discriminatory performance, with an area under the receiver operating characteristic curve of 0.53 for REACH-B, 0.55 for mREACH-B, and 0.39 for PAGE-B.

CONCLUSION

REACH-B, mREACH-B, and PAGE-B offer alternative options for risk stratification. However, this study suggests that the accuracy of these risk scores when applied to the Australian population may be limited, likely due to local virological, socio-demographic and lifestyle related factors which need to be taken into account. In heterogenous populations, these risk scores should be used cautiously and alongside individualized clinical assessment rather than as standalone decision tools.

Key Words: Chronic hepatitis B; Hepatocellular carcinoma surveillance; Risk prediction models; Australian population; Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B; Platelet Age Gender-Hepatitis B; Modified Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B

Core Tip: Hepatocellular carcinoma (HCC) surveillance in people with chronic hepatitis B (CHB) aims to detect cancer early, however, there is a risk of over- or under-surveillance. In this Australian cohort, the validated HCC risk scores Risk Estimation for HCC in CHB, modified Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B, and Platelet Age Gender-Hepatitis B identified substantially fewer patients for surveillance compared with the current Gastroenterological Society of Australia guidelines. Importantly, these risk scores failed to identify all patients who developed HCC, highlighting the limitations of applying population-specific tools to Australia’s heterogeneous CHB population.

INTRODUCTION

Chronic hepatitis B (CHB) is a global public health concern with significant morbidity and mortality[1,2]. It affects nearly 1% of the Australian population, of which approximately half were born in the Asia-Pacific region[3]. Overall, its prevalence since discovery in 1967 has been reducing in endemic regions through the use of vaccination programmes, targeted global health policies, and through antiviral treatment[2,4]. The hepatitis B virus (HBV) causes a variable degree of clinical manifestations. If left unmanaged, CHB may progress to cirrhosis or hepatocellular carcinoma (HCC) which place a significant burden on the healthcare system, with an estimated annual cost in 2012 of 50.2 million Australian dollars for patients admitted to hospital[5]. Individuals with CHB have a 10-fold to 25-fold lifetime risk of developing HCC compared to those without the virus, and hepatitis B contributes to half of all HCC cases globally[1]. In the absence of cirrhosis, the annual incidence of HCC development is 1% in people with CHB, with this risk rising to 2%-3% if cirrhosis is present[6]. Through the use of surveillance programmes, early detection of HCC allows for prompt treatment and has shown a reduction of HCC-related mortality by 37%, with improved quality of life[7].

In Australia, surveillance for HCC in the CHB population is based on the 2022 Gastroenterological Society of Australia (GESA) consensus guidelines. This is a proven cost-effective strategy in Australia, as the incidence of HCC in CHB patients ranges between 2%-7%[1]. GESA recommends biannual screening for HCC in CHB patients with a liver ultrasound and a serum alpha foetoprotein (AFP) in patients with cirrhosis, or in patients without cirrhosis who have risk factors for HCC development[1]. In the non-cirrhotic CHB population, the risk of development of HCC is based on established risk factors (Table 1). These include numerous viral factors including presence of hepatitis B e antigen (HBeAg), rising alanine transaminase values, hepatitis B surface antigen seropositivity, and HBV viral DNA load[8,9].

Table 1 Risk factors for hepatocellular carcinoma in patients with chronic hepatitis B.

Risk factors for HCC in chronic hepatitis B
Asian men over the age of 40
Asian women over the age of 50
Sub-Saharan Africans over the age of 20
Aboriginal and Torres Strait Islander people over the age of 50
Co-infection with hepatitis delta virus
Family history of a 1st degree relative with HCC

HCC: Hepatocellular carcinoma.

An alternative to current Australian HCC surveillance practices in non-cirrhotic patients is to utilize validated HCC risk scoring systems. They provide an opportunity for clinicians to use easily accessible tools to quantify HCC risk and determine which patients would benefit from surveillance, thus preventing unnecessary testing. A variety of different scoring systems exist. The most commonly used are the Risk Estimation for HCC in CHB (REACH-B) score and the modified REACH-B (mREACH-B) score which are validated for Asian CHB patients; as well as the Platelet Age Gender-Hepatitis B (PAGE-B) HCC risk scoring tool which is validated for the European Caucasian population. The REACH-B score uses patient sex, age, alanine transaminase, HBeAg and viral DNA status in non-cirrhotic CHB patients not on antiviral therapy. Whilst antiviral therapy reduces the risk of HCC progression, it does not completely mitigate the risk. As such, the mREACH-B score is validated for both treated and untreated CHB patients, substituting HBV DNA levels for liver stiffness measurement (LSM). Using LSM instead of viral DNA levels demonstrated better predictive performance compared to contemporary scoring tools in CHB treated patients[10]. On the other hand, the PAGE-B score utilizes platelet count, age, and sex, and has been developed for both untreated and treated Caucasians. In all three scoring systems, scores equating HCC risk of < 0.2% per year are considered low and would indicate the patient does not require surveillance. There are currently no validated HCC scoring systems for the Australian population. Our aim was to evaluate the differences in surveillance patterns between the REACH-B, mREACH-B, and PAGE-B HCC risk scoring tools against the current GESA surveillance recommendations for HCC surveillance in non-cirrhotic CHB patients.

