Health economic evaluation of a vaccine for the prevention of herpes zoster (shingles) and post-herpetic neuralgia in adults in Belgium

Abstract

Objective:

To determine the cost-effectiveness of vaccination against herpes zoster (HZ) and post-herpetic neuralgia (PHN) in individuals aged 60 years and older in Belgium.

Methods:

A Markov model was developed to compare the cost-effectiveness of vaccination with that of a policy of no vaccination. The model estimated the lifetime incidence and consequences of HZ and PHN using inputs derived from Belgian data, literature sources, and expert opinion. Cost-effectiveness was measured by the incremental cost-effectiveness ratio (ICER), expressed as cost per quality-adjusted life-year (QALY) gained.

Results:

Vaccination in individuals aged 60 years and older resulted in ICERs of €6,799 (third party payer perspective), €7,168 (healthcare perspective), and €7,137 (societal perspective). The number needed to vaccinate to prevent one case was 12 for HZ, and 35 or 36 for PHN depending on the definition used. Univariate sensitivity analyses produced ICERs of €4,959–19,052/QALY; duration of vaccine efficacy had the greatest impact on cost-effectiveness. Probabilistic sensitivity analysis showed at least a 94% probability of ICERs remaining below the unofficial €30,000 threshold.

Discussion:

Key strengths of the model are the combination of efficacy data from a pivotal clinical trial with country-specific epidemiological data and complete sensitivity analysis performed. Main limitations are the use of non country-specific PHN proportion and non Belgian disease-specific utilities. Results are comparable with those recently published.

Conclusions:

HZ vaccination in individuals aged 60 years and older would represent a cost-effective strategy in Belgium.

Key words:

• Belgium
• Cost-effectiveness
• Herpes zoster
• Post-herpetic neuralgia
• Vaccination

Introduction

Herpes zoster (HZ; ‘shingles’) is caused by reactivation of the varicella zoster virus (VZV) that has remained latent in dorsal root ganglia after primary infection acquired in childhood (varicella or chickenpox). The lifetime risk of HZ has been estimated to be 25%1, and the risk of having a recurrent HZ has been assessed between 1 and 5%2 for those having had a first HZ episode. The incidence rises rapidly after the age of 50 years, resulting in a lifetime risk of 25–50% in individuals reaching their eighth decade3,4. VZV reactivation and the resulting HZ is attributable to the progressive decrease in the VZV-specific cell-mediated immunity that occurs with aging (‘immunosenescence’) or other conditions or therapies that cause immune impairment5.

HZ6 is characterised by a painful, unilateral, vesicular rash localised to the dermatome corresponding to the affected ganglion. Rash eruption may be preceded by itching, numbness or pain during a prodromal phase. The pain and rash usually resolve spontaneously within 3–4 weeks7, but the former may persist for months or years after the rash has healed in 20–25% of cases8. This persistent pain is referred to as ‘post-herpetic neuralgia’ (PHN), and is the commonest complication of HZ, occurring in 25–50% of individuals over the age of 50 years8. PHN is usually defined as significant pain that occurs or persists three months after rash onset9, although it may also be defined as pain that occurs or persists one month after rash onset10. PHN can cause severe physical, functional, social and psychological impairment as a consequence of the continuing pain11–16. As a result, this condition is associated with a marked reduction in health-related quality of life17,18.

A new live attenuated vaccine, Zostavax, has been developed for the prevention of HZ and PHN in older adults. The efficacy of Zostavax has been demonstrated in the Shingles Prevention Study (SPS)19, a randomized, double-blind, placebo-controlled trial involving 38,546 immunocompetent participants aged 60 years and older. In this study, the incidence of HZ and PHN was reduced by 51.3% and 66.5%, respectively, and the burden of illness due to HZ (defined as the average severity of illness, which was measured as the area under the curve of HZ pain scores divided by time), reduced by 61.1%. Moreover, vaccinated patients who developed HZ had less severe illness than unvaccinated patients with HZ. The vaccine was approved by the Food and Drug Administration (FDA) in the USA in 2006 for use in individuals aged 60 years or older, and was recommended for use in this age group later the same year20,21. The European Medicines Agency (EMEA) approved the vaccine in 2007 for individuals aged 50 years and older and it was recommended for use in this age group in Austria in 200822.

Ideally, vaccination programmes that aim to promote healthy ageing (including the prevention of HZ and PHN) should begin in middle age, before the age-related decline in cell-mediated immunity has begun. HZ vaccination at the age of 50 years may offer benefits from a public health and health economic perspective because productivity losses resulting from HZ in individuals of working age would be diminished. The incidence and severity of HZ increase with age, so vaccination later in life (e.g., 60 years) may be more cost-effective than earlier vaccination. Cost-effectiveness data are required for a medicinal product to be approved for reimbursement in Belgium23. A health economic model was therefore developed to evaluate the cost-effectiveness of HZ vaccination in Belgium.

Methods

A Markov model was developed to evaluate the cost-effectiveness of the HZ vaccine in Belgium. Two policies were compared in the model: adoption of an HZ vaccination policy with coverage of 20% among individuals aged 50 years and older in the Belgian population, and the current policy of no vaccination. The model was developed in accordance with guidelines for healthcare evaluations issued by the Belgian Health Care Knowledge Centre (KCE)23, and with published guidelines for modelling studies24.

