Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Chapter 1: Modeling the Impact of Adjuvant Therapy and Screening Mammography on U.S. Breast Cancer Mortality Between 1975 and 2000: Introduction to the Problem

Chapter 1: Modeling the Impact of Adjuvant Therapy and Screening Mammography on U.S. Breast... BACKGROUND The Cancer Intervention and Surveillance Modeling Network (CISNET) (http://cisnet.cancer.gov) is a consortium of National Cancer Institute (NCI)–sponsored investigators whose focus is modeling the impact of cancer control interventions on population trends in incidence and mortality for breast, prostate, colorectal, and lung cancer. These models are also used to project future trends and to help determine optimal cancer control strategies. Although each investigator has pursued research questions of individual interest, the breast group, consisting of seven principal investigators and their coinvestigators, agreed to collaborate to answer the following question: “What is the Impact of Adjuvant Therapy and Screening Mammography on U.S. Breast Cancer Mortality, 1975–2000”? The idea of this comparative modeling approach was to have each group synthesize existing, and sometimes conflicting, knowledge to build models of breast cancer treatment and screening, using certain specified common population-level inputs. Because there was considerable room for variation in approaches, the desire was to use the palette of modeling to bring into better focus areas of consensus and disagreement. U.S. breast cancer mortality for women was rising slightly until 1990 but since declined 20% through 2000 and continued to decline a total of 23% (about 2.3% per year) through 2002 (1). To better understand why these declines have occurred.(i.e., where we have been) and to target new cancer control strategies (i.e., where we are going), decomposing the population impact of interventions that began in the 1970s and 1980s is important. Population impact represents the final phase of cancer research, because even if interventions have been evaluated in randomized controlled trials (RCTs), its impact in the population setting (effectiveness) may differ from that in a trial setting (efficacy), and the dissemination of the intervention in the target population may be less than complete. The consortium used a common set of assumptions about the dissemination patterns of mammography and adjuvant therapy. Effective multiagent chemotherapy regimens were first introduced in the mid 1970s for premenopausal women with node-positive disease (2). As its safety and efficacy was further established, a series of National Institutes of Health consensus conferences and NCI clinical alerts eventually recommended multiagent chemotherapy in earlier stage disease and for postmenopausal women (3–6). Using data from patterns of care studies drawn from the Surveillance, Epidemiology, and End Results (SEER) population-based registry program of the NCI, Mariotto (7) found that for node-positive women younger than 70, dissemination reached more than 50% by the early 1980s and more than 70% by 2000 (with higher levels for those under age 50). For stage II node-negative and stage I women under age 50 years, usage was low prior to a rapid increase in 1988 triggered by an NCI clinical alert (8) reaching 70 and 40% dissemination, respectively, by 2000. For stage II node-negative women aged 50–69 years, the increase in usage was gradual starting in 1984 but reached 60% by 2000. The use of multiagent chemotherapy has remained limited in patients aged 70 or more years regardless of stage, as well as in stage I patients aged 50–69 years (7). Tamoxifen was introduced in the early 1980s, and its use has grown steadily among estrogen receptor–positive patients, especially among older women. Usage is more than 50% for women aged 50 or more years, regardless of stage. Longer administration of tamoxifen is more effective, and the length of administration has grown from 2 to 5 years, with longer use becoming common by 1990. The combined use of tamoxifen and multiagent chemotherapy is more effective than either modality used individually, and most of the continued increases in the use of multiagent chemotherapy from the late 1980s through the 1990s can be attributed to the use of the combined therapies (7). Meta-analyses by the Early Breast Cancer Trialists' Collaborative Group (EBTCG) place the reduction of the annual odds of death for multiagent chemotherapy at .27, .14, and .08 for women younger than 50, aged 50–59 years, and aged 60–69 years, respectively, regardless of nodal status. (9). The reduction in the annual odds of death for tamoxifen are 0.18 if administered for 2 years and 0.28 if administered for 5 years (10) among women with estrogen receptor–positive tumors. The benefits of tamoxifen only accrue to estrogen receptor–positive women and again are the same regardless of nodal status and age. Multiagent chemotherapy and tamoxifen are posited to act independently of one another. When adjuvant therapy was becoming established, the use of screening mammography became more widespread starting in the early 1980s. Early dissemination was limited by fears of the adverse effects of radiation exposure (11–12). However, these fears faded as more modern equipment was introduced, and in the late 1980s and early 1990s there was a spurt of rapid dissemination in the United States. Dissemination continued in the mid and late 1990s, although not at its prior pace. National Health Interview Surveys (NHIS) indicate that in 1987 only 29% of women aged 40 or more years had received a mammogram in the past 2 years, whereas these rates where 56%, 67%, and 70% by 1992, 1998, and 2000, respectively (13–14). There have been eight randomized controlled trials (RCTs) of mammography started between 1961 and 1980, and reported between 1966 and 1999 (15). Although these trials have provided a rich resource of information, conflicting meta-analyses have attempted to estimate the precise benefit of mammography excluding different trials on the basis of their putative shortcomings (15–16). The U.S. Preventative Task Force (15), which excluded only one trial judged of poor quality (Edinburgh), estimates a 16% (95% confidence interval [CI] = 9% to 23%) reduction in breast cancer mortality, with a 15% (95% CI = 1% to 17%) reduction for women under 50. Olson and Gotzsche (16), as part of a Cochrane review, excluded more trials because of baseline imbalances and postrandomization exclusions and used all-cause mortality rather than breast cancer mortality because of biases potentially caused by unblinded cause-of-death attribution. In the end, they used only three trials (the two Canadian studies and the Malmo study) and concluded that the reduction in all-cause mortality was zero. Fletcher et al. (17) summarizes that “in-depth reviews of the criticisms [raised by Olson and Gotzche] concluded that they do not negate the effectiveness of mammography, especially for women older than 50 years of age,” (p. 1674). There are unlikely to be any further RCTs of mammography, although there will be further follow-up of existing trials. Goodman (18), in an editorial summarizing the controversy surrounding these two meta-analyses, states, “Observational evidence—for example, pathophysiologic research or evidence from studies of population that made their own mammography choices—is given little consideration in the USPTF or Cochrane reviews. This is unfortunate; such evidence can provide useful ancillary information when RCTs are not definitive,” (p. 364). Given the complex dissemination patterns and the debates about the benefits of screening, the role that both of these interventions have played