Cost-effectiveness analysis of human papillomavirus DNA testing and Pap smear for cervical cancer screening in a publicly financed health-care system

Objective: To evaluate the long-term cost-effectiveness of different strategies for human papillomavirus (HPV) DNA testing combined with Pap smear for cervical cancer screening in Taiwan. Methods: This study adopts a perspective of Department of Health in cost-effectiveness analysis to compare a no-screening strategy with nine different screening strategies. These strategies comprise three screening tools (Pap smear alone, HPV DNA testing followed by Pap smear triage, and HPV DNA testing combined with Pap smear), and three screening intervals (annually, every 3 years, and every 5 years). Outcomes are life expectancy, quality-adjusted life years (QALYs), lifetime costs, and incremental cost-effectiveness ratios (ICERs). Probabilistic sensitivity analyses (PSAs) were conducted to assess parameter uncertainty. Results: When three times gross domestic product per capita is used as the decision threshold, all nine screening strategies were cost-effective compared with the no-screening strategy. Compared with the current screening strategy (an annual Pap smear), HPV DNA testing followed by Pap smear triage every 5 years and every 3 years were cost-effective. Results of PSA also indicated that a HPV DNA testing followed by Pap smear triage every 5 or every 3 years achieved the highest expected net benefits. Conclusions: Possible economic advantages are associated with extending the cervical cancer screening interval from one Pap smear annually to HPV DNA testing followed by Pap smear triage every 5 years with an ICER $1 247 000 per QALY gained, especially in a country with a publicly financed health-care system.

Cervical cancer is one of the most common cancers in Taiwan, currently ranked as the fifth leading cancer causing death for females (Chen et al, 2009). Although the incidence and mortality associated with cervical cancer have declined steadily since the National Health Insurance (NHI) program was instituted in 1995, the incidence of cervical cancer in Taiwan is still one of the highest in Asia and is a huge burden to the health-care system.
All females in Taiwan aged X30 years are now provided with a free Pap smear annually by the NHI system (Koong et al, 2006). The annual Pap smear screening rate for females aged 30 -69 years increased to 27.4% in 2007 from 14.3% in 1996. The 3-year coverage rate reached 52.4% during 2005 -2007; 5-year coverage was roughly 60% during -2007(Bureau of Health Promotion, 2007. In France, 60% of the female population had a regular Pap smear within the last 3 years (Schaffer et al, 2000), 80% of Dutch females and 85% of English females had a Pap smear within the last 5 years (Patnick, 2000;Van Ballegooijen and Hermens, 2000). Notably, the screening program in Taiwan has not yet achieved satisfactory coverage. The health authorities in Taiwan currently face the challenge of implementing cervical cancer screening programs with increased effectiveness.
A population-based randomised trial conducted in Sweden indicated that primary the human papillomavirus (HPV) DNA testing with Pap smear triage and repeat HPV DNA testing for those who test positive in HPV DNA testing with negative Pap smear results were the most effective cervical cancer screening strategy (Naucler et al, 2009). Recent findings from a cohort study on long-term outcomes of HPV infection in Taiwan demonstrated that, for a female who had a normal Pap smear result, the 6-year cumulative risk of developing high-grade squamous intraepithelial lesions (HSILs) was 10.0% when the HPV DNA testing result was positive, whereas the risk for those with negative HPV DNA testing was only 0.34%. These data support the efficacy of HPV DNA testing for cervical cancer screening program with intervals of at least 5 years (Huang et al, 2008). Another European study of six countries also suggested that screening intervals could safely be lengthened to 6 years among females with a negative HPV DNA testing result .
