Healthcare resource use and costs associated with cervical, vaginal and vulvar cancers in a large U.S. health plan

Methods We estimated incremental ambulatory visits, hospitalizations, prescription fills and healthcare costs for cancer cases relative to population controls. Data for cervical ( n = 2788), vulvar ( n = 621) and vaginal cancer ( n = 254) cases and an identical number of controls were obtained from a large U.S. health plan. Cases were identified via diagnostic codes on a healthcare claim and matched to controls. Incremental resource use was assessed using a two-stage regression method developed by Carides, with costs analyzed using Lin's regression method. Results Through 4 years of follow-up, cervical cancer patients had incremental resource use of 12.0 ambulatory visits, 0.6 hospital admissions and 7.0 prescription fills per case. Cumulative 4-year incremental healthcare costs per case ranged from $8236 for vulvar cancers to $18,799 for cervical cancers. When adjusted to cervical, vulvar and vaginal cancer excess mortality rates observed within the U.S. Surveillance Epidemiology and End Results program, estimated incremental costs were $29,649 for cervical, $11,356 for vulvar and $21,963 for vaginal cancers. There was a significant upward trend in costs with increasing age for cervical cancer, however trends were less consistent for vulvar and vaginal cancers. Conclusions Direct medical costs associated with cervical, vulvar and vaginal cancers were observed to be substantial. These data can help inform evaluations of the economic burden and cost-effectiveness of prevention of these cancers, particularly for vulvar and vaginal disease, where such data have not been previously reported. Keywords Cervix Vulva Vagina Cancer Cost Resource use Background Human papillomavirus (HPV) infections contribute to the development of female genital cancers of the cervix, vulva and vagina [1] . Cervical cancers will be diagnosed in approximately 11,150 women in the United States in 2007, with 3670 dying from this disease [2] . HPV infection has been found to be a necessary cause in nearly 100% of cervical cancers [3] . There is evidence of a mixed etiology for vulvar cancers [4] . The vast majority of vulvar cancers among younger women are associated with HPV infections [5] . However, among women over the age of 70, approximately 50% of cases, especially of the less common keratinizing form of squamous cell carcinoma, are HPV-negative and of unknown etiology [4,6] . It is estimated that 3490 U.S. women will be diagnosed with vulvar cancer in 2007, with 880 deaths [2] . Vaginal cancers are also of mixed etiology and, like vulvar cancers, appear to have an age gradient associated with the proportion caused by HPV [7] . Overall, approximately 60–65% of U.S. vaginal cancers have been found to be infected with HPV [7,8] . In addition, intrauterine diethylstilbestrol (DES) exposure has been linked to clear cell vaginal adenocarcinomas, and chronic vaginitis, prior hysterectomy for benign disease, endometriosis and cervical irradiation have also been cited as pre-disposing factors for vaginal cancers [9] . Approximately 1100 U.S. women are diagnosed with vaginal cancer annually [10] , with 400 deaths [11] . Although estimates of the U.S. incidence and mortality associated with these cancers are reported annually [2,11,12] , data on the healthcare resource use and costs associated with their management in clinical practice are relatively sparse. For instance, we are unaware of prior U.S. studies describing the total healthcare costs per case associated with vulvar or vaginal cancers in clinical practice. In this paper we estimate healthcare resource utilization and costs associated with the management of cervical, vulvar and vaginal cancers in a large U.S. health plan. These data can be of value to policymakers wishing to understand the economic burden of these cancers as well as for policy evaluations of emerging technologies for their prevention such as HPV tests and vaccines [13-15] . Methods Study design We conducted a retrospective case-control study to assess the incremental healthcare resource utilization and costs associated with cervical, vulvar and vaginal cancers. Data sources Women included in this study were enrolled within a large fee-for-service U.S. health plan affiliated with i3 Innovus. During the period covered by our study, this health plan had more than 22 million enrollees from throughout the United States. Compared to the general U.S. population in 2000, health plan enrollees were represented in higher proportion in the Midwest (32.6% vs. 22.9%) and South (43.6% vs. 35.7%) and lower proportion in the Northeast (10.3% vs. 19.3%) and West (13.5% vs. 22.1%) [16] . Health plan members were insured through commercial or dual commercial and Medicare/Medicaid insurance coverage. Administrative data were available