in the observed population mortality declines is unclear. In 2000, Peto and colleagues (19) described mortality declines in the United Kingdom that are even larger than the declines observed in the United States. They posited that because of the suddenness, these declines are attributable chiefly to the way breast cancer is diagnosed and treated rather than to any changes in the causes of breast cancer or the way it is registered. However, the attribution of the decline in the United States remains uncertain. PURPOSE The purpose of this modeling effort is to partition observed mortality trends into components associated with the increased use of adjuvant therapy, mammography, and background changes in underlying risk. A modeling effort of this type serves as a centerpiece for further analysis: 1) targeting future cancer control efforts (e.g., promoting more women to have an initial mammogram versus promoting women to have more regular mammograms), 2) quantifying the mortality impact of reducing health disparities in access to care and use of screening, 3) studying alternative models of the natural history of breast cancer and its impact on the overdiagnosis of disease and the efficacy of screening, and 4) extrapolation of future trends as a function of posited future dissemination patterns and the introduction of new therapies, screening, and prevention modalities. It can help us better understand fundamental relationships between upstream (i.e., screening, treatment, and prevention) and downstream (i.e., mortality) Healthy People 2010 goals (http://www.healthypeople.gov and http://progressreport.cancer.gov). The controversies surrounding mammography have made this a timely topic, especially as we attempt to weigh any mortality gains against its downsides (e.g., overdiagnosis, false positives). Given the uncertainties associated with this issue, the comparative approach supported by CISNET provides an ideal forum for this modeling effort. In a recent article reviewing good practices for decision analytic modeling in health care evaluation, CISNET was applauded for its role in setting up a forum that allows modelers to compare results and articulate reasons for discrepancies (20). Similar comparative modeling efforts have occurred for global warming (21). This effort is similar in spirit to previous efforts that have been conducted in cardiovascular disease (22), which partitioned declines in cardiovascular mortality into components associated with lowering diastolic blood pressure, cholesterol, and smoking rates, as well as improved treatment for acute myocardial infarction and coronary artery disease. Unlike the cardiovascular disease effort, most CISNET models explicitly include a preclinical natural history component (which is especially helpful in modeling the impact of screening). MODEL INPUTS This modeling effort is defined by a prescribed set of calendar years and age groups (i.e., a rectangle in the year x age continuum) and for this analysis we chose 1975–2000 and ages 30–79. All birth cohorts that intersect this box were included, and thus birth cohorts from 1894 to 1970 were modeled (Fig. 1). Although the modelers share a common set of inputs, the basic model structure and parameter estimates are unique to each investigator. The issue of what to make common inputs and what to allow individual investigator free hand in shaping engendered much debate within the consortium. Too much conformity would essentially make this effort seven replicates of the same model, whereas too much variation would have each of the investigators answering different variations of the same question. In the end, the CISNET consortium developed the following common inputs, which are described in this monograph: dissemination of adjuvant therapy [(7) and further adapted in chapter 2 to include estrogen receptor status (23)], mortality from all causes other than breast cancer [chapter 3 (24)], cohort trends in the underlying risk of breast cancer [chapter 4 (25)], dissemination and patterns of usage of screening mammography [(26) and summarized in chapter 5 (27)], and prevalence, stage distribution, survival, and mortality of breast cancer prior to 1975 [chapter 5 (27)] Fig. 1. View largeDownload slide Definition of the Calendar Year-Age Box Modeled for the CISNET Breast Base Case. Fig. 1. View largeDownload slide Definition of the Calendar Year-Age Box Modeled for the CISNET Breast Base Case. Examples of how these inputs are generated can be found at the CISNET web site (http://cisnet.cancer.gov/interfaces). Not all models could accommodate all of these as inputs. Some models had to calibrate to them, whereas others, because of the way the model was formulated, could not incorporate them at all. Model-specific inputs and assumptions included the efficacy of treatment [although most used some form of the results from the meta analyses of the Clinical Trialists Collaborative Group (9,10)], tumor growth rates and metastatic spread, operating characteristics of screening, and postdiagnosis survival by tumor characteristics (Fig. 2). Fig. 2. View largeDownload slide CISNET Breast Cancer Base Question: What is the Impact of Mammography, Adjuvant Therapy, and the Combination on U.S. Breast Cancer Mortality: 1975–2000? Fig. 2. View largeDownload slide CISNET Breast Cancer Base Question: What is the Impact of Mammography, Adjuvant Therapy, and the Combination on U.S. Breast Cancer Mortality: 1975–2000? Each modeler had a free hand in deciding what data sources to use for parameter estimation and calibration. Philosophical differences about the nature of modeling became apparent as this work progressed. Some believed that despite the comprehensive approach to modeling all the major factors influencing mortality trends, calibrating to mortality in such a complex system could introduce more bias than it could possibly eliminate, potentially overfitting the model to the observed data because of missing or misspecified components. Others believed that calibration to mortality is reasonable because of the prior knowledge (either through prior calibration steps or expert opinion) that limits both the number and range of parameters to be considered. An advantage of calibration to observed mortality is the potential to estimate the attenuation of population effectiveness relative to trial efficacy. Cohort trends in the underlying risk of breast cancer (chapter 4) was considered a nuisance parameter in these models in that no attempt was made to explicitly understand which risk factors (such as fewer children, increasing age at first pregnancy, increased use of hormone replacement therapy) contributed to this changing risk. Instead, an age–period–cohort model was fitted to the Connecticut Historical Tumor Registry incidence data from 1940 and SEER incidence data from 1975. Because cohort risk usually represents changing risk factors, and period risk represents the introductions of interventions such as screening or treatment (treatment does not influence incidence), it was assumed that period trends were flat prior to 1982 when systematic programs for breast cancer screening, including mammography, began to take place. Making this assumption allowed us to uniquely estimate the age–period–cohort parameters and provided the modelers estimates of both pre- and postmenopausal cohort risk separated from the period effects of mammography. The model indicates that birth cohorts from about 1900 to 1925 had fairly rapid and similar increases in risk for each successive birth cohort regardless of age. For those under 50 the risk increased moderately for each birth successive birth cohort from about 1925 to 1950 and then declined for subsequent cohorts. For