Bidus, Michael A. MD1; Maxwell, G Larry MD1; Kulasingam, Shalini PhD2; Rose, G Scott MD1; Elkas, John C. MD, JD1; Chernofsky, Mildred MD1; Myers, Evan R. MD, MPH2
Despite the decline in the incidence of cervical cancer in developed countries in the past 4 decades, cervical cancer remains the second most common female malignancy worldwide.1 In the United States, an estimated 10,370 new cases and 3,710 deaths are expected to occur annually.2 Over the past 20 years, cervical cancer screening with the Pap test has resulted in a significant reduction in the mortality from this disease.3 Analysis of Surveillance, Epidemiology, and End Results registry data has shown a 46% reduction in mortality related to cervical cancer from 1973 to 1995.4 However, screening guidelines for the detection of preinvasive and invasive disease are varied, and have included conventional or liquid-based cytology alone, testing for oncogenic human papillomavirus (HPV) types alone, or cytologic evaluation in conjunction with HPV testing or with reflex HPV testing in cases of atypical squamous cells of undetermined significance (ASC-US).5 In 2001 the results of the Atypical Squamous Cells of Undetermined Significance Low-Grade Squamous Intraepithelial Lesion Triage Study trial confirmed the use of HPV DNA testing as an effective ASC-US triage algorithm for the detecton of high-grade preinvasive disease.6,7 Recently, the United States Food and Drug Administration approved HPV DNA testing with cytology screening for women aged older than 30 years provided that screening occurs no more frequently than every 3 years.8
Based in part on the results of the Atypical Squamous Cells of Undetermined Significance Low-Grade Squamous Intraepithelial Lesion Triage Study trial and a cost-effectiveness analysis comparing different triage strategies for ASC-US using cost data from the Army,9 the U.S. Army recently completed a transition to liquid-based cytology with HPV triage of equivocal results in accordance with the American Society for Colposcopy and Cervical Pathology (ASCCP) guidelines for the management of abnormal cervical cytology and cervical cancer precursors. A limitation of the previous analyses is that they did not account for indirect costs to the parent organization associated with time away from duty or work, which for a military population can affect readiness enhancement and deployment. Time away from duty is defined as the amount of time that a patient is either physically incapable of or otherwise unavailable to perform employment-related activities. This is significant because other models have potentially underestimated the economic importance of the “time away from duty” component of the analysis to the parent organization. In this analysis, we examine the effectiveness and cost-effectiveness of multiple cervical cancer screening strategies using cytology and HPV data from an Army population and incorporating both direct and indirect costs.
MATERIALS AND METHODS
A previously described Markov state transition model9 was used to simulate the natural history of HPV infection, preinvasive disease, and invasive cervical cancer for a cohort of 100,000 active-duty U.S. Army service women. Healthy, infected with HPV, low-grade intraepithelial lesions, high-grade intraepithelial lesions, invasive cervical cancer, and death states were represented in the model. Patients entered the model in 1 health state and could transition between the different states based on defined probabilities. The model was updated with algorithms reflecting current ASCCP treatment guidelines10 and the Bethesda System 2001 reporting guidelines.11 Specifically, we assumed that screening began at age 18 and ended at age 85. All women undergoing colposcopy were assumed to undergo endocervical curettage. We assumed that women with biopsy-proven cervical intraepithelial neoplasia (CIN) 1 were followed with repeat cytology and that those with persistent CIN 1 (greater than 12 months) were treated with loop electrosurgical excision procedure. Women with CIN 2-3 were assumed to be referred for immediate treatment. Because all patients were active duty military, compliance with screening, diagnosis, and treatment was assumed to be perfect. The states and transition probabilities are listed in Table 1.
Women were grouped according to age (18–21 years, 22–24 years, 25–29 years, 30–34 years, and 35+ years) and enlisted rank (resulting in rank groups of e1-e3, e4, e5, e6, and e7-e9). Enlisted rank structure starts with e1 and progresses with skill level, job expertise, and leadership abilities until e9 is reached. Enlisted ranks are as follows: e1 and e2, Private; e3, Private First Class; e4, Specialist or Corporal; e5, Sergeant; e6, Staff Sergeant; e7, Sergeant First Class; e8, Master or First Sergeant; e9, Sergeant Major. Age and rank groups were developed to better reflect potential differences in screening history, natural history of HPV infection, cervical precancer, cancer, and salary for calculations of indirect costs. Women were assumed to enter the model based on the average age of women in a specific rank. The distribution of underlying CIN (prevalence) was calculated based on cytology data12 from the army used in conjunction with data from Washington State13 and is described below. We calculated outcomes and costs based on retention and promotion rates. We assumed that if women reached the rank of e5 and subsequently were not promoted, they were allowed to stay in the model until retirement. Female soldiers of a lesser rank (e1 to e4) who were not promoted were allowed to stay in their respective rank until 8 years had passed, at which point they were assumed to leave the army due to failure to promote. Women who retired from the U.S. Army after 20 years of service were entitled to life-time health care benefits through the Department of Veterans Affairs and therefore were kept in the model until age 85 years.