Perspectives

Three perspectives were considered: a third-party payer (TPP) perspective (includes only direct healthcare costs covered by the Belgian National Institute for Health and Disability Insurance; RIZIV-INAMI); a healthcare perspective (includes patient co-payments in addition to the costs considered in the TPP perspective); and a societal perspective (includes productivity losses in addition to the costs included in the healthcare perspective).

Model design

The Markov model was used to simulate the lifetime incidence and consequences of HZ and PHN in the current Belgian population aged 50 years and older. It was developed in Microsoft Office Excel 2003 (Microsoft, Seattle, WA, USA). Several health states were described in the model (Figure 1). These included healthy (i.e., no HZ or PHN), dead, HZ and PHN. HZ was subdivided into four levels of pain severity: no pain, mild, moderate or severe and PHN was subdivided into three levels of pain severity: mild, moderate or severe. Single episodes of recurrent HZ and associated neuropathic pain were also allowed. The population was analysed over the lifetime of 5-year age groups (50–54 years, 55–59 years, etc.) The lifetime of the cohort was divided into 1-month Markov cycles. Within each cycle, cohort members could remain in their current health state, or move into one of the permitted states (Figure 1); these transitions were governed by a matrix of probability values. During each successive cycle, an increasing proportion of the cohort would move through the HZ and PHN states and eventually to death. These cycles were repeated until the entire cohort had died.

Figure 1.  Health states used in the Markov model, and permitted transitions between states.

Movement from the healthy state to HZ was based on monthly age-specific HZ incidence data. The Markov model applied a monthly cycle length and the 1-month definition of PHN, i.e. pain occurring or persisting 1 month after HZ onset was adopted in the model. PHN is more commonly defined as pain occurring within 3 months of HZ onset9, so the model was adjusted to produce results according to both definitions. In particular, the model was calibrated to ensure that the calculated proportion of PHN cases occurring 3 months after the onset of the HZ rash matched that observed in the SPS19.

Because the likelihood of developing PHN is higher in patients with severe zoster-related pain than in those with milder pain, the proportion of HZ patients developing PHN was derived by applying odds ratios of 2.39 for severe pain and 0.88 for mild pain (both relative to an odds ratio of 1.0 for moderate pain)25. The model assumed that PHN diminishes over time and eventually ceases, and hence all patients must pass through the mild PHN state to reach the healthy (post-zoster) state. As a result, the mean duration of PHN will depend on the initial split of the population (i.e., proportions of patients with mild, moderate or severe pain): patients with severe pain will take longer to return to health than patients with mild pain. It was therefore essential to calibrate the transition probabilities between PHN severity states and the healthy state to reflect the mean duration of PHN from the literature. The model assumed that the transition rates from severe to moderate, moderate to mild, and mild to healthy (post-zoster) are identical, although this constant rate differs between age groups: older patients experience a longer duration of pain than younger patients19.

Recurrent HZ was considered to be a distinct state in the model because it is uncommon. The characteristics of recurrent HZ, such as duration, initial pain split (severity), and occurrence of subsequent PHN, were assumed to be identical to those of first episodes. The monthly probability of an individual reaching the death state was derived from age-specific mortality data.

Model validation

The model was validated using clinical data from the SPS19. For the observed number of cases of HZ and PHN over 3 years (duration of the SPS), the figures from the placebo arm of the SPS (n = 642 and n = 80, respectively were used, corresponding to incidence rates of 11.1 and 1.4 per 1000 person-years, respectively). Among vaccinated individuals, the number of HZ cases was the same as in the vaccine arm of the SPS (n = 315, corresponding to incidence rates of 5.4 per 1000 person-years), but the number of PHN cases was 15% lower (23 versus 27). Vaccine efficacy was therefore adjusted downwards according to the PHN pain state transition probabilities. Vaccination reduced the mean duration of PHN in the SPS by 2.2 months in patients aged 60–69 years and 3.3 months in older patients19, so the transition probability for moving to the next lowest pain state (severe-to-moderate; moderate-to-mild; mild-to-no pain) was calculated by modelling an initial cohort with the SPS pain split. A sufficient number of cycles were simulated until >99% of patients were no longer in pain, and the mean duration of pain calculated for this population. This constant probability was adjusted until the mean duration of pain was equal to the PHN duration observed in the SPS, to yield the PHN pain state transition probability. Among vaccinated individuals, the calculated transition probabilities were 23.0% in those aged 60–69 years, and 22.0% in older individuals; the corresponding figures in unvaccinated persons were 18.2% and 16.2%, respectively. Following this adjustment, the model results exactly matched those reported in the SPS, which suggests that the model provides a valid estimate of clinical benefit.

Model inputs

Data used in the model were derived from the SPS19 and other published sources, in addition to expert opinion (Tables 1–3). Expert opinion was obtained from a Delphi panel comprising four primary care physicians, three dermatologists, four ophthalmologists, and eight pain specialists or neurologists. The panel was conducted in two rounds using written questionnaires26.

Table 1.  Model inputs: epidemiology, vaccine characteristics and utilities.