In two systematic review studies, Holmes et al (2005) and Muhlberger et al (2008) summarised existing evidence for the costeffectiveness of incorporating the HPV DNA testing into current cervical cancer screening programs. Their analytical results showed the HPV DNA testing with a Pap smear as a screening strategy for females aged X30 years was cost-effective. With an incremental cost-effectiveness ratio (ICER) threshold of h50 000 per quality-adjusted life year (QALY), Muhlberger et al (2008) concluded that, in comparison with annual Pap smear, or Pap smear every 2, 3, or 5 years, the probabilities of HPV DNA testing strategies being cost-effective were 83, 25, 55, and 92%, respectively. These analytical findings illustrate that when evaluating the cost-effectiveness of the HPV DNA testing, the screening interval should be considered.
Several studies utilised computer-based models to estimate the cost-effectiveness of adding the HPV DNA testing to current cervical cancer screening programs in Sweden, Canada, the United States, the United Kingdom, the Netherlands, France, Italy, and Germany. Adding the HPV DNA testing can reduce cancer incidence and is less costly than Pap smear alone (Mandelblatt et al, 2002b;Mittendorf et al, 2003;Goldie et al, 2004;Kim et al, 2005;Bistoletti et al, 2008;Chuck, 2009).
In Taiwan, which has a poor Pap smear participation rate, policy makers face the problems of whether to combine the HPV DNA testing with Pap smear for cervical cancer screening and determining the optimal screening interval. This study evaluates the long-term cost-effectiveness of different strategies for HPV DNA testing combined with Pap smear for cervical cancer screening in Taiwan for healthy females aged 30 years.

Cost-effectiveness analysis and decision analytic model
This cost-effectiveness analysis (CEA) adopts the perspective of Department of Health. A decision analytic model was developed using TreeAge Pro Suite 2009 software (TreeAge Software, Inc., Williamstown, MA, USA) for evaluating the long-term costs and effectiveness of various cervical cancer screening strategies. This study used clinical outcome measures, including cervical cancer incidence, mortality, life expectancy, and QALYs. All costs, life expectancy, and QALYs were discounted to their present values at an annual rate of 3%. The ICERs were calculated as the difference in cost divided by the difference in QALYs for each strategy compared with the no-screening strategy, with the strategy of an annual Pap smear, or with the next-best strategy. Currently, no consensus exists for a specific cost-effectiveness threshold in Taiwan. However, this study considered strategies with ICER values less than three times per capita gross domestic product (GDP) as cost-effective (Eichler et al, 2004). In 2008, Taiwan's per capita GDP was NT$538 790 (Directorate-General of Budget, Accounting and Statistics, 2009a). Therefore, the cost-effectiveness threshold assumed in this model was NT$1 620 000/QALY.

Natural history model
The natural history model for cervical cancer was adopted from literature (Debicki et al, 2008;Rogoza et al, 2008;Suárez et al, 2008). The health states of the decision analytic model for the natural history of cervical cancer among females have the following seven categories: normal (healthy); HPV infection; grades of cervical intraepithelial neoplasia (CIN) (CIN I, CIN II or III); persistent CIN II or III; stage of cervical cancer; death from other causes; and, death from cervical cancer. This model simulates a cohort of 30-year-old healthy females; the hypothetical cohort moves between different health states according to prespecified probabilities in each Markov cycle (1 year). The analytical time horizon was lifetime (85 years old).
The model starts with a cohort of 30-year-old healthy females (normal state). During simulation, individuals may remain in the normal state or progress to the HPV infection state in the next year (or cycle). Individuals with HPV infection may remain in the HPV infection state, regress to the normal state, or progress towards the CIN I state. The CIN I patients may remain in the CIN I state, regress to the normal state, or progress towards the CIN II or III state. The CIN II or III patients may remain in the CIN II or III state, regress to the normal state, or progress towards the persistent CIN II or III state. From there, patients may remain in the persistent CIN II or III state, regress to the normal state, or progress towards the cervical cancer state. When individuals reach the cervical cancer state, we assume no regression is possible; that is, they can only progress to a cured cancer state when alive at 5 years after diagnosis or to the cervical cancer death state. In this model, a female in any state can die because of other causes.

