Disparity of Strategies
There were remarkable disparities in the definitions of strategies used by the authors. The cocoon strategy was defined as a "vaccination of both parents immediately after birth of the child" by Scuffham et al. and Westra et al., and as "targeted vaccination of parents of newborn" by Coudeville et al. Lee et al. considered it to be "vaccination of mothers after birth plus another adult caregiver after the firstborn child", and Purdy et al. analyzed a subgroup of "adults ≥15 years of age who are the primary caretakers of infants <1 year of age". All have been considered as referring to the cocoon strategy in this review. Concerning the adolescent strategy, several ages were considered for vaccination.
There was also a disparity in the choice of strategies assessed, which can potentially be explained by the difference in vaccination schedules already in place in the studied countries as well as the type of market (e.g., public or private). For example, while the primary vaccination has been implemented in all countries, preschool vaccination is not recommended everywhere and implementation of a strategy such as 'cocoon' might be facilitated in a public market with tender process compared with a private market. Consequently, the authors had to evaluate different vaccination strategies depending on the local context. Moreover, some authors chose to evaluate vaccination strategies for different population groups separately (only adults, only adolescents and so on) whereas others used combinations of those groups.[15,16]
Lack of Data
Cost–effectiveness analyses are limited by the available data. First, the incidence and extent of under-reporting are not precisely known, and authors would benefit from more data about the true incidence of symptomatic cases in all age groups. Also, it is not clear whether the cost of the nonreported symptomatic cases is the same as the cost of those reported. Moreover, incidence data relative to asymptomatic cases is clearly not sufficient. Second, data about mortality is also rare. As assessed by Edmunds et al., the disease-related mortality rate can have a substantial impact on the cost–effectiveness results, and this is of importance in models considering protection of infants. Third, there remain crucial questions regarding the efficacy of the vaccine. Currently, there is a lack of knowledge concerning the duration of immunity acquired by individuals after vaccination, which may vary between strategies. In addition, several facts were not known at the time of analysis: efficacy and tolerability when more than five doses are assessed; safety and efficacy in nonpediatric populations; maternal immunization efficacy; and efficacy of vaccination at birth. Fourth, only one study collected health-state utilities for pertussis disease, and utility values may not be suitable for use in infants. Decision makers in different countries recommend different instruments for measuring utilities. For example, in the UK, the NICE has a preference for the EQ-5D valued using the UK TTO value set. Such values are not available for pertussis disease. Health-state valuations are not exhaustive, and most of the studies used assumptions. Further evaluations would benefit from new data. Last, authors often have access to limited data about costs, which ideally should be country and age specific (especially relative to mild pertussis and indirect costs). Major obstacles to the accurate assessment of the costs of pertussis are the severity and cost consequences of unreported and undiagnosed cases.
There is uncertainty regarding the degree of indirect protection provided by vaccination. The fact that booster vaccination may generate herd immunity seems to be well accepted, yet data about its extent are scarce. Level of infectiousness of different types of cases and mixing patterns are often assumed by authors. In dynamic models, many parameters are probably obtained from calibration, but it is not clearly reported whether all those calibrations are performed with appropriate and sufficient data. There are two possible ways to estimate herd immunity. One can analyze the impact of previously implemented programs, but the impact of herd immunity is largely dependent on local parameters, and is difficult to estimate because the changes in incidence following the implementation of a new vaccination program are generally not entirely attributable to this program. One can also develop and use transmission dynamic models, which require many input data but appear to be useful. Nowadays, the validity of transmission models could be assessed based on the impact of recent vaccination programs.
