Katrina Armstrong - US grants

DESCRIPTION: (Applicant's Description) Dr. Katrina Armstrong, a National Research Service Award Fellow in the Division of General Internal Medicine at the University of Pennsylvania, is applying for a Preventive Oncology Career Development Award to develop and enhance the skills necessary for an academic career as an independent investigator. The candidate's career goal is to develop a multidisciplinary program to (1) study the use and outcomes of genetic cancer susceptibility testing and (2) develop and assess strategies for effectively incorporating genetic cancer susceptibility testing into primary care medicine. The Preventive Oncology Award will provide the means to take two important steps towards implementing this goal. First, it will allow for additional education in epidemiology, cancer genetics, and health behavior that will augment the candidate's previous training in epidemiology, biostatistics and general internal medicine, and lead to a Ph.D. in Epidemiology. Second, it will support an original clinical research program using rigorous epidemiologic methods to examine how risk perception, risk communication and other factors influence the actual utilization of testing for BRCA1/2 mutations. In the last ten years, the identification of cancer susceptibility genes has transformed genetic susceptibility testing for common cancers from a distant promise to an immediate clinical reality. Despite the simple appeal of defining individual risk based on genetic susceptibility, the practical reality of genetic susceptibility testing is complex. For testing to succeed in identifying high risk individuals in an environment where cost and other concerns are likely to preclude population based screening, counseling and testing must be directed to individuals at risk of carrying a mutation. Very little is known about the factors affecting the utilization of genetic testing outside a research setting. Understanding the role of risk perception and risk communication in this process is critical to our ability to direct limited counseling and testing resources to individuals at risk, and, thereby, realize the maximum clinical and public health benefit from genetic susceptibility testing. In this application, the candidate proposes a two-part research plan that is an innovative approach to important, unanswered questions about genetic testing for cancer susceptibility. The first study will use a case-control design to compare women who undergo genetic counseling for breast cancer susceptibility to women from the same population who do not. The second study will use a prospective cohort design to measure the effect of genetic counseling on risk perception and the relationship between risk perception, predicted risk and decisions about genetic susceptibility testing. Together, these studies will provide a comprehensive overview of the utilization of BRCA 1/2 testing and form the foundation for the candidate's transition to an independent researcher in preventive oncology.

[unreadable] DESCRIPTION (provided by applicant): Use of chemoprevention, such as tamoxifen, among high-risk African American women could significantly reduce mortality among this group. However, cultural barriers may exist to African-American women adopting chemoprevention. An African American health maintenance model, characterized by the concept of the body as a fragile homeostatic system that is easily disrupted by impure elements such as unnecessary medications, negative thinking and stress, may impact chemoprevention uptake and use. In order to understand the dimensions of this model and its potential impact on chemoprevention among African American women, we intend to: 1) Verify the elements of the above model, identify any other contributing elements prior research did not uncover, and clarify dynamic relationships among the elements to define a culturally specific model of health maintenance and disease prevention among African American women; 2) Discover whether this model may affect preventive medical decision making about tamoxifen use among African American women who are at high risk of breast cancer; and 3) Lay the groundwork for future research examining how this model functions relative to other models known to be important to medical decision making. These goals will be best achieved through a two-step qualitative study, to reduce complexity and allow emerging insights to guide inquiry throughout the research process. Study 1 will guide the accurate representation of the above model and inform development of the data manual to be used in Study 2. Data from existing literature, focus groups and in-depth interviews will inform this step. Study 2 will explore what role the model defined in Study 1 plays in decision making about chemoprevention use among women at risk for breast cancer. A purposive sample of 120 African American women with a first-degree relative diagnosed with breast cancer will be sufficient to conduct analyses. Semi-structured interviews will elicit how this model might play a role in decision making about tamoxifen and consensus analysis will estimate how strongly the overall group knows about and believes in the model. Analysis will provide preliminary information regarding points of convergence and divergence between the biomedical and African American models of health maintenance. This understanding may lay the groundwork for improved understanding and trust between biomedical practitioners and African American patients. [unreadable] [unreadable]

