Discussion of Process Issues or Recommendations

Engaging stakeholders to shape research so that it is more responsive to what consumers, patients, clinicians, and decision-makers need is a national priority. Because this practice is relatively new, there is no clear guidance for the optimal methods of stakeholder engagement or distribution of stakeholders. While approximately half of invited stakeholders agreed to participate in this project, we were still able to achieve adequate participation within each stakeholder group. We were able to accomplish this because we defined the optimal stakeholder groups before extending invitations and then carefully monitored acceptance and refusals, adjusting recruitment accordingly. Through this process, we documented our a priori stakeholder groups, compared them with our final participation, and ensured participation of all five major stakeholder groups. Ultimately, through close monitoring of recruitment, 29 percent of participants were consumer advocates, followed by researchers (24%), clinicians (19%), funders (19%), and policymakers (9%). Factors that may have negatively influenced response rates include short time frame between invitation and Webinar, summer vacations, and transitions to other professional positions. Processes that facilitated stakeholder engagement included personal contact with organizations and individuals, existing relationships with individuals, and stakeholders being so invested in the topic that they either wanted to participate themselves or personally recommended others. This personal championship not only encouraged engagement but also created momentum around the dissemination of the final product.

The goal of this project was to prioritize research relating to the prevention of primary breast cancer. Quantitative methods such as questionnaires, Delphi processes, and voting are frequently used as efficient and equitable processes to obtain priorities. In this project we asked all stakeholders to rate items as high, medium, or low priority to determine the top ten priorities in rank order based upon the frequency that individual stakeholders rated the item as high priority. We plan to compare these results to priorities by stakeholder group to have a deeper understanding of whether and how stakeholders differ in their priorities. When using a questionnaire, offering different administration formats for the questionnaire (choice of email delivery of PDF form or Web-based form) may have promoted broader participation as two-thirds of stakeholders chose the Web-based option and 1/3 chose either electronic completion of PDF sent by email or printed PDF returned by fax.

Framework for Future Research and Reflections on CERs

Analytic frameworks have been used to structure reviews but were not designed to guide discussions of future research. In the usual format for an analytic framework, interventions and actions are represented by arrows and events are represented in boxes. This format facilitates discussion of outcomes, but it makes it very difficult to focus attention on the range and choice of interventions. For future research discussions, graphical frameworks need to clearly communicate ideas, linkages, and assumptions in an organized way that demonstrates that the research proposed is well-integrated, well-reasoned, and appropriately designed to advance a field of research.

Figure 4 presents a conceptual framework for future research in the primary prevention of breast cancer. Similar to analytic frameworks of CERs the framework is read from left to right starting with the population of interest on the left and ending with health outcomes on the right. Arrows are used to indicate actions and squares to indicate health outcomes. Circular symbols (circles and ovals) are used to indicate events, whether benefits (e.g., intermediate outcomes) or harms. We have used rounded boxes to highlight important topics for future research discussion and we have added the diamond that has been used (e.g., behavioral intervention¹¹ and Vaginal Birth After Caesarean frameworks^6,12) to indicate influencing factors.

Figure 4

Conceptual framework: future research needs to reduce the risk of primary breast cancer.

The future research needs framework demonstrates the three major areas for future research in the primary prevention of breast cancer. Stakeholders consistently agreed that one of the highest priorities is answering the question: Who is at highest risk for developing breast cancer and most likely to benefit from preventive therapies (Future Research Question A)? This question combines risk for breast cancer and susceptibility to benefits and harms of therapies. See Table 7 for details on research gaps identified by stakeholders and corresponding study designs. Investigations in this area could include determining all the possible markers and tests that should be considered to classify women regarding their candidacy for preventive therapies. Specifically, which molecular, genetic, and demographic characteristics and/or blood or imaging tests predict who is at highest risk of developing the most aggressive forms of breast cancer? Stakeholders frequently mentioned wanting more epidemiological research in premenopausal women. We are aware of a study underway that combines literature synthesis and epidemiological methods to examine which factors increase a premenopausal woman’s (ages 40–49) risk for primary breast cancer and the magnitude of these risk factors.¹³ For example, the study will estimate a woman’s risk of primary breast cancer if she has a history of smoking.

Table 7

Future research agenda for breast cancer prevention.

Another suggestion was to conduct intervention studies in special populations. For example, intervention studies among women with hyperplasia in breast biopsy and repeated biopsy to see if the tissue changes. Which factors predict who is most susceptible to harms of therapy vs. benefits? In general, the two groups of stakeholders emphasized different features of the population. Consumers and policymakers emphasized demographic features of the population that may reflect access to care and create additional vulnerabilities that worsen prognosis, whereas researchers, funders, and clinicians were interested in the molecular and genetic basis that places a woman at higher risk, causes the development of more aggressive disease, and/or predicts better response to therapies. One informational interview mentioned that the Gail model, which is often used to calculate risk, does not predict risk for special populations such as the Puerto Rican population in New York City and Mexican-American population on the west coast.

