STRENGTHS AND LIMITATIONS OF THIS STUDY

Diabetic patients with metabolic dysfunction-associated steatotic liver disease (MASLD) were divided into six groups, risk-stratified according to the degree of obesity and progression to cirrhosis.

One-way, two-way and probabilistic sensitivity analyses were performed to assess the impact of model uncertainty.

The sensitivity and specificity of abdominal ultrasound (US) with alpha-fetoprotein (AFP) and US with AFP and lectin-reactive alpha-fetoprotein in obese diabetic patients with MASLD were not obtained.

Other hepatocellular carcinoma risk factors, such as viral hepatitis infection, alcohol use, tobacco use, genetic susceptibility and male gender were not considered.

Introduction

Diabetes and metabolic dysfunction-associated steatotic liver disease (MASLD) are two of the most important increasing risk factors for hepatocellular carcinoma (HCC) in the world.1–8 In 2021, 537 million adults aged 20–79 years had diabetes worldwide, and the number of people with diabetes is estimated to reach 643 million by 2030 and 784 million by 2045.9 The prevalence of MASLD is estimated to be up to 75% of patients with type 2 diabetes and increases with obesity.1 4 MASLD is also often associated with metabolic comorbidities, such as obesity (51.34%), type 2 diabetes (22.51%), hyperlipidaemia (69.16%), hypertension (39.34%) and metabolic syndrome (42.54%).10 Metabolic dysfunction-associated steatohepatitis (MASH), the progression to cirrhosis and advanced liver fibrosis accelerate the development of HCC.10 HCC is the sixth most common cancer in new patients and the third most common cancer in deaths in 2022 in the world, and the number of new HCC cases and deaths are estimated to be 866 136 and 758 725, respectively.11 Asia is reported to account for 70.1% of the world’s liver cancer cases and 70.0% of the world’s liver cancer deaths in 2022.11 Elderly, male and Asian populations remain the highest risk groups for HCC.12 In Japan, a total of 37 296 liver cancers were diagnosed in 2019, of which HCC accounted for 91.1% and intrahepatic bile duct cancer 6.4%.13 The 22nd nationwide follow-up survey of primary liver cancer in Japan showed that hepatitis C virus antibody and the hepatitis B surface antigen-positivity as major pathogenic promoters of HCC have decreased to 49.2% and 13.8%, replacing diabetes and severe obesity (BMI≥30 kg/m2) which have increased to 32.5% and 5.2%.14 The 5 year survival rate for patients with advanced HCC is very poor at 8%, and even the 5 year survival rate for patients with early-stage HCC stands at 69.6% in Japan.13 Surveillance programmes for detecting early-stage HCC provide an increased chance of curative treatment options including surgical liver resection, liver transplantation, radiofrequency ablation and transarterial chemoembolisation and improve prognosis.7

Currently, six screening methods for HCC are available: abdominal ultrasound (US), US with alpha-fetoprotein (AFP), US with AFP and lectin-reactive alpha-fetoprotein (AFP-L3), CT, extracellular contrast-media-enhanced MRI (ECCM-MRI) and gadoxetic acid-enhanced MRI (EOB-MRI). The accuracy of US is greatly influenced by the degree of obesity.15 For obese diabetic patients with fatty liver, CT and MRI are preferred as primary screening tools compared with US because US reflects off fatty tissue and has low sensitivity for the diagnosis of HCC.15 EOB-MRI has a superior sensitivity for the diagnosis of HCC compared with US.16–18 Japanese evidence-based clinical practice guidelines for MASLD/MASH recommend US and tumour marker measurements every 6 months in MASLD with liver cirrhosis.19 American Gastroenterological Association (AGA) clinical practice guidelines recommend that HCC screening should be considered in all patients with cirrhosis due to non-alcoholic fatty liver disease and that prospective screening with or without AFP should be performed every 6 months with CT or MRI if US quality is suboptimal for HCC screening, for example, due to obesity.20 The risk of developing HCC is 2.18 and 3.28 times higher for type 2 diabetes and type 2 diabetes combined with obesity, respectively.21–23 With the estimated increase in obesity, the risk of HCC in diabetic patients is increasing and HCC surveillance in diabetic patients with MASLD is gaining epidemiological importance and attention along with its clinical significance. However, the effectiveness and cost-effectiveness of HCC screening for diabetic patients with MASLD have not yet been assessed. Cost-effectiveness studies considering the costs and the benefits of HCC screening for diabetic patients with MASLD warrant evaluation.

