A cost-utility analysis was conducted to compare the “one-time universal screening in the Korean population aged 40–65 years using anti-HCV,” provided as a part of the NHE program, to “no screening.”

An established model reflecting the actual clinical setting and natural history of HCV infection was used, which was developed with the TreeAge Pro program (TreeAge Software, Williamstown, MA, USA) [9]. Some input parameters were updated, and the model was partially modified to reflect current practices such as prescribing trends and reduced prices of DAA agents. The model had two parts: a decision tree model and a Markov model. In the decision tree model, the population was divided into subgroups of “treatment state (F0-F4),” “diagnosed state (F0-F4),” “undiagnosed state (F0-F4),” and “alive without HCV” according to the prevalence of HCV antibodies, test performance of HCV antibody, HCV RNA positivity, and acceptance rates of screening and treatment (Fig. 1A). Then, the population was entered into the Markov model and moved to 28 predefined health states, including chronic hepatitis with fibrosis stage 0–3, compensated cirrhosis (F4), decompensated cirrhosis (DC), HCC, SVR, and liver transplantation (LT). In every health state, patients were at the risk of death from non-liver disease (age-related mortality) or liver disease (disease-specific mortality in DC, HCC, or LT). The cycle length of the Markov model was 1 year, and the lifetime horizon was selected. In each cycle, patients either remained in their current health state (recursive arrow) or progressed to another health state (straight arrow) according to the transition probability (Fig. 1B).

The following assumptions were made in our analysis. 1) Only one-time screening for HCV infection was provided to the Korean population aged 40–65 years during the biannual NHE. 2) Half of the patients were treated in the year of diagnosis, while the other half were treated in the following year. 3) Patients diagnosed from screening would have mild or no symptoms; hence, there were no symptomatic cases of decompensated liver cirrhosis at the time points of simulation entry. 4) Although there is insufficient data on HCV re-infection, it is considered to be very rare in Korea. Therefore, we assumed an absence of re-infection in our model.

Studies in South Korea and other countries on epidemiology and treatment effects on patients with chronic HCV, analysis of health insurance claim data, and government statistics were extensively used. The optimal values were determined after discussion with the researchers. Data from South Korea were primarily used. The input parameters used in the model are listed in Table 1 and Supplementary Table 2.

Screening was performed by testing anti-HCV with 98.1% sensitivity and 99.8% specificity. The assumed acceptance rate of screening was 77.4% based on a report from the Korean National Health Insurance Service (KNHIS) [14]. HCV viremia was estimated to be present in 46.5% of people showing a positive anti-HCV [9,15]. The population structure by age was extracted from the 2019 population census [16]; the prevalence of anti-HCV by age group was estimated to be 0.38%, 0.63%, and 1.06% in 40s, 50s, and 60s, respectively [6]. The annual detection rate of HCV infection without screening was applied as 3.8–5.6% [17,18].

The current prescription profiles of various DAA regimens were estimated using the data extracted from the 2019 Korean HCV cohort study. Specifically, glecaprevir/pibrentasvir for 8 weeks was the most commonly prescribed regimen (74.8%), followed by ledipasvir/sofosbuvir (11.2%), elbasvir/grazoprevir (10.5%), and sofosbuvir and ribavirin (3.5%) for 12 weeks each. The SVR rates of each regimen were obtained from the literature review [17,19-22].

Transition probabilities for movement between health states were determined as the best optimal values by authors through previous literature reviews or from the Korean government databases [23-34].

The direct medical costs related to each health state were extracted from the KNHIS claim data analysis and reimbursement price list from HIRA. Direct medical costs included total fees for hospitalization, consultation, medication, examination, and other management. The operational definitions for each health state to estimate medical costs using claims data are presented in Supplementary Table 3. Indirect costs associated with productivity loss due to premature deaths were calculated using the average number of working days, average wage, and employment rate by age group from the Korean government database. Transportation costs resulting from hospital visits were also estimated according to the health states. All costs were measured in Korean won (KRW) and adjusted for inflation using the Korean consumer price index to reflect the 2020 KRW; they were converted to United States dollars ($) using the annual average currency rate in 2020 ($1=1,180 KRW) (Supplementary Table 3).

Utility weights for health states were obtained from published studies [9,23,26]. Age-specific utility weights in the general population were derived from the EQ-5D values in the Korean National Health and Nutrition Examination Survey 2018 (Supplementary Tables 2, 4).