on the duration of member enrollment within the health plan, demographics and diagnoses, and inpatient, outpatient and pharmacy healthcare utilization and costs (from both commercial and public payors). Enrollee vital status was not recorded within health plan administrative data. Dates of death for enrollees dying during the course of follow-up were queried from the National Death Index (NDI) of the Centers for Disease Control's (CDC) National Center for Health Statistics ( http://cdc.gov/nchs/ndi.htm ). Analysis populations Cancer cases were identified from healthcare claims covering the period from January 1, 1998 to December 31, 2003. To qualify as an incident case, a woman must have had a primary or secondary International Classification of Disease, Ninth Revision (ICD-9) code for cervical (180.0–180.9), vulvar (184.1–184.4) or vaginal (184.0) cancer recorded on a healthcare claim during the study period. The first date upon which such a code appeared during the study period was labeled the "index date." Cancer cases were also required to lack evidence of a primary or secondary ICD-9 diagnosis code for cancer at any site (140.0–208.9) during the 12 months prior to their index date and to have been continuously enrolled within the health plan over that period. This restriction was imposed to eliminate cases in which cervical, vulvar or vaginal cancers represented metastases from other body sites as U.S. cancer statistics reflect reporting of data for primary tumors only [12] . Population controls without a claim-based ICD-9 code for cervical, vulvar or vaginal cancer were also selected for each respective cancer sample from among females enrolled in the health plan during the same period as cases. An initial group of population controls was selected based on four matching criteria with respect to cancer cases: 1) Index date (identical to case); 2) No cancer ICD-9 code during the 12 months prior to index date, with continuous enrollment over that period; 3) Age (± 5 years); and 4) Region (Northeast, South, Midwest, West). From this initial group, a single population control was then selected for each cancer case using propensity score matching on co-morbidities and healthcare costs observed during the 12 months prior to each subject's index date. This propensity score matching was designed to yield a sample of controls that was relatively similar to cases with respect to the number and healthcare cost impact of co-morbid conditions and to also reduce the number of controls selected for analysis to a manageable number. First, health conditions observed during the 12 months prior to the index date were tabulated from medical claims using an algorithm maintained by the Agency for Healthcare Research and Quality's (AHRQ) Healthcare Cost and Utilization Project (HCUP) [17] . This algorithm creates binary variables indicating whether individuals have or do not have each of 131 health conditions. Second, healthcare costs were tabulated from claims data during the 12 months prior to each subject's the index date. Third, propensity scores were estimated for each case and control through unconditional logistic regression using group status (case vs. control) as the dependent variable and co-morbidities and healthcare costs as independent variables. Fourth, the control groups were then formed for each of the three cancer groups by selecting the control with the closest propensity score to each cancer case. Once selected as a control for a given case, a control was removed from the eligible population of controls for subsequent cases. Economic outcome measures For each disease category (cervical, vulvar or vaginal cancer) we estimated incremental counts of ambulatory visits, hospitalizations and prescription fills, and healthcare costs, for cases vis a vis controls post-index date. Healthcare costs reflected payments for all health plan care from all sources (primary insurance, co-insurance, patient deductibles and co-payments and other payments and adjustments). While all patients were commercially insured, some patients (mainly among those age 65 and over) were also insured through Medicare, and these payments were also included. While a case's index date may represent the date on which a cancer diagnosis is confirmed, for some patients, this is preceded by a period of diagnostic work-up. This may reflect follow-up for an abnormal Pap smear, colposcopy and biopsy prompted by reporting of disease symptoms, or an initially incomplete classification of disease (e.g., as a high-grade pre-cancer) prior to the discovery of invasion. In addition to examining healthcare resource use and costs during the post-index date period, we also separately characterized these pre-index date resources and costs based on the presence of the following pertinent administrative codes during the 12 months prior to the index date: Cervical cancer (Principal ICD-9: 079.4, 233.1, 236.3, 239.5, 622.10–622.12, 795.00–795.09, V10.41, V72.3-V72.32, V76.2, 91.46; Current Procedural Terminology-4th edition [CPT-4]: 88141–88154, 88164–88167; Healthcare Common Procedure Coding System [HCPCS]: P3000-P3001; Universal Billing [UB92] Form: 923) Vulvar cancer (Principal ICD-9: 233.3, 236.3, 239.5, 624.8) Vaginal cancer (Principal ICD-9: 233.3, 236.3, 239.5, 623.0, 795.1, V76.47) These codes include those for pre-cancer ous neoplasms at each site, as well as lesions of uncertain malignant behavior, and routine and follow-up Pap smears (for cervical and vaginal cancers). Statistical analysis Baseline characteristics for cases and controls are reported descriptively through numbers and percentages for categorical variables and means and standard deviations for continuous variables. The comparability of baseline characteristics between cases and controls was assessed using chi-square tests for categorical variables and t-tests for continuous variables. We used multiple regression analysis methods to estimate incremental healthcare resource utilization and costs. Incremental healthcare resource utilization during 1-year, 2-year, 3-year and 4-year follow-up periods for cancer cases versus controls was analyzed using a two-stage regression method developed by Carides [18] . In the first stage, a linear regression was run for each time period of interest to predict mean utilization rates for each category of care for cases and controls. The regression included covariates for case status (case or control), age, pre-index period Charlson co-morbidity index score [19] and pre-index period healthcare costs. The Charlson co-morbidity index score was chosen over the richer AHRQ algorithm assessment of co-morbidities for analysis purposes because the latter measure would have prohibitively reduced regression degrees of freedom for some analyses. In the second stage, mean estimates of resource utilization for each time period evaluated in the first stage, for those surviving to that time or dying prior to that time, were respectively weighted by the probability of survival or death to that time point based on Kaplan-Meier estimation, with separate weighting conducted for cases and controls. Administrative censoring 225 was defined to occur on the date of insurance disenrollment and accounted for in estimating survival probabilities. Incremental healthcare utilization was then estimated by calculating the difference in mean resource use between cases and controls. Estimates of resource use for individual subjects were bootstrapped to obtain 95% confidence intervals [20] . Incremental healthcare costs during 1-year, 2-year, 3-year and 4-year follow-up periods for cancer cases versus controls were analyzed using Lin's regression method [21] . With Lin's regression method, case and control cost histories were divided into monthly intervals and a weighted linear regression model, with the same covariates as in the analyses of resource utilization, was then used to estimate the incremental costs for cases versus controls within each monthly interval. These monthly incremental costs were weighted by the probability of surviving and not being administratively censored by that interval, and summed over the relevant time period of interest. In addition we also used Lin's regression method to estimate the incremental costs of cancer patients who died prior to the end of each year of follow-up and those who survived that year. Due to differential follow-up periods dictated by varying index dates (e.g., a woman diagnosed with cancer in December 2002 could have a maximum of 1 year of follow-up until the end of the study period), there was progressively increasing administrative censoring within the data with longer follow-up times. We report economic outcomes through 4 years of follow-up, along with the size of the eligible sample at the start of each yearly time point, unless otherwise noted. All costs were healthcare inflation adjusted to 2003 U.S. dollars using the Medical Care component of the U.S. Bureau of Labor Statistics Consumer Price Index [22] . To enhance external generalizability, overall healthcare resource use and cost results across all ages were age-adjusted to the 2000 U.S. female population [23] . Weights for age adjustment were derived from the underlying population from which cases and controls were drawn; reflecting women with at least 12 months of continuous health plan enrollment during the analysis period. Results A total of 2788 cervical, 621 vulvar and 254 vaginal cancer cases met the study eligibility criteria and were 1:1 matched to an equal number of population controls. Baseline characteristics of these samples are described in Table 1 . There were no substantive differences in any baseline characteristic between cancer cases and population controls, with the exception of for