Incremental net monetary benefit of herpes zoster vaccination: a systematic review and meta-analysis of cost-effectiveness evidence

Abstract

Objectives

To perform a systematic review and meta-analysis to pool the incremental net benefit (INB) of each herpes zoster vaccine [i.e. Zoster Vaccine Live (ZVL) and Recombinant Zoster Vaccine (RZV)].

Methods

We initially identified individual studies by hand-searching reference lists of the relevant systematic review articles. An updated comprehensive search was performed in Medline, Scopus, and Embase until June 2020 for additional studies. Studies were eligible if they assessed the cost-effectiveness/utility of any pair among ZVL and RZV, and no vaccine and reported economic outcomes. Details of the study characteristics, economic model inputs, costs, and outcomes were extracted. INB was calculated with monetary units adjusting for purchasing power parity for 2019 US dollars and pooled by meta-analysis.

Results

A total of 37 studies were pooled for meta-analysis stratified by perspectives [i.e. societal (SP) and third-party payer (TPP)] and vaccine types. In SP, ZVL was cost-effective compared to no vaccine when vaccinated at ages of 50–59 and 70–79 years with INBs (95% CI) of $0.61 (0.37, 0.85) and $9.67 (5.20, 14.14), respectively. RZV was cost-effective for those aged 60–69 and 70–79 years with INBs of $75.61 (17.98, 133.23) and $85.01 (30.02, 140.01), respectively. In TPP, ZVL was cost-effective compared to no vaccine when vaccinated at age 70–79 years with INB of $7.57 (0.27, 14.86) and RZV was cost-effective at 60–69 years with INB $220.87 (47.80, 393.93). The cost-effectiveness of RZV was robust across a series of sensitivity analyses, but ZVL differs on different vaccination ages.

Conclusions

RZV may be cost-effective for vaccination in ages of 60–79 years for both SP and TPP perspectives, while ZVL might be cost-effective in some age groups, but results are not robust.

Keywords:

• Herpeszoster
• vaccine
• systematic review
• meta-analysis
• economic evaluation

JEL CLASSIFICATION CODES:

• A10
• A1
• A
• C00
• C

Introduction

Herpes zoster (HZ), also known as shingles, caused by reactivation of latent varicella-zoster virus (VZV) or chickenpox reservoir, occurs frequently in individuals aged 50 years or older^1,2. HZ and its complications, such as post-herpetic neuralgia (PHN), can have a substantial negative impact on quality of life. The incidence of herpes zoster has been increasing throughout the world, causing substantial morbidity¹. In 2013, there were more than 1 million cases of HZ in the United States (US) alone, with the direct and indirect medical costs of infection estimated at $1.3 billion and $1.7 billion, respectively, annually^3,4. This burden is expected to rise significantly due to an aging population and an increasingly large pool of at-risk individuals worldwide^5,6. Hence, the prevention of HZ is an important public health objective.

Currently, two HZ vaccines are available [i.e. the non-live subunit Recombinant Zoster Vaccine (RZV) and the attenuated Zoster Vaccine Live (ZVL)] which are approved by the Food and Drug Administration (FDA) for adults aged 50 years or older. ZVL is a single-dose regimen, whereas RZV is a two-dose regimen, in which a booster should be administered 2–6 months after the first dose. Current recommendations from the Advisory Committee on Immunization Practices (ACIP) prefer RZV over ZVL for most patients because it provides greater protection, especially among people in their seventh to the ninth decade⁷. Moreover, ZVL is gradually being replaced by RZV in many countries⁸. For example, as of November 2020, ZVL is no longer available in the US⁹; however, ZVL continues to be used in many high-income countries since RZV is not yet widely available because of its limited supply⁸. Additionally, a few countries still incorporated ZVL in their immunization programs funded through National Health System (e.g. Australia, Italy, France, etc.)⁸. Meanwhile, HZ vaccination has not become routine practice in many low- and middle-income countries (LMICs) because of the limited availability of both vaccines, and ZVL remains the first option in these countries^10,11.