Model input	Base case analysis			Deterministic sensitivity analysis
	HZ	PHN		HZ	PHN
Epidemiology
HZ incidence (per 1,000 people) and PHN proportion per HZ case (%)	Belgian Health Authorities27,28	GPRD (1 month)29*	GPRD (3 months)	Opstelten et al.7
Age 50–54 years	4.0	10.3 (14.1)	7.6	3.6	3.9 (5.3)**
Age 55–59 years	3.5	13.7 (18.8)	9.8	5.8	6.5 (8.9)
Age 60–64 years	5.6	15.7 (21.5)	11.0	5.8	6.5 (8.9)
Age 65–69 years	5.8	18.7 (26.5)	13.1	6.5	10.7 (14.7)
Age 70–74 years	7.1	22.5 (30.4)	15.2	6.5	10.7 (14.5)
Age 75–79 years	10.2	26.6 (36.0)	18.0	9.1	18.0 (24.3)
Age 80–84 years	11.6	28.9 (39.0)	20.6	9.1	18.0 (24.3)
Age 85–89 years	13.9	25.9 (35.0)†	18.5	9.1	18.0 (24.3)
Age 90–94 years	11.2			9.1	18.0 (24.3)
Age ≥ 95 years	18.2			9.1	18.0 (24.3)
Mean duration (months)	SPS data19
Age ≤ 69 years	1.0	8.3
Age ≥ 70 years	1.0	10.9
Annual HZ mortality (%)	GPRD29 and expert opinion			Edmunds et al.40
Age 50–54 years	0	0		0.0009	–
Age 55–59 years	0	0		0.0009	–
Age 60–64 years	0	0		0.0027	–
Age 65–69 years	0	0		0.0027	–
Age 70–74 years	0	0		0.0035	–
Age 75–79 years	0	0		0.0092	–
Age 80–84 years	0	0		0.0487	–
Age 85–89 years	0	0		0.2018	–
Age 90–94 years	0	0		0.2018	–
Age ≥ 95 years	0	0		0.2018	–
Pain split (%)	SPS data19			GPRD29
Age ≤ 69 years
No pain	27	–		65
Mild pain	41	42		24	48
Moderate pain	18	9		4	42
Severe pain	14	49		8	10
Age ≥ 70 years
No pain	26	–		45
Mild pain	32	17		41	37
Moderate pain	23	16		5	54
Severe pain	19	67		9	9
Incidence of recurrent HZ (%)	Cunningham et al.2
Age 50–54 years	0.09	–‡
Age 55–59 years	0.10	–
Age 60–64 years	0.12	–
Age 65–69 years	0.14	–
Age 70–74 years	0.18	–
Age 75–79 years	0.23	–
Age 80–84 years	0.30	–
Age 85–89 years	0.42	–
Age 90–94 years	0.60	–
Age 95–99 years	0.87	–
Age 100 years	1.1	–
Vaccine characteristics
Efficacy: reduction in number of cases (%)	SPS data19
Age 50–69 years	63.9	65.7
Age ≥ 70 years	37.6	66.8
Efficacy: reduction in PHN duration (months)
Age 60–69 years	–	−2.2 (26.5)
Age ≥ 70 years	–	−3.3 (30.3)
Vaccine cost, including copayments (€)	141.18			100.38–160.38
Quality of life
Age-specific utilities (males/females)	EuroQoL group34,35
Age 50–54 years	0.798/0.801	0.798/0.801
Age 55–59 years	0.798/0.801	0.798/0.801
Age 60–64 years	0.773/0.800	0.773/0.800
Age 65–69 years	0.773/0.800	0.773/0.800
Age 70–74 years	0.714/0.733	0.714/0.733
Age 75–79 years	0.714/0.733	0.714/0.733
Age ≥ 80 years	0.730/0.686	0.730/0.686
Utility decrements	Oster et al.18			SPS data19§	Bala et al.42§
No pain	0.00	–		0.14
Mild pain	0.31	0.31		0.23	0.27
Moderate pain	0.42	0.42		0.32	0.40
Severe pain	0.75	0.75		0.45	0.53

GPRD, General Practice Research Database.

*PHN proportion is shown as the incidences derived from Gauthier et al.25 using the 1-month definition of PHN, and values adjusted by applying the pain split observed in the SPS to ensure that the GPRD data reflect all HZ cases with pain.

**Data from the study by Opstelten et al.6, adjusted in the same way.

†Age ≥ 85 years.

‡It was assumed that the proportion of patients with PHN after recurrent HZ would be the same as that following a first episode.

§Same increments apply to both HZ and PHN.

Table 2.  Model inputs: healthcare resource use (2007 prices). Monthly costs.

