The target population for this analysis was all adolescent girls in the United States. Our base-case analysis considered a hypothetical cohort of 12-year-old girls. A recent study by the Centers for Disease Control and Prevention (CDC), indicated that although 3% of girls have had sexual intercourse before reaching age 13, 18.6% are sexually active by age 15, and 59.2% by age 18 (12). We therefore believed that vaccinating 12-year-old girls would capture most girls before they are sexually active and are at risk for HPV infection. We examined the optimal vaccination age in sensitivity analyses.

Our analysis assumes a universal vaccination strategy for adolescent girls. Although risk factors for HPV infection are identifiable, we chose to evaluate a universal vaccination program for several reasons. Previous vaccination programs aimed to reduce incidence of Hepatitis B virus (HBV) infection have tried to target the risk groups that account for most cases (13). These high-risk groups, however, are difficult to vaccinate for a variety of reasons, including inaccessibility, noncompliance, and the inability to identify people at risk. Also, because >30% of HBV-infected persons show no identifiable risk factor for infection (13,14), they would not be included in such a targeted immunization strategy. Similarly for HPV infection, the broad range of risk factors and the difficulty identifying these behaviors inhibit targeting such risk groups. We evaluated the cost-effectiveness of targeting high-risk girls (assuming a reduced compliance) in sensitivity analyses.

We used Decision Maker software (Pratt Medical Group, Boston, MA, v2002.07.2) to develop a Markov model that followed the girls over their lifetimes. For each strategy, our model included probabilities of occurrence and progression of HPV, of squamous intraepithelial lesions (SIL), and of cervical cancer, as well as the probability of death, costs, and quality of life associated with the various health states. Whenever possible, we based our probability estimates (Appendix) on large, high-quality studies reported in the literature.

Our model (Figure 1A) tracks a cohort of girls who are either vaccinated against specific HPV types or who receive the current standard of care. Based on hepatitis B vaccination completion rates among U.S. adolescents, we assumed that 70% of the targeted girls would be vaccinated successfully (Appendix). We assumed that girls who were not vaccinated would receive the current standard of care.

Every month, each girl is at risk of developing high- or low-risk HPV, SIL, or cervical cancer. Over time, an infected woman’s HPV infection can regress, persist, or progress to either low- or high-grade SIL. SIL can also exist independent of an HPV infection. High-grade SIL can progress to cervical cancer. The diagnosis, treatment, and natural history of cervical cancer are modeled in Figure 1B.

We assumed that the current standard of care included routine Pap tests for compliant patients every 2 years starting at age 16. Throughout a woman’s lifetime, her HPV, SIL, or cervical cancer status can be discovered and treated either because symptoms have developed or through routine Pap tests (Figure 1C). We assumed that 10% of woman diagnosed with low-grade SIL would undergo cryotherapy and that all would receive a 6-week reexamination, and Pap tests at 3, 6, 12, and 18 months after cryotherapy. Treatment of high-grade SIL was assumed to include loop electrosurgical excision procedure (LEEP), and subsequent reexamination and Pap tests (15,16).

A woman may also choose to have a benign hysterectomy reducing her risk of cervical cancer. In addition to being at risk for death because of cervical cancer, all women are at risk for age-specific death unrelated to HPV or cervical cancer.

Because of a lack of significant HPV genotype cross-immunity, any vaccine developed probably will be effective against a limited number of HPV types (18,19). HPV types 16, 18, 45, and 31 together are the most commonly associated with cervical cancer, with evidence of these four types apparent in >75% of women who have cervical cancer (20). In our model, HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68 were considered high risk; all other HPV types were categorized as low risk (20–22). Based upon this classification in the general population, 59% of HPV infections are caused by high-risk types (Appendix).

To evaluate potential vaccination strategies, we modeled different disease-progression rates in women infected with low- and high-risk HPV. By combining data on overall progression rates of HPV infection to cancer, with prevalence data on women infected with low- and high-risk HPV who had low- or high-grade SIL or cervical cancer, we estimated separate progression rates for low- and high-risk HPV infections. Data from seven articles were considered of high enough quality to be included in our analysis (1–7).

High-risk HPV infections were significantly more common in women who had cervical cancer than in women who had precursor lesions. Based on the results of seven studies (N=1609), high-risk HPV infection was detected in 56% of women who had low-grade SIL, in 83% of women who had high-grade SIL, and in 90% of women who had cervical cancer. Low-risk HPV infection was detected in 22%, 8%, and 3% of these women, respectively. No evidence of HPV infection was found in 22%, 9%, and 7% of these women, respectively (1–7). We then calculated relative progression rates for transition from high-risk, low-risk, or no HPV infection to low-grade SIL; from low-grade to high-grade SIL; and from high-grade SIL to cervical cancer (Appendix).