Cost-effectiveness analyses (CEAs) provide information to decision-makers for efficient use of available resources for maximizing health benefits. A previous systematic review of CEAs of HZ vaccination summarized that most included studies found HZ vaccination cost-effective¹². However, the evidence was largely available for only ZVL because of the limited number of CEAs on RZV in these reviews. The cost-effectiveness of HZ vaccination seemed to be influenced by vaccine types and their price, age groups, perspectives of analyses, region or country-level income, and many other factors¹². It is important for global decision-makers from both resource-rich and resource-poor countries to know how these factors influence the cost-effectiveness of HZ vaccination. However, previous reviews provided only an overall descriptive summary of evidence from individual cost-effectiveness studies. Recently, Crespo et al. suggested a meta-analysis method of CEAs¹³. This approach is now getting recognition by the wider scientific community^14–18. The evidence generated from a meta-analysis of CEAs was found to be useful to decision-makers in several aspects as it provides (1) an overall summary of all economic evidence quantitatively, and (2) a summary stratified by perspectives [e.g. societal (SP), third-party payer (TPP), etc.], vaccine types, age groups, and other factors as mentioned above, to provide specific evidence for each group with less heterogeneity. Immunization and Vaccine-related Implementation Research Advisory Committee (IVIR-AC)-WHO (March 2021) agrees that meta-analysis of CEAs is useful and could facilitate decision-making in countries where context-specific CEAs are not available¹⁹.

Therefore, we conducted a systematic review and meta-analysis of CEAs by pooling incremental net benefit (INB) to assess the cost-effectiveness of ZVL and RZV compared to no vaccine (NoV) and each other in patients aged 50 years and older. The purpose of this research is to emphasize the economic values of HZ vaccines and inform policymakers especially in those countries where economic evidence is limited for the selection of two choices of HZ vaccine.

Methods

This study was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement (Supplementary Appendix I). The review protocol was registered with Open Science Framework (OSF) and can be found at osf.io/4f867.

Data source and searches

We performed an updated comprehensive search to identify additional individual studies using MEDLINE via PubMed, Scopus, and EMBASE from the date of the last search conducted by Chiyaka et al. in March 2018 to 9 June 2020¹². Search terms used were “herpes zoster,” “vaccination,” and “economic evaluation” (Supplementary Appendix II).

Study selection

Titles, abstracts, and full texts of the studies identified by the updated search were screened by two independent reviewers (SU and MR). Individual studies from the updated search and Chiyaka et al. were included if they met the following criteria: conducted CEA or cost-utility analysis (CUE) by comparing any pair among ZVL, RZV, and NoV and reported economic outcomes. Studies were excluded if they did not provide sufficient economic data. Any disagreement in study selection was discussed with a third reviewer (NC).

Data extraction and quality assessment

We extracted data including vaccine data (i.e. type, dose, and administration schedule), comparator, World Health Organization (WHO)-epidemiological regions, country income level or GDP (per the World Bank 2019 report), time horizon, study perspective, discount rate for costs and effects, model used, population of interest, data source, sensitivity analysis results, cost (C), incremental cost (ΔC), willingness-to-pay threshold (K), effectiveness/utility (E), incremental effectiveness/utility (ΔE), and the incremental cost-effectiveness ratio (ICER), along with their dispersions [standard deviation (SD), standard error (SE), or 95% confidence interval (CI)]. Effectiveness was measured as quality-adjusted life years (QALYs) or life-years (LYs) gained. The risk of bias was independently assessed at the study level by the same reviewers above using the modified Economic Evaluations Bias (ECOBIAS) checklist, which consists of 22 items²⁰. Each item was graded as yes, partly, unclear, no, or not applicable.

Data analysis and synthesis

Interventions

There were three comparisons of interest: ZVL vs. NoV, RZV vs. NoV, and RZV vs. ZVL. ZVL was a single-dose regimen, while RZV was a two-dose regimen. Other regimens of ZVL and RZV were also included, such as single-dose ZVL vs. ZVL with booster and RZV in the previous ZVL vs. ZVL.

Outcome of Interest

An incremental net benefit (INB) (also recognized as an incremental net monetary benefit) was the primary outcome of interest which was calculated using Equation (1) or (2) depending on the parameters reported in the individual study^13,21,22. The variances of INB were calculated using Equations (3) or (4).

