Dr EJ Dasbach, Health Economic Statistics, Merck Research Laboratories, UG1C-60, PO Box 1000, North Wales, PA 19454-1099, USA. Email email@example.com
Objective To assess the potential epidemiological and economic impact of a prophylactic quadrivalent human papillomavirus (HPV) (6/11/16/18) vaccine for preventing cervical cancer, cervical intraepithelial neoplasia grades 2 and 3 (CIN2/3), CIN1 and genital warts.
Design Cost-utility analysis.
Population Female and male UK population 12 years or older.
Methods We adapted a previously developed multi-HPV type dynamic transmission to compare four female vaccination strategies, routine vaccination at age 12 years, and routine vaccination at age 12 years combined with temporary catch-up vaccination at ages 12–14, 12–17 and 12–24 years.
Main outcomes measures Costs, cases avoided, incremental cost per quality-adjusted life year (QALY).
Results The model projected that at year 100, each vaccination strategy could reduce the number of HPV 6/11/16/18-related cervical cancer, CIN2/3, CIN1 and genital wart cases among women by 86, 85, 79 and 89% respectively. Over 25 years, routine vaccination at age 12 years combined with a 12- to 24-year-old catch-up programme was the most effective strategy, reducing the cumulative number of cases of cervical cancer, CIN2/3, CIN1 and genital warts by 5800, 146 000, 28 000, and 1.1 million respectively. Over 100 years, the incremental cost-effectiveness ratios across all strategies ranged from £5882 to £11,412 per QALY gained.
Conclusion In the UK, a quadrivalent HPV vaccination programme that includes a catch-up strategy can reduce the incidence of cervical cancer, CIN and genital warts at a cost per QALY ratio within the range typically regarded as cost-effective.
The prophylactic quadrivalent vaccine, Gardasil® (Merck & Co. Inc., Whitehouse Station, NJ, USA) has been approved for use in the European Union for the prevention of high-grade cervical dysplasia (cervical intraepithelial neoplasia grades 2 and 3 [CIN2/3]), cervical carcinoma, high-grade vulval dysplastic lesions (vulval intraepithelial neoplasm grades 2 and 3) and external genital warts (condyloma acuminata) causally related to human papillomavirus (HPV) types 6, 11, 16 and 18.1,2 To formulate HPV vaccination guidelines and to support reimbursement decisions, policy makers will seek information on the epidemiological and economic consequences of immunisation programmes using this quadrivalent HPV vaccine.3 One source for such information will be projections from mathematical models.4 For instance, in a previous analysis based on a mathematical model, we projected the epidemiological and economic consequences of a vaccination programme using a quadrivalent HPV vaccine in the USA and found that the programme can be cost-effective.5 In 2007, in the UK, the Department of Health recommended HPV vaccination for the NHS immunisation programme.6 This HPV immunisation schedule recommends targeting girls aged 12–13 years of age for routine vaccination and girls up to 18 years of age for a catch-up programme that would last for 2 years. The purpose of this article was to review the results from a model developed to project and examine the epidemiological and economic consequences of a quadrivalent HPV vaccine in the UK. Specifically, this study answers the following questions: What is the potential impact of a prophylactic quadrivalent HPV vaccination on CIN, cervical cancer, cervical cancer mortality and genital warts in the population in the UK? What is the cost-effectiveness of a quadrivalent HPV vaccine immunisation programme when added to the current standard of care (i.e. cervical cancer screening and clinical management of CIN, cervical cancer and genital warts) from the perspective of the healthcare system in the UK?
Materials and methods
We adapted a previously developed mathematical model for evaluating the public health impact of immunisation programmes using a quadrivalent HPV vaccine in the USA for application in the UK.5 Details on the model structure and equations have been previously published.5 Components of the model that were modified for the UK included the demographic characteristics (e.g. mortality), screening,7 treatment and vaccination strategies as well as clinical, behavioural (i.e. sexual mixing),8 and economic input parameters. The following describes the screening and vaccination strategies evaluated, model parameters, model output, validation analyses and sensitivity analyses.