The distribution of liquid-based cytology (LBC) results from a population of military health care beneficiaries screened during 2003 in the Washington, DC area was used12 in conjunction with screening and histology data from a large population-based study in Washington State13 to estimate underlying prevalence of disease by age (Table 1). Data from the Northeast region of the United States regarding the performance of HPV testing among women with ASC-US (grouped by age) was used to calibrate the model (unpublished data from Walter Reed Army Medical Center, Washington, DC).
Several strategies of cervix cancer screening were compared: LBC with testing for HPV irrespective of cytologic results (DNA/PAP), compared with LBC with HPV detection for cytologic results of atypical cells of undetermined significance (reflex HPV). The costs and outcomes of these screening methods were evaluated separately as well as in combination (LBC and reflex HPV before age 30 years and DNA/PAP every 3 years thereafter). In addition, each of these 3 screening methods was evaluated at 1-, 2-, and 3-year intervals. Screening test performance was based on a review of the literature14,15; estimates of sensitivity and specificity for LBC and reflex HPV were assumed to be 80% and 95%, respectively.14 Sensitivity and specificity of DNA/PAP conducted in women aged 30 years and older was based on literature that corrected for verification bias and attendance at colposcopy, and was assumed to be 89% and 90%, respectively.15 Sensitivity and specificity of DNA/PAP in women aged younger than 30 years were assumed to be 93% and 80%, respectively.15 Colposcopy was assumed to be 100% sensitive and 100% specific. All estimates of sensitivity and specificity were based on a cytology threshold of ASC-US and higher and for detection of CIN 2-3.
Health-related costs were obtained from the U.S. Army and TRICARE allowable reimbursables when possible but taken from the literature when unavailable from military resources.16 All health-care related costs were measured in 2004 U.S. dollars and are listed in Table 2. Time costs (estimated as time away from work to undergo screening, diagnosis, and treatment) were calculated by applying salaries from the Department of Defense pay tables (2004 data)17 to the groupings described previously. Time associated with screening was assumed to be 0.5 days; diagnosis 0.5 days; treatment for CIN 2-3 3.0 days; treatment for cancer less than stage IIA was 0.90 at 5 weeks and 0.10 at 8 weeks; and cancer greater than stage IIB, 8 weeks. All costs and outcomes are discounted at 3% annually.
We calculated incremental cost-effectiveness ratios in which the additional costs of a strategy, divided by the additional savings in life expectancy or quality-adjusted life expectancy were compared with the next, least-costly strategy. Strategies were considered to be dominated if they were more costly and less effective in terms of life expectancy or had a higher cost-effectiveness ratio than an adjacent strategy or strategies.
Sensitivity analyses were performed in which we varied the prevalence of CIN and HPV by 50% to 200% of base prevalence to determine whether our use of data from Washington State to estimate the distribution of underlying disease conditional on cytology results for the Army population would affect our findings. Additionally, the sensitivity of LBC was varied down to 60%, and the specificity of combined HPV and LBC varied from 70% (for women aged younger than 30 years) to 80% (for women aged older than 30 years) to account for differences in test performance that might affect the ranking of the different strategies tested. Finally, sensitivity analyses were conducted to determine whether the inclusion of fringe benefits (allowance for sustenance and allowable housing costs) in addition to base salary would affect outcomes. Time costs with and without fringe benefits are shown in Tables 3 and 4. We also varied other costs, including the cost of the HPV test ($5 to $40).
