Recent guidelines published by the American College of Obstetricians and Gynecologists1 and the American Cancer Society2 state that once women aged older than 30 years have had 3 or more prior normal cytology tests, screening intervals can be lengthened to every 2 or 3 years. Despite similar recommendations to extend screening intervals published more than 15 years ago, many practitioners continue to screen annually.3 Although continued annual screening of women after several normal cytology tests likely averts a few additional cases of cervical cancer, these benefits must be weighed against the costs and harms incurred by overscreening. Because the prevalence of preinvasive cervical neoplasia decreases as the number of prior normal Pap tests increases,4 repeat testing in a woman with many normal tests substantially increases the likelihood of a false-positive test. These false positives often lead to unnecessary follow-up diagnostic testing, including additional Pap tests, colposcopic biopsies, and sometimes surgical procedures. These interventions are not only costly but also cause needless patient anxiety and discomfort.
In 1990, the U.S. Congress passed the National Breast and Cervical Cancer Mortality Prevention Act, which led to the creation of the National Breast and Cervical Cancer Early Detection Program. The National Breast and Cervical Cancer Early Detection Program is administered by the Centers for Disease Control and Prevention (CDC) and since 1991 has provided clinical breast examinations, mammograms, pelvic examinations, and Pap tests to more than 2 million low-income, underinsured and uninsured women throughout the United States.5 The program is charged with decreasing the incidence of cervical cancer and related mortality in this population. Given that most cervical cancer in the United States is diagnosed among never-screened and rarely screened women, interventions designed to decrease cervical cancer morbidity and mortality among this group deserve high priority. Providing access to screening services to this hard-to-reach population can be logistically difficult and expensive. Thus, the cost and efficacy of clinical practices must be estimated if limited resources are to be allocated efficiently. Cost-effectiveness and cost–utility analyses that incorporate clinically derived estimates of screening outcomes after recent normal cytology tests can identify screening intervals that maximize benefits and minimize potential harms.
Using data on the prevalence of cervical intraepithelial neoplasia (CIN) from the National Breast and Cervical Cancer Early Detection Program, we estimated the incremental cost-effectiveness of various screening intervals in women with a history of normal cervical cytology tests. We also performed cost–utility analyses to estimate the cost per quality-adjusted life year saved associated with each interval.
MATERIALS AND METHODS
The study was conducted under a data-use agreement approved by the institutional review board of the CDC, and approval was obtained from the Committee on Human Research at the University of California, San Francisco. Methods for data collection and preparation are detailed elsewhere.4 Briefly, we estimated the prevalence of CIN among women enrolled in the National Breast and Cervical Cancer Early Detection Program. Data collected for each woman screened included demographic characteristics, screening results, diagnostic procedures, and histologic outcomes. Although data collection forms vary by screening location, local program officials standardize data categories before submitting information to the CDC semiannually.
Cytology tests were read at laboratories throughout the United States, and results were reported using the Bethesda System categories.6 We defined a normal Pap test as one interpreted as “normal” or “infection/reactive changes” and defined “consecutive” as tests performed within 36 months of one another. Abnormal histologic results were classified as: CIN grade 1, grade 2, grade 3, or invasive cancer. We grouped women into 4 screening categories: women with no prior program Pap tests, those with an initial normal Pap test followed by a second consecutive Pap test, those with 2 prior normal Pap tests followed by a third, and those with 3 or more prior normal Pap tests followed by another Pap test. We only counted Pap tests recorded by the Program; some women with several normal Pap tests documented in the Program, however, may have had a history of prior abnormal tests or dysplasia. We also grouped women into 4 categories based on the age at the most recent Program Pap test during the study period (aged younger than 30, 30 to 44, 45 to 59, and 60 to 65 years) and determined the prevalence of CIN for each age group. When the observed CIN prevalence in our population was zero, we estimated the prevalence to be the upper limit of the 99% confidence interval (applicable to 1 category only: CIN grade 2 in women under age 30 with 3 or more prior normal tests).
For women with invasive cancer found by their first program Pap test, we estimated a distribution of prevalent cancer cases of 56% stage I, 24% stage II, 12% stage III, and 8% stage IV based on Surveillance, Epidemiology, and End Results data.7 Whenever women with some history of normal cervical cytology results were considered, we adjusted the stage distribution toward an early-stage disease for incident cancer cases based on findings from a large study of U.S. women in a prepaid health plan.8 We assumed that 90% of these cases were stage I, 5% stage II, 3% stage III, and 2% stage IV.