The main output was the incremental cost-effectiveness ratio (ICER) of the “universal HCV screening” strategy compared to the “no screening” strategy. The ICER was calculated by dividing the incremental costs by the incremental QALY (or LY) between the comparative strategies. When the estimated ICER was less than $25,000/QALY, implicitly accepted as a willingness-to-pay (WTP) threshold in Korea, the universal screening was considered to be cost-effective.

One-way deterministic and probabilistic sensitivity analyses were performed to explore the uncertainty of the input parameters and applied assumptions. For one-way sensitivity analyses, the following variables related to HCV screening and the natural course of HCV infection were included: the prevalence of anti-HCV, detection rate without HCV screening, age, acceptability of treatment, SVR rate of DAA therapy, test fee, medical cost, utility weight, and discount rate.

A probabilistic sensitivity analysis, using a second-order Monte Carlo simulation, was performed to evaluate the overall effect of uncertainty. The simulation was run 10,000 times with parameter values randomly generated from the relevant distributions. We applied a beta distribution for transition probabilities and utility weights and a gamma distribution for the costs (Supplementary Table 5). The results are presented as a cost-effectiveness acceptability curve (CEAC).

In the base-case analysis from a healthcare system perspective, including direct medical costs, the universal HCV screening and DAA treatment group gained 0.0014 QALY (and 0.0010 LY) per patient compared to the “no screening” group over the lifetime horizon. The difference in the total costs per patient was $3.6, and the estimated ICER were $2,666/QALY and $3,653/LY (Table 2).

The cost-effectiveness results from a societal perspective, including the cost of productivity loss due to premature deaths and transportation cost, as well as direct medical costs, showed that universal HCV screening and DAA treatment compared to no screening would raise the QALY by 0.0014 (and LY by 0.0010) and the cost by $0.58, resulting in an ICER of $431/QALY and $590/LY. Therefore, the results from the societal perspective showed lower ICER than those from the healthcare system perspective. As the estimated ICERs ($2,666/QALY and $431/QALY) were far less than the threshold of $25,000/QALY, universal screening was highly cost-effective compared to the no screening strategy.

The results of the one-way sensitivity analyses are presented in Table 3 and Figure 2. The most influential parameter on the ICERs was the anti-HCV prevalence. As the prevalence of anti-HCV antibodies increased, the ICER decreased remarkably (Fig. 2). The changes in ICER according to various anti-HCV prevalence are shown in Figure 3; in the analysis from a healthcare system perspective, the ICER was $15,170/QALY with 0.2% anti-HCV prevalence and $527/QALY with 2.0% of it. Similarly, in the analysis from a societal perspective, the trend of ICER changed according to anti-HCV prevalence. With the anti-HCV prevalence over 0.179–0.186%, universal screening was cost-effective (ICER <$25,000/QALY) from both the healthcare system and societal perspectives (Fig. 3).

Regarding the screening costs, it was observed that when the high-quality immunoassay test (National Health Insurance [NHI] fee code: D7026, $13.61) was substituted for the anti-HCV test (D7005, $3.91), the ICER increased to $8,199/QALY. Although our baseline anti-HCV test is adequate for population screening, many healthcare centers in Korea currently use the high-quality immunoassay test. For the detection rate without HCV screening, the ICER was $1,841/QALY with 2% and $5,174/QALY with 12% in the analysis from healthcare system perspective (in a societal perspective, -$0.54 and $3,731, respectively).

Other variables affecting the ICERs included the acceptance rate of DAA treatment, discount rate, utility weight for SVR, cohort age, SVR rate of DAA agent, utility weight for F0–F3, medical costs for HCC and DC, and HCV genotype testing.

The results of the probabilistic sensitivity analysis simulated 10,000 times are illustrated in a CEAC (Fig. 4) and a cost-effectiveness scatter plot (Supplementary Fig. 1). CEAC indicated that the probability of a one-time screening and DAA therapy being cost-effective was 60.3%, 81.8%, and 97.7% with a WTP of $1,000, $2,000, and $5,000, respectively. In the cost-effectiveness scatter plot, the probability was 99.9% at a WTP threshold of $25,000. This confirmed the robustness of the cost-effectiveness of universal screening strategy.

Under the base-case analytic conditions, the one-time universal screening for Korean populations aged 40–65 years (total 21,099,926) was estimated to detect 32,148 HCV infection cases additionally. Moreover, it was estimated to reduce 4,081 HCV-related deaths over a lifetime (19.4/100,000). The number of preventable HCCs, DCs, and LTs over a lifetime was 3,156, 1,939, and 554, respectively (Supplementary Table 6).