healthcare costs during the 12 months pre-index date, in which, for the vaginal cancer analysis, cases were observed to have higher costs than controls ( p = 0.05). Through 4 years of follow-up, cervical cancer patients were observed to have incremental resource use resulting in 12.0 ambulatory visits, 0.6 hospital admissions and 7.0 prescription fills per case ( Table 2 ). The vast majority of incremental ambulatory visits and hospital admissions (≥ 80%) were observed to have occurred by the end of the first year post-index. Resource use during the pre-index period, representing cervical cancer screening and diagnostic work-up for cervical abnormalities, comprised 17.5% of the total ambulatory visits observed during the course of follow-up, but less than 5% of hospital admissions. While incremental ambulatory visits and hospital admissions were statistically significantly greater for cases than controls, 95% confidence intervals for prescription fills overlapped zero at all time points. Further investigation revealed that cases and controls averaged 15–16 prescription fills unrelated to cervical disease during the 12 month pre-index period, compared to 12 such ambulatory visits. This somewhat higher background rate of utilization combined with the observed smaller magnitude of prescription fills relative to ambulatory visits during follow-up may have contributed additional noise resulting in the observed lack of statistical significance. For vulvar cancer patients, incremental resource use over 4 years of follow-up was 24.4 ambulatory visits, 0.9 hospital admissions and 15.9 prescription fills per case. A potentially different pattern emerged from that observed for cervical cancer cases, as a much smaller fraction of incremental resource use was observed through the first year of follow-up relative to cumulative 4-year totals for vulvar cancers. Caution should be exercised in interpreting some of this time trend as 95% confidence intervals for resource use observed during years 1–3 overlapped with each other. Confidence intervals for incremental prescription fills again overlapped zero. A lack of observations in certain age strata prevented a complete analysis of incremental resource use for vaginal cancer cases through 4 years of follow-up. Based on limited data reported through 1 year of follow-up however, the intensity of resource use appeared more similar to that observed for cervical as compared to vulvar cancers. Cumulative 4-year incremental healthcare costs per case for cervical cancer were $18,799 as compared to $8236 for vulvar cancer ( Table 3 ) . The majority (61.6%) of cervical cancer costs were incurred during the first year of follow-up whereas just over half (50.5%) of vulvar cancer costs were observed during years 2–4 post-index. Incremental costs associated with vaginal cancers were similar in magnitude and trajectory to those reported for cervical cancers, but due to the small number of cases for which data were available, 95% confidence intervals at all time points overlapped zero. When stratified by age, there was a statistically significant upward trend ( p < 0.05) in cervical cancer costs with increasing age, with 4-year cumulative costs of $12,220 (95% CI, $8949–$15,491) among women under age 40 versus $21,967 ($16,748–$27,185) among those ages 40–59 and $28,953 ($17,442–$40,464) among those ages 60 and over. Some increasing trends in costs with age were also seen with vulvar and vaginal cancers, however patterns were less consistent and confidence intervals across age groups frequently overlapped. Incremental healthcare costs among cancer cases dying during the course of follow-up were observed to be 2–4 times higher than for cases surviving to a given follow-up time ( Fig. 1 ). Confidence intervals around cost estimates for those cases that died were wide, due to the small number of women with cervical ( n = 63), vulvar ( n = 14) and vaginal ( n = 7) cancers who died during the course of follow-up. As a result, confidence intervals overlapped between the two groups. The small number of deaths among the cancer groups in our study population was unexpected. To provide further perspective, we compared absolute excess mortality rates observed among cancer cases over controls within our study population to those reported from the Surveillance Epidemiology and End Results (SEER) Program for women diagnosed with cervical, vulvar and vaginal cancers from 1998–2003 ( Table 4 ) [10] . The SEER program is comprised of 17 state or local registries covering approximately one-fourth of the U.S. population. Excess mortality rates for each cancer and at each time point were substantially higher within SEER than in our study population. For instance, after 3 years, the cumulative mortality rates for cervical, vulvar and vaginal cancer patients within our study population were 7.0%, 7.6% and 6.3% respectively as