(1) $$\mathrm{INB}=\text{ΔE }\left(\mathrm{K}-\mathrm{ICER}\right)$$(1) (2) $$\mathrm{INB}=(\mathrm{K}\times \mathrm{\Delta }\mathrm{E})-\mathrm{\Delta }\mathrm{C}$$(2) (3) $$\mathrm{Var}\left(\mathrm{INB}\right)={\mathrm{K}}^2{\mathrm{\sigma }}_{\text{ΔE}}^2+{\mathrm{\sigma }}_{\mathrm{ICER}}^2$$(3) (4) $$\mathrm{Var}\left(\mathrm{INB}\right)={\mathrm{K}}^2{\mathrm{\sigma }}_{\text{ΔE}}^2+{\mathrm{\sigma }}_{\text{ΔC}}^2-2\mathrm{K}{\mathrm{\rho }}_{\text{ΔCΔE}}\mathrm{ }$$(4) K is the willingness-to-pay threshold, ΔE and ΔC are the incremental effectiveness and cost, and ΔCΔE is the covariance between ΔC and ΔE. A positive INB indicates the intervention is cost-effective, whereas a negative INB indicates the intervention is not cost-effective and the comparator is preferred.

Data preparation

Data were prepared according to five scenarios described elsewhere and in Supplementary Appendix IIIa^14–18,22. INB and its variance were then calculated accordingly. All cost data including K, ΔC, ICER, and their variances in each currency were standardized by converting to 2019 United States Dollar (USD) using consumer price index (CPI)²³ and purchasing power parity (PPP)²⁴ (Supplementary Appendix IIIb).

Statistical analysis

A meta-analysis was applied to pool INBs across studies using the Dersimonian-Laid random-effects model if heterogeneity was present (I² > 25% or the Q test p < .1)²⁵; otherwise, a fixed-effect inverse-variance model was applied (Supplementary Appendix IIIc). Heterogeneity was assessed by Cochrane’s Q test and I² statistic²⁶. Sources of heterogeneity were explored by considering the cost of vaccine and comparator covariables in a meta-regression model one by one. If the I² decreased by 50% or more, such a covariable was considered a source of heterogeneity. Publication bias was assessed using a funnel plot and Egger’s test. The source of asymmetry was further explored using a contour-enhanced funnel plot by contouring the funnel as significant (i.e. <0.01, <0.05) and non-significant areas (i.e. >0.05–0.1)²⁷. If missing studies were in a non-significant area, publication bias might be a source of heterogeneity; if studies were mostly in both significant and non-significant areas, heterogeneity might play a role.

If studies reported the cost-effectiveness of vaccines for different perspectives, vaccine types, and age groups, the studies were considered and accounted for more than once according to the number of comparisons reported. All analyses were stratified by country-level income, perspectives (SP and TPP), and vaccine types (ZVL and RZ). Vaccination age was used in subgroup analyses. A series of prespecified sensitivity analyses were performed as follows: (1) imputation of variance values by borrowing values from other studies with similar characteristics, (2) analysis using the average values of vaccination age and INB across different vaccination age groups reported within-study, (3) excluding studies with missing variances from the analyses, and (4) excluding studies with a high risk of bias. Analyses were performed using Microsoft Excel and STATA version 16.0 and results were considered statistically significant for all analyses at p < .05 (two-sided).

Results

Study selection

Twenty-seven individual studies were identified from a previous systematic review¹². We then performed an updated search and identified additional 387 studies. After screening 387 studies, 10 met the inclusion criteria and were added to the 27 previously identified studies. Therefore, a total of 37 studies consisting of 110 comparisons were included in the meta-analysis (Figure 1). However, 13 comparisons were not included in the quantitative analysis because 12 were comparisons of various schedules of ZVL and RZV, and only one study compared RZV vs. ZVL in TPP, therefore there was not an adequate number of included studies for pooling. Finally, 97 comparisons from 37 studies were included in the synthesis (Figure 1).

Figure 1. Study selection process.

Study characteristics

According to WHO-epidemiological regions²⁸, 17 studies were from European Region (EUR), 14 from the Americas Region (AMR), and 6 from Western Pacific Region (WPR) (Table 1). All studies were conducted in high-income countries. Most studies used decision-analytic or Markov model (35, 94.6%) and lifetime time horizon (33, 89.2%). Non-lifetime time horizons of 7.5, 10, and 15 years were used in 10.8% of the studies^31,37,56.

Table 1. Characteristics of included studies (n = 37).

