Cost-effectiveness analysis of COVID-19 booster vaccination with BNT162b2 in Japan

ABSTRACT

Background

The aim of this study was to evaluate the public health and economic impact of the COVID-19 booster vaccination with BNT162b2 in Japan during an Omicron-dominant period from early 2022.

Research design and methods

A combined cohort Markov decision tree model estimated the cost-effectiveness of annual or biannual booster vaccination strategies compared to no booster vaccination for those aged 65 years and above, and those aged 60–64 years at high risk as the base case. The societal perspective was primarily considered. We also examined other target populations with different age and risk groups. Sensitivity and scenario analyses with alternative inputs were performed.

Results

Annual and biannual vaccination strategies were dominant from the societal perspective in the base case. Incremental Cost Effectiveness Ratios (ICERs) from the payer perspective were JPY 1,752,499/Quality Adjusted Life Year (QALY) for annual vaccination and JPY 2,831,878/QALY for biannual vaccination, both less than the threshold value in Japan (JPY 5 million/QALY). The results were consistent even when examining other target age and risk groups. All sensitivity and scenario analyses indicated that ICERs were below JPY 5 million/QALY.

Conclusions

Booster vaccination with the COVID-19 vaccine BNT162b2 is a dominant strategy and beneficial to public health in Japan.

KEYWORDS:

• COVID-19
• SARS-CoV-2
• vaccine
• booster
• cost-effectiveness
• Japan

1. Introduction

The World Health Organization (WHO) has reported approximately 771 million confirmed cases of COVID-19, including 6.96 million COVID-19 related deaths worldwide as of September 2023 [1]. In Japan, there have been more than 33.8 million confirmed cases of COVID-19 and nearly 74,700 deaths [1]. Since 2020, the prevention and control of the disease have disrupted societies and caused an enormous public health and economic burden in all countries, with Japan being no exception [1].

The contributions of vaccination to the prevention of COVID-19 infection and severe symptoms have been evident [2]. Globally, a total of 13 billion vaccine doses have been administered. Emerging evidence also indicates the need for booster vaccinations, which reduce symptomatic infections, hospital admissions, and severe COVID-19 outcomes compared to not having a booster vaccination [3–6] which is also consistent in Japan [7,8]. Further efforts are needed to encourage eligible individuals to be vaccinated and staying up to date with the recommended COVID-19 vaccines is also encouraged from a public health perspective [6].

The voluntary lockdown measures promoted by the government and voluntary behavioral changes successfully limited transmission rates, but came with a cost. The impact of COVID-19 went beyond the disease itself; the health service delivery was significantly impacted in Japan, especially during the early phase, disrupting both preventive and curative services for other diseases. The delayed essential healthcare services due to the pandemic affected many patients beyond COVID-19 [9,10].

In Japan, COVID-19 vaccinations have been categorized as ‘Special Temporary Vaccinations’ under the Immunization Act, and the cost has been entirely covered by the government [11]. The free vaccination program is scheduled to end in March 2024, with routine annual vaccinations for those aged 65 years and above, and those aged 60–64 years at high risk beginning in the fall of 2024 [12]. No economic evaluation of the COVID-19 vaccination in the Japanese population has been published to date. This study estimates the public health impacts and economic value of booster vaccination with BNT162b2, mRNA COVID-19 vaccine after the Omicron dominant period in Japan which is ongoing since early 2022.

2. Methods

2.1. Overview

A cost-effectiveness analysis was conducted comparing the booster vaccination with BNT162b2 with no booster vaccination during the Omicron dominant period starting in early 2022 in Japan. Since COVID-19 infection may cause home care or hospitalization, which could significantly affects work productivity for the patients and their caregivers [13], we primarily analyzed from the societal perspective to capture the broader impact of productivity loss. The model incorporated the costs and effects on the COVID-19 disease and the vaccination. The cost-effectiveness model builds on a combined cohort Markov model (Figure 1) and a decision tree [14]. This static model was developed based on the model previously used in Di Fusco et al. [15] and was programmed in Microsoft Excel. The intervention modeled was either an annual or biannual dose of the BNT162b2 booster. In Japan, an annual COVID-19 vaccination program is scheduled to begin in the fall of 2024. However, a study recommended a biannual vaccination in view of the severity of COVID-19 infection [16]. Therefore, our study evaluated the cost-effectiveness of both the annual and biannual vaccinations. The time horizon was set to 3 years to fully capture the impact of immunization through the booster vaccination. A 2% discount rate was applied to both the costs and effectiveness, in line with the Japanese cost-effectiveness evaluation guideline [17]. Ethical approval was not needed since this study was a model analysis with published secondary data and without individual patient data.

Figure 1. Markov cohort model structure.

In the Markov model component, individuals transition weekly through health states related to COVID-19 protection and infection. If individuals transition into any infection state, they immediately enter. Abbreviations. FBV, first booster vaccination; FPV, full primary vaccination.

We set up multiple hypothetical scenarios, including one scheduled for fall 2024 in Japan, in order to examine the cost-effectiveness of vaccination in Japan in a broad range of situations. The primary scenarios are as follows (Table 1).

1. Base case (those aged 65 years and above, and those aged 60−64 years at high risk): In the fall of 2024, COVID-19 vaccination is scheduled for those aged 65 years and above and those aged 60−64 years who are considered ‘at high risk,’ which aligns with the target population for the seasonal influenza vaccination [18]. The ‘at high risk’ population is defined as those who have heart, kidney, or respiratory function disorders that severely limit their daily activities or immune function caused by the human immunodeficiency virus that make daily life almost impossible. It is roughly equivalent to the definition of Grade 1 disability for the physically disabled in Japan. Certification is given in levels from 1 to 6 according to the degree of the impairment, with Grade 1 being the most severely impaired. This study used the MHLW survey and government statistics as the data source [18,19]. The detailed definition of an ‘at high risk’ population is provided in Supplemental Table S1.

