Economic evaluation of hepatitis A vaccines by income level of the country: A systematic review

Hepatitis A is a liver disease caused by hepatitis A virus (HAV) infection, which belongs to the Picornaviridae family1. HAV is transmitted through the ingestion of contaminated food and water or even by close physical contact with an infected person2. Once a person gets infected with HAV, lifelong immunity develops3,4. A person with hepatitis A infection may have an asymptomatic state, or may develop symptoms such as fever, nausea or vomiting, abdominal discomfort, jaundice and acute liver failure. Nevertheless, it does not progress to chronic hepatitis2. Unlike hepatitis B and C, hepatitis A is rarely fatal2.

The World Health Organization (WHO) estimates have suggested an increase in the number of acute hepatitis A cases from 117 million in 1990 to 126 million in 2005 with increase in deaths due to hepatitis A from 30,283 (in 1990) to 35,245 (in 2005)5,6. A global seroprevalence study on hepatitis A estimates an intermediate or low, level of endemicity in middle-income countries (MICs) from Asia, Eastern Europe, Latin America and the Middle East6. On the other hand, high-income countries (HICs) generally have low levels of HAV endemicity7.

The severity of hepatitis A infection increases with age, leading to a higher rate of severe disease and death in adults2. In low-income countries (LICs), which usually have a high level of endemicity, nearly all children get infected at an early age and are usually asymptomatic2.

In regions with intermediate endemicity, improved sanitary conditions may lead to the accumulation of adults who have never been infected, hence, have no immunity. These individuals in older age groups, therefore, are at a high risk of symptomatic hepatitis A infection8. Recently, the increasing burden of hepatitis A disease is noted in the regions with intermediate endemicity; thus, the countries in these regions may benefit from new/expanded vaccination programmes8.

HAV vaccination is considered as an effective and safe method to prevent hepatitis A infection2. Worldwide, two types of HAV vaccines (formaldehyde inactivated and live attenuated vaccines) are available2. The WHO recommends HAV vaccination to be integrated into the national immunization schedule for children aged more than one year based on the incidence of hepatitis A, change in endemicity from high to intermediate and considering the cost-effectiveness of the vaccination strategy8.

Economic evaluation (EE) is the comparative analysis of two or more interventions in terms of their costs and consequences9. Three main types of EE methods are cost-effectiveness analysis (CEA), cost-utility analysis (CUA) and cost-benefit analysis (CBA)9. In CEA, cost of each intervention is measured against its effectiveness (e.g. cost per case prevented, cost per life year gained). For CUA, cost incurred in the intervention is measured against the common unit, called quality-adjusted life year (QALY) (e.g. cost per 1 QALY gained). One QALY means one year in full health. For CBA, both cost and consequences of an intervention are expressed in monetary units. Then, the net benefit can be calculated as the difference between cost and consequences9.

To compare the alternative intervention over a long timeframe, modelling techniques have been adopted. Modelling offers several advantages including extrapolation beyond data generated through a trial, synthesizing head-to-head comparisons wherever relevant and linking intermediate endpoint to final outcomes. The most common modelling approaches used in EE studies are decision tree and Markov model10. Unlike decision tree model, Markov model is suitable when timeframe is long, process of disease is complex and events may repeat10.

Evidence generated through EE is important to inform effective healthcare resource allocation. Nevertheless, the capacity to conduct economic studies in many countries is limited11. To date, three systematic reviews on EE of HAV vaccination have been published12–14. The most comprehensive study12 published in 2018 included four studies from MICs and 27 studies from HICs. Another systematic review included nine studies conducted in MICs, which were published till 201213. The other identified 11 EE studies, were published between 1995 and 201014. It should also be noted that methodological characteristics were not fully described in the previous reviews, making it challenging to assess the transferability of the results. Therefore, the aim of this study was to systematically review evidences on cost-effectiveness of hepatitis A vaccination along with epidemiologic parameters and methodological characteristics. Cost-effectiveness evidences were also summarized by the types of population, intervention and income level of the countries.

Material & Methods

This systematic review was conducted as per the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines15. The protocol for this review was registered with the International Prospective Register of Systematic Reviews (PROSPERO; CRD42018105279).

Search strategy: Relevant studies were identified from PubMed and Scopus database without language restriction from inception to May 31, 2021. For studies other than the English language, help of the language translator from Mahidol University, Bangkok, Thailand was sought. Reference lists of the included studies and previous systematic reviews12–14,16 were also screened. Search terms were constructed based on intervention (I), outcome (O) and study design (S). These were combined using Boolean operators ‘OR’, ‘AND’ for within the same and between the domains, respectively. Both keywords and MeSH terms were used. The full details of search terms and strategies are given in Supplementary Appendix I.