We used a Markov model that simulates the natural history of cervical cancer in a theoretical cohort of women to estimate lifetime costs and life expectancy of different screening strategies among cohorts of women grouped by age and prior history of documented normal Pap tests. This 20-state Markov model has previously been described in detail.4,7,9 Briefly, a population advances from 1 state to another based on predefined probabilities during simulations. These health states are chosen to reflect the progression from healthy and disease-free through precancer to cancer and death. Women are initially distributed among the different states (well, low-grade dysplasia, high-grade dysplasia, cancer stage I, cancer stage II, cancer stage III, cancer stage IV) based on the prevalence of disease. Each year, women can either remain in the same state, progress to a more advanced disease state, or regress to a less severe disease state. Each year, women can either die from causes other than cervical cancer or have hysterectomies for noncancerous uterine conditions.
Among women with 0, 1, 2, and 3 or more prior normal program Pap tests stratified by the 4 age groups, we compared 3 strategies: conventional Pap test screening performed every 3 years, every 2 years, or every year. We assumed conventional cytology to have a sensitivity of 51% and a specificity of 97% for detecting CIN grade 1 or worse at a threshold of atypical squamous cells of undetermined significance (ASCUS), based on data from large population-based studies corrected for verification bias and on meta-analyses.10,11 We modeled conventional cytology with the assumption that potentially more sensitive strategies, such as liquid-based cytology with human papillomavirus DNA triage of ASCUS12–15 would further increase the number of cancers averted; we examined this assumption in sensitivity analyses. Women with ASCUS cytology results were assumed to undergo repeat testing using conventional cytology with colposcopy used for those with repeated abnormalities (≥ ASCUS). Women with low-grade squamous intraepithelial lesions or greater (≥ LSIL) on cytology were triaged to immediate colposcopy. Colposcopy with cervical biopsy was assumed to have 100% sensitivity and specificity. Treatment of CIN was assumed to be curative, and adherence to screening, follow-up, and treatment was assumed to be 100%. These assumptions were varied in sensitivity analyses. Women were assumed to be screened until age 65 and followed up until age 85. Selected parameters and ranges for sensitivity analyses are presented in Table 1.
We used the mean age of each group as the age at which women within a given cohort (defined by Pap test history and age) enter the model. These “start ages” varied by screening history because ages vary as the number of prior normal Pap tests increases (ie, older women are more likely to have had a longer history of normal Pap tests). To compare differences directly, we selected the “start ages” using the mean ages from the group with no prior Pap tests (ages 22, 40, and 52 years for the first 3 age groups). For the 60 to 65 years age group (mean age, 63 years), we assumed a start age of 60 years, so that we could adequately compare differences in screening interval. The prevalence of disease in each group was estimated as described and used to distribute the cohort across the different states in the model.4 Based on summary evidence, the model estimates that 40% of CIN grade 2 and 3 lesions, if untreated, will progress over 10 years to cancer.7 We assumed that all CIN grade 2 lesions would progress to invasion at the same rate as CIN grade 3 lesions, thereby favoring frequent screening. In sensitivity analyses, we explored the effect of assuming that CIN grade 2 progressed like CIN grade 1.
Base-case direct medical cost estimates and ranges for screening and diagnoses were derived from Medicare and the Medstat MarketScan database7 (Table 1). The latter was used for the base case to account for wide variability in treatment and diagnostic costs. In sensitivity analyses we examined the use of Medicare costs for treatment and diagnosis.7 All costs were adjusted to 2004 U.S. dollars using the medical care component of the consumer price index from the Bureau of Labor statistics.16
Utility analyses incorporate measures of women’s preferences for different health states ranging from perfect health (1.0) to death (0). In the context of cervical cancer screening, these analyses incorporate measures of preferences for avoiding both adverse health outcomes associated with developing cervical cancer as well as adverse outcomes inherent to overscreening (eg, additional interventions, false-positive test results). Little is known about the utilities associated with cervical cancer screening and prevention. Because the states included, and utilities for these states, can substantially affect cost-effectiveness ratios (Kulasingam S, Harper DM, Tosteson AT, Myers ER. Impact of quality of life assumptions on cost-effectiveness of cervical cancer screening. Presented at the 20th International Papillomavirus Conference, Paris, France, October 5–9, 2002), we did not include health-related quality of life in the base case. However, in sensitivity analysis, we examined whether the rankings of the strategies would change if we used cost per quality-adjusted life year. To determine how robust our findings were to assumptions about disease severity, we also included an analysis in which the utilities for cancer and cancer survivors were equivalent to 0, indicating death. Utilities were based on the literature and applied to women diagnosed with cancer or who had false-positive screening results.14,15 Utilities were applied for 5 years for those diagnosed with cancer, a lifetime for cancer survivors who had lived longer than 5 years and 1 month for women with false-positive screening test results.