Screening and vaccination strategies
We assumed that all evaluated quadrivalent HPV vaccination strategies would be combined with current cervical cancer screening and HPV disease treatment practices in the UK. We defined the reference vaccination strategy for this analysis to be routine HPV vaccination of girls at age 12 years. We also examined the following HPV vaccination strategies in combination with screening: (1) routine female vaccination at age 12 years and catch-up female vaccination for ages 12–14 years, (2) routine female vaccination at age 12 years and catch-up female vaccination for ages 12–17 years and (3) routine female vaccination at age 12 years and catch-up female vaccination for ages 12–24 years. The catch-up programmes were assumed to be temporary, lasting only for a period of 2 years. Routine vaccination of 12-year-old girls was assumed to be permanent.
The model requires input values for demographic, behavioural, epidemiological, screening, treatment, vaccine and economic parameters. A comprehensive search of the literature was conducted to obtain values for these parameters. The baseline values and sources for key model parameters for the UK are summarised in Tables 1 and 2 and Tables S1 and S2.
Table 1. Costs of diagnosing and treating HPV disease (£) in the UK
Screening and vaccination programme strategy parameters
We used data from the 2005–2006 Cervical Cancer Screening Programme in England to estimate compliance rates for cytology screening.7 The prophylactic efficacy of the vaccine against incident HPV 6, 11, 16 and 18 infections was assumed to be 90%.14 We assumed the prophylactic efficacy of the vaccine against HPV 6-, 11-, 16- and 18-related CIN and genital warts to be 95.2 and 98.9% respectively.1 Vaccination was assumed not to have an effect on the natural course of HPV infection or disease prevalent at the time of vaccination or occurring after vaccination. We assumed the duration of protection of HPV vaccination to be lifelong for the base case consistent with previous models.4 The percentage of girls in the population at age 12 years receiving the three-dose vaccine (i.e. coverage) was assumed to be 80%.15 Vaccine coverage for the catch-up programme was assumed to vary by the age of the target group. In particular, we assumed that the coverage for 12–14, 15–17 and 19–24 years of age would be 40, 30 and 25%, respectively, in the first year of the catch-up programme. In the second year, the coverage rates were assumed to be double those of the first year.
All costs are reported in 2006 pounds sterling (£) (Table 1). Total costs of each strategy included the cost of cytology screening; cost of vaccination; total cost of diagnosing and treating detected invasive cervical cancer, CIN or genital wart cases; and the cost of following false-positive results of the Pap screening test. Although the NHS price of the HPV vaccine is £80.50, a volume-based discount is usual for vaccines procured centrally in bulk, as is the case in the UK. For the purposes of this analysis, we assumed a cost per dose for the vaccine of £75. A vaccination programme is likely to be predominantly school based and the cost for administration was assumed to be £3.40 per dose for the base case.16 Hence, the total cost of the HPV vaccine for three doses and administration was assumed to be £236. We did not include productivity costs (i.e. indirect costs) in the analyses. The planning horizon used in the analysis was 100 years. We assumed the size of the population of over 12-year olds at any given point in time over the 100-year horizon to be 100 000, with a female–male sex ratio of 1. Finally, all costs and effects were discounted to present value at an annual rate of 3.5%.17
Health utility parameters
We measured the impact of screening and vaccination on quality of life and survival using quality-adjusted life years (QALYs). Table 2 shows the health utility estimates used to construct the QALYs. In the absence of UK-specific population data, health utility data collected in the USA were used.12,13
The model simulates the rate at which susceptible individuals acquire HPV infection by accounting for the age-specific frequency and assortativeness of sexual partnerships in the population, the fraction of infected partners and the likelihood of transmission of infection per partnership. We used population data from the 2nd National Survey of Sexual Attitudes and Lifestyles in the UK to model the frequency of sexual partnerships by age and within sexual activity groups.8
We used a number of measures to assess the epidemiological impact and cost-effectiveness of each vaccination strategy. Epidemiological output included CIN2/3, CIN1, invasive cervical cancer, genital wart cases and cervical cancer-related deaths. The economic output of interest from the model included total costs, quality-adjusted survival and incremental cost per QALY. We measured quality-adjusted survival time by weighting survival time by the quality-of-life adjustment weights associated with each health state (i.e. the health utility values) and integrating the sum of all these weights over the planning horizon. We measured the cost per QALY ratio as the incremental cost difference between two strategies divided by the incremental QALY difference between the two strategies.