Author	Country	WHO region	Time horizon	Model	Currency year	Threshold	Perspective	Vaccine comparison	Vaccination age
Hornberger 2006^29	USA	AMR	Lifetime	Decision theoretical model with health states	2006	$100,000	Societal	ZVL vs. NoV	Base case 69 60, 65, 70, 75, 80
Pellissier 2007^30	USA	AMR	Lifetime	Decision analytic model with pain states	2006	$50,000	Societal TPP	ZVL vs. NoV	60+
Rothberg 2007^31	USA	AMR	10 years	Markov	2005	$100,000	Societal	ZVL vs. NoV	60+
Brisson 2008^32	Canada	AMR	Lifetime	Markov	2005	CA$40,000	TPP (MoH)	ZVL vs. NoV	Base case 65 50, 60, 70, 80
Najafzadeh 2009^33	Canada	AMR	Lifetime	Discrete-event simulation	2008	CA$50,000	TPP (HCP)	ZVL vs. NoV	60+
van Hoek 2009^34	UK	EUR	Lifetime	Markov	2006	£30,000	HCP	ZVL vs. NoV	Base case 65 60, 70, 75
Annemans 2010^35	Belgium	EUR	Lifetime	Markov	2007	€30,000	Societal HCP TPP	ZVL vs. NoV	60+
Moore 2010^36	UK	EUR	Lifetime	Markov	2006	£30,000	TPP (NHS) Societal	ZVL vs. NoV	50+
van Lier 2010^37	Netherlands	EUR	7.5 years	Markov	2008	€20,000	Societal	ZVL vs. NoV	Base case 70 60, 65, 75, 80
Szucs 2011^38	Switzerland	EUR	Lifetime	Markov	2011	CHf30,000	Societal TPP	ZVL vs. NoV	Base case 70–79
Bresse 2013^39	France	EUR	10 years	Markov	2010	€28,000	HCP TPP	ZVL vs. NoV	65+, 70–79
de Boer 2013^40	Netherlands	EUR	Lifetime	Markov	2010	€50,000	Societal	ZVL vs. NoV	Base case 60 65, 70, 75
Ultsch 2013^41	Germany	EUR	Lifetime	Markov	2010	€30,000	Societal	ZVL vs. NoV	60
Preaud 2015^42	Germany	EUR	Lifetime	Markov	2013	€45,000	Societal HCP (SHI)	ZVL vs. NoV	50+, 60+
Le 2015^43	USA	AMR	Lifetime	Markov	2014	$100,000	Societal	ZVL vs. NoV	50
Belchoir 2016^44	France	EUR	Lifetime	Markov	2012	€50,000	Societal	ZVL vs. NoV	60–64, 65–69, 70–74, 75–79
Bilde 2016^45	Denmark	EUR	Lifetime	Markov	2011	Kr270,000	Societal	ZVL vs. NoV	65+
Coretti 2016^46	Italy	EUR	Lifetime	Markov	2014	€30,000	Societal TPP (I-NHS)	ZVL vs. NoV	60–79
Lopez-Belmonte 2016^47	Spain	EUR	Lifetime	Markov	2013	€30,000	Societal TPP	ZVL vs. NoV	50+
Blank 2017^48	Switzerland	EUR	Lifetime	Markov	2014	CHf40,000	TPP	ZVL vs. NoV	65–79
Hoshi 2017^49	Japan	WPR	Lifetime	Markov	2016	¥5,000,000	Societal	ZVL vs. NoV	Base case 65–84 70–84, 75–84, 80–84
Pan 2017^50	Singapore	WPR	Lifetime	Decision analytic Markov	2013	SG$65,000	Societal	ZVL vs. NoV	50+
Le 2017^51	USA	AMR	Lifetime	Markov	2014	$100,000	Societal	ZVL vs. NoV ZVL + booster vs. NoV ZVL + 2 boosters vs. NoV	60, 70
Le 2018^52	USA	AMR	Lifetime	Markov	2016	$50,000	Societal	RZV vs. ZVL RZV vs. NoV	60, 70, 80
Carpenter 2019^53	USA	AMR	Lifetime	Markov	2018	$100,000	Societal	ZVL vs. NoV RZV vs. NoV	50, 60, 70
Curran 2018^54	USA	AMR	Lifetime	Markov	2016	$100,000	Societal	RZV vs. NoV RZV vs. ZVL	60+
Curran 2019^55	USA	AMR	Lifetime	Markov	2016	$100,000	Societal	RZV in previous ZVL vs. ZVL RZV vs. ZVL + booster	60+
de Boer 2018^56	Netherlands	EUR	15 years	Markov with decision tree	2017	€20,000	Societal	ZVL vs. NoV ZVL + booster vs. NoV RZV vs. NoV	50, 60, 70, 80
Drolet 2019^57	Canada	AMR	Lifetime	Decision analytic static cohort	2018	CA$45,000	HCP	RZV vs. NoV ZVL vs. NoV	65
Hoshi 2019^58	Japan	WPR	Lifetime	Markov with decision tree	2017	¥5,000,000	TPP	RZV vs. NoV ZVL vs. NoV	65–84, 70–84, 75–84, 80–84
McGirr 2019^59	Canada	AMR	Lifetime	Markov	2016	CA$50,000	HCP	RZV vs. NoV RZV vs. ZVL	60+
Melegaro 2018^60	Italy	EUR	Lifetime	Stochastic individual based model	2015	€40,000	Societal	ZVL vs. NoV ZVL + booster vs. NoV	65
Prosser 2019^61	USA	AMR	Lifetime	Simulation (state-transition)	2016	$100,000	Societal	RZV vs. NoV ZVL vs. NoV	50–59, 60–69, 70–79, 80–89, 90–99
Shiragami 2019^62	Japan	WPR	Lifetime	Markov	2017	¥5,000,000	TPP Societal	RZV vs. NoV	65+
van Oorschot 2019^63	Germany	EUR	Lifetime	Markov	2017	€50,000	Societal	RZV vs. NoV	Base case 60+, 70+ 60, 65, 70
You 2018^64	Hong Kong, China	WPR	Lifetime	Markov	2016	$130,590	Societal	RZV vs. NoV	50, 60, 70
You 2019^65	Hong Kong, China	WPR	Lifetime	Markov	2017	$46,153	Societal	RZV vs. NoV	50, 60, 70, 80

