Optimization of Screening Strategies and Sensitivity Analyses

The focus of this investigation is to compare the effects of different breast cancer screening policies and the costs directly related to these policies, based on the models introduced in the last sections. The health outcome of interest is the expected gain in quality-adjusted survival. We interpret this quality adjustment to be relative to a typical health history rather than that of a state of perfect health [PBW99]. Quality adjustments are important because they allow, with certain limitations, to account for the effects of medical intervention on morbidity as well as mortality. In screening this is especially important becasue of the so-called overdiagnosis problem. While beneficial to many women, screening leads to discovering cancer that would have not otherwise affected certain womens health. While lenght of life may be unaffected, this is a considerable loss of quality of life. Also, early detection can prolong the portion of one's life spent as a cancer survivor. The specific quality adjustments used in our model are the same as Parmigiani [PARM02].

The marginal effectiveness for each screening strategy is calculated based on the difference between the expected quality-adjusted life in years for women in a cohort undergoing screening versus the same cohort of women without screening. The summaries of interest are the expected gain in quality life years (QALYS) and the expected total monetary cost for each screening strategy. Marginal cost is the difference in total cost between the screened and unscreened cohorts. The marginal effectiveness for each screening strategy is the difference between the expected QALYS in the screened and unscreened cohorts. The ratio is marginal cost per year of quality-adjusted life saved (MCYQLS).

Three important issues to consider for screening policies are the age at which a woman should start a screening program, the screening frequency, and what screening modalities are to be used. In this study, we will evaluate a total of 48 screening strategies with the following combinations:

• The age to begin and end periodic screening: 40-79, 45-79, and 50-79 years;

• The interval between consecutive examinations: 0.5, 1, 1.5 and 2 year(s);

• The combined use of MM and CBE: whether mammogram or CBE is given for every one or every two exams.

Using the model described earlier, we generate a cohort of women and their natural histories of disease, and assess how the screening strategies interact

Table 1. Balance sheet for two alternative screening strategies: annual MM and CBE screening and biennial MM and annual CBE. In both cases screening starts at 40 years of age and stops at age 79. Values are increments compared to no screening for a cohort of 10000 breast cancer women.

Screening Strategy MM/1,CBE/1 MM/2,CBE/1

Table 1. Balance sheet for two alternative screening strategies: annual MM and CBE screening and biennial MM and annual CBE. In both cases screening starts at 40 years of age and stops at age 79. Values are increments compared to no screening for a cohort of 10000 breast cancer women.

Screening Strategy MM/1,CBE/1 MM/2,CBE/1

Additional number of MM per woman



Additional number of CBE per woman



Additional number of false positives per woman



Additional years of life saved per woman



Additional women detected in preclinical state



Women treated unnecessarily



with the disease process and the survival after diagnosis. The quantities of interest are estimated using 100,000 Monte Carlo replicates, for each of the screening strategies.

In summary, we simulate a birth cohort of 100,000 women and follow them through the years. A fraction of them will develop breast cancer according to the age-specific incidence of pre-clinical breast cancer. For those women, we generate the natural histories of their disease, which include their ages at the onset of the preclinical disease, the pre-clinical durations (via tumor growth rates), and ages at the clinical onset of the disease. When a screening strategy is provided to a woman during a pre-clinical disease state, the probability that her cancer will be detected by this screening strategy is generated using the equations in Section 2.3, based on her age and tumor size at the time of the exam. If the diagnosis is missed during the exam, her breast cancer may be detected at her next scheduled exam or it may clinically manifest before the next exam depending on the sojourn time of her preclinical disease state. Once a woman is diagnosed to have breast cancer, we obtain her tumor size and age at the time of detection. The information is then used to predict the woman's survival and quality-adjusted survival after the detection using models developed in Section 2.2. The expected cost is estimated based on the average cost of screening exams from the 100,000 women for each screening strategy in the simulation.