MATERIALS AND METHODS

A prospectively collected health registry identified all adult patients with CHB from a regional tertiary Australian hospital in South East Queensland from January 2019 to December 2022. The final data cut-off was from December 2022. The study was approved by the Gold Coast Health Service Human Research Ethics Committee, No. EX/2022/QGC/91272. CHB patients without evidence of cirrhosis were included in the study, whereas those with a diagnosis of cirrhosis or already part of cirrhotic HCC screening programme were excluded. Clinical, biochemical, and radiological data were retrospectively collected via electronic medical records to identify host, viral and environmental risk factors associated with HCC development in CHB. Follow-up was determined from the first clinic encounter. Non-cirrhotic status was determined by LSM under the F4 threshold, and by the absence of signs of cirrhosis on imaging.

REACH-B, mREACH-B, and PAGE-B scores were calculated to predict the percentage of risk of developing HCC as shown in Tables 2, 3 and 4 respectively. REACH-B and PAGE-B scores ≤ 9 along with mREACH-B score ≤ 10 are considered low-risk and HCC surveillance is not recommended[1,10]. Descriptive statistics were used to summarize baseline patient characteristics. Graphical summaries, including bar charts, were used to illustrate the distribution of HCC surveillance recommendations for each surveillance tool. Discriminatory performance of HCC risk prediction tools was assessed using receiver operating characteristic (ROC) curve analysis, with calculation of the area under the curve (AUC). Sensitivity and specificity were calculated at pre-specified cut-points defined by the original tool derivation study and guideline recommendations. All statistical analyses were performed using Stata version 19 (StataCorp, College Station, TX, United States).

Table 2 Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B scoring.

Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B
Sex	Female	0
	Male	2
Age (years)	30-34	0
	35-39	1
	40-44	2
	45-49	3
	50-54	4
	55-59	5
	60-65	6
ALT (U/L)	< 15	0
	15-44	1
	≥ 45	2
HBeAg status	Negative	0
	Positive	2
HBV virus DNA level (copies/mL)	< 300-9999	0
	10000-99999	3
	100000-999999	5
	≥ 106	4

ALT: Alanine aminotransferase; HBeAg: Hepatitis B e antigen; HBV: Hepatitis B virus.

Table 3 Modified Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B scoring.

Modified Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B scoring
Sex	Female	0
	Male	2
Age (years)	30-34	0
	35-39	1
	40-44	2
	45-49	3
	50-54	4
	55-59	5
	60-65	6
ALT (U/L)	< 15	0
	15-44	1
	≥ 45	2
HBeAg status	Negative	0
	Positive	2
LSM (kPa)	< 8.0	0
	8.0-13.0	2
	> 13.0	4

ALT: Alanine aminotransferase; HBeAg: Hepatitis B e antigen; LSM: Liver stiffness measurement.

Table 4 Platelet Age Gender-Hepatitis B scoring.

Platelet Age Gender-Hepatitis B scoring
Age (years)	16-29	0
	30-39	2
	40-49	4
	50-59	6
	60-69	8
	≥ 70	10
Sex	Female	0
	Male	6
Platelet count (mm3)	≥ 200000	0
	100000-199999	6
	< 100000	9

RESULTS

A total of 405 patients were initially screened, with 314 patients meeting the inclusion criteria. Baseline characteristics are included in Table 5. The study population comprised 110 (35.0%) males and 204 (65.0%) females with a median age of 50 (22, 85) years. Most patients originated from Asia (69.7%), followed by Oceania (20.7%). Notably, 37.9% of the patients were on antiviral therapy, 6.7% had a first-degree family history of HCC, and 87.9% did not report clinically significant alcohol consumption. The median LSM was 5.3 ± 1.8 kPa. 30 (9.6%) patients were HBeAg positive. Only 2 (0.6%) patients had progression to HCC, the baseline characteristics of these patients can be found in Table 6. Patients were followed up at approximately three to six-monthly intervals.