Perspective	No pain		Mild pain			Moderate pain		Severe pain
	TPP	Healthcare/ societal	TPP		Healthcare/ societal	TPP	Healthcare/ societal	TPP	Healthcare/ societal
HZ ambulatory management costs (€)
Primary and secondary ambulatory care	22.72	28.71	25.17		32.20	45.45	61.09	78.93	59.72
Medications	31.99	43.74	39.70		57.54	66.11	98.72	79.93	129.26
Non-pharmacological   treatments	0.00	0.00	5.76		6.40	36.55	37.59	89.34	90.99
Diagnostic tests	2.63	2.68	2.57		2.65	6.39	6.73	19.55	21.04
Total HZ costs	57.34	75.13	73.20		98.79	154.49	204.12	265.75	301.01
HZ hospitalization costs
Hospitalisation rate (%)	0.11		0.22			8.33		15.56
Mean length of hospital stay (days)	3		3			4.67		6.29
Mean unit cost of hospitalisation (societal perspective (€)	317.24
Mean HZ total cost (€)	0.97	1.06	1.97	2.11		113.99	123.37	288.66	310.19
PHN* ambulatory management costs (€)
Primary and secondary ambulatory care			52.22	67.87		53.29	120.15	68.40	94.18
Medications			6.93	22.63		27.32	61.67	56.20	130.28
Non-pharmacological   treatments			10.64	11.17		43.22	43.97	120.99	124.10
Diagnostic tests			8.14	8.55		12.17	12.87	25.37	27.34
Total PHN costs			77.93	110.22		136.00	238.66	270.96	375.90
PHN hospitalisation costs
Hospitalisation rate (%)			0.42			2.33		11.75
Mean length of hospital stay (days)			4.5			6.75		6.44
Mean unit cost of hospitalisation (societal perspective (€)			317.24
Mean total PHN costs (€)			5.49	5.95		46.57	49.96	227.90	240.22

TPP, third party payer.

*1-month definition.

Table 3.  Costs of lost productivity due to HZ and PHN.

	Pain level	Days lost from work	Proportion of patients (%)	Cost of lost productivity by age group (€)
				50–54 years	55–59 years	60–64 years	65–69 years
HZ	None	6	4.38	36.77	22.56	7.45	2.67
	Mild	6.33	12.50	110.90	68.05	22.48	8.04
	Moderate	6.33	39.38	496.42	304.61	100.62	35.99
	Severe	11.07	78.75	1,221.35	749.43	247.56	88.55
PHN	Mild	6.17	9.17	8.67	5.28	1.72	0.61
	Moderate	11.4	33.33	58.29	35.52	11.58	4.13
	Severe	19.18	52.58	154.71	94.28	30.73	10.95

Epidemiology of HZ and PHN

For the base-case analysis, data on the incidence of HZ in Belgium27 were derived from a 2006 survey conducted by the Belgian Sentinel of General Practitioners28. This is an epidemiological surveillance system that involves approximately 180 physicians selected to be representative of Belgian general practitioners in terms of age, sex and geographical location. The survey covered 183,895 individuals; the reported age-specific HZ incidence rates ranged from 0.35% to 1.82% (Table 1). This survey did not provide data on the proportion of patients with HZ developing PHN, and hence data were derived from the General Practice Research Database (GPRD) in the UK29. This yielded a PHN proportion between 10.3% and 28.9% using the 1-month definition of PHN; these numbers were adjusted by applying the pain split observed in the SPS19 to ensure that GPRD data reflected only HZ cases with pain (i.e., the dominator used to calculate the conditional probability of PHN was the number of patients with painful HZ). As described above, the proportion of PHN cases obtained using the 1-month definition was calibrated to ensure that the proportion of patients calculated in the model matched that observed in the SPS (in which the 3-month definition was used). The mean durations of HZ and PHN were derived from the SPS19. Data on mortality associated with HZ or PHN for Belgium were not available. Disease-specific mortality was therefore set at zero, based on data from the GPRD29 and supported by expert opinion. The pain split used in the model, both for HZ and PHN, was that observed in the SPS19. The incidence of recurrent HZ was calculated from Belgian life expectancy data for each 5-year age group30, applying a conditional lifetime risk of 3%, which is the midpoint of the 1–5% range reported by Cunningham et al.2. It was assumed that the proportion of patients with PHN after recurrent HZ would be the same as that after a first episode.

Vaccine characteristics

Data on vaccine efficacy were derived from the SPS19. This trial included individuals aged 60 years and older. Two clinical trials showing that the immunological response to the vaccine in the younger 50–59 years age group was not inferior to that observed in older persons suggest that efficacy data may be extrapolated to the entire population aged 50 years and older31,32. Both direct and indirect effects of vaccination were included in the model. It was assumed that vaccination would have a direct effect in reducing the number of cases of HZ and PHN. As HZ precedes PHN, the number of PHN cases was assumed to be reduced indirectly through the reduction in the HZ cases; hence the indirect effects of vaccination were included implicitly as additional consequences of the direct effects.

It was assumed that vaccine efficacy would have a lifelong duration. Current available data do not show any decline of vaccine efficacy over time33. After having reviewed those data, EMEA confirmed that vaccine efficacy was maintained for up to 7 years and the labelling of Zostavax, one-dose vaccine, has not been changed. To anticipate a possible future decline of efficacy over time, a waning function was built into the model. This function was set at zero for the base-case analysis; alternative efficacy durations were examined in sensitivity analyses (see below). In the absence of literature to validate the choice of vaccine coverage rate the authors assumed a hypothetical rate of 20% in all age groups, but they also included other coverage rates in the sensitivity analyses. The unit cost of vaccination was set at €141.18 in the base-case analysis, including co-payments. Different vaccination costs were included in sensitivity analyses.

Utilities

The base-case analysis used applied an additive method to calculate utilities in which disease-specific decrements were subtracted from age-specific utilities. Age-specific Belgian utility data were derived from a report of the EuroQoL Group34 which used data from a 2004 survey of 2,754 individuals in the Flemish-speaking Belgian population (Table 1)35. In the absence of data specific to Belgium, utility decrements associated with HZ and PHN were derived from a study by Oster et al. in the USA18. The applicability of these data to the Belgian population is supported by their similarity to reported data from patients with neuropathic pain in five European countries36.

