Universal varicella vaccination in Denmark: Modeling public health impact, age-shift, and cost-effectiveness

We modeled the long-term clinical and economic impact of two-dose universal varicella vaccination (UVV) strategies in Denmark using a dynamic transmission model. The cost-effectiveness of UVV was evaluated along with the impact on varicella (including age-shift) and herpes zoster burden. Six two-dose UVV strategies were compared to no vaccination, at either short (12/15 months) or medium (15/48 months) intervals. Monovalent vaccines (V-MSD or V-GSK) for the 1st dose, and either monovalent or quadrivalent vaccines (MMRV-MSD or MMRV-GSK) for the 2nd dose were considered. Compared to no vaccination, all two-dose UVV strategies reduced varicella cases by 94%-96%, hospitalizations by 93%-94%, and deaths by 91%-92% over 50 years; herpes zoster cases were also reduced by 9%. There was a decline in the total number of annual varicella cases in all age groups including adolescents and adults. All UVV strategies were cost-effective compared to no vaccination, with ICER values ranging from €18,228-€20,263/QALY (payer perspective) and €3,746-€5,937/QALY (societal perspective). The frontier analysis showed that a two-dose strategy with V-MSD (15 months) and MMRV-MSD (48 months) dominated all other strategies and was the most cost-effective. In conclusion, all modeled two-dose UVV strategies were projected to substantially reduce the clinical and economic burden of varicella disease in Denmark compared to the current no vaccination strategy, with declines in both varicella and zoster incidence for all age groups over a 50-year time horizon.


Introduction
Varicella (chickenpox) is an acute and highly infectious disease caused by the varicella zoster virus. Varicella infections commonly present as a generalized pruritic vesicular rash with fever and malaise. The infections are generally mild and self-limiting, but can sometimes result in complications (e.g., bacterial infection of skin and the soft tissue, pneumonia, and encephalitis) or, rarely, death [1][2][3]. The risk of complication and death is greatest among the unvaccinated, the immunocompromised, and other high-risk groups (infants, pregnant women, and adults) [1][2][3]. Lowered  addition, this model used the latest data on vaccine performance parameters derived from 10-year clinical trial data for two different vaccine formulations [22]. The detailed vaccine parameters included in this model are described in Table D in S1 Text and their derivation from 10 years of randomized controlled trial data are described elsewhere [22]. The model was calibrated to age-stratified varicella seroprevalence and HZ incidence from a proxy country, Norway, via maximum likelihood estimates method. The observed and fitted varicella seroprevalence and HZ incidence plots are shown in Fig A in S2 Text. Model assumptions regarding exogenous boosting were updated based on real-world evidence showing the impact of contact with persons with infectious varicella on rates of HZ [19]. Using data on relative incidence of HZ following household exposure to varicella, we estimated the proportion of individuals boosted and the duration of boosting (see S1 Text for additional details) to describe the temporary immunity conferred by exogenous boosting. Additional details are provided in the Supplement regarding the model structure (S1 Text) along with the epidemiological, health resource utilization, cost and health utility parameters used (S2 Text).

Vaccination strategies
A total of six strategies (A-F) for UVV, each involving two doses of varicella vaccine, were compared to the no vaccination strategy over a 50-year time horizon (Table 1). Four varicella vaccines from two manufacturers were considered: monovalent Varivax (V-MSD) and quadrivalent ProQuad (MMRV-MSD), both manufactured by Merck & Co., Inc., Rahway, NJ, USA; and monovalent Varilrix (V-GSK) and quadrivalent Priorix-Tetra (MMRV-GSK), both manufactured by GSK, Belgium. Two vaccine doses were provided at either short (12 and 15 months) or medium (15 months and 48 months) intervals, aligning with the Danish National Immunization Program's current schedule [4]. For all UVV strategies, vaccination coverage rates were set to 94% of those eligible for the first dose, and 89% of those eligible for the second dose, consistent with current measles-mumps-rubella (MMR) vaccination coverage rates [23]. Children who were 2-12 years of age at the time of UVV introduction were eligible for catchup vaccination with two doses of monovalent vaccine, with coverage assumed to be 90% for each dose.