Table 5 Baseline characteristics of the study population, n (%)/mean ± SD.

Characteristics
Demographics	Total (n)	314
	Age	50.6 ± 13.1
	Male sex	110 (35.0)
Continent of origin	Asia	219 (69.7)
	Oceania	63 (20.0)
	Europe	17 (5.4)
	Africa	10 (3.2)
	Unknown	5 (1.6)
Clinical	LSM (kPa)	5.3 ± 1.8
	Current antiviral use	119 (37.9)
	Diabetes	16 (5.1)
	Excess alcohol use	38 (12.1)
	1st degree family history HCC	20 (6.7)
	2nd degree family history HCC	13 (4.1)
Biochemical	ALT (U/L)	45 ± 196.3
	Platelet count (× 109/L)	240.9 ± 64.3
	AFP (μg/L)	4.3 ± 21.3
	HBeAg positive	30 (9.6)
	HBV DNA < 2000 IU/mL	188 (59.9)
HCC development	Progression to HCC	2 (0.63)

HCC: Hepatocellular carcinoma; LSM: Liver stiffness measurement; HBeAg: Hepatitis B e antigen; ALT: Alanine aminotransferase; AFP: Alpha foetoprotein; HBV: Hepatitis B virus.

Table 6 Clinical characteristics of patients who developed hepatocellular carcinoma during follow-up.

	Patient 1	Patient 2
Age (years)	40	50
Sex	Male	Female
Country of birth	Uganda	China
HBV treatment status	Untreated	Treated
HBV DNA (IU/mL)	1.3 × 103	0
HBeAg status	Negative	Negative
ALT (U/L)	63	14
Platelet count	360	221
AFP	28	315
LSM (kPa)	5.2	5.1
Diabetes	No	No
Excess alcohol use	No	No
Family history hepatocellular carcinoma	No	No

HBV: Hepatitis B virus; HBeAg: Hepatitis B e antigen; ALT: Alanine aminotransferase; AFP: Alpha foetoprotein; LSM: Liver stiffness measurement.

Within the surveillance eligible cohort, 278 (88.5%) patients underwent at least one surveillance ultrasound during the study period, and 221 (70.4%) had an AFP measurement. However, 11 (3.5%) had no documented surveillance over the study period, despite meeting eligibility criteria. The mean follow-up duration was 4 years. A total of 36 patients (11.5%) were lost to follow up.

Implementing surveillance for CHB patients with significant risk factors as per GESA recommendations would encompass 57% of the included patients in this cohort. In comparison, however, the other HCC risk scores reveal a much lower percentage of patients who would qualify for surveillance, including those established on antiviral therapy. REACH-B recommended surveillance in 23.2% of patients, mREACH-B recommended 12.4% and PAGE-B recommended 46.8% (Figure 1). Table 7 highlights the different HCC risk scores for the two patients who developed HCC. These two patients obtained scores under the threshold for HCC surveillance in all three scoring systems, however, they did qualify for surveillance according to GESA guidelines.

Figure 1 Comparison of hepatocellular carcinoma surveillance recommendations. HCC: Hepatocellular carcinoma; GESA: Gastroenterological Society of Australia; REACH-B: Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B; mREACH-B: Modified Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B; PAGE-B: Platelet Age Gender-Hepatitis B.

Table 7 Hepatocellular carcinoma risk score results of patients who developed hepatocellular carcinoma during follow-up.

Patient	REACH-B score	mREACH-B score	PAGE-B score	Surveillance eligibility according to GESA
1	9	9	8	Yes
2	4	4	6	Yes

REACH-B: Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B; mREACH-B: Modified Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B; PAGE-B: Platelet Age Gender-Hepatitis B; GESA: Gastroenterological Society of Australia.

The AUC for the ROC curves for REACH-B, mREACH-B, and PAGE-B were 0.53 [95% confidence interval (CI): 0.12-0.95], 0.55 (95%CI: 0.13-0.97), and 0.39 (95%CI: 0.25-0.53) respectively, indicating poor discriminatory performances across all three models (Figure 2). The REACH-B scoring system recommends HCC surveillance for a score of 10 or over, and at this cut-off point the sensitivity and specificity were 0% and 76.3% respectively for our population. Using the pre-specified cut off point of 11 for mREACH-B, sensitivity was 0% and specificity was 86.5%. When applying the pre-defined PAGE-B cut off point of 10, sensitivity was 0% and specificity was 51.6% for our cohort.

