Publications were primarily identified through Medline and EconLit searches and from citations from identified publications. Search terms included “papillomavirus, human”, “cost-benefit analysis”, “models, theoretical”, and “vaccination”. This process identified five English language cost-effectiveness analyses of HPV vaccination, four of which were included in the study. An early, crude, explorative assessment, which was made for the US Institute of Medicine,23 was not reviewed here in view of
ReviewCost-effectiveness analyses of human papillomavirus vaccination
Introduction
Worldwide, 500 000 new cases of cervical cancer are estimated to occur each year, resulting in 250 000 deaths.1 In recent years, the link between human papillomavirus (HPV) and cervical cancer has been conclusively proven.2 HPV is now thought to be a necessary but not sufficient cause of cervical cancer.3 This type of necessary causal relation offers substantial scope for both primary and secondary prevention strategies.4
HPV is primarily spread through sexual contact,5 and is associated with a wide range of diseases, including genital warts6 and many forms of cancer.7 Although several HPV types are defined as highly carcinogenic (known as high-risk or oncogenic types), those most commonly responsible for cervical cancer are HPV16 and HPV18.8 Worldwide, HPV16 and HPV18 have been estimated to account for approximately 70% of cervical cancers.9 The non-oncogenic types HPV6 and HPV11 are the main cause of condylomata acuminata (genital warts).
The incidence of cervical cancer differs between regions, particularly between high-income and low-income countries.10 The variation is mainly a function of cytological screening efforts and the quality of the screening programmes. Through Papanicolaou screening, cervical cancer is largely preventable.9 However, the precursors of cancer and ambiguous cytology results still represent a major burden to health-care systems.9 In the USA, the costs of screening represent up to two-thirds of the direct economic burden of cervical HPV-related disease.11
To date the most promising prophylactic vaccines have been based on virus-like particles.4 Currently, there are two vaccine candidates: a bivalent vaccine targeted at the oncogenic HPV16 and HPV18,12 and a quadrivalent vaccine targeted at the oncogenic HPV types and the HPV types primarily responsible for genital warts (HPV6 and HPV11).13 Both vaccines have been shown to be safe, immunogenic, and highly effective against type-specific persistent infection.12, 13 Several countries have licensed the quadrivalent vaccine.
Section snippets
Models of cost-effectiveness
Mathematical models can play an important role in our understanding of the effect of a new intervention, as well as identifying the best strategies for its introduction. In the context of HPV vaccination, an additional complexity is introduced by the existence of effective cytological screening programmes. Since there are more oncogenic HPV types than those targeted by current vaccine candidates, vaccination cannot yet replace screening in high-income countries, and must be assessed as a
Epidemiological assumptions
Two of the studies14, 15 adapted a previous model by Myers and colleagues18 to simulate both high-risk and low-risk HPV types. Taira and colleagues17 generated infection rates through the dynamic modelling process, and then incorporated these rates into the natural history model of Sanders and Taira.14 Therefore, three studies adapted the same progression and screening model.18 Kulasingam and Myers15 gave no detail of progression rates, instead referring readers to previous studies.18, 19, 20
Vaccine and screening assumptions
Details of vaccine and screening base-case assumptions are shown in table 1. In their base-case models, all four studies assumed that vaccination occurred at age 12 years and examined various vaccination ages in their sensitivity analyses. In the only dynamic model,17 ICERs were sensitive to vaccination coverage assumptions, particularly when vaccination of both boys and girls was considered. The ICERs in the three static models were insensitive to the level of coverage assumed, since
Health outcome measures
All four studies included quality-of-life (QOL) measures. However, Kulasingam and Myers15 reported base-case results in cost per life-year saved. There was a wide variation in the grouping and definition of health states to which QOL weights were applied, which makes comparison difficult. However, health states that do correspond show differences. For instance, the range of stage I cervical cancer follow-up utility weights was 0·90–0·97.14, 16 This lack of consistency is surprising given that
Economic factors
Goldie and colleagues' base-case analysis16 adopted the widest costing perspective (table 1). Direct cost estimates are shown to vary widely for some disease categories (table 2). All four studies used an annual discount rate of 3% in their base-case analyses for both costs and benefits. Only one study presented their results over a range of discount rates (0–5%).14 As in any economic evaluation of a prevention strategy with long-term effects, the initial intervention costs and the choice of
Model validation
Ideally, we would have liked to see a comparison of the model results to type-specific and age-specific data on rates of HPV infection, cervical intraepithelial neoplasia, and cancer. However, none of the studies provided such a comparison. Two studies15, 16 reported that their models gave approximately a 3·5% lifetime risk of developing cancer in the absence of control, but no other comparison was presented. Even the original model provides no comparison of age-specific results to the data
Methodological issues and limitations
Uncertainty, particularly around vaccine efficacy and duration of protection, has been accounted for through the wide range of values used in sensitivity analyses. However, although all four studies did sensitivity analyses, only two of the studies did two-way15 or multi-way analyses.14 All studies found that their results were fairly robust around changes in vaccine efficacy, although Taira and colleagues17 found that when they modelled female plus male vaccination, cost-effectiveness was
Modelling design
Three of the economic analyses used static Markov models.14, 15, 16 This type of modelling design is unable to take into account the dynamics of viral transmission within a host population, and therefore is unable to properly assess herd immunity (ie, the protective effect conferred on a population by immune individuals within the population).22, 35, 36 If the contribution of herd immunity is ignored, then the effectiveness and cost-effectiveness of a vaccination programme is likely to be
Future directions
Ideally, future HPV vaccine cost-effectiveness studies should be based on dynamic modelling. However, the validity of future dynamic models will be dependent on the accuracy of the data used to determine the transmission dynamics. Given the uncertainty around many of the variables, studies should do comprehensive sensitivity analyses, involving both univariate sensitivity analysis and multivariate probabilistic sensitivity analysis. The results should be presented as incremental (ie, compared
Conclusions
Overall, and with the assumption that the main model input variables (eg, HPV incidence, disease progression, QALY weights) are accurate, the three static models are likely to have underestimated the cost-effectiveness of HPV vaccination.14, 15, 16 Their base-case results therefore suggest that the introduction of HPV vaccination could be considered cost effective compared with current practice in the USA. The only published economic analysis based on a dynamic model found vaccination (of girls
Search strategy and selection criteria
References (46)
- et al.