Selection criteria: Duplicate articles were removed by EndNoteX9 software [Camelot UK Bidco Lmtd. (Clarivate analysis), Bangalore, Karnataka, India]. Study selection was performed independently by two authors. Titles and abstracts were screened for potential eligibility. The following criteria were used for screening: (i) full EE comparing HAV vaccine (inactivated or attenuated) to no vaccine or immunoglobulin and, (ii) reported findings in terms of cost per case prevented or incremental cost-effectiveness ratio (ICER) or benefit-to-cost ratio. Studies were excluded if HAV vaccine was investigated in combination with other vaccines, animal studies or studies which reported only clinical effectiveness or disease burden or outbreak investigations or if their fulltext were unavailable. In addition, narrative reviews, systematic reviews, editorial publications, and conference proceedings were also excluded.

Data extraction and quality assessment: Data were extracted independently by two authors using a predesigned data extraction form (Supplementary Appendix II). Any disagreement was resolved by discussion and consensus with a third author. The data extracted included were study and population characteristics, vaccination and comparator details (i.e. vaccine efficacy, vaccination approach), epidemiological parameters (i.e. incidence of HAV), methodological details (i.e. perspective, time horizon, discounting and sensitivity analysis) and EE results.

Risk of bias assessment was performed using the ECOBIAS checklist, which was developed for assessing bias in EE studies17. This 22-item checklist consists of two parts. Part A is related to overall bias, while Part B focusses on model-specific aspects of bias. Results for each item were recorded as ‘yes’, ‘partly addressed’, ‘unclear’, ‘no’ and ‘not applicable’.

Statistical analysis: Descriptive synthesis and narrative summary of study characteristics, participants, interventions, methodology and EE findings were reported according to the income level of the country studied as per the World Bank Report18. Countries were also classified into regions according to the WHO19. According to the World Bank Report 2017, the world’s economies were classified into four income groups based on Gross National Income per capita (current US $) as: LICs (<1005 $), lower-middle-income countries (LMICs) (1006-3955 $), upper-middle income countries (3956-12,235 $) and HICs (>12,235 $)18.

Results

Search results and study characteristics: Of the 1984 studies identified, a total of 43 eligible studies (40 English language and 3 Chinese language) were included in this review. The PRISMA flow diagram for study selection is shown in the Figure with preferred reporting items (Supplementary Appendix III). Selected studies were from 17 different countries: Argentina (2)20,21, Belgium (3)22–24, Brazil (1)25, Canada (1)26 Chile (2)27,28, China (4)29–32, France (1)33, Germany (1)34, Indonesia (1)35, Israel (3)36–38, Jordan (1)39, Mexico (2)40,41, Netherlands (2)42,43, Spain (1)44, Thailand (2)45,46, United Kingdom (1)47, USA (15)25,48–61 and a multi-country study from developed countries62. One study25 was conducted in the USA and Brazil (Table I). Among these, the majority (27/43, 62.8%) were from HICs, followed by MICs (15/43, 34.9%) and LICs (1/43, 2.3%) (Table I).

Close

Figure

PRISMA flow diagram for study selection.

Export to PPT

As per the WHO regions, the majority of the studies were from America (21/43, 48.8%) followed by Europe (9/43, 20.9%), Eastern Mediterranean (4/43, 9.3%), Western Pacific (4/43, 9.3%) and South-East Asia (3/43, 7.0%). It was not possible to classify a multi-country study from developed countries (1/43, 2.3%). Four studies were published before the licensure of hepatitis A vaccine in 1995 (Table I).

In terms of population, 27 studies were conducted in the general population while 16 were conducted in the specific risk group populations. Of those conducted in the general population, 19, five and three focused on children, adolescents and adults, respectively. Studies conducted among specific risk group population included military personnel (n=4), travellers (n=5), medical students (n=1), healthcare workers (n=3), people with hepatitis B infection (n=1), people with hepatitis C infection (n=2), day-care personnel (n=1), food-handlers (n=1) and homosexuals (n=1) (Table I).

Type of EE studies were CUA (18/43, 41.9%), CEA (14/43, 32.6%), both CUA and CEA (3/43, 7%), CBA (5/43, 11.6%) and CBA and CEA (3/43, 7%). Most studies used Markov model (13/43, 30.2%) followed by Markov model with decision tree (6/43, 14.0%), decision tree (7/43, 16.3%), dynamic model (6/43, 14.0%) and decision tree with dynamic model (2/43, 4.7%) (Table I). Most studies adopted societal (28/43, 65%) and healthcare provider perspective (14/43, 33%). However, nine studies (21%) did not mention the perspective.

