Introducing HPV vaccine and scaling up screening procedures to prevent deaths from cervical cancer in Japan: a cost-effectiveness analysis


N Yamamoto, Department of Global Health Policy, Graduate School of Medicine, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan. Email


Please cite this paper as: Yamamoto N, Mori R, Jacklin P, Osuga Y, Kawana K, Shibuya K, Taketani Y. Introducing HPV vaccine and scaling up screening procedures to prevent deaths from cervical cancer in Japan: a cost-effectiveness analysis. BJOG 2012;119:177–186.

Objective  To assess the cost-effectiveness of universal vaccination of 11-year-old girls against human papillomavirus (HPV) infection and increased screening coverage to prevent cervical cancer in Japan where the coverage of Papanicolaou smears is very low.

Design  A cost–utility analysis from a societal perspective.

Setting  Japan, 2010.

Population  The female Japanese population aged 11 years or older.

Methods  A Markov model of the natural history of cervical cancer was constructed to compare six strategies: i.e. a screening coverage rate of 20, 50 and 80% with and without routine vaccination at age 11.

Main outcome measures  Cervical cancer incidence, quality-adjusted life years (QALYs), costs and incremental cost-effectiveness ratios.

Results  Expanding the coverage of Papanicolaou smears from the current level of 20–50 and 80% yields a 45.5 and 63.1% reduction in cervical cancer incidence, respectively. Impact of combined strategies increases with coverage. Coverages of 20, 50 and 80% showed a 66.1, 80.9 and 86.8% reduction in disease, respectively. The costs of strategies with vaccination are four times higher than the cost of strategies without vaccination. Vaccinating all 11-year-old girls with bivalent vaccines with a Papanicolaou smear coverage rate of 50% is likely to be the most cost-effective option among the six strategies.

Conclusions  The introduction of HPV vaccination in Japan is cost-effective as in other countries. It is more cost-effective to increase the coverage of the Papanicolaou smear along with the universal administration of HPV vaccine.


Cervical cancer is the fifth leading cause of female cancer death in the world.1 The overall frequency of cervical cancer in Japan, including carcinoma in situ, was reported as 17 000 per year.2 In Japan, it is the third leading cause of cancer death among women <40 years of age.2 The age-adjusted mortality rate of cervical cancer in Japan has remained at almost the same level for the past two decades, although it has declined in the USA and UK.2–4

Screening with cervical cytology [i.e. Papanicolaou (Pap) smear] has been the key national strategy for early detection and treatment of cervical cancer to reduce its burden.5 However, the coverage of Pap smear screening in Japan remains between 10 and 20%,6 much lower than in other countries such as the UK (81%), France (54%) and the USA (>82%).7

Persistent human papillomavirus (HPV) infection, particularly with oncogenic types 16, 18, 52 and 58, is associated with a higher risk of incident cervical cancer precursor lesions.8 A prophylactic vaccine to prevent infection from HPV16 and 18 to reduce the burden of cervical cancer has been developed and implemented in some countries.9 The idea of introducing HPV vaccine in the Japanese population has evoked public debate and become a huge political issue. The bivalent vaccine was officially approved for use in Japan by the end of 2009, but there has been an on-going debate on whether the use of HPV vaccine should be underpinned by tax-payers’ money and if so, how much the government should spend.9

Despite its proven cost-effectiveness in other settings10–12 a simple extrapolation of the costs and effectiveness of HPV vaccine in countries other than the Japanese setting is not appropriate because of the differences in cervical cancer epidemiology and health systems. The prevalence of HPV types differs between geographic regions. In the case of squamous cell carcinoma, HPV16 was the predominant type (46–63%) followed by HPV18 (10–14%), 45 (2–8%), 31 (2–7%) and 33 (3–5%) in all regions except Asia, where HPV types 58 (6%) and 52 (4%) were more frequently identified.13 In Japan, HPV52 and HPV58 are most frequently found in squamous intraepithelial lesion following HPV16.14 A relatively lower prevalence of HPV16 and HPV18 in Japan has cast doubt on the effectiveness of the current HPV vaccine when compared with other countries.15