We estimated that 71% of the adult female population received biennial Pap testing. Pap test sensitivity and specificity results were based on a meta-analysis conducted by the Duke Evidence-Based Practice Center (17,23). Diagnosis of asymptomatic cervical lesions depended on a woman’s likelihood of having a Pap test and on the sensitivity and specificity of this test.

Assessment of treatment effectiveness for cervical lesions was based on a review of 13 studies that detailed treatment effectiveness by lesion stage (Appendix). Initial treatment effectiveness was estimated at 97% and 94% for low-grade and high-grade SIL, respectively. Unsuccessful initial treatment of cervical lesions was followed with a repeat treatment (cryotherapy or LEEP) (77%), cone biopsy (18%), or hysterectomy (5%) (24), increasing treatment success. We based cancer progression rates, annual patient survival rates, and probability of symptoms by cancer stage on an analysis by Myers et al. (Appendix) (17). Myers et al. validated their data by comparing predicted distribution of cancer by stage for an unscreened population with data from studies of women who had had no prior screening.

We considered women who did not have cervical cancer but who had hysterectomies to be fully protected from cervical cancer. We tested this assumption in sensitivity analyses. Age-specific hysterectomy rates were based on data from the Hospital Discharge Survey of the National Center for Health Statistics (Appendix).

In our model the HPV vaccine was administered by using a series of three injections in a school-based immunization program. Because vaccine longevity is uncertain, we assumed that successful vaccination conferred immunity for 10 years but that repeated booster shots every 10 years were required to maintain the vaccine’s efficacy. We evaluated the need for more frequent booster shots or a vaccine that conferred lifetime immunity in sensitivity analysis. For our base-case analysis, vaccine efficacy against high-risk HPV types was estimated at 75%. We tested the complete range of vaccine effectiveness (from 0% to 100%) because of the absence of efficacy data from Phase III clinical trials and because future marketed vaccines may target only a subset of the high-risk HPV types.

HPV infection and cervical cancer can markedly affect quality of life and therefore can affect a woman’s quality-adjusted life expectancy. Accordingly, we incorporated adjustments for quality of life associated with current health, HPV, SIL, and with cervical cancer and its treatment.

Utilities for health states were based on a report by the Institute of Medicine on Vaccines for the 21st Century, which used committee-consensus Health Utility Indices levels for relevant health states (Appendix). Undiagnosed HPV and cervical lesions were considered to be asymptomatic and to have no utility decrement. Diagnosed and treated low- and high-grade SIL were assigned lower utilities (0.97) for a 1-year duration. Treatment for locally invasive cancer was assigned a low utility (0.79–0.80) during 4 months of initial treatment, with a moderate utility (0.90–0.97) during a 2- or 3-year follow-up. For more advanced cancer, a woman’s utility was decreased to 0.62 during both treatment and follow-up to reflect the severity of her disease and its effects on quality of life. We based current health utilities on the gender- and age-specific data from the Beaver Dam study (25).

We converted all costs to 2001 U.S. dollars by using the gross-domestic-product deflator. Pap-testing costs were $81 per test, including a 10% rescreen rate. We estimated the cost of the vaccine materials, personnel, and administration at $300, based on school-based HBV vaccination programs (Appendix). We assumed a three-injection protocol with a booster shot ($100) required every 10 years.

Treatment costs of low- and high-grade SIL were based on Medicare average reimbursements and resource-based cost estimates. We estimated the cost of treatment of low-grade SIL from the cost of an initial colposcopy and biopsy, cryotherapy (in 10% of patients), a 6-week reexamination, and Pap tests at 3, 6, 12, and 18 months after treatment. The cost of treatment of high-grade SIL was based on cost of initial colposcopy and biopsy, LEEP, and subsequent reexamination and Pap tests. Cost of cancer treatment varied, depending on the stage at which cancer was diagnosed. Costs were based on Medicare average reimbursement rates (26) and cross-checked with a 1999 HMO case-control full-cost analysis (27) (Appendix).

We performed one-way and multi-way sensitivity analyses to account for important model uncertainties. For clinical variables, our ranges for sensitivity analyses represent our judgment of the variation likely to be encountered in clinical practice, based on the literature and on discussion with experts. The ranges for costs represent variation by 25% above and below the base-case estimate. To determine ranges for utilities, we used clinical judgment.