This study aimed to evaluate the cost-effectiveness of risk-stratified HCC screening according to the degree of obesity and progression to cirrhosis in diabetic patients with MASLD.

MethodsStudy design

A state-transition model was developed from a healthcare payer perspective on a lifetime horizon. Decision branches led directly to one Markov node for each intervention strategy, and the initial event was modelled within a Markov cycle tree. Six HCC screening strategies were compared with no screening: US, US with AFP, US with AFP and AFP-L3, CT, ECCM-MRI and EOB-MRI. A cycle length of half a year was chosen24 25 and all probabilities were adjusted for the 6 month cycle length. The half-cycle correction was applied. The target population was a hypothetical cohort of 50-year-old diabetic patients with MASLD.25 HCC incidence depends on the degree of obesity and progression to cirrhosis in diabetic patients. Diabetic patients with MASLD were divided into six groups, risk-stratified according to the degree of obesity and progression to cirrhosis: non-obese metabolic dysfunction-associated steatotic liver (MASL), non-obese MASH, non-obese MASH cirrhosis; obese MASL, obese MASH and obese MASH cirrhosis (figure 1). MASL, MASH and MASH cirrhosis are progressive manifestations of this specific type of liver disease.

Figure 1

Risk-stratified classification according to the degree of obesity and progression to cirrhosis in diabetic patients with MASLD. MASH, metabolic dysfunction-associated steatohepatitis; MASL, metabolic dysfunction-associated steatotic liver; MASLD, metabolic dysfunction-associated steatotic liver disease.

The main outcomes were costs, quality-adjusted life-years (QALYs), incremental cost-effectiveness ratios (ICERs), early-stage HCC cases, advanced-stage HCC cases and HCC-related deaths. The willingness-to-pay (WTP) threshold was $50 000 per QALY gained.26 The model was constructed with TreeAge Pro 2024 (TreeAge Software, Williamstown, MA).

Ethical approval was not required because this was a modelling study with all inputs and parameters derived from the published literature and Japanese statistics. This study is reported following the Consolidated Health Economic Evaluation Reporting Standards 2022 (CHEERS 2022) statement.27

Model structureHCC screening strategy

A patient undergoes HCC screening every 6 months. Six HCC screening methods were considered in the models: US, US with AFP, US with AFP and AFP-L3, CT, ECCM-MRI and EOB-MRI. When US, US with AFP or US with AFP and AFP-L3 is selected as HCC screening, the patient with a true-positive result confirms the presence of HCC by ECCM-MRI and CT and then receives liver biopsy before liver resection, radiofrequency ablation, transarterial chemoembolisation, liver transplantation or palliative care, according to the stage of cancer. When a liver biopsy reveals a complication of liver bleeding, the patient is treated for liver bleeding. When US, US with AFP or US with AFP and AFP-L3 is selected as HCC screening, the patient with a false-positive result undergoes CT and ECCM-MRI to confirm the absence of HCC and is screened the following year. When ECCM-MRI or CT is selected for HCC screening, CT or ECCM-MRI is added to confirm HCC diagnosis. The probability of detecting and treating HCC in the patient with a false-negative result was assumed to be the same as in patients who were not screened. The patient with a true-negative result proceeds directly to the next year’s screening programme. The recurrence of HCC was considered in the model. The adherence rate of HCC screening was assumed to be 32% from the literature.28

No screening

In no screening, a diabetic patient has no opportunity for HCC screening. When HCC is suspected in a clinical setting, the patient is treated for HCC after confirmation of HCC diagnosis by CT and ECCM-MRI. The model is set up to detect HCC in 24.3% of early-stage HCC, 73.4% of intermediate-stage HCC and 2.3% of advanced-stage HCC.29

Model inputsProbability

The probabilities were collected using MEDLINE and Japanese statistics from 2000 to August 2024 to estimate input parameters for the models (online supplemental table S1). The incidence of HCC in non-obese diabetic patients with MASLD was calculated by multiplying the incidence of HCC in patients with MASLD by the relative excess risk of developing HCC with type 2 diabetes mellitus.10 21 22 30 31 The incidence of HCC in obese diabetic patients with MASLD was calculated by multiplying the incidence of HCC in patients with MASLD by the relative excess risk of developing HCC with type 2 diabetes mellitus and obesity.10 21–23 30 31 Stage-specific probability of HCC detection by US, MRI, CT and no screening,29 32 stage-specific 5 year HCC survival rate,14 increased risk of all-cause mortality due to MASLD,33 adherence rate of HCC screening28 and the probability of liver bleeding complication by percutaneous liver biopsy34 were obtained from the literature and Japanese cancer statistics. The sensitivity and specificity of US, US with AFP, US with AFP and AFP-L3, CT, ECCM-MRI and EOB-MRI were derived from the literature.15 35–37 All-cause mortality rates were obtained from Japanese life tables.

