Our economic analysis indicates that inclusion of RV vaccination in the Dutch NIP could be considered cost-effective depending on the exact cost of the vaccine and the impact of RV on children's quality of life. Assumptions which have a major impact on the ICER and which are also associated with a relatively large degree of uncertainty are (i) the QALY losses associated with RVGE, particularly in children treated at home and in caregivers, (ii) inclusion of potential herd protection, and (iii) the mortality rate in hospitalised RVGE cases. Along with assumed differences in underreporting, these aspects also explain the variation in the outcomes of the cost-effectiveness analyses performed by other research groups.
Strengths and weaknesses
In the base-case analysis, we chose not to include indirect protective effects for unvaccinated individuals within (approximately 5% in the Netherlands) and outside the vaccinated cohort. Recent epidemiological studies do, however, provide some evidence for the existence of such herd protection benefits [
32‐
38]. Additionally, several so-called dynamic models have been published which also predict an indirect protective effect in unvaccinated children [
39,
40]. However, further evidence is required before definite interpretations can be made. Therefore, we did not include these indirect protective effects in our base-case analysis as we did not want to present a too optimistic picture on the cost-effectiveness, which has been the case previously with pneumococcal vaccination [
29,
30]. Yet, conservative inclusion of limited potential herd protection effects in children (those aged less than 5 years of age) could improve cost-effectiveness considerably.
QALY losses of caregivers were not included in the base-case analysis. The impacts on the quality of life of caregivers are generally not included in Dutch cost-effectiveness evaluations, and including them here would have made a comparison with other interventions difficult (see below). Nevertheless, the Dutch pharmacoeconomic guidelines indicate that from a societal perspective, all costs and benefits should be considered, irrespective of who pays or loses, and who benefits [
42]. This would certainly provide an argument in favour of including all QALY impacts, such as those on caregivers. If we did incorporate QALY losses of caregivers, the ICER decreased considerably. It is not unlikely that a child suffering from RV has a similar QALY impact on both parents, and including two caregivers in the analysis could even be advocated. However, given that not all families consist of two caregivers and our approach is to remain conservative, this analysis was not pursued here.
Besides the assumed QALY decrements for caregivers, especially the assumed QALY losses for children treated at home had a major impact on the ICER. We based our QALY decrement for children on two published studies performed in the UK and Canada - the only ones currently available in the literature [
25,
26]. In the UK study, the utility of infants suffering from an RV infection was determined by health care providers, while in the Canadian study the utility decrements were estimated by the caregivers of children visiting a GP or paediatrician because of RVGE. Which of these estimates is more appropriate is not easy to determine. The Canadian study based their estimates on parents, which might be more suitable when performing a cost-effectiveness study from a societal perspective than estimates from GPs. On the other hand, the UK study provided age- and disease-severity-specific estimates, which might be more appropriate than one overall QALY decrement. We therefore chose to base our estimates on combining both studies (as described in the Methods). We do, however, note that combining the data from these different studies come with limitations. For example, the utility estimates and the duration of illness, which were used to estimate the QALY loss per case came from different foreign countries where one would ideally wanted to have those from one and the same study. Given differences in health-care systems and treatment patterns, combining information on duration and utilities from different countries may provide non-optimal estimations.
In contrast to previous studies, which used 50% of the QALY loss of cases attending primary care for cases not seeking medical care having, we assumed that the QALY loss in cases that would be treated at home would be 31% lower than for cases requiring a GP visit, based on illness durations. However, we feel that the former approach is likely to underestimate the QALY loss in these cases as only the most severe RVGE cases are expected to visit a primary care facility in the Netherlands (see also below). Although we are aware of the limitation of our approach, we do feel that this is the best approach. To anticipate on the uncertainty, we performed extensive sensitivity analyses on the QALY losses per case not seeking medical care.