In the base-case analysis, the data are presented as total costs, incremental costs, total effects (life expectancy), incremental effects and cost-effectiveness (defined using the incremental cost-effectiveness ratios, which is the difference in cost divided by the difference in effectiveness). Strategies that were more costly and less effective or had a higher incremental cost-effectiveness ratio than an adjacent strategy were considered dominated (less desirable strategies). Screening women with LBC and reflex HPV testing of ASC-US every 3 years was the least costly strategy compared with no screening. Liquid-based cytology with reflex HPV testing of ASC-US conducted every 2 years was associated with incremental cost-effectiveness ratios ranging from $50,000 to $57,000, depending on whether all enlisted women were being screened or just those in the ranks with an average age younger than 30 years. Liquid-based cytology and reflex HPV testing of ASC-US conducted yearly was not cost-effective, independent of underlying assumptions regarding age, prevalence of underlying disease, and inclusion of fringe benefits. Combined HPV and LBC testing conducted yearly was associated with incremental cost-effectiveness ratios in excess of $1 million regardless of underlying assumptions (Table 5).
The rankings of strategies were essentially unchanged after sensitivity analysis. Inclusion of fringe benefits increased the incremental cost-effectiveness ratios but did not affect the rankings of the strategies. The rankings also remained the same whether restricted to younger women with average age younger than 30 years or all enlisted women, and when the cost of the HPV test was varied (down to a low of $5). However, the results were sensitive to assumptions about underlying prevalence of disease, with lower incremental cost-effectiveness ratios associated with higher prevalence of disease. When the sensitivity of LBC was lowered and the specificity of the combined LBC and HPV testing was lowered to reflect variation in these measures in different populations, the rankings of the strategies changed. However, in all cases, the results were consistent with the base case analysis in that the least costly option remained LBC with reflex HPV testing of ASC-US conducted every 2 to 3 years.
Short-term outcomes evaluated in our model included cancers detected, CIN 2–3 detected, CIN 1 detected, and office visits. When expected office visits were calculated for each strategy, the strategies that were associated with less frequent screening had the fewest associated number of office visits. In addition, use of reflex HPV testing for cases of ASC-US instead of routine HPV testing irrespective of cytologic results (DNA/PAP) decreased requirements for follow-up. The ranking of strategies was consistent with the base case analysis for the detection of high-grade dysplasia only (Table 6). When the detection of cancer or CIN 1 was the outcome of interest, the rankings of the strategies changed from the base case analysis, reflecting the effect of test sensitivity and specificity over time for the detection of CIN 1 and cancer.
Previously, several authors have published data showing that performing more sensitive cervical cancer screening tests at extended frequencies may be more cost-effective than conventional cytology on an annual basis. In particular, combining HPV DNA testing in conjunction with liquid-based cytology offers the promise of improvement in sensitivity for the detection of preinvasive and invasive disease, but at an increased cost if screening interval remains unchanged.9,18 However, due to the high age-related prevalence rate of HPV, this algorithm may not be the most cost-effective algorithm for cervical cancer screening in all age groups. Therefore, the most cost-effective screening program might be one in which different screening methodologies are used in different decades of life. Our results show that algorithms that extend the screening interval can be cost-effective when using tests with high sensitivity and specificity, although we did not find that altering the screening strategy on the basis of age was cost-effective. These conclusions were robust over a wide range of assumptions, including use of fringe benefits in the calculation of indirect costs, varying the prevalence of underlying disease in the population, restricting the analysis to women aged younger than 30 years in addition to all women, and accounting for potential variation in sensitivity and specificity.
Although our reported findings are restricted to enlisted women in the U.S. Army, we also performed preliminary outcomes models (over a 10-year period) for warrant officers and officers. In these models we calculated expected cancers for the 3 groups. Of the 3, more cases of cancer were expected for enlisted women than for warranted officers and officers, a finding consistent with data reported by Ollayos and Peterson,19 who showed that enlisted women have a higher prevalence of cervical abnormalities as detected by cytology than other groups of women in the army. We also estimated the costs and outcomes associated with screening the army as whole by using a weighted average of the costs and outcomes from each group. Because a large proportion of women are enlisted (85%), our findings were consistent with our base case model that includes only enlisted soldiers, which suggests that our results are broadly applicable to all women in the U.S. Army.