compared to 23.6%, 18.2% and 39.4% observed within SEER. Given the magnitude of this differential mortality, we conducted a sensitivity analysis to explore the potential impact on our results for cancer costs. For each cancer, and at each time point, we multiplied incremental costs estimated using the Lin method for women surviving or dying to that time point in our results by the respective fraction of women estimated to die or not to die from each cancer by that time point within the SEER data. Because costs were observed to be higher among cancer patients who died than among those surviving to a given time point, costs for each cancer and at each time point were observed to increase when adjusted to the higher excess cancer mortality rates within SEER ( Table 4 ). The largest absolute differences with the adjustment were seen for cervical cancer, where costs through 4 years of follow-up were $29,649, compared to $18,799 without mortality adjustment. On a relative basis, at the longest follow-up time, costs for the three cancers ranged from 38–59% higher with the adjustment. Discussion This study has examined the incremental healthcare resource use and costs associated with cervical, vulvar and vaginal cancers within a large U.S. health plan. Management of these cancers within the health plan population was estimated to result in 12–24 ambulatory visits, 0.4–0.9 hospital admissions and $8000–19,000 in healthcare costs per case, reflective of a high economic burden at the patient level. To our knowledge this is the first U.S. population-based study to report resource use and costs associated with the management of vulvar and vaginal cancers in general clinical practice. These data can be helpful for providing needed inputs for economic evaluations of interventions for the early diagnosis and prevention of these cancers. For instance, several previous studies have evaluated the cost-effectiveness of prophylactic HPV vaccines targeting types responsible for a substantial portion of vulvar and vaginal cancers [14,24] . However, these studies have exclusively focused on the prevention of cervical cancers and its precursors or, in one instance incorporated prevention of anogenital warts (for a quadrivalent HPV 6/11/16/18 vaccine), and have not examined the health and economic implications of vulvar or vaginal cancer prevention. This analysis has also provided an updated perspective on the resource use and healthcare costs associated with cervical cancers. Several previous U.S. studies have also analyzed patient-level data in reporting resource use and/or healthcare costs per case for cervical cancers. Baker et al. analyzed Medicare charge data on inpatient, outpatient and nursing facility care for cervical cancer patients from 1974–1981 and reported charges (1984 dollars) of $8979 for the initial 3 months post-diagnosis, $493/month for continuing care and $16,414 for terminal care [25] . Information on the global costs of cervical cancer per case was not reported as the distribution of patients in each of these stages of care was not described. Fahs et al. assigned Medicare average allowable charges to inpatient resource use among elderly women observed in the 1987 National Hospital Discharge Survey [26] . However, outpatient cervical cancer costs were based on pre-specified treatment protocols, rather than observed patient resource use. Reported costs were $9216 for Stage I and $13,359 for Stages II-IV (1988 dollars). Mandelblatt et al. also reported costs for initial, continuing and terminal care using Medicare-SEER linked data from 1986–1998, however methods of analysis were not fully reported, and some costs also appeared to be based on pre-specified treatment algorithms [27] . While Medicare data are helpful for describing the experiences of women primarily over the age of 65, they are not representative of the majority of women diagnosed with cervical cancer in the U.S. Based on SEER data for 2004, 19% of women diagnosed with cervical cancer were over the age of 65(10) and prior studies have reported a minority of U.S. women diagnosed with cervical cancer to be covered by Medicare, with private insurance representing the most commonly observed type of coverage [28,29] . Two previous U.S. studies have reported cervical cancer costs in other settings. Helms et al. utilized 1990–1991 data from Group Health Cooperative health maintenance organization (Puget Sound, WA) to estimate incremental costs for 98 cervical cancer cases as compared to 133,058 population controls. Similar to in our study, their population was younger than the general population of U.S. women diagnosed with cervical cancer and they adjusted their results to SEER data (using stage distribution rather than mortality) to estimate a global cost for cervical cancer of $30,136 (1996 dollars) [30] . McCrory et al. used 1992–1994 healthcare claims from the