Fellowships, Grants, & Awards

Network (CISNET) The Division of Cancer Control and Population Sciences (DCCPS) of the National Cancer Institute (NCI) invites applications from domestic and foreign applicants to support collaborative research using simulation and other modeling techniques to describe the impact of interventions in population-based settings that will shed light on U.S. population-based trends. It is well known that great progress in the war against cancer is possible by the complete use and adequate delivery of existing modalities of cancer control. The primary goals of this research are to determine the impact of cancer control interventions on observed trends in incidence and/or mortality, and to determine if recommended interventions are having their expected population impact by examining discrepancies between controlled cancer intervention study results and the population experience. Once a general understanding of the various factors influencing current trends has been achieved, a number of secondary goals may be addressed. Applicants may propose secondary goals of modeling the potential impact of new interventions on future national trends, and/or evaluating optimal cancer control strategies. The NCI has a long-standing function of providing answers to critical policy questions, which can only be answered through an indirect synthesis of available information and assumptions. A commitment to modeling of this type will allow the NCI to apply the most sophisticated tools available for evidence-based planning to several areas: 1) Be responsive to challenges due to the increasing pace of technology, and to provide short-term answers while randomized controlled trials (RCTs) are still in progress. In the future we will be increasingly faced with new interventions, biomarkers, and diagnostic and genetic tests that will become widely disseminated prior to rigorous testing in controlled settings, and therefore the evaluation of population impact will become even more important. 2) Address emerging questions while they are still being debated in the policy forum. For example, new smokeless tobacco products are coming on the market, and modeling of their potential impact can benefit the Federal Trade Commission and other policy makers. 3) Translate RCT evidence of quantities to the population setting. 4) Provide estimates of quantities that will never be derived from RCTs. For example, half of Americans alive today who ever smoked are ex-smokers. It is important to understand the patterns of quitting, the process of carcinogenesis for ex-smokers, and the implications for future lung cancer trends. DCCPS, which fulfills a federal-level function to respond to evolving surveillance questions of national policy relevance, helps focus research questions and acts as a conduit to national data resources necessary for parameter estimation, model calibration, validation, and population trends. An emergent property of this collaborative agreement is progress toward a comprehensive understanding of the determinants of sitespecific cancer trends at the population level and a better understanding of the science of modeling. Modeling is the use of mathematical and statistical techniques within a logical framework to integrate and synthesize known biological, epidemiological, clinical, behavioral, genetic, and economic information. Prior to the Cancer Intervention and Surveillance Modeling Network (CISNET), many of the simulation and other modeling techniques had been utilized to describe the impact of cancer interventions (i.e., primary prevention, screening, treatment) for hypothetical cohorts or in trial and other clinical settings. The goal of this request for applications (RFA) is to promote the application and extension of these models to population-based settings in order to ascertain determinants of cancer trends. This information is critical to the NCI because of the necessity of understanding whether recommended interventions are having their expected population impact, and of predicting the potential impact of new interventions on national trends. These studies will often involve extrapolation of results of controlled cancer intervention studies to estimates of U.S. population and community effectiveness. This type of modeling addresses issues of population-based policies and programs, and is distinct from individual-level models of risk and models of clinical decision making used at the individual patient–physician level. An additional goal of this concept is to advance methodology for modeling and to develop more uniform criteria for model validation in the population setting. It is not the purpose of this RFA to focus on the analysis of hypothetical or trial-based cohorts and/or cost-effectiveness analyses, but rather to support analyses based on realistic scenarios of population impact. Projects will focus on models describing the population impact of the observed dissemination of cancer control interventions as well as other factors on observed national incidence and/or mortality trends. CISNET was originally funded as a cooperative agreement (U01) for two phased-in rounds of funding. In September 2000, RFA CA-99-013 funded seven grants in breast cancer, one in prostate cancer, and one in colorectal cancer. A second round, funded under RFA CA-02-010 in August 2002, funded five grants in lung cancer as well as two additional grants for colorectal cancer and one in prostate cancer. CISNET investigators are currently engaged in a wide range of policy-relevant modeling studies including the following: 1) Development of base case questions. A major strength of having a consortium of modelers is the ability to employ a comparative modeling approach. While each modeler has areas of individual focus, whenever possible, common “base” questions have been developed that allow for comparisons across models. The sometimes widely different results from models are often difficult to resolve, and base cases provide a chance to reach consensus on important questions, and to better understand differences between models. In these base case questions, a set of common population inputs is used across all models (e.g., dissemination patterns of screening and treatment, mortality from noncancer causes), and a common set of intermediate and final outputs is developed to help understand differences and similarities across models. 