Use and cost of healthcare resources

Data on the use and cost of healthcare resources associated with HZ and PHN (Table 2) were obtained from a burden of illness study conducted by IMS Health (Brussels, Belgium)26. This used the previously mentioned expert panel comprising four primary care physicians, three dermatologists, four ophthalmologists, and eight pain specialists. This study examined the use of primary and secondary care, oral and topical medications, non-pharmacological treatments (transcutaneous electrical nerve stimulation, acupuncture, infiltrations, hypnosis), diagnostic tests, and hospitalisations from societal and TPP perspectives. HZ was classified according to the localisation and number of dermatomes affected. For analyses of medications and non-pharmacological treatments, a distinction was made between HZ diagnosed within 72 hours of rash onset and HZ diagnosed later. This generated two sets of data; because >50% of patients contact their general practitioners within 72 hours of rash onset7, mean resource use and cost values were derived by applying a 50% weighting to the two datasets.

Costs of lost productivity

Productivity costs (Table 3) were applied in the societal perspective and were calculated by combining data on work absenteeism (obtained from the IMS report) with national Belgian employment statistics37, with adjustment for the higher proportion of women affected by HZ and PHN compared with men (Table 3). It was assumed that nobody was employed after the age of 70 years, based on reported employment rates for individuals aged 65−69 years (the official retirement age in Belgium is 65 years). Costs resulting from absence from work were calculated by multiplying the number of workdays lost by the proportion of patients who missed work because of HZ or PHN in each pain category, and multiplying the result by the mean daily cost of absenteeism in Belgium in 2004 (€200)38.

Model outputs

The model outputs were total costs, quality-adjusted life-years (QALYs), number of cases of HZ and PHN under vaccination and no-vaccination policies, and the number needed to vaccinate (NNV) to prevent one case of HZ or PHN. The cost-effectiveness of vaccination was assessed by comparing the total costs and effectiveness in terms of the number of cases of HZ or PHN avoided or QALYs gained under both policies. The incremental cost-effectiveness ratio (ICER) was calculated by dividing the difference in cost between the two policies by the difference in effectiveness. In Belgium there is no official threshold below which an intervention would be considered cost-effective, although ICERs below €30,000/QALY have previously been considered to be cost-effective39. This threshold was therefore used as a criterion for cost-effectiveness in the present study.

Discount rates

Based on the KCE recommendations23, discount rates of 3% were applied for costs and 1.5% for outcomes (QALYs). Other discount rates were investigated in sensitivity analyses (see below).

Sensitivity analyses

Deterministic sensitivity analyses were carried out to assess the sensitivity of the ICER obtained in the base-case analysis to variations of several key input variables within feasible ranges. These analyses were univariate, i.e. only one variable was changed in each analysis.

The impact of a reduction in the incidence of HZ on the cost-effectiveness of vaccination was investigated by applying data from an epidemiological study carried out in the Netherlands7, in which the incidence of HZ was lower than that reported in Belgium (Table 1). In addition, the effect of varying durations of PHN and changes in the level of pain used in the base-case analysis were examined using GPRD data29. A further analysis examined the impact of potential HZ-related mortality because, although it is widely considered that there is no mortality risk associated with HZ, an epidemiological study in the UK reported evidence of a low risk40. Data from this study were therefore included in the sensitivity analysis (Table 1).

As noted above, a waning function was incorporated in the model to account for uncertainty around vaccine efficacy decline over the time. In the sensitivity analysis, this function was set at 8.3% per year, which has been reported to be the upper limit of the waning rate from the first month after vaccination41. The authors also conducted analyses assuming that the efficacy of the vaccine was limited to 10 years with an additional vaccine dose, or 20 years with no additional dose. The former analysis assumed that a repeat dose of the vaccine would be given after 10 years to 50% of patients who had received the vaccine and who were healthy and without HZ. This enabled the impact of coverage rates of 10–75% to be determined. In addition, the sensitivity analysis allowed vaccine prices to be determined (€100.38–160.38).

The effects of differing utility decrements were analysed using HZ-specific data from the SPS19, and a study by Bala et al.42. (Table 1). As the Bala et al. study provided data for only two pain states, mild and severe, the sensitivity analysis used an extrapolated value of 0.6 for moderate pain. A further analysis explored the impact of the methodology adopted for the calculation of utilities. As noted above, the base case applied an additive method whereby disease-specific decrements were subtracted from age-specific utilities. The additive method of QALY calculation was selected on equity grounds as it treats all individuals as equal, irrespective of their age whereas a multiplicative approach values people with higher age-specific utilities more than those with lower age-specific utilities. To evaluate the impact of the choice of methodology, the sensitivity analysis used the multiplicative technique in which the final utility value was obtained by multiplying age-specific utilities with disease-specific utilities.

As noted above, the analysis of the use and cost of healthcare resources used a 50% weighting to distinguish between HZ diagnoses made within and later than 72 hours after rash onset. The uncertainty surrounding the assumption that most patients contact their physician within this time was tested by applying weightings of 25% and 75% in sensitivity analyses. Further analyses explored the impact of 0%, 3% and 5% discount rates for costs and benefits.