those aged 50 or more years, the risk increased rapidly for each successive cohort after 1925 until about 1950. Data for those aged 50 or more years for those born after 1950 is just starting to accumulate. Follow-up efforts to tie these cohort risks to specific changes in risk factors are under way. The consortium developed dissemination models for both adjuvant therapy (chapter 2) and mammography (chapter 5). The idea behind the treatment model was to be able to simulate treatment choice as a function of calendar year, age, stage, and estrogen receptor status. Similarly, for mammography, a model was developed to simulate the lifetime history of screening mammograms as a function of year of birth, in the absence of breast cancer symptoms or diagnosis, and death from other causes. MODELS The seven breast cancer models spanning a wide range of modeling philosophies and approaches are described individually in chapters 6–12 (28–34). Chapter 13 (35) provides a comparison of the models' structure and function. Feuer et al. (36) provide two useful dimensions of the types of surveillance models used here. The first dimension incorporates microsimulation models at one end of the spectrum, where individuals are run through the model one at a time, where at each transition a random number is generated and individual life histories are generated, to mechanistic or analytic models, where a set of analytically derived equations describe the relationships between key health states and/or tumor growth and metastasis. The University of Texas M. D. Anderson Cancer Center, University of Wisconsin, Georgetown, and Erasmus models could be characterized as microsimulation models; the Dana-Farber model could be characterized as analytic; and the remaining two models (University of Rochester and Stanford) could be described as having some aspects of each. The second dimension of model characterization runs from biologic, where the model goes beyond observable quantities to model the underlying disease onset, growth, and progression of disease, to epidemiologic, where only a portion of the disease process is modeled (usually the observable portion). All the models except M. D. Anderson could be characterized as biologic. MODEL RUNS Final results (i.e., a partitioning of breast cancer mortality trends) are compared across models, which is accomplished through runs of the model that reflect various counterfactual scenarios. In most of the simulation models, random number generation is kept consistent across the runs, so that an individual can be tracked across the runs. This approach allows estimation of various counterfactual quantities such as lead time (i.e., the time between diagnosis from screen detection and the date the person would have been clinically diagnosed in the absence of screening). The six basic runs included the following: (B) background risk—only changes in incidence in the absence of screening, non–breast cancer mortality, and survival trends prior to 1975 could influence mortality. (SB) screening and background risk. (TB) treatment and background risk, including runs for chemotherapy only (TBc), tamoxifen only (TBt), and both (TBct). (TctSB) treatment (with both chemotherapy and tamoxifen), screening, and background. Also, several diagnostic runs were conducted, including the following: (B̄) All six of the basic runs, but with a flat background trend in incidence (at 1975 levels) (S * B and S * B̄) Background/no background with annual screens every year for 100% of women starting in 1976 MODEL COMPARISONS For each calendar year (y) from 1975 to 2000, the measure of effect for a particular intervention is the percent change in mortality from the background run (or alternatively the B̄ run), e.g., {(TctSB(y) – B(y))/B(y)} * 100. The effect of screening and treatment together is generally hypothesized to be less than the effect of sum of them each modeled individually (i.e., negative synergism). This assumption is true because screening diagnoses patients at an earlier stage of disease than they would have been diagnosed with symptoms, and therapy is thought to have the same proportional effect across stages. The absolute mortality benefit of a proportional increase in survival applied to a later-stage case is larger than that same benefit applied to a screen detected earlier stage case. This negative synergism can be measured as follows:  \begin{eqnarray*}&&[{\{}(\mathrm{T}_{\mathrm{ct}}\mathrm{SB}(\mathrm{y}){-}\mathrm{B}(\mathrm{y}))/\mathrm{B}(\mathrm{y}){\}}{\ast}100]\\&&{-}{\{}[{\{}(\mathrm{T}_{\mathrm{ct}}\mathrm{B}(\mathrm{y}){-}\mathrm{B}(\mathrm{y}))/\mathrm{B}(\mathrm{y}){\}}{\ast}100]{+}[{\{}(\mathrm{SB}(\mathrm{y}){-}\mathrm{B}(\mathrm{y}))/\mathrm{B}(\mathrm{y}){\}}{\ast}100]{\}}\\&&{=}{\{}(\mathrm{T}\mathrm{ct}\mathrm{SB}(\mathrm{y}){-}\mathrm{T}\mathrm{ct}\mathrm{B}(\mathrm{y}){-}\mathrm{SB}(\mathrm{y}){+}\mathrm{B}(\mathrm{y}))/\mathrm{B}(\mathrm{y}){\}}{\ast}100\end{eqnarray*} A useful summary of various approaches to presenting model results as a function of model runs can be found in Feuer et al. (25). Results were compared in a series of semiannual meetings, and results generally became more concordant over time as investigators searched for errors, used new data sources, and examined their model structure and assumptions. Despite these discussions and comparisons, there is a large variation in the basic approaches used, as well as some variation in results. To help understand the differences and similarities across models, a set of intermediate outcomes were developed, including incidence, lead time, overdiagnosis, detection rates, program sensitivity, changes in stage distribution over time. Also, the CISNET collaborative group has developed a state-of-the-art interactive Web site, called the Model Profiler, which allows modelers to put components of their model into templates to facilitate comparisons of model structure. Because the core documentation format is the same for each group, the published profile information is readily comparable among models. The model profiling system exists to support a framework in which modelers describe their models, and consumers read about, compare, and contrast simulation models. The CISNET Model Profiler provides a platform for managing both the content and the form of complex model documentation. A public version of the model profiler containing information on all seven breast models is available on the CISNET public site (http://cisnet.cancer.gov/profiles) . MONOGRAPH OUTLINE The monograph is divided into three basic sections. Section I [chapters 2–5 (23–25,27)] describes the common inputs. Section II [chapters 6–12 (28–34)] includes a description of each model, including its structure, data sources used for parameters estimation, calibration, and validation, and individual modeling results and interpretation (with any sensitivity analyses). Section III [chapters 13–16 (35,37–39)] starts with a chapter comparing the seven models, highlighting differences in basic assumptions, modeling strategies, and use of the common inputs. The next three chapters provide a head-to-head comparison of intermediate results, mortality results, and a final chapter with conclusions and discussion. We think that this comparative modeling effort will add an important perspective in understanding the impact of cancer control interventions for breast cancer that have occurred over the last 25 years. More generally, we hope that our systematic approach will serve as a milestone in the use of simulation in health care. References (1) Ries LA, Eisner MP, Kosary CL, Hankey BF, Miller BA, Clegg L, et al. (eds). SEER Cancer Statistics Review, 1975–2002. Bethesda (MD): National