Although the current epidemiological situation for pertussis is not comparable with the prevaccine era, potential negative externalities of booster vaccination have to be appropriately evaluated. For example, implementation of adolescent vaccination is expected to reduce circulation of the virus among adolescents. Nonvaccinated individuals will therefore be less likely to be infected (symptomatically or asymptomatically) during adolescence, and may then not benefit from natural immunity when they become young adults, at an age when they are most likely to transmit infection to infants. In addition, the duration of immunity after vaccination may be shorter than the duration of natural immunity, and adolescents or young adults with low immunity levels may be less likely to benefit from natural boosting conferred by asymptomatic infections, due to lower prevalence of the disease around them if an adolescent booster strategy is implemented. These combined facts could result in a greater risk of infection among parents of young infants, leading to transmission of infection to the key risk group. Depending on the level of vaccination coverage and duration of immunity after vaccination, it is possible that the negative externalities associated with booster vaccination exceed indirect protection conferred to infants. This issue was not discussed in reviewed publications based on transmission dynamic models and would merit investigation. Also, it is important to report outcomes of vaccination policies by age group and evaluate the risk of negative externalities as a function of vaccination coverage and the duration of immunity conferred by vaccination, relative to natural immunity. Both positive and negative externalities must be discussed, and considered accordingly to the large extent of vaccination benefit.
Sensitivity Analyses & Key Drivers
As the lack of data is a major issue in evaluations of pertussis vaccination, it is necessary to conduct appropriate sensitivity analyses. The authors mostly used deterministic univariate sensitivity analysis. Probabilistic sensitivity analysis was reported in only one article. Although the conclusions were generally quite robust to sensitivity analyses, disease incidence and the associated issue of under-reporting were reported as key cost–effectiveness drivers. Case fatality rates might be expected to have a strong influence on ICERs as one of the reasons for using booster vaccination is to prevent risk of fatal complications in infants. However, only two studies found this parameter as a cost–effectiveness driver.[19,22]
One pecularity of pertussis vaccination is the multitude of strategies being compared by the authors, and some authors performed questionable comparisons.
For example, Lee et al. considered adolescent vaccination, adult routine vaccination, combined adolescent and adult routine vaccination and postpartum vaccination. The latter strategy is said to be 'dominated', presumably because it is more expensive and less effective than adolescent vaccination. However, these strategies have different objectives: postpartum vaccination aims at protecting infants, and adolescent vaccination aims at protecting adolescents. More relevant comparisons would have been postpartum vaccination versus no booster vaccination, or combined postpartum vaccination and adolescent vaccination versus adolescent vaccination. Coudeville et al. evaluated the adolescent vaccination strategy, then combined with cocoon and then with both cocoon and one-time adult vaccination. The one-time adult vaccination was not assessed separately and therefore this analysis does not address the issue of whether vaccinating adults only would be cost effective.
When more than two strategies are considered, strategies must be ranked from the least effective to the most effective, and ICERs must be calculated for each strategy versus the previous one after exclusion of dominated and weakly dominated strategies.[34–36] Unfortunately, this methodology was not used by most of the authors, leading to inappropriate conclusions.
Concerning models, the choice between static and dynamic approach is of utmost importance. A static approach may be considered as conservative, since a dynamic model will often (although not always) predict a more favorable outcome, due to a more rapid effect and because the effect of herd immunity accumulates over time. Nevertheless, the choice between static and dynamic approach should mostly depend on the type of strategy considered. When the primary objective of the strategy is to reduce transmission to infants – that is, when herd immunity development is not the primary objective of the strategy (e.g., for cocoon strategy) – a static approach is likely to be sufficient. That approach was appropriately used by Westra et al., which only considered strategies aimed at directly protecting infants. When adolescent or adult immunization strategies are considered (i.e., when the strategy implementation can potentially produce significant changes in the severity of infection), then a transmission dynamic model is necessary. There are many issues associated with dynamic models, but estimates of indirect effects are required, and unless there is any other way to account for them, dynamic models should be preferred in order to determine an optimal strategy.
Outbreaks may be identified via a stochastic approach; nevertheless, a major drawback of the stochastic approach is that the model running time is long. For example, De Vries et al. ran multiple simulations, but as the computational time required was considerable, they could perform only a limited number of simulations.
