Cost-effectiveness analysis of adult pneumococcal conjugate vaccines for pneumococcal disease in Japan

ABSTRACT

Background

The 23-valent pneumococcal polysaccharide vaccine (PPSV23) is used in the Japanese National Immunization Program for older adults and adults with increased risk for pneumococcal disease, however, disease incidence and associated burden remain high. We evaluated the cost-effectiveness of pneumococcal conjugate vaccines (PCVs) for adults aged 65 years and high-risk adults aged 60–64 years in Japan.

Research design and methods

Using a Markov model, we evaluated lifetime costs using societal and healthcare payer perspectives and estimated quality-adjusted life-years (QALYs), and number of prevented cases and deaths caused by invasive pneumococcal disease (IPD) and non-IPD. The base case analysis used a societal perspective.

Results

In comparison with PPSV23, the 20-valent PCV (PCV20) prevented 127 IPD cases 10,813 non-IPD cases (inpatients: 2,461, outpatients: 8,352) and 226 deaths, and gained more QALYs (+0.0015 per person) with less cost (-JPY22,513 per person). All sensitivity and scenario analyses including a payer perspective analysis indicated that the incremental cost-effectiveness ratios (ICERs) were below the cost-effectiveness threshold value in Japan (JPY5 million/QALY).

Conclusions

PCV20 is both cost saving and more effective than PPSV23 for adults aged 65 years and high-risk adults aged 60–64 years in Japan.

KEYWORDS:

• Cost-effectiveness
• invasive pneumococcal disease
• Japan
• PCV20
• pneumococcal disease
• vaccination

1. Introduction

Pneumococcal disease (PD) is a common illness caused by infection with the bacteria Streptococcus pneumoniae (S. pneumoniae), and is a significant source of morbidity and mortality in adults aged ≥50 years [1]. Two classifications of PD exist: invasive PD (IPD) and non-IPD (including non-bacteremic pneumococcal pneumonia (NBP)) based on the site of infection. The National Institute of Infectious Disease (NIID) reported IPD incidence rates in a Japanese population aged ≥65 years as 5.57 per 100,000 population with associated mortality rates of 8.1% in 2019 [2]. Miyazaki et al. reported an annual incidence rate of pneumococcal community-acquired pneumonia (CAP, it was assumed in this study that CAP and NBP refer to the same disease) as 89, 260, and 645 per 100,000 population in adults aged 18 to 64 years, 65 to 79 years, and ≥80 years, respectively, in their prospective, population-based study in Goto city from 2015 to 2020 [3]. Given this, prevention of PD is especially important in Japan, a country with a considerable elderly population, where 29% of the population are aged ≥65 years [4].

As of January 2024, 3 vaccines, the 23-valent pneumococcal polysaccharide vaccine (PPSV23) and the 13- and 15-valent pneumococcal conjugate vaccines (PCV13 and PCV15) are available in Japan. Only PPSV23 is included in the routine national immunization program (NIP) since October 2014 for all adults aged 65, 70, 75, 80, 85, 90, 95, and 100 years, as well as adults aged 60 to 64 years with increased risk such as impairment of heart, kidney, or respiratory function, or impairment of immune function due to human immunodeficiency virus (HIV) [5]. From April 2024, only adults aged 65 years and adults aged 60 to 64 years with increased risk will be eligible for routine vaccination of PPSV23 under the NIP [6]. Despite the introduction of the PPSV23 vaccination in the NIP, the percentage of IPD caused by PPSV23-serotypes out of total IPD cases has remained high in adults (67% and 47% in 2014 and 2022) [7]. One of the reasons for this may be the low pneumococcal vaccination rates in adults [8,9]. This contrasts with the change in IPD since the introduction of pediatric PCV13. Indeed, the percentage of IPD caused by PCV13-serotypes was dramatically decreased after the introduction of PCV13 in the NIP system in November 2013 (53.2% and 2.3% in 2013 and 2022) [7] with nearly a 100% pneumococcal vaccination rate [10].

Unlike the immune response to plain polysaccharide vaccines such as PPSV23, the immune response to conjugate vaccines such as PCV13 and PCV15 provides long-term immunity through the production of new memory B cells [11]. The efficacy of PPSV23 is known to decline from year 1, while no decline of PCV13 efficacy is observed in the first 5 years [12–15]. Additionally, a new vaccine, the 20-valent pneumococcal conjugate vaccine (PCV20) has been licensed for the prevention of PD in the United States and has replaced PCV13 in prescribing practice. In Japan, an application to approve the use of PCV20 has been submitted to the Pharmaceuticals and Medical Devices Agency (PMDA) [16]. PCV20 was investigated in a phase II study in adults aged 60 to 64 years and in two phase III randomized controlled trials in adults aged ≥18 years [17–20]. The results of these studies demonstrate that PCV20 is associated with an improved immune response compared with PPSV23 and is anticipated to provide more robust protection compared with PPSV23 for the serotypes both vaccines have in common. The introduction of PCV20 could therefore prevent more IPD and NBP cases; it could also reduce deaths related to those diseases among the elderly and immunocompromised persons.

In advance of PCV20 approval in Japan, an economic evaluation of PCV20 and PCV15 compared with PPSV23 in an elderly population aged 65 years was conducted by Hoshi et al. [21]. The study found both PCV20 and PCV15 were cost-effective compared with PPSV23, with PCV20 being the most cost-effective. However, since PCV20 and PCV15 were not available in Japan when they conducted the analysis, the US Centers for Disease Control and Prevention (CDC)’s adult vaccine prices were used. Furthermore, the authors considered a restricted population of adults aged 65 years. Given this, the findings may not be fully reflective of a Japanese population eligible for the NIP. To inform potential gaps in knowledge about the impact of PCV20 in a Japanese healthcare setting, this paper evaluates the cost-effectiveness of PCV20 versus PPSV23 in adults aged 65 years and high-risk adults aged 60 to 64 years from both payer and societal perspectives using the latest available data from Japanese sources. Additionally, an assessment of the cost-effectiveness of PCV15 versus PPSV23 and PCV20 versus PCV15 in the same population are explored as secondary objectives.

