Introduction of birth dose of hepatitis B virus vaccine to the immunization program in Ethiopia: An economic evaluation

Background: Hepatitis B virus (HBV) infection is an important cause of morbidity and mortality with a very high burden in Africa. The risk of developing chronic infection is marked if the infection is acquired perinatally, which is largely preventable through a birth dose of HBV vaccine. We examined the cost-effectiveness of a birth dose of HBV vaccine in a medical setting in Ethiopia. Methods: We constructed a decision analytic model with a Markov process to estimate the costs and effects of a birth dose of HBV vaccine (the intervention), compared with current practices in Ethiopia. Current practice is pentavalent vaccination (DPT-HiB-HepB) administered at 6, 10 and 14 weeks after birth. We used disability-adjusted life years (DALYs) averted to quantify the health benefits while the costs of the intervention were expressed in 2018 USD. Analyses were based on Ethiopian epidemiological, demographic and cost data when available; otherwise we used a thorough literature review, in particular for assigning transition probabilities. Results: In Ethiopia, where the prevalence of HBV among pregnant women is 5%, adding a birth dose of HBV vaccine would present an incremental cost-effectiveness ratio (ICER) of USD 110 per DALY averted. The estimated ICER compares very favorably with a willingness-to-pay level of 0.31 times gross domestic product per capita (about USD 240 in 2018) in Ethiopia. Our ICER estimates were robust over a wide range of epidemiologic, vaccine effectiveness, vaccine coverage and cost parameter inputs. Conclusions: Based on our cost-effectiveness findings, introducing a birth dose of HBV vaccine in Ethiopia would likely be highly cost-effective. Such evidence could help guide policymakers in considering including HBV vaccine into Ethiopia’s essential health services package. We constructed a decision analytic model with a Markov process (Additional file 1 and Fig. 1) to estimate intervention costs and health impact of an infected individual over a lifetime. We used TreeAge Pro 2018 software for the analysis. As demonstrated in the Additional file 1 and Fig. 1, the model reflects the natural history of perinatally acquired HBV-infection [13]. We compared two strategies. In the novel strategy (HepB-BD vaccine-plus) all infants receive HepB-BD vaccine (monovalent) within 24 h of delivery and continue on with the pentavalent vaccine (DPT-HiB-HepB) series starting at the age of 6 weeks. We assumed 50% of the target birth cohort will be born in health care facilities and receive the birth dose in a medical setting; following the latest skilled birth attendance rate in Ethiopia [14]. We further assumed that the government rolls-out HepB-BD vaccine to the whole birth cohort and therefore incur costs but vaccine effectiveness would vary depending on the skilled birth attendance rate in Ethiopia. In the current strategy, the birth cohort will only receive the existing pentavalent vaccination schedule at 6, 10 and 14 weeks after birth. Markov process showing the different health states In the model, we assumed HepB-BD vaccine would prevent against MTCT of HBV infection during birth/delivery (vertical transmission), otherwise the two strategies would have similar efficacy on infections that occur later in life (horizontal transmissions). We used data on HBV prevalence among pregnant women, vaccine effectiveness, and the risk of perinatal transmission to calculate the percentage of children born with HBV infection. Infected infants either develop acute symptomatic infection with a risk of fulminant hepatitis or remain asymptomatic (Additional file 1). Even though symptomatic acute infections occur less frequently in infants, we accounted for their costs and health consequences in our model (see Additional file 1) [15, 16]. Perinatally infected infants develop acute symptomatic infections early in life, therefore we calculated disability-adjusted life years (DALYs)2 (based on life expectancy at birth adjusted for health state valuations from WHO-CHOICE) for only fatal cases and did not account for the lost quality of life for the duration of acute illness [17, 18]. The majority (90%) of the asymptomatic cases and about 33% of those surviving fulminant hepatitis will develop chronic HBV infection [4, 15]. Most perinatally infected individuals enter the immunoactive phase and develop HBeAg positive chronic hepatitis with elevated liver enzyme (alanine aminotransferase) levels only after 10–30 years of infection [19]. Therefore, we started the Markov process at age 20 years. Individuals with chronic HBV infection are simulated within the Markov process either as an inactive carrier or with chronic active hepatitis (CAH). Individuals who do not require antiviral therapy are considered inactive carriers and CAH are those who fulfill the treatment criteria (irrespective of the HBeAg status). A study in Ethiopia has evaluated treatment eligibility and response to antiviral management of chronic hepatitis B infection [20]. In this study, among individuals 18–25 years of age with chronic HBV infection, nearly a quarter of patients (25%) were eligible to antiviral treatment (ALT > 80 U/L and viral load > 2000 IU/mL). Therefore, in our model 75% of the individuals with chronic HBV infection began the Markov process as inactive carriers while 25% as CAH. Even though there are studies that document clearance of HBsAg (in 6–17% of the cases) in Caucasian children who acquired chronic HBV infection horizontally, such data for perinatally infected children in sub-Saharan Africa were lacking where there is limited access to antiviral treatment [21, 22]. Therefore, our model did not account for state transitions from either inactive carriers or CAH to HBsAg clearance (no infection). We ran the Markov process for 51 cycles, which corresponds to the average life expectancy of both males and females at age 20 years in Ethiopia [17]. Every Markov cycle lasts 1 year allowing infected individuals to pass through different morbid states based on their transition probabilities. In each annual cycle, infected individuals could incur costs related to medical care and health outcomes depending on their