Vaccine intervention: The summary of vaccination parameters is reported in Table II. All studies used attenuated hepatitis A vaccine as an intervention. Ten studies disclosed the name of the manufacturer. Vaccine efficacy ranged from 87.3 to 100 per cent (Table II).

Epidemiological parameters: As shown in Table II, the incidence of HAV was reported in about half of the studies (22/43), while the seroprevalence was reported in 12 studies. The incidence of HAV varied widely from 1.5 per 100,00033 to 1130 per 100,000 population53. The seroprevalence varied from 0.1-4 per cent44 to 91-94 per cent24. Only 10 studies considered herd immunity in the analysis20,27,28,36,39,40,49,50,54,61.

Risk of bias assessment: The risk of bias assessment for this study is shown in Supplemantary Appendix IV. All included studies had adequate comparators. Only 23.3 per cent (10/43) of the studies adopted a lifetime horizon, while 27.9 per cent (12/43) did not specify a time horizon. In terms of perspective, only 67.4 per cent (29/43) adopted a societal perspective, while about 16.3 per cent (7/43) did not specify the perspective. The discounting rate was not specified in 18.8 per cent of the studies (8/43). Of the 20 studies that disclosed funding sources, 11 were funded by pharmaceutical companies. Eight studies were subjected to risk of bias related to sensitivity analysis. One-way sensitivity analysis was adopted in 72.1 per cent (31/43) of the studies, while probabilistic sensitivity analysis was conducted in only 16.3 per cent (7/43) of the studies. Among CUA studies, 85.7 per cent (18/21) had a partial risk of bias related to quality of life weight. Eleven studies (25.6%) had an unclear risk of double-counting biases. Double-counting occurred when a parameter was counted more than once. It usually occurs in CUA, when consequences of an intervention (i.e. productivity loss/time loss) get incorporated on the cost side (numerator) as well as on the consequences side, i.e. QALY (denominator). All studies in this review had an unclear risk of biases related to internal consistency.

Cost-effectiveness findings: These are summarized in Table III. Summary of cost-effectiveness results by income level of the country, type of population and vaccination strategies is shown in Table IV.

For universal vaccination strategy, 70 (7/10), 86.7 (13/15) and 100 per cent (1/1) of the studies conducted respectively, in HICs, MICs and LICs were found to be cost-effective. When examining the types of population, universal vaccination among children was more likely to be cost-effective than the other age groups. About 63 per cent (17/27) of the studies conducted in HICs found that universal vaccination more cost-effective as compared to no vaccination, i.e. 86.7 per cent (13/15) in MICs in contrast to the adult population where, universal vaccination was not found to be cost-effective in both HICs (0/3) and MICs (0/1). Only 50 per cent (1/2) of the studies, comparing screening and vaccination to no vaccination among children in MICs, were found to be cost-effective. On the other hand, screening and vaccination among children in HICs were not cost-effective (0/1).

Hepatitis A vaccine was proven to be cost-effective as compared to no immunization among hepatitis C virus patients, food handlers and the homosexual population in studies conducted in high-income nations that used a targeted vaccination strategy. The results from travellers, healthcare staff and military personnel were mixed. In studies comparing the cost-effectiveness of vaccines vs. no vaccine among travellers, healthcare workers and military people, 40, 33 and 75 per cent were shown to be cost-effective, respectively. In studies comparing screening and vaccination versus no vaccination, 50, 50 and 66.7 per cent were found to be cost-effective among the same categories, respectively (Table IV).

Discussion

The present study revealed that universal hepatitis A vaccination without screening among children, especially in MICs, was more likely to be cost-effective than no vaccine strategy. This finding was consistent with that of earlier studies12,14. This might probably be due to the fact that countries such as Argentina20, Brazil25, Chile28, China30–32, and Indonesia35 have intermediate endemicity. Only half of the studies with data from MICs found that screening and vaccination among children were cost-effective. However, only two such studies were identified in this review. Because of the high seroprevalence of HAV infection among children in MICs, the cost-effectiveness of screening and immunisation was less favoured. In both HICs and MICs, universal hepatitis A vaccination among adults either with or without screening was less likely to be cost-effective. Consistent findings as that was seen in the previous study12, cost-effectiveness evidences among specific risk group populations varied widely depending on the risk of HAV infectivity. It was found that among people with greater risk of acquiring an infection due to a particular occupation or lifestyle, hepatitis A vaccination was found to be economically attractive22,23,43,44,51–53,63,64.