Model outcomes
Epidemiological and clinical outcomes included: the annual incidence rates of natural varicella, breakthrough varicella, and HZ; the cumulative numbers of varicella cases, outpatient visits, hospitalizations, and deaths over 50 years; and the cumulative numbers of wild (resulting from prior infection with natural varicella virus) and vaccine-related HZ cases and deaths. Age-stratified annual varicella incidence rates and the age distribution of cases were also reported to evaluate if there is a potential risk of age-shift of varicella infection to older populations. This study also assessed the impact of UVV on exogenous boosting and HZ incidence. The current model included assumptions for exogenous boosting, modeled from a recent realworld study with 20 years of follow-up conducted in the UK (S1 Text).
Cost outcomes for all strategies were calculated from both payer (direct costs) and societal (direct and indirect costs) perspectives, along with incremental cost-effectiveness ratios (ICERs) for each UVV strategy compared to no vaccination. In addition, a frontier analysis was conducted to compare the effect and cost associated with the various UVV strategies, with comparisons made between strategies lying on the effectiveness frontier in the cost-effectiveness plane. All prices were updated to 2020 Euros. Discounting for costs and quality-adjusted life years (QALYs) followed the time-variable discount rate prescribed by the Danish Ministry of Finance of 3.5% for years 0-35, 2.5% for years 36-70, and 1.5% after 70 years [24].
In the absence of a formal cost-effectiveness threshold for Denmark, we used the per capita gross domestic product (GDP) approach recommended by the World Health Organization and also compared results to the cost-effectiveness threshold recommended by UK's Joint Committee on Vaccination and Immunisation (JCVI; €23,964/QALY gained equivalent to £20,000/QALY gained), which is much lower than Denmark's per capita gross domestic product (€53,552) [25,26].
Additional scenario analyses explored outcomes at 25-and 100-year time horizons and with 3% or 5% annual discount rates (S3 Text). Additionally, the model was re-calibrated to explore outcomes in the absence of exogenous boosting. Deterministic and probabilistic sensitivity analyses were conducted on a subset of parameter values. Cost parameters were varied by ±20% and other parameters by ±5% of the baseline value. For the probabilistic analysis, 500 random parameter sets were drawn from uniform distributions.

Impact of UVV on varicella age-shift
In the absence of vaccination, annual varicella incidence was highest among children aged 1-5 years (10,859 per 100,000 persons) followed by incidence among children aged 5-10 years (9,065 per 100,000 persons). Immediately following the introduction of UVV, total varicella incidence declined significantly in all age groups including adolescents and adults (Fig 3). After the initial drop following the introduction of UVV, a small increase in total varicella incidence (Fig 3) was observed in children aged 5-10 years and 10-15 years in the first two decades, after which incidence again declined; the incidence in both age groups was still lower compared to no vaccination strategy. The magnitude of the increase in these age groups varied with strategies and short vs medium interval. For example, varicella incidence for MSD medium interval Strategy E for 5-10-year-olds dropped from pre-UVV values of 9,065 per 100,000 to 218 per 100,000 (263 per 100,000 for Strategy F) and then peaked at 343 per 100,000 versus 440 per 100,000 for the comparative GSK Strategy F (Fig 3). However, varicella incidence in all age groups was substantially lower for all modeled strategies compared to no vaccination, for every year following UVV introduction. Fig A in S3 Text shows annual varicella cases by age groups over time. There was significant decline in varicella cases after UVV including among children aged >10 years, even though their relative proportion was higher.

Impact of UVV on the clinical burden of herpes zoster
All six UVV strategies projected a decline in HZ incidence of 25%-26% at 50 years post-UVV, compared to no vaccination ( Table 2). The model projected a small, transient increase of less than 1% in HZ incidence, peaking about 3-4 years after UVV introduction, followed by a consistent decline beginning at about 5-10 years (Fig B in S3 Text). The cumulative number of HZ cases over 50 years was projected to decrease by approximately 9%, regardless of UVV strategy (Fig 2), from 723,945 HZ cases under no vaccination strategy to 656,802-660,784 HZ cases with UVV ( Table 2).

