Overview
We developed an individual-level Markov model to simulate HBV infection outcomes in individuals with chronic HBV infection from 2006 to 2050. The model projects the prevalence of total chronic HBV infection, HBsAg loss population, incidences of progression to cirrhosis, decompensated cirrhosis (DC), HCC, liver transplantation (LT), and HBV-related deaths. Model construction was performed using TreeAge Pro 2011 Suite (TreeAge Software, Williamstown, MA) (see Supplementary Figure 1). Data analysis was performed using R version 3.5.3. Figures were drawn with R version 3.5.3 plus Tableau Desktop 2018 (Tableau Software, Inc. Seattle, WA).
Patient demographics
Our study constructed a simulated chronic HBV-infected population representing a nationwide baseline in 2006 using the results of the national survey of chronic HBV infection epidemiologic study [20, 21] (Supplementary Tables 1-2). New HBV infections after 2006 were added to the simulated population annually based on reports by the Chinese Centers for Disease Control and Prevention, and the age distribution was derived from published data. Since there are no data regarding age structure after 2017, we adopted the same distribution as that adopted in 2017 [22-24] (Supplementary Table 3).
Model schematics
In the natural history scenario, we simplified and organized chronic HBV infection status into the following: three ‘CHB states’, including HBsAg-positive inactive carriers, HBeAg-negative hepatitis, HBeAg-positive hepatitis; four ‘complication states’, including cirrhosis, DC, HCC, and LT; and three absorbing states, including HBsAg losses, background deaths, and HBV-related deaths (Supplementary Figure 2). Each ‘CHB state’, at predefined probability rates, could either progress into cirrhosis and HCC states or undergo HBsAg loss or background death as absorbing states. All of the ‘complication states’ could transfer to each other or undergo HBV-related death (Supplementary Table 4). We also modeled the annual probability of background death caused by non-liver disease, and the background mortality of the age-standardized rate (ASR) was estimated from the National Bureau of Statistics of China [25].
We did not consider immune tolerance as one of the states here, which was common among infants and children, due to less epidemiologic data being available and the largest population comprising chronic HBV infection in adults who were inactive carriers in China [5, 26-28].
We also assumed that those who met the treatment criteria in the Chinese Medical Association and Asian Pacific Association for the Study of the Liver (APASL) guidelines for HBeAg-negative hepatitis, HBeAg-positive hepatitis and cirrhosis states could receive NA therapy and achieve a virological response (defined as undetectable serum HBV DNA during therapy), thereafter progressing at different probability rates (see Supplementary Table 5) [4, 29]. For those who did not achieve a virological response, we assumed that they had the same progression as that observed in the natural history scenario. For DC, HCC, or LT patients, we assumed that the progression rate was relatively constant regardless of NA therapy.
We did not use different strategies for HBV vaccination in our model because it has already been covered universally, with more than 95% coverage at present, and the blocking of vertical transmission is also highly effective, with an incidence of less than 10/100,000 new HBV infections [30, 31]. Meanwhile, the vaccine’s effect has already been shown as the new infection number every year in our model. We simulated only NA therapy, electing not to include interferon therapy due to its finite coverage, numerous side effects and contraindications [13].
HBV diagnosis, awareness, and treatment
During the simulation, we assumed that newly enrolled individuals were neither diagnosed nor aware at first, and they could become diagnosed with a certain probability. Once diagnosed, patients were assumed to accept their condition and awareness of HBV infection. A certain proportion of diagnosed patients could receive therapy. Newly treated individuals were divided into virological response or non-virological response states, which were permanent until simulation ended.
Simulation scenarios
Considering the current problems and improvement plans of diagnosis and therapy, we designed five scenarios to simulate different diagnosis rates, treatment rates, treatment eligibility, and new infection numbers. The clinical characteristics of the simulated scenarios are summarized in Table 1, and the simulation parameters are shown in Table 2.
***Table 1 should appear here in the text file***
he five scenarios were as follows: 1) Natural history scenario: No interventions (diagnosis or treatment) were applied; 2) Base case scenario: The current diagnosis and treatment rate simulated were applied from 2004 to 2050; 3) WHO-proposed target scenario: Gradually increased diagnosis and treatment rates and a reduction in the number of new infections were proposed by the WHO, which illustrated an improved diagnosis/treatment rate (Supplementary Tables 3 and 6); 4) Ideal scenario 1—full capacity of diagnosis and treatment: This scenario simulated that all existing HBV-infected patients would be diagnosed, and all treatment-eligible patients could receive and benefit from NA therapy, but annual new infection cases and treatment eligibility were not changed compared with those of the WHO target scenario (Supplementary Table 3); and 5) Ideal case 2—full treatment eligibility: This scenario simulated that all hepatitis and cirrhosis patients were treatment-eligible, but the annual new infection number, diagnosis rate, and treatment rate were unchanged compared with those of the WHO target scenario (Supplementary Tables 3 and 6). In brief, among all the scenarios, the base case scenario represented the current problems of diagnosis and therapy. The WHO targeted scenario and ideal scenario 1 represented a gradually and rapidly increased diagnosis and treatment rate, respectively. Ideal scenario 2 mainly represented expanded treatment eligibility.
***Table 2 should appear here in the text file***
Validation
We validated our model’s results using authoritative public health data sources, which included the following: the annual cirrhosis and DC incidence from the Institute for Health Metrics and Evaluation (IHME) Global Health Data Exchange (GHDx) online database; the annual HCC incidence from WHO CI5plus/IARC 2010-2012, WHO Globocan 2018, Polaris online database; the annual HBV-related death with WHO Globocan 2018 online database; the annual LT incidence from the China Liver Transplant Registry (CLTR) online database [32-36]; the annual number of cirrhosis deaths from published global disease burden studies [37, 38]; and the total chronic HBV infection prevalence from studies published by Chinese hepatology experts for the base case scenario [22]. In addition, in the natural history scenario, we compared our predicted cumulative 10-year probabilities of HBsAg loss and chronic hepatitis with those in published studies on inactive carriers [39] (Supplementary S 1.5).
Sensitivity analysis
We performed a 1-way sensitivity analysis on the model in base-case parameters by considering uncertainty in all probability rates. We defined each annual transition probability using the upper range and lower range values and then compared the upper or lower values of cirrhosis, DC, HCC, and LT cumulative incidence and cumulative death with base-case values (Supplementary S 1.6).