2. Patients and methods

2.1. Model description

The cost effectiveness of a single dose of the various pneumococcal vaccines was assessed using a decision-analytic Markov model. The model was adapted from a model previously published by Mendes et al., calibrated to reflect Japanese healthcare settings [22]. The model utilized a deterministic framework and a Markov-type process to depict the risks and costs of IPD and all-cause NBP over remaining years of life in a hypothetical population of Japanese adults. The model population was stratified by age and risk profile (i.e. low, medium, or high-risk) at model entry.

Expected clinical outcomes and economic costs were calculated on an annual basis, using age, risk profile, vaccination status, vaccine type, and time since vaccination. IPD was assumed to include bacteremia and meningitis; all-cause NBP was stratified by setting of care (inpatient or outpatient). People vaccinated at model entry had a differential risk of future IPD and all-cause NBP. The magnitude of vaccine-associated risk reduction was dependent on clinical presentation (i.e. IPD or all-cause NBP), in addition to the type of vaccine used, time since vaccination, and risk profile. Risk of death from IPD, all-cause NBP requiring inpatient care, and other causes (i.e. other than IPD and all-cause NBP) were dependent on age and risk profile.

At each model cycle, individuals were modeled to remain in their current health state or transition to another health state, based on health state transition probabilities. In the original model published by Mendes et al., individuals were permitted to transition to a higher risk group (i.e. from low risk to medium risk or medium risk to high risk) but not to a lower risk group [22]. However, due to a lack of Japanese specific evidence to inform this transition, individuals were not permitted to transition to lower or higher risk groups in the base case analysis. General population mortality risk was applied to all individuals, with additional mortality risk applied for IPD or inpatient NBP. Death was an absorbing health state. A model schematic is presented in Figure 1.

Figure 1. Model structure.

IPD: invasive pneumococcal disease; NBP: non-bacteremic pneumococcal pneumonia.

Note: Transitions between risk categories were not included in this study.

Expected costs of medical treatment for IPD and all-cause NBP were generated based on event rates and unit costs for the setting of care (i.e. inpatient or outpatient), age, and risk profile. Costs of vaccination (including vaccine and administration costs) were applied at model entry. The value of morbidity- and mortality-related work loss was included in the model. Clinical outcomes and economic costs were projected over the model lifetime horizon (up to 100 years) for each vaccination strategy. As with previous studies in this field, the model used a one-year cycle length and applied a half-cycle correction [21,23]. Cost-effectiveness was calculated in terms of cost per quality-adjusted life-year (QALY) gained and cost per life-year (LY) gained. Costs and outcomes were discounted at a rate of 2% per year [24]. A cost-effectiveness threshold of JPY5,000,000 per QALY gained was used in the analysis to assess cost-effectiveness [25]. The incremental net monetary benefit (INMB) was calculated as follows:

INMB = incremental QALYs gained with the intervention strategy (e.g. PCV20) compared to the comparator strategy (e.g. PPSV23) × threshold – incremental costs associated with the intervention strategy compared to the comparator strategy

The base case analysis was conducted from a societal perspective (inclusive of medical, productivity loss, and vaccination costs) as it was important to consider productivity loss, given the high burden of the disease [24]. Results from a payer perspective are presented as a scenario.

2.2. Model estimation

The main model parameters are presented in Table 1 and further presented in the supplemental online material. Additional details on the targeted literature review (TLR) are provided in the supplemental online material. If data from Japan were not available, other appropriate data were considered.

Table 1. Base-case model inputs.

Parameters	Values				Source
	50–64 years	65–74 years	75–84 years	85–99 years
Population
Age distribution	126,565 (60–64 high risk only)	1,417,000 (65 only)	–	–	E-stat statistics of Japan [26], Imai et al. [27]
Risk stratifications
Risk stratifications by age group: low/moderate/high	0.0%/0.0%/100.0% (High risk only)	88.1%/10.1%/1.7%			Imai et al. [27]
Allow switching between risk groups	‘No’				Assumption
Vaccine history
Percentage of persons previously vaccinated with					Assumption
PCV13 only	0.0%
PPSV23 only	0.0%
PCV13 + PPSV23	0.0%
Years since vaccination	N/A				Assumption
Disease incidence per 100,000 persons (JMDC [32]and MDV [33] data, 2017–2019)
Incidence of IPD bacteremia: low/moderate/high	1.03/5.89/18.81	3.67/11.01/11.38	4.42/13.26/13.70	4.42/13.26/13.70	NIID [34], e-stat statistics of Japan [26], IASR 2023 [2], Imai et al. [27]
Incidence of IPD meningitis: low/moderate/high	0.09/0.51/1.64	0.32/0.96/0.99	0.38/1.15/1.19	0.38/1.15/1.19	NIID [34], e-stat statistics of Japan [26], IASR 2023 [2], Imai et al. [27]
Incidence of NBP hospitalization: low/moderate/high	140.07/700.33/ 1,428.68	501.53/1,103.37/ 902.76	1,287.96/2,833.51/ 2,318.33	2,237.61/4,922.73/ 4,027.69	JMDC [32], MDV [33], Morimoto et al. [77], Imai et al. [27]
Incidence of NBP outpatient: low/moderate/high	1,061.18/5,305.89/ 10,824.02	2,123.22/4,671.09/ 3,821.80	2,619.56/5,763.03/ 4,715.21	1,075.40/2,365.87/ 1,935.72	JMDC [32], MDV [33], Morimoto et al. [77], Imai et al. [27]
Vaccine serotypes
Vaccine serotype distribution: IPD					IPD information [7]
PCV13 serotypes	29.0%
PCV15 serotypes	35.0%
PCV20 serotypes	61.0%
PPSV23 serotypes	61.7%
Serotype 3	10.7%
Serotype 6A	1.0%
Vaccine serotype distribution: NBP pneumococcal					Assumption (same to IPD)
PCV13 serotypes only	29.0%
PCV15 serotypes only	35.0%
PCV20 serotypes only	61.0%
PPSV23 serotypes only	61.7%
Serotype 3 only	10.7%
Serotype 6A only	1.0%
Proportion of all-cause NBP due to S. pneumoniae	16.5%				Miyazaki et al. [3]
Vaccine coverage
Percentage of persons vaccinated	40.4%	40.4%	N/A		IASR 2018 [9]
Mortality rates per 100 persons
Annual general mortality	0.38	1.23	3.24	11.13	e-stat [26], MHLW [61]
IPD meningitis case fatality	7.14	12.10	12.10	12.10	Chang et al. [62]
IPD bacteremia case fatality	13.82	19.24	19.24	19.24	Chang et al. [62]
NBP hospitalization case fatality	5.52	7.41	10.12	10.12	Miyazaki et al. [3]
Utility
Utility for the general population	0.9280	0.9030	0.8420	0.7890	e-stat [26], Shiroiwa et al. [66]
QALY loss for meningitis	0.1741				Blommaert et al. [67]
QALY loss for bacteremia	0.1741				Blommaert et al. [67]
QALY loss for hospitalized NBP	0.0434				Calculated values*
QALY loss for outpatient NBP	0.0289				Calculated values*
Direct cost
Vaccination cost
PCV20	JPY8,102				Assumption
PCV15	JPY7,200				MHLW [69]
PPSV23	JPY4,735				NHI price revisions [70]
Administration cost	JPY3,250				MHLW medical fee 2022 [72]
IPD meningitis cost	JPY1,636,140	JPY2,178,263	JPY2,178,263	JPY2,178,263	JMDC [32], MDV [33]
IPD bacteremia cost	JPY1,261,855	JPY1,349,205	JPY1,349,205	JPY1,349,205	JMDC [32], MDV [33]
Hospitalized NBP cost	JPY907,530	JPY1,125,977	JPY1,125,977	JPY1,125,977	JMDC [32], MDV [33]
Outpatient NBP cost	JPY33,040	JPY43,646	JPY43,646	JPY43,646	JMDC [32], MDV [33]
Cost related to productivity loss
Employment rate	100%	0%	0%	0%	Assumption
Daily wage: patient	JPY30,402	–	–	–	Wage Structure Basic Statistical Survey [76], assumption
Caregiver utilization rate	0%	100%	100%	100%	Assumption
Daily wage for caregiver	–	JPY30,402	JPY30,402	JPY30,402	Wage Structure Basic Statistical Survey [76], assumption
Days lost because of IPD hospitalization for bacteremia	22	24	24	24	JMDC [32], Konomura et al. [64]
Days lost because of IPD hospitalization for meningitis	32	24	24	24	JMDC [32], Konomura et al. [64]
Days lost because of NBP hospitalization	16	17	17	17	JMDC [32], Konomura et al. [64]
Days lost because of NBP outpatient care	7	7	7	7	Konomura et al. [64]