We calculated incremental cost-effectiveness ratios in which the costs of a strategy for each additional case of cancer detected, divided by the additional savings in life expectancy or quality-adjusted life expectancy, were compared with the next, less costly strategy. Strategies that were more costly and either less effective or less cost-effective than an adjacent strategy were considered to be dominated. We adjusted future costs and life expectancy to current values by discounting them at 3% annually.17 In the base case, we did not include nonmedical costs or quality-of-life measures; analyses were conducted from a payer perspective. In sensitivity analyses, however, we included quality-of-life measures.
As the number of prior normal Pap tests increased, gains in life expectancy decreased and costs per discounted years of life saved increased (Table 2; Fig. 1). As age advanced, cost-effectiveness ratios increased because both disease prevalence and years of life available to be saved decreased. The more cost-effective strategies involved screening younger women with no prior Pap tests or 1 prior normal Pap test at 2- and 3-year intervals. Incremental cost-effectiveness ratios associated with screening every year compared with screening every 2 years all exceeded $100,000 per life year saved, regardless of age and screening history. Among women aged 60 or older, screening at any interval, regardless of history, was associated with incremental cost-effectiveness ratios that ranged from $100,000 to almost $2.5 million (screening every year in women with 3 or more prior normal consecutive Pap tests). Because 44% of women in the National Breast and Cervical Cancer Early Detection Program are aged 45 to 59 years, results in this age group apply to the average program participant.
Efficiency curves depict the tradeoffs in costs as a function of gains in life expectancy. As Figure 1 shows, for women aged 30 to 44 years, the least-expensive strategies for women already screened were to conduct no further screening. Regardless of screening history, as the frequency of screening increased from every 3 years to every year, the lines flattened, indicating high incremental costs associated with diminishing gains in life expectancy. Results were similar for women aged less than 30, 45 to 59, and 60 to 65 years.
We determined the robustness of our findings with sensitivity analyses (analyses available upon request from the authors). The magnitude of the incremental cost-effectiveness ratios remained unchanged when costs were varied over a wide range or Medicare costs were used for diagnosis and treatment instead of Medstat MarketScan data (data not shown). If CIN grade 2 was assumed to progress like CIN grade 1, the incremental cost-effectiveness ratios associated with screening increased, especially if women were screened annually, regardless of screening history. If the prevalence of disease was doubled, the incremental cost-effectiveness ratios decreased, but the magnitude remained similar to the base case. As expected, with a test that is more sensitive but less specific, incremental cost-effectiveness ratios associated with screening interval increased markedly, especially for women screened every 1 or every 2 years. Assuming that colposcopy is only 85% sensitive for detecting CIN grade 2 or 3 lesions and 75% sensitive for CIN grade-1 lesions resulted in decreases in incremental cost-effectiveness ratios but did not change strategy rankings. We also examined whether adherence to screening would affect our results (for women with 3 or more Pap tests) by varying screening adherence from 95% for screening every year, to 85% for screening every 3 years: the magnitude of the incremental cost-effectiveness ratios remained unchanged.
Figure 2 presents the incremental cost-effectiveness ratios associated with the different screening intervals among women aged 30 to 44 years with 3 or more prior, normal, consecutive Paps, using cost per life year or cost per quality-adjusted life year as outcomes. Each group of bars represents different values for the utilities. Including utilities for cancer and false-positive cytology testing minimally changed the magnitude of the incremental cost-effectiveness ratios. Even if the utilities for women with all stages of cancer were set to 0, which is equivalent to death—a conservative estimate favoring frequent screening—the incremental cost-effectiveness ratio for annual screening, compared with screening every 2 years, still exceeded $100,000.
