We developed a static Markov-model and conducted various SA to assess the cost-effectiveness of routine HZ-vaccination in Germany and to identify the most cost-effective strategy when targeting specific age-groups for vaccination. HZ-vaccination of 20% individuals of a cohort of 1 million 50 year old persons at age 60 would avoid 20,000 HZ-cases and reduce the overall HZ-treatment costs by 10 million Euros. However, with a BCR of 0.34 for the base-case, our results show that vaccination against HZ and PHN is not a cost-saving measure. Univariate analyses revealed that the price had the greatest impact on ICERs in the DSA. In the break-even analysis with all other variable set at base-case scenario, a price per vaccine dose of <26.50 Euros caused cost-savings.
Varying several variables in favor for the vaccine (including a 50% reduction in the price per vaccine dose to 70 Euros) the intervention becomes almost cost-saving. However, under least favorable conditions (low estimates for HZ/PHN-related incidence and treatment costs, low HZ-recurrence rate, high utilities, low VE, a high waning immunity rate, high vaccine administration costs, and a high vaccine dose price) the model accounts very high ICERs. In general, health economic evaluations of new vaccines are often impacted by two important input factors: incidence of the target disease and VE. However, since the incidence data are rather reliable and range within narrow CIs, the wide ranges of VE (especially for the protection against PHN), which are based on trial data, mainly force the differences of these two extreme scenarios. Furthermore, the variation of the vaccine price (-50% vs. +50%) and the variation of disease recurrence seem to influence the ICERs, too. However, these two extreme scenarios of varying several input parameter either in favor for the vaccine or under least favorable conditions, can be seen as unlikely outer boundaries of the whole SA’s scope.
Due to decreasing VE by age, an increasing age at vaccination usually leads to higher ICERs indicating a lower cost-effectiveness. However, before the age of 60 the ICERs are not necessarily more favorable, due to lower HZ and PHN-related incidence and assumed waning immunity. Hence, targeting persons aged 60 years is likely the most cost-effective vaccination strategy. The variation in the annual waning rate of vaccine-induced immunity had rather limited impact on the ICERs. Altering the period of stable VE following vaccination from 0 to 20 years produces a relative wide array of ICERs. The combined analysis of varying waning immunity rates and durations of stable VE has a rather high impact on ICERs by age at vaccination. However, this analysis confirms that age 60 seems to be the optimal age at vaccination if the annual waning rate is ≥5%. The booster scenarios showed an increasing similarity between best and worst booster scenario with increasing age at vaccination. This confirms the assumption that independently from how a fictitious booster scenario is designed age of vaccination is one of the most relevant factors. The most cost-effective age of vaccination changes from 50 to 60, when moving from best to worst booster scenario. A variation of PHN-duration (6–15 month) had a moderate impact in ICERs. Whereas an extreme extension of PHN-duration up to 3 years downsizes the costs per QALY gained enormously. However, since recent studies confirm an average PHN-duration of several month, our base-case assumption of 9 month seems to be justified [29
In terms of internal validity, we compared the epidemiology reported in the literature for Germany (used as model input parameter) and the model results. The model slightly overestimated HZ and accounted less than one percent more HZ-cases per age-group than represented by input-data. Regarding PHN-cases, the model calculated 4% more PHN-cases in age-groups 50 to 60 years. In older age-groups the overestimation was less than 1%. When implementing vaccination into the model, the accounted cases were reduced according to the VE implemented into the model. Thus, the model’s internal validity can be considered as good.