Only one study has evaluated the cost-effectiveness of HPV vaccination in the Japanese setting.16 However, the study did not compare strategies with a variable screening rate. Nor did it consider the effect of HPV type prevalence by age in Japan. Therefore, a cost-effectiveness analysis of screening coverage and vaccination, taking into account the age-specific prevalence by HPV type in the Japanese setting is urgently needed to inform and support policy decisions. Healthcare resources are limited; resources dedicated to screening and vaccination are no longer available for alternative healthcare uses and therefore the chosen strategy should represent a cost-effective use of scarce resources. The major objective of the present study is to assess the cost-effectiveness of universal vaccination against HPV in Japan from a societal perspective where the coverage of Pap smears is low and HPV oncogenic types are different from in other settings.


Natural history model of HPV infection

We developed a state-transition Markov model that simulates the natural history of HPV infection and carcinogenesis, in which transitions take place from one state to another at 1-month intervals (Figure 1). The model has 25 Markov states. The entry point into the model is girls aged 11 years with no previous exposure to HPV. We assumed that when girls/women enter the model, they start sexual activities, so acquiring a risk for HPV with the currently observed probabilities. In each cycle, they proceed to one of the four states: HPV16 and 18 DNA-positive group (HPV16 and 18), the other high-risk HPV DNA-positive group (other HR), the low-risk HPV DNA-positive group (LR), and the non-infected group (Normal) using monthly transition probabilities based on the systematic review of published literature.17

Figure 1.

 HPV natural history model.

Each group follows a natural history unless they are screened. When cervical intraepithelial neoplasia (CIN) 2, 3 or invasive cancer is identified by the screening, a treatment intervention conditional on cancer stage is implemented. After the treatment for an invasive cancer, a certain proportion of patients die whereas others survive according to the survival probabilities compiled from the cancer registry data.18,19 Age-dependent transition rates of disease progression from the susceptible to those infected with either HPV16/18, other HR (excluding HPV16 and 18) or LR were estimated from recent Japanese data.20 All individuals are followed up for 50 years until they reach 60 years of age, which is the average retirement age in Japan, unless they die earlier.

Vaccine efficacy

The vaccine efficacy was evaluated in eight randomised controlled trials.21,22 There was a substantial variation in follow-up periods and outcome measures among the studies. The World Health Organization adopted CIN2/3 as surrogate endpoints for cervical cancer in trials assessing vaccine efficacy.21 In our model we used the relative risk of HPV16 and 18 persistent infection risk, as the vaccine immunises against the contraction of HPV. The hypothesis that persistent infection with one of the 15 carcinogenic HPV types is the fundamental cause of cervical cancer is clearly supported by scientific evidence.23 We assumed relative risks of 0.12 (95% CI 0.03–0.48) for persistent HPV16 and 18 infection and 0.5 (range 0.3–0.7) for persistent HPV high-risk type excluding 16, 18.21,22,24,25 Additionally, we assumed 100% lifetime protection against HPV16 and 18 once fully vaccinated.

Intervention strategies

The bivalent vaccine was approved for use in women and girls who are over 10 years old. The Japan Society of Obstetrics and Gynaecology recommended administering HPV vaccine among girls from 11 to 14 years of age as a priority, partly because they are old enough to understand the meaning of the vaccination and partly because the vaccination in this age group is efficient and ensures early protection against HPV with high immunogenicity. Therefore, for strategies which include vaccination, all 11-year-old girls are vaccinated at the entry point into the model. We assumed that there is no exposure to any HPV types before the entry to the model. At 20 years of age, they start receiving screening every 2 years according to the current Japanese recommendations.5 Our reference strategy is the screening programme only with the current level of coverage (i.e. 20%).6,7Table 1 summarises six strategies that were analysed in the present study. The sensitivity of the Pap smear was assumed to be 94.7% as previously reported.26 The specificity (reported to be 98.9%) is not included in the model because screening will be repeated in false positives as determined by the cytology results.