CDC: US Centers for Disease Control; NHI: National Health Insurance; NIID: National Institute of Infectious Diseases; IASR: Infectious Agents Surveillance Report; IPD: invasive pneumococcal disease; JPY: Japanese yen; MDV: Medical Data Vision Co., Ltd; MHLW: Ministry of Health, Labour and Welfare; NBP: non-bacteremic pneumonia; PCV13: 13-valent pneumococcal vaccine; PCV15: 15-valent pneumococcal vaccine; PCV20: 20-valent pneumococcal vaccine; PPSV23: 23-valent pneumococcal polysaccharide vaccine; N/A: Not Applicable.

*Detailed information on the calculations made to estimate those values are described in the supplemental online material.

2.2.1. Population

In the base case, the target population was adults aged 65 years and high-risk adults aged 60 to 64 years. Total population numbers and age distribution were informed by the official age structure of Japan (as of 2022) [26]. The proportion of adults with high-risk (1.7%) was informed by Imai et al. [27]. Further information on risk categories is provided below. Other scenarios were tested for the modeled population: adults aged ≥65 years and high-risk adults aged 60 to 64 years (scenario A), adults aged 65 years only (scenario B), adults aged 65 years and moderate and high-risk adults aged 60 to 64 years (scenario C), and adults aged 66 to 84 years only (scenario D).

2.2.2. Risk definition

In the base case, the risk definition from Imai et al. was used as this study considered guidelines and recommendations from the United States, England, and Japan [27–31]. Medium-risk conditions included chronic heart disease, chronic lung disease, diabetes mellitus, chronic liver disease, and alcoholism. High-risk conditions included chronic renal disease, cancer, HIV/AIDS, functional or anatomic asplenia, organ transplant, and cerebrospinal fluid leakage. The same source was used to inform the proportion of at-risk adults and risk ratio for disease incidence. Imai et al. sourced the proportion of at risk adults from 2 Japanese claims databases, the database of JMDC Inc. (JMDC database) for adults aged <65 years and Medical Data Vision Co., Ltd. (MDV database) for adults aged ≥65 years [27,32,33]. The JMDC database consists of approximately 16 million beneficiaries of health insurance associations and the MDV database contains Diagnosis Procedure Combination (DPC) data from approximately 27% facilities out of the nationwide DPC target hospitals.

2.2.3. Disease incidence and serotype distribution

Disease incidence and serotype distribution before the COVID-19 pandemic (2017–19) were used for the base-case analysis as PD incidence is expected to return to pre-pandemic levels. Incidence data during the COVID-19 pandemic (2020–22) were used for scenario analysis to consider the impact of COVID-19. Incidence rates for IPD and NBP were calculated based on surveillance data from NIID and the JMDC database, respectively [32,34]. Detailed information on the methods employed are described in the supplemental online material.

The proportion of NBP due to S. pneumoniae of any serotype was assumed to be 16.5% in the base case, based on data from Miyazaki et al. [3]. In scenario analysis, a value of 20% was applied, sourced from a study by Kobayashi et al. [35]. Percentages of vaccine serotypes for IPD were obtained from the latest data reported by the Epidemiological Information on Invasive Pneumococcal Infections in Adults and Children Research Group [7]. Values were assumed to remain constant throughout the time horizon. Vaccine serotypes for NBP were assumed to be the same as IPD in the base-case and scenario analyses.

2.2.4. Vaccine coverage and history

Vaccine coverage was assumed to be 40.4% of those eligible for vaccination; consistent with the value used in a previously published economic evaluation, sourced from the Japanese Infectious Agents Surveillance Report (IASR) from 2018 [9,21]. This value was applied to the modeled population. Two additional scenarios for vaccine coverage were explored: 52%, consistent with data available from the IASR 2023 and 20%, which is an assumption to test the effects of poor uptake [2].