2) Breast base case spin-off questions. The breast base case serves as a jumping-off point for each grantee as they vary the basic formulation to focus on areas of individual interest. Spin-off issues that are actively being pursued include a) modeling the impact of using alternative, more biologically based natural disease history formulations, especially continuous time tumor growth models (which include microscopic fatal metastases that are initially undetectable); b) analyses for different racial, ethnic, and insurance-status groups; c) a unique Bayesian approach to update its prior estimates of treatment efficacy to obtain posterior estimates of community effectiveness of adjuvant therapy and mammography that best reproduce national mortality trends; d) geographically based analyses; e) the role of risk factors in breast cancer trends; and f) the potential impact of optimal screening intervals. 3) Prostate cancer. CISNET researchers have published an analysis of trends in the use of the prostatespecific antigen (PSA) test for modeling prostate cancer incidence trends to obtain estimates of overdiagnoses associated with PSA screening. In addition, these researchers are investigating the use of modeling to better understand the results of ecologic analyses of the effectiveness of PSA screening. 4) Special issue of Statistical Methods in Medical Research. CISNET was invited to sponsor a special issue of the journal Statistical Methods in Medical Research titled “Uses of Stochastic Models for the Early Detection of Cancer,” with articles submitted in spring 2003. Articles in the issue include 1) “Distribution of Clinical Covariates at Detection of Cancer: Stochastic Modeling and Statistical Inference,” 2) “Planning Public Health Programs and Scheduling: Breast Cancer,” 3) “Planning of Randomized Trials,” 4) “The Use of Modeling to Understand the Impact of Screening on U.S. Mortality: Examples from Mammography and PSA Screening,” 5) “Parameter Estimation for Stochastic Models via Simulation,” and 6) “Diversity of Model Approaches.” 5) Linkages with other cancer surveillance and control activities. CISNET has sought linkages to be integrated with and responsive to situations where modeling may play an important role. For example, the Agency for Health Research and Quality and the Center for Medicare and Medicaid Studies approached the NCI for assistance in studying a reimbursement decision related to the immunochemical fecal occult blood test (iFOBT) (http://cisnet. cancer.gov/reports/medicare.html). CISNET modelers have also been asked to aid in a midcourse (2005) evaluation to help determine whether reaching Healthy People 2010 upstream goals for cancer treatment, screening, and prevention will enable us to fall short of, meet, or exceed the downstream 2010 cancer mortality goals, and to retarget our efforts if necessary. This reissuance of CISNET will be limited to modeling applications focusing on breast, prostate, lung, and colorectal cancer. Although the reissuance of CISNET will not be limited to grantees previously or currently funded, CISNET will no longer fund models that either are starting from scratch or have not been previously applied to the analysis of population trends. This means that models should have been applied to multiple real birth cohorts representing the actual population experience. Models that have been applied only to hypothetical cohorts, as is sometimes done to model trial data or estimate cost-effectiveness, will not be considered. The emphasis in this reissuance is in the application of already developed models to study the population impact of existing or emerging cancer control interventions. In addition, applications are being solicited for cancer site–specific coordinating centers for breast, prostate, colorectal, and lung cancer. Areas of application will include more refined analyses of current trends, and a renewed emphasis on future trends and optimal cancer control planning. While the original issuance focused primarily on discovery (basic mathematical and statistical relationships necessary for the development of multi-cohort population models) and development (data sources and realistic scenarios to evaluate past intervention impact in the population setting and project future impact), the reissuance will continue development efforts and will greatly enhance the delivery element (synthesizing relevant scenarios for informing policy decisions and cancer control planning implementation). While some new mathematical and statistical derivations may be necessary, they should not be the centerpiece of these applications. Instead, the focus of the application should be on identifying important cancer surveillance and control questions, obtaining the data sources and making model modifications as