Table 1.   Strategies
1. 20% coverage rate of screening with no vaccination
2. 50% coverage rate of screening with no vaccination
3. 80% coverage rate of screening with no vaccination
4. 20% coverage rate of screening with vaccination for all 11-year-old girls
5. 50% coverage rate of screening with vaccination for all 11-year-old girls
6. 80% coverage rate of screening with vaccination for all 11-year-old girls

Survival rates of women with cervical cancer

We used the data from life tables of Japanese vital registration to estimate the population-based mortality rates by age from cervical cancer and other competing risks.27 Cumulative nationwide survival rates by cancer stages of FIGO classification were not available in Japan. We adopted the data from the US SEER programme (Surveillance Epidemiology and End Results),18,19 which were calibrated using data from an existing Japanese regional cancer registry.28

Transition probabilities

Several natural history models of HPV have been developed and used in policy evaluations.29,30 Whereas a particular parameter has been common to several natural history models, there is a huge variation in the structure and parameters used in the previous models.29 We used age-dependent type-specific HPV prevalence data from Japanese women20 to derive transition probabilities from the susceptible to those infected with HPV16/18, other HR types and LR types.

Other model parameters were estimated from systematic literature reviews and then calibrated to the Japanese setting (Table 2).17 We simulated the model by using the transition rates of CIN2, 3 to the undetected stage I cancer of HPV16/18 and other HR depending on their age groups in Japan. Then we adjusted them by using the data of age-dependent incident rates of cervical cancer. We validated the model by goodness-of-fit statistics using age-dependent mortality rates of cervical cancer.

Table 2.   Transition rates
VariableBaseline valuesRange
  1. *Multiplied by age-dependent rate derived from calibration.

 Low-risk HPV0.02640.0245–0.0284
 High-risk 16, 18 HPV0.01500.0026–0.0274
 High-risk other HPV0.03760.0271–0.0480
HPV DNA to CIN2, 3
 Low-risk HPV0.000030.000003–0.00006
 High-risk 16, 18 HPV0.00120.000014–0.0024
 High-risk other HPV0.0000250.000002–0.00005
CIN1 to CIN2, 3
 Low risk HPV0.00030.00002–0.0005
 High-risk 16, 18 HPV0.00420.0001–0.0082
 High-risk other HPV0.00150.0001–0.0028
CIN2, 3 to undetected stage 1 cancer
 High-risk 16, 18 HPV0.0049*0.00001–0.0098
 High-risk other HPV0.0088*0.00004–0.0176
Progression rates in unscreened women with cancer
 Stage I to stage II0.0188 
 Stage II to stage III0.0250 
 Stage III to stage IV0.0375 
HPV DNA to Normal
 Low-risk HPV0.1951 
 High-risk 16, 18 HPV0.1951 
 High-risk other HPV0.1951 
CIN1 to Normal
 Low-risk HPV0.0854 
 High-risk 16, 18 HPV0.14060.1316–0.1497
 High-risk other HPV0.0430 
CIN2, 3 to Normal (70% of women)
 Low-risk HPV0.01450.0052–0.0238
 High-risk 16, 18 HPV0.00450.0010–0.0080
 High-risk other HPV0.00820.0029–0.0134
CIN2, 3 to HPV DNA or to CIN1 (15% of women each)
 Low-risk HPV0.00310.0011–0.0051
 High-risk 16, 18 HPV0.00100.0002–0.0017
 High-risk other HPV0.00180.0006–0.0029

Cost estimation

A societal perspective was adopted for this cost analysis. Cost estimates are presented in Table 3 that include programme costs and time costs. We approximated the programme costs by using the current national tariff used by the national health insurance scheme.31 These data were cross-validated by the cost of treatments and care for gynaecological patients at the University of Tokyo Hospital between August 2007 and November 2009. Both variable costs and doctor’s fees are included in the programme costs according to the fee schedule set by the national tariff.31 We estimated patients’ time cost by using the national average hourly wage of part-time workers from a national survey.32

Table 3.   Cost data
Costs involving patient’s time costsYen
Screening visit (Pap-test) per event7460
CIN1 detected patient per month4228
CIN2, 3 detection per event28,360
Conisation cost per case310,900
Treatment cost for stage I cancer case664,300
Treatment cost for stage II cancer per case2,869,900
Treatment cost for stage III cancer per case3,066,500
Treatment cost for stage IV cancer per case2,940,200
Average monthly wage for a Japanese case226,100
Vaccination cost (for three doses/visits)58,000