Cost-effectiveness of UVV
In the absence of varicella vaccination, the cumulative cost to manage varicella and HZ disease in Denmark over 50 years was projected to be €1.1 billion from the payer perspective and €1.2 billion from the societal perspective (Table 3). Compared to no vaccination, each strategy increased total costs (11.0%-12.1% from the payer perspective and 2.1%-3.2% from the societal perspective), and led to projected gains in QALYs of 6,750-6,813, to give ICER values of   €18,228-€20,263/QALY from the payer perspective and €3,746-€5,937/QALY from the societal perspective. All UVV strategies were cost-effective compared to no vaccination (Table 3). In frontier analysis, Strategy E dominated all other vaccination strategies and was found to be cost-effective at a threshold of 1 x GDP per capita as well as JCVI threshold from both payer (ICER: €18,228/QALY) and societal perspectives (ICER: €3,746/QALY).

Uncertainty analyses
In the one-way deterministic sensitivity analysis, the cost-effectiveness ratios compared with no vaccination remained relatively stable to parameter variation, with values ranging between €15,975 and €20,481/QALY from the payer perspective for Strategy E (Fig 4A; and  also remained below the 1 x GDP threshold for all parameter variations. The cost-effectiveness acceptability curve for Strategy E is shown in Fig 5; the probabilistic sensitivity analysis point clouds for all strategies from payer and societal perspectives are shown in Figs E and F in S3 Text, respectively.

Scenario analyses
Under the assumption of no exogenous boosting, there was no transient increase in HZ cases following introduction of varicella vaccination as observed in the base case. All strategies resulted in slightly greater reductions in HZ cases (

Discussion
This modeling study evaluated the cost-effectiveness of UVV in Denmark and addressed common concerns for implementing UVV in Nordic countries, namely the impact on age-shift and exogenous boosting. Our study found that all six strategies for introducing 2-dose UVV in Denmark would substantially decrease the clinical burden of varicella and were shown to be cost-effective compared to the no vaccination strategy, with Strategy E being the dominant strategy, from both payer and societal perspectives. Our study further reported no varicella age-shift to older age groups and showed a reduction in HZ incidence over 50 years which is consistent with recent literature [2,10,[15][16][17][18][19]. Compared to no vaccination, all two-dose UVV strategies reduced varicella cases by 94%-96%, hospitalizations by 93%-94%, and deaths by 91%-92% over 50 years while HZ cases decreased by 9%. We found that all three strategies with MSD vaccines resulted in lower varicella incidence, breakthrough cases, and mortality when compared with equivalent strategies with GSK vaccines. This is likely due to the higher first dose permanent protection or "take" of the MSD vaccines (90.3%) compared to the GSK vaccines (61.7%) [22], and is consistent with observational studies [27][28][29] as well as data reported in randomized controlled trials (1 st dose efficacy of V-MSD: 94.4% [30]; V-GSK: 67.2% [31]).
All six short and medium interval strategies were cost-effective compared to no vaccination. The cost-effectiveness frontier analysis further provided information on which strategy was cost-effective among all strategies under consideration. However, the medium interval Strategy E (V-MSD at 15 months and MMRV-MSD at 48 months) was the only UVV strategy on the cost-effectiveness frontier and deemed to be cost-effective at a threshold of 1 x GDP per capita from both payer and societal perspectives. The medium interval strategies align with the current MMR schedule in Denmark, thus possibly giving these strategies greater weight when considering adopting UVV in Denmark.
This study assessed whether an age-shift in varicella cases would occur following the introduction of UVV, a common concern of implementing UVV in many countries [2,10,17]. Our model showed that varicella incidence in all age groups declined by 97%-99% for all strategies, compared with no vaccination over 50 years and did not predict any significant age-shift after UVV introduction. There was a small increase in incidence in children aged 5-15 years with all vaccination strategies during the first two decades which was still well below pre-UVV incidence. These results align with real-world studies showing significant declines in varicella incidence and varicella-related hospitalizations following UVV in all age groups, further establishing that there is no evidence of age-shift after UVV [2,10,[15][16][17].
Some previously published mathematical models, including those by Brisson et al and van Hoek et al, had predicted that UVV would lead to an increase in the burden of HZ due to assumed pronounced exogenous boosting effect, and further concluded that UVV may not be a cost-effective strategy [14,32,33]. The effect of UVV on HZ is strongly correlated with assumptions about the magnitude and duration of exogenous boosting, both of which are subject to significant uncertainty [34]. Brisson et al and van Hoek et al both assumed magnitude of exogenous boosting effect is 100% (versus 33.45% in our model) with the duration of immunity against zoster ranging from 20 years to lifetime versus 81.3 years in our model. Recent literature has shown that the impact of UVV on exogenous boosting may not be as significant as previously thought [2,18,19]. For example, a study presenting 25 years of real-world data from the US (which introduced 1 dose UVV in 1995 and 2 dose UVV in 2007) did not show an increase in HZ incidence attributable to UVV and seems to indicate that rates of HZ will decline as vaccinated children age [17,35]. The exogenous boosting assumptions used by our model were calculated from a recently published real-world study with 20 years of follow-up that examined the impact of contact with persons with infectious varicella on rates of HZ (See S1 Text for more details) [19]. In contrast with previous models, due to use of latest data, our results showed total HZ incidence was projected to decrease by about 25% by 50 years, with 9% fewer total HZ cases over the 50-year period compared to no UVV. Using current data, our model estimated all vaccination strategies to be cost-effective compared with no vaccination after accounting for the impact on exogenous boosting and HZ incidence, from both the payer and societal perspectives, considering both the JCVI and GDP thresholds. Another strength of our model compared to previous models, is the use of recently updated vaccine performance parametrization by Pillsbury et al [22], which used a new deterministic model and 10 years of follow-up to clinical trial data for both V-MSD [30] and V-GSK [31] rather than van Hoek et al's model which only modeled V-MSD, thus allowing for the direct comparison of vaccine performance [32].
The results of our model are robust to variation in input parameters, and are consistent with recently published results for Turkey, Italy, Switzerland, UK and Norway, which showed significant reductions in the clinical and economic burden of disease after the introduction of UVV [20,[36][37][38][39]. This is also consistent with a cost analysis done by the Danish Chamber of Commerce, which found annual cost savings of 169 million Danish Krone (€22.7 million) with a two-dose varicella vaccination strategy from the societal perspective [40]. In addition to the cost-effectiveness of vaccination strategies, several factors need to be considered when implementing UVV [41]. These include the ability to maintain high vaccination coverage (>80%), the flexibility of the pediatric vaccination schedule to accommodate an additional vaccination visit (if required), fiscal vaccination budget, as well as the programmatic goals of the country [3].

