Cost-Effectiveness of Hepatitis B Mass Screening and Management in High-Prevalent Rural China: A Model Study From 2020 to 2049

Background: Chronic hepatitis B (CHB) is highly prevalent among adults in rural China and better management of those populations is of vital importance for viral hepatitis elimination. Adult immunization has been the subject of much controversy in previous studies. This study estimates the cost-effectiveness of population-based hepatitis B screening, treatment, and immunization strategy (comprehensive strategy) in rural areas with high prevalence under the national policy of sharp-drop drug prices. Methods: We constructed a Markov model comparing 4 strategies in a 30-year horizon from the healthcare payer perspective: (1) the conventional pattern; (2) screening and treating infected (treatment); (3) screening and immunizing susceptible individuals (immunization); and (4) the comprehensive strategy. Screening intensity ranged from 50% to 100%. Outcomes were measured by costs, quality-adjusted life-years (QALYs), incremental cost-effectiveness ratios (ICERs), and clinical outcomes. Results: The costs for the conventional pattern, treatment strategy, immunization strategy, and comprehensive strategy were US$ 341, 351, 318, and 323, respectively. In addition, effects were 17.45, 17.57, 17.46, and 17.58 QALYs, respectively. The ICER of the comprehensive strategy was US$ 35/QALY gained at 50% screening intensity and 420 US$/QALY gained at 100%. The net monetary benefit increased with increasing screening intensity and declined after 90%, with the highest value of US$40 693. All new infections and 52.5% mortality could be avoided from 2020 to 2049 if all patients were properly treated and all susceptible individuals were immunized. The results were stable within a wide range of parameters. Conclusion: It was cost-effective to implement the mass hepatitis B screening, treatment, and immunization strategy in areas of rural China with high prevalence, and the strategy gained the most net monetary benefit at a screening intensity of 90%. Although it was impractical to fulfill 100% coverage, efforts should be made to obtain more people screened.


Model assumptions
Modeling natural history: (1) Except for the mutual transition between CHB and inactive HBsAg carriers, the transition of other disease states was irreversible ( Figure S1), 1 and the transition probability between each state was fixed each year. (2) Individuals with HBsAb titer above 10 mIU/mL would not be infected with HBV.
(3) Adult-acquired HBV infection entered the acute hepatitis B state and transitioned into either CHB or HBsAg clearance state. Asymptomatic and symptomatic acute hepatitis, fulminant hepatitis, and death from fulminant hepatitis were modeled in this state. Liver transplantation caused by fulminant hepatitis was not considered in this state because of the low incidence in this population.  (6) HBV reoccurrence in patients who suffered liver transplantation was not modeled.
Modeling screening: 3 (1) The baseline screening intensity was 50%, representing half of the population were willing to participate in screening currently.
(2) If more people were mobilized to participate in screening, more resources must be consumed, which was reflected by the higher cost of screening. We assumed screening cost increased by 50% with each additional 10% population screened.
(3) We didn't consider the probability of false negatives or false positives with screening tests.
Modeling treatment: (1) In the conventional pattern, rural patients don't go to a hospital for health examinations on their initiative until exacerbation. While symptoms were obscure in carriers and CHB patients which were discovered in the later stage in most cases, so it was assumed that there was no treatment cost in these groups in the conventional pattern.
(2) In the treatment or comprehensive strategy, carriers and CHB patients were discovered and managed according to the guideline. There would be outpatient costs from these patients, but no hospitalization.
(3) We hypothesized that comprehensive management of HBV carriers with regular follow-up could reduce their incidence of CHB, liver cirrhosis, and HCC by 50%. Modeling immunization: (1) By default, those who were willing to be vaccinated would accomplish all three doses, and those with HBsAb titer > 10 mIU/mL would not be infected. 4 According to the HBV serological markers, HBV DNA quantification, alpha fetoprotein, alanine transaminase or aspartate aminotransferase abnormality, liver fibrosis degree, and liver occupation or not, the progression after HBV infected was divided into 10 states and the simplified Markov model was shown in Figure S2. The disease distribution of the initial state of the cohort entered the model was assumed with similar data in former studies. Ratios for immune tolerant, chronic hepatitis B (CHB), inactive HBsAg carriers, compensated cirrhosis, and decompensated cirrhosis were 1.5%, 5%, 90.5%, 2.5%, and 0.5%, respectively. 2 And the transition probability between above states were referred from published researches. We used the following formula to calculate the annual transition rate when it was unavailable.

Model calibration and validation
We compared the HBV infection progress in natural history and after antiviral treatment with published epidemiological surveys or research articles. The distribution for CHB patients at the initial state of the infection cohort was set to 1 to simulate the incidence of liver cirrhosis and HCC in the population after 5 and 48 years, respectively.
Similarly, we set the distribution for inactive HBsAg carriers to 1 and calculate the cumulative incidence of HBsAg loss and hepatitis recurrence. The distribution for compensated cirrhosis was set to 1 to calculate the cumulative incidence of decompensation and HCC. Besides, cumulative mortalities in CHB, cirrhosis, and HCC were also simulated. To be mentioned, we assumed that the annual mortality rate of HCC kept the same in the 4 comparators for patients with HCC would seek for help from doctors initiatively in the conventional pattern, and treatment was unnecessary for inactive carriers. Hence, we didn't simulate the incidence after antiviral treatment in these circumstances. We conducted the Markov cohort analysis of a fixed cohort of 1000 people with the following results in Table S4. It was consistent with the results of published cohort studies or observational studies, indicating that the Markov model we established was in line with real-world data and feasible for the subsequent costeffectiveness analysis.