Cost

Costs were calculated as uninsured costs from the Japanese medical fee schedule,38 adjusted to 2021 Japanese yen, using the medical care component of the Japanese consumer price index and converted to 2021 US dollars, using the Organisation for Economic Co-operation and Development (OECD) purchasing power parity rate ($1=¥100.412) (online supplemental table S1).39 Stage-specific treatment costs of HCC were calculated based on Japanese clinical practice guidelines for HCC.40 Direct costs were based on the perspective of Japanese healthcare payers; however, indirect costs were not included in the cost calculation. All costs were discounted by 3%.41 42

Health state utility

A Markov model was constructed with a Markov cycle tree which included 10 health states: precancer, liver bleeding complication, early-stage HCC, intermediate-stage HCC, advanced-stage HCC, recurrent early-stage HCC, recurrent intermediate-stage HCC, recurrent advanced-stage HCC, post-treatment HCC and dead (figure 2). Health state utility values were obtained from the literature (online supplemental table S1).43–45 Health state utility of liver bleeding complication was assumed. The annual discounting of health state utility values was set at a rate of 3%.41 42

Figure 2

Schematic depiction of a Markov cycle tree in a state-transition model. This figure displays the health states in the model as ovals, and the possible transitions between one health state and another during a 6 month model cycle are indicated by arrows. HCC, hepatocellular carcinoma.

Sensitivity analysis

One-way sensitivity analyses were performed on probabilities, costs and health state utilities to determine which strategy would be more cost-effective if certain variables were considered in the widest possible range, holding all other variables constant. Two-way sensitivity analyses were performed to assess the impacts of a decision on simultaneous changes in semiannual HCC incidence and adherence rate of HCC screening. Probabilistic sensitivity analyses using a second-order Monte Carlo simulation for 10 000 trials were conducted to assess the impact of model uncertainty on the base-case estimates. The uncertainty had a beta distribution for probability, health state utility and accuracy, a gamma distribution for cost and relative risk and a Dirichlet distribution for stage-specific probability of HCC detection.

Markov cohort analysis

Markov cohort analysis was performed to determine the cumulative lifetime probability of early-stage HCC cases, advanced-stage HCC cases and HCC-related deaths with and without HCC screening. The cumulative lifetime probabilities of early-stage HCC cases, advanced-stage HCC cases and HCC-related deaths were multiplied by 100 000 to obtain the cumulative lifetime number of early-stage HCC cases, advanced-stage HCC cases and HCC-related deaths in 100 000 diabetic patients with MASLD.

Patient and public involvement

Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of my research.

ResultsBase-case analysis

In non-obese diabetic patients with MASL and MASH, and obese diabetic patients with MASL, no HCC screening method is more cost-effective than no screening (online supplemental table S2). In obese diabetic patients with MASH, EOB-MRI was more cost-effective than no screening (ICER, US$49 160 per QALY gained) (online supplemental table S2). In non-obese and obese diabetic patients with MASH cirrhosis, EOB-MRI yielded the greatest cost-saving and the greatest QALYs of any HCC screening (online supplemental table S2).

Sensitivity analysisOne-way sensitivity analysis

Cost-effectiveness was sensitive to HCC incidence in non-obese diabetic patients with MASH cirrhosis and obese diabetic patients with MASH, and the adherence rate of HCC screening in obese diabetic patients with MASH. When the semiannual HCC incidence was between 0.008 and 0.0138, US with AFP was more cost-effective than EOB-MRI in non-obese diabetic patients with MASH cirrhosis (figure 3A,B). When the semiannual HCC incidence was over 0.0086 and the adherence rate of HCC screening was under 0.49, EOB-MRI was more cost-effective than no screening in obese diabetic patients with MASH (figure 3C).

Figure 3

Tornado diagram showing results of one-way sensitivity analysis. (A) No screening vs US with AFP in non-obese diabetic patients with MASH cirrhosis. (B) US with AFP vs EOB-MRI in non-obese diabetic patients with MASH cirrhosis. (C) No screening vs EOB-MRI in obese diabetic patients with MASH. AFP, alpha-fetoprotein; EOB-MRI, gadoxetic acid-enhanced MRI; EV, expected value; HCC, hepatocellular carcinoma; MASH, metabolic dysfunction-associated steatohepatitis; NMB, net monetary benefit; US, abdominal ultrasound.