Cost-effectiveness analysis

We calculated quality-adjusted life-years (QALYs) from the model outputs on incidence, duration and mortality. The quality-of-life weights for different health states were based on those used in previous studies (Table 4).33–35 All costs and benefits were discounted at 3%, a frequently used rate for cost-effective analysis done in Japanese settings.36

Table 4.   Quality of life weights
VariableBaseline valuesRange
Quality of life weights for CIN
CIN2, 30.930.93–1.00
Quality of life weights for invasive cancer
Stage I0.650.49–0.81
Stage II0.560.42–0.70
Stage III0.560.42–0.70
Stage IV0.480.36–0.60
Quality of life weights after treatment for invasive cancer
Stage I0.970.73–0.99
Stage II0.90.68–0.98
Stage III0.90.68–0.98
Stage IV0.620.47–0.78

In line with a standard health economic evaluation, strategies are ranked in order of effectiveness after excluding dominated strategies.37 Incremental cost-effectiveness ratios (ICERs) are then calculated for each strategy relative to the next best alternative. The preferred strategy is the most effective strategy with an ICER within the willingness to pay threshold of 4.5 million yen. A commonly applied threshold for acceptable cost-effectiveness in the USA is $50,000;38 it is often used as a basis of cost-effective analysis in a Japanese setting.


Reduction in lifetime risk of cancer

Figure 2 shows the lifetime risk of cervical cancer by strategy estimated from a two-dimensional probabilistic sensitivity analysis. The range represents the minimum and maximum numbers of cervical cancer incidence per 100 000 population and its interquartile range (IQR). The bars represent the median value. Increasing the coverage of screening from the current level of 20–50 and 80% will substantially reduce the number of incident cervical cancer cases by 45.5% (IQR 42.0–48.7) and 63.1% (IQR 60.5–65.7), respectively. Combined strategies of 20, 50 and 80% screening coverage rate yields, respectively, a 66.1% (IQR 68.3–64.2), 80.9% (IQR 78.6–83.3) and 86.8% (IQR 85.4–87.9) reduction in cervical cancer incidence.

Figure 2.

 Lifetime risk of cancer for each strategy.

Total costs and QALYs of vaccination and screening programmes

Total QALYs gained per 100 000 population for each strategy showed slight increase as the screening coverage increases and the universal vaccination is added (Figure 3). Figure 4 shows cost per person for each strategy. The squares represent average values and the range represents average value ± 2 SD. Costs of strategies including vaccination are approximately four times higher than that of strategies without vaccination. Increasing the screening coverage rate was cheaper than introducing vaccination for all 11-year-old girls.

Figure 3.

 Total QALYs per 100,000 people for each strategy.

Figure 4.

 Cost per person for each strategy.

Incremental cost-effectiveness ratio

Table 5 shows the ICER of each strategy compared with its next best alternative strategy. Using the default model values, 50% screening coverage with a vaccination strategy was the most cost-effective when using a willingness to pay for a QALY threshold of 4,500,000 yen (≅ US$500,000) (Figure 5).

Table 5.   Cost effectiveness of alternative screening and vaccination strategies
StrategyIncremental cost effectiveness Ratio* (Yen/QALY)
  1. *Ratio of additional costs and benefits of a particular strategy compared with the previous strategy.

20% Screening– Dominated
50% Screening658
80% Screening571 015
20% Screening + vaccination– Extended Dominance
50% Screening + vaccination2 920 636
80% Screening + vaccination8 568 182 not cost effective
Figure 5.

 Cost and QALYs per 1000 people.

Sensitivity analysis on vaccine efficacy

We performed a sensitivity analysis on vaccine efficacy. The vaccine efficacy is determined by the combination of risk ratios to acquire HPV16/18 and other HR in our model. Table 6 shows cost and QALYs derived from the reference vaccine efficacy, minimum and maximum vaccine efficacy per 1000 people. Differences in vaccine efficacy would result in the differences in programme costs ranging from approximately 4,000,000–8,000,000 yen (≅US$480,000–960,000).