Cancer Institute. Available at: http://seer.cancer.gov/csr/1975_2002/. Based on November 2004 SEER data submission, posted to the SEER Web site, 2005. Google Scholar (2) Bonadonna G, Brusamolino E, Valagussa P, Rossi A, Brugnatelli L, Brambilla C, et al. Combination chemotherapy as an adjuvant treatment in operable breast cancer. N Engl J Med  1976; 294: 405–10. Google Scholar (3) National Institutes of Health Consensus Development Conference Statement: Adjuvant Chemotherapy for Breast Cancer. September 9–11, 1985. CA Cancer J Clin  1986; 36: 42–7. Google Scholar (4) NCI alert on node-negative breast cancer. Breast Cancer Res Treat  1988; 12: 3–5. Google Scholar (5) NIH Consensus Conference. Treatment of early-stage breast cancer. JAMA  1991; 265: 391–5. Google Scholar (6) National Institutes of Health Consensus Development Conference Statement: Adjuvant Therapy for Breast Cancer, November 1–3, 2000. J Natl Cancer Inst Monogr  2001; 30: 5–15. Google Scholar (7) Mariotto A, Feuer EJ, Harlan LC, Wun LM, Johnson KA, Abrams J. Trends in use of adjuvant multi-agent chemotherapy and tamoxifen for breast cancer in the United States: 1975–1999. J Natl Cancer Inst  2002; 94: 1626–34. Google Scholar (8) Johnson TP, Ford L, Warnecke RB, Nayfield SG, Kaluzny A, Cutter G, et al. Effect of a National Cancer Institute Clinical Alert on breast cancer practice patterns. J Clin Oncol  1994; 12: 1783–8. Google Scholar (9) Polychemotherapy for early breast cancer: an overview of the randomized trials Early Breast Cancer Trialists' Collaborative Group. Lancet  1998; 352: 930–42. Google Scholar (10) Tamoxifen for early breast cancer: an overview of the randomized trials. Early Breast Cancer Trialists' Collaborative Group. Lancet  1998; 351: 1451–67. Google Scholar (11) Howard J. Using mammography for cancer control: an unrealized potential. CA Cancer J Clin  1987; 37: 33–48. Google Scholar (12) Bailer JC Mammography: a contrary view. Ann Intern Med  1976; 84: 77–84. Google Scholar (13) Breen N, Wagener DK, Brown ML, Davis WW, Ballard-Barbash R. Related progress in cancer screening over a decade: results of cancer screening from the 1987, 1992, and 1998 National Health Interview Surveys. J Natl Cancer Inst  2001; 93: 1704–13. Google Scholar (14) Swan J, Breen N, Coates RJ, Rimer BK, Lee NC. Progress in cancer screening practices in the United States: results from the 2000 National Health Interview Survey. Cancer  2003; 97: 1528–40. Google Scholar (15) Humphrey LL, Helfand M, Chan BK, Woolf SH. Breast cancer screening: a summary of the evidence for the U.S. Preventative Task Force. Ann Intern Med  2002; 137: 347–60. Google Scholar (16) Olsen O, Gotzche PC. Cochrane review on screening for breast cancer with mammography. Lancet  2001; 358: 1340–2. Google Scholar (17) Fletcher SW, Elmore JG. Mammographic screening for breast cancer. N Engl J Med  2003; 348: 1672–80. Google Scholar (18) Goodman, SN. The mammography dilemma: a crisis for evidence-based medicine? Ann Intern Med  2002; 137: 363–5. Google Scholar (19) Peto R, Boreham J, Clarke M, Davies C, Beral V. UK and USA breast cancer deaths down 25% in year 2000 at ages 20–69 years. Lancet  2000; 355: 1822. Google Scholar (20) Weinstein MC, O'Brien B, Hornberger J, Jackson J, Johannesson M, McCabe C, et al.; ISPOR Task Force on Good Research Practices—Modeling Studies. Value Health  2003; 6: 9–17. Google Scholar (21) The Costs of the Kyoto Protocol: A multi-model evaluation. The Energy Journal Special Issue. Weyant JP, Jacoby H, Edmunds J, Richels R (eds.); 1999. Google Scholar (22) Hunink MG, Goldman L, Tosteson AN, Mittleman MA, Goldman PA, Williams LW, et al. The recent decline in mortality from coronary heart disease, 1980–1990. The effect of secular trends in risk factors and treatment. JAMA  1997; 277: 535–42. Google Scholar (23) Mariotto AB, Feuer EJ, Harlan LC, Abrams J. Dissemination of adjuvant multiagent chemotherapy and tamoxifen for breast cancer in the United States using estrogen receptor information: 1975–1999. J Natl Cancer Inst Monogr  2006; 36: 7–15. Google Scholar (24) Rosenberg MA. Competing risks to breast cancer mortality. J Natl Cancer Inst Monogr  2006; 36: 15–9. Google Scholar (25) Holford TR, Cronin KA, Mariotto AB, Feuer EJ. Changing patterns in breast cancer incidence trends. J Natl Cancer Inst Monogr  2006; 36: 19–25. Google Scholar (26) Cronin KA, Yu B, Krapcho M, Miglioretti DL, Fay MP, Izmirlian G, et al. Modeling the dissemination of mammography in the United States. Cancer Causes Control  2005; 16: 701–12. Google Scholar (27) Cronin KA, Mariotto AB, Clarke LD, Feuer EJ. Additional common inputs for analyzing impact of adjuvant therapy and mammography on U.S. mortality. J Natl Cancer Inst Monogr  2006; 36: 26–9. Google Scholar (28) Berry DA, Inoue L, Shen Y, Venier J, Cohen D, Bondy M, et al. Modeling the impact of treatment and screening on U.S. breast cancer mortality: a Bayesian approach. J Natl Cancer Inst Monogr  2006; 36: 30–6. Google Scholar (29) Fryback DG, Stout NK, Rosenberg MA, Trentham-Dietz A, Kuruchittham V, Remington PL. The Wisconsin breast cancer epidemiology simulation model. J Natl Cancer Inst Monogr  2006; 36: 37–47. Google Scholar (30) Mandelblatt J, Schechter CB, Lawrence W, Yi B, Cullen J. The SPECTRUM population model of the impact of screening and treatment on U.S. breast cancer trends from 1975 to 2000: principles and practice of the model methods. J Natl Cancer Inst Monogr  2006; 36: 47–55. Google Scholar (31) Tan SYGL, van Oortmarssen GJ, de Koning HJ, Boer R, Habbema JDF. The MISCAN-Fadia continuous tumor growth model for breast cancer. J Natl Cancer Inst Monogr  2006; 36: 56–65. Google Scholar (32) Hanin LG, Miller A, Zorin AV, Yakovlev AY. The University of Rochester model of breast cancer detection and survival. J Natl Cancer Inst Monogr  2006; 36: 66–78. Google Scholar (33) Lee S, Zelen M. A stochastic model for predicting the mortality of breast cancer. J Natl Cancer Inst Monogr  2006; 36: 79–86. Google Scholar (34) Plevritis SK, Sigal BM, Salzman P, Rosenberg J, Glynn P. A stochastic simulation model of U.S. breast cancer mortality trends from 1975 to 2000. J Natl Cancer Inst Monogr  2006; 36: 86–95. Google Scholar (35) Clarke LD, Plevritis SK, Boer R, Cronin KA, Feuer EJ. A comparative review of cisnet breast models used to analyze U.S. Breast cancer incidence and mortality trends. J Natl Cancer Inst Monogr  2006; 36: 96–105. Google Scholar (36) Feuer EJ, Etzioni R, Cronin KA, Mariotto A. The use of modeling to understand the impact of screening on U.S. mortality: examples from mammography and PSA testing. Stat Methods Med Res  2004; 13: 421–42. Google Scholar (37) Habbema JDF, Tan SYGL, Cronin KA. Impact of mammography on U.S. breast cancer mortality, 1975–2000: are intermediate outcome measures informative? J Natl Cancer Inst Monogr  2006; 36: 105–11. Google Scholar (38) Cronin KA, Feuer EJ, Clarke LD, Plevritis SK. Impact of adjuvant therapy and mammography on U.S. mortality from 1975 to 2000: comparison of mortality results from the cisnet breast cancer base case analysis. J Natl Cancer Inst Monogr  2006; 36: 112–21. Google Scholar (39) Habbema JDF, Schechter CB, Cronin KA, Clarke LD, Feuer EJ. Modeling cancer natural history, epidemiology, and control: reflections on the CISNET breast group experience. J Natl Cancer Inst Monogr  2006; 36: 122–6. Google Scholar © The Author 2006. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png JNCI Monographs Oxford University Press

Chapter 1: Modeling the Impact of Adjuvant Therapy and Screening Mammography on U.S. Breast Cancer Mortality Between 1975 and 2000: Introduction to the Problem

JNCI Monographs , Volume 2006 (36) – Oct 1, 2006

Loading next page...