We programmed all model equations and inputs in Mathematica® (Wolfram Research, Champaign, IL, USA). We used the NDSolve subroutine in Mathematica version 5.2 to generate numerical solutions for the differential equations making up the model. The following strategy for simulations was followed: First, the baseline parameter estimates were used to solve the model for the prevaccination steady-state values of the variables. Second, we used the prevaccination data as initial values for the vaccination model. Next, we solved for the entire time path of the variables until the system approached a steady state (approximately 100 years). Finally, we used this solution to generate the previously described output for each of the screening and vaccination strategies.
Validation and calibration of the natural history component of the model has been described in detail previously.5 For the UK adaptation, we assessed the predictive validity of the model by comparing model predictions to observed epidemiological data from the literature on the incidence of cervical cancer and genital wart cases in the UK.
Extensive sensitivity analyses with previous models5,18 have identified those parameters that most influenced the results. Hence, we used these prior analyses, as well as those parameters with the largest uncertainty, to guide our sensitivity analyses. These parameters included duration and degree of vaccine protection, vaccine coverage rates, quality-of-life weights, vaccine series cost and the discount rate. In addition, we conducted a pessimistic scenario sensitivity analysis. For this analysis, we assumed that the costs and quality-of-life decrements for HPV diseases and duration of vaccine protection would be less than the values used for the reference case analysis. Finally, given the interest from policy makers on the additional value of protecting against infection with type HPV 6 and 11, we conducted a sensitivity analysis on the effect of protecting against infection with HPV 6 and 11.19
The model predicted with current screening, in the absence of vaccination, an annual incidence of HPV 16/18-related cervical cancer of 7.0 per 100 000 among girls and women aged ≥12 years. The overall incidence of cervical cancer observed in the UK among girls and women aged ≥12 years is estimated to be 10.2 per 100 000,20 of which approximately 70% is estimated to be attributable to HPV 16/18.21 In addition, the model predicted an incidence of HPV 6/11-related female genital warts of 127 per 100 000 among girls and women aged ≥12 years in the absence of vaccination. The overall incidence of genital warts in women observed in the UK is estimated to be 159 per 100 000,22,23 of which approximately 90% are attributable to HPV 6/11.24
Epidemiological impact of the HPV vaccination strategies (reference case)
Figure 1A shows the incidence of HPV 16/18-related cervical cancer cases over time by vaccination strategy. Across all strategies, the effect of vaccination was to steadily reduce the incidence of cervical cancer cases and deaths until the system approached a steady state, about 100 years after vaccination is initiated in the population. Compared with no vaccination, all vaccination strategies reduced the incidence of cervical cancer cases and deaths by 86% at year 100. A similar dynamic can be seen for the reductions in CIN2/3, CIN1 and genital wart cases as displayed in Figure 1B–D. These curves share similar qualitative features with those of cervical cancer. However, because CIN2/3, CIN1 and genital warts occur sooner following HPV infection, the curves are shifted to the left compared with the cervical cancer curves. In the long run, the model predicted that all vaccination strategies would reduce the incidence of CIN2/3, CIN1 and genital warts by 85, 79 and 89% respectively. The primary reason the long-run impact did not differ among the vaccination strategies was because the catch-up vaccination programmes were temporary.
Although the long-run reductions in HPV-related disease do not differ significantly among the vaccination strategies, the catch-up vaccination strategies realised both earlier and greater reductions in HPV-related diseases than the routine vaccination strategy that targets girls at the age of 12 years. Moreover, broadening the age range of the catch-up vaccination strategies resulted in greater and earlier reductions in disease. This is evidenced by the area between the curves in Figure 1 and is summarised quantitatively in Table 3. In particular, Table 3 shows the cumulative cases of HPV 6/11/16/18 disease events prevented overall and by HPV type with the most effective vaccination strategy (i.e. routine vaccination at age 12 years combined with a 12- to 24-year-old catch-up programme) relative to no vaccination 25 years after the introduction of the vaccine into the population. For example, under the overall heading in row 3, the model projected that 146 366 women of CIN2/3 could be prevented over 25 years in the UK by vaccinating girls and women aged 12–24 years compared with no vaccination. Table 3 also highlights that the benefits of vaccination could be realised early with a large portion of the HPV disease events avoided attributable to preventing infection of HPV types 6 and 11. For example, 97% of the HPV disease events prevented in the first 5 years were attributable to preventing infection of HPV types 6 and 11.