Screening strategies
This study was designed to compare a no-screening strategy with nine different screening strategies, which varied using three different screening algorithms -Pap smear alone, HPV DNA testing followed by Pap smear triage, and HPV DNA testing combined with Pap smear, with three screening intervalsannually, every 3 years, and every 5 years. We assume the ceiling age for screening is 69 years. Table 1 lists the primary screening and triage tool, screening ages, screening interval, and coverage for each strategy. We assume a 30% coverage rate for annual screening, based on the reported 28.4% of females aged 30 -69 years who had a Pap smear in 2007. Similarly, 3-year (60%) and 5-year coverage rates (75%) were assumed as the reported percentages of 52.5% and 64.9% of those who underwent at least one Pap smear during a 3-or 5-year period, respectively.
When a female is invited to undergo cervical cancer screening, she is stratified into the never-screened or regularly screened categories ( Figure 1). According to research by Chen et al (2009), after implementing the screening program in 1995, roughly 75% eligible females in Taiwan participated in the program at least once by 2007. We infer that the percentage of those never screened was 25%. Considering the uncertainty, 12% and 30% as low and high estimate values were changed during sensitivity analysis.
In the regular screening strata, a female may or may not be willing to participate in an appropriate year. If a female chooses to undergo screening, the following three combinations of screening tools were evaluated: (1) Pap smear as the primary screening test followed by colposcopy-guided biopsy for positive Pap smear results (i.e., Pap); (2) HPV DNA testing as the primary screening test, using a Pap smear as a triage strategy for those who test positive for high-risk HPV types in HPV DNA testing (i.e., HPV-Pap); and, (3) HPV DNA testing in combination with Pap smear as the primary screening test, using a Pap smear as a follow-up screening strategy for those who have positive Pap smear results or test positive for high-risk HPV types (i.e., combination).
Experts and existing guidelines were consulted before making the following assumptions for management of females with different screening results: (1) women with negative primary screening results are asked to undergo regular screening; (2) when a woman had a positive final screening result, she is urged to undergo a colposcopyguided biopsy. All women with a confirmed histology of CIN I, CIN II or III and cancer are assumed to be treated with the therapy recommended by the Taiwan Cooperative Oncology Group (Taiwan Cooperative Oncology Group, 2007). These patients are assumed treated immediately after disease confirmation. When treatment was completed, annual Pap smear screening followed; (3) in the HPV-Pap and combination strategies, those negative after follow-up screenings are suggested to undergo annual Pap smear screening for the rest of their lifetime.
We further assume that compliance with follow-up and treatment is 100% in the model. Table 2 summarises the probabilities, costs, and data sources used in the base-case analysis.

Model parameters
Parameters of the natural history model Parameters of HPV infection rates and disease transitions were obtained from the study by Debicki et al (2008). They estimated disease parameters by reviewing published reports of Taiwanese screening coverage and cancer incidence, and further calibrated disease parameters to fit cervical endpoints. The succinct model, which adopted fewer complex processes and a short cycle length of 1 year, was used in this study.
Survival rate and death probability The annual probabilities of dying from cervical cancer for patients were obtained from survival rates published by the Taiwan Cancer Registry (2007). The 1-, 2-, 3-, 4-, and 5-year survival for patients with invasive cervical cancer (all stages) in Taiwan    Utility weight data The health utility data for the base case and ranges were adopted from a CEA evaluating HPV vaccination programs in Taiwan (Suárez et al, 2008), which was compiled using health utility data collected in the USA (Insinga et al, 2007).
Costs data From a perspective of Department of Health, only costs associated with direct medical costs were utilised.
Furthermore, this study included only the direct medical costs associated with use of health-care services covered by the NHI program.
The category for costs associated with cervical cancer consists of costs of screening tests, costs of diagnosis, costs of treatment for CIN I, CIN II or III, and cancer treatment costs.

Sensitivity analysis
Considering uncertainty is crucial in a decision analytical study.
