Health economics aspects of primary liver cancer in Australia

Background: Primary liver cancer (PLC), with hepatocellular carcinoma (HCC) being the most frequently diagnosed subtype, is the fastest increasing cancer in many countries in terms of incidence and mortality. HCC is mainly caused by viral hepatitis, excessive alcohol consumption, metabolicassociated fatty liver disease (MAFLD), diabetes, hereditary haemochromatosis, and exposure to aflatoxin. It is frequently diagnosed at late stages due to its asymptomatic nature and thereby, results in the constraint of treatments with curative intent for patients. Surveillance has been shown to be effective at detecting HCC early, and thereby improve the survival chances of patients. However, a large-scale implementation of surveillance would likely be resource-intensive and therefore, decision makers need to be confident that the increased cost is reasonably offset by the benefits the program can achieve. This thesis investigates the health and economic burden of PLC and the costeffectiveness of HCC surveillance in Australia. The results from this thesis provide robust data for the future decision making process in terms of resource allocation for PLC and HCC early-diagnosis and management programs that represent best value for money. Objectives: The principal aims of this thesis were to: ‚ÄövѬ¢ Assess current health economic evidence for HCC surveillance by systematically synthesising published studies to provide an overview of health economic evaluations for HCC surveillance; ‚ÄövѬ¢ Comprehensively assess the resource utilisation and costs associated with PLC in public hospitals in Australia; ‚ÄövѬ¢ Compare survival of patients with PLC from different socio-economic statuses, region of residence and ethnicities, and explore factors influencing survival; ‚ÄövѬ¢ Develop and validate a disease simulation model for HCC to assess costs and outcomes associated with HCC surveillance in Australia. Thesis outline, methods, and summary of key results Chapter 1 provides a general introduction to PLC/HCC and concepts of health economics. Chapter 2 presents a systematic review and narrative synthesis of the cost and/or effectiveness of HCC surveillance. Forty-three studies were included in the review and most of them were modelling studies that employed either decision tree analysis or Markov models. The majority of studies evaluated surveillance on people with liver cirrhosis and most concluded HCC surveillance using either ultrasound or the combination of ultrasound and alpha-fetoprotein (AFP) to be cost-effective. However, several limitations within these studies necessitates the cost-effectiveness results to be interpreted with care. Computed tomography (CT) and magnetic resonance imaging (MRI) scans have been used more commonly for HCC surveillance but using these methods may not be cost effective. In addition, CT also poses the risk of radiation exposure whilst MRI can be unacceptable to patients due to discomfort. This study was published in Value in Health in May 2021 Chapter 3 presents the hospital-related costs of PLC in Australia. The study cohort consisted of all individuals with PLC notification to the Victorian Cancer Registry (VCR) from 01/2018 to 12/2015. The Centre for Victorian Data Linkage (CVDL) linked the data of these patients with three other datasets, including the Victorian Admission Episode Dataset (VAED), Victorian Emergency Minimum Dataset (VEMD) and Victorian Death Index (VDI). Costs were measured during the first and second year after cancer notification. Direct medical costs based on the admissions and emergency presentations were estimated as total, annualised and annual per-patient costs in 2017 Australian Dollar (AUD). The annual per-patient costs and first-year survival rate of the cohort were used to extrapolate the costs to the Australian population. Factors associated with costs were explored using the linear mixed model with Box-Cox transformation of the cost. The results from this study showed PLC placed a heavy financial burden on the Australian health system with the total cost of AUD 137 million for the 12 months post diagnosis and AUD 42.6 million for the subsequent 12 to 24 months after cancer notification. This study has been accepted for publication in Australian Health Review in April 2022. Chapter 4 presents a survival analysis of PLC that took into account the culturally and linguistically diverse (CALD) backgrounds, region of residence, socio-economic, and PLC aetiologies of patients. Using the same dataset as outlined for Chapter 3, with additional hospital admission data from 01/2000. The Kaplan-Meier method, log-rank test and Cox proportional hazard model was used to explore the survival of PLC patients. Assessment of the proportional hazard assumption was conducted using the Schoenfeld residuals and by testing the inclusion of the interaction terms between each covariate with time in the model. This study demonstrated that PLC is a cancer with very high mortality rates. Australian and New Zealand cases had lower survival than cases born in Asia and Africa as PLC in Asian and African cases was mostly caused by viral hepatitis. Higher survival was observed for cases living in metropolitan areas, not only due to accessibility to surveillance and healthcare services but also because the majority of overseas-born patients reside in metropolitan areas. Intervention programs for PLC need to place more emphasis on the socioAbstract economically disadvantaged populations and those whose PLC is caused by alcoholic liver disease, MAFLD or obesity to achieve better outcomes for patients. This study is under revision in Cancer Epidemiology. Chapter 5 documents the construction and validation of a state-transition microsimulation model that evaluates the cost-effectiveness of HCC surveillance using ultrasound with or without AFP at 6- month intervals in Australia. Based on the recently published consensus statement for HCC management in Australia, the model was constructed to take into account the limitations that previous models have not addressed and was validated for its face, internal and external validities. It demonstrated a good level of accuracy and, therefore, decision makers can interpret the results generated from this model with high confidence. This study has been submitted to Global Health Research and Policy. Chapter 6 presents the main findings of the model. The model used a hypothetical cohort of individuals at different ages having non-cirrhotic chronic hepatitis B (CHB), compensated cirrhosis and decompensated cirrhosis. Multiple one-way and probabilistic sensitivity analyses as well as scenario and threshold analyses were conducted to account for uncertainties existing around the parameters, characteristics of population, process of microsimulation and structure of the model. The results showed HCC surveillance based on current Australian recommendations using either ultrasound or ultrasound with AFP was cost-effective, but the former strategy was dominated by the latter strategy due to its higher incremental cost-effectiveness ratios (ICER). Surveillance was also not cost-effective if it was limited to the CHB population due to the much lower risk of developing HCC than those of compensated and decompensated cirrhosis population. As obesity can decrease the diagnostic performance of ultrasound, this was also considered in the model: based on Australian obesity prevalence data, HCC surveillance across all at-risk groups remained cost-effective. This study has been submitted to The Lancet Regional Health - Western Pacific. Chapter 7 presents the summary and conclusion of this thesis and directions for future research.