Epidemiological impact and cost-effectiveness of universal vaccination with Bexsero ® to reduce meningococcal group B disease in Germany. Vaccine, 34(29),

of


Introduction
In Germany, an average of 243 cases and 20 deaths of invasive meningococcal disease (IMD) due to serogroup B (MenB) were reported to the Robert Koch Institute (RKI) each year between 2009 and 2012 (statutory surveillance data, RKI, personal

Routine vaccination
Months protection b One-off catch-up c Months protection a Strategies involving routine adolescent vaccination were implemented in the dynamic model only.
b Waning protection from vaccination was implemented as a rate equal to 1/months protection.Where two values are specified this is the duration of protection following the priming course and then the booster, for example there is waning protection following the 3 dose course at 2, 3, 4 months at a monthly rate of 1/18 and following the booster at 12 months there is waning protection at a monthly rate of 1/36.
c For one-off catch-up campaigns the months of administration are provided to indicate the dosing schedule e.g.1-4 years: 0, 2 indicates 2 vaccine doses given 2 months apart in children aged 1-4 years.
in Germany, but is recommended for those at increased risk after individual risk assessment, such as household contacts of cases, laboratory workers and immunocompromised persons [5].
In January 2013 Bexsero ® became the first vaccine to be licensed in the EU to provide broad protection against MenB disease.This vaccine is based upon a number of surface proteins and an outer membrane vesicle component, and is thus potentially immunogenic against strains with sufficient expression of the vaccine antigens regardless of the capsular group [6].In Germany the Standing Committee on Vaccination (Ständige Impfkommission, STIKO) is the independent advisory group whose recommendations are required for inclusion of a vaccine in the national vaccination schedule and for reimbursement by statutory health insurance.Currently STIKO recommends Bexsero ® for persons at increased risk of acquiring IMD, but not for universal childhood vaccination [7].Modelling the potential impact of a new vaccine on disease burden provides valuable evidence to STIKO and while assessment of the cost-effectiveness of a new vaccine is not obligatory for development of a STIKO recommendation, results are valuable for deciding on an overall immunisation strategy.
To support decision making in Germany we adapted the independently developed model for England [8] to the German setting to predict the potential health impact and cost-effectiveness of universal vaccination with Bexsero ® against MenB disease.

Models
We used two models to estimate the potential impact of universal Bexsero ® vaccination in Germany due to the uncertainty about the effect of the vaccine on carriage [9]: a cohort model allowing for direct vaccine protection against disease only, and a dynamic transmission model that includes additional vaccine protection against carriage.These models are described fully elsewhere [8].Due to existent universal MenC vaccination in Germany and an extremely low incidence of meningococcal disease due to non-B serogroups (0.15 cases per 100,000 from 2009 to 2012), we considered MenB disease exclusively in the models.
Both models are age-structured with yearly age classes; individuals are born susceptible.Upon disease, quality of life losses for the acute episode were included.Following disease, individuals have three possible outcomes: survival without sequelae, survival with sequelae (with a reduced quality of life) or death.Those dying from the disease are assumed to lose the average life expectancy for the age at which they die.Individuals may die from other causes; published mortality rates were adjusted to remove deaths due to meningococcal disease as these are explicitly modelled.Vaccine induced protection was assumed to start one month after the second vaccine dose and we allowed for waning protection (modelled as a constant rate set to the reciprocal of the average duration of vaccine protection).We considered several vaccination strategies (Table 1), comparing these to no universal vaccination against MenB and treating cases as they arise, over a 100 year time horizon.

Cohort model specific details
A Markov model with monthly cycles was used (Appendix).Disease cases were generated through applying the age-specific probability of disease to the susceptible population; survivors of disease were removed from the susceptible pool.Years of life were weighted by the age-specific quality of life.Cohort sizes were based upon 2011 population statistics.Single birth cohorts were considered for routine infant or toddler vaccination; multiple cohorts were considered for strategies with catch-up vaccination.

Dynamic model specific details
Transmission of meningococcal carriage was represented using a Susceptible-Infected-Susceptible (SIS) model [10] without considering co-infection [11] and using a daily time step (Appendix).Disease cases were generated by applying an agespecific case:carrier ratio to the number of new carriage acquisitions.Vaccinated individuals with immunity could have protection against carriage acquisition (initially assumed to be 30% reduction in carriage acquisition) as well as disease.a For a single birth cohort, without vaccination against MenB the cost of treating and caring for the estimated 224 cases that would occur over the lifetime of the cohort is D 5.1 M; vaccinating the birth cohort would cost an estimated D 191.9 M and would result in an estimated D 873,500 in healthcare savings.