Note. All studies conducted cost-utility analysis (CUA).

Of the 37 studies, 97 comparisons of HZ vaccines were identified according to different vaccines, perspectives, and age groups reported within the same study. Of the 97 comparisons, 74 (76.3%) used SP and 23 (23.7%) used TPP (Supplementary Appendix IV). Comparison pairs were single-dose ZVL with NoV (n = 61), 2-dose RZV with NoV (n = 30), and RZV with ZVL (n = 6). Vaccination ages were diversely reported as distinct ages, i.e. 50, 60, and up to 99 years, in ranges of 50–59, 60–69, and 70–79, and in open ranges of 50+, 60+, and 70+. Therefore, vaccination ages from all studies were combined and grouped into five subgroups including 50+ to 70+, 50–59, 60–69, 70–70, and 80–99 years. Data pooling was finally performed to compare INBs of different perspectives and vaccines; these included SP and TPP of ZVL vs. NoV, RZV vs. NoV, and RZV vs. ZVL, resulting in a total of 97 comparisons from 37 studies.

Risk of bias assessment

Risk of bias assessment was performed using the ECOBIAS checklist (Supplementary Appendix V). Overall, most studies had a similar bias profile. Bias related to reporting and dissemination were unclear for most studies. Studies that did not demonstrate bias related to cost measurement, structural assumptions, treatment comparator, model type, time horizon, data identification, baseline data, treatment effects, quality-of-life weights (utilities), data incorporation, and limited scope were determined to have a low risk of bias [15 studies (39.5%)].

Pooled INBs based on societal perspective

ZVL vs. NoV

Twenty-one studies with 42 comparisons of single-dose ZVL vs. NoV were included in pooling INBs (Figure 2). All studies used the decision-analytic or Markov model and lifetime time horizon. One study used a stochastic individual-based model⁶⁰ and three studies^31,37,56 used non-lifetime time horizon. Overall INB indicated that ZVL was not cost-effective relative to NoV [pooled INB (95% CI) −$5.78 (−8.71, −2.85)] (Table 2). For age subgroups, ZVL was cost-effective compared to NoV in ages 50–59 and 70–79 years with INBs of $0.61 (0.37, 0.85) and $9.67 (5.20, 14.14), respectively. Egger’s test showed asymmetry of the funnel in only 70–79 years (coefficient = 1.08, SE = 0.446, p = .0158), and a contour-enhanced funnel plot showed that most fell in the non-significant area, which indicated that asymmetry was likely due to heterogeneity rather than publication bias (Supplementary Appendix VIIa).

Figure 2. Pooling the INB of ZVL vs. NoV for societal perspective by age groups.