Probabilistic sensitivity analyses were carried out using Monte Carlo simulations in Microsoft Excel 2003. The model was run 1000 times using inputs randomly selected from their statistical distributions. In these analyses, vaccine efficacy was tested by varying the mean efficacy value of the base case using a beta distribution, and by applying a waning function ranging between 0% and 8.3% using a uniform distribution. HZ incidence and PHN distribution were also varied over a beta distribution. Resource use mean costs were varied across known maximum and minimum values by means of a triangular distribution, whereas utility decrements were tested by a log-normal distribution.

Results

Base-case results

The base-case analysis examined the cost-effectiveness of vaccination in members of the Belgian population aged 60 years and older from societal, healthcare, and TPP perspectives. Subgroup analyses included individuals aged ≥50 or ≥65 years, and those aged 60–64, 65–69 and 60–69 years. Discount rates of 3% for costs, and 1.5% for utilities, were applied in the base-case analysis.

The model estimated that vaccination against HZ in individuals aged 60 years and older would cost between €51,640,533 (TPP perspective) and €54,439,238 (healthcare perspective), and would prevent 39,480 cases of HZ, and 13,121 (1-month definition) or 13,442 (3-month definition) cases of PHN, resulting in a gain of 7,595 QALYs (assuming 20% coverage and a lifetime duration of efficacy). ICERs showed that vaccination was a cost-effective strategy from all three perspectives (Table 4): the cost per QALY was €6,799 from a TPP perspective, €7,168 from a healthcare perspective, and €7,137 from a societal perspective. The NNV to prevent one case of disease in this population was 12 for HZ, 36 for PHN defined using the 1-month criterion, and 35 for PHN defined using the 3-month criterion.

Table 4.  Base-case analysis: lifetime cost-effectiveness in individuals aged 60 years and older.

	ICER (€)
	TPP	Healthcare	Societal
Cost per QALY	6,799	7,168	7,137
Cost per HZ case avoided	1,308	1,379	1,373
Cost per PHN case avoided (1-month definition)	3,936	4,149	4,131
Cost per PHN case avoided (3-month definition)	3,842	4,050	4,033

Cost-effectiveness results for the subgroup analyses are summarized in Table 5. Vaccination was consistently found to be a cost-effective strategy for the prevention of HZ and PHN, with ICERs below the unofficial but widely accepted threshold of €30,000, and NNVs of 8–14 for HZ and 27–39 for PHN.

Table 5.  Subgroup analyses: lifetime cost-effectiveness by age group.

Subgroup	≥50 years	≥65 years	60–64 years	65–69 years	60–69 years
HZ cases prevented	78,522	26,185	13,296	11,792	25,088
PHN cases prevented
1-month definition	23,303	9421	3700	3476	7176
3-month definition	23,943	9783	3659	3448	7107
QALYs gained	12,396	5687	1908	1906	3814
ICERs
Cost per QALY (€)	6,614*	€7,173*	€5,686*	€5,404*	€5,545*
	€6,966†	€7,575†	€5,953†	€5,647†	€5,800†
	€6,812‡	€7,566‡	€5,858‡	€5,620‡	€5,739‡
Cost per HZ case avoided (€)	€1,044*	€1,558*	€816*	€873*	€843*
	€1,100†	€1,645†	€855†	€913†	€882†
	€1,075‡	€1,643‡	€841‡	€908‡	€873‡
Cost per PHN case avoided (€)
1-month definition	€3,518*	€4,330	€2,932*	€2,963*	€2,947*
	€3,706†	€4,573†	€3,070†	€3,096†	€3,083†
	€3,624‡	€4,567‡	€3,021‡	€3,082‡	€3,050‡
3-month definition	€3,490*	€4,170*	€2,965*	€2,987*	€2,976*
	€3,676†	€4,403†	€3,105†	€3,122†	€3,113†
	€3,595‡	€4,398‡	€3,055‡	€3,107‡	€3,080‡
NNV to avoid one HZ case	9	14	8	8	8
NNV to prevent one PHN case
1-month definition	32	39	27	28	28
3-month definition	32	37	28	28	28

*,†,‡Third party payer, healthcare and societal perspective, respectively.

Sensitivity analyses

The results of the deterministic sensitivity analyses are summarised in the tornado diagrams shown in Figure 2, which present the cost per QALY associated with each intervention.

Figure 2.  Tornado diagrams showing the outcome of deterministic sensitivity analyses from third party payer (a), healthcare (b) and societal (c) perspectives. The vertical line represents the base-case incremental cost-effectiveness ratio (ICER; cost per quality-adjusted life year [QALY]), as shown in Table 4. The bars represent the change in ICER resulting from alterations in the variable shown on the left of the figure. GPRD, General Practice Research Database; HZ, herpes zoster; PHN, post-herpetic neuralgia; SPS, Shingles Prevention Study.

Epidemiology of HZ and PHN

Applying the lower HZ incidence rates and PHN proportion, and the lower proportion of women affected by HZ and PHN, reported by Opstelten et al.7 resulted in an increase in the ICERs to €12,897, €13,721 and €13,669 from a TPP, healthcare and societal perspective, respectively. NNVs were also higher than in the base-case analysis (14 for HZ, and 69 and 67 for PHN defined using the 1-month and 3-month criteria, respectively). The use of HZ duration and pain split data from the GPRD had a marked impact on cost-effectiveness, mainly due to a reduction in the number of patients with moderate or severe pain. The resulting ICERs were €14,418 from a TPP perspective, €15,382 from a healthcare perspective, and €15,352 from a societal perspective.