Table 3. Cumulative HPV 6/11/16/18 disease events prevented over time with routine female vaccination at age 12 years combined with a 12- to 24-year-old female catch-up programme relative to no vaccination in the UK
HPV disease event
Years since start of vaccination programme
HPV types 6/11 related
Genital warts (women)
Genital warts (men)
1 139 308
HPV types 16/18 related
Cervical cancer deaths
1 129 660
Cervical cancer deaths
1 311 735
Finally, one visual difference among the curves for the different HPV diseases can be seen in Figure 1D. In particular, the incidence of genital warts increases slightly after year 19, decreases again after year 29 and reaches equilibrium after year 70. The reason for this increase in genital wart cases is because the reduction in the pool of susceptible individuals in the population is not great enough to eradicate infection and disease in the population. As a result, the pool of individuals susceptible to infection slowly accrues. This pool of susceptible individuals eventually reaches a critical threshold size by year 19, which results in a slight, temporary increase in cases before decreasing again at year 29. This phenomenon of low disease levels followed by an increase in disease is a recognised dynamic that occurs in vaccinated populations, with coverage levels that are close to the critical coverage level necessary for eradication.25,26
Economic impact of HPV vaccination strategies (reference case)
For each vaccination strategy, we also measured the costs from the HPV diseases avoided relative to no vaccination. Figure 2 summarises the annual, discounted, HPV disease treatment costs prevented in the population in the UK by the most effective vaccination strategy (i.e. the 12- to 24-year-old female vaccination programme) relative to the screening programme without vaccination, stratified by HPV disease. For the first 20 years, the majority of costs avoided are attributable to the prevention of genital wart cases. This is consistent with the notion that genital wart cases occur sooner following HPV infection than cervical cancer and CIN.27 However, after year 20, the majority of costs avoided through vaccination are attributable to the prevention of cervical cancer.
In addition to examining HPV disease costs avoided, we examined total costs incurred and the cost-effectiveness of each strategy in preventing disease. To assess the cost-effectiveness of the vaccination strategies in preventing disease with respect to costs, we began by estimating the total discounted costs and effects (i.e. QALYs) accrued over a 100-year period for each strategy. These total costs and QALYs are shown in columns 2 and 3, respectively, in Table 4. The strategies in the table are ordered from least effective at the top to most effective at the bottom as measured in QALY units. Next, we calculated the incremental cost incurred to achieve an incremental gain in benefit of each successive strategy. These incremental costs and effects are shown in columns 4 and 5. The final column shows the ratio of the incremental costs to incremental QALYs gained (i.e. the incremental cost-effectiveness ratio). Note that the most effective strategy in the bottom row (i.e. vaccinating girls and women 12–24 years of age) has the highest cost-effectiveness ratio, £11,412 per QALY gained relative to vaccinating girls and women 12–17 years of age. In addition, the least effective vaccination strategy (i.e. routine vaccination of girls by the age of 12 years) has a slightly higher cost-effectiveness ratio (i.e. £5890 per QALY gained) than the cost-effectiveness ratio for the 12- to 14-year-old female vaccination strategy. Hence, routine female 12-year-old vaccination combined with a 12- to 14-year-old female catch-up vaccination strategy would be considered the preferred strategy relative to routine vaccination of girls by the age of 12 years because it is both more effective and more efficient. Thus, we note in Table 4 that routine vaccination of girls at age 12 years combined with catch-up vaccination of 12- to 14-year-old girls and women ‘weakly dominates’ the routine female 12-year-old vaccination strategy.
Table 4. Cost-effectiveness analysis of alternative vaccination strategies*
Assumes cost of vaccination series is £236 and duration of protection is lifelong. All costs are measured in 2006 pounds sterling (£), and costs and QALY are discounted at 3.5%. Time horizon is 100 years.
Compared with the preceding nondominated strategy. Strategy A is weakly dominated if there is another strategy; B, that is both more effective and more efficient than strategy A.