To better understand the impact on the results for different levels of participation, along with the purpose to present the validity of the model results, a scenario sensitivity analysis was performed by approximating the highest coverage rate at 100% for all screening intervals. Probabilistic sensitivity analyses (PSAs) were conducted to assess parameter uncertainty. The ranges of parameters were mostly from published studies (Table 2). A 50% increase or decrease in the base-case value for costs data were changed during sensitivity analysis. Probability distributions were defined for the following six sets of model parameters: (1) screening rate; (2) disease transition probabilities in the natural history model; (3) survival rate for cervical cancer; (4) test characteristics; (5) utility; and, (6) costs. A beta distribution was fitted for screening rate, the natural history model, survival rate for cervical cancer, test characteristics, and utility, those parameters were between 0 and 1. A gamma distribution can be used to represent uncertainty in cost data, which was constrained to be non-negative and characterised in the interval of 0 to positive infinity (Briggs et al, 2007).
The PSA results in this model were analysed using costeffectiveness acceptability frontier (CEAF) constructed from cost-effectiveness acceptability curves (CEACs) for the different strategies. The CEACs were calculated from net monetary benefit over a range of willingness-to-pay (WTP) thresholds for each strategy, and produced by plotting the probability of being costeffective for a strategy under different WTP thresholds (Barton et al, 2008). To further determine the uncertainty associated with a priori intervention over all WTP thresholds for each strategy (Fenwick et al, 2001), this study produced a CEAF by identifying the probability of the optimal strategy in each of 1000 iterations based on the strategy with the highest expected net benefits at different WTP thresholds (Barton et al, 2008). Table 3 presents the lifetime effectiveness and costs for a hypothetical cohort of 30-year-old females for all nine screening strategies. Predicted effectiveness outcomes were lifetime incidence and mortality for cervical cancer, total life expectancy, and total QALYs; additionally, total lifetime costs, and ICERs calculated by comparing different bases (i.e., the no-screening strategy, Pap annual strategy, and next-best strategy) are presented.

Reduction in lifetime risk
Compared with the no-screening strategy, all nine screening strategies were more effective ( Table 3). The estimated reduction in lifetime incidence and related mortality for cervical cancer when screened at 30 years of age varies at 35.8 -71.1%, depending on screening tests, screening intervals, and screening strategies. The HPV-Pap and combination strategies at annual, 3-, or 5-year intervals prevented more cases than the Pap strategy at annual, 3-, or 5-year screening intervals. The model estimated cervical cancer cases at 87.6 -118.2 per 100 000 women and a related mortality of 25.6 -34.8 per 100 000 patients with the Pap strategy; 54.6 -70.0 cases and 16.3 -20.7 deaths for the HPV-Pap strategy; and 53.2 -65.2 cases and 15.9 -19.3 deaths with the combination strategy (Table 3). Figure 2 presents the cost-effectiveness results that indicate that all nine screening strategies Abbreviations: ICER ¼ incremental cost-effectiveness ratio; QALY ¼ quality-adjusted life year. a Pap refers to Pap smear as the primary screening test followed by colposcopyguided biopsy for positive Pap smear results. HPV-Pap refers to HPV DNA testing as the primary screening test, using a Pap smear as a triage strategy for those who test positive for high-risk HPV types in HPV DNA testing. Combination refers to HPV DNA testing in combination with Pap smear as the primary screening test, using a Pap smear as a followup screening strategy for those who have positive Pap smear results or test positive for high-risk HPV types. 