 
/lp/oxford-university-press/chapter-1-modeling-the-impact-of-adjuvant-therapy-and-screening-Af9z2HZPJD
Publisher
Oxford University Press
Copyright
© The Author 2006. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org.
ISSN
1052-6773
eISSN
1745-6614
DOI
10.1093/jncimonographs/lgj002
pmid
17032887
Publisher site
See Article on Publisher Site

Abstract

BACKGROUND The Cancer Intervention and Surveillance Modeling Network (CISNET) (http://cisnet.cancer.gov) is a consortium of National Cancer Institute (NCI)–sponsored investigators whose focus is modeling the impact of cancer control interventions on population trends in incidence and mortality for breast, prostate, colorectal, and lung cancer. These models are also used to project future trends and to help determine optimal cancer control strategies. Although each investigator has pursued research questions of individual interest, the breast group, consisting of seven principal investigators and their coinvestigators, agreed to collaborate to answer the following question: “What is the Impact of Adjuvant Therapy and Screening Mammography on U.S. Breast Cancer Mortality, 1975–2000”? The idea of this comparative modeling approach was to have each group synthesize existing, and sometimes conflicting, knowledge to build models of breast cancer treatment and screening, using certain specified common population-level inputs. Because there was considerable room for variation in approaches, the desire was to use the palette of modeling to bring into better focus areas of consensus and disagreement. U.S. breast cancer mortality for women was rising slightly until 1990 but since declined 20% through 2000 and continued to decline a total of 23% (about 2.3% per year) through 2002 (1). To better understand why these declines have occurred.(i.e., where we have been) and to target new cancer control strategies (i.e., where we are going), decomposing the population impact of interventions that began in the 1970s and 1980s is important. Population impact represents the final phase of cancer research, because even if interventions have been evaluated in randomized controlled trials (RCTs), its impact in the population setting (effectiveness) may differ from that in a trial setting (efficacy), and the dissemination of the intervention in the target population may be less than complete. The consortium used a common set of assumptions about the dissemination patterns of mammography and adjuvant therapy. Effective multiagent chemotherapy regimens were first introduced in the mid 1970s for premenopausal women with node-positive disease (2). As its safety and efficacy was further established, a series of National Institutes of Health consensus conferences and NCI clinical alerts eventually recommended multiagent chemotherapy in earlier stage disease and for postmenopausal women (3–6). Using data from patterns of care studies drawn from the Surveillance, Epidemiology, and End Results (SEER) population-based registry program of the NCI, Mariotto (7) found that for node-positive women younger than 70, dissemination reached more than 50% by the early 1980s and more than 70% by 2000 (with higher levels for those under age 50). For stage II node-negative and stage I women under age 50 years, usage was low prior to a rapid increase in 1988 triggered by an NCI clinical alert (8) reaching 70 and 40% dissemination, respectively, by 2000. For stage II node-negative women aged 50–69 years, the increase in usage was gradual starting in 1984 but reached 60% by 2000. The use of multiagent chemotherapy has remained limited in patients aged 70 or more years regardless of stage, as well as in stage I patients aged 50–69 years (7). Tamoxifen was introduced in the early 1980s, and its use has grown steadily among estrogen receptor–positive patients, especially among older women. Usage is more than 50% for women aged 50 or more years, regardless of stage. Longer administration of tamoxifen is more effective, and the length of administration has grown from 2 to 5 years, with longer use becoming common by 1990. The combined use of tamoxifen and multiagent chemotherapy is more effective than either modality used individually, and most of the continued increases in the use of multiagent chemotherapy from the late 1980s through the 1990s can be attributed to the use of the combined therapies (7). Meta-analyses by the Early Breast Cancer Trialists' Collaborative Group (EBTCG) place the reduction of the annual odds of death for multiagent chemotherapy at .27, .14, and .08 for women younger than 50, aged 50–59 years, and aged 60–69 years, respectively, regardless of nodal status. (9). The reduction in the annual odds of death for tamoxifen are 0.18 if administered for 2 years and 0.28 if administered for 5 years (10) among women with estrogen receptor–positive tumors. The benefits of tamoxifen only accrue to estrogen receptor–positive women and again are the same regardless of nodal status and age. Multiagent chemotherapy and tamoxifen are posited to act independently of one another. When adjuvant therapy was becoming established, the use of screening mammography became more widespread starting in the early 1980s. Early dissemination was limited by fears of the adverse effects of radiation exposure (11–12). However, these fears faded as more modern equipment was introduced, and in the late 1980s and early 1990s there was a spurt of rapid dissemination in the United States. Dissemination continued in the mid and late 1990s, although not at its prior pace. National Health Interview Surveys (NHIS) indicate that in 1987 only 29% of women aged 40 or more years had received a mammogram in the past 2 years, whereas these rates where 56%, 67%, and 70% by 1992, 1998, and 2000, respectively (13–14). There have been eight randomized controlled trials (RCTs) of mammography started between 1961 and 1980, and reported between 1966 and 1999 (15). Although these trials have provided a rich resource of information, conflicting meta-analyses have attempted to estimate the precise benefit of mammography excluding different trials on the basis of their putative shortcomings (15–16). The U.S. Preventative Task Force (15), which excluded only one trial judged of poor quality (Edinburgh), estimates a 16% (95% confidence interval [CI] = 9% to 23%) reduction in breast cancer mortality, with a 15% (95% CI = 1% to 17%) reduction for women under 50. Olson and Gotzsche (16), as part of a Cochrane review, excluded more trials because of baseline imbalances and postrandomization exclusions and used all-cause mortality rather than breast cancer mortality because of biases potentially caused by unblinded cause-of-death attribution. In the end, they used only three trials (the two Canadian studies and the Malmo study) and concluded that the reduction in all-cause mortality was zero. Fletcher et al. (17) summarizes that “in-depth reviews of the criticisms [raised by Olson and Gotzche] concluded that they do not negate the effectiveness of mammography, especially for women older than 50 years of age,” (p. 1674). There are unlikely to be any further RCTs of mammography, although there will be further follow-up of existing trials. Goodman (18), in an editorial summarizing the controversy surrounding these two meta-analyses, states, “Observational evidence—for example, pathophysiologic research or evidence from studies of population that made their own mammography choices—is given little consideration in the USPTF or Cochrane reviews. This is unfortunate; such evidence can provide useful ancillary information when RCTs are not definitive,” (p. 364). Given the complex dissemination patterns and the debates about the