The correct time horizon is also subject to discussion, and presenting analyses over very long time horizons only may not be very informative for decision-makers. Indeed, it is very unlikely that models are able to predict disease impact in 50 or 100 years, since the transmission dynamics of pertussis are not fully understood, and some external factors are unpredictable. Many changes or events could happen in the long term, such as demographic changes or medical innovations, which would make the results in the very long term irrelevant. This raises concerns about the steady-state approach, as the time needed to achieve equilibrium is very long, approximately 80 years according to Coudeville et al. The choice of estimating ICERs at steady state might only seem insufficient for a decision maker, since they might be more interested in average costs and effects over a short- or medium-term time horizon, as well as in the long term.
Recommendations for Future Studies
Since this review was performed, an abstract has already been published, and we anticipate that a number of new models will be published within a few years.
Based on the review of published economic evaluations of pertussis booster vaccination, we would recommend the following for future studies:
To assess the predictive validity of existing models against recent data;
To use a timeframe over which disease incidence can realistically be predicted;
To compare each strategy versus the next best strategy, not versus no booster;
To consider the advantages and the feasibility of improving coverage rates in currently targeted age groups versus adding new age groups;
To consider research to get further insight on management of outbreaks.
As stated above, it is highly desirable to collect new data to improve the future evaluations of pertussis booster vaccination, especially costs of different types of pertussis cases, utility and transmission of pertussis, and incidence of symptomatic and asymptomatic cases for transmission dynamic models.
The rationale for the choice of strategies assessed is one of the key points that should be discussed in new studies. Indeed, it is very important to consider the possibility of combining strategies, and to consider relevant population subgroups (healthcare workers, or individuals with chronic obstructive pulmonary disease, as considered by Purdy et al.). Also, it could be useful to use the model to search for the optimal age of vaccination. Nevertheless, the choice of age of vaccination is likely to be influenced or constrained by the country-specific context.
It may be valuable to explore the cost–effectiveness of other subgroups of population. Nevertheless, authors will face incomplete data or even lack of information; for example, efficacy of booster during pregnancy. Therefore, the evaluation of such strategies would require strong assumptions.
Comparisons between epidemiological data in countries with different vaccination programs could be useful, and may provide clues about the potential extent of herd immunity effects. However, the disparity in vaccination policies (in terms of number of doses or age groups) inevitably has an impact on epidemiology, making the comparison difficult to interpret.
The majority of reviewed publications reported cost–effectiveness analyses (i.e., they addressed the issues of the value for money of booster vaccination). Budget impact analyses would also provide valuable information for decision makers about the affordability of vaccination. Although several articles provided costs with and without vaccination at country level, no full budget impact analysis on pertussis booster vaccination was found in the literature.
Moreover, it is expected that existing models will be adapted for future evaluations in other countries. Nevertheless, predictive validity of existing models, in particular of transmission dynamic models, is still to be assessed and should now be performed prior to conducting further adaptations. For example, it would be interesting to compare the disease incidence observed in the USA following implementation of booster vaccination to that predicted by Coudeville et al. However, this comparison may not be straightforward due to geographical variations in vaccination uptake.
Last, we noticed the need for a consistent terminology across studies to refer to the different age groups, to distinguish between 'young adults' and 'adolescents', or 'young children' and 'infants'.
Despite the differences at several levels between the analyses, economic evaluations were generally favorable to pertussis booster vaccination, especially in adults and adolescents, with some divergences concerning the age groups to vaccinate and exact epidemiological conditions under which vaccination would be cost effective. The existence of such divergences is hardly surprising given the lack of information available, and the assumptions authors had to make about epidemiological, vaccine-specific, costs and utility inputs. Furthermore, the epidemiology of pertussis varies over time, between countries, and sometimes between regions within countries. There remains uncertainty concerning the optimal vaccination policy to adopt, and future cost–effectiveness studies should explore and compare a wider range of strategies.
Expert Rev Pharmacoeconomics Outcomes Res. 2012;12(1):71-94. © 2012 Expert Reviews Ltd.