In the base case, it was assumed that the studied population is vaccine-naïve, as there were limited data to inform this parameter. Data on vaccine history for persons with CAP were available from a study by Nakashima et al., who reported the proportion of persons aged ≥65 years who had been vaccinated within the previous 5 years (55.6% with only PPSV23 and 1.4% with both PPSV23 and PCV13) in Japan. These values were used in scenario analysis [36].

2.2.5. Effectiveness of PCVs

Methods to derive the effectiveness of PCVs against vaccine type invasive pneumococcal disease (VT-IPD) and vaccine type non-bacteremic pneumonia (VT-NBP) were described in Mendes et al. [22]. Values used for vaccine effectiveness are summarized in Supplementary Table S1. The same serotype-specific effectiveness was assumed for all PCVs and for all included serotypes.

Efficacy inputs for the initial period of effectiveness of PCVs against VT-IPD for low- and medium-risk people were based on data for adults aged ≥65 years in the Community-Acquired Pneumonia Immunization Trial in Adults (CAPiTA) per-protocol population and from post hoc analyses for people aged 60 to 64 years [37,38]. For people aged ≥65 years with low or medium risk, initial vaccine effectiveness was assumed to be 75% [37]. For people aged 60 to 64 years, low- and medium-risk effectiveness was derived using age-specific relative changes in vaccine effectiveness against VT-IPD [37]. High-risk people (all ages) were assumed to have effectiveness equal to 80% of the corresponding values for low- and medium-risk persons. This assumption was based on a study by Klugman et al., in which PCV9 vaccine efficacy against VT-IPD among HIV-positive children was 65%, versus 83% among HIV-negative children (relative efficacy, 78%) [39,40].

Efficacy inputs for the initial period of effectiveness of PCV vaccines against VT-NBP for low- and medium-risk people were based on data for adults aged ≥65 years in the CAPiTA per-protocol population and from post hoc analyses of CAPiTA data for adults aged <65 years [37]. For people aged ≥65 years with low or medium risk, initial vaccine effectiveness was assumed to be 45%, based on the overall study findings. For people aged 60 to 64 years with low or medium risk, initial vaccine effectiveness estimates were derived using age-specific relative changes in effectiveness against VT-NBP (versus age 65 years) [37]. Initial vaccine effectiveness against VT-NBP for people of all ages with high risk was assumed equal to 80% of the corresponding values for low- and medium-risk persons, using a similar approach as for VT-IPD in the high-risk category [39,40].

2.2.6. Effectiveness of PPSV23

Methods to derive the effectiveness of PPSV23 against VT-IPD and VT-NBP were also described in the publication of Mendes et al. [22]. Efficacy inputs for the initial period of effectiveness of PPSV23 vaccine against VT-IPD for low-, medium-, and high-risk persons was estimated using data from Djennad et al. [41]. Initial effectiveness was derived for all ages by fitting a logarithmic curve to values for people aged 65 to 74, 75 to 84, and 85 to 99 years, respectively, from data reported in Djennad et al., and then allocating estimated age-specific values across groups using relative risks and corresponding population weights [26,41].

For the base case, efficacy inputs of PPSV23 against VT-NBP was assumed to be zero, based on various published sources and consistent with base-case assumptions used in several similar economic evaluation studies [42–57]. A scenario analysis using data reported by Suzuki et al. was performed to test moderate effectiveness of PPSV23 against VT-NBP [15].

2.2.7. Vaccine waning

For all PCV vaccines, initial vaccine effectiveness was assumed to persist for the first 5 years of the modeling time horizon for all ages and risk categories, which is consistent with CAPiTA, in which no waning of vaccine efficacy was observed in the 5-year follow-up period [37]. Beyond year 5, vaccine effectiveness was assumed to wane across all age and risk groups by 5% annual decline during years 6 to 10, 10% annual decline during years 11 to 15, and an assumption of no efficacy from year 16 onwards [38].

For PPSV23, vaccine effectiveness was assumed to wane across all age and risk groups, based on data reported by Djennad et al. as follows: linear decline to 76.2% of initial vaccine effectiveness by year 5, then linear decline to no efficacy by year 10 and through to the end of the modeling time horizon [41].

2.2.8. Herd effects from pediatric PCV15 and PCV20 use

The percentage reduction in total disease due to indirect effects (herd effects) from pediatric pneumococcal vaccines was not considered in the base case. In scenario analyses, two scenarios testing herd effects from pediatric PCV15 and PCV20 use were explored using data from Mendes et al. and Japanese epidemiology data [7,22]. The method used to estimate herd effect from Japanese epidemiology data is described in the supplemental online material and values used in this scenario are summarized in Supplementary Table S2.

2.2.9. Long-term consequences

Long-term consequences due to all-cause NBP were not considered in the base-case analysis. Previous Japanese studies and cost-effectiveness analyses of PCV20 in other countries did not consider long-term consequences due to all-cause NBP [21–23,58]. Furthermore, it was concluded that it was difficult to set the rate of long-term consequences due to all-cause NBP, which would be influenced by antibiotic use and age. Including long-term consequences due to all-cause NBP, would lead to better cost-effectiveness results as PCV20 is anticipated to reduce the number of NBP cases and associated sequelae. As a conservative approach, we have not included this parameter due to lack of evidence and expected positive impact on the results [22,59,60].

2.2.10. Mortality

Age-specific general population mortality rates were estimated using the number of Japanese population in 2022 from Japan statistics data (e-stat) and mortality rates by age and sex from Ministry of Health, Labour and Welfare (MHLW) [26,61].

Case fatality for IPD (bacteremia and meningitis) was based on a publication from Chang et al. [62]. Case fatality for NBP (hospitalized) for adults aged 50 to 64 years and ≥65 years was sourced from Miyazaki et al. [3]. Alternative values for IPD case fatality were available from Yanagihara et al. and explored in scenario analysis [63]. One scenario for NBP case fatality was explored using values from two published papers (Konomura et al. and Tashiro et al.) [64,65].

2.2.11. Utilities

Age- and sex-specific EQ-5D-5L values to inform baseline utilities for a Japanese population were sourced from Shiroiwa et al. [66]. Utility values in adults aged 50 to 64, 65 to 74, 75 to 84 and ≥85 years of age were calculated by weighting the Japanese population per age group [26]. Annual utility decrements of 0.1741, 0.0434, and 0.0289 were applied for IPD, inpatient NBP, and outpatient NBP, respectively [67,68].