benefits of screening, the role that both of these interventions have played in the observed population mortality declines is unclear. In 2000, Peto and colleagues (19) described mortality declines in the United Kingdom that are even larger than the declines observed in the United States. They posited that because of the suddenness, these declines are attributable chiefly to the way breast cancer is diagnosed and treated rather than to any changes in the causes of breast cancer or the way it is registered. However, the attribution of the decline in the United States remains uncertain. PURPOSE The purpose of this modeling effort is to partition observed mortality trends into components associated with the increased use of adjuvant therapy, mammography, and background changes in underlying risk. A modeling effort of this type serves as a centerpiece for further analysis: 1) targeting future cancer control efforts (e.g., promoting more women to have an initial mammogram versus promoting women to have more regular mammograms), 2) quantifying the mortality impact of reducing health disparities in access to care and use of screening, 3) studying alternative models of the natural history of breast cancer and its impact on the overdiagnosis of disease and the efficacy of screening, and 4) extrapolation of future trends as a function of posited future dissemination patterns and the introduction of new therapies, screening, and prevention modalities. It can help us better understand fundamental relationships between upstream (i.e., screening, treatment, and prevention) and downstream (i.e., mortality) Healthy People 2010 goals (http://www.healthypeople.gov and http://progressreport.cancer.gov). The controversies surrounding mammography have made this a timely topic, especially as we attempt to weigh any mortality gains against its downsides (e.g., overdiagnosis, false positives). Given the uncertainties associated with this issue, the comparative approach supported by CISNET provides an ideal forum for this modeling effort. In a recent article reviewing good practices for decision analytic modeling in health care evaluation, CISNET was applauded for its role in setting up a forum that allows modelers to compare results and articulate reasons for discrepancies (20). Similar comparative modeling efforts have occurred for global warming (21). This effort is similar in spirit to previous efforts that have been conducted in cardiovascular disease (22), which partitioned declines in cardiovascular mortality into components associated with lowering diastolic blood pressure, cholesterol, and smoking rates, as well as improved treatment for acute myocardial infarction and coronary artery disease. Unlike the cardiovascular disease effort, most CISNET models explicitly include a preclinical natural history component (which is especially helpful in modeling the impact of screening). MODEL INPUTS This modeling effort is defined by a prescribed set of calendar years and age groups (i.e., a rectangle in the year x age continuum) and for this analysis we chose 1975–2000 and ages 30–79. All birth cohorts that intersect this box were included, and thus birth cohorts from 1894 to 1970 were modeled (Fig. 1). Although the modelers share a common set of inputs, the basic model structure and parameter estimates are unique to each investigator. The issue of what to make common inputs and what to allow individual investigator free hand in shaping engendered much debate within the consortium. Too much conformity would essentially make this effort seven replicates of the same model, whereas too much variation would have each of the investigators answering different variations of the same question. In the end, the CISNET consortium developed the following common inputs, which are described in this monograph: dissemination of adjuvant therapy [(7) and further adapted in chapter 2 to include estrogen receptor status (23)], mortality from all causes other than breast cancer [chapter 3 (24)], cohort trends in the underlying risk of breast cancer [chapter 4 (25)], dissemination and patterns of usage of screening mammography [(26) and summarized in chapter 5 (27)], and prevalence, stage distribution, survival, and mortality of breast cancer prior to 1975 [chapter 5 (27)] Fig. 1. View largeDownload slide Definition of the Calendar Year-Age Box Modeled for the CISNET Breast Base Case. Fig. 1. View largeDownload slide Definition of the Calendar Year-Age Box Modeled for the CISNET Breast Base Case. Examples of how these inputs are generated can be found at the CISNET web site (http://cisnet.cancer.gov/interfaces). Not all models could accommodate all of these as inputs. Some models had to calibrate to them, whereas others, because of the way the model was formulated, could not incorporate them at all. Model-specific inputs and assumptions included the efficacy of treatment [although most used some form of the results from the meta analyses of the Clinical Trialists Collaborative Group (9,10)], tumor growth rates and metastatic spread, operating characteristics of screening, and postdiagnosis survival by tumor characteristics (Fig. 2). Fig. 2. View largeDownload slide CISNET Breast Cancer Base Question: What is the Impact of Mammography, Adjuvant Therapy, and the Combination on U.S. Breast Cancer Mortality: 1975–2000? Fig. 2. View largeDownload slide CISNET Breast Cancer Base Question: What is the Impact of Mammography, Adjuvant Therapy, and the Combination on U.S. Breast Cancer Mortality: 1975–2000? Each modeler had a free hand in deciding what data sources to use for parameter estimation and calibration. Philosophical differences about the nature of modeling became apparent as this work progressed. Some believed that despite the comprehensive approach to modeling all the major factors influencing mortality trends, calibrating to mortality in such a complex system could introduce more bias than it could possibly eliminate, potentially overfitting the model to the observed data because of missing or misspecified components. Others believed that calibration to mortality is reasonable because of the prior knowledge (either through prior calibration steps or expert opinion) that limits both the number and range of parameters to be considered. An advantage of calibration to observed mortality is the potential to estimate the attenuation of population effectiveness relative to trial efficacy. Cohort trends in the underlying risk of breast cancer (chapter 4) was considered a nuisance parameter in these models in that no attempt was made to explicitly understand which risk factors (such as fewer children, increasing age at first pregnancy, increased use of hormone replacement therapy) contributed to this changing risk. Instead, an age–period–cohort model was fitted to the Connecticut Historical Tumor Registry incidence data from 1940 and SEER incidence data from 1975. Because cohort risk usually represents changing risk factors, and period risk represents the introductions of interventions such as screening or treatment (treatment does not influence incidence), it was assumed that period trends were flat prior to 1982 when systematic programs for breast cancer screening, including mammography, began to take place. Making this assumption allowed us to uniquely estimate the age–period–cohort parameters and provided the modelers estimates of both pre- and postmenopausal cohort risk separated from the period effects of mammography. The model indicates that birth cohorts from about 1900 to 1925 had fairly rapid and similar increases in risk for each successive birth cohort regardless of age. For those under 50 the risk increased moderately