Medstat MarketScan database to estimate costs of Stage I ($17,645), Stages II–III ($27,069) and Stage IV ($40,280) cervical cancers (1994 dollars) among women aged 20–64 with private health insurance [31] . None of the prior studies reported the number of ambulatory visits, hospitalizations or prescription fills associated with cervical cancer. Placing the results of the present study in the context of previous work, we observed a steadily increasing trend in cervical cancer costs with patient age, suggesting that resource use and costs among older cervical cancer patients, such as those covered by Medicare, may well be higher than among younger ones. Cervical cancer screening rates among older U.S. women are very low compared to among younger women [32] and they are much more likely than to be diagnosed with late-stage cervical cancer [12,28] and to die from disease [12] which would render their care to be more expensive. The majority of U.S. women over the age of 65 are enrolled in some supplementary form of private insurance [33] , as in our study population, however whether this is the case for the subset of U.S. women diagnosed with cervical cancer is not known. Among privately insured women below age 65, our estimates of healthcare costs are lower than those reported by McCrory et al. for 1992–1994 [31] , particularly were one to adjust for healthcare inflation over time [22] . This is somewhat surprising, as both studies identified cancer cases from healthcare claims using similar ICD-9 codes, and the reasons behind the differential are unclear. Our cost results adjusted to mortality rates observed within SEER are very similar to those reported by Helms et al., who stage adjusted to SEER data, albeit from a much smaller population of cases and in an earlier year [30] . In placing our results in the context of costing studies of other cancers, it is important to note that many such studies have been conducted among a selected group of cancer patients according to age (e.g., > 65 years), stage of diagnosis or phase of care (e.g., initial or terminal) [34] . However, similar to the present analyses, Fireman et al. reported undiscounted cancer costs (1992 dollars) for patients across all ages, stages and phases of care of $39,670 for breast cancer, $34,442 for lung cancer and $45,045 for colon cancer within the Kaiser Permanente Northern California health plan [35] . These costs are somewhat higher than the costs estimated for female gynecologic cancers reported in this paper. This could be a real effect, or relate in part to the longer follow-up period used (15 years), which may have captured additional cancer recurrences. Also, patient co-morbidities/baseline healthcare use were not directly adjusted for in comparison to controls in the analysis. Our study has several limitations. First, cancer cases were identified from ICD-9 codes on a healthcare claim as we did not have access to patients' medical records for this study. A previous study assessed the accuracy of inpatient ICD-9 codes for cervical cancer through a comparison with medical records and found a reasonably high positive predictive value (86%) for administrative coding [36] . The accuracy of ICD-9 codes for outpatient cervical cancer care and for vulvar and vaginal cancers have not been evaluated. Second, because the pathogenesis of cancer, and complications associated with its treatment, can have a diverse impact on health and functioning [37] , we did not attempt to directly attribute resource use and costs to cancer cases, but rather to estimate these endpoints incrementally relative to data for population controls. We attempted to match and control for variables such as age, region, baseline co-morbidities and baseline intensity of healthcare costs, however, it is possible that further differences may have existed between cases and controls beyond those apparent from within our data. In the vaginal cancer analysis, there was observed to be some imbalance in baseline period costs between cases and controls, even after propensity score matching, but we were able to control for this variable in estimating incremental vaginal cancer costs using the Lin method. Third, although national data on the age-specific incidence of cervical, vulvar or vaginal cancers are not available, because our study population had private health insurance, which is often employer based, it tended to be younger than populations of patients with these cancers from U.S. cancer registries such as SEER. In an attempt to enhance the generalizability of our results, we initially performed age adjustment to the 2000 U.S. female population, with weights derived from the underlying population from which cases and controls were drawn. However, even within age strata, there were relatively large differences in mortality rates between our study population and that of SEER. This could have