2. Scenario 1 (those aged 65 years and above, and those aged 12−64 years at risk): During the period when the COVID-19 vaccine was categorized as a ‘Temporary Vaccination,’ the prioritized population was those aged 65 years and above and those aged 12−64 years ‘at risk’ [20]. ‘at risk’ population included those with underlying diseases such as chronic respiratory disease and chronic heart disease. The definition is further detailed in Supplemental Table S1. The age-dependent proportion of the population at risk was derived from the 2020 patient survey from the Ministry of Health, Labour, and Welfare of Japan (MHLW) and government statistics [19,21] by matching the disease classification code used in the survey to the ‘at risk’ definitions.

3. Scenario 2 (those aged 5 and above): We acknowledge the importance of vaccinating a broader population, i.e. age 5 years and above in Japan to reduce transmission. This target population was also aligned with that of Di Fusco et al. [14], which assessed the cost-effectiveness of the one-time BNT162b2 booster in the United States of America (U.S.A.). Neither the ‘at risk’ nor the ‘at high risk’ population was considered under this scenario.

Table 1. Base case and the main scenarios explored.

Scenario	Target population	Annual number of vaccination
Base case 1	Age 65 and above + Age 60–64 at high risk	1
Base case 2	Age 65 and above + Age 60–64 at high risk	2
Scenario 1–1	Age 65 and above + Age 12–64 at risk	1
Scenario 1–2	Age 65 and above + Age 12–64 at risk	2
Scenario 2–1	Age 5 and above	1
Scenario 2–2	Age 5 and above	2

See Supplemental Table S1 for the ‘at high risk’ and ‘at risk’ population definitions.

2.2. Model structure

The model structure was almost the same as the model used in the U.S. setting [14]. The Markov model (Figure 1) simulated the transition of the target population through mutually exclusive health states: susceptible, infected, recovered, and vaccinated states. For the vaccinated states, ‘Full primary vaccination,’ ‘First booster vaccination,’ and ‘Further booster vaccination’ were considered, the results of which led to vaccine-induced immunity. The susceptible state indicates that an individual is no longer protected through vaccine-induced immunity nor infection-induced immunity.

All cohorts entering the model were fully vaccinated with a primary series in line with the current booster authorized indication [14] including those who the received monovalent or bivalent booster regimen. The model assumed that at the model start, the included population belonged to one of the following three categories: fully vaccinated by the primary series, susceptible, or has infection-induced immunity (‘Recovered’). The cohorts moved through these health states in weekly cycles over the three-year time horizon. The cohorts transitioning to other health states are determined by the age-dependent probabilities in each cycle. Those in the recovered state were assumed to be protected by infection-induced immunity which was assumed to wane at a rate of 1/(duration of protection) before they transition to ‘Susceptible.’ A proportion of those in the susceptible state would transition to the infected state based on the age-dependent annual attack rate. Vaccine effectiveness (VE) reduced the age-dependent attack rate in determining the proportion of those in the vaccinated state transitioning to the infected state. The booster uptake was defined based on age-dependent vaccine coverage data to date. Once vaccinated, vaccine effectiveness was assumed to wane at a rate of 1/(duration of protection) and eventually transitioned to the susceptible state with a probability determined by the attack rate and vaccine effectiveness against infection. Infection could either lead to recovery or death, and the recovery would eventually lead to the susceptible state after infection-induced immunity waned. Once infected, the cohorts entered through the decision tree component which is described in Di Fusco et al. [14], which was the model used in the U.S. setting.

2.3. Model parameters/base case model inputs

The main model parameters are presented in Table 2 and further described in Supplemental Table S2. To the extent possible, relevant data from Japan during the Omicron strain dominant period were prioritized during the parameter search and used in the analyses.

Table 2. Main model parameters for the base case.