Vaccine characteristics

The duration of vaccine efficacy had the greatest impact on cost-effectiveness results (Figure 2). The use of a 20-year duration increased ICERs to €9,184 from a TPP perspective, €9,739 from a healthcare perspective, and €9,699 from a societal perspective. Use of the lower 95% limit of the confidence interval of vaccine efficacy reported in the SPS19 increased ICERs to €7,741 (TPP perspective), €8,200 (healthcare perspective) and €8,170 (societal perspective). Conversely, use of the upper 95% limit decreased ICERs to €6,129, €6,433 and €6,401 from a TPP, healthcare and societal perspective, respectively. Inclusion of a repeat dose after 10 years, at an identical unit cost as the initial vaccine, resulted in lifetime ICERs of €13,511, €14,426 and €14,382 from a TPP, healthcare and societal perspective, respectively. Vaccine coverage rate had no effect on cost-effectiveness because outcomes and costs change proportionally at each coverage rate. Analysis of the impact of vaccine price on cost-effectiveness showed that each €10 increase in vaccine price would increase the cost per QALY by €613 from all three perspectives.

Utilities

The use of alternative utility values, derived from the study by Bala et al.42, had minimal impact on the ICERs from all three perspectives (Figure 2). Use of a multiplicative (rather than additive) method to calculate utilities resulted in a reduction in the number of QALYs gained from 7,595 to 5,416 (data not shown). As a result, ICERs increased to €9,535 (TPP perspective), €10,052 (healthcare perspective) and €10,009 (societal perspective).

Use of healthcare resources

Weighting resource use by 25% or 75% to allow for different proportions of diagnoses made within 72 hours of rash onset had little effect on ICERs (Figure 2).

Discount rates

Applying a discount rate of 5% resulted in increased ICERs (Figure 2), reflecting vaccination costs incurred during the first year that are not affected by discounting.

Probabilistic sensitivity analysis using Monte Carlo simulation methods showed that the cost-effectiveness of vaccination was maintained under various conditions. The mean (±SE) costs per QALY from TPP, healthcare and societal perspectives were €8,057 ± 324, €8,624 ± 347 and €8,853 ± 408, respectively. The higher ICERs obtained in these analyses, compared with the base-case analysis, are largely attributable to the inclusion of variations in vaccine efficacy rates and the HZ or PHN resource utilisation costs. As the maximum value for these variables was skewed far from the mean, the probability of a higher monthly cost of HZ or PHN was more likely. Thus the benefits of a vaccination that prevented these high costs associated with the disease were more pronounced. Scatter plots of the Monte Carlo simulations on the cost-effectiveness plane are shown in Figure 3. A cost-effectiveness acceptability curve derived from these analyses (Figure 4) shows that the probability of not exceeding the unofficial Belgian cost-effectiveness threshold of €30,000 is 94.5% from a TPP perspective, 94.3% from a healthcare perspective, and 94.0% from a societal perspective.

Figure 3.  Scatter plots showing the outcome of probabilistic sensitivity analyses using Monte Carlo simulations. Results are presented as costs per quality-adjusted life year (QALY) from third party payer (a), healthcare (b) and societal (c) perspectives.

Figure 4.  Cost-effectiveness acceptability curve. The dotted line indicates the unofficial Belgian threshold for cost-effectiveness of €30,000.

Discussion

The results of this study indicate that vaccinating members of the Belgian population aged 60 years and older against HZ is likely to yield substantial benefits in terms of the number of cases of HZ and PHN prevented, and that this strategy is likely to be cost-effective. The costs per QALY gained were considerably lower than the threshold of €30,000 which is widely accepted to indicate cost-effectiveness39, and the model was robust to changes in key epidemiological and vaccine-related variables. According to this model, it would be necessary to vaccinate only 12 people to prevent one case of HZ, whereas vaccination of 36 or 35 people would prevent one case of PHN defined according to 1-month or 3-month criteria, respectively.

The model has several strengths that support its validity as a measure of the clinical and economic benefits of HZ vaccination. It combines efficacy data from a pivotal clinical trial19 with country-specific epidemiological data, and takes into account the ageing of the population over the duration of the model, allowing for mortality, likelihood of developing HZ or PHN, and vaccine efficacy. It also incorporates a high level of detail, taking into account the duration of HZ and PHN, and differing levels of pain severity for both conditions. This allowed HZ, PHN, and their treatment to be modelled accurately. The direct effect of vaccination on the incidence of HZ and PHN, and the indirect effect whereby the frequency of PHN decreases as a result of the reduced number of HZ cases, are included in the model.

A further strength of the model is that it was subjected to comprehensive and robust univariate sensitivity analysis in which ICERs consistently remained below the €30,000 threshold for cost-effectiveness. It is noteworthy that cost-effectiveness was maintained when utilities were calculated by either additive or multiplicative techniques. The additive method of calculating QALYs was chosen for the base-case analysis because it treats all members of the population as equal, irrespective of their age. The multiplicative approach would value people with higher age-specific utilities more highly than those with lower age-specific utilities (i.e., the elderly). The sensitivity analysis showed that vaccination remained cost-effective from the healthcare and societal perspectives if the multiplicative approach was used. The validity of these findings is further supported by the probabilistic sensitivity analysis, which showed that there was at least a 94% probability of the ICER remaining below the €30,000 threshold.