Two-way sensitivity analysis

Figure 4 shows the results of two-way sensitivity analyses for semiannual HCC incidence and adherence rate of HCC screening. In non-obese diabetic patients with MASH cirrhosis, there was a brown area where US with AFP was more cost-effective than EOB-MRI when HCC incidence was low (figure 4A). In obese diabetic patients with MASH, stratification into EOB-MRI or no screening was mainly based on HCC incidence (figure 4B).

Figure 4

Two-way sensitivity analysis for semiannual HCC incidence and adherence rate of HCC screening in non-obese diabetic patients with MASH cirrhosis (A) and obese diabetic patients with MASH (B). AFP, alpha-fetoprotein; AFP-L3, lectin-reactive alpha-fetoprotein; ECCM-MRI, extracellular contrast-media-enhanced MRI; EOB-MRI, gadoxetic acid-enhanced MRI; HCC, hepatocellular carcinoma; MASH, metabolic dysfunction-associated steatohepatitis; US, abdominal ultrasound.

Probabilistic sensitivity analysis

Cost-effectiveness acceptability curves showed that no screening was 100% cost-effective in non-obese and obese diabetic patients with MASL at a WTP level of $50 000 per QALY gained (figure 5A,B). In non-obese diabetic patients with MASH, US with AFP was 25.9% cost-effective (figure 5C). In non-obese diabetic patients with MASH cirrhosis, obese diabetic patients with MASH and obese diabetic patients with MASH cirrhosis, EOB-MRI was 96.0%, 50.7% and 99.9% cost-effective (figure 5E, D and F).

Figure 5

Cost-effectiveness acceptability curve for diabetic patients with MASLD. (A) Non-obese diabetic patients with MASL. (B) Obese diabetic patients with MASL. (C) Non-obese diabetic patients with MASH. (D) Obese diabetic patients with MASH. (E) Non-obese diabetic patients with MASH cirrhosis. (F) Obese diabetic patients with MASH cirrhosis. AFP, alpha-fetoprotein; AFP-L3, lectin-reactive alpha-fetoprotein; ECCM-MRI, extracellular contrast-media-enhanced MRI; EOB-MRI, gadoxetic acid-enhanced MRI; MASH, metabolic dysfunction-associated steatohepatitis; MASL, metabolic dysfunction-associated steatotic liver; MASLD, metabolic dysfunction-associated steatotic liver disease; QALY, quality-adjusted life-year; US, abdominal ultrasound.

Cumulative lifetime health outcome

Compared with no screening in 100 000 non-obese diabetic patients with MASH cirrhosis and obese diabetic patients with MASH cirrhosis, EOB-MRI reduced total costs by US$69 million and by US$142 million, increased lifetime effectiveness by 12 546 QALYs and by 15 815 QALYs, detected 17 873 and 21 014 early-stage HCC cases and averted 2068 and 2471 HCC-related deaths, respectively (table 1). In other words, EOB-MRI increased the detection of early-stage HCC by 4.2-fold, decreased the detection of advanced-stage HCC by 29% and reduced HCC-related deaths by 21% compared with no screening (table 1). The findings support that EOB-MRI provides the greatest cost-saving and benefits in diabetic patients with MASH cirrhosis.

Table 1

Cumulative lifetime economic and health outcomes of HCC screening compared with no screening

Discussion

This study showed that EOB-MRI is the most cost-effective screening method for non-obese diabetic patients with MASH cirrhosis and for obese diabetic patients with MASH and MASH cirrhosis. Particularly, in diabetic patients with MASH cirrhosis, EOB-MRI yields the greatest cost-saving, detects the greatest number of early-stage HCC cases and averts the greatest number of advanced-stage HCC cases and HCC-related deaths. There are two reasons for the superior cost-effectiveness of EOB-MRI. First, EOB-MRI has high specificity along with high sensitivity. Second, the probability of detecting early HCC on MRI-based screening is higher than US-based screening and no screening. EOB-MRI prevents delays in therapeutic intervention and improves the prognosis of patients with HCC. The earlier HCC cases are detected, the lower the cost of treatment and the better the quality of life for patients.