2.2.12. Vaccination and medical care costs

Vaccine acquisition costs were JPY4,735 and JPY7,200 for PPSV23 and PCV15, respectively [69,70]. As of January 2024, PCV20 has not been approved in Japan. Therefore, in the base-case analysis, we assumed a cost of JPY8,102, which was calculated by multiplying the cost of PCV13 by 1.13. In the absence of other data, a value of 1.13 was used, calculated based on the ratio of pediatric PCV20 cost (USD 178.00) and PCV13 cost (USD 158.18), with vaccination costs sourced from the US Centers for Disease Control [71]. In scenario analysis, the vaccine cost for PCV20 was also assumed to be the same as PCV13 (JPY7,200), as the vaccine cost for PCV20 is anticipated to be similar to the cost of PCV13 or higher [71]. A vaccine administration cost of JPY3,250 including initial consultation, injection, and biological preparations was applied, based on the MHLW published list [72].

Medication costs per episode of each disease condition were derived using the JMDC Japanese claims database for adults aged <65 years and the MDV database for ages ≥65 years [32,33]. Details on costing methods are described in the supplemental online material and in Supplementary Table S3. Alternative costs reported in the studies by Fukuda et al. and Konomura et al. were explored in scenario analysis [64,73].

2.2.13. Costs related to productivity loss

Productivity losses were assumed to be incurred by all patients aged 60 to 64 years and by caregivers of patients aged ≥65 years in the base-case analysis. In scenario analysis, working rate and caregiver utilization rate were considered based on Japanese statistics (Population Estimates Annual Report 2022 and Labour Force Survey 2022) and MHLW data [26,74,75]. Costs related to productivity loss were derived by multiplying the number of working days missed due to PDs and vaccination. The average daily wage of a patient and caregivers was calculated based on the monthly wage (20.4 working days per month assumed) and included an annual bonus compensation, both sourced from the Wage Structure Basic Statistical Survey [76]. The number of days lost due to hospitalization, stratified by age and risk category, were obtained from Konomura et al. and JMDC data [32,64]. An assumption was made to add 3 additional days to work loss for hospitalization. The number of days lost by age and risk category due to outpatient care for all-cause NBP were obtained from Konomura et al., corresponding to the median number of prescription days of antibiotics [64]. Two scenarios for the number of days lost due to outpatient care for all-cause NBP in adults aged 60 to 64 years and ≥65 years were explored: 5 days and 10 days. The number of days lost due to vaccination was assumed at 0.5 day.

2.3. Sensitivity and scenario analyses

Deterministic sensitivity analyses (DSA) were performed to evaluate the robustness of base case results to changes in key model parameters, including disease incidence, utility, mortality, vaccine efficacy, and costs. Probabilistic sensitivity analysis (PSA) was performed to estimate the level of uncertainty surrounding the model’s incremental cost and effectiveness results. These analyses are described in detail in the supplemental online material and PSA input parameters are summarized in Supplementary Table S4. The generated DSA results are presented in the form of a tornado diagram. The PSA results are presented on a cost-effectiveness plane. In addition to sensitivity analyses, scenario analyses were also conducted (described in Supplementary Table S5). The input parameters used in scenario analyses are presented in Supplementary Table S6.

3. Results

3.1. Base-case results

The base-case results using societal and payer perspectives are presented in Tables 2 and 3. For the comparison of PCV20 with PPSV23, assuming the PCV20 price of JPY8,102, there was an incremental QALY gain of 0.0015 per person, with reduced costs of JPY22,513 per person and JPY477 per person from a societal and payer perspective respectively (Table 2).

Table 2. Results of base case in 65-year-old adults and adults at high risk aged 60 to 64 years.

Outcomes	PCV20	PCV15	PPSV23	Differences
				PCV20 vs PPSV23	PCV20 vs PCV15	PCV15 vs PPSV23
Incremental QALY (per person)	14.331	14.331	14.330	0.0015	0.0007	0.0009
Incremental costs (JPY, per person)
Medical care + vaccination	262,316	262,752	262,793	−477	−436	−41
Medical care + productivity loss + vaccination	2,787,585	2,797,827	2,810,098	−22,513	−10,242	−12,271
ICER, JPY/QALY
Societal perspective				Dominant	Dominant	Dominant
Payer perspective				Dominant	Dominant	Dominant
INMB, QALY data
Societal perspective				30,205	13,673	16,532
Payer perspective				8,170	3,867	4,302

ICER: incremental cost-effectiveness ratio; INMB: incremental net monetary benefit; JPY: Japanese yen; PCV15: 15-valent pneumococcal vaccine; PCV20: 20-valent pneumococcal vaccine; PPSV23: 23-valent pneumococcal polysaccharide vaccine; QALY: quality-adjusted life year.

When a new strategy (i.e. PCV20) is both more effective and cost saving, it is referred to as an economically ‘dominant’ strategy.

Table 3. Breakdown of health outcomes and costs in base case and scenario analysis 1 in 65-year-old adults and adults at high risk aged 60 to 64 years.

Outcomes	PCV20	PCV15	PPSV23	Differences
				PCV20 vs PPSV23	PCV20 vs PCV15	PCV15 vs PPSV23
Health outcomes, number of cases
IPD
Bacteremia	1,704	1,773	1,821	−117	−69	−47
Meningitis	148	154	158	−10	−6	−4
All-cause NBP
Hospitalized	456,176	457,223	458,637	−2,461	−1,048	−1,413
Requiring outpatient care only	790,825	794,384	799,177	−8,352	−3,559	−4,794
Number of deaths	43,581	43,681	43,807	−226	−100	−126
Costs per population in millions, JPY
Medical care	397,822	399,058	400,659	−2,837	−1,236	−1,601
Productivity loss	3,897,917	3,913,053	3,931,931	−34,013	−15,136	−18,877
Vaccination
Base case	7,079	6,517	4,979	2,100	562	1,537
Scenario 1	6,517	6,517	4,979	1,537	0	1,537

IPD: invasive pneumococcal disease; JPY: Japanese yen; NBP: non-bacterial pneumonia; PCV15: 15-valent pneumococcal vaccine; PCV20: 20-valent pneumococcal vaccine; PPSV23: 23-valent pneumococcal polysaccharide vaccine; QALYs: quality-adjusted life years.