[unreadable] DESCRIPTION (provided by applicant): Scholarly interest in trust and distrust has increased sharply over the last several decades. This multidisciplinary attention has highlighted how trust facilitates and sustains human relationships and organizations and how distrust can destroy these same relationships and organizations. Concurrently, progress in human genetic research has led to the development of genetic tests that predict the risk of future disease. The uncertainty, vulnerability, and implications for the future inherent in predictive genetic testing suggest that trust and distrust will play important roles in determining whether this new technology is successfully translated into improved health outcomes. Furthermore, given the historical context and the current media coverage, distrust related to predictive genetic testing may already exist, particularly among racial and ethnic minorities. [unreadable] [unreadable] The overall goals of this project are to increase understanding of the nature, distribution, and correlates of trust and distrust related to predictive genetic testing and to investigate the relationship among trust/distrust, attributes of a genetic test and its delivery, and willingness to undergo predictive genetic testing. In our preliminary conceptual model, trust and distrust related to predictive genetic testing are multidimensional constructs that encompass beliefs about whether or not a predictive genetic test will be used in one's best interest. The components of these constructs are defined by how and by whom the genetic test may be used. The project focuses on African-American, Caucasian and Latino populations in order to elucidate the relationship between distrust and ethnicity/race and to address the legacy of distrust among historically disadvantaged minorities. [unreadable] [unreadable] We propose a three phase project using qualitative and quantitative methods. In Phase One, we will refine our conceptual model through comprehensive literature review, focus groups, interviews and consultation with an expert advisory panel. In Phase Two, we will develop a psychometrically sound instrument to measure trust and distrust related to predictive genetic testing. In Phase Three, we will conduct a national survey of African-American, Caucasian, and Latino adults to measure distrust, potential correlates of distrust, and the relationship between distrust, attributes of the genetic test and its delivery, and willingness to undergo testing. To accomplish these goals, we have assembled an investigator team of experts in bioethics, genetics, psychometrics, and survey research, as well as a multidisciplinary advisory panel that includes stakeholders from the African-American and Latino communities. [unreadable] [unreadable]

[unreadable] DESCRIPTION (provided by investigator): Survival after the diagnosis of endometrial cancer varies significantly between African-American and Caucasian women. Between 1992 and 1998, five-year survival for African-American women after endometrial cancer diagnosis was 58.9%, compared to 85.8% for Caucasian women. The disparity in survival is greatest among older women, with an absolute difference in five year survival of 10% for women under 50 compared to 30% for women 50 years of age and older. Prior studies have identified several factors that contribute to the observed racial disparity in endometrial cancer survival, including stage at diagnosis and tumor grade. However, significant differences in survival between Caucasian and African-American women persist even after adjusting for these factors. There are several reasons to believe that differences in the prevalence and characteristics of treatment may contribute to this residual survival disparity. African- Americans have been shown to be less likely to undergo definitive treatment for many different medical and surgical conditions. Characteristics of treatment (including provider characteristics, hospital characteristics and intensity of therapy) are associated with outcome for other surgical conditions, including surgery for lung, pancreatic and breast cancer. Understanding the prevalence and outcomes of differences in treatment characteristics between African-American and Caucasian women offers a potentially promising new approach to improving endometrial cancer survival among African-American women. [unreadable] [unreadable] In this application we propose to use SEER-Medicare linked data to examine the outcomes of African- American and Caucasian women diagnosed with endometrial cancer between 1991 and 1999. The primary outcome will be overall and disease-specific survival times (which may be censored) as assessed by Medicare vital statistics and SEER linkage to the National Death Index respectively. Analyses will adjust for comorbidity, socioeconomic status and tumor characteristics using information provided in the SEER-Medicare database. Provider and hospital characteristics will be determined by linkage to the AMA practitioner database and AHA annual survey respectively. [unreadable] [unreadable] Our three specific aims explore the contribution of differences in treatment to the higher mortality among African-American women diagnosed with endometrial cancer. We group differences in treatment into three categories: (1) differences in the rates of treatment; (2) differences in the extent/intensity of treatment; and (3) differences in the providers and hospitals/facilities who deliver the treatment. For each category, we will explore differences between African-American and Caucasian women, their association with outcome, and to what extent variations in treatment explain the excess mortality among African-American women. In addition, within each category, we will investigate differences related to primary surgery and adjuvant radiation therapy. [unreadable] [unreadable] [unreadable]