Category	Input description	5–11 years	12–17 years	18–29 years	30–59 years	60–64 years	65–74 years	≥75 years	Ref
Vaccine inputs	Booster vaccine coverage (%)	8.7%	45.9%	55.3%	68.6%	86.6%	88.2%	93.8%	[23]
	Vaccine effectiveness (VE) (infection/symptomatic/hospitalization)	60%/70%/80%							[14,22]
	Vaccine-induced duration of protection against infection	6 months							[14]
Infection inputs	Population with infection-induced immunity at the start (%)	30.3%	30.3%	27.1%	22.3%	14.4%	14.4%	14.4%	[27,28]
	Annual attack rate (%)	32.7%	30.5%	27.2%	21.1%	11.1%	9.1%	9.1%	[30]
	Duration of protection from infection-induced immunity	3 months							[14]
Clinical inputs	Hospitalization rate among symptomatic patients (%)	0.8%	0.6%	1.1%	1.3%	2.9%	6.0%	22.8%	[31]
	Among hospitalized patients:
	ICU admission rate (%)	0.1%	0.1%	1.3%	7.2%	7.2%	10.9%	10.9%	[34]
	IMV rate in general ward (%)	0.0%	0.0%	0.4%	4.2%	4.2%	6.8%	6.8%	[34]
	IMV rate in ICU (%)	0.0%	0.0%	0.4%	4.2%	4.2%	6.8%	6.8%	[34]
	Probability of COVID-19 death among infected individuals:
	General ward without IMV (%)	0.0%	0.0%	0.0%	0.1%	0.1%	0.7%	0.7%	[34]
	General ward with IMV (%)	0.0%	0.0%	6.7%	9.1%	9.1%	34.8%	34.8%	[34]
	ICU without IMV (%)	0.0%	0.0%	6.7%	9.1%	9.1%	34.8%	34.8%	[34]
	ICU with IMV (%)	0.0%	0.0%	6.7%	9.1%	9.1%	34.8%	34.8%	[34]
	Outpatient care setting (%)	0.00%	0.00%	0.00%	0.01%	0.08%	0.20%	1.13%	[33]
	Probability of long COVID among infected individuals:
	Asymptomatic patients (%)	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	Assumption
	Patients having received  outpatient care (%)	15.5%	25.3%	34.5%	42.8%	33.1%	32.2%	29.9%	[38]
	Patients having received  inpatient care (%)	35.6%	35.6%	35.6%	46.9%	66.7%	66.7%	66.7%	[39]
Risk inputs	Share of ‘at high risk’ population within age groups	0.1%	0.1%	0.1%	0.4%	0.6%	0.9%	2.3%	[18,19]
	Share of ‘at risk’ population within age groups	1.3%	0.7%	0.6%	1.6%	3.3%	4.8%	6.1%	[19–21]
	Hospitalization risk ratio – ‘at high risk’	1.83							[35]
	Hospitalization risk ratio – ‘at risk’	1.57							[36]
	Mortality risk ratio – ‘at high risk’	1.78							[35]
	Mortality risk ratio- ‘at risk’	1.57							[36]
Vaccine costs	Vaccine acquisition cost per dose for 30 mcg bivalent for individuals 12 years and older (JPY)	13,340							[14,40]
	Vaccine acquisition cost per dose for 10 mcg bivalent for individuals 5 to 11 years (JPY)	8,932							[14,40]
	Vaccine administration cost per dose (JPY)	3,250							[41]
Direct costs in the acute phase	COVID-19 test (JPY)	8,500							[42]
	Number of tests/case (asymptomatic/outpatient/inpatient/ICU)	0/1/3/3							Assumption
	Outpatient visit unit cost (JPY)	28,054							[42]
	Inpatient treatment (JPY)
	General ward  (without/with IMV)	562,650/ 1,449,227	561,632/ 1,449,227	845,761/ 1,449,227	937,469/ 1,449,227	965,039/ 205,148	1,239,964/ 854,644	1,446,865/ 1,986,158	[43]
	ICU (without/with IMV)	3,534,277/ 2,263,806	3,534,277/ 2,263,806	3,534,277/ 3,870,947	1,857,751/ 3,870,947	5,259,124/ 7,413,719	4,079,808/ 6,364,921	3,788,960/ 3,768,221	[43]
Direct costs of sequelae	Long COVID treatment cost (JPY)	107,725							[15,42]
Productivity loss	Asymptomatic (JPY)	36,439							[13,43]
	Symptomatic outpatient (JPY)	60,732							[13,43]
	General ward without IMV (JPY)	255,075							[41,43]
	General ward with IMV (JPY)	485,857							[41,43]
	ICU without IMV (JPY)	485,857							[41,43]
	ICU with IMV (JPY)	485,857							[41,43]
	Patients with long COVID (JPY)	728,786							[46]
QALY outcomes	Utility Decrement (symptoms, general ward, general ward with IMV, ICU, ICU with IMV)	−0.11, −0.17, −0.17, −0.18, −0.18							[48–50]
	Duration (days)
	Symptoms	6							[51]
	General ward  (without/with IMV)	6.0/23.5	4.0/23.5	9.0/23.5	11.0/23.5	16.0/1.0	21.0/48.0	28.0/42.0	[43]
	ICU (without/with IMV)	32.0/16.5	32.0/16.5	32.0/17.0	12.5/17.0	39.5/77.0	40.0/65.0	35.0/36.0	[43]

Refer to the Supplemental Material for further details on the model inputs for the different scenarios.

Abbreviations. COVID-19, coronavirus disease 2019; GP, general physician; ICU, intensive care unit; IMV, invasive mechanical ventilation; JPY, Japanese yen; PASC, post-acute sequelae SARS-CoV-2 infection; QALY, quality-adjusted life year.

2.3.1. Population

Age stratified estimates of population size were adopted from the Statistics Bureau of Japan census data conducted in 2020 [22]. The population eligible for a booster was derived from the reported coverage tracked by the Prime Minister’s Office of Japan of those who had completed full primary vaccination by May 2023 [23]. The coverage includes those who received their first and/or further booster vaccinations. Most of these booster vaccines were the previously authorized monovalent vaccines. Japan age-specific population estimates were used to adjust for the age-stratified percentages to match the model age groups.

2.3.2. Vaccine inputs

Regarding coverage, data were taken from the Prime Minister’s office of Japan website which tracks the number of vaccinations by age groups [23]. VE data on BNT162b2 from randomized controlled trials are unavailable, therefore, these data and duration of protection data were sourced from real world studies used in previous studies [14]. One study in Japan estimated the VE against symptomatic infection to be 71% [24] which was close to the value used in the aforementioned study at 70%. As of writing, the peer-reviewed journal publications to date offer limited data on the effectiveness of vaccines in Japan [25]. Moreover, most of the global studies on the effectiveness of COVID-19 vaccines from the Omicron period captures monovalent vaccines [26]. Duration of protection is also an important parameter for this analysis due to questions remaining around the vaccines and its effect on the recommended frequency of the booster. We assumed 6 months for the duration of protection which is consistent with previous studies and literature [14].

2.3.3. Infection inputs

Multiple data sources (Cabinet Secretariat, National Institute of Infectious Diseases data, etc.) were combined with the assumptions to define the distribution of the starting population for the model (Supplement Table S2) [27,28]. The seroprevalence data was taken from the latest survey conducted by the MHLW in May 2023 [29].

The annual attack rate that determines the proportion of the ‘Susceptible’ who were infected was derived from MHLW data that tracked age-stratified [30] weekly infection rates for the period covering May 2022 to May 2023, which was an Omicron strain dominant period. The reported number of infections is likely under-reported. Conservatively, no adjustments were made for the attack rate. The tracking of weekly infection stopped as of May 2023.

2.3.4. Clinical inputs

Clinical inputs on symptomatic infections, hospitalizations, and death were derived from prefectural evidence by age group, due to unavailability of national-level data [31–33]. Intensive care unit (ICU) admission rates of those hospitalized, invasive mechanical ventilation (IMV) rate among those hospitalized, and death rate among those hospitalized were taken from a nationwide observational study on clinical characteristics of the first three waves of hospitalized COVID-19 patients [34].

The risk ratio for hospitalization, including ICU and IMV, of the ”at high risk” population was taken from the cohort study conducted in New York State [35] while the mortality risk ratio was taken from the existing meta-analysis [36]. The mortality risk ratio of the ‘at risk’ population was calculated based on data from the MHLW Health Center Real-time information-sharing System on COVID-19 [37]. Due to data restriction, the same value as for the mortality risk ratio applied to the hospitalization risk ratio for the ‘at risk’ population.