for each birth successive birth cohort from about 1925 to 1950 and then declined for subsequent cohorts. For those aged 50 or more years, the risk increased rapidly for each successive cohort after 1925 until about 1950. Data for those aged 50 or more years for those born after 1950 is just starting to accumulate. Follow-up efforts to tie these cohort risks to specific changes in risk factors are under way. The consortium developed dissemination models for both adjuvant therapy (chapter 2) and mammography (chapter 5). The idea behind the treatment model was to be able to simulate treatment choice as a function of calendar year, age, stage, and estrogen receptor status. Similarly, for mammography, a model was developed to simulate the lifetime history of screening mammograms as a function of year of birth, in the absence of breast cancer symptoms or diagnosis, and death from other causes. MODELS The seven breast cancer models spanning a wide range of modeling philosophies and approaches are described individually in chapters 6–12 (28–34). Chapter 13 (35) provides a comparison of the models' structure and function. Feuer et al. (36) provide two useful dimensions of the types of surveillance models used here. The first dimension incorporates microsimulation models at one end of the spectrum, where individuals are run through the model one at a time, where at each transition a random number is generated and individual life histories are generated, to mechanistic or analytic models, where a set of analytically derived equations describe the relationships between key health states and/or tumor growth and metastasis. The University of Texas M. D. Anderson Cancer Center, University of Wisconsin, Georgetown, and Erasmus models could be characterized as microsimulation models; the Dana-Farber model could be characterized as analytic; and the remaining two models (University of Rochester and Stanford) could be described as having some aspects of each. The second dimension of model characterization runs from biologic, where the model goes beyond observable quantities to model the underlying disease onset, growth, and progression of disease, to epidemiologic, where only a portion of the disease process is modeled (usually the observable portion). All the models except M. D. Anderson could be characterized as biologic. MODEL RUNS Final results (i.e., a partitioning of breast cancer mortality trends) are compared across models, which is accomplished through runs of the model that reflect various counterfactual scenarios. In most of the simulation models, random number generation is kept consistent across the runs, so that an individual can be tracked across the runs. This approach allows estimation of various counterfactual quantities such as lead time (i.e., the time between diagnosis from screen detection and the date the person would have been clinically diagnosed in the absence of screening). The six basic runs included the following: (B) background risk—only changes in incidence in the absence of screening, non–breast cancer mortality, and survival trends prior to 1975 could influence mortality. (SB) screening and background risk. (TB) treatment and background risk, including runs for chemotherapy only (TBc), tamoxifen only (TBt), and both (TBct). (TctSB) treatment (with both chemotherapy and tamoxifen), screening, and background. Also, several diagnostic runs were conducted, including the following: (B̄) All six of the basic runs, but with a flat background trend in incidence (at 1975 levels) (S * B and S * B̄) Background/no background with annual screens every year for 100% of women starting in 1976 MODEL COMPARISONS For each calendar year (y) from 1975 to 2000, the measure of effect for a particular intervention is the percent change in mortality from the background run (or alternatively the B̄ run), e.g., {(TctSB(y) – B(y))/B(y)} * 100. The effect of screening and treatment together is generally hypothesized to be less than the effect of sum of them each modeled individually (i.e., negative synergism). This assumption is true because screening diagnoses patients at an earlier stage of disease than they would have been diagnosed with symptoms, and therapy is thought to have the same proportional effect across stages. The absolute mortality benefit of a proportional increase in survival applied to a later-stage case is larger than that same benefit applied to a screen detected earlier stage case. This negative synergism can be measured as follows:  \begin{eqnarray*}&&[{\{}(\mathrm{T}_{\mathrm{ct}}\mathrm{SB}(\mathrm{y}){-}\mathrm{B}(\mathrm{y}))/\mathrm{B}(\mathrm{y}){\}}{\ast}100]\\&&{-}{\{}[{\{}(\mathrm{T}_{\mathrm{ct}}\mathrm{B}(\mathrm{y}){-}\mathrm{B}(\mathrm{y}))/\mathrm{B}(\mathrm{y}){\}}{\ast}100]{+}[{\{}(\mathrm{SB}(\mathrm{y}){-}\mathrm{B}(\mathrm{y}))/\mathrm{B}(\mathrm{y}){\}}{\ast}100]{\}}\\&&{=}{\{}(\mathrm{T}\mathrm{ct}\mathrm{SB}(\mathrm{y}){-}\mathrm{T}\mathrm{ct}\mathrm{B}(\mathrm{y}){-}\mathrm{SB}(\mathrm{y}){+}\mathrm{B}(\mathrm{y}))/\mathrm{B}(\mathrm{y}){\}}{\ast}100\end{eqnarray*} A useful summary of various approaches to presenting model results as a function of model runs can be found in Feuer et al. (25). Results were compared in a series of semiannual meetings, and results generally became more concordant over time as investigators searched for errors, used new data sources, and examined their model structure and assumptions. Despite these discussions and comparisons, there is a large variation in the basic approaches used, as well as some variation in results. To help understand the differences and similarities across models, a set of intermediate outcomes were developed, including incidence, lead time, overdiagnosis, detection rates, program sensitivity, changes in stage distribution over time. Also, the CISNET collaborative group has developed a state-of-the-art interactive Web site, called the Model Profiler, which allows modelers to put components of their model into templates to facilitate comparisons of model structure. Because the core documentation format is the same for each group, the published profile information is readily comparable among models. The model profiling system exists to support a framework in which modelers describe their models, and consumers read about, compare, and contrast simulation models. The CISNET Model Profiler provides a platform for managing both the content and the form of complex model documentation. A public version of the model profiler containing information on all seven breast models is available on the CISNET public site (http://cisnet.cancer.gov/profiles) . MONOGRAPH OUTLINE The monograph is divided into three basic sections. Section I [chapters 2–5 (23–25,27)] describes the common inputs. Section II [chapters 6–12 (28–34)] includes a description of each model, including its structure, data sources used for parameters estimation, calibration, and validation, and individual modeling results and interpretation (with any sensitivity analyses). Section III [chapters 13–16 (35,37–39)] starts with a chapter comparing the seven models, highlighting differences in basic assumptions, modeling strategies, and use of the common inputs. The next three chapters provide a head-to-head comparison of intermediate results, mortality results, and a final chapter with conclusions and discussion. We think that this comparative modeling effort will add an important perspective in understanding the impact of cancer control interventions for breast cancer that have occurred over the last 25 years. More generally, we hope that our systematic approach will serve as a milestone in the use of simulation in health care. References (1) Ries