Cost-Effectiveness of a Potential Vaccine for Human papillomavirus

Gillian D. Sanders, Al V. Taira

Disclosures

Emerging Infectious Diseases. 2003;9(1) 

In This Article

Data and Methods

We used a decision model to estimate the length of life and expenditures for vaccination of adolescent girls for high-risk HPV types (Figure 1). We adhered to the recommendations of the Panel on Cost Effectiveness in Health and Medicine[11] for conducting and reporting a reference-case analysis. We expressed our results in terms of costs, life-years, quality-adjusted life-years (QALYs), and incremental cost-effectiveness (ICE) ratios. We performed one-way sensitivity analyses on all model variables, as well as multi-way sensitivity analyses on selected variables.

Schematic representation of the decision model. In panel A, the square node at the left represents the vaccination decision. The woman's health thereafter is simulated by a Markov model. Each month, women are at risk of developing Human papillomavirus (HPV) infection, SIL (squamous intraepithelial lesions), or cervical cancer. Women who contract HPV may be infected by a low- or high-risk type. Panel B demonstrates cervical-cancer diagnosis, treatment, and natural history. Throughout a woman's lifetime, her HPV, SIL, or cervical cancer status can be discovered either through development of symptoms or through routine Pap tests. Panel C shows that women who undergo a Pap test may test negative or positive for SIL.

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.

HPV Infection. Incidence of HPV infection was based on Myers' mathematical model of HPV infection (Appendix)[17]. In our base-case analysis, annual incidence began at age 15 (10%), peaked at age 19 (18%), and dropped off quickly after age 29 (1%). We assumed that no prevalent HPV infections existed in the initial cohort of 12-year-old girls but varied this assumption in sensitivity analyses. Given HPV infection, regression rates were highest for women <25 years (46%/yr) and lowest for women >30 years (7%/yr), reflecting a preponderance of more persistent infections in the older age group (Appendix).

Low- Versus High-Risk HPV. 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[20,21,22]. Based upon this classification in the general population, 59% of HPV infections are caused by high-risk types (Appendix).

Low- and High-Risk Rates of HPV Progression. 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[1,2,3,4,5,6,7].

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[1,2,3,4,5,6,7]. 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).

Cancer Surveillance, Treatment, and Progression. 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.

Benign Hysterectomy. 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).

HPV Vaccine Characteristics. 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.

Quality of Life. 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].

Costs. 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.

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as:

processing....