The probability of post-acute sequelae SARS-CoV-2 infection (PASC) was differentiated among outpatient, inpatient, and asymptomatic cases. We use the terms PASC, long COVID, and post COVID-19 condition interchangeably in this manuscript. For outpatient cases, a survey conducted by the Yamanashi prefecture authority among those who were infected during the June 2021–July 2022 period was used as a proxy for the national representation [38]. For inpatient cases, a study reporting the prevalence of PASC among hospitalized patients were used [39]. For asymptomatic cases, we conservatively assumed that the probability of PASC was 0% across all ages.

2.3.5. Direct medical cost inputs

Vaccine related costs consisted of vaccine acquisition and vaccine administration costs. For the vaccine acquisition costs, considering the lack of a list price in Japan, we used the US-based price used by Di Fusco et al. [14] and applied the average exchange rate over the past 3 years between the U.S. dollar and Japanese yen (JPY) [40]. The vaccine administration costs were derived from the MHLW guidelines [41].

Symptomatic COVID-19 patients could incur medical costs including COVID-19 tests, outpatient visits, and hospitalization. For costs of COVID-19 testing, outpatient care, and treatment of PASC, we used the medical fee information provided by MHLW for calculation [42]. The hospitalization costs (general ward with or without IMV, ICU with or without IMV) were estimated from MDV claims data, one of the largest administrative data sources available in Japan from over 460 hospitals that covers over 40 million patients [43]. The data provided included persons hospitalized in the same year and month they became infected with COVID-19 to estimate the hospitalization cost. The data period was from 1 January 2022 to 31 December 2022. To avoid higher costs affected by outliers, median values were used over average values. The medical costs of PASC were also considered through outpatient care and number of tests.

Even though the reported probability of serious adverse events related to BNT162b2 intake has been extremely low, for example with the expected incidence of myocarditis being lower than 0.00001% [44], the risk and costs of myocarditis and pericarditis were considered in the model [44,45].

2.3.6. Productivity losses

Productivity losses from COVID-19 were calculated based on the age-dependent workforce participation rate, average labor cost per week (JPY 85,025), and number of work days lost [46]. The workdays lost estimates were derived from MHLW guidelines for isolation for those that received outpatient care [13]. The working days lost among hospitalized patients in the general ward (21 days) and ICU with and without IMV (40 days) were taken from the aforementioned large administrative hospital data (MDV data above [43]). The working days lost among PASC patients (60 days) were derived from the WHO clinical case definition [47]. Productivity loss for the patient herself/himself was considered. Moreover, we assumed that if the patient was not working (e.g. those aged 65 and above), the productivity loss for caregivers would occur (set at 100%).

2.3.7. Health utilities

Inputs on general population utility weights and utility decrements were taken from published articles based on the Japanese population [43,48–51].

2.4. Model analyses

The model estimated the health outcome gains achieved by booster vaccination over 3 years compared to no booster vaccination. The health outcomes included COVID-19 symptomatic cases, outpatient cases, hospitalizations, PASC, and total deaths. Total deaths included both COVID-19-related and non-COVID-19-related deaths. The model also predicted the associated direct healthcare costs and productivity losses saved by booster vaccination along with quality-adjusted life years (QALYs) gained.

Incremental cost-effectiveness ratio (ICER) was calculated by dividing the differences in costs by differences in QALYs, comparing the booster vaccination versus no booster vaccination. The formula of ICER:

$\mathit{ICER}=\frac{\left(\begin{array}{c}\mathit{Cost}\mathit{of}\mathit{intervention}\mathit{arm}-\\ \mathit{Cost}\mathit{of}\mathit{comparator}\mathit{arm}\end{array}\right)}{\left(-1\right)\times \left(\begin{array}{c}\mathit{QALY}\mathit{lost}\mathit{of}\mathit{intervention}\mathit{arm}-\\ \mathit{QALY}\mathit{lost}\mathit{of}\mathit{comparator}\mathit{arm}\end{array}\right)}$

The analysis is conducted from both the societal perspective and the healthcare payer perspective. The societal one is the primary perspective investigated in our study and requires the examination of productivity loss. The cost per QALY reference values (thresholds) commonly used in Japan is JPY 5 million (Japanese Yen), and this value was used in our analyses [41,52]. Incremental net monetary benefits (INMBs) were also calculated with the threshold from the societal perspective. In addition, we compared annual and biannual booster in terms of cost-effectiveness and evaluated multiple vaccination target population scenarios. The formula of INMB:

$\mathit{INMB}=\left(-1\right)\times \left(\mathit{QALY}\mathit{lost}\mathit{of}\mathit{intervention}\mathit{arm}-\mathit{QALY}\mathit{lost}\mathit{of}\mathit{comparator}\mathit{arm}\right)\times \mathit{JPY}5\mathit{million}-\left(\mathit{Cost}\mathit{of}\mathit{intervention}\mathit{arm}-\mathit{Cost}\mathit{of}\mathit{comparator}\mathit{arm}\right)$

Additional scenario analyses were conducted using alternative values for main model inputs. The selection of parameters was made considering the uncertainty of the parameter setting and the impact on the INMB. Moreover, various differing target populations for vaccinations were explored as scenario analyses, which were then compared on the cost-effectiveness plane.

Both deterministic and probabilistic sensitivity analyses were conducted to explore the robustness of the base case results. The deterministic sensitivity analysis (DSA) investigated the impact of individual parameter variability on INMB, and the input parameters were varied from −20% to + 20% from the base case value when there were no standard deviation (error) and 95% confidence interval in the sources. Therefore, most of the data from governmental reports were varied by ± 20%. In the probabilistic sensitivity analysis (PSA), the number of simulations was set at 1,000 times, calculating the ICER based on variable parameter input, according to probability distributions.