Researchers highlighted that SERMs do not completely prevent even estrogen receptor-positive cancers and they have considerable side effects and adverse events. They felt that molecular biomarkers such as PARP inhibitors offered promise to both target the most lethal types of breast cancer and focus the medications. Stakeholders’ comments such as, “Would prefer identification of molecular or genetic predictors of response to chemopreventive interventions as this would enable a more individualized approach to women at increased risk of breast cancer,” demonstrate the importance of these features in individuals and populations of individuals for patient-centered care.

Moreover, researchers suggested using stored biologic samples from participants in the SERM trials who had events vs. those who did not to explore the genetic (SNPs) and molecular characterization to better predict risk and benefit. While biological factors may be implicit in the model, the emphasis of stakeholders on their importance not only to discovery but to individualized care caused us to highlight these items in the population. The wide range of factors thought to contribute to population risks require a wide range of investigator skills ranging from basic science to epidemiology to clinical researchers.

Progressing through the framework, the next major research question is: What interventions are most effective to reduce the risk of breast cancer and improve short and long-term outcomes (Future Research Question B)? Overall, when discussing interventions, the I (intervention), C (comparison), and O (outcome) of PICO were often inextricably intertwined in the responses of stakeholders such that it was not possible to accurately distinguish the relative priority of the benefits (outcomes) of an intervention from the intervention itself. While the scope of the CER focused on traditional medications to prevent breast cancer, stakeholders’ interests in interventions were much broader, extending to lifestyle changes, diet, exercise, dietary supplements, and other interventions.

“This [weight loss as therapy] is an intervention that carries essentially no harm and great preventive/therapeutic benefits for many diseases”

“I like this idea [weight loss as therapy] because of the huge public health burden of the obesity epidemic and the biologically plausible effectiveness without drugs.”

“The evidence for this is weak and varies by menopausal status and too many confounders.”

“Wow!!! If this works we could have labels on some foods “eating this may be hazardous to your breast”’

“Exercise and dietary modification may be interesting to study in young females (children and adolescents).”

“One published paper in African American women... none for Hispanic women. The challenge for this... lies in the fact that longitudinal studies like the Women’s Healthy Eating and Living Project (experimental design) are needed to adequately address these areas. The potential benefits and long term health care savings would far outweigh the costs of doing the studies.”

“This (exercise as therapy) is an intervention that carries essentially no harm and great preventive/therapeutic benefits for many diseases. Better understanding the benefits of exercise on breast cancer prevention would provide clinicians with additional rationale for recommending it and would motivate more women to be active.”

“I'm ready for a trial--but the logistics of a trial and its size make this a hard sell. I do not know any evidence from other trials of exercise (which is always confounded by weight loss) that did show cancer reduction. The Women’s Health Initiative diet arm did cause weight loss but no cancer risk changes were observed. I think the weight and exercise trial should be combined, given the known difficulty of sustaining weight loss without increasing physical activity”

Several stakeholders wanted a study comparing lifestyle changes to medications. Comparisons mentioned from questionnaire responses included tamoxifen or aromatase inhibitors compared with diet/exercise and an arm combining medications and lifestyle changes. A basic science researcher commented on research in other fields demonstrating that exercise up-regulated certain gene expressions (for example in depression) and that it would be good to understand at a physiological level whether exercise has similar effects on breast cancer genes. However, as mentioned in questionnaires and informational interviews, diet and exercise are complicated interventions and it is important to understand what specific factors are necessary for the intervention. For these reasons, an evidence review that would review the literature on the effectiveness of lifestyle interventions to reduce the risk of primary breast cancer may be particularly helpful both to inform current patient decisionmaking and future research in this area. Such a review could evaluate the effectiveness of multiple lifestyle interventions (weight loss, exercise, diet, green tea, and fish oil) that were mentioned by stakeholders and suggest promising interventions to reduce the risk of primary breast cancer. The findings could help in the design of the interventions as well as study designs by highlighting important limitations in prior work, barriers, and specifying individual or combination therapies to be considered in future studies. A preliminary search of the literature found there are likely to be sufficient studies to inform a systematic review of the effectiveness of non-medication based interventions to prevent breast cancer with over 800 abstracts and a handful of comparative studies.

The third high priority research question for the prevention of breast cancer is: What factors influence the acceptability and effectiveness of risk reduction treatment (Research Question C)? The biggest difference between the CER analytic framework and the future conceptual framework is the addition of a diamond (used in behavioral intervention frameworks) to represent influencing factors, with action arrows extending both to the arrow between the population and intervention and the intervention and outcomes. Influencing factors, such as patient-provider communication and decisionmaking, attitudes and prescribing practices, insurance status, community, and exposures on risk and availability and susceptibilities to treatment were consistently among the highest priority items mentioned among stakeholders and were the leading priority for clinicians, research funders, and researchers.