1-y, 3-y, and 5-y refer to three screening intervals of annually, every 3 years, and every 5 years for women 30 -69 years of age, respectively. b Life expectancy, QALYs, and costs are discounted at 3% per annum. c Cost-effectiveness ratio calculated as the difference in cost divided by the difference in QALYs for each strategy compared with the no screening strategy. d Cost-effectiveness ratio calculated as the difference in cost divided by the difference in QALYs for each strategy compared with next best strategy. e Cost-effectiveness ratio calculated as the difference in cost divided by the difference in QALYs for each strategy compared with annual Pap smear alone strategy. generated greater effectiveness (i.e., total QALYs) and total costs than the no-screening strategy from the perspective of Department of Health. Using three times Taiwan's per capita GDP as the costeffectiveness threshold, all these strategies were cost-effective. However, when each strategy compared with the next mosteffective non-dominated strategy, an annual Pap, combination every 5 years and every 3 years were extended-dominated. Pap every 5 years and every 3 years, HPV-Pap every 5 years and every 3 years had ICERs of NT$660 000, NT$889 000, NT$1 323 000, NT$1 358 000 per QALY gained, respectively. These strategies were cost-effective at three times per capita GDP threshold. If the current screening strategy, an annual Pap, is used for comparison, HPV-Pap every 3 years and every 5 years cost NT$1 303 000 and NT$1 247 000 per QALY gained, respectively. These strategies are oNT$1 620 000/QALY. However, extending screening intervals for Pap annually to every 3 or 5 years was less costly but also less effective. Changing Pap annually to every 3 or 5 years save NT$1197 to NT$1890 while loss 0.33 -0.62 of discounted quality-adjusted life days per woman (see Table 3 and Figure 2).

Sensitivity analysis
The coverage rate changes into 100% for all screening intervals in the scenario sensitivity analysis. Results show that screening strategies on the sensitivity analysis efficiency frontier in this analysis are Pap every 5 years and 3 years, HPV-Pap every 5 years, every 3 years and annually, and combination annually (Figure 3). The ranking, and the magnitude of the discounted QALYs and costs, of the alternatives are as expected (i.e., programs with the annual screening are the most efficacious and the most costly), suggesting the validity of this model.

DISCUSSION
Screening has proven cost-effective in preventing cervical cancer according to many studies conducted in different countries (Mandelblatt et al, 2002a, b;Mittendorf et al, 2003;Goldie et al, 2004Goldie et al, , 2005Kim et al, 2004Kim et al, , 2005Holmes et al, 2005;Siebert et al, 2006;Anderson et al, 2008;Andres-Gamboa et al, 2008;Bistoletti et al, 2008;Muhlberger et al, 2008;Chuck, 2009;Vijayaraghavan et al, 2009); the model results in this study demonstrate the costeffectiveness of screening compared with that of a no-screening strategy, regardless of which screening tool or interval is used. Compared with the annual Pap strategy, the HPV-Pap strategy at 3-or 5-year intervals are cost-effective. Annual HPV-Pap strategy, combination strategy annually or at 3-or 5-year intervals generated the greatest effectiveness, while cost a substantial amount of total costs; thus, the annual HPV-Pap and the combination strategies are not cost-effective. For a nation with a publicly financed health-care system, such as Taiwan, this study shows that extending cervical cancer screening interval from Pap annually to HPV-Pap strategy every 3 or 5 years costs NT$1 303 000 or NT$1 247 000 per QALY gained, respectively.
Compared with the no-screening strategy, Pap annually and Pap every 3 and 5 years reduced the incidence of cervical cancer in a lifetime by 52.4, 44.1, and 35.8% under the coverage assumptions in this model. Arbyn et al (2006) and Mayrand et al (2007) indicated that the HPV DNA testing has greater sensitivity and is more accurate than Pap smear for detecting HSILs. As expected, analytical results demonstrate that the HPV-Pap or combination strategy annually or at 3-or 5-year screening intervals prevented more cases than the Pap strategy annually or at 3-or 5-year screening intervals. The reductions in lifetime incidence of cervical cancer with the HPV-Pap strategy annually or at 3-or 5-year screening intervals were 70.3, 67.4, and 62.0%; the values for the combination strategy were 71.1% (annually), 69.0% (every 3 years), and 64.6% (every 5 years). Two studies undertaken in the United States reported a decline in lifetime incidence of cervical cancer of 70 -87% (Mandelblatt et al, 2002b) and 81 -90%  for Pap smear alone with different screening intervals. The HPV DNA testing with a Pap smear in a cervical cancer screening program decreased risk by 78 -93% (Mandelblatt et al,   Figure 2 Base-case efficiency frontier depicting costs and QALYs for cervical cancer screening strategies. The 1-year (30%), 3-year (60%), and 5-year coverage rates (75%) were assumed those who underwent at least one Pap smear during a 1-, 3-, or 5-year period.