Vaccination against human papillomavirus infection: a new paradigm in cervical cancer control
Vaccine
(2005) - et al.
Efficacy and other milestones for human papillomavirus vaccine introduction
Vaccine
(2004) - et al.
The health care costs of cervical human papillomavirus-related disease
Am J Obstet Gynecol
(2004) - et al.
Efficacy of a bivalent L1 virus-like particle vaccine in prevention of infection with human papillomavirus types 16 and 18 in young women: a randomised controlled trial
Lancet
(2004) - et al.
Prophylactic quadrivalent human papillomavirus (types 6, 11, 16, and 18) L1 virus-like particle vaccine in young women: a randomised double-blind placebo-controlled multicentre phase II efficacy trial
Lancet Oncol
(2005) - et al.
Costs and effectiveness of alternative strategies for cervical cancer screening in military beneficiaries
Obstet Gynecol
(2002) - et al.
Setting the target for a better cervical screening test: characteristics of a cost-effective test for cervical neoplasia screening
Obstet Gynecol
(2000) - et al.
Health economic guidelines—similarities, differences and some implications
Value Health
(2001) - et al.
Vaccination against multiple HPV types
Math Biosci
(2005)
Human papillomavirus is a necessary cause of invasive cervical cancer worldwide
J Pathol
Epidemiologic evidence and human papillomavirus infection as a necessary cause of cervical cancer
J Natl Cancer Inst
Natural history of anogenital human papillomavirus infection and neoplasia
J Natl Cancer Inst Monogr
Epidemiology of genital human papillomavirus infection
Am J Med
Role of mucosal human papillomavirus in nongenital cancers
J Natl Cancer Inst Monogr
Human papillomavirus and cervical cancer—burden and assessment of causality
J Natl Cancer Inst Monogr
International trends in incidence of cervical cancer: II. Squamous-cell carcinoma
Int J Cancer
Cost-effectiveness of a potential vaccine for human papillomavirus
Emerg Infect Dis
Potential health and economic impact of adding a human papillomavirus vaccine to screening programs
JAMA
Projected clinical benefits and cost-effectiveness of a human papillomavirus 16/18 vaccine
J Natl Cancer Inst
Evaluating human papillomavirus vaccination programs
Emerg Infect Dis
Mathematical model for the natural history of human papillomavirus infection and cervical carcinogenesis
Am J Epidemiol
Economic evaluations of hepatitis B immunization: a global review of recent studies (1994–2000)
Health Econ
Cited by (118)
HPV-FRAME: A consensus statement and quality framework for modelled evaluations of HPV-related cancer control
2019, Papillomavirus ResearchCitation Excerpt :The actual price paid for vaccines at a regional or country level is often an outcome of confidential tender negotiations and may not be available to use in an analysis; in this situation it is recommended that threshold analysis be performed to determine the maximum price that should be paid. The key drivers of uncertainty in cost-effectiveness evaluations of HPV vaccination are duration of vaccine protection [14,16,34], natural immunity duration [34,35], model type (static/dynamic) [17,28], vaccine effectiveness [28] (measured by how effective the vaccine is in preventing persistent HPV-infection or HPV-related disease, or as applicable to the particular research question), comparators (screening/no screening) [27,28,36], vaccine price [27,36], discount rate [14] and HPV prevalence [14]. Furthermore, a recent meta-analysis of models predicting the long-term population-level effectiveness of quadrivalent vaccination in women and men found that key drivers of vaccination impact include how models capture heterogeneity in mixing between risk groups (e.g., age, level of sexual behaviour) and the level of natural immunity among women [35].
Commercial Aspects of Vaccine Development
2017, Micro- and Nanotechnology in Vaccine DevelopmentHerd immunity effect of the HPV vaccination program in Australia under different assumptions regarding natural immunity against re-infection
2013, VaccineCitation Excerpt :This would be sufficient to motivate inclusion of multiple model types in evaluations of vaccination programmes, as assessment of herd immunity effect is one of the key factors to potentially influence any public health decision making regarding improving the effectiveness of a vaccination programme. Theoretically, given the same coverage, dosage, vaccine delivery schedules, etc., the impact of vaccination with a prophylactic vaccine under the non-SIR models should be greater than under an SIR model [2,13,19,20]. This argument, however, does not clarify the magnitude of these differences, which depends on a particular modelled population, properties of the vaccine and implemented vaccination strategies.