Limitations
Modeling studies incorporate various assumptions about which there is uncertainty. We used proxy data from Norway and Sweden when recent Danish data were unavailable, though the similarity between these Nordic countries suggests they can be appropriate substitutes. Since the list prices for MMRV formulations in Denmark were not available, we estimated list prices from the International Reference Pricing basket using an approach recommended by the Danish Ministry of Health [42]. List prices tend to be higher than tender prices; hence ours is a more conservative estimate of cost-effectiveness. It was assumed that, when the second vaccination occurred as a combination MMRV vaccine, the effectiveness of the quadrivalent vaccines was the same as that of a monovalent varicella vaccine dose. We used the latest realworld evidence to estimate the impact of exogenous boosting, however different data or exogenous boosting assumptions might lead to different HZ outcomes. Additionally, our model did not account for changes in the age distribution of the population over time. This may impact the results especially since Denmark, like other European countries, has an aging population. Finally, the model assumed zero productivity loss for both varicella and HZ in adults 65 years and older, possibly underestimating true societal cost savings of UVV.

Conclusions
All modeled two-dose strategies of universal childhood varicella vaccination were projected to substantially reduce the clinical burden of varicella and were cost-effective compared to no vaccination in Denmark. Our model also showed no significant age-shift, and varicella incidence was substantially reduced across all age groups including adolescents and adults compared to no vaccination. Although there was a small increase in HZ incidence within the first few years of UVV introduction, overall HZ incidence decreased over 50-year time horizon. While all vaccination strategies were cost-effective compared to no vaccination, the medium interval strategy, V-MSD at 15 months and MMRV-MSD at 48 months, was the most costeffective from payer and societal perspectives. These results provide evidence to support the introduction of universal childhood two-dose varicella vaccination in Denmark.