Preventive screening for any condition is rarely, if ever, cost saving. The goal of cost-effectiveness analysis is to identify strategies that use resources most efficiently to reduce morbidity and mortality while minimizing screening harms. In this analysis, we used information from the National Breast and Cervical Cancer Early Detection Program to determine the extent to which the frequency of screening well-screened women affects cost-effectiveness estimates. As the number of prior consecutive normal Pap tests increases, the costs per discounted year of life saved increase substantially. This increase is especially high when screening frequency is taken into account. The most cost-effective strategies for cervical-cancer screening involve screening previously unscreened women younger than 30 years of age every 2 or 3 years and those 30 years of age and older every 3 years.
If a preventive intervention is designed to lengthen life, then considering the cost of this benefit in dollars per life year saved or quality-adjusted life year allows it to be compared with other health interventions also designed to lengthen life if all ratios are determined applying similar methods. In our analysis, screening every year was associated with cost-effectiveness ratios in excess of $100,000 per year of life saved; among women age 45 years and older with a history of normal Pap tests, cost-effectiveness ratios exceeded $1 million per year of life saved. Our findings are consistent with other analyses14,15 showing that screening well-screened women frequently is associated with small gains in life expectancy and high costs, resulting in large associated incremental cost-effectiveness ratios. We extend these findings by quantifying the life-expectancy gains and costs using data from a large racially, ethnically, and geographically diverse U.S. population stratified by age and, uniquely, screening history.
In sensitivity analyses we included extreme utilities for cancer (equivalent to death) to explore whether severe disutility among a relatively small proportion of women who get cancer despite having 3 or more prior normal Pap tests could justify a policy of annual screening. Our finding that annual screening of these women is still associated with large increases in the incremental cost-effectiveness ratios highlights the need for cervical cancer specific utilities. Although the quality-adjusted life year analysis is also limited by a lack of published data on the disutility of a false-positive screening test result, recent findings18,19 of an association between excisional treatment of CIN and adverse pregnancy outcomes suggest that if we could more accurately quantify the negative aspects of overscreening, our conclusions regarding the use of more sensitive and less-specific tests, especially when combined with frequent screening, might differ; the incremental cost-effectiveness ratios associated with frequent screening might be higher because of the disutility associated with having a false-positive screening test result.
A limitation of our analysis is determining how the performance of colposcopy in detecting cervical disease affects the overall number of cancers averted through screening. Findings from the ASCUS-LSIL Triage Study suggest that colposcopy performance may be less than ideal, even under rigorous standards as applied within a clinical trial.13 For our main analysis we assumed that colposcopy had 100% sensitivity for detecting CIN lesions, but we varied this assumption in sensitivity analyses. Among women reporting no prior screening with the National Breast and Cervical Cancer Early Detection Program, screening every year was still associated with an incremental cost-effectiveness ratio in excess of $100,000 per life year saved for women in the youngest age group, a cost that increased to more than $800,000 for women in the older age groups.
We assumed perfect adherence to screening, follow-up and treatment in our initial analysis. In sensitivity analyses, we used estimates of screening adherence from the Behavioral Risk Factor Surveillance System (www.cdc.gov)20 to determine impact of nonadherence. Because a concern with using a less frequent screening interval for women with a history of normal Pap tests would be nonadherence to screening rather than follow-up,21 we varied adherence with screening from 95% (for women screened every year) to 85% (for women screened every 3 years). Lower adherence increased costs (due to more cancer cases) and decreased life-expectancy, resulting in decreased incremental cost-effectiveness ratios; however, the magnitude of the incremental cost-effectiveness ratios remained unchanged.
In conclusion, few analyses have examined the potential harms and costs of frequent screening in well-screened populations. These screening effects can be weighed against benefits by estimating the cost-effectiveness of frequent screening in women with many prior normal tests. Our findings show that less-than-annual screening has advantages for both individual women and health systems. Women screened less often than every year may be spared the unavoidable harms of false-positive testing and the inconvenience of frequent testing. Health systems may use limited resources better by targeting coverage to screen women with little or no prior screening or by improving the follow-up of patients with abnormal results rather than frequently screening women with multiple consecutive normal tests.