The findings support the promotion of risk-stratified HCC screening in diabetic patients with MASLD for the best secondary prevention of HCC. Lee and colleagues suggested that changing the current ‘one-size-fits-all’ HCC screening to a personalised and precise approach for early HCC detection could ultimately improve the poor prognosis of HCC patients.46 Not only cost-effectiveness but also the individual clinical profile of the patient and the imaging characteristics of abnormal lesions are important in selecting HCC screening methods. Obesity, one of the risk factors for HCC, is preventable for patients. Physicians, diabetic patients and healthcare providers should pay close attention to focus on primary prevention for HCC through obesity prevention and lifestyle modification efforts.

To the best of my knowledge, this is the first study in the world to evaluate the cost-effectiveness of risk-stratified HCC screening according to the degree of obesity and progression to cirrhosis for diabetic patients with MASLD.

US is non-invasive, inexpensive and easy to use in clinical practice, but early detection of HCC in obese patients is difficult and less accurate.15 47 Both US and MRI have the advantage of no radiation exposure associated with the examination, unlike CT. CT has the disadvantage of contrast-induced nephrotoxicity. EOB-MRI is very sensitive for early and small lesions of HCC compared with ECCM-MRI and can help differentiate early HCCs from cirrhosis-associated benign nodules.48 Recently, EOB-MRI has been reported to provide useful imaging features for HCC subtype differentiation according to the WHO Classification of Digestive System Tumours, fifth Edition.18

There are several cost-effectiveness studies of US versus MRI for HCC screening for patients with liver cirrhosis. Goossens and colleagues demonstrated that risk-stratified HCC surveillance strategies applying MRI and/or US targeting high- and intermediate-risk patients with cirrhosis were cost-effective.49 Kim and colleagues showed that MRI was more cost-effective than US for HCC surveillance in high-risk patients with cirrhosis.50 Nahon and colleagues found that semiannual surveillance with MRI in patients with cirrhosis could enable the cost-effective detection of 5× very early HCCs compared with US monitoring.32 Nishie and colleagues demonstrated that EOB-MRI could be the first-choice imaging modality for medical care of HCC among patients with hepatitis or liver cirrhosis compared with ECCM-MRI and CT.17 The results of this study are consistent with those of previous cost-effectiveness studies. Parikh and colleagues showed that US with AFP was more cost-effective than US and no screening in patients with compensated cirrhosis and addressed the importance of considering both the benefits and harms of HCC surveillance.51 Considering patients’ burdens, including the cost of unnecessary additional imaging examinations due to false positive results, reduced quality of life due to liver biopsy and treatment costs due to liver biopsy, this study showed that EOB-MRI was superior to US with AFP in diabetic patients with a high incidence of HCC. The superior cost-effectiveness of EOB-MRI for diabetic patients with MASH cirrhosis in this study has potential implications for future amendment and expansion of Japanese evidence-based clinical practice guidelines for MASLD/MASH19 and AGA clinical practice guidelines.20

This study has several limitations. First, stage-specific probabilities of HCC detection using US, MRI and CT were obtained from the literature.32 Large-scale prospective studies are needed to determine stage-specific probabilities of HCC detection by each HCC screening method. Second, obesity is often correlated with other risk factors such as alcohol, dyslipidaemia and cardiovascular diseases. Third, the sensitivity and specificity of US with AFP and US with AFP and AFP-L3 in obese diabetic patients with MASLD were not obtained. Fourth, indirect costs such as lost productivity, work absenteeism and income loss were not included in the costs in this study. Fifth, as assumed from the literature,28 the adherence rate of HCC screening was low. Sixth, the health state utilities were obtained from the literature.43–45 Seventh, other HCC risk factors, such as viral hepatitis infection, alcohol use, tobacco use, genetic susceptibility and male gender were not considered in this study.52–54 Finally, each country has different costs for HCC screening, epidemiological parameters and healthcare and insurance systems. Further cost-effectiveness studies are needed based on country diversity.

In conclusion, of all HCC screening methods for diabetic patients with MASH cirrhosis, EOB-MRI yields the greatest cost-saving with the highest QALYs, detects the greatest number of early-stage HCC cases and averts the greatest number of advanced-stage HCC cases and HCC-related deaths. For diabetic patients with a high HCC incidence, the high accuracy and early HCC detection rate for EOB-MRI make EOB-MRI the most cost-effective option. The findings provide important insights for the precise implementation of risk-stratified HCC surveillance to reduce HCC mortality and improve the quality of life in diabetic patients with MASLD.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information. The data are available in the current study.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Not applicable.