3. Results

3.1. Base case analysis

The base case results for annual and biannual vaccinations are presented in Table 3 and Supplemental Table S3. In the annual vaccination scenario, over 3 years, the booster program for those aged 65 years and above, and those aged 60–64 years at high risk was estimated to have averted 1,915,588 symptomatic cases 23,398 total deaths, and 319,066 hospitalizations related to COVID-19, resulting in an estimated JPY 673 billion additional direct medical costs, and JPY 1.94 trillion productivity loss saved, and JPY 1.27 trillion net cost savings relative to the counterfactual scenario of no booster vaccination. The vaccine-related costs were estimated to be JPY 1.39 trillion for annual vaccination. The annual vaccination resulted in discounted gains of 383,965 QALYs. Therefore, the annual booster base case model prediction indicated that from the societal perspective, the BNT162b2 booster was a dominant strategy compared to no booster vaccination.

Table 3. Results of the cost-effectiveness analysis of booster vaccination with the BNT162b2 vaccine in Japan for the base case over 3 years.

Outcomes	No Vaccination	Annual vaccination	Biannual vaccination	Difference (Annual vs. No vaccination)	%	Difference (Biannual vs. No vaccination)	%
	Base case (Age 65 and above + Age 60–64 at high risk)
Number of symptomatic cases, n	48,544,247	46,628,659	45,623,059	−1,915,588	−3.9%	−2,921,188	−6.0%
Number of total deaths, n	5,902,693	5,879,295	5,867,120	−23,398	−0.4%	−35,573	−0.6%
Number of outpatient cases, n	46,989,653	45,393,130	44,550,805	−1,596,523	−3.4%	−2,438,848	−5.2%
Number of hospitalizations, n	1,554,595	1,235,529	1,072,255	−319,066	−20.5%	−482,340	−31.0%
QALY lost, discounted	4,624,534	4,240,568	4,042,336	−383,965	−8.3%	−582,197	−12.6%
				This value is the difference in QALY lost. Therefore, the incremental QALY is the reversed-sign of this value.
Total medical cost, JPY	7,185,392,203,802	7,858,291,076,096	8,834,103,127,080	672,898,872,293	9.4%	1,648,710,923,277	22.9%
Vaccine-related	0	1,388,130,938,736	2,732,631,287,334	1,388,130,938,736	–	2,732,631,287,334	–
Testing	430,258,155,953	408,994,542,955	397,885,156,142	−21,263,612,998	−4.9%	−32,372,999,812	−7.5%
Adverse event	0	17,645,855	34,737,080	17,645,855	–	34,737,080	–
Inpatient	2,241,244,855,546	1,710,017,210,010	1,438,104,187,429	−531,227,645,537	−23.7%	−803,140,668,117	−35.8%
Outpatient	2,583,497,860,932	2,495,767,555,561	2,449,407,832,555	−87,730,305,371	−3.4%	−134,090,028,377	−5.2%
Long COVID	1,930,391,331,371	1,855,363,182,979	1,816,039,926,539	−75,028,148,392	−3.9%	−114,351,404,831	−5.9%
Productivity loss, JPY	21,683,064,958,283	19,743,705,250,059	18,736,353,819,653	−1,939,359,708,224	−8.9%	−2,946,711,138,630	−13.6%
Overall cost, JPY	28,868,457,162,085	27,601,996,326,154	27,570,456,946,733	−1,266,460,835,931	−4.4%	−1,298,000,215,352	−4.5%
ICER – societal perspective				Dominant		Dominant
ICER – payer perspective, JPY				1,752,499		2,831,878
Net monetary benefit from societal perspective, JPY				3,186,287,995,770		4,208,985,435,324

The reference scenario is no COVID-19 booster vaccination with the BNT162b2 COVID-19 Vaccine. Vaccine-related costs include acquisition and administration. The societal perspective includes productivity losses as well as direct medical costs. The payer perspective includes only direct medical costs.

Abbreviations. ICER, incremental cost-effectiveness ratio; JPY, Japanese yen; QALY, quality-adjusted life year.

In the biannual vaccination scenario, the booster program was estimated to have averted 2,921,188 symptomatic cases 35,573 deaths, and 482,340 hospitalizations related to COVID-19 over three years, resulting in an estimated JPY 1.65 trillion additional direct medical costs, JPY 2.95 trillion productivity loss saved compared to no vaccination. The net cost savings was JPY 1.30 trillion. The vaccine-related costs were estimated to be JPY 2.73 trillion for biannual vaccination. The biannual vaccination resulted in discounted gains 582,197 QALYs. Therefore, the base case model predictions indicated that from the societal perspective, the BNT162b2 booster was a dominant strategy compared to no booster vaccination, both in the case of annual and biannual vaccination. Detailed results of the cost-effectiveness analysis of booster vaccination with the BNT162b2 vaccine in Japan for the base case over 3 years are shown in Supplemental Table S3 at the population (3–1) and the individual (3–2) levels. On the other hand, ICERs for annual and biannual vaccinations versus no vaccination in terms of healthcare payer perspective were JPY 1,752,499/QALY and JPY 2,831,878/QALY which were less than the Japanese threshold of JPY 5 million/QALY (Table 3). Comparing with annual vaccination among the base case population, biannual vaccination was dominant from the societal perspective (Supplemental Table S4).

3.2. Scenario analyses

The results of the main scenario analyses are presented in Table 4 and Supplemental Tables S5-1 and S5-2. In Scenario 1, the target population was those aged 65 years and above, and those aged 12–64 years at risk. Generally, the differences were small between Scenario 1 and the Base case, and ICER from the societal perspective remained dominant. From the payer perspective in Scenario 1, ICERs for annual and biannual vaccinations were JPY 1,752,192/QALY and JPY 2,796,259/QALY respectively, which are less than the Japanese threshold JPY 5 million/QALY. When comparing the Scenario 1 results with the respective base case based on the number of vaccinations per year, the benefits of the vaccine increased in Scenario 1 in terms of symptomatic cases averted, total deaths averted, outpatient and hospitalization cases averted, and QALY saved. Scenario 1 also had larger productivity losses saved, and overall costs saved compared to the respective base case, however, the incremental direct medical costs were larger in Scenario 1 than in the base cases.

Table 4. Results of the cost-effectiveness analysis of booster vaccination with the BNT162b2 vaccine in Japan under two scenarios over 3 years.