The Seeking and Scanning Behavior Project under CECCR I (SSB I), established the frequent use of nonclinician information sources among a large representative sample of patients with colon, breast and prostate cancer, as they made post-diagnosis decisions related to treatments and quality of life. There is evidence from that study that use of these sources (e.g., personal sources, such as friends and families; impersonal sources, such as internet) is related to several important behaviors. However, while the specific aims of SSB I focused on the role of non-clinician information sources, the study also makes clear that the information exchange between clinicians and patients was central for these patients. The proposed project builds on this finding and focuses on the role of medical information sources and patient-clinician communication in an oncology setting. There is a substantial literature about patient-clinician communication: describing it, making normative claims about what patterns of communication are most productive, testing short term effects of experimental interventions meant to improve such communication, and, more rarely, examining longer term health effects of variation in such communication [1].There are, to our knowledge, no studies that test the longer term effects of natural variation in physician-patient communication on important health outcomes among a representative sample of cancer patients. The proposed study will be the first to undertake this task. Our prior CECCR study randomly drew a sample of cancer patients from the Pennsylvania Cancer Registry: 2013 breast, prostate and colon cancer patients provided baseline data within 12-20 months of their initial cancer diagnosis; 1296 provided follow-up data one year later. The baseline and follow-up questionnaires document the complex information exchange between patients and their physicians: the great majority of patients (75%) reported actively seeking treatment and other information from their treating physicians. However patients varied in the extent and content of other interactions with their physicians: whether and from what sources they brought information to their physicians; how often their physicians directed them to other interpersonal and mediated sources; whether they sought quality-of-life information from their physicians; whether they received recommendations from their physicians about lifestyle behavioral changes and other follow-up behaviors; whether they could get information they wanted from their physicians; whether they sometimes avoided talking with their physicians about their questions; and whether their physicians shared decision-making with them to the extent that the patient desired. Patients also varied sharply in their use of non-clinician information sources, including personal sources and mediated sources. The proposed longitudinal study will focus on variations in patient-clinician information exchange around cancer diagnosis and treatment (a component of patient-clinician communication). On the one hand we propose studying whether patient-clinician information exchange is associated with and predictive of important health management decisions, behaviors and outcomes, including treatment choices, adherence with physician recommendations for tests and medications, lifestyle behavior changes, maintenance visits to physicians, and treatment-associated morbidity (Study One) and, on the other hand, whether physician-patient information exchange is a function of different patterns of physician visits (as documented in Medicare administrative claims data) and other socio-demographic and illness characteristic determinants (Study Two). The proposed Study One also will test whether the level of patient-clinician information exchange influences the effect of exposure to non-clinician information on these outcomes. Both studies will use three data sources: (1) we will fully exploit the available data from the first two rounds of the cancer patient survey from SSB I, a unique resource, to examine new research questions; (2) we will match the data from the surveys to Medicare administrative claims for those patients ¿ 65 years, which will allow us to move from dependence on patient self-report alone to a more objective form of evidence about patient-clinician visit patterns and about health outcomes; and (3), we will obtain a third round of questionnaire data from cancer patients who provided data in the first two rounds - of the 1296 patients who provided data at the follow-up round, 91% have agreed to provide data for a follow-up questionnaire. The proposed study will (a) provide previously unavailable evidence to assess the relationship between one form of patient-clinician communication and hypothesized effects of such communication and (b) if there is such evidence, allow development of recommendations to encourage and facilitate the types of patient-clinician communication associated with improved health outcomes.