1.ACOG Committee on Practice Bulletins. ACOG Practice Bulletin: clinical management guidelines for obstetrician-gynecologists. Number 45, August 2003. Cervical cytology screening (replaces committee opinion 152, March 1995). Obstet Gynecol 2003;102:417–27.
2.Screening for cervical cancer: recommendations and rationale. Am Fam Physician 2003;67:1759–66.
3.Noller KL, Bettes B, Zinberg S, Schulkin J. Cervical cytology screening practices among obstetrician-gynecologists. Obstet Gynecol 2003;102:259–65.
4.Sawaya GF, McConnell KJ, Kulasingam SL, Lawson HW, Kerlikowske K, Melnikow J, et al. Risk of cervical cancer associated with extending the interval between cervical-cancer screenings. N Engl J Med 2003;349:1501–9.
5.Centers for Disease Control and Prevention. National Breast and Cervical Cancer Early Detection Program. Available at: http://www.cdc.gov/cancer/nbccedp
. Retrieved October 9, 2004.
6.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.
7.McCrory DC, Matchar DB, Bastian L, Datta S, Hasselblad V, Hickey J, et al. Evaluation of cervical cytology. Evid Rep Technol Assess (Summ) 1999:1–6.
8.Sawaya GF, Sung HY, Kearney KA, Miller M, Kinney W, Hiatt RA, et al. Advancing age and cervical cancer screening and prognosis. J Am Geriatr Soc 2001;49:1499–504.
9.Myers ER, McCrory DC, Nanda K, Bastian L, Matchar DB. Mathematical model for the natural history of human papillomavirus infection and cervical carcinogenesis. Am J Epidemiol 2000;151:1158–71.
10.Nanda K, McCrory DC, Myers ER, Bastian Hasselblad V, Hickey JD, et al. Accuracy of the Papanicolaou test in screening for and follow-up of cervical cytologic abnormalities: a systematic review. Ann Intern Med 2000;132:810–9.
11.Ratnam S, Franco EL, Ferenczy A. Human papillomavirus testing for primary screening of cervical cancer precursors. Cancer Epidemiol Biomarkers Prev 2000;9:945–51.
12.Arbyn M, Buntinx F, Van Ranst, M, Paraskevaidis E, Martin-Hirsch P, Dillner J. Virologic versus cytologic triage of women with equivocal Pap smears: a meta-analysis of the accuracy to detect high-grade intraepithelial neoplasia. J Natl Cancer Inst 2004;96:280–93.
13.ASCUS-LSIL Triage Study (ALTS) Group. Results of a randomized trial on the management of cytology interpretations of atypical squamous cells of undetermined significance. Am J Obstet Gynecol. 2003;188:1383–92.
14.Mandelblatt JS, Lawrence WF, Womack SM, Jacobson D, Yi B, Hwang YT, et al. Benefits and costs of using HPV testing to screen for cervical cancer. JAMA 2002;287:2372–81.
15.Kim JJ, Wright TC, Goldie SJ. Cost-effectiveness of alternative triage strategies for atypical squamous cells of undetermined significance. JAMA 2002;287:2382–90.
16.U.S. Bureau of Labor Statistics. Consumer Price Index, 2003. Available at: http://www.bls.gov/cpi/home.htm
. Retrieved on September 9, 2004.
17.Gold MR, Siegel JE, Russel LB, Weinstein MC, editors. Cost-effectiveness in health and medicine.
New York (NY): Oxford University Press Inc; 1996.
18.Sadler L, Saftlas A, Wang W, Exeter E, Whittaker J, McCowan L. Treatment for cervical intraepithelial neoplasia and risk of preterm delivery. JAMA 2004;291:2100–6.
19.Samson SL, Bentley JR, Fahey TJ, McKay DJ, Gill GH. The effect of loop electrosurgical excision procedure on future pregnancy outcome. Obstet Gynecol. 2005;105:325–32.
20.Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System. Trends data: nationwide. Available at: http://apps.nccd.cdc.gov/brfss/Trends/trendchart.asp?qkey=10070&state=US
. Retrieved October 7, 2004.
21.Leyden WA, Manos MM, Geiger AM Weinmann, S, Mouchawar J, Bischoff K, et al. Cervical cancer in women with comprehensive health care access: attributable factors in the screening process. J Natl Cancer Inst 2005;97:675–83.