Outcomes	No vaccination	Annual vaccination	Biannual vaccination	Difference (Annual vs. No vaccination)	%	Difference (Biannual vs. No vaccination)	%
	Scenario 1 (Age 65 and above + Age 12–64 at risk)
Number of symptomatic cases, n	48,544,172	46,525,917	45,418,026	−2,018,255	−4.2%	−3,126,146	−6.4%
Number of total deaths, n	5,903,358	5,879,744	5,867,443	−23,615	−0.4%	−35,915	−0.6%
Number of outpatient cases, n	46,989,592	45,292,145	44,349,238	−1,697,447	−3.6%	−2,640,354	−5.6%
Number of hospitalizations, n	1,554,580	1,233,772	1,068,788	−320,808	−20.6%	−485,793	−31.2%
QALY lost, discounted	4,631,773	4,236,915	4,028,760	−394,858	−8.5%	−603,013	−13.0%
				This value is the difference in QALY lost. Therefore, the incremental QALY is the reversed-sign of this value.
Total medical cost, JPY	7,186,773,360,914	7,878,640,649,640	8,872,952,984,827	691,867,288,726	9.6%	1,686,179,623,914	23.5%
Vaccine-related	0	1,420,086,667,232	2,795,826,077,613	1,420,086,667,232	–	2,795,826,077,613	–
Testing	430,257,300,974	408,112,056,127	396,124,439,029	−22,145,244,847	−5.1%	−34,132,861,944	−7.9%
Adverse event	0	18,052,075	35,540,410	18,052,075	–	35,540,410	–
Inpatient	2,242,633,226,095	1,709,204,753,361	1,435,304,524,641	−533,428,472,735	−23.8%	−807,328,701,454	−36.0%
Outpatient	2,583,494,548,713	2,490,231,250,523	2,438,355,870,973	−93,263,298,190	−3.6%	−145,138,677,740	−5.6%
Long COVID	1,930,388,285,132	1,850,987,870,323	1,807,306,532,162	−79,400,414,809	−4.1%	−123,081,752,970	−6.4%
Productivity loss, JPY	21,720,972,844,961	19,732,185,841,315	18,680,345,109,773	−1,988,787,003,645	−9.2%	−3,040,627,735,188	−14.0%
Overall cost, JPY	28,907,746,205,874	27,610,826,490,955	27,553,298,094,600	−1,296,919,714,919	−4.5%	−1,354,448,111,274	−4.7%
ICER – societal perspective, JPY				Dominant		Dominant
ICER – payer perspective, JPY				1,752,192		2,796,259
Net monetary benefit from societal perspective, JPY				3,271,210,317,922		4,369,511,056,568
	Scenario 2 (Age 5 and above)
Number of symptomatic cases, n	48,544,499	39,920,686	34,352,514	−8,623,812	−17.8%	−14,191,984	−29.2%
Number of total deaths, n	5,900,475	5,876,457	5,863,850	−24,018	−0.4%	−36,625	−0.6%
Number of outpatient cases, n	46,989,759	38,787,542	33,449,808	−8,202,217	−17.5%	−13,539,951	−28.8%
Number of hospitalizations, n	1,554,739	1,133,144	902,706	−421,595	−27.1%	−652,034	−41.9%
QALY lost, discounted	4,605,430	3,668,385	3,097,546	−937,046	−20.3%	−1,507,884	−32.7%
				This value is the difference in QALY lost. Therefore, the incremental QALY is the reversed-sign of this value.
Total medical cost, JPY	7,181,749,224,491	9,102,975,597,734	11,568,969,159,802	1,921,226,373,243	26.8%	4,387,219,935,311	61.1%
Vaccine-related	0	3,435,300,852,336	6,809,727,032,169	3,435,300,852,336	–	6,809,727,032,169	–
Testing	430,262,621,078	351,566,686,960	301,354,708,945	−78,695,934,119	−18.3%	−128,907,912,133	−30.0%
Adverse event	0	43,687,913	86,601,462	43,687,913	–	86,601,462	–
Inpatient	2,237,578,029,345	1,604,545,368,324	1,266,537,210,950	−633,032,661,021	−28.3%	−971,040,818,395	−43.4%
Outpatient	2,583,503,649,919	2,133,538,305,440	1,840,137,836,564	−449,965,344,479	−17.4%	−743,365,813,355	−28.8%
Long COVID	1,930,404,924,149	1,577,980,696,761	1,351,125,769,713	−352,424,227,388	−18.3%	−579,279,154,436	−30.0%
Productivity loss, JPY	21,581,908,765,147	17,220,791,146,959	14,577,411,148,019	−4,361,117,618,188	−20.2%	−7,004,497,617,127	−32.5%
Overall cost, JPY	28,763,657,989,638	26,323,766,744,693	26,146,380,307,821	−2,439,891,244,945	−8.5%	−2,617,277,681,816	−9.1%
ICER – societal perspective, JPY				Dominant		Dominant
ICER – payer perspective, JPY				2,050,302		2,909,520
Net monetary benefit from societal perspective, JPY				7,125,119,220,929		10,156,699,224,815

The reference scenario is no COVID-19 booster vaccination with the BNT162b2 COVID-19 Vaccine. Vaccine-related costs include acquisition and administration. The societal perspective includes productivity losses as well as direct medical costs. The payer perspective includes only direct medical costs.

Abbreviations. ICER, incremental cost-effectiveness ratio; JPY, Japanese yen; QALY, quality-adjusted life year.

In Scenario 2, the target population was age 5 and above. Due to the expanded vaccination target population in Scenario 2, the direct healthcare costs incurred were JPY 1.92 trillion (annual vaccination) and JPY 4.39 trillion (biannual vaccination) compared to no vaccination. The annual vaccination resulted in discounted gains of 937,046 QALYs while the biannual vaccination resulted in 1,507,884 QALYs. ICER remained dominant and therefore cost-effective from the societal perspective. ICERs for annual and biannual vaccinations were JPY 2,050,302/QALY and JPY 2,909,520/QALY from the payer perspective, respectively. These results were also less than the threshold value used in Japan. When comparing the Scenario 2 results with the respective Base case according to the number of vaccinations per year, the benefits of the vaccine increased in Scenario 2 in terms of symptomatic cases averted, total deaths averted, outpatient and hospitalization cases averted, and QALY saved.