related to differences in factors such as patient behaviors (e.g., greater participation in gynecologic screening with diagnosis at an earlier cancer stage) or less rigorous case identification methods (e.g., misclassification of some patients actually diagnosed with carcinoma in situ). For example, among cervical cancer patients aged 20–39 in our study, the 4-year absolute excess mortality rate relative to controls was 3.7% as compared to 15.6% within SEER data from 1998–2003 [10] . As a result, we found that age adjustment of our results produced relatively little impact on estimated resource use and costs, while adjustment to mortality rates observed within SEER had a larger effect. For policy analysis purposes, we would therefore view the data presented in our primary results as most applicable to commercially insured populations with employer-based coverage, with the mortality-adjusted data more reflective of the general population of cervical, vulvar and vaginal cancer patients. Even so, generalizability to the latter group is imperfect. For instance, although individuals with commercial insurance or dual/Medicare commercial insurance coverage are estimated to account for the majority of cervical cancer patients [28,29] , our results may be less generalizable to the minority of patients who are insured through Medicare or Medicaid alone, or who are uninsured. It would have also been desirable to have had information on other clinical variables such as cancer stage at diagnosis and tumor histology, however, these data were unfortunately not available to us within the healthcare claims. Finally, costs reported within this study reflect direct medical costs and do not include lost wages and productivity due to cancer morbidity and mortality. A previous study of women with cervical cancer estimated these indirect costs to account for more than three times the direct costs of disease [38] . Similar comparisons have not been reported for vulvar or vaginal disease. In conclusion, this study has described the substantial direct medical costs associated with cervical, vulvar and vaginal cancers. Were the mortality-adjusted estimates of cancer costs per case in this study to be extrapolated to the U.S. population [2] , the resultant annual direct costs associated with cervical cancers would be $300–400 million as compared to $40 million for vulvar cancers and $25 million for vaginal cancers. Conflict of interest statement This study was supported by Merck & Co., Inc., where authors RPI, PKS and GWC were employed. XY participated in this study under contract to Merck & Co., Inc. Acknowledgments We dedicate this paper to the late Dr. George W. Carides whose professional contributions to the field of healthcare resource use and cost analysis and positive personal impact upon those he worked with will not be forgotten. We also thank Dr. Yiyu Fang for programming assistance with the data analysis and Dr. Henry J. Henk for reviewing an earlier version of this manuscript. References [1] L. Denny H.Y. Ngan Prevention and treatment of HPV associated disease in the HPV vaccine era. Section B: malignant manifestations of HPV infection — Carcinoma of the cervix, vulva, vagina, anus, and penis Int J Obstet Gynecol 94 Suppl 1 2006 S50 S55 [2] A. Jemal R. Siegel E. Ward T. Murray J. Xu M.J. Thun Cancer statistics, 2007 CA Cancer J Clin 57 1 2007 43 66 [3] J.M. Walboomers M.V. Jacobs M.M. Manos F.X. Bosch J.A. Kummer K.V. Shah Human papillomavirus is a necessary cause of invasive cervical cancer worldwide J Pathol 189 1 1999 12 19 [4] S.K. Tyring Vulvar squamous cell carcinoma: guidelines for early diagnosis and treatment Am J Obstet Gynecol 189 3 Suppl 2003 S17 S23 [5] B.J. Monk R.A. Burger F. Lin G. Parham S.A. Vasilev S.P. Wilczynski Prognostic significance of human papillomavirus DNA in vulvar carcinoma Obstet Gynecol 85 5 Pt 1 1995 709 715 [6] G.J. Nuovo P. Delvenne P. MacConnell E. Chalas C. Neto W.J. Mann Correlation of histology and detection of human papillomavirus DNA in vulvar cancers Gynecol Oncol 43 3 1991 275 280 [7] J.R. Daling M.M. Madeleine S.M. Schwartz K.A. Shera J.J. Carter B. McKnight A population-based study of squamous cell vaginal cancer: HPV and cofactors Gynecol Oncol 84 2 2002 263 270 [8] M.T. Kiyabu D. Shibata N. Arnheim W.J. Martin P.L. Fitzgibbons Detection of human papillomavirus in formalin-fixed, invasive squamous carcinomas using the polymerase chain reaction Am J Surg Pathol 13 3 1989 221 224 [9] M.J. Merino Vaginal cancer: the role of infectious and environmental factors Am J Obstet Gynecol 165 4 Pt 2 1991 1255 1262 [10] SEER ⁎ Stat 6.3.5- SEER Cancer Incidence Limited-Use Database, 1973–2004 and Associated Populations in Binary Format for Use by SEER ⁎ Stat. 4-16-2007. Bethesda, MD, U.S.A., Cancer Statistics Branch, National Cancer Institute, National Insititutes of Health. [11] U.S. Cancer Statistics Working Group. United States Cancer Statistics: 1999–2002 Incidence and Mortality Web-based Report. Atlanta: U.S.Department of Health and Human Services, Centers for Disease Control and Prevention and National Cancer Institute; 2005.Available at: www.cdc.gov/cancer/npcr/uscs . 