A balance sheet is a summary of the expected benefits and harms of an intervention. Its goal is to inform decision makers, and enable them to weigh benefits and harms according to their individual values [MS99, BIG99]. Table 1 is a balance sheet for evaluating two alternative screening strategies, based on the model of this chapter. We consider annual MM and CBE screening (denoted by MM/1, CBS/1) and biennial MM and annual CBE (MM/2, CBE/1). Differences between the two columns can inform decision makers about whether annual or biennial MM are to be preferred once annual CBE is planned. Elmore and colleagues [EBM98] collected data on a retrospective cohort study of breast cancer screening and diagnostic evaluations among 2400

women who were 40 to 69 years old at study entry. False positive results occurred in 6.5% of the mammograms, an estimate that was used here to translate the estimated number of additional tests into estimated false positives. In addition we assume that positive CBE's would be followed by a mammography, that 10% of CBE are false positive, and the two tests are independent of each other. Then the overall false positive number per woman for the 1st strategy is: (0.065+0.1—0.1*0.065)*33 = 5.2; and the overall false positive number per woman for the 2nd strategy is 0.065 * 17 + 0.1 * 33 — 0.1 * 0.065 * 17 = 4.3.

In Section 2.1, three model specification are discussed for the distribution of sojourn times in the preclinical state of the disease. It is of interest to investigate how these different models may impact the QALYs and expected cost of each screening strategy reported by [SP05]. We find that the analyses are fairly robust for the three model assumptions. The marginal QALYS is slightly higher (about 1-2%) for the lognormal model than for the exponential model for a given screening strategy. The relative marginal costs and QALYS among the screening strategies under evaluation are similar for the three model choices.

4 Discussion

Much attention has been focused on the early detection capabilities of new breast cancer screening technologies, including advances in mammography and MRI. The importance of clinical breast examination in breast cancer screening programs seems to be unclear. Even though some recent studies have indicated that regular CBE in addition to MM can be important in the early detection of breast cancer, few studies have investigated the optimal use of both mammography and clinical breast exam to reduce the mortality of breast cancer while balancing the associated burdens and costs to women and to the health care system.

Developing early detection guidelines and making public health policy requires careful consideration of the long-term benefits, costs, and feasibility associated with the screening strategies. In Shen and Parmigiani [SP05], we explore the trade-off between the QALYS and costs related to each screening strategy among several combinations of starting ages of screening, frequencies of screening, and the use of two screening modalities. The study indicates that starting from 40 years of age, a biennial mammogram is often cost-effective for women who undergo annual clinical breast exams. Given the cost to women who are already receiving care for other health issues or regular check-ups in a clinic, an annual CBE as part of their routine examination should not add much burden. Our analyses also indicate that CBE alone cannot replace regular mammography in screening practice, but can be used complementarily or alternatively in a screening program.

The decision analysis methodology and simulation techniques developed for this study can be directly applied to investigate other screening strategies, and even to other chronic diseases with certain modifications to the models. We have modeled screening sensitivity for MM and CBE, respectively, through age and tumor size at diagnosis. We have also introduced random variations for the parameters to incorporate uncertainty of data input and population heterogeneity. We have considered various models and parameters, and have derived them based on data from the large randomized breast cancer screening trials of the HIP [Sha97], CNBSS [MTB97], and the Nijmegen Trial [PVH93]. We have performed sensitivity analyses to assess the robustness of the patterns of benefit and cost with the alternative models.

Our study has several limitations. The cost of a biopsy following a CBE or MM that is positive for breast cancer has not been considered in the analysis. Moreover, we have not included the potential costs of false-positive exams, such as the anxiety, fear and discomfort that are associated with a biopsy. In fact, it is often difficult to convert these factors into dollar amounts [EBM98]. In addition, we have not included important cost components, which are the costs of follow-up procedures undertaken after the detection of breast cancer. This is in part due to the great variation in treatment protocols and in the cost of treating breast cancer that has existed over the years. Finally, we have used a hypothetical birth cohort of women with 100% compliance in the simulations for each screening strategy. In reality, it is rare to have 100% compliance for any screening program, and a real cohort would be dynamic, which would include changes in the cohort due to migration.