3.3. Further scenario analyses

Further scenario analyses were conducted using alternative values for main model inputs as presented in Table 5 and Supplemental Tables S6–1 and S6–2. The inputs explored included percentage of population with infection-induced immunity at the start of the model horizon, annual attack rates, infected-induced duration of protection, vaccine effectiveness and duration, probability of symptomatic infection, and hospitalization rate and cost. In all scenarios, ICER from the societal perspective remained dominant, both for annual and biannual booster vaccination. All ICER values from the healthcare payer perspective were below the threshold of JPY 5 million/QALY.

Table 5. Scenario analyses using alternative values for main model inputs.

	Annual vaccination (Base case 1)			Biannual vaccination (Base case 2)
	Incremental QALY, discounted	Incremental total costs, JPY	ICER – societal perspective	Incremental QALY, discounted	Incremental total costs, JPY	ICER – societal perspective
Base case	383,965	−1,266,460,835,931	Dominant	582,197	−1,298,000,215,352	Dominant
Percentage of population with infection-induced immunity at the start of the model horizon
−25% (75% of the base case)	388,563	−1,297,936,743,143	Dominant	586,788	−1,329,421,424,384	Dominant
+25% (125% of the base case)	379,368	−1,234,984,928,719	Dominant	577,606	−1,266,579,006,321	Dominant
Yearly attack rate in susceptible population
Uniformly 30% for all age groups	1,251,250	−7,273,655,159,290	Dominant	1,918,021	−10,569,015,539,456	Dominant
Infection-induced duration of protection against infection
2 months	393,756	−1,335,104,301,926	Dominant	594,347	−1,383,004,117,738	Dominant
4 months	373,828	−1,196,199,595,117	Dominant	570,032	−1,213,709,609,553	Dominant
Efficacy of vaccine against infection and symptomatic cases
50%, 60%	351,617	−1,077,020,423,318	Dominant	533,933	−1,015,413,118,921	Dominant
60%, 60%	351,517	−1,088,169,774,050	Dominant	533,618	−1,030,715,275,204	Dominant
70%, 80%	416,523	−1,457,131,319,426	Dominant	630,664	−1,581,777,737,228	Dominant
Efficacy of vaccine against hospitalization
70%	360,832	−1,072,576,069,039	Dominant	547,669	−1,008,208,106,649	Dominant
90%	407,101	−1,460,362,622,104	Dominant	616,730	−1,587,837,975,812	Dominant
Vaccine-effectiveness duration of protection against infection
5 months	336,616	−938,075,327,063	Dominant	536,679	−981,411,847,922	Dominant
7 months	424,299	−1,546,476,133,510	Dominant	618,045	−1,547,569,460,099	Dominant
Probability of symptomatic infection
Applying highest value for all generations  uniformly (96.9%)	410,539	−1,448,902,225,573	Dominant	622,403	−1,574,454,826,753	Dominant
Applying lowest value for all generations  uniformly (89.7%)	379,982	−1,242,412,654,323	Dominant	576,070	−1,260,928,939,630	Dominant
Hospitalization rate among symptomatic patients
−25% (75% of the base case)	341,186	−906,639,901,909	Dominant	517,502	−752,809,522,509	Dominant
+25% (125% of the base case)	426,704	−1,625,944,354,746	Dominant	646,835	−1,842,714,092,041	Dominant
Hospitalization cost (general ward without IMV)
25%ile value	383,965	−1,098,900,131,119	Dominant	582,197	−1,044,518,542,894	Dominant
75%ile value	383,965	−1,522,206,547,083	Dominant	582,197	−1,684,861,559,305	Dominant

Abbreviations. ICER, incremental cost-effectiveness ratio; IMV, invasive mechanical ventilation; JPY, Japanese yen, QALY, quality-adjusted life year.

Moreover, various differing target populations were explored as additional scenarios. The description of each target population and their respective results are shown in Supplemental Table S7. The target groups included various age groups with and without the ”at high risk” population consideration, and both annual and biannual vaccinations were considered. The results were robust, with all target population groups showing ICER from the societal perspective being dominant. All ICER values from the payer perspectives were below the threshold of JPY 5 million/QALY. Cost-effectiveness planes for the scenarios explored in Supplemental Table S7 are presented in Supplemental Figures S1 and S2.

3.4. Sensitivity analyses

The results of the DSA were presented as Tornado diagrams, and the most impactful parameters for INMB are shown in Figure 2 for Base case 1 (annual vaccination) & 2 (biannual vaccination). Both in Base case 1 and 2, the probability of symptomatic infection for the eldest group (i.e. age 75 and above) was the most impactful parameter. This was followed by parameters related to the duration of protection associated with the vaccine and the hospitalization rate among symptomatic patients aged 75 and above. The INMBs were positive in all cases examined in the DSA.

Figure 2. Deterministic sensitivity analyses for annual and biannual groups: Tornado diagrams for incremental net monetary benefit – societal perspective.

Color of the bar indicates whether the parameter is at its lower or upper bound. 647 parameters were tested, and this figure includes only the 10 most impactful parameters. Abbreviations. JPY, Japanese yen; Q1, first quartile; Q3, third quartile.

The PSA showed that the results were robust. Figure 3 shows the cost-effectiveness acceptability curves. For Base case 1 & 2, they show that an average 92.7% and 88.7% probability of the booster vaccine being cost-effective at JPY 5 million per QALY gained, respectively. Figure 4 shows the cost-effectiveness plane for the societal perspective in Base case 1 & 2 and the simulation results. For Base case 1 and 2, 72.0% and 61.4% of the simulations were in the fourth quadrant (lower cost, higher efficacy versus no booster vaccination), respectively.

Figure 3. Probabilistic sensitivity analysis: cost-effectiveness acceptability curve for societal perspective.

The results were calculated by individual level. Abbreviations. JPY, Japanese yen.

Figure 4. Probabilistic sensitivity analysis: cost-effectiveness plane for societal perspective.

The results were calculated by individual level. Abbreviations. JPY, Japanese yen, QALY, quality-adjusted life year.