Some stakeholders mentioned wanting to understand what barriers prevented providers from prescribing SERMs and patients from taking them. The results of such research would be helpful not only for existing medications but also for upcoming medications such as aromatase inhibitors. While we identified a questionnaire of family medicine providers, obstetrician/ gynecologists, and internal medicine primary care providers regarding their practices for breast cancer prevention screening and prescribing SERMs, the study has a number of methodological limitations.¹⁴ Providers were asked to self- report to questions specific to screening for breast cancer and prescribing SERMs. The degree to which providers were prescribing SERMs specifically for breast cancer prevention compared with osteoporosis was not discussed. Given the inclination to provide positive responses (e.g. higher prescribing of SERMs), creative scenario-based questioning or questionnaires that combine characteristics of the patient or prevention conditions might provide a better understanding of providers’ behaviors and attitudes towards the use of SERMS to reduce the risk of breast cancer. This also presents an area where an evidence review may be helpful to inform and guide future research as well as clinical practice. Furthermore, stakeholders wanted to know about the best strategies to communicate risks to patients, how to have discussions about harms and benefits of preventive therapies, and how to ensure that both patients and providers were up to date on current research. We conducted a preliminary search which identified 400 abstracts relating to breast cancer and communication and attitudes.

Consumers and policymakers were particularly interested in the degree to which environmental, economic, community, and social factors influenced decisionmaking, options, and outcomes. Stakeholder comments such as below reflect that influencing factors are critical to patient-centered care and comparative effectiveness research.

“We need studies that go beyond racial and ethnic disparities. As we all know, disparities just means “difference.” What matters is what leads to those differences and is often social and economic and racial inequities. Studies should look at what societal changes would have most impact on risk reduction in communities of color.”

“What social, economic, medical barriers prevent high risk women from using chemo-preventive agents?”

“I wonder if you want to do studies about other influences…because I just personally feel that clinicians aren’t that influential anymore. It’s more CNN and my neighbor and my cousin and my mom with cancer… the social network theory around health and disease. Social networks have a lot to do with how we do things.”

“How can physicians or other health care providers best support a persons’ decisionmaking process who is considering preventive care for breast cancer. How can we ensure the provider is up-to-date on the latest research, that he/she has explained that information to a patient in a way he/she will understand and then provide an opportunity for the patient to ask questions and seek additional information, guidance and support?”

“The provider-patient communication dynamic is imperative to good decision-making. If we can understand this better, then we can find those populations where communication can be improved.”

“There is a primary disconnect between patient and physician understanding/perception of risk-benefit rations for chemoprevention agents that is both poorly documented and clearly not understood. Well designed studies are needed that integrate health literacy and communication and target patients AND physicians.”

“Even more critical would be the development of tools that would facilitate this communication in a busy primary care practice. Such tools should communicate the patient’s breast cancer risk, the benefits and harms of risk reduction therapy and lead the patient through a decision-making process.”

As demonstrated by comments above, clinicians, consumer, research funders, researchers and policymakers were concerned about how to best disseminate information to ensure that patients and clinicians were able to make informed decisions based on high-quality evidence. They also wanted to understand the patient-provider communication process and the most effective method for communicating risks and facilitating decisionmaking. Because they are important to stakeholders and integral to patient-centered care, we believe that influencing factors should be depicted when appropriate in CER frameworks. Depicting influencing factors in frameworks encourages the reviewers to look for related evidence, and raises the readers’ awareness of their importance. For those reasons, we propose the addition of “Influencing factors (I)” to PICO as I PICO. The paradigm of research embodied in this framework promotes interdisciplinary and translational research teams that have been endorsed nationally.

Future Research Study Designs

To activate and inform future research, Table 7 lists all priority research topics that arose from narrative and structured responses, ongoing and completed research relating to that topic, and potential study designs for future research in that area organized according to the conceptual framework. From searches described above (See Methods: Identification of Ongoing Studies), the research team identified approximately 200 ongoing, recently completed, and/or funded studies from clinical trial registries, grant databases, and individual funders’ Web sites. These studies were listed according to stakeholder identified priorities.

Ongoing and completed studies still remain underpowered to assess the differential risk and effectiveness of preventive therapies based upon race or ethnicity. Similarly, while there are several studies of biomarkers, intervention studies do not appear to be collecting biomarker data which could advance our understanding of responses to treatment. In general, ongoing and completed studies focus on short-term intermediate outcomes such as mammographic density changes, hormone levels and precancerous lesions. Recognizing this limitation, some large interventions studies such as the STAR trial have added long-term followup. This is critical to understanding benefits and risks, to understand whether therapy prevents or delays the development of breast cancer, and to understand which population is most likely to accrue benefits rather than harms.