The calculated ICERs for the probabilistic analysis were all located the south-east quadrant of the cost-effectiveness plane, which indicates PCV20 was dominant (more effective and cost saving) compared with PPSV23. The INMB at a cost-effectiveness threshold of JPY5,000,000 per QALY was JPY30,205 and JPY8,170 using a societal and payer perspective respectively. The positive INMBs indicate that PCV20 was cost-effective compared with PPSV23 at this threshold using both analysis perspectives.

Compared with PPSV23 for the base-case population, PCV20 resulted in reduced numbers of bacteremia, meningitis, hospitalized NBP and outpatient NBP cases of 117, 10, 2,461 and 8,352 respectively (Table 3). Mortality for PCV20 was reduced by 226 cases compared to PPSV23. Medical costs for the population were reduced by JPY2.84 billion and productivity loss costs by JPY34.01 billion per population, offsetting the incremental vaccination costs (JPY2.10 billion per population). Breakdown of health outcomes and costs for the other populations tested in scenarios are shown in Supplementary Table S7.

At a PCV20 price of JPY7,200 (par with PCV13), PCV20 was dominant compared with PPSV23 (Supplementary Table S8) and more cost-saving than the base-case, as the incremental vaccination cost was JPY1.54 billion per population (Table 3). The cost-effectiveness results of scenarios examining the PCV20 price are provided in Supplementary Table S9. Furthermore, analyses comparing PCV20 with PCV15 and PCV15 with PPSV23 were found to be more effective and incurred less costs (Table 2 and Supplementary Table S8).

Threshold analysis was conducted to determine the price at which the PCV20 vaccine would return an ICER of JPY5,000,000 per QALY. From a societal perspective, a PCV20 unit cost of JPY82,867 was required (JPY74,765 higher than PCV20 cost used in the base case). From a payer perspective, a PCV20 unit cost of JPY28,324 was required (which is JPY20,222 higher than the cost of PCV20 used in the base-case analysis).

3.2. Scenarios and sensitivity analyses results (PCV20 vs PPSV23)

Detailed scenario analyses results are presented in Tables 4 and 5, and Supplementary Tables S10 to S13. Using a societal perspective, PCV20 was dominant (less costly and more effective) in all scenario analyses compared with PPSV23 at a cost-effectiveness threshold of JPY5,000,000 per QALY gained. PPSV23 NBP vaccine effectiveness (scenario 13), productivity loss (scenario 18), and vaccine coverage (scenario 6) had the highest impact on the analysis results. The INMB was JPY30,205 in the base case analysis versus JPY6,180 in scenario 13, JPY11,066 in scenario 18 and JPY14,953 in scenario 6. For the payer perspective, PCV20 was either dominant or cost-effective compared with PPSV23 at a cost-effectiveness threshold of JPY5,000,000 per QALY gained. PPSV23 NBP vaccine effectiveness (scenario 13), vaccine coverage (scenarios 6) and vaccine target population (scenario C) were the top three most impactful scenarios on the INMB.

Table 4. Scenario analysis (PCV20 vs PPSV23) with different vaccine target population.

Scenario	Vaccine target population	Incremental values			ICER, JPY/QALY gained		INMB, JPY
		QALY	Medical care + productivity losses + vaccination costs (JPY)	Medical care + vaccination costs (JPY)	Societal perspective	Payer perspective	Societal perspective	Payer perspective
	Base case results: reference (65 years and those aged 60–64 years with high risk)	0.0015	−22,513	−477	Dominant	Dominant	30,205	8,170
A	≥65 years and those aged 60–64 years with high risk	0.0015	−26,518	−1,349	Dominant	Dominant	33,967	8,798
B	65 years	0.0015	−21,556	−416	Dominant	Dominant	28,864	7,724
C	65 years and those aged 60–64 years with moderate or high risk	0.0020	−28,346	−897	Dominant	Dominant	38,161	10,712
D	66–84 years	0.0017	−29,983	−1,432	Dominant	Dominant	38,424	9,873

INMB: incremental net monetary benefit; JPY: Japanese yen; PCV20: 20-valent pneumococcal vaccine; PPSV23: 23-valent pneumococcal polysaccharide vaccine.

Table 5. Scenario analysis (PCV20 vs PPSV23) in 65-year-old adults and adults at high risk aged 60 to 64 years.

Scenario	Parameters and scenario settings	Incremental values			ICER, JPY/QALY gained		INMB, JPY
		QALY	Medical care + productivity losses + vaccination costs (JPY)	Medical care + vaccination costs (JPY)	Societal perspective	Payer perspective	Societal perspective	Payer perspective
	Base-case result: reference	0.0015	−22,513	−477	Dominant	Dominant	30,205	8,170
1	Vaccination cost: PCV20 cost equal to PCV13 cost	0.0015	−22,877	−842	Dominant	Dominant	30,569	8,534
2	Vaccine history: Data from Nakashima et al. [36]	0.0015	−22,444	−472	Dominant	Dominant	30,112	8,141
3	Incidence and serotype distribution: Data from COVID-19 era [2,7,26,27,32,34,77]	0.0011	−16,355	−13	Dominant	Dominant	21,993	5,651
4	Proportion of NBP due to S. pneumoniae: 20.0% from Kobayashi et al. [35]	0.0018	−27,110	−847	Dominant	Dominant	36,278	10,015
5	Vaccine coverage: 52.0% from IASR 2023 [2]	0.002	−28,977	−615	Dominant	Dominant	38,878	10,515
6	Vaccine coverage: 20.0%	0.0008	−11,145	−236	Dominant	Dominant	14,953	4,044
7	IPD case fatality: 14.70 per 100 persons from Yanagihara et al. [63]	0.0015	−22,121	−478	Dominant	Dominant	29,686	8,044
8	NBP case fatality: Data from Konomura et al. [64] and Tashiro et al. [65]	0.0015	−21,507	−480	Dominant	Dominant	28,844	7,817
9	Vaccine effectiveness: Increase of 10%	0.0017	−24,759	−655	Dominant	Dominant	33,173	9,069
10	Vaccine effectiveness: Decrease of 10%	0.0014	−20,267	−300	Dominant	Dominant	27,237	7,270
11	Vaccine effectiveness: Increase of 5%	0.0015	−22,329	−467	Dominant	Dominant	29,963	8,101
12	Vaccine effectiveness: Decrease of 5%	0.0013	−19,388	−225	Dominant	Dominant	26,098	6,935
13	PPSV23 effectiveness against NBP: 33.5% from Suzuki et al. [15]	0.0004	−4,261	997	Dominant	2,596,979	6,180	922
14	Herd effects from pediatric PCV20: Data from Mendes 2022 [22]	0.0015	−21,206	−389	Dominant	Dominant	28,491	7,674
15	Herd effects from pediatric PCV20: Data from Japanese epidemiology data [7]	0.0013	−18,388	−160	Dominant	Dominant	24,783	6,555
16	Medication costs: Data from Fukuda et al. [73] and Konomura et al. [64]	0.0015	−21,838	197	Dominant	128,186	29,530	7,495
17	Productivity loss: Exclude vaccine productivity loss	0.0015	−15,168	−477	Dominant	Dominant	22,860	8,170
18	Productivity loss: Apply employment and caregiver utilization rate based on data from public sources [26,74,75]	0.0015	−3,374	−477	Dominant	Dominant	11,066	8,170
19	Productivity loss: 5 days lost because of NBP outpatient care	0.0015	−22,222	−477	Dominant	Dominant	29,914	8,170
20	Productivity loss: 10 days lost because of NBP outpatient care	0.0015	−22,949	−477	Dominant	Dominant	30,641	8,170