health state. Subsequently for each strategy, the costs and effects accrued in each of the decision trees and the Markov cycles are added and compared to calculate the incremental cost-effectiveness ratio (ICER). The overall prevalence of HBsAg among pregnant women in Ethiopia was estimated at 4.7% [3]. The risk of perinatal transmission varies by the maternal HBeAg status, where HBeAg-positive mothers carry a higher risk of transmission [4]. Prevalence of HBeAg among HBsAg-positive pregnant women in Ethiopia is unknown; therefore we used the mean prevalence from several sub-Saharan African countries [4]. Table 1 summarizes epidemiologic and probability data inputs used in our model. Individuals infected with HBV may develop a spectrum of disease conditions with different probabilities (see Fig. 1 and Additional file 1). Local data on transition probabilities among different health states were not available. Therefore we used data from settings that are similar to the Ethiopian context when available (Table 1) [23, 24]. Similar to what has been used in acute symptomatic infections, DALYs averted were the measure of effectiveness for chronic HBV states, inline with WHO recommendations since chronic HBV infection results in both premature mortality and morbidity [23, 25]. Evidence on disability weights for the different disease states were gathered from the Global Burden of Diseases (GBD) study database but other sources were also sought when such data were not available in the GBD database (Table 2) [24, 26]. Epidemiologic parameters and annual transition probabilities used in the model aIn our model, patients with chronic active hepatitis (CAH) are eligible for antiviral treatment therefore we used transition probabilities of treated patients bThe highest vaccine utilization rate was based on administrative report for facility delivery in Ethiopia Disability weights for different disease states used in the model aFan L, presented quality adjusted life years and we simply subtracted these values from one to compute disability weights bWe just took the average disability weights of the different HCC stages to calculate the annual HCC disability weight of 0.43 Vaccine effectiveness data based on a randomized controlled trial (RCT) were not available locally and from other comparable sub-Saharan African countries. Therefore we used data from other settings that were based on a systematic review of the efficacy of hepatitis B immunization for newborn infants of HBsAg-positive mothers where the protective efficacy of the vaccine was 72% (95% CI 60 to 80%) in preventing perinatal HBV transmission [27]. Vaccine adverse effects that are usually mild were not factored into the model [6]. Deaths due to other causes (background mortality) were integrated in the model using the World Health Organization’s life tables for Ethiopia [17]. Costs were estimated from a health provider perspective and only included direct medical costs. We estimated the incremental cost of introducing HepB-BD vaccine in its monovalent form. We preferred the monovalent form of the vaccine in order to minimize wastage rates. Both recurrent and capital costs were included using an ingredients-based approach following WHO guidelines [28]. Recurrent costs included costs of vaccines, syringes, safety boxes, transport and maintenance, and cold chain storage while social mobilization and training costs were under capital costs (Table 3). We used the latest UNICEF price data for vaccines, auto-disable (AD) syringes and safety box adjusted for wastage rates and freight costs [29, 30]. Data on wastage rates and freight costs were retrieved from a study on economic evaluation of HBV vaccine in low-income countries [23]. Local data on other aspects of recurrent costs (transport and maintenance and cold chain storage) and capital costs for HepB-BD vaccine were not available therefore we used 2008 estimates from Mozambique adjusted to the 2018 USD values [12, 31]. Vaccination cost estimates and intervention costs in 2018 US$ aVaccine wastage rate was assumed to be 20% and a freight rate of 6% bWastage and freight rates were 10% and 15%, respectively The medical care costs included initial assessment and diagnosis costs, antiviral drug costs, costs associated with monitoring those on treatment and not on treatment, and cost of managing decompensated cirrhosis (DCC) and hepatocellular carcinoma (HCC) (Table 3). In order to calculate the costs of initial assessment and diagnosis, we used the national guidelines on viral hepatitis to identify required laboratory tests and imaging modalities and frequency of health care visits [11]. We collected laboratory cost data from local sources. The median supplier price for the antiviral drug of choice (Tenofovir) was used to compute for annual drug cost after accounting for transportation costs [32, 33]. Hospital admission cost data for DCC and HCC were not locally available; therefore we used estimates from The Gambia adjusted to the 2018 USD values [31, 34]. Costs for acute symptomatic conditions were based on expert estimates. All costs were expressed in 2018 USD. Both future costs and health outcomes were discounted at 3% annual rate following WHO recommendations [25]. We conducted a series of one-way sensitivity analyses where we varied key input parameters one at a time over plausible ranges to test the robustness of our findings (Table 1). Based on the findings in a one-way sensitivity analysis, we proceeded and conducted two-way and three-way sensitivity analyses for parameters that are likely to change the result in a critical way. Furthermore, a multivariate sensitivity analysis was conducted using Monte Carlo simulations with n = 10,000 simulation runs. We varied all key parameters (vaccine effectiveness, vaccine utilization, risk of perinatal transmission in HBeAg-mothers, risk of perinatal transmission in HBeAg + mothers, prevalence of HBV infection among mothers, cost of medical care, average cost per vaccinated child, prevalence of HBeAg in pregnant women, transition probability of CAH to inactive carrier state) simultaneously [35]. Lastly, we also ran the model without discounting future health benefits or costs (one at a time), and without discounting both health benefits and costs [35].