One limitation of this study was that, due to the lack of country-specific data for Belgium, the proportion of patients with HZ who developed PHN was derived from UK data29. This proportion is a major driver of the model results, and having country-specific data would clearly have been an advantage.

Another limitation was that the model did not take into account the effect of the vaccine in reducing pain in individuals who develop HZ despite vaccination. In the SPS, such patients reported a 22% reduction in HZ pain19, which may translate into a subsequent reduction in PHN. However, vaccine effects on PHN were included indirectly in the model by shortening PHN duration. A related issue is that the model used the pain split reported in the SPS19 rather than that used in the GPRD29, and sensitivity analysis showed that this had a marked impact on the observed ICERs. Nevertheless, pain assessments in the SPS are probably more reliable than those in the GPRD because the former measured pain by means of the Zoster Brief Pain Inventory (a specific tool for assessing zoster-related pain)43, whereas the latter assessed pain retrospectively based on the level of analgesic drugs prescribed29.

A further limitation of the study is that resource use and cost data were derived from a burden of disease study26 that was based on expert opinion, rather than population data. As a result, the predictive strength of the model and the generalisability of its results could be limited.

The utility weights for severe pain in the studies by Oster et al.18 and Bala et al.42 differ markedly (0.25 vs. 0.45, respectively). The weights used by Oster et al. were preferred for the base-case analysis because these included a specific weight for moderate pain, whereas those of Bala et al. covered only mild and severe pain. Moreover, Bala et al. assessed HZ pain, which is of shorter duration than PHN, and may therefore have a smaller impact on quality of life. Support for the use of the utility weights of Oster et al. comes from the finding that similar values were reported in a study of neuropathic pain in patients from five European countries36.

The results of the present analysis are comparable with those of a recent evaluation of the cost-effectiveness of HZ vaccination in England and Wales44. This study reported an ICER of £20,400 (€22,843) per QALY gained, which is markedly higher than that obtained in the base-case analysis of the present study. However, there were several important differences between the two models. The model used in the UK study included clinically relevant pain, which incorporated moderate and severe pain, and assumed a limited duration of vaccine efficacy by incorporating a waning rate. The two models also differed in input variables such as HZ incidence rates, disease-specific utilities, and vaccine price. Direct comparison of the base-case results of the two analyses is not possible, but sensitivity analysis applying the maximum vaccine protection duration (100 years) for the cohorts aged 60, 65, 70, and 75 years in the UK study resulted in ICERs ranging between £5,660 and £9,891 (€6,017–10,515†) per QALY gained, which are of a similar order of magnitude to those obtained in the present study. In the Recommendations of the Advisory Committee on Immunization Practices for the Prevention of Herpes Zoster published in 2008 by Harpaz et al.6, the cost-effectiveness of such a program has been discussed. The estimated cost per QALY ratios for a zoster vaccination program for people aged 60 and over was presented acceptable in comparison to established interventions or standard thresholds, ranging from US$27,000 to 112,000/QALY, under a societal perspective and at a $150 per dose. Two main limitations in this comparison are to be addressed such as the fact that models and most of data used to set-up this economic review were non-European and that vaccine efficacy against PHN was not always taken into account in the models, making any robust conclusion difficult.

Conclusion

Results from the present study suggest that vaccination against HZ and PHN in the Belgian population aged 60 years and older is cost-effective. This is an important finding because as the general population ages the number of cases of HZ and PHN may be expected to rise45. HZ vaccination would be expected to offer important public health benefits in such an ageing population by preventing the deterioration in quality of life due to HZ and PHN in otherwise healthy individuals. Further clinical data are needed to determine the duration of vaccine efficacy and the potential impact of this on the cost-effectiveness of vaccination. Currently available data support the hypothesis that the duration of efficacy is lifelong. Indeed, sustainable vaccine efficacy has been demonstrated for at least 7 years with a single dose33; this length of follow-up is one of the longest reported for a vaccine at European launch and an efficacy follow-up study remains ongoing. The deterministic sensitivity analysis in the present study provided an initial estimation of long-term cost-effectiveness by assessing the impact of a shorter duration of efficacy (10 years with a repeat dose, or 20 years). A budget impact analysis of a vaccination for the prevention of HZ and PHN in older adults will be required.

Transparency

Declaration of funding

This study was funded by Sanofi Pasteur MSD, France.

Declaration of financial/other relationships

L.A. has disclosed that he received grants from Sanofi Pasteur to conduct this study. X.B. and C.G. have disclosed that they are employees of Sanofi Pasteur. M.P. has disclosed that he is an employee of i3 Innovus, a company who was received funding to conduct this study.

Acknowledgements

The authors gratefully acknowledge the contribution of M. Lamotte and K. Caekelbergh from IMS Health, Belgium, and Communigen Limited, Oxford, UK for their assistance in preparing the manuscript.

Notes

*Zostavax is a registered trademark of Sanofi Pasteur MSD, Lyon, France.

†Calculated using the exchange rate in force on 15 November 2009 (£1 = €1.11974).