Figure 2 Receiver operating characteristic curves for hepatocellular carcinoma prediction scores. A: Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B; B: Platelet Age Gender-Hepatitis B; C: Modified Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B. REACH-B: Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B; mREACH-B: Modified Risk Estimation for Hepatocellular Carcinoma in Chronic Hepatitis B; PAGE-B: Platelet Age Gender-Hepatitis B; ROC: Receiver operating characteristic.

DISCUSSION

Validated HCC risk scoring tools were created to enhance risk stratification and identify CHB patients who would benefit from regular HCC surveillance. Based on their HCC risk score, patients are classified into low, medium, or high risk. Patients in the low-risk group have a negligible risk for HCC and thus surveillance is not required. While current GESA guidelines recommend risk factor-based surveillance to ensure broader coverage, this can result in over-surveillance, inefficient use of healthcare resources and considerable financial costs. Our results demonstrated reduced numbers of eligible patients for HCC surveillance when using these scoring tools compared with current Australian guidelines. This would suggest a proportion of patients may be undergoing unnecessary surveillance, subsequently leading to an increase in healthcare costs. Importantly, however, all the risk scores failed to recommend surveillance for the two patients who went on to develop HCC, which raises questions on their application in the Australian population.

The three hepatitis B scoring tools (REACH-B, mREACH-B and PAGE-B) use different markers to estimate HCC risk in the non-cirrhotic CHB population. The REACH-B risk tool was specifically validated for use in treatment-naive Asian patients with CHB[9]. Their study involved 3584 CHB non-cirrhotic Taiwanese patients and was validated with Hong Kong and South Korean cohorts, revealing a negative predictive value of 99.2% for 5-year risk of HCC[9]. However, in the era of nucleotide analogues (NA) whereby patients are expected to have lower HBV DNA levels, the accuracy of REACH-B is diminished, bringing the generalisability of REACH-B into question[10,11]. In contrast, mREACH-B and PAGE-B were validated in both treated and untreated CHB patients. In the mREACH-B model, HBV DNA was replaced with LSM which is especially relevant because cirrhosis is another major determinant of HCC risk[12]. The predictability of mREACH-B compared with REACH-B demonstrated greater 3-year and 5-year HCC prediction in treated patients[10]. Whilst PAGE-B was originally modelled around treated Caucasian CHB patients, it has since been validated in treated Asian patients, showing superior results to REACH-B[13,14]. In our cohort, we observed that our HCC surveillance eligibility reduced from 57% using current GESA recommendations to 23.2%, 12.4%, and 46.8% using REACH-B, mREACH-B, and PAGE-B tools respectively. This stark contrast underscores the potential impact of adopting these tools into clinical practice. However, ROC analysis demonstrated poor discriminatory performance for all three tools. While they offer a more targeted approach to surveillance, their use in our cohort notably decreased the number of patients identified. The drastic reduction in eligibility for surveillance also raises concerns of potential missed opportunities for early HCC detection among those classified as low risk. There is uncertainty as to the exact reason behind this difference, but this is likely attributed to the complex interplay of viral, host, and environmental factors which contribute to HCC development in the CHB cohort.

Applying contemporary HCC risk scores universally to the Australian population poses challenges as they currently fail to capture the complex interplay of factors influencing hepatocarcinogensis. Host characteristics such as hepatitis B genotype and metabolic and lifestyle factors all modify the risk for HCC development. It is important to note that these tools were developed and validated in specific homogenous populations globally, often lacking diverse representation, in contrast to the migrant-rich Australian population. Within this population, multiple HBV genotypes circulate, each associated with distinct biological behavior and oncogenic risk. Genotype C, the most prevalent genotype in Australia, is associated with delayed HBeAg seroconversion leading to prolonged viral replication and a persistent inflammatory response in the liver. This genotype is therefore associated with an increased incidence of HCC compared to genotypes A (more commonly seen in Africa, Europe and North America), and genotype B (which predominates in Asia along with genotype C)[15-17]. Immigration related factors, including age at migration, duration of untreated infection, and type of genotype acquired, may result in variation in cumulative carcinogenic exposure that is not accounted for in existing models[1,16,17].

In addition, metabolic and lifestyle related factors such as obesity, diabetes mellitus, cigarette smoking and alcohol consumption further modify HCC risk in patients with HBV infection. Diabetes is independently associated with an increased HCC risk, likely mediated through insulin resistance, chronic inflammation, and hepatic steatosis, which may act synergistically with viral induced liver injury from HBV. Similarly, alcohol exposure increases the risk of cirrhosis and HCC development in a dose-dependent fashion through accelerated fibrosis and possibly through promotive HBV replication[18]. As no HCC risk scores have been specifically derived or validated in the Australian HBV cohort, their generalisability is limited, potentially resulting in an inaccurate risk assessment.