Recently investigators have suggested that DNA/ PAP testing performed at 2- to 3-year intervals is cost-effective compared with annual conventional cytology. Human papilloma virus DNA testing in conjunction with cytology in patients aged older than 30 years was more cost-effective when performed at intervals of every 2 to 3 years than annual conventional cytology.18 We did not, however, find that differing strategies on the basis of age alone resulted in more cost-effective screening strategies. In contrast, our data revealed that DNA/PAP testing in women aged older than 30 years was less cost-effective even at extended screening intervals than LBC and reflex HPV testing at the same interval. There are multiple possible explanations for these results in women aged older than 30 years. First, the prevalence of HPV infection is lower in older relative to younger women. Second, women in the Army will have been screened frequently before age 30 years, allowing detection and treatment of preexisting disease. Third, as women progress in age, they also move up in rank. This progression results in higher indirect costs that ultimately translate into higher incremental cost-effectiveness ratios. Fourth, unlike previous models, we modeled multiple cohorts, starting at the ages determined by the Army age distribution, rather than one starting from a single age. A recent study using a model based on this one found that cost-effectiveness estimates were substantially different when a single cohort was simulated compared with a multiple cohort.20 Other differences include use of strategies that are based on LBC only and use of a 3-year screening interval for the LBC and HPV testing for women aged 30 years or older.
Of the screening strategies tested in our model, LBC and reflex HPV at a 2- or 3-year screening interval resulted in incremental cost-effectiveness ratios near or below the $50,000 per life year saved threshold widely used by health care economists. A strategy of LBC and reflex HPV every 3 years would result in 60% fewer cases of CIN 2–3 detected than the same strategy performed on a 2-year interval. Although the percentage difference is large, the actual difference in terms of expected cancers prevented is small and should be considered, given that the 2-year screening interval is associated with a 10-fold increase in cost per life year saved compared with the 3 year interval.
Our study differs from previous studies in that we have explicitly considered time costs associated in our analysis. In previously published articles by Goldie et al18 and Shireman et al,21 time costs were included, based upon previously published data from survey responses and comparisons to national gender- and age-adjusted hourly wages. Although data are available for wages in the civilian sector, the fact that women earn less compared with men in similar positions may lead to an underestimate of the true costs associated with time away from work. One advantage of using military salary data are that salary is based solely on rank, regardless of gender; thus avoiding an underestimate of salary that can sometimes exist for women compared with men. Our time costs are robust in that salaries are well defined, independent of hours worked (assumes a forty hour work week), and time costs for screening are preserved throughout all age and rank groupings. Although we did not model the screening strategies with time costs excluded to estimate the degree of effect of time costs, we would expect that including time costs results in an increase in costs approaching 25% for each strategy21 while not significantly changing the rankings of strategies. In a similar analysis by Maxwell et al9 that did not include indirect costs or simultaneous HPV and cytology testing, LBC with reflex HPV testing of ASC-US on a 2- and 3-year interval were the most cost-effective strategies using a $70,000 threshold (dominated and dissimilar strategies excluded). Our current model is more closely calibrated to actual disease-specific data and population data from the Army. Including time costs in the current model has the effect of making frequent screening even more expensive, and therefore, making less frequent but more sensitive testing strategies even more attractive. Current Army guidelines require that an abnormal Pap result be evaluated before deployment or, if detected in the field, evaluated at a facility capable of management per ASCCP guidelines. If we had included these costs, annual screening would be even less attractive.
It should also be noted that we included some health care costs, particularly the costs of cancer treatment, that in reality might not be borne by the Army. For women who do not remain in the Army, detection and treatment of high-grade CIN while in the Army likely prevents cases that would occur and be treated within the civilian or Veterans Administration health systems. Technically, these costs could be excluded; if they were, they would likely result in even higher cost-effectiveness ratios for all strategies.
Our findings have a significant potential effect on the delivery of health care to active duty service women in the U.S. Army. Current screening guidelines for the U.S. Army incorporate yearly testing in a manner consistent with established ASCCP recommendations. The American College of Obstetrics and Gynecology has endorsed extending the screening interval to every 2 or 3 years for selected low-risk populations.22 Our mathematical model provides an economic rational for implementing such strategies in the military. For the U.S. Army, a strategy of LBC with HPV testing of ASC-US on a 2-year basis seems safe and cost-effective and directly enhances mission readiness by eliminating screening during deployments as well as the resultant evacuation of female soldiers to military treatment facilities for treatment of cervical abnormalities detected “in the field.” Continued evaluation and reassessment as new data appears on the effectiveness of different screening modalities, as well as the potential effect of HPV vaccines,23 will help to define further the optimal strategy for preventing cervical cancer in female soldiers.