Model parameters
Details of the data sources used to estimate parameters are summarised below with full details provided in Appendix.
National surveillance data from RKI were used to estimate agespecific disease incidence (data from 2009 to 2012) and case fatality (2002 to 2012) for MenB disease; the longer time period was used for case fatality due to the small annual number of meningococcal deaths.For the dynamic model MenB carriage prevalence estimates were based on a systematic review of all serogroup carriage combined with serogroup specific information from a carriage study in Germany [12].
Each case was assumed to be hospitalised, with 48% requiring ambulance transfer.The proportion of survivors with mild and severe sequelae was estimated from the literature [13][14][15][16][17]1].Quality of life losses for survivors with sequelae were based on currently unpublished data from the MOSAIC study, a case-control study of MenB survivors in the UK [1]; losses for carers of a person with sequelae were also considered [18].
Acute health care costs included the cost of: ambulance transfer; hospitalisation; hearing assessment; and public health management.Costs due to loss of work were also included.The costs of aftercare included one follow-up appointment for those aged under 5 years, cochlear implants (0.4% of survivors), scarring treatment (4%), physical therapy (1.9%) and logopaedics treatment (3.7% of survivors under 19 years) for the year following illness.Annual support costs were included for mild sequelae (unilateral hearing loss) and severe sequelae (which included amputations, major [bilateral] hearing loss, and epilepsy).We assumed that all cases with an amputation would result in a 50% work loss over their lifetime, either for a parent or for themselves at a later time.
We considered several vaccination strategies including routine infant immunisation at varying ages with or without a catch-up campaign (Table 1).In the dynamic model we investigated routine adolescent vaccination (12 year olds) alone, or in combination with an infant programme.Vaccination uptake was estimated based on the uptake of other vaccines with similar age-specific schedules in current use.Vaccine strain coverage was estimated using results of the Meningococcal antigen typing system (MATS) assay on German strains [6,19].The 2015 pharmacy retail price of D 96.96 was used as the cost per vaccine dose.Costs of vaccine administration were estimated from administration costs for other vaccines in Germany.We included the costs of hospitalisation for severe fever and anaphylaxis as possible adverse events following vaccination, but did not include possible quality of life losses associated with adverse events, which were assumed to resolve quickly.

Effectiveness analyses
We calculated the number needed to vaccinate (NNV) to prevent one case by dividing the number of persons vaccinated by the number of cases averted under various model assumptions.

Cost-effectiveness analyses
Health outcomes were defined as cases averted, deaths averted and quality adjusted life years (QALYs) gained under vaccination.All costs were measured in Euros at 2013 prices, with previous costs adjusted based on the German consumer price index [20].In the base case, future costs and benefits were discounted back to their present value at a rate of 3.0% as recommended in Germany [21] and the analysis was undertaken from the payer perspective.

Scenario analyses
Parameter uncertainty was handled through scenario analyses and by probabilistic sensitivity analyses (PSA).Factors considered in scenario analyses included: disease incidence, population mixing, vaccination uptake, strain coverage, vaccine price, societal perspective (with and without the addition of quality of life losses for carers and costs for work loss) and discount rates.The PSA was used to characterise the uncertainty around other model parameters (Appendix).

Cohort model: direct effects (no vaccine effects on carriage)
Table 2 shows the predicted impact of vaccination in birth cohorts (663,026 individuals in a single birth cohort) over their lifetime.In the absence of MenB vaccination the model estimates 224 cases of MenB disease and 19 deaths would occur over a cohort's lifetime.Assuming 65% vaccine uptake and 82% strain coverage, vaccinating infants with a 2, 3, 4 + 12 months schedule is estimated to avert 34 (15%) of these cases and 3 deaths, with a similar number prevented under a 2, 4, 6 + 12 months programme (Fig. 1).Vaccination at 6, 8, 12 months of age averted 25 cases as the assumed  increased duration of protection does not compensate for missing the cases that occur before vaccination.To consider catch-up strategies additional birth cohorts are included.Adding a large one-off catch-up strategy for 1-17 year olds to the routine infant schedule averted more cases.However, the percentage averted is reduced (from 15% to 6%) because incidence and assumed vaccine uptake are lower in 1-17 year olds compared to under one-year olds.

Dynamic transmission model: incorporating herd effects following vaccination
We assumed a 30% vaccine efficacy against acquisition.When considering routine infant vaccination alone, strategies starting earlier in life remained most favourable in reducing cases.The greatest health benefit in the short term, however, is achieved through routine infant vaccination with large-scale catch-up, which could reduce cases by 24.9% after 5 years and 27.9% after 10 years (Fig. 3).In the long term (20 years or more) policies including routine vaccination of 12 year olds are most favourable; after 50 years routine adolescent vaccination leads to an annual case reduction of 37.9% compared to no vaccination (Fig. 3).