The authors thank Professor Marvin Zelen for his encouragement. This work was partially supported by the National Institutes of Health Grants R01 CA-79466 (YS), and by the NCI under the Johns Hopkins SPORE in Breast cancer P50CA88843 (GP).


[WTHR03] Weir HK, Thun MJ, Hankey BF, Ries LA, Lowe HL, Wingo PA, Jemal A, Ward E, Anderson RN, Edwards B.: Annual report to the nation on the status of cancer, 1975-2000, featuring the uses of surveillance data for cancer prevention and control. J Natl Cancer Inst, 95(17):1276-1299, 2003

[SEER98] National Cancer Institute: The surveillance, epidemiology, and end results (seer) program. www.seer.cancer.gov, 1998. [IARC99] Research on Cancer (IARC) IA. The cancer-mondial website.

www.dep.iarc.fr, 1999. [FE03] Fletcher SW; Elmore JG. Mammographic screening for breast cancer. N Engl J Med, 348(17):1672-1680, 2003.

[Wan03] Wang L. Mammography and beyond: Building better breast cancer screening tests. J Natl Cancer Inst, 94(18):1346-1347, 2002.

[Ber9] Berry DA. Benefits and risks of screening mammography for women in their forties: A statistical appraisal. J Natl Cancer Inst, 90:1431-1439, 1998.

[GBC97] Gordis L; Berry D; Chu S; et al. Breast cancer screening for women ages 40-49. J Natl Cancer Inst, 89:1015-1026, 1997.

[CTF01] Canadian Task Force on Preventive Health Care. Preventive health care, 2001 update: screening mammography among women aged 40-49 years at average risk of breast cancer. CMAJ, 164(4):469-476, 2001.

[Par02] Parmigiani G. Measuring uncertainty in complex decision analyses models. Statistical Methods in Medical Research, 11(6):513-37, 2002.

[VBH95] van Oortmarseen G; Boer R; Habbema J. Modeling issues in cancer screening. Statistical Methods in Medical Research, 4:33-54, 1995.

[PARM02] Parmigiani G. Modeling in Medical Decision Making. Wiley, Chichester, 2002.

[Par93] Parmigiani G. On optimal screening ages. J Amer Stat Assoc, 88:622628, 1993.

[Dek00] De Koning H. Breast cancer screening; cost-effectiveness in practice. Eur J Radiol, 33:32-7, 2000.

[LR95] Lindfors KK; Rosenquist CJ. The cost-effectiveness of mammographic screening strategies. J Am Med Assoc, 274:881-884, 1995.

[Bro92] Brown ML. Sensitivity analysis in the cost-effectiveness of breast cancer screening. Cancer, 69 (7 Suppl):1963-1967, 1992.

[MF92] Mushlin AI; Fintor L. Is screening for breast cancer cost-effective? Cancer, 69 (7 Suppl):1957-1962, 1992.

[VVD93] van Ineveld BM; van Oortmarssen GJ; de Koning HJ; Boer R; van der Maas PJ. How cost-effective is breast cancer screening in different EC countries? European Journal of Cancer, 29:1663-1668, 1993.

[BF93] Brown ML; Fintor L. Cost-effectiveness of breast cancer screening: preliminary results of a systematic review of the literature. Breast Cancer Res Treat, 25:113-118, 1993.

[Eli91] Elixhauser A. Costs of breast cancer and the cost-effectiveness of breast cancer screening. Int J Technol Assess Health Care, 7:604-615, 1991.

[Cla92] Clark RA. Economic issues in screening mammography. Am J of Roentgenology, 158:527-534, 1992.