2002b) and 90 -93% . This study identified an increasing trend of a reduction in the incidence of cervical cancer that is consistent with findings obtained with other models comparing the HPV DNA testing with Pap smear alone. Several studies demonstrated that using the HPV DNA testing as primary screening or combined with Pap smear with a screening interval of every 2, 3, or 5 years was more cost-effective than Pap smear alone (Mandelblatt et al, 2002b;Mittendorf et al, 2003;Goldie et al, 2004;Kim et al, 2005;Bistoletti et al, 2008;Chuck, 2009). Analytical results in this study demonstrate that the HPV-Pap strategy at 3-or 5-year intervals was cost-effective; this finding is generally consistent with those studies. However, the results of this model demonstrate that the combination strategy, regardless of screening intervals, were not cost-effective compared with current cervical cancer screening program (Pap annually). Three important factors were considered in this study. First, the combination strategy annually and at 3-and 5-year intervals were more effective (i.e., total QALYs) than Pap annually. The combination annually strategy saves 0.69 days of discounted QALY per woman compared with Pap annually. Whereas total costs of the combination strategy is more than 1.4-fold (combination every 5 years) to 2.0-fold (combination annually) that of the annual Pap strategy (NT$4242) (Table 3). Thus, extending the cervical cancer screening interval from Pap annually to HPV-Pap strategy every 3 or 5 years can be considered.
Second, differences in cost structures may account the difference between cost-effectiveness results in this study and those obtained by other studies using the HPV DNA testing as primary screening or combined with Pap smear for cervical cancer screening. The unit cost of the HPV DNA testing is 40% more than that of a Pap smear in Taiwan. Conversely, the unit cost of the HPV DNA testing is roughly identical to that of Pap smear in many European countries and in the United States (Mandelblatt et al, 2002b;Kim et al, 2004Kim et al, , 2005Bistoletti et al, 2008). In comparison, the  180 000 360 000 540 000 720 000 900 000 1080 000 1260 000 1440 000 1620 000 1800 000 Figure 4 Cost-effectiveness acceptability frontier curves.
CEA of HPV DNA testing and Pap smear in Taiwan I H-I Chow et al screening cost of the HPV-Pap and combination strategies are higher in Taiwan than in other countries. Third, using the HPV DNA testing increased disease cases by detecting more precancerous lesions than Pap smear alone (Naucler et al, 2009). The costs of treatment with the HPV-Pap or combination strategy may be lower than Pap smear alone theoretically. In fact, the costs of treatments for precancerous lesions and cervical cancer in Taiwan are lower than those in many countries Kim et al, 2004Kim et al, , 2005Tang et al, 2010). For instance, treatment cost was roughly NT$6956 for CIN II or III patients. These figures are more than three-fold in the UK (US$678, about NT$22 463) and seven-fold in the United States (US$2 833, about NT$93 863). Moreover, overall invasive cervical cancer treatment costs are 1.6-fold higher in the Netherlands (US$17 603, about NT$583 223) to 2.8-fold higher in the UK (US$31 494, about NT$1 043 443) than in Taiwan (NT$375 322). The difference in treatment cost between the HPV-Pap or combination strategy and Pap smear alone is less in Taiwan than that in other countries. Consequently, using the HPV DNA testing has the potential to improve health effectiveness at a reasonable cost compared with the Pap smear in European countries and in the United States is not surprising. Early detection of precancerous lesions may not reduce treatment costs largely; thus, it may have difficulty being cost-effective when treatment costs are generally lower in Taiwan than in other countries.