Table 6.   Costs and QALYs per 1000 people of varied vaccine effect
StrategyMinimum vaccine effect*Baseline vaccine effect*Maximum vaccine effect*
Cost (¥)QALYsCost (¥)QALYsCost (¥)QALYs
  1. *Minimum vaccine effect means relative risks of 0.48 for persistent HPV16 and 18 infection and 0.7 for persistent HPV high-risk type excluding 16, 18 infection. Baseline vaccine effect means relative risks of 0.12 for persistent HPV16 and 18 infection and 0.5 for persistent HPV high-risk type excluding 16, 18 infection. Maximum vaccine effect means relative risks of 0.03 for persistent HPV16 and 18 infection and 0.3 for persistent HPV high-risk type excluding 16, 18 infection.

Screening 20% + vaccination69,561,00025 933.8866,114,00025 942.6462,628,00025 950.77
Screening 50% + vaccination70,300,00025 937.3367,334,00025 945.5464,300,00025 953.22
Screening 80% + vaccination72,129,00025 940.8169,219,00025 945.7666,277,00025 953.07

Table 7 shows the ICERs derived from the sensitivity analysis. The current strategy is dominated by strategies with a higher screening rate. A screening rate of 20% with a vaccination strategy is always ruled out because of extended dominance. The ICER for a screening rate of 50 and 80% with vaccination strategies was sensitive to the differences in incremental costs and effectiveness given by the result of a two-dimensional probabilistic sensitivity analysis of the model with each vaccine efficacy. With the minimum efficacy, a combined strategy of 80% screening and universal vaccination is most cost-effective. On the other hand, with the maximum and baseline vaccine efficacy, a combined strategy of 50% screening and universal vaccination remains most cost-effective.

Table 7.   Sensitivity analysis on vaccine effect (ICER)*
StrategyMinimum vaccine effect**Baseline vaccine effect**Maximum vaccine effect**
  1. *Incremental cost effectiveness ratio (Yen/QALY).

  2. **Minimum vaccine effect means relative risks of 0.48 for persistent HPV16 and 18 infection and 0.7 for persistent HPV high-risk type excluding 16, 18 infection. Baseline vaccine effect means relative risks of 0.12 for persistent HPV16 and 18 infection and 0.5 for persistent HPV high-risk type excluding HPV16, 18 infection. Maximum vaccine effect means relative risks of 0.03 for persistent HPV16 and 18 infection and 0.3 for persistent HPV high-risk type excluding 16, 18 infection.

Screening 20%DominatedDominatedDominated
Screening 50%658658658
Screening 80%Extended dominance571 015571 015
Screening 20% + vaccinationExtended dominanceExtended dominanceExtended dominance
Screening 50% + vaccinationExtended dominance2 920 6361 874 867
Screening 80% + vaccination3 745 4428 568 182Dominated


The introduction of HPV vaccine to the Japanese population has been controversial because the coverage of Pap smear screening is low and the prevalence of HPV types is different from that observed in Western countries.

To date there has been only one study that has assessed the impact of introducing HPV vaccine in Japan.16 However, this study suffered from several major limitations. It did not distinguish health status related to HPV type 16 and 18 from other high-risk types. We modelled the natural history of each HPV type status; HPV16/18, other HR, and LR. We used different vaccine efficacies depending on the HPV types with a range that was derived from a meta-analysis of the available evidence. The previous study also did not include strategies of varied screening rates without vaccination. The authors analysed the effect of screening atthe currently observed levels ranging from 13.6 to 24.7%, and so the impact of increasing Pap smear coverage was not considered. Instead, the present study compared the strategies of varied screening rates ranging from 20 to 80%.

Our analysis suggests that increasing cervical cancer screening coverage to 50% would halve the incidence of cervical cancer and save programme costs and that the introduction of HPV vaccination would reduce the incidence by two-thirds but result in a four-fold increase in programme costs. Using the model’s default values, a combined strategy to expand the coverage of cancer screening up to 50% and the introduction of universal vaccination would be most cost-effective. The results are robust with sensitivity analysis in which the optimum coverage level most likely lies somewhere between 50 and 80%. Our result confirms the need for expanding coverage for Pap smears in Japan as previously suggested,39 to maximise the impact of the cervical cancer strategy regardless of whether a national vaccine programme is also implemented.