Vaccine considered in the analysis, it should be noted that all studies used inactivated hepatitis A vaccine. However, live-attenuated hepatitis A vaccine has been developed in China since 200765. The vaccine is mainly marketed in China and India65. It was shown to have similar efficacy to that of inactivated vaccine66, but only one dose was required. With the assumption of similar price per dose and similar efficacy, cost-effectiveness evidence would likely favour live-attenuated vaccine.

The present review found that the most common biases identified were related to internal inconsistency in terms of methodological quality. This is similar to other studies67,68, which found that mathematical logic was not evaluated in most of the investigations. Although experts generally recommend using societal perspective, as it is more comprehensive69, a societal perspective was adopted only in 68 per cent of the studies; hence, the direct non-medical cost was not included in the analysis. In addition, we found that only 23 per cent of the studies adopted lifetime horizons. Furthermore, probabilistic sensitivity analysis was rarely conducted.

It should be noted that most of the studies did not use a dynamic model. Furthermore, herd immunity was not taken into account. In fact, a dynamic model was a necessity in deciding on implementing realistic universal vaccination strategies70. However, due to the unavailability of large epidemiological parameters in a local context, complex study design and lack of expertise, the dynamic models were not used widely by researchers. In addition, it should be noted that when herd immunity is not taken into account, cost-effectiveness evidences of vaccine may be underestimated.

Our review found that most included studies had partial bias related to quality of life weight. This was because most of the studies used secondary data with limited information on the methods used to estimate utility weight, as well as characteristics of the sample. In addition, data on utility weights for symptomatic and asymptomatic hepatitis A infection were limited.

It should be noted that cost-effectiveness studies need to be conducted using locally available epidemiological data as such data from other settings have low transferability13. Although the age-specific incidence of hepatitis A infection had a significant impact on cost-effectiveness finding12, we found that many included studies20,21,27,28,31 adopted such data from the US study54,55,71. However, it was suggested that if the hepatitis A incidence data were not available, seroprevalence data of the country could be used to estimate the incidence71,72. In the absence of local data, it is recommended that data from countries with similar endemicity may be used cautiously73. On the other hand, some parameters could be adopted from other countries or other studies. As the natural history of hepatitis A infection is similar across the countries, the probability of symptomatic infection (presented with jaundice) among infected individuals may be transferable from other studies13. Since the efficacy of HAV vaccination was not affected by ethnicity variation, vaccine efficacy data could be adopted from other studies.

In terms of study perspective, most of the studies with societal perspective indicated that HAV vaccination was cost-effective. Studies with societal perspectives, in which HAV vaccination was not found to be cost-effective, were conducted in HICs24,34,42,56,58. For the studies that used both societal perspective and healthcare provider perspective, the results from societal perspective were more favourable towards cost-effectiveness or even cost-saving.

The present systematic review could not identify any EE study on HAV conducted in India. India is considered as a LMIC with wide variation in terms of socio-economic status. Due to rapid improvement in sociodemographic development in India during the past decade, there is evidence of a shift from high to intermediate endemicity, especially in the high-income region. In such region, with the decreasing number of adolescents with prior exposure to HAV, several hepatitis A outbreaks have been reported74–76. According to our review, almost all studies conducted among children in MICs, which were also facing improvement in sociodemographic development, found that HAV vaccination was cost-effective. Therefore, it is likely that HAV vaccination would be cost-effective in India, especially in the regions with reported shift from high to intermediate endemicity. In these regions, policymakers working on HAV vaccination may consider inclusion of HAV vaccination in public insurance schemes.

In summary, our study provides updated cost-effectiveness evidences of hepatitis A vaccination. Based on the existing evidence, we found that universal vaccination among children was more likely to be cost-effective, especially in MICs. Nevertheless, our study had some limitations. First, evidences on LICs and live-attenuated vaccines were limited. Second, as the presented ICERs varied by type of currency, year of valuation and types of outcome, direct comparisons could not be made. Third, most studies had partial biases on both epidemiological parameters and quality of life weights; therefore, further studies that aim to estimate such parameters are warranted to ensure the accuracy of cost-effectiveness evidences. Finally, transferability of the cost-effectiveness findings of hepatitis A vaccine should be made after careful consideration of epidemiological parameters, resource utilization, unit cost data, as well as structure of healthcare delivery system, and country-level income.