We based the mortality rate (0.02% in hospitalised cases) on a study performed in England and Wales [
18]. Applying this rate in our model resulted in 0.65 deaths for all children less than 5 years of age (assuming a birth cohort of 180,000 infants each year). Increasing the mortality rates to 0.055% or 0.09% decreased the ICER to 27% and 42%, respectively. When we applied higher mortality rates, changes in the discount rate for health effects had a larger impact on the ICER, since the life years were obtained over a long time period.
The GP incidence used in our study is lower compared to those observed and used in other countries. As previously argued by Mangen
et al. this is likely to be related to the fact that in the Netherlands it is common practice to advise persons with GE to consult a GP only if symptoms remain for a longer period, or if the patient's health state gets worse [
9]. Furthermore, in the Netherlands it is not required to obtain a medical certificates from a GP to prove sickness or having a sick child at home. In other European countries like Germany, France and Spain such a certificate is required within 1-3 days off work in order to take care of a sick person, consequently GP's in these countries will be consulted more often [
9].
In contrast to previous studies, we based our efficacy estimates on specific European vaccine efficacy data wherever possible [
21,
23,
24]. Using these data instead of the general efficacy data (which were based on 11 countries throughout the world) probably gives more reliable estimates. We used efficacy estimates based on the latest available data for RotaTeq
®. These data show that the efficacy estimates after the second dose of RotaTeq
® are much more similar to the efficacy of Rotarix
® after the second dose than previously assumed [
11]. Also, remaining differences between both vaccines' efficacy estimates are based on clinical trials performed in different regions of the world and case definitions for disease were different between clinical trials performed with Rotateq
® and Rotarix
® [
19]. Strictly considered, our analysis - building on Rotateq
® clinical trials - is an economic evaluation for that specific vaccine, yet we expect the results for a Rotarix
®-specific analysis to be highly similar given the similarities between both vaccines.
Comparison with other studies
Our calculated cost-effectiveness ratio for RV vaccination is in between the estimates of previous Dutch studies [
7‐
10]. This is due to a combination of factors: (i) we used a lower total cost per vaccinee, (ii) we used higher QALY decrements in our study than in three of the four previous studies [
8‐
10], (iii) we used more realistic disease incidence data (including mortality rates) compared to all previous studies, and finally (iv) we estimated efficacy based on the most suitable data.
On the one hand, our results indicate that RV vaccination is probably more cost-effective than the current Dutch pneumococcal vaccination programme with the seven valent pneumococcal vaccine [
30]. On the other hand, our cost-effectiveness results show that RV vaccination is likely to be more expensive per QALY gained than other routine vaccination programs recently implemented such as HPV [
43] (€30,000 per QALY). It is as yet unclear how RV vaccination compares to other vaccination programs not yet implemented in the Netherlands, such as for varicella [
44]. Yet, the cost-effectiveness crucially depends on the exact vaccination costs of the RV vaccine if included within the Dutch NIP.
Implications and future research
Increasingly crowded infant vaccination schedules and restrained national budgets highlight the importance of cost-effectiveness analyses in the decision-making process on which vaccines should be included in national immunisation programmes. We show that RV vaccination in the Netherlands can be considered cost-effective depending on the total cost per vaccinated child. We also describe the main drivers for cost-effectiveness outcomes. In order to make an accurate appraisal of the RV vaccine and other currently available - but not yet introduced - vaccines as well as upcoming vaccines such as respiratory syncytial virus (RSV) vaccines, more accurate data regarding the main uncertain cost-effectiveness drivers are necessary.
Future research should, therefore, focus in particular on the number of deaths due to RV infections in the Netherlands as accurate data for the Netherlands and most other European countries are currently lacking. In addition, the relatively old cohort studies conducted at the population and GP level [
12‐
15] should ideally be updated, in combination with a cohort study conducted at hospital level. Furthermore, more research is needed on the quality of life of infected children. Consensus should be obtained regarding the question whether or not to incorporate the effect of childhood disease on the quality of life of caregivers [
19]. Finally, as potential herd effects have a large impact on the cost-effectiveness, continued surveillance and additional epidemiological studies in those countries in which an RV vaccination schedule has already been introduced should provide more insights into the epidemiology of RV over time, including such potential indirect effects.