Table 2. Meta-analysis results for main groups.

Analysis groups and vaccination age (years)	n Studies	n Analyses	INB (2019 USD)	95% CI	I^2	p-Value	Egger’s test
							Coefficient	SE	p-Value
SP–ZVL vs. NoV
50+ to 70+		11	19.429	−7.194, 46.052	0.00	.925	0.43	0.540	.4257
50–59		3	0.610	0.369, 0.850	0.00	.387	−0.76	0.708	.2842
60–69		13	−17.945	−25.030, −10.861	99.90	<.001	0.73	0.431	.0884
70–79		12	9.673	5.204, 14.142	99.75	<.001	1.08	0.446	.0158
80–99		3	−16.864	−56.883, 23.154	99.89	<.001	0.73	1.088	.5010
Overall	21	42	−5.776	−8.706, −2.847	99.78	<.001	0.55	0.204	.0076
SP–RZV vs. NoV
50+ to 70+		4	24.056	−34.704, 82.816	0.00	.658	0.76	0.916	.4050
50–59		5	30.582	−8.821, 69.985	62.48	.031	0.93	0.678	.1714
60–69		6	75.607	17.981, 133.233	78.01	<.001	1.23	0.686	.0734
70–79		6	85.013	30.022, 140.005	85.67	<.001	2.48	0.645	.0001
80–99		5	24.863	−15.750, 65.476	71.78	.007	2.25	0.610	.0002
Overall	9	26	49.504	28.261, 70.747	79.21	<.001	1.52	0.292	<.001
SP–RZV vs. ZVL
50+ to 70+		1	228.408	−0.795, 457.611	–	–	–	–	–
50–59		1	105.706	−139.734, 351.147	–	–	–	–	–
60–69		1	222.315	−22.748, 467.378	–	–	–	–	–
70–79		1	388.518	25.515, 751.521	–	–	–	–	–
80–99		2	528.237	231.837, 824.637	0.00	1.00	–	–	–
Overall	2	6	266.671	143.377, 389.965	6.14	.257	3.50	1.622	.0352
TPP–ZVL vs. NoV
50+ to 70+		11	8.625	3.622, 13.629	0.00	.773	0.40	0.349	.2480
50–59		1	−24.548	−153.530, 104.435	–	–	–	–	–
60–69		4	12.465	−45.485, 70.415	0.00	.999	0.18	1.543	.9080
70–79		2	7.566	0.268, 14.863	0.00	.785	0.29	1.057	.7866
80–99		1	−36.684	−145.272, 71.905	–	–	–	–	–
Overall	14	19	8.210	4.099, 12.321	0.00	.913	0.19	0.264	.4754
TPP–RZV vs. NoV
50+ to 70+		3	54.703	−109.168, 218.575	0.00	.895	−0.73	1.592	.6487
60–69		1	220.869	47.804, 393.935	–	–	–	–	–
Overall	4	4	133.255	14.263, 252.248	0.00	.172	−1.50	1.359	.2708

SP: societal perspective; TPP: third-party payer perspective; ZVL: zoster vaccine live; RZV: recombinant subunit vaccine; NoV: no vaccine; SE: standard error.

RZV vs. NoV

Nine studies with 26 comparisons were included in pooling INBs of RZV vs. NoV (Figure 3). Most of the studies used the Markov model and lifetime time horizon, except one study used simulation (state-transition model)⁶¹ and another study⁵⁶ used non-lifetime time horizon. Overall INB indicated that RZV was cost-effective compared to NoV with the INB (95% CI) of $49.5 (28.2, 70.7) (Table 2). In addition, RZV was more cost-effective in age groups 60–69 and 70–79 years with INBs of $75.61 (17.98, 133.23) and $85.01 (30.02, 140.01), respectively. Egger’s test showed significant publication bias for 70–79 (coefficient = 2.48, SE = 0.645, p = .001), and 80–89 years (coefficient = 2.25, SE = 0.610, p = .0002). The contour-enhanced funnel plot showed studies in both significant and non-significant areas which suggested that asymmetry was likely due to heterogeneity rather than publication bias (Supplementary Appendix VIIb).

Figure 3. Pooling the INB of RZV vs. NoV for societal perspective by age groups.