We evaluated outcomes in the current practice arm of the model to ensure that they reflected the frequency of events from the Surveillance, Epidemiology and End Results (SEER) registry. Our model’s annual rates of cervical-cancer cases and cervical-cancer-related deaths match 2001 SEER estimates, as well as those calculated by the Myers model (17) (data available from the authors).

A prophylactic vaccine against high-risk HPV types is more expensive than current practice but results in greater quality-adjusted life expectancy (Table 1). HPV vaccination of 12-year-old girls improves their life expectancy by 2.8 days or 4.0 quality-adjusted life days at a cost of $246 relative to current practice (ICE of $22,755/QALY).

Vaccinating the present U.S. cohort of 12-year-old girls (population approximately 1,988,600) averts >224,255 cases of HPV, 112,710 cases of SIL, 3,317 cases of cervical cancer, and 1,340 cervical-cancer deaths over the cohort’s lifetime. Prevention of one case of cervical cancer would require vaccination of 600 girls (Table 2).

Figure 2 shows the ICE ratios of one-way sensitivity analyses of the vaccination strategy compared to current practice. We explored those variables with the greatest effect on the ICE ratio by running more extensive sensitivity analyses. Given the uncertainty surrounding the vaccine efficacy, pricing, and mechanism, we performed extensive sensitivity analyses using vaccine-related variables.

In our base-case analysis, we estimated that an HPV vaccine would provide immunity against high-risk HPV types in 75% of the girls vaccinated. At early stages of vaccine development or given a vaccine that targets only selected high-risk HPV types, the efficacy may prove to be lower. Sensitivity analyses on the vaccine efficacy and cost showed that even if the efficacy was reduced to 40% or the vaccine cost was increased to $600, vaccination costs <$50,000/QALY, relative to current practice (Figures 3 and 4).

We assumed that vaccination required a one-shot booster every 10 years. We also considered that vaccination could provide lifetime immunity, in which case the ICE improved to $12,682/QALY. Vaccinating the present U.S. cohort of 12-year-old girls with such a lifetime vaccine would avert >272,740 cases of HPV, 174,208 cases of SIL, 7,992 cases of cervical cancer, and 3,093 cervical-cancer deaths over the cohort’s lifetime. Prevention of one case of cervical cancer would require vaccination of 250 girls. Even if a booster shot is required every 3 years, the vaccine compared to current practice remained fairly cost effective ($45,599/QALY) (Figures 3 and 4). Our model assumes that a vaccination program would target 12-year-old girls for vaccination. Waiting until girls are 15 years old to provide vaccination results in a slightly lower life expectancy (reducing quality-adjusted life expectancy by 0.2 days) though at a reduced cost ($20). Vaccination of 12-year-old girls as compared to 15-year-old girls costs $40,440 per additional quality-adjusted life year gained.

Although the estimates used in our analysis reflect current Pap-test characteristics and compliance, if every woman obtained a Pap test every 2 years (base-case estimate is 71% compliance every 2 years), the ICE of vaccination increases to $33,218/QALY. Our base-case analysis assumes that vaccinated women would continue to receive Pap tests at the same frequency as unvaccinated women. HPV vaccination and the resulting reduction in cervical-cancer risk, however, might decrease frequency of Pap testing. Figure 5 shows how the costs and quality-adjusted life expectancy are influenced by the frequency of Pap tests in the vaccinated cohort. A strategy in which vaccinated women have Pap testing every 4 years increases life expectancy while reducing costs compared to current practice. While providing more frequent Pap tests to vaccinated women does increase a woman’s quality-adjusted life expectancy, it also increases costs. The cost-effectiveness ratios of more frequent testing are shown in Figure 5.

Our results were sensitive to several of our base-case assumptions (Figure 2). Vaccination saved 11.4 quality-adjusted life days and cost $290 over current practice when costs and benefits were not discounted (ICE of $9,286/QALY). At a discount rate of 5%, vaccination cost $37,752/QALY. Some women may be quite alarmed by being diagnosed with high-grade SIL; decreasing the utility of high-grade SIL to 0.8 lowers the cost-effectiveness ratio to $16,927/QALY. Varying the underlying incidence of HPV from 0.5 to 2 times our base-case values resulted in cost-effectiveness ratios ranging from $43,088 to $12,664 per QALY, respectively. Sensitivity analyses with other variables did not change our results substantially (Figure 2).