LA, Eisner MP, Kosary CL, Hankey BF, Miller BA, Clegg L, et al. (eds). SEER Cancer Statistics Review, 1975–2002. Bethesda (MD): National Cancer Institute. Available at: http://seer.cancer.gov/csr/1975_2002/. Based on November 2004 SEER data submission, posted to the SEER Web site, 2005. Google Scholar (2) Bonadonna G, Brusamolino E, Valagussa P, Rossi A, Brugnatelli L, Brambilla C, et al. Combination chemotherapy as an adjuvant treatment in operable breast cancer. N Engl J Med  1976; 294: 405–10. Google Scholar (3) National Institutes of Health Consensus Development Conference Statement: Adjuvant Chemotherapy for Breast Cancer. September 9–11, 1985. CA Cancer J Clin  1986; 36: 42–7. Google Scholar (4) NCI alert on node-negative breast cancer. Breast Cancer Res Treat  1988; 12: 3–5. Google Scholar (5) NIH Consensus Conference. Treatment of early-stage breast cancer. JAMA  1991; 265: 391–5. Google Scholar (6) National Institutes of Health Consensus Development Conference Statement: Adjuvant Therapy for Breast Cancer, November 1–3, 2000. J Natl Cancer Inst Monogr  2001; 30: 5–15. Google Scholar (7) Mariotto A, Feuer EJ, Harlan LC, Wun LM, Johnson KA, Abrams J. Trends in use of adjuvant multi-agent chemotherapy and tamoxifen for breast cancer in the United States: 1975–1999. J Natl Cancer Inst  2002; 94: 1626–34. Google Scholar (8) Johnson TP, Ford L, Warnecke RB, Nayfield SG, Kaluzny A, Cutter G, et al. Effect of a National Cancer Institute Clinical Alert on breast cancer practice patterns. J Clin Oncol  1994; 12: 1783–8. Google Scholar (9) Polychemotherapy for early breast cancer: an overview of the randomized trials Early Breast Cancer Trialists' Collaborative Group. Lancet  1998; 352: 930–42. Google Scholar (10) Tamoxifen for early breast cancer: an overview of the randomized trials. Early Breast Cancer Trialists' Collaborative Group. Lancet  1998; 351: 1451–67. Google Scholar (11) Howard J. Using mammography for cancer control: an unrealized potential. CA Cancer J Clin  1987; 37: 33–48. Google Scholar (12) Bailer JC Mammography: a contrary view. Ann Intern Med  1976; 84: 77–84. Google Scholar (13) Breen N, Wagener DK, Brown ML, Davis WW, Ballard-Barbash R. Related progress in cancer screening over a decade: results of cancer screening from the 1987, 1992, and 1998 National Health Interview Surveys. J Natl Cancer Inst  2001; 93: 1704–13. Google Scholar (14) Swan J, Breen N, Coates RJ, Rimer BK, Lee NC. Progress in cancer screening practices in the United States: results from the 2000 National Health Interview Survey. Cancer  2003; 97: 1528–40. Google Scholar (15) Humphrey LL, Helfand M, Chan BK, Woolf SH. Breast cancer screening: a summary of the evidence for the U.S. Preventative Task Force. Ann Intern Med  2002; 137: 347–60. Google Scholar (16) Olsen O, Gotzche PC. Cochrane review on screening for breast cancer with mammography. Lancet  2001; 358: 1340–2. Google Scholar (17) Fletcher SW, Elmore JG. Mammographic screening for breast cancer. N Engl J Med  2003; 348: 1672–80. Google Scholar (18) Goodman, SN. The mammography dilemma: a crisis for evidence-based medicine? Ann Intern Med  2002; 137: 363–5. Google Scholar (19) Peto R, Boreham J, Clarke M, Davies C, Beral V. UK and USA breast cancer deaths down 25% in year 2000 at ages 20–69 years. Lancet  2000; 355: 1822. Google Scholar (20) Weinstein MC, O'Brien B, Hornberger J, Jackson J, Johannesson M, McCabe C, et al.; ISPOR Task Force on Good Research Practices—Modeling Studies. Value Health  2003; 6: 9–17. Google Scholar (21) The Costs of the Kyoto Protocol: A multi-model evaluation. The Energy Journal Special Issue. Weyant JP, Jacoby H, Edmunds J, Richels R (eds.); 1999. Google Scholar (22) Hunink MG, Goldman L, Tosteson AN, Mittleman MA, Goldman PA, Williams LW, et al. The recent decline in mortality from coronary heart disease, 1980–1990. The effect of secular trends in risk factors and treatment. JAMA  1997; 277: 535–42. Google Scholar (23) Mariotto AB, Feuer EJ, Harlan LC, Abrams J. Dissemination of adjuvant multiagent chemotherapy and tamoxifen for breast cancer in the United States using estrogen receptor information: 1975–1999. J Natl Cancer Inst Monogr  2006; 36: 7–15. Google Scholar (24) Rosenberg MA. Competing risks to breast cancer mortality. J Natl Cancer Inst Monogr  2006; 36: 15–9. Google Scholar (25) Holford TR, Cronin KA, Mariotto AB, Feuer EJ. Changing patterns in breast cancer incidence trends. J Natl Cancer Inst Monogr  2006; 36: 19–25. Google Scholar (26) Cronin KA, Yu B, Krapcho M, Miglioretti DL, Fay MP, Izmirlian G, et al. Modeling the dissemination of mammography in the United States. Cancer Causes Control  2005; 16: 701–12. Google Scholar (27) Cronin KA, Mariotto AB, Clarke LD, Feuer EJ. Additional common inputs for analyzing impact of adjuvant therapy and mammography on U.S. mortality. J Natl Cancer Inst Monogr  2006; 36: 26–9. Google Scholar (28) Berry DA, Inoue L, Shen Y, Venier J, Cohen D, Bondy M, et al. Modeling the impact of treatment and screening on U.S. breast cancer mortality: a Bayesian approach. J Natl Cancer Inst Monogr  2006; 36: 30–6. Google Scholar (29) Fryback DG, Stout NK, Rosenberg MA, Trentham-Dietz A, Kuruchittham V, Remington PL. The Wisconsin breast cancer epidemiology simulation model. J Natl Cancer Inst Monogr  2006; 36: 37–47. Google Scholar (30) Mandelblatt J, Schechter CB, Lawrence W, Yi B, Cullen J. The SPECTRUM population model of the impact of screening and treatment on U.S. breast cancer trends from 1975 to 2000: principles and practice of the model methods. J Natl Cancer Inst Monogr  2006; 36: 47–55. Google Scholar (31) Tan SYGL, van Oortmarssen GJ, de Koning HJ, Boer R, Habbema JDF. The MISCAN-Fadia continuous tumor growth model for breast cancer. J Natl Cancer Inst Monogr  2006; 36: 56–65. Google Scholar (32) Hanin LG, Miller A, Zorin AV, Yakovlev AY. The University of Rochester model of breast cancer detection and survival. J Natl Cancer Inst Monogr  2006; 36: 66–78. Google Scholar (33) Lee S, Zelen M. A stochastic model for predicting the mortality of breast cancer. J Natl Cancer Inst Monogr  2006; 36: 79–86. Google Scholar (34) Plevritis SK, Sigal BM, Salzman P, Rosenberg J, Glynn P. A stochastic simulation model of U.S. breast cancer mortality trends from 1975 to 2000. J Natl Cancer Inst Monogr  2006; 36: 86–95. Google Scholar (35) Clarke LD, Plevritis SK, Boer R, Cronin KA, Feuer EJ. A comparative review of cisnet breast models used to analyze U.S. Breast cancer incidence and mortality trends. J Natl Cancer Inst Monogr  2006; 36: 96–105. Google Scholar (36) Feuer EJ, Etzioni R, Cronin KA, Mariotto A. The use of modeling to understand the impact of screening on U.S. mortality: examples from mammography and PSA testing. Stat Methods Med Res  2004; 13: 421–42. Google Scholar (37) Habbema JDF, Tan SYGL, Cronin KA. Impact of mammography on U.S. breast cancer mortality, 1975–2000: are intermediate outcome measures informative? J Natl Cancer Inst Monogr  2006; 36: 105–11. Google Scholar (38) Cronin KA, Feuer EJ, Clarke LD, Plevritis SK. Impact of adjuvant therapy and mammography on U.S. mortality from 1975 to 2000: comparison of mortality results from the cisnet breast cancer base case analysis. J Natl Cancer Inst Monogr  2006; 36: 112–21. Google Scholar (39) Habbema JDF, Schechter CB, Cronin KA, Clarke LD, Feuer EJ. Modeling cancer natural history, epidemiology, and control: reflections on the CISNET breast group experience. J Natl Cancer Inst Monogr  2006; 36: 122–6. Google Scholar © The Author 2006. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org.

Journal

JNCI MonographsOxford University Press

Published: Oct 1, 2006

There are no references for this article.