IASR: Infectious Agents Surveillance Report; INMB: incremental net monetary benefit; IPD: invasive pneumococcal disease; JMDC: Japan Medical Data Center Inc.; JPY: Japanese yen; MDV: Medical Data Vision Co., Ltd; NBP: non-bacterial pneumonia; OP: Outpatient; PCV13: 13-valent pneumococcal vaccine; PCV20: 20-valent pneumococcal vaccine; PPSV23: 23-valent pneumococcal polysaccharide vaccine.

The parameters with the greatest impact on INMB in the DSA when comparing PCV20 to PPSV23 were disease incidence of inpatient NBP; PCV vaccine effectiveness against NBP and inpatient NBP case fatality (see Figure 2 for detailed results).

Figure 2. Deterministic sensitivity analysis results: tornado diagram.

INMB: incremental net monetary benefit; NBP: non-bacterial pneumonia; PCV20: 20-valent pneumococcal vaccine; PPSV23: 23-valent pneumococcal polysaccharide vaccine.

For PCV20 cost, a lower value of 11.13% was applied as we assumed the lowest PCV20 cost would be par to PCV13 cost.

PCV20 was less costly and more effective (cost-saving) compared with PPSV23 in 100% of the 1,000 PSA simulations from a societal perspective (Figure 3(a)). The probability of PCV20 being cost-effective against PPSV23 was 100% at a cost-effectiveness threshold of JPY5,000,000 per QALY gained. PCV20 was less costly and more effective (cost-saving) compared with PPSV23 in 63% of the 1,000 PSA simulations using a payer perspective (Figure 3(b)). The remaining model simulations were in the northeast quadrant, indicating that PCV20 was more costly and more effective compared to PPSV23.

Figure 3. Probabilistic sensitivity analysis results for the cost-effectiveness plane (PCV20 vs PPSV23).

PCV20: 20-valent pneumococcal vaccine; PPSV23: 23-valent pneumococcal polysaccharide vaccine; QALY: quality-adjusted life year

The red line corresponds to the cost-effectiveness threshold of JPY5,000,000 per QALY gained used in this analysis.

4. Discussion

We demonstrated that PCV20 would be a dominant strategy in comparison with PPSV23 (current NIP vaccine) from a societal perspective when the PCV20 price was set as JPY8,102 in adults aged 65 years and high-risk adults aged 60 to 64 years. Use of a payer perspective and other scenario analysis support this conclusion, where PCV20 was dominant or cost effective compared to PPSV23 at a cost-effectiveness threshold of JPY5,000,000 per QALY gained. PCV15 was also shown to be a dominant strategy in comparison with PPSV23, but when compared with PCV20, PCV20 was the dominant strategy because it provided greater health benefits at a lower cost than PCV15.

In Japan, PCV20 was submitted to PMDA in 2023 for approval. Therefore, the price of PCV20 has not yet been established as of January 2024. In this study, we used a cost of JPY8,102 for the base case and JPY7,200 (parity to PCV13) in scenario analysis. Sensitivity analysis revealed the price of PCV20 was one of the key drivers of the ICER. Therefore, threshold analysis was conducted and shows that using a higher PCV20 cost of JPY87,133 (societal perspective) or JPY28,324 (payer perspective) would still return an ICER below the JPY5,000,000 per QALY threshold. This result will be informative for discussions about PCV20 price and the NIP for PCV20.

During the COVID-19 pandemic, the incidence rate of PD reduced markedly due to several safety measures that were applied globally to curb the spread of coronavirus infection. For example, the incidence rate of IPD was 5.43 and 5.57 per 100,000 in 2018 and 2019 (before the COVID-19 pandemic), and 2.77 and 2.21 per 100,000 in 2020 and 2021 (during COVID-19 pandemic) in adults aged ≥65 years [2]. Therefore, this study employed the incidence and serotype distribution data before the COVID-19 period (2017 to 2019) for the base-case analysis. In the scenario analysis that utilized a lower PD incidence and serotype distribution data during the COVID-19 era (2020 to 2022), the results showed that the base-case cost-effectiveness outcomes remained unchanged as a dominant strategy.

In addition to the adult PCV20 vaccine, the use of PCV20 for a pediatric population has been submitted for approval in Japan [78]. Several reports suggest that pneumococcal vaccines exert an indirect effect (i.e. herd immunity) [79]. Therefore, a scenario analysis examining herd immunity from use of pediatric PCV20 was considered by assuming that PCV20 would be widely used in the pediatric population (scenarios 14 and 15). These scenarios showed that PCV20 was still dominant compared with PPSV23. At present, only PCV13 is a NIP target for a pediatric population since 2013. In the model, we assumed that the proportion of IPD due to PCV13 serotypes should be assumed to have reached a steady state prior to model entry, and thus remain constant for the entirety of the modeling horizon. As the scenario we previously mentioned indicates, it is anticipated, however, that future pediatric use of PCV20 would result in a decline in the proportion of adult disease attributable to additional serotypes included in the novel vaccines.