DESCRIPTION (Provided by the applicant): Advances in genomics have the potential to improve the delivery of health care by targeting interventions to individuals who will receive the greatest benefit and experience the lowest risk of adverse events. More effective targeting of interventions is particularly appealing because it is one of the few approaches that can both improve outcomes and reduce costs- the unquestioned sweet spot of health care reform. However, genomic medicine has a long way to go before it gets to that sweet spot. Some of what is needed is discovery of gene-disease associations that lead to the development of a clinical genetic test. But increasingly, the needs lie not in developing the test but in deciding whether using the test is better than doing what we are currently doing - i.e. comparative effectiveness. The overarching goal of this project is to develop a coordinated, multidisciplinary center for the generation and synthesis of evidence to support the translation of genomic tests into improvements in cancer prevention, screening, diagnosis, treatment and survivorship. The center (entitled the Center for Comparative Effectiveness in Genomic Medicine or CCEGM) will involve two primary scientific components: (1) evidence generation;and (2) evidence synthesis and modeling. The goal of the evidence generation component is to conduct observational and experimental studies of the comparative effectiveness of genomic test that are clinically available or nearly clinically available. Four proof of principle, pilot studies are included in this proposal: (1) pharmacogenomics of nicotine addiction treatment;(2) incremental information from breast cancer SNP panels in breast cancer risk screening and prevention;(3) personalized treatment for non small cell lung cancer;and (4) CDKN2A/p16 testing and adherence to melanoma prevention behaviors. The goal of the evidence synthesis component is to use statistical and modeling methods to bring together existing evidence to inform recommendations about clinical practice. Three corresponding evidence synthesis pilot projects are included: a systematic review of the clinical validity of EGFr and K-ras mutations in predicting response to treatment, a decision model of the use of SNP panels in breast cancer screening, a cost- effectiveness model of nicotine metabolism markers in treatment of nicotine addiction. PUBLIC HEALTH RELEVANCE: Advances in genomics have the potential to improve the delivery of health care. Increasingly, there is a need to decide whether the use of genomic tests is better than doing what we are currently doing - i.e. comparative effectiveness. The overarching goal of this project is to develop a coordinated, multidisciplinary center for the generation and synthesis of evidence to support the translation of genomic tests into improvements in cancer prevention, screening, diagnosis, treatment and survivorship.

DESCRIPTION (provided by applicant): Genetic tests for cancer susceptibility have the potential to reduce cancer mortality by targeting intensive cancer risk reduction interventions to high risk individuals. Testing for mutations in the breast cancer susceptibility genes, BRCA1and BRCA2, is one of the earliest examples of cancer genetic susceptibility testing to become incorporated into clinical practice and be linked to multiple evidence based options for cancer risk reduction. Although recommendations for consideration of BRCA1/2 testing among high risk groups have existed for nearly ten years, evidence suggests that testing uptake has been limited and that substantial racial disparities in testing utilization may exist. Understanding the underlying causes of these disparities is critical for the development of strategies to ensure that advances in genomics are translated into equitable improvements in clinical care. The determinants of utilization of genetic testing are complex and encompass patient, provider and system factors. Although racial disparities in health care have traditionally been traced to differences in patient characteristics or physician bias, a growing body of literature indicates that differences in providers across patient racial groups are an important and often fundamental of health care disparities. Whether the sorting of patients of different races across different providers is an important determinant of racial differences in the use of genetic or genomic technologies is currently not known. In this proposal, we outline a population based, prospective cohort study that will determine the rates of BRCA1/2 testing among Black and White women with early onset breast cancer and investigate the relative contribution of individual patient characteristics (within provider effects) vs. provider characteristics (between provider effects) to racial differences in utilization. We will use both multi-level models to examine the contribution of specific patient and provider characteristics and conditional logistic regression and decompositional techniques to determine the relative proportion of the racial difference in testing explained at the patient vs. the provider level. Building on our prior work in this area, we will test specific hypotheses about the contribution of health care related distrust and insurance coverage at the patient level and specific hypotheses about the contribution of attitudes towards innovation and access to resources at the provider level. We will compare these results to the contribution of the same factors to other known racial disparities in breast cancer management, asking whether the patient and provider characteristics that lead to disparities in BRCA1/2 testing also lead to disparities in use of adjuvant therapy and whether a provider's performance on one outcome also predicts their performance on the other outcomes. Public Health Relevance: Genetic susceptibility tests have the potential to reduce cancer mortality by targeting intensive cancer risk reduction interventions to high risk individuals. BRCA1/2 testing is one of the earliest examples of cancer genetic susceptibility testing to become incorporated into clinical practice and be linked to multiple evidence based options for cancer risk reduction. This project will increase our understanding of the use of testing among Black and White women with early onset breast cancer and the patient and provider determinants of disparities in that use.