Number needed to vaccinate
Considering direct effects only (no herd protection) 12,668 children would need to receive the vaccine to prevent a single case over a cohort's lifetime with a 2, 3, 4 + 12 months schedule.Assuming 30% vaccine effectiveness against carriage, this reduces to 8461 children and becomes even more favourable if older children are also vaccinated, reducing to 6373 children for the vaccination strategy 6, 8, 12 months +12 years.

Economic impact and cost-effectiveness
At a vaccine price per dose of D 96.96 vaccination of infants at 2, 3, 4 + 12 months within the cohort model is expected to cost D 191.9 M annually (Table 2).The predicted reduction in healthcare costs over a cohort's lifetime as a result of direct vaccine effects is D 873,500 with a resulting incremental cost-effectiveness ratio (ICER) of D 2.0 M per QALY gained.Assuming direct vaccine effects only, all vaccination strategies considered resulted in very high ICERs, with strategies that included catch-up being least favourable (Table 2, Fig. 2).Allowing for herd effects improves the cost-effectiveness of vaccination, however, the ICER remains  over D 500,000 for all considered strategies (Table 3).The inclusion of herd effects makes catch-up in addition to routine infant immunisation more economically favourable than routine infant immunisation alone.The lowest ICERs in this context are produced by strategies with routine adolescent immunisation (Table 3), due to the reduced dosing schedule and therefore lower costs for vaccination, and consistent targeting of those with high meningococcal carriage prevalence.

Sensitivity analyses
Increasing vaccine uptake in infants from 65% to 70% resulted in an estimated 1% point increase in cases averted assuming direct protection only or 2% point increase when including herd effects.Increasing the strain coverage to 92% resulted in a 2% point increase in averted cases assuming direct protection only or 3% point increase allowing for herd effects.Allowing for lower vaccine strain coverage in infants compared to older age groups (<1 year 68%; 1-9 years 88%; 10-19 years 86%; 20-49 years 79%; 50+ years 76%, see Appendix) reduced the estimated cases directly averted from 34 to 32 over the cohort's lifetime.Altering the assumption about population mixing to one based on self-reported contacts in Germany rather than a simple structure also reduced the proportion of predicted cases averted through vaccination from 22% to 19% in the dynamic model (Appendix).Both models were also sensitive to disease incidence (Appendix).ICERs remained very high even when using vaccine favourable assumptions or allowing for herd effects.From the societal perspective ICERs were lower, but remained over one million Euros per QALY gained even when allowing for herd effects (Appendix).Reducing the cost of the vaccine considerably reduced the ICER, however the cost per QALY gained remained over D 100,000 even with a vaccine price of D 0 and including herd effects (up to 60% efficacy against carriage acquisition) for the infant strategies.Routine adolescent vaccination strategies assuming indirect protection were more economically favourable, but the vaccine would have to be priced at less than D 1 a dose for the ICER to fall below D 30,000 per QALY gained.Of the parameters considered probabilistically in the cohort model, the incremental costs of vaccination were most sensitive to the vaccine administration costs and the rates of adverse vaccine reactions; incremental QALYs gained were most sensitive to the quality of life loss utilities and to a lesser extent the proportion of people with sequelae associated with disease and long-term sequelae (Appendix).

Principal findings
Model predictions suggest that only a small proportion and low absolute number of MenB cases could be prevented each year in Germany if Bexsero ® vaccination was introduced at 2, 3, 4 + 12 months and if the vaccine had no impact on carriage.This low absolute impact is due to the very low MenB incidence and only moderate anticipated vaccination uptake in Germany.Delaying the age at which the vaccine course is started reduces the potential health impact because young infants are at greatest risk of disease.The limited impact of MenB vaccination in the German setting is also reflected in very high NNVs.High NNVs (over 30,000) were also estimated for MenB infant vaccination in Canada [22].For comparison, much lower NNVs have been estimated to prevent one influenza-related hospitalisation when vaccinating children aged 6-23 months with an influenza vaccine at 50% efficacy (NNV 1031-3050) [23] and an estimated 80 children would need to receive rotavirus vaccination to prevent one hospitalisation [24].
In terms of economic impact, all modelled strategies for the use of Bexsero ® vaccination in Germany were associated with ICERs over D 500,000 per QALY gained under base case conditions.This was driven by the low absolute number of preventable cases predicted, particularly by the models that assumed no herd effects, however, evidence for an impact on carriage is uncertain [25,26].