The detection rate of HPV16 and 18 among women with cervical cancer in Japan is reported to be lower than that in other countries.14 We used the latest age-dependent prevalence data, which consistently show that the younger population has a higher detection rate of HPV16 and 18 than the older population.20 The prevention of cervical cancer in a young person shows larger QALYs gained than that of an older person because of the longer remaining life expectancy. Hence the effect of vaccine on cancer incidence or QALYs is not as low as might otherwise be expected.

The present study has several limitations. First, we assumed life-long lasting immunity acquired by the vaccine. The vaccine has only been recently introduced, and the latest evidence shows 7.3 years of efficacy and immunogenicity of the vaccine, which was derived from the population of the initial placebo-controlled study.40 If additional vaccination is required to maintain immunity in the future, then programme costs are slightly underestimated. Second, there is no population-based survival data of women with cervical cancer by stages of FIGO. These data are essential when building a model. However, we managed to adopt and validate data from an existing Japanese regional cancer registry. Third, we did not incorporate the preferences of girls and their parents and the subsequent uptake of vaccine as a result of their preferences. Both effects and costs may be overestimated in that sense. Finally, we did not include the cost for campaigns to increase the coverage of screening and/or vaccination in this analysis, which may underestimate the programme costs but such a bias is minimal given the fact that the majority of costs is incurred by screening, vaccination and treatment interventions.

Vaccination for HPV is attracting considerable policy attention now as a strategy for cervical cancer prevention in Japan. Our analysis showed that increasing the rate of the current screening strategy would halve cancer incidence with a similar cost to the current screening strategy, though vaccination strategies may also be cost effective. We suggest further efforts to expand the current screening programme regardless of what support is provided for vaccination.

Some of the reasons why Pap smear coverage is so low in Japan relate to a lack of knowledge and from the fact that the financial support of the screening programme from the Ministry of Health, Labour and Welfare was discontinued because it was included in the general ones in 1998. Most cities, towns and villages decided to reduce the cost for the screening programme.41–43 Free tickets for the Pap smear were provided under supplemental budgets for 2009. Distributing free tickets to a target population of certain ages showed a significant increase in the coverage rate by 2.8 times.44 We need to continue endeavours to increase coverage by effective interventions such as providing free tickets and undertaking awareness campaigns. The involvement of gynaecologists in school education will also support the enhancement of knowledge about cervical cancer prevention and help to increase the coverage rate of screening as has been seen in other countries.45,46

Our analysis showed that introducing the HPV vaccination for all 11-year-old girls would reduce cervical cancer incidence to 33.9% with a net cost of only 49,000 yen per person (taking into account the social burden of cancer). Vaccinating all 11-year-old girls would cost 33.7 billion yen. Our analysis showed the cost-effectiveness of vaccination and that it would save future costs. It is important to give priority to policy which is evidence based medically and economically. If the prevalence of HPV infection is reduced as a result of universal vaccination, as our model predicts, then it may be possible to extend the interval between routine screens or to increase the age at which screening is first offered, as suggested in other cost-effectiveness studies.34,47 The use of the HPV-DNA test in the screening programme is one choice that should be evaluated in the future.

In conclusion, the introduction of HPV vaccine in Japan is cost-effective as in other countries. It is more cost-effective to increase the coverage of the Pap smear along with the universal administration of HPV vaccine. Only by doing so, can the scarce healthcare resources be efficiently and effectively used to reduce the burden from cervical cancer in Japan.

Disclosure of interests

None of the authors have any conflicts of interest to declare.

Contribution to authorship

NY contributed to the study design of the current paper, model construction, data acquisition, data analysis and interpretation, drafting and revising the manuscript. RM contributed to the study design of the current paper, model construction, results interpretation and revising the manuscript. PJ contributed to the model construction, results interpretation and the critical review of the manuscript. YO contributed to the study design of the current paper. KK contributed to the model construction, data acquisition and interpretation of the results. KS and YT contributed to the study design of the current paper and interpretation of the results. All authors approved the final version of the manuscript.


This work was funded by two grants; Health and Labour Sciences Research Grants from the Japanese Ministry of Health, Labour and Welfare and a Research accomplishment cooperation system grant for doctoral courses at University of Tokyo.


We would like to thank Mitsuhiro Sado and Makoto Kobayashi for their technical advice on model building, and Koji Matsumoto for providing data on type-specific HPV prevalence.