DESCRIPTION (provided by applicant): The ongoing controversy regarding mammography screening recommendations highlights the need for improved performance across the breast cancer screening process. While there is little debate that breast cancer screening can reduce breast cancer mortality, it is also clear that it is time for better approaches to screening that leverage advances in breast imaging modalities, breast cancer risk assessment and patient centered care to achieve better outcomes at lower cost. This application for a Penn Center for Personalized Breast Cancer Screening will advance a personalized breast cancer screening paradigm by developing a new tool (breast complexity index) for predicting individual screening outcomes, evaluating the comparative effectiveness of a new imaging modality (digital breast tomosynthesis) on screening process and outcomes, and developing effective strategies for communicating individual estimates of benefit and risk of alternative screening approaches to better link individualized information to informed patient and provider decision making. In addition to these three highly integrated research projects, the Center will bring together comprehensive screening process and outcome data on a diverse population of 74,000 women who undergo breast cancer screening at 6 sites within the Penn Medicine integrated health network. In addition to clinical, socio demographic and screening data from EMR and screening reporting systems, this Screening Process Documentation Unit will include risk factor and other patient reported information collected through web portal and point of care surveys, screening images, cancer outcomes from state cancer registries and pathology records, patient neighborhood characteristics from the Penn Cartographic Modeling Laboratory, as well as an ongoing DNA biobank. The proposed Center leverages the substantial expertise at Penn in the multiple disciplines needed to advance a personalized screening paradigm (including breast imaging, primary care, communication, computer science, biostatistics, health services research, bioinformatics, medical oncology, and cancer genetics) as well as the commitment of the clinical leadership (including the Center Principal Investigators) to Penn Medicine's role in the evaluation and implementation of such a paradigm. The Center will address questions with immediate scientific, clinical and policy impact and create an infrastructure for continuous learning about the breast cancer screening process and enabling collaboration through the PROSPR network.

Summary: The Biostatistics Core (the designated Shared Research Resources Core) will provide expert guidance and support in the biostatistical aspects of the design, conduct and analysis of research projects generated under the Penn PROSPR Research Center grant. Core staff members will collaborate with project investigators at every stage: They have already assisted in the design of proposed projects, and they will continue to work with investigators in refining design and analysis plans. As studies proceed. Core staff members will monitor study databases, participate in the evaluation of data quality, and conduct any designated interim analyses. When studies are finished. Core staff will conduct correct and efficient data analyses, create graphs and tables, assist investigators with the preparation of presentations and manuscripts, and consult on the design of subsequent research. The Core staff members have extensive experience supporting research projects of all kinds in breast cancer and other forms of cancer. In particular, they have substantial expertise in comparative effectiveness research, cancer diagnostics, health economics, and the statistical evaluation of screening programs. Thus the Core is superbly qualified to provide biostatistical support of the highest quality to PCIPS.

In this project, we propose to develop and test a communication strategy to support clinical decision making about a key step in a personalized breast cancer screening paradigm, the age of initiating breast cancer screening. The absolute benefit of mammography screening varies according to risk of developing breast cancer, a risk that depends upon multiple factors and particularly age. Women at lower risk of breast cancer will experience smaller absolute benefits from screening, but are likely to have the same risk of false positives and other adverse effects. This variation in the risk-benefit ratio of a particular screening strategy according to individual characteristics is fundamental to the personalized screening paradigm. However, there is little evidence about how best to communicate this information to women considering screening. The proposed project includes two phases. In Phase 1, we will conduct a series of experimental studies to examine the effect of alternative communication strategies on comprehension and the correlation between screening intentions and the absolute benefit and risk in the scenarios. Based on recent advances in communication theory and our preliminary work, we will focus on the potential impact of exemplars and of information addressing the expertise and motives of the information sources. In Phase 2, we will test a decision aid based upon the results of Phase 1 in an cross-over, clustered randomized controlled trial in a primary care population, examining its effects on knowledge, decision satisfaction and screening.focusing on the correlation between the absolute benefit of screening and screening utilization. PHS