Analytical results demonstrated that extending screening intervals from Pap annually to every 3 or 5 years would save costs but be less effective; this finding is generally consistent with those in studies by Anderson et al (2008) and Siebert et al (2006). The population of women aged X30 years, that is, those targeted for cervical cancer screening, was estimated at approximately 7 million in Taiwan in 2008 (Directorate-General of Budget, Accounting and Statistics, 2009b). However, the budget for the screening program (free annual Pap smear-approximately NT$430 for one Pap smear) from the Bureau of Health Promotion, Department of Health, was only NT$860 million annually (Liberty Times, 2009). Although this figure seems quite small, it hardly meets the expenses under low participation. As a result of the limited resources available for cervical cancer screening programs, implementing an annual screening program is barely affordable. The cost-effectiveness results obtained by this study suggest that HPV-Pap every 3 or 5 years represent an optimal use of resources, considering improve participation rate and extending screening intervals.
Some studies have explored factors associated with low participation for the cervical cancer screening program in Taiwan (Lee et al, 1997;Lin et al, 2003Lin et al, , 2007. These factors may be summarised as Taiwanese women having a more lax attitude toward cervical cancer screening programs, perhaps because of societal values or those in traditional culture. This is also at play in the difficulty policy-makers have faced during the last decade in determining how to improve coverage for Pap smear screening. One alternative to screening may be self-sampling, which may make cervical cancer screening more accessible to women not participating in the current screening program or those rarely contacting health-care centres, offering an important method for increasing cervical cancer screening coverage (Sellors et al, 2000;Wright et al, 2000;Agorastos et al, 2005;De Alba et al, 2008).
Future studies should evaluate the cost-effectiveness of selfsampling for HPV DNA testing.
This study has several limitations. First, women treated for CIN I, CIN II or III may have a different (i.e., higher or lower) probability of recurrent cervical lesions than those who have negative test results. However, we did not have good evidence on this difference. Thus, we assumed that women treated for CIN I, CIN II or III were cured and returned to a normal state and acquired a new HPV infection at similar rates to those without previous abnormal results. The model assumed that women must have been infected with HPV to develop further lesions or diseases. This study did not consider disease progression without previous HPV infection.
Second, while the natural history model was based on Taiwan calibrated data sources from studies, most transition probabilities were obtained from literature with no age stratification for all Pap, HPV-Pap, and combination strategies. The model also used the same cervical cancer survival rate for all ages, which is not the case in reality. This may add to model uncertainties and parameter uncertainties in analytical results; however, sensitivity analyses demonstrate the robustness of study results.
Third, this study did not consider women age o30 years who have already infected with high-risk HPV. According to a cohort study in Taiwan, the prevalence of HPV infection was 9.09% in women age o30 years (Huang et al, 2008). For women with HPV infection, they are recommended to undergo a Pap smear as a triage screening. If necessary, further tests or treatments will be required. Thus, the total costs estimates in this study may have been underestimated; however, the magnitude is not likely to be high.
Finally, this study also assumed that all women with negative primary screening results were asked to undergo regular screening until 69 years of age. Although the recent scientific evidence suggests that women aged 50 or above with negative HPV test results will have a lower risk to develop cervical cancer, it is the general consensus in Taiwan that aged women should also be covered in screening programs. It will be of value, however, to explore the cost-effectiveness of polices that considered the strategies for different age groups.
In conclusion, this study demonstrates that, compared with a no-screening or an annual Pap strategy, using the HPV DNA testing in the cervical screening program in Taiwan is costeffective. This study also identified as cost-effective for extending the cervical cancer screening interval from Pap annually to HPV-Pap strategy every 3 or 5 years had ICERs of NT$1 303 000 or NT$1 247 000 per QALY gained, respectively. Considering the potential economic advantages, HPV-Pap strategy every 5 years would seem to be more attractive, especially in Taiwan, a country with a publicly financed health-care system.