2 352 770
Routine 12-year olds
2 353 242
+12- to 14-year-old catch-up
2 353 268
+12- to 17-year-old catch-up
2 353 294
+12- to 24-year-old catch-up
2 353 327
We conducted a variety of sensitivity analyses on the incremental cost-effectiveness ratios. Table 5 summarises these results. The two most influential parameters on the results were the health utility values and the duration of vaccine protection. Assuming a shorter duration of vaccine protection or smaller quality-of-life benefits decreased the efficiency and the cost-effectiveness of all the vaccination strategies. In fact, when examined collectively with the pessimistic scenario, the cost-effectiveness ratio for the 12- to 24-year-old vaccination strategy increased threefold compared with the reference case cost-effectiveness ratio for this strategy using a scenario in which duration of efficacy was limited, quality-of-life benefits were less and HPV disease costs were lower.
Table 5. Summary of incremental cost-effectiveness ratios for sensitivity analyses
Routine + 12- to 14-year-old strategy (£)
Routine + 12- to 17-year-old strategy (£)
Routine + 12- to 24-year-old strategy (£)
Compared with the preceding nondominated strategy. Strategy A is weakly dominated if there is another strategy, B, that is both more effective and more efficient than strategy A.
HPV disease costs decreased by 25%, health utility for any HPV-related diseases equals 0.97, and duration of protection equals 10 years.
The results from this model can be useful for deciding which age groups are appropriate to target for introducing a quadrivalent HPV (types 6/11/16/18) vaccination programme in the UK. The model showed that during the long run, all vaccination strategies reduced the incidence of HPV-related disease similarly compared with screening without vaccination. However, one of the key findings from this analysis was that introducing a temporary, catch-up vaccination programme may reduce the incidence of HPV-related diseases earlier, more efficiently, and more effectively than a vaccination strategy that does not have a catch-up programme and only targeted girls at the age of 12 years. Moreover, we found that increasing the age band targeted by the catch-up programme resulted in greater reductions in HPV-related disease. Hence, the 12- to 24-year-old vaccination programme had the greatest impact on reducing HPV-related disease among the vaccination strategies evaluated.
Targeting broader age groups for vaccination, however, will also result in greater costs. To address this issue, we assessed the cost-effectiveness of each of the vaccination strategies. Based on this assessment, we found that the cost-effectiveness ratios for all of the vaccination strategies fell within the range of what would be considered cost-effective.28,29 In particular, the cost-effectiveness ratios of the four vaccination strategies were £5890, £5882, £5971 and £11,412 per QALY gained for the routine vaccination strategy at the age of 12 years and the routine plus catch-up vaccination strategies at the ages of 12–14, 12–17 and 12–24 years, respectively. Thus, although the current NHS HPV immunisation programme recommendation targets girls up to 18 years of age for catch-up vaccination, broadening the recommendation to target girls and women up to the age of 24 can reduce HPV-related disease further at a cost-effectiveness ratio that falls within the range of what has typically been considered cost-effective.30
One factor that contributed to this favourable cost-effectiveness ratio was the economic and quality-of-life benefits conferred by preventing genital warts. Figure 2 shows that preventing genital warts had the greatest impact on reducing costs during the first 20 years following introduction of vaccination in the population. This is further illustrated in a sensitivity analysis that excluded the benefits of preventing genital warts (i.e. eliminated protection against HPV 6/11). Specifically, we found that the cost-effectiveness ratio for the 12- to 24-year-old strategy when compared with vaccinating girls and women aged 12–17 years increased nearly 50% to £16,989 per QALY gained.
Another influential parameter in the model included the duration of vaccine protection. High sustained efficacy of the quadrivalent HPV vaccine has been demonstrated through 5 years of follow up to date.31 Decreasing the duration of vaccine protection to 10 years increased the cost-effectiveness ratios for the 12- to 24-year-old vaccination strategy. An interesting finding from this sensitivity analysis is that all the less-effective vaccination strategies were dominated, and hence were less efficient from a cost-effectiveness standpoint than the most effective vaccination strategy. The reason for this is that when the duration of protection is not lifelong, a more sustained impact on reducing infection and disease in the population can be achieved by targeting a broader age group for vaccination. Finally, we would like to note that the vaccination strategy that did not include a catch-up programme was always dominated by a more effective vaccination strategy that included a catch-up programme in all analyses. Unless coverage is more than 95% for routine vaccination of girls before the age of 12 years, catch-up vaccination programmes appear to be a more effective and efficient vaccination strategy.