PCIPS CORE SUPPORT The PCIPS will be supported by three cores: the Screening Process Documentation Unit, the Biostatistics Core, and the Administrative Core, each described briefly below Screening Process Documentation Unit: The foundation of the PCIPS is the Screening Process Documentation Unit (SPDU), an innovative data core that integrates data from a range of sources to create a comprehensive picture of the screening process across a diverse population of women in the Delaware Valley. This comprehensive picture includes longitudinal information about the screening process as well as multi-level information about the systems in which women are seen and the communities in which they live. By pulling these data together into a cohesive whole (entitled the Breast Screening Data Repository or BSDR) (Figure 2), insight can be gained into the determinants of screening outcomes across a population, and perhaps even more importantly, into where this process can be improved.

PROJECT SUMMARY (See instructions): Racial disparities in cancer treatment are well described but poorly understood. Although most experts believe that access to care contributes to these disparities, access for vulnerable populations is a complex, multidimensional process that is inadequately captured by measures of insurance coverage or travel time. In this application, we propose to build upon our prior researching racial disparities, prostate cancer treatment and access to care to investigate the contribution of a multidimensional picture of health care access to racial differences in prostate cancer treatment. We will collect primary data on a population based sample of men diagnosed with localized prostate cancer (N=3,000) in 8 counties in the Greater Philadelphia area and on the urologists and radiation oncologists who form their access to prostate cancer treatment. We will focus on this area as it is the primary focus of the proposed center and because the multiple dimensions of cancer care in urban populations are poorly understood. Building on the access framework of Penchansky and Thomas, we will assess the accessibility /availability, accommodation, affordability, & acceptability of prostate cancer care, using data from a patient survey, administrative records, & a provider inventory. Following the framework of Aday & Andersen, we will create spatial measures of potential access & individual measures of experienced access (a form of realized access). We will incorporate information about the characteristics of providers using data from patients about their experiences, as well as data from the AMA masterfile about provider board certification and training. We will then examine the relationships between potential access, experienced access, & differences in treatment as well as the degree to which these effects are mediated by insurance coverage, socioeconomic status & neighborhood deprivation. To our knowledge, this will be the first study to construct a comprehensive, multi-dimensional picture of access to cancer care by collecting information from patients, providers & administrative sources & to use this picture to improve our understanding of a racial difference in cancer treatment.

The goal of this project is to determine the benefit of incorporating novel imaging markers of breast tissue composition in guiding decisions for personalized breast cancer screening recommendations. Specifically, we propose to determine the predictive value of a new Breast Complexity Index (BCI), defined as a comprehensive descriptor of breast density and parenchymal texture, in ;) improving breast cancer risk estimation and //) assessing the risk of a false-positive or a false-negative screening exam. Studies have identified breast density as a strong independent risk factor for breast cancer. In addition, clinical trials have shown that mammographic sensitivity and specificity drops significantly in women with dense breasts, when compared to women with fatty breasts, suggesting that increased breast density is also independently associated with cancers missed by mammography (i.e., false negatives) and an increase in false positive recalls. The proposed research plan is designed as a phased imaging biomarker validation study with the Intention to determine the predictive value of BCI in improving individualized assessment in two important aspects of personalized risk: a woman's risk of developing breast cancer and her risk of having unnecessary call-back or mammographically occult disease. Our main hypothesis is that, by integrating quantitative breast density information with parenchymal texture features, BCI can provide a more comprehensive and objective descriptor of breast tissue composition than the commonly used BIRADS density and breast percent density (PD%) estimates and therefore result in more accurate predictions at the individual level. Improving risk assessment in both of these aspects can result in more informed decisions for guiding breast cancer screening recommendations, recognizing that personalized approaches to screening not only intensify surveillance to those most likely to benefit {i.e., high-risk women), but should also reduce intensity or provide alternative approaches to those less likely to benefit. Within this setting we propose compare mammography to breast tomosynthesis, an emerging tomographic breast imaging modality with the potential to significantly reduce false-positive call-backs in the screening setting, especially for women with very complex breast tissue. Our study holds the promise to shift the current paradigm in breast health care delivery by providing improved clinical decision-making tools for guiding personalized breast cancer screening recommendations.