RZV vs. ZVL

Two studies^54,61 with six comparisons were included in pooling INBs of RZV vs. ZVL with SP (Figure 4). The overall pooling indicated that RZV was cost-effective compared to ZVL with the pooled INB (95% CI) of $266.67 (143.38, 389.97) (Table 2). Egger’s test showed significant publication bias (coefficient = 3.50, SE = 1.662, p = .0352) but contour-enhanced funnel plot showed studies in both significant and non-significant areas which suggested that asymmetry was likely due to heterogeneity rather than publication bias (Supplementary Appendix VIIc).

Figure 4. Pooling the INB of RZV vs. ZVL for societal perspective by age groups.

Pooled INBs based on third-party perspective (TPP)

ZVL vs. NoV

Fourteen studies with 19 comparisons of ZVL vs. NoV were included with TPP (Table 2). Overall pooling indicated that ZVL was cost-effective relative to NoV with the pooled INB (95% CI) of $8.21 (4.10, 12.32) (Supplementary Appendix VIa). Subgroup analysis by age groups indicated that ZVL was cost-effective in 50+ to 70+ and 70–79 years with INBs of $8.63 (3.62, 13.63) and $7.57 (0.27, 14.86), respectively. Egger’s test resulted in non-significant publication bias (coefficient = 0.19, SE = 0.264, p = .4754) (Supplementary Appendix VIId).

RZV vs. NoV

Four comparisons from four studies were included in pooling overall INBs showing RZV as cost-effective with the pooled INB (95% CI) of $133.25 (14.26, 252.25) (Supplementary Appendix VIc). Egger’s test resulted in non-significant publication bias (coefficient = −1.50, SE = 1.359, p = .2708) (Supplementary Appendix VIIe).

Sources of heterogeneity

Sources of heterogeneity were explored considering the cost of vaccines and comparators according to the comparisons ZVL vs. NoV, RZV vs. NoV, and RZV vs. ZVL (Supplementary Appendix VIII). None of them were identified as a source of heterogeneity from meta-regression results.

Sensitivity analyses

Sensitivity analysis showed that cost-effective results of RZV compared to NoV were robust in all sensitivity analyses, especially in the 70–79 age group (Supplementary Appendix IX). Cost-effectiveness results of ZVL differ across sensitivity analyses, especially when restricted to studies with non-missing variances and those with a low risk of bias. In addition, vaccination in 50–59 and 70–79 age groups was no longer cost-effective.

Discussion

In this study, we reviewed 37 published cost-effectiveness studies of HZ vaccine and performed meta-analyses by pooling INB to assess the cost-effectiveness of ZVL and RZV in patients aged 50 years and older. All included studies were conducted in high-income countries. Our findings demonstrated that RZV was found to be cost-effective as seen from positive INBs compared with NoV in both SP and TPP, especially for age 60 years and above. Comparing the two vaccines, RZV was dominant over ZVL in SP. However, the number of individual studies that directly compared two vaccines was limited, and even lacking in TPP. Meanwhile, ZVL was also cost-effective compared with NoV when vaccinated at ages 70–79 years from SP and TPP. ZVL from SP was also cost-effective in 50–59 years.

In our main analysis, we stratified the studies by different perspectives and vaccine types with subgroups of age resulting in more than one comparison and weights from one study. One of the sensitivity analyses performed to assess the robustness of our findings was utilizing the average values of age and INB to weight one study once and to avoid violation of the independence assumption. The results from sensitivity analyses were robust for RZV, but not for ZVL. In addition, we found RZV’s superiority over ZVL with positive INBs for all the age groups analyzed. Our findings for RZV are consistent with recommendations from the AICP as the preferred type of vaccine⁷. However, It is important to note that the age for which we consistently saw greater INB benefit in RZV vaccination, 60 and above, is older than the current ACIP recommendation of 50 years and older. Additional research should be conducted to assess the impact of delaying vaccination, as aging is an important risk factor for HZ infection. As the world population continues to age and an increasing number of individuals are at risk for developing HZ, extending the vaccination age may be more cost-effective in those countries that authorized the use of RZV but confront limited supply.