We believe the dynamic modelling approach used in this evaluation has a number of strengths. First, modelling transmission dynamics can account for both the direct and indirect benefits of vaccination.25,32 For example Kohli et al.33 noted that the indirect, secondary benefits of herd immunity were not accounted for in their cohort modelling approach used to evaluate routine, prophylactic bivalent HPV vaccination strategies in the UK. Second, given that the output from this dynamic model is population based, the comparison with population statistics is better aligned than comparison of cohort model output with national registry data. Finally, a dynamic modelling approach is better suited for evaluating vaccination programmes that target heterogeneous populations (e.g. routine vaccination of 12-year olds combined with a 12- to 24-year-old catch-up population) than cohort models that tend to focus on specific age cohorts (e.g. routine vaccination of 12-year olds).
However, as with any model, there are limitations. First, we limited the vaccination strategies evaluated to those within the approved age range for vaccination.1 If the indication for vaccination expands to other age groups, we can expand the model to evaluate these strategies.
Second, we did not model any potential interactions between demand for vaccination and demand for cervical cancer screening. In particular, we assumed that the sexually active population had equal access to health care, be it vaccination, screening or treatment. Moreover, if coverage rates for screening decrease with time, then the cost-effectiveness of vaccination will improve.
Third, we limited the scope of the model to cervical diseases and genital warts. The quadrivalent HPV vaccine has also demonstrated the ability to reduce precancerous vulval and vaginal lesions.34 In addition, the quadrivalent vaccine has demonstrated evidence of cross protection (i.e. protecting against HPV diseases associated with infection from other HPV types not included in the vaccine).35 Incorporating these other benefits into the model will most likely improve the cost-effectiveness ratios further.
Fourth, we based vaccine degree of protection on what had been published in the label for Gardasil at the time of the analyses.2 More recent data have been presented showing that vaccine degree of protection is slightly higher than what was assumed in our analyses.36 Hence, incorporating the latest efficacy data would slightly improve the cost-effectiveness ratios for vaccination.
Fifth, we based administration costs of the vaccine on a study that had estimated what the cost of administration of a vaccine would be for a school-based vaccination programme. Not all individuals targeted for vaccination, however, will be school based. Hence, we conducted a sensitivity analysis on the cost of the vaccine series. We found that even with a 25% increase in the cost of the vaccine series, the cost-effectiveness ratios for all the strategies would still fall within a range that would be considered cost-effective.
Sixth, we did not account for any potential adverse effects associated with vaccination or screening. Quadrivalent HPV vaccination has been found to be generally well-tolerated.37 However, HPV vaccination has been associated with a higher incidence of injection site adverse events as well as self-reported fever.37 Given the nature of these adverse events were brief and not serious, we did not account for these effects in the model. We would expect had we accounted for the adverse effects associated with vaccination as well as screening in the model the impact on the cost-effectiveness ratios would have been minor.
Other areas of future development for the model include accounting for mortality and productivity costs (i.e. indirect costs), as has been performed in other vaccine cost-effectiveness analyses.38 Including these costs would further reduce the cost-effectiveness ratios. Including other diseases associated with HPV infection in the model would also reduce the cost-effectiveness ratios for vaccination. For instance, HPV infection has also been associated with cancers of the anus, penis, vagina, vulva and head and neck as well as recurrent respiratory papillomatosis. As evidence becomes available to support modelling the potential effects of vaccination against these other HPV conditions, the scope of the model will be broadened to incorporate them. Finally, we did not account for homosexual transmission of HPV in the model. Future refinements of the model will also explore this.
In summary, the results from this model suggest that in a setting of organised cervical cancer screening in the UK, a prophylactic quadrivalent HPV (6/11/16/18) vaccine programme that includes a catch-up strategy can (1) substantially reduce cervical cancer, CIN and genital warts, (2) improve quality of life and survival and (3) be cost-effective when implemented as a vaccination strategy that targets girls and women 12–24 years of age.
Conflict of interest
Gardasil was developed and is marketed by Merck & Co. Inc. E.J.D., E.H.E. and R.P.I. are all employees of Merck & Co. Inc.
Contribution to authorship
The authors contributed to development of the model, analyses and writing of the manuscript.
Development of the model and analyses were funded by Merck & Co. Inc.
The authors wish to thank Nathalie Largeron for her support and Karen R. Collins of JK Associates, Inc., for editorial assistance with this manuscript.