[PK93] Parmigiani G; Kamlet M. Cost-utility analysis of alternative strategies in screening for breast cancer. In C Gatsonis; J Hodges; RE Kass; N Singpurwalla, eds., Case Studies in Bayesian Statistics, 390-402. Springer, New York, 1993.

[EHM89] Eddy DM; Hasselblad V; McGivney W; Hendee W. The value of mammography screening in women under age 50 years. J Am Med Assoc, 259:1512-1519, 1989.

[CGV93] Carter R; Glasziou P; van Oortmarssen G; de Koning H; Stevenson C; Salkeld G; Boer R. Cost-effectiveness of mammographic screening in Australia. Austr J Pub Health, 17:42-50, 1993.

[SKP97] Saltzmann P; Kerlikowske K; Phillips K. Cost-effectiveness of extending screening mammography guidelines to include women 40 to 49 years of age. Ann Intern Med, 127:955-965, 1997.

[BBE01] Burnside E; Belkora J; Esserman L. The impact of alternative practices on the cost and quality of mammographic screening in the United States. Clinical Breast Cancer, 2(2):145-152, 2001.

[Fet01] Fett M. Computer modelling of the Swedish two country trial of mammographic screening and trade offs between participation screening interval. J Med Screen, 8(1):39-45, 2001.

[YRP03] Yasmeen S; Romano P; Pettinger M; Chlebowski R; Robbins J; Lane D; Hendrix S. Frequency and predictive value of a mammographic recommendation for short-interval follow-up. J Natl Cancer Inst, 95(6):429-436, 2003.

[KSR03] Kerlikowske K; Smith-Bindman R; Sickles E. Short-interval follow-up mammography: Are we doing the right thing? J Natl Cancer Inst, 95(6):418-419, 2003.

[BHF99] Barton MB; Harris R; Fletcher SW. Does this patient have breast cancer? The screening clinical breast examination: Should it be done? How? J Am Mec Assoc, 282:1270-80, 1999.

[BMB99] Baines CJ; Miller AB; Bassett AA. Physical examination. its role as a single screening modality in the canadian national breast screening study. Cancer, 63:1816-22, 1989.

[BLT00] Bobo J; Lee N; Thames SF. Findings from 752081 clinical breast examinations reported to a national screening program from 1995 through 1998. J Natl Cancer Inst, 92:971-6, 2000.

[SZ01] Shen Y; Zelen M. Screening sensitivity and sojourn time from breast cancer early detection clinical trials: mammograms and physical examinations. J Clin Oncol, 19:3490-9, 2001.

[SP05] Shen Y; Parmigiani G. A model-based comparison of breast cancer screening strategies: Mammograms and clinical breast examinations. Cancer Epidemiology, Biomarkers and Prevention, in press, 2005.

[MSL79] Moolgavkar SH; Stevens RG; Lee JAH. Effect of age on incidence of breast cancer in females. J Natl Cancer Inst, 62:493-501, 1979.

[PS01] Parmigiani G; Skates S. Estimating the age of onset of detectable asymptomatic cancer. Mathematical and Computer Modeling, 33:13471360, 2001.

[AGL78] Albert A; Gertman P; Louis T. Screening for the early detection of cancer: I. the temporal natural history of a progressive disease state. Mathematical Biosciences, 40:1-59, 1978.

[DW84] Day NE; Walter SD. Simplified models of screening for chronic disease: Estimation procedures from mass screening programmes. Biometrics, 40:1-13, 1984.

[BDM86] Brookmeyer R; Day NE; Moss S. Case-control studies for estimation of the natural history of preclinical disease from screening data. Statistics in Medicine, 5:127-138, 1986.

[ES97] Etzioni RD; Shen Y. Estimating asymptomatic duration in cancer: the AIDS connection. Stat in Med, 16:627-644, 1997.

[SPV97] Straatman H; Peer PG; Verbeek AL. Estimating lead time and sensitivity in a screening program without estimating the incidence in the screened group. Biometrics, 53:217-229, 1997.