Although our findings are specific to the U.S. Army, this study serves as a model that accounts for indirect costs of health care such as salary costs and time away from work costs in cost-effectiveness analysis. These results may be generalizable to health care systems at large and specifically to organizations that both provide health care and are forced to absorb the economic impact of continuing to provide salary benefits to employees who are consuming health care resources.
1.Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin 2005;55:74–108.
2.Jemal A, Murray T, Ward E, Samuels A, Tiwari RC, Ghafoor A, et al. Cancer Statistics, 2005 [published erratum appears in CA Cancer J Clin 2005;55:259]. CA Cancer J Clin 2005; 55:10–30.
3.Gustafsson L, Ponten J, Zack M, Adami HO. International incidence rates of invasive cervical cancer after introduction of cytological screening. Cancer Causes Control 1997;8:755–63.
4.Ries L, Kosary C, Hankey B, Miller BA, Edwards BK. SEER cancer statistics review, 1973–1995. Bethesda (MD): National Cancer Institute; 1998.
5.Smith RA, Cokkinides V, Eyre HJ. American Cancer Society guidelines for the early detection of cancer, 2005. CA Cancer J Clin 2005;55:31–44.
6.Solomon D, Schiffman M, Tarone R, ALTS Study Group. Comparison of three management strategies for patients with atypical squamous cells of undetermined significance: baseline results from a randomized trial. J Natl Cancer Inst 2001;93:293–9.
7.Wright TC Jr, Schiffman M. Adding a test for human papillomavirus DNA to cervical-cancer screening. N Engl J Med 2003;348:489–90.
9.Maxwell GL, Carlson JW, Ochoa M, Krivak T, Rose GS, Myers ER. Costs and effectiveness of alternative strategies for cervical cancer screening in military beneficiaries. Obstet Gynecol 2002;100:740–8.
10.Wright TC Jr, Cox JT, Massad LS, Twiggs LB, Wilkinson EJ, ASCCP-Sponsored Consensus Conference. 2001 Consensus guidelines for the management of women with cervical cytological abnormalities. JAMA 2002;287:2120–9.
11.Solomon D, Davey D, Kurman R, Moriarty A, O'Connor D, Prey M, et al. The 2001 Bethesda System: terminology for reporting results of cervical cytology. JAMA 2002;287:2114–9.
12.Stany MP, Bidus MA, Reed EJ, Kaplan KJ, McHale MT, Rose GS, et al. The prevalence of HR-HPV DNA in ASC-US Pap smears: a military population study. Gynecol Oncol 2005 Nov 9; [Epub ahead of print].
13.Kulasingam SL, Hughes JP, Kiviat NB, Mao C, Weiss NS, Kuypers JM, et al. Evaluation of human papillomavirus testing in primary screening for cervical abnormalities: comparison of sensitivity, specificity, and frequency of referral. JAMA 2002;288:1749–57
14.Kim JJ, Wright TC, Goldie SJ. Cost-effectiveness of alternative triage strategies for atypical squamous cells of undetermined significance. JAMA 2002;287:2382–90.
15.Franco EL. Chapter 13: primary screening of cervical cancer with human papillomavirus tests. J Natl Cancer Inst Monogr 2003;31:89–96.
16.McCrory DC, Matchar DB, Bastian L, Datta S, Hasselblad V, Hickey J, et al. Evaluation of cervical cytology. Evid Rep Technol Assess (Summ) 1999;5:1–6.
18.Goldie SJ, Kim JJ, Wright TC. Cost-effectiveness of human papillomavirus DNA testing for cervical cancer screening in women aged 30 years or more. Obstet Gynecol 2004;103:619–31.
19.Ollayos CW, Peterson M. Relative risks for squamous intraepithelial lesions detected by the Papanicolaou test among Air Force and Army beneficiaries of the Military Health Care System. Mil Med 2002;167:719–22.
20.Dewilde S, Anderson R. The cost-effectiveness of screening programs using single and multiple birth cohort simulations: a comparison using a model of cervical cancer. Med Decis Making 2004;24:486–92.
21.Shireman TI, Tsevat J, Goldie SJ. Time costs associated with cervical cancer screening. Int J Technol Assess Health Care 2001;17:146–52.
22.Cervical cytology screening. Clinical Management Guidelines for Obstetrician–Gynecologists. Number 45. American College of Obstetricians and Gynecologists. Obstet Gynecol 2003;102:417–27.
23.Kulasingam SL, Myers ER. Potential health and economic impact of adding a human papillomavirus vaccine to screening programs. JAMA 2003;290:781–9.
Figure. No caption available.