In the base case, we assumed that PPSV23 had no effectiveness against NBP. A similar assumption was made in other studies evaluating PCV20 [22,58,80,81]. We used that assumption here because data to support the argument that PPSV23 affects NBP are inconsistent, and any inclusion of this effect would be uncertain and should be considered with caution [15,22,50]. However, assuming effectiveness of PPSV23 against NBP does not change the conclusions of the analysis. Suzuki et al. reported that PPSV23 showed low to moderate effectiveness against vaccine serotype pneumococcal pneumonia in people aged 65 years or older [11]. In the scenario analysis, we assumed that vaccine effectiveness of PPSV23 against NBP was 33.5% but PCV20 was still cost saving from a societal perspective and cost effective from a payer perspective at a cost-effectiveness threshold of JPY5,000,000 per QALY gained.

Our findings are consistent with a published study considering the use of PCV20 compared to PPSV23 in Japan. Hoshi et al. evaluated the cost-effectiveness of PCV20 and PCV15 compared with PPSV23 in an elderly Japanese population aged 65 years, using a payer perspective [21]. Higher costs than those considered in the current study for PCV15 (JPY16,489) and PCV20 (JPY19,030) were used in that study, based on the US CDC vaccination costs [71]. However, the results of that analysis found that the PCV20 vaccination was cost saving compared with PPSV23, supporting the findings of our analysis. Other key differences include different model structures (Hoshi et al. developed a decision tree linked to a Markov model, whereas our analysis used a Markov model structure) and use of a third-party payer perspective rather than a societal perspective [21]. There was also a difference in key inputs, including the modeled population age (our analysis included adults aged 60 to 64 years who were at high risk), and Hoshi et al. did not stratify the model population by risk severity [21]. The findings indicate that the introduction of PCV20 in adult NIP would be meaningful in Japan.

PCV20 has been approved in other jurisdictions, including the United Kingdom, Denmark, and Germany. Several publications evaluated PCV20 compared with the current standard of care, PPSV23, in these countries, using the same model structure as our analysis [22,58,80,81]. Despite differences in perspective, population age, and model inputs, each study found that PCV20 would be a cost saving and effective strategy compared with PPSV23. The findings of these studies support the argument for the introduction of PCV20 as a cost-saving alternative treatment in comparison to PPSV23.

The model analysis has several limitations, including data limitations. First, a comparison of PCV20 with a ‘no vaccine’ scenario was not considered in this analysis because several vaccines are already available, such as PPSV23, which is the current standard vaccine in the NIP. We considered this comparison would be non-informative given there are vaccines already used in practice.

Secondly, we made our best effort to use robust inputs estimates for Japan collected via targeted literature reviews. However, there were limited available data to inform several inputs, including vaccine history, caregiver work loss, percentage of NBP due to vaccine serotypes, and reduction in total disease, which was assumed to be zero. To test the uncertainty associated with these parameters, we ran scenario analyses; PCV20 remained the dominant strategy, suggesting it is likely to be cost saving in a Japanese healthcare setting.

Thirdly, we applied the risk distribution to the age groups without any adjustment to distribution and without taking real-world observations into account. We also used assumptions where data on several parameters, such as IPD case fatality, case fatality for hospitalized NBP, hospitalization cost data, productivity loss, and utility data, were available but incomplete for our target population because they were not specified by age group or risk stratification. To address data gaps, we assumed the same inputs were applicable across age groups and risk groups. However, we expect more granular data would have had little impact on the results, given the robustness of the analysis findings.

Additionally, there is uncertainty regarding vaccine effectiveness. Vaccine effectiveness estimates of all PCVs were based on studies showing non-inferior immunogenicity comparing PCV13 or PPSV23 [17,82]. Therefore, PCV20 and PCV15 effectiveness against PD was assumed to be the same as PCV13 effectiveness against PD in the CAPiTA trial [37]. In clinical practice, PCV20 and PCV15 effectiveness may differ from PCV13 effectiveness. To examine this, we tested several scenarios, which had minimal impact on results.

Finally, given that economic models are a simplification of reality, we could not add all relevant inputs, and we did not include long-term clinical costs and consequences of PD, although evidence shows that these sequelae are common and costly [83–87]. Likewise, we did not consider adverse events, given that these are expected to be rare and similar across all vaccines. Additionally, because the price for PCV20 in Japan is unknown, we assumed that the cost of PCV20 in Japan would be JPY3,367 higher than that of PPSV23. The analysis showed that even at the assumed higher cost, PCV20 was a cost-saving strategy compared with PPSV23.

5. Conclusions

Our analysis supports the case for the introduction of PCV20 in Japan, with the findings demonstrating that PCV20 is both cost saving and more effective than currently available vaccines for all adults aged 65 years and those aged 60 to 64 years at high risk. Given the prolonged duration of protection and greater effectiveness associated with PCV20, the introduction of PCV20 has the potential to substantially reduce healthcare burden and positively affect morbidity and mortality related to PD in Japan.

Declaration of interest

S Nakamura declares that he receives honoraria from Pfizer Japan Inc., for delivering promotional lectures and a consultation fee. M Mikami, T Hayamizu, N Yonemoto, and K Kamei are employees of Pfizer Japan Inc., and J Vietri is an employee of Pfizer Inc. C Crossan, C Moyon, M Gouldson, Y Onishi, M Nomoto, and F Khrouf are employees of Putnam. Putnam was paid by Pfizer Japan Inc. for conducting the study and medical writing support. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or material discussed in the manuscript apart from those disclosed.

Reviewer disclosures

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

Author contributions

All authors have substantially contributed to the conception, design, and execution of this study. In addition, all authors have substantially contributed to the analysis and interpretation of the study findings. Finally, all authors have read and agreed to the published version of the manuscript.

Acknowledgments

The authors would like to thank Ataru Igarashi of Department of Public Health, Yokohama City University School of Medicine, and Department of Health Economics and Outcomes Research, Graduate School of Pharmaceutical Sciences, The University of Tokyo for providing valuable scientific advice. The authors also thank Yoshie Onishi, Mariko Nomoto, and Fatma Khrouf of Putnam, who conducted targeted literature reviews and desk research to inform model input parameters for Japanese healthcare setting.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14760584.2024.2350246

Additional information

Funding

This manuscript was funded by Pfizer Japan Inc.