3-21-2007. [12] Ries LAG, Melbert D, Krapcho M, Mariotto A, Miller BA, Feuer EJ, et al. SEER Cancer Statistics Review, 1975–2004, National Cancer Institute. Bethesda, MD, http://seer.cancer.gov/csr/1975_2004/ , based on November 2006 SEER data submission, posted to the SEER web site. 2007. [13] A.D. Brown A.M. Garber Cost-effectiveness of 3 methods to enhance the sensitivity of Papanicolaou testing JAMA 281 4 1999 347 353 [14] E.J. Dasbach E.H. Elbasha R.P. Insinga Mathematical models for predicting the epidemiologic and economic impact of vaccination against human papillomavirus infection and disease Epidemiol Rev 28 2006 88 100 [15] J. Holmes L. Hemmett S. Garfield The cost-effectiveness of human papillomavirus screening for cervical cancer. a review of recent modelling studies Eur J Health Econ 6 1 2005 30 37 [16] Population pyramids and demographic summary indicators for U.S. Regions and Divisions. U.S.Census Bureau ( http://www.census.gov/population/www/projections/regdivpyramid.html ) . 8-17-2007. [17] Comorbidity Software, Version 3.2. Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality (AHRQ), Rockville, MD, http://www.hcup-us.ahrq.gov/toolssoftware/comorbidity/comorbidity.jsp , 2006. [18] G.W. Carides J.F. Heyse B. Iglewicz A regression-based method for estimating mean treatment cost in the presence of right-censoring Biostatistics 1 3 2000 299 313 [19] M. Charlson T.P. Szatrowski J. Peterson J. Gold Validation of a combined comorbidity index J Clin Epidemiol 47 11 1994 1245 1251 [20] T. Hesterberg S. Monaghan D.S. Moore A. Clipson R. Epstein Bootstrap Methods and Permutation Tests 2003 W. H. Freeman New York, N.Y. [21] D.Y. Lin Linear regression analysis of censored medical costs Biostatistics 1 1 2000 35 47 [22] Consumer Price Index-Medical Care. U.S. Department of Labor. Bureau of Labor Statistics. 2006. http://data.bls.gov/cgi-bin/surveymost , Accessed. 2-15-2007. [23] 2000 resident population estimates of the United States by age and sex., http://wonder.cdc.gov/census.shtml . 3-15-2004. U.S. Census Bureau. [24] A.T. Newall P. Beutels J.G. Wood W.J. Edmunds C.R. MacIntyre Cost-effectiveness analyses of human papillomavirus vaccination Lancet Infect Dis 7 4 2007 289 296 [25] M.S. Baker L.G. Kessler R.C. Smucker Site-specific treatment costs for cancer: an analysis of the Medicare continuous history sample file R.M. Sheffler N.C. Andrews Cancer Care and Costs 1989 Health Administration Press Ann Arbor, MI 127 138 [26] M.C. Fahs J. Mandelblatt C. Schechter C. Muller Cost effectiveness of cervical cancer screening for the elderly Ann Intern Med 117 6 1992 520 527 [27] J.S. Mandelblatt W.F. Lawrence S.M. Womack D. Jacobson B. Yi Y.T. Hwang Benefits and costs of using HPV testing to screen for cervical cancer JAMA 287 18 2002 2372 2381 [28] J.M. Ferrante E.C. Gonzalez R.G. Roetzheim N. Pal L. Woodard Clinical and demographic predictors of late-stage cervical cancer Arch Fam Med 9 5 2000 439 445 [29] C.A. Leath III J.M. Straughn Jr. T.O. Kirby A. Huggins E.E. Partridge G.P. Parham Predictors of outcomes for women with cervical carcinoma Gynecol Oncol 99 2 2005 432 436 [30] L.J. Helms J. Melnikow Determining costs of health care services for cost-effectiveness analyses: the case of cervical cancer prevention and treatment Med Care 37 7 1999 652 661 [31] McCrory DC, Myers ER, Nanda K, Bastian LA, Hasselblad V, Matchar DB. Evaluation of cervical cytology. Evidence report/Technology assessment No. 5. AHCPR Publication No. 99-E010. 1999. Rockville, MD, Agency for Healthcare Policy and Research. [32] R.P. Insinga A.G. Glass B.B. Rush Pap screening in a U.S. health plan Cancer Epidemiol Biomarkers Prev 13 3 2004 355 360 [33] National Center for Health Statistics. Health, United States, 2006: With Chartbook on Trends in the Health of Americans. Hyattsville, MD: 2006. [34] K.R. Yabroff J.L. Warren M.L. Brown Costs of cancer care in the USA: a descriptive review Nat Clin Pract Oncol 4 11 2007 643 656 [35] B.H. Fireman C.P. Quesenberry C.P. Somkin A.S. Jacobson D. Baer D. West Cost of care for cancer in a health maintenance organization Health Care Financ Rev 18 4 1997 51 76 [36] L. Penberthy D. McClish A. Pugh W. Smith C. Manning S. Retchin Using hospital discharge files to enhance cancer surveillance Am J Epidemiol 158 1 2003 27 34 [37] M.L. Brown G.F. Riley N. Schussler R. Etzioni Estimating health care costs related to cancer treatment from SEER-Medicare data Med Care 40 8 Suppl 2002 IV 17 [38] W. Max D.P. Rice H.Y. Sung M. Michel W. Breuer X. Zhang The economic burden of gynecologic cancers in California, 1998 Gynecol Oncol 88 2 2003 96 103