[SZ99] Shen Y; Zelen M. Parametric estimation procedures for screening programmes: Stable and nonstable disease models for multimodality case finding. Biometrika, 86:503-515, 1999.

[SGH86] Spratt JS; Greenberg RA; Heuser LS. Geometry, growth rates, and duration of cancer and carcinoma in situ of the breast before detection by screening. Cancer Research, 46:970-974, 1986.

[PVH93] Peer P; van Dijck JAAM; Hendriks J; Holland R; Verbeek ALM. Age-dependent growth rate of primary breast cancer. Cancer, 71:35473551, 1993.

[PVS96] Peer P; Verbeek A; Straatman H; Hendriks J; Holland R. Age-specific sensitivities of mammographic screening for breast cancer. Breast Cancer Research and Treatment, 38:153-160, 1996.

[NIH02] National Cancer Institute: Chemoprevention of Estrogen Receptor (ER) Negative Breast Cancer Preclinical Studies. NIH homepage. http://grants1.nih.gov/grants/guide/rfa-files/RFA-CA-03-005.html, 2002.

[WWT85] Wood W; Weiss R; Tormey D; Holland J; Henry P; Leone L; et al. A randomized trial of CMF versus CMFVP as adjuvant chemotherapy in women with node-positive stage ii breast cancer: a CALGB study. World J Surg, 9:714-718, 1985.

[PNK96] Perloff M; Norton L; Korzun A; Wood W; Carey R; Gottlieb A; et al. Postsurgical adjuvant chemotherapy of stage ii breast carcinoma with or without crossover to a non-cross-resistant regimen: a cancer and leukemia group b study. J Clin Oncol, 14:1589-98, 1996.

[WBK94] Wood W; Budman D; Korzun A; Cooper M; Younger J; Hart R; et al. Dose and dose intensity of adjuvant chemotherapy for stage ii, node-positive breast carcinoma. N Engl J Med, 330:1253-1259, 1994.

[PBW99] Parmigiani G; Berry DA; Winer EP; Tebaldi C; Iglehart JD; Prosnitz L. Is axillary lymph node dissection indicated for early stage breast cancer—a decision analysis. J Clin Oncol, 17(5):1465-1473, 1999.

[SWZ01] Shen Y; Wu D; Zelen M. Testing the independence of two diagnostic tests. Biometrics, 57:1009-1017, 2001.

[KGB98] Kerlikowske K; Grady D; Barclay J; Frankel SD; Ominsky SH; Sickles EA; Ernster V. Variability and accuracy in mammographic interpretation using the american college of radiology breast imaging reporting and data system. J Natl Cancer Inst, 90(23):1801-1809, 1998.

[PRE02] PREVENTION. Cancer tests worth paying for.

http://www.prevention.com/cda/ feature/0,1204,876,00.html,


[MS99] Matchar DB; Samsa GP. Using outcomes data to identify best medical practice: the role of policy models. Hepatology, Jun;29(6 Suppl):36S-39S, 1999.

[BIG99] Barratt A; Irwig L; Glasziou P; et al. Users' guides to the medical literature: XVII. how to use guidelines and recommendations about screening. evidence-based medicine working group. J Am Mec Assoc, 281(21):2029-2034, 1999.

[EBM98] Elmore JG; Barton MB; Moceri VM; Polk S; Arena PJ; Fletcher SW. Ten-year risk of false positive screening mammograms and clinical breast examinations. N Engl J Med, 338:1089-1096, 1998.

[Sha97] Shapiro S. Periodic screening for breast cancer: The HIP randomized controlled trial. Monogr Natl Cancer Inst, 22:27-30, 1997.

[MTB97] Miller A; To T; Baines C; Wall C. The Canadian national breast screening study: Update on breast cancer mortality. Monogr Natl Cancer Inst, 22:37-41, 1997.

0 0

Post a comment