Strengths and limitations
Our models use the latest available German specific data where possible and the use of a transmission dynamic model allows for indirect vaccine benefits (herd effects).Both payer and societal perspectives were explored.There is considerable uncertainty in some of the parameters used in the models and this was addressed using a partial probabilistic approach in the cohort model and scenario analyses in both models.
The models here consider MenB disease only, as we considered the impact on other serogroups would be very limited given their low incidence (there were only 42, 6 and 7 cases annually of MenC, W and Y, respectively in under 20 year olds from 2009 to 2012) and MenC vaccination coverage in targeted cohorts is already very high [27].There were limited available data on the incidence and costs associated with long term sequelae in Germany.Consequently, we did not include long-term costs for mild learning disability or institutional care for patients with severe disability, making our cost estimate for severe sequelae rather conservative [8,[28][29][30][31][32].We did include costs for rehabilitation, physical therapy, and speech therapy in the year after illness for a proportion of the patients.Not including the full range and costs of possible sequelae from meningococcal disease will have increased the estimated cost per QALY gained of the vaccination strategies, however in sensitivity analyses ICERs remained high even when the proportion of patients with sequelae and their associated costs were increased.In other aspects the model parameters were potentially vaccine favourable.For instance, we did not include quality of life losses from adverse vaccine reactions, allowances for strain replacement or potential deleterious effects of reducing meningococcal transmission.In addition, duration of protection in scenarios that included catch-up vaccination of toddlers may be overoptimistic based on a recently published small study of hSBA persistence [33].

Comparison with other studies
Modelling and cost-effectiveness studies on the use of Bexsero ® have been published for England [8,34], the Netherlands [30], France [35], Belgium [32] and Canada [29].In Spain the direct health impact alone was considered [36].As for the German models presented here, the England and Belgian analyses included the use of dynamic transmission models to appropriately allow for any herd effects.In France herd effects were estimated through incorporation into a Markov model and direct protection was principally considered in the Dutch and the Canadian studies primarily due to limited evidence of the effect of Bexsero ® on meningococcal carriage and transmission.The predictions here for vaccination in Germany are in line with those estimated elsewhere, namely that in the absence of herd effects routine immunisation early in life offers the greatest health impact, but with the inclusion of herd effects routine immunisation of teenagers becomes the best longterm strategy.Although the ICERs under base case conditions have been found to be high in all countries considered thus far, those presented for Germany are amongst the highest to date.This is in part explained by a higher vaccine price, the lower sequelae costs assigned to MenB patients as well as the very low MenB incidence (lower only in the Canadian model that also estimated high ICERs > $CDN 3 Million for infant vaccination).

Implications for policy makers
Our models suggest that maximal health impact in the short term could be achieved in Germany by vaccinating infants early in life.However, a recent study of paediatricians in Germany suggested only 13.4% of physicians preferred this strategy, in contrast to the 66.7% who preferred vaccination at 6, 8, 12 months (14% chose neither schedule) [37].Paediatricians were concerned about acceptance and safety of concomitant vaccination and possible parental refusal of other recommended vaccines since vaccinating MenB early in life would usually involve three vaccine shots per appointment.Thus, any immunisation decision will need to balance the potential benefits of any given vaccination strategy, the likelihood of the strategy being adopted in practice, as well as potentially unfavourable effects on the uptake of other vaccines.

Conclusions
Given the current very low incidence of MenB disease in Germany, implementation of universal infant vaccination with Bexsero ® would prevent only a small absolute number of cases.If the vaccine has an effect on carriage, the prevented number of cases and deaths increase significantly when vaccinating adolescents alone or -even more -when adding adolescent vaccination to a routine infant vaccination strategy.Whilst cost-effectiveness is not a central requirement for immunisation decision-making in Germany, the majority of scenarios considerably exceeded commonly used economic willingness to pay thresholds.

Fig. 1 .
Fig. 1.Predicted cumulative cases by year over the lifetime of a single birth cohort under different vaccination scenarios.

Fig. 2 .
Fig. 2. Results of 1000 Monte Carlo simulation predictions of difference vaccination strategies presented on the cost-effectiveness plane.

Fig. 3 .
Fig. 3. Predicted annual cases of MenB disease under different vaccination scenarios, assuming herd effects (30% vaccine efficacy against carriage, note the scaling of the Y-axis).

Table 1
Vaccination strategies against group B meningococcal disease modelled with base case vaccination parameters.a

Table 2
Epidemiological impact and cost-effectiveness of Bexsero vaccination against MenB disease in Germany assuming direct vaccine protection only, estimated using a cohort model.

Table 3
Epidemiological impact and cost-effectiveness of Bexsero vaccination against MenB disease in Germany allowing for herd effects, estimated using a dynamic transmission model.