To date there is no other health economic evaluation of the HZ-vaccine from Germany. However, we identified 14 studies from other European countries [17
] and North-America [56
]. We found a wide range in vaccine prices from 43.85 to 438.5 PPP-Euros in one US study [72
] and 81 to 147.32 PPP-Euros in the other studies. From SP for vaccination of individuals of 60 years results ranged from cost-savings [72
] to 130,097 PPP-Euros per QALY gained [73
]. From PP ICERs ranged from 6,809 [34
], 1,200–46,968 [54
] to 40,050 PPP-Euros per QALY gained [68
]. Our base-case results range well within these international findings. However country-specific issues like vaccine-price, disease epidemiology, price levels, and treatment pathways as well as model-specific issues like model structure, specific assumptions, and dealing with uncertainty hinder a full comparability. Hence, more uniform methods in studies are needed to make results more comparable [76
One limitation of our health economic model is the absence of utility-data considering the impact on health related QoL caused by HZ and PHN specific for Germany. Instead, we used data derived from a Canadian study without country-specific adaptation, which might not necessarily represent the real impact of HZ/PHN on QoL in Germany [14
]. Other studies have reported a higher limitation on QoL caused by PHN [13
]. Thus, the utilized values in our model might underestimate the impact on QoL due to PHN, whereas the HZ utilities used in our model might overestimate the impact of HZ on QoL [29
]. However, we reduced this uncertainty by varying respective utility values within SAs. Furthermore, we set the baseline utility value for healthy individuals in base-case to 1. Thus, the impact of HZ and PHN and therefore the effect of the vaccine might be overestimated, since age-specific utilities among elderly tend to be lower than 1. Even though for health economic evaluation the utilization of accurate utility data for QoL assessments is critical, in Germany age-specific utility data for healthy but also for the most indications are scarce since cost-utility analyses do not have the relevance in decision-making in Germany as in other countries. We identified in a literature search in total five studies evaluating the QoL among healthy individuals in Germany [61
]. Since all studies have certain limitations (e.g. only visual analog scale values was reported, no age-specific values reported, small study sample) we decided against using values of one of these studies for baseline utility values within the base-case analysis. However, by considering age-specific baseline utility values in SA we provided insights into this factor. The cost data used in the model were derived from a database from one large regional SHI in Germany and might not necessarily be representative for the whole population living in Germany. However, since countrywide treatment guidelines for HZ and PHN exist and since prices are mandatory for all SHIs in Germany, we believe that the utilization of these input-data is justified and representative. Furthermore our incidence data included both immune-competent and incompetent individuals. Since the live-attenuated vaccine is licensed for immune-competent individuals only, the incidence figures used in the model might be slightly overestimated. Our treatment cost input-data did not include privately covered expenditures for health-care services and over the counter drugs. Hence, costs from SP might be underestimated. However, since HZ and PHN-treatments are usually covered by the SHI, we suspect that the level of underestimation is low. As evidence was lacking concerning the duration and waning of vaccine-induced immunity, we had to make a few assumptions but included them in the SA. Finally, a real cost-effectiveness threshold concerning the WTP does not exist for the German health-care system. Therefore especially the ICERs and PSA results on costs per QALY gained have to be interpreted with caution. However, when comparing different age-groups to be targeted by routine HZ-vaccination and when comparing different scenarios (e.g. with and without booster vaccination), the lack of a cost-effectiveness threshold for Germany does not constitute a relevant limitation.
Our model provides valuable analyses and insights when considering the implementation of an efficient strategy for the prevention of HZ and PHN, and it will guide decision-making when developing a vaccination recommendation for Germany. First, this model reflects the efficacy of the vaccine quite well, since the HZ- and PHN-definitions as well as age-strata used in this model were similar to those used in the clinical-trials [21
]. HZ-cases with a clinical diagnosis but also with a ‘suspected’ diagnosis were included in both data-sets. Second, the definition of PHN in the clinical trial was pain persisting more than 90 days after rash onset. This matches exactly with the definition in our input-data, in which HZ-cases became PHN-cases when they were diagnosed or received a PHN-specific medication at least three month after HZ-diagnosis. Second, a further strength of our analysis is the intensive parameterization during modeling. The base-case and the SAs demonstrate that on the one side a careful selection of reliable input-data is important, but on the other side a wide range of SAs has to be conducted to reduce uncertainty within model results. Especially HZ- and PHN-related incidence and VE have to be incorporated with caution, since these factors influence the analyses considerably. However, for our model incidence data were utilized from a large study recently conducted in Germany. These data correspond well with results from a nationwide incidence study in Germany and another retrospective data analysis in Germany, but also with study results from other countries [17
]. Thus, this incidence input-data can be regarded as rather robust. SAs with variations of these data within the respective CIs have little impact on our results. However, vaccine characteristics are based on a less rich data fundament. While data on the VE in individuals from the age of 50 years is available, data on the duration of vaccine-induced protection and waning rates of vaccine-induced immunity is lacking and assumptions had to be made. Therefore, VE data were analyzed in SAs to assess the associated uncertainty. Based on one clinical-trial we assumed the period of stable VE following vaccination to be ten years. Since this assumption carries some uncertainty, we conducted a structural SA in which we varied the period of stable VE in order to assess the associated impact on results. Furthermore, evidence on the exact relative annual waning rate was utilized from literature. To analyze the impact of waning on ICERs we neglected the existence of a waning rate in one scenario and then subsequently increased it within SAs. A combined sensitivity analysis of varying waning immunity rates and durations of stable VE illuminated their combined impact on the optimal age at vaccination. Going one step beyond, we analyzed also the potential impact of differently designed fictitious booster scenarios on the results. This enables to assess booster scenarios way before evidence on the potential need for booster is established [86
]. Finally, the SA in respect to HZ-recurrence shows a considerable impact of this parameter on the results. However neglecting HZ-recurrence has a lower impact than considering a high HZ-recurrence. This analysis shows that there is an urgent need to establish more evidence on HZ-recurrence on the long term view. Since HZ-recurrence but also HZ-booster vaccination are important parameters when conducting a health economic evaluation of HZ-vaccination in a given country, we believe that our results are also of high interest to other countries that consider the introduction of routine HZ-vaccination in their health-care systems. For our study we were able to use numerous country-specific input-data of high quality.