4. Discussion

Comparing the cost-effectiveness of having a booster vaccination compared to no booster vaccination was the primary focus of this study. The base case was the main target population, i.e. those aged 65 years and above, and those aged 60–64 years at high risk. The ICERs in the base case analyses were found to be dominant from the societal perspective. The findings remained consistent in Scenarios 1, 2, and other target populations. Moreover, all ICERs from the payer perspective remained under the threshold of JPY 5 million commonly applied in Japan. Further, the findings were supported by additional scenario analyses and sensitivity analyses. The analysis from the societal perspective leads to an understanding of the broader value of vaccines.

From a cost-effectiveness perspective, our study suggested that a biannual COVID-19 vaccination strategy may be more beneficial than an annual vaccination (Supplemental Table S4). This is in agreement with a previous study on the severity of COVID-19 infection where biannual vaccinations are recommended [16]. In Japan, however, an annual vaccination is scheduled in the fall of 2024, and to the best of our knowledge, there are no countries currently recommending the use of biannual vaccinations [53–55]. Biannual vaccination may be operationally burdensome and unfeasible as a measure for the governments.

COVID-19 infection and its spread have significant impacts on loss of productivity due to having to stay at home or in hospitals. Vaccine-related costs overall were estimated to be JPY 1.39 trillion for annual vaccination and JPY 2.73 trillion for biannual vaccination, however, they were offset by the reductions in productivity losses (JPY 1.94 trillion and JPY 2.95 trillion for annual and biannual vaccinations, respectively). Therefore, the booster vaccinations are cost-saving from the societal perspective. This study supported the value of booster vaccination from the cost-effectiveness aspect.

Infection rates had an impact on all analyses. They especially had a significant impact on the cost-effectiveness among the elderly population. The base case had a setting where the attack rate was assumed to be lower as the age increased based on the data obtained, and the rate was less than 10% for those over 65 years old. The scenario analysis (Table 5) showed that when the annual attack rate was set at 30% for all age groups, the ICERs improved significantly, and the ICER from the payer perspective became dominant. The spread of the infection would affect the value of the vaccine. As of July 2023, there were reports that the infection is still spreading in some areas of Japan [27]. It remains challenging to predict how the epidemic will evolve in the future.

For both annual and biannual booster vaccination, the results were sensitive to clinical and vaccine efficacy inputs for those aged 75 years and older (Figure 2). Even when these input parameters were varied by some ranges, the INMBs remained positive. The INMBs still remained positive with a 20% increase in the vaccine acquisition costs in the analysis. Consequently, the results showed robust cost-effectiveness in all sensitivity and scenario analyses, despite sensitivities to some inputs.

The after-effects of infection as long COVID also lead to stagnations in social and economic activities. COVID-19 vaccinations are also effective for preventing long COVID [56]. As patients who have experienced hospitalization are more likely to develop long COVID, our study indirectly observed the effects of vaccination on long COVID. Our results demonstrate that treatment costs related to long COVID were reduced by 3.9% and 5.9% for annual and biannual vaccinations respectively (Table 3).

Previous research on the cost-effectiveness of one-shot only booster vaccination in the context of the U.S., the United Kingdom, and Australia, all indicate that it would be cost-saving to invest in the booster vaccination [14,57–59]. Although these studies differed in settings from this study, their cost-effectiveness results also indicated being dominant or under the threshold of the ICER which is in agreement with this present study.

This study had some limitations. First, the strains continue to evolve, and the challenge was to obtain the timeliest data available. Due to limited clinical data (symptomatic case rate, hospitalization rate, etc.) from the Omicron period, select pre-Omicron period data were also used. There was also a lack of appropriate data at the national level in Japan, and we used data from the prefecture-level. Therefore, there may be issues associated with representativeness and validity of the data for Japan as a whole. Second, our model did not account for seasonal or epidemic variation in infection as we used the proportion of those infected during the preceding year. To complement the analyses, we performed additional scenario analyses and sensitivity analyses (DSA and PSA) and showed the robustness of the main results. Third, COVID-19 had broad societal impacts on activities such as going to school and accessing public services including healthcare. Our model, however, only included loss of productivity for patients and caregivers. Advanced models that consider the broader impact of COVID-19 May be needed for future research. Lastly, our study was based on situations in September 2023 in Japan. CEAs of booster vaccinations in future situations and settings in Japan and in other countries will be needed based on the appropriate assumptions.

5. Conclusions

This study indicated that booster vaccination with COVID-19 vaccine BNT162b2 was cost-saving and beneficial to public health in Japan, compared to no vaccination during the Omicron predominance among those aged 65 years and above, and those aged 60–64 years at high risk. The results were consistent with other plausible target populations from both the societal and payer perspectives. This evidence supports the policy of booster vaccination program by the government of Japan.

Declaration of interests

M Nagano, K Kamei, C Takahashi and N Yonemoto are full-time employees of Pfizer Japan Inc. K Wada was a full-time employee of Pfizer Japan Inc. when the study was conducted. M Nagano and K Kamei hold Pfizer stock, and C Takahashi and N Yonemoto hold stock and stock options from Pfizer Inc. H Matsuda is a full-time employee of IQVIA Solutions Japan G.K., was employed and compensated for conducting the study and medical writing support from Pfizer Japan Inc. Y Jingyan is a full-time employee of Pfizer Inc.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

All named authors take responsibility for the integrity of the work as a whole and have given final approval for this version to be published. All authors conceived and designed the study, populated the model inputs, and reviewed the manuscript.

Acknowledgments

The authors’ appreciation goes to Karin Matsumoto of IQVIA Solutions G.K. for the project management of this study, Kanae Togo and Masashi Mikami of Pfizer Japan Inc. for data analysis for cost parameters, Kosuke Tanabe of Pfizer Japan Inc. for providing his medical opinions, and Ataru Igarashi of the Department of Public Health, Yokohama City University School of Medicine, and the Department of Health Economics and Outcomes Research, Graduate School of Pharmaceutical Sciences, The University of Tokyo for providing valuable scientific advice.

Supplementary Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14760584.2024.2323133

Additional information

Funding

This paper was funded by Pfizer Japan Inc.