In our cohort, all tested HCC risk scoring tools would not have recommended surveillance for the two CHB patients who went on to develop HCC. In contrast, both patients met the surveillance criteria under the current GESA consensus recommendations (Table 7). The first patient was an Asian, non-cirrhotic individual receiving NA therapy, while the second was an African, non-cirrhotic, treatment-naive patient. The failure to identify the latter for surveillance may, in part, reflect the lack of validation of the HCC risk scores in African populations. Although the former patient was on NA therapy, which reduces the predictive accuracy of the REACH-B score, alternative scores like the mREACH-B and PAGE-B also failed to identify this patient as requiring surveillance, highlighting the lack of complete accuracy of these risk stratification tools.

Our findings also suggest that surveillance effectiveness is not solely determined by eligibility definitions, but also by degree of implementation. Surveillance uptake in our cohort was relatively high, with nearly 90% of eligible patients receiving at least one ultrasound, and 70% undergoing AFP testing throughout the study period. A small but clinically relevant proportion (3.5%) of surveillance eligible patients received no monitoring at all. This compares favourably to other real-world studies where adherence rates to HCC surveillance imaging were poorer[19]. Although uptake was high, 11.5% of our patients were lost to follow up. Accordingly, improving longitudinal surveillance delivery rather than simply redefining eligibility criteria may represent a more impactful target for quality improvement. Taken together, these findings underscore that optimisation of both risk stratification accuracy and surveillance execution is critical to improving HCC detection outcomes in Australia.

Despite providing important real-world insights, this study has several limitations that must be considered. Firstly, this study was conducted at a single tertiary center with a limited sample size, which may limit the generalisability to the broader Australian CHB population. Although the cohort reflects the ethnic heterogeneity commonly encountered in a major Australian urban center, referral patterns to tertiary centers may introduce selection bias and limit generalisability to regional populations. To overcome these limitations, our findings would need to be validated by a multi-centre study capturing a greater sample of the Australian population. Secondly, as this study is retrospective it is inherently subject to limitations, including selection bias, incomplete data capture and confounding factors. While data was obtained from a prospectively maintained registry and supplemented by review of medical records, the retrospective nature of this study precludes causal inference and limits control over unmeasured variables that may influence HCC risk. Lastly, the number of HCC events observed during the follow-up period was extremely small. This ultimately limits statistical power and the ability to draw definitive conclusions on the discriminative performance of individual HCC risk scoring systems. Our calculated AUC for each scoring tool indicated poor discriminatory performance, which needs to be interpreted cautiously due to the small number of HCC events. The low event rate likely reflects the relatively short follow-up period and the non-cirrhotic nature of the cohort, rather than an absence of long-term HCC risk. Despite these limitations, our findings provide real-world insights into the applicability of commonly used international HCC risk scores within an Australian non-cirrhotic CHB population, and highlight important considerations for future refinement of surveillance strategies.

CONCLUSION

While HCC risk scoring tools hold promise for improving surveillance in CHB patients, their universal application requires careful consideration, particularly within diverse migrant-rich populations such as Australia. Their integration into clinical practice carries significant implications, most notably potentially underestimating the risk of HCC in individuals with unaccounted virological, metabolic, and socio-demographic risk factors. Such underestimation may translate into delayed or missed surveillance opportunities, ultimately compromising early HCC detection and curative treatment eligibility. These risk tools offer the potential to enhance early detection of HCC; however, their limitations in generalisability and lack of validation in certain populations must not be overlooked. At present, no international guidelines recommend their adoption into clinical practice, however, the European Association for the Study of the Liver and GESA acknowledge their potential[1,20]. Further studies and multi-centre validation are needed before these tools can be recommended for diverse ethnic populations, for which additional patient factors and social determinants should be taken into account.

In the absence of Australian specific HCC risk models, a pragmatic and integrated surveillance approach may be more appropriate for clinical practice. Internationally derived risk scores can serve as an initial framework for risk stratification; however, relying on their sole use should be considered with caution. Instead, a combination of local risk factors, including genotype, lifestyle exposures, metabolic comorbidities, and family history of HCC should be considered. Ultimately, a balanced approach of using risk scores as a supportive tool in adjunct with targeted patient risk assessment is central to improve early HCC detection in clinical practice. In the current Australian setting, the greatest immediate clinical gains are likely to be achieved not through widespread adoption of a particular risk score, but through optimised adherence to existing guideline recommended surveillance programmes.

ACKNOWLEDGEMENTS

We thank the hepatology clinical nurses for their invaluable contribution to the patient registry. We also thank Dr. Ian Hughes for his assistance with statistical analysis.