With FDA approval and ACIP’s recommendation for the preference of RZV over ZVL, the market share of the RZV is predicted to change drastically and is expected to be more expensive. Moreover, RZV is now only available in a few countries including the United States, Canada, Belgium, Germany, Sweden, the Netherlands, etc. because of limited supply⁶⁶. Although it is licensed in the European Union, Japan, New Zealand, Hong Kong, China, and Singapore, it is not yet available, and ZVL continues to be used in these countries and in LMICs. From ZVL comparisons, we have recognized that it is cost-effective compared to NoV when vaccinated at ages 70–79 years. Although we found a positive INB for ZVL for patients aged 50–59 years, the ACIP does not recommend ZVL in this age group⁶⁷. This is partly due to the limited long-term protection provided by ZVL and because of the lower rate of herpes zoster in this age group compared with persons ≥60 years old⁶⁸. Although high and rising healthcare costs of treating HZ documented in several LMICs suggest that it would probably be cost-effective to immunize elders from this region than others¹⁰. Due to the specific economic evaluation data are lacking in these countries; our findings on HZ vaccines at different age groups may provide information to decision-makers for efficient use of available resources for maximizing health benefits in the future and to resolve current uncertainties. In the context of unavailability and the expected high cost of RZV, our findings on ZVL at different age groups may support researchers from LMICs to conduct country-specific CEA to inform evidence-based decision-making.

Our study is the first to perform a synthesis on economic outcomes by quantifying INBs across studies for HZ vaccination. Although vaccine prices and costs differ between countries, we considered country-specific thresholds embedded into INB calculation. Additionally, meta-regression that incorporated both vaccine and comparator costs found neither of the variables to be a significant source of heterogeneity. Meta-analyses of economic evaluation studies do not aim to conclude the cost-effectiveness profiles for every country universally. Instead, we aim to broadly summarize how precise the cost-effectiveness values of HZ vaccines are based on all existing studies answering the same question across different settings. Our study is also strengthened by many included economic evaluation studies compared to previous reviews. Further, we were able to standardize monetary units to 2019 USD and used INB instead of ICER to indicate cost-effectiveness and non-cost-effectiveness for decision-making or policy analysis^69,70.

Our study has some limitations. Our study used INB calculation to quantify the cost-effectiveness of the vaccines. It is assumed that the ICER is calculated from QALYs or life-years gained. Some of the cost-effectiveness studies may use different health outcome measures that do not fit the INB formula we used. However, these studies were not included in our pooled analyses and may not represent all existing evidence related to health economic evaluation studies investigating Herpes Zoster vaccination.

Some studies assessed various schedules of ZVL and RZV, including ZVL with boosters in varying time periods and additional RZV vaccination in those previously vaccinated with ZVL^{51,52,54–56,60,61}. However, the number of these studies was limited, thus prohibiting us from pooling them using meta-analysis. We could not validate the ACIP’s recommendation to vaccinate RZV in individuals who previously received ZVL. Based on the efficacy of clinical trials, the 2017 recommendations suggested that additional vaccination would be beneficial with a more pronounced benefit in those over the age of 70. One of the included CEAs evaluated this revaccination strategy with five years between administering ZVL and RZV⁵⁶. Vaccination of adults aged 60+ with RZV compared to no additional vaccination resulted in an ICER of $58,793 and was cost-saving compared to a ZVL booster. While these results originated from a single study, it added to the evidence behind the ACIP recommendation. More research on the direct comparison of RZV and ZVL and other vaccination schedules may be needed to determine the cost-effectiveness profiles for the elderly population.

Conclusions

Our study showed that RZV might be cost-effective compared to no vaccine as quantified by positive incremental net benefits across different age groups. Further, RZV was cost-effective for 60–79 years from the societal perspective, while ZVL compared to no vaccine may be cost-effective for 70–79 years, but the cost-effectiveness profile differs across sensitivity analyses. Future health care policies should be reviewed to determine whether vaccination at 60 may be more appropriate than the currently recommended 50 or older.

Transparency

Declaration of funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Declaration of financial/other relationships

The authors declare that they have no known conflicts of interest that are directly relevant to the content of this article.

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

Author contributions

S.U. and M.G.R. reviewed literatures, extracted data for the study, synthesis, and quality assessment with input from N.C., A.T., and T.A. S.U. did the meta-analysis. S.U., M.G.R., and S.V. wrote the first drafts of the manuscript and all authors made substantial contribution. All authors contributed to the study design, interpretation of findings, and critical revision of the manuscript.

Acknowledgements

None reported.

Consent for publication

All authors approved the final version of the manuscript for submission.

Data availability statement

All data are available from the corresponding author upon request.

Code availability

The data analysis file is available from the corresponding author upon request.