Study population
The target group for analysis was adults who received a medical check-ups or health check-ups for various reasons in the US. Subjects with a clear risk for liver diseases such as viral hepatitis or significant alcohol consumption were excluded. All subjects had their FIB-4 calculated, and if it was higher than 1.3, VCTE test was sequentially performed. If the liver stiffness value was higher than 8.8kpa, and there was no evidence of chronic viral hepatitis, and there was no history of significant alcohol consumption (intake greater than 210 g for male and 140 g for female), the ILI applied to all subjects.
To identify the burden of fibrosis in the US population, we estimated the prevalence of advanced hepatic fibrosis from data of VCTE of the US National Health and Nutrition Examination Survey (NHANES) from 2017-March 2020. According to previous study, the liver stiffness cut off based on VCTE results of 6.2 kPa, 7.6 kPa, 8.8 kPa and 11.8 kPa were used for F0, F1, F2, F3 and F4, respectively19,20,21. Stratifying by age groups ranging from the 30–39 years to the 60–69 years, the prevalence of advanced fibrosis was 6.3–10.7% (Table 1).
Construction of a decision analytic model
A cost-utility analysis was conducted to evaluate the cost-effectiveness of advanced fibrosis screening for long-term clinical outcomes. We constructed a hybrid decision tree and Markov model using Excel 2019 (Microsoft Corp) to assess the difference in quality-adjusted life-year (QALY) and expected cost between the’ screening’ and ‘no screening’ groups.
Decision tree model: screening strategy
The decision tree represents the detailed structure of the events and outcomes associated with the two comparison strategies (Fig. 1A). Those with a high FIB-4 (> 1.3) were referred to a referral center and VCTE testing. The sensitivity and specificity of FIB-4 was obtained from previous studies22. In the aged 40–49 years, FIB-4 sensitivity and specificity were 56.6% and 86.2%, respectively22 For the performance of VCTE, sensitivity was 80.0%, specificity was 74.0%, respectively23 After screening, patients diagnosed with advanced fibrosis (both true positive (TP) and false positive (FP) cases) participated in an intensive lifestyle intervention (ILI) program. On the other hand, in the case of the no screening strategy, the simulation cohort with the same risk as the screening group did not receive ILI.
Overview of decision analytic model (A) decision tree model, (B) Markov model. CVD, cardiovascular disease; DC, decompensated cirrhosis; FN, false negative; FP, false positive; HCC, hepatocellular carcinoma; ILI, intensive lifestyle intervention; MRE, magnetic resonance elastography; NIT, non-invasive test; TN, true negative; TP, true positive; VCTE, vibration controlled transient elastography.
* It represents three states of CVD in H0, F0-F2, F3, and F4.
Markov model
The population classified according to the diagnostic results was subdivided from F0 to F4 according to the distribution of fibrosis. We assumed half of the F0 population as healthy population (H0) who do not have the risk of fibrosis. And then the classified population entered into the Markov model to estimate long-term cost and obtain life-year (LY) and QALY (Fig. 1B). The model included H0 (healthy population), fibrosis stages; F0 to F4 (and each pair condition undergoing ILI), cardiovascular disease (CVD) complications (and each pair condition undergoing ILI), decompensated cirrhosis (DC), and hepatocellular carcinoma (HCC), extrahepatic malignancy and death. The CVD status was subdivided into three stages (CVD with F0-F2/F3/F4) according to the stage of fibrosis. HCC can occur in F3, F4 and DC. The disease-specific mortality (F0-F4, CVD, HCC, and extrahepatic malignancies) and age-specific all-cause mortality were applied for all health conditions. Details of parameters related on disease progression risk and mortalities applied to the model are described in Supplementary Table 1, Supplementary Tables 2, and Supplementary Methods. The cycle length of the model was 1 year, and the time horizon was set to 30 years.
Assumption
The following assumptions were applied to this analytical model. (i) The population classified by screening results was subdivided into H0 and F0-F4 according to the ratio of fibrosis stage. Patients with H0 (healthy population who do not have the risk of fibrosis) maintain a healthy state that does not progress to fibrosis, but they have a risk of CVD and incurred all-cause mortality. (ii) In the H0 and F0-F4, all-cause mortality in the general population was applied; mortality due to liver disease was not additively applied. There was no HCC risk in the H0 and F0-F2 population. (iii) All patients diagnosed with advanced fibrosis received the ILI program. While receiving ILI, patients were monitored for advanced fibrosis regardless of TP or FP. (iv) The effect of ILI was reflected in the newly development of CVD and extrahepatic malignancies, and hepatic fibrosis regression. The regression rate of hepatic fibrosis (F1-F3) due to ILI was applied at 17.0%, and it was assumed that hepatic fibrosis regression would occur at 7.4% even in the control group24. The difference of hepatic fibrosis regression rate between in the ILI group and control group was 9.6%12,25,26 (v). The gap in regression rate by ILI between ‘screening’ and ‘no-screening’ groups lasted only for 10 years, the duration of receiving ILI. And after stopping the ILI program, the population of screening group had the same risk as no-screening group.
Cost and quality of life
We performed the analysis from the healthcare system perspective, and all the cost parameters were adjusted to the 2022 US dollars. An annual discount rate of 3% was applied to all costs and outcomes. Detailed cost parameters and utility weights of health states to calculate QALY were described in Supplementary Tables 3 and Supplementary Methods section.
Analysis
The primary outcome was the incremental cost effectiveness ratio (ICER) between each strategy, which was calculated by dividing the incremental cost by the incremental QALY (or LY) between comparison strategies. We simulated our model by applying different screening ages between 30 and 39, 40–49, 50–59 and 60–69 years. According to their age group, distribution of fibrosis and performance of FIB-4 were applied differently (Table 1). The results were interpreted as cost-effective when the expected ICER was less than $100,000/QALY, which was implicitly accepted as the willingness-to-pay (WTP) threshold in the US27.
Deterministic and probabilistic sensitivity analyzes (PSA) were performed to explore the impact of uncertainty in parameter estimates and assumptions applied to the analytic model. One-way sensitivity analyses were conducted by varying values over clinically relevant ranges (Supplementary Tables 1 and Supplementary Table 3). We also performed a two-way sensitivity analysis to investigate the impact of simultaneous variation by the ILI effect (regression rate) and duration on cost-effectiveness. And PSA using second-order Monte Carlo simulations was performed to evaluate the overall impact of uncertainty. Beta distributions were applied for transition probabilities and utility weights, and gamma distributions for costs.
link