In addition to the general challenges associated with conducting RCTs, there are a number of unique characteristics associated with breast cancer which make some aspects of care difficult to assess with a RCT (Table 1). These largely reflect our success in treating breast cancer (with an associated improved survival) and the increasing role patients and their preferences play in treatment decision-making.

Implications of Prolonged Survival: With improvements in cancer detection and treatment, the prognosis for breast cancer patients has markedly improved in the past 20 years. The 3-year survival for the 3 million breast cancer survivors currently living in the United States exceeds 97 % for localized (node negative) and 84 % for regional (node positive) disease (Fig. 2) [4]. The increased overall survival and decreased local recurrence associated with modern breast cancer treatment represents a challenge in conducting efficacy RCTs, as predicted differences between treatment groups will be small and require an extended period of follow-up to identify. Conducting such trials would therefore require large patient cohorts to have adequate power which will be prohibitively expensive and time-intensive, and may not yield a meaningful clinical outcome.

One option to address this challenge is to limit inclusion criteria for clinical trials to select “higher-risk” sub-groups for whom differences may be more easily observed. However, such inclusion criteria may increase the difficulty in patient accrual, and would not represent a feasible method for studying processes with high cure rates like DCIS.

Patient Preference: When compared to other cancers, decision-making surrounding breast cancer management is especially preference-sensitive (decisions regarding mastectomy versus breast conservation, reconstruction choices, etc.). These preferences may also determine how patients perceive and ultimately decide to participate in, a clinic trial. Investigators must therefore address this in their trial design to ensure adequate trial accrual. For example, NSABP-06 randomized women between one of the three treatment arms: mastectomy, segmental mastectomy followed by breast irradiation, or segmental mastectomy alone. These treatment arms represent therapies with relevant differences in patient-centered outcomes including quality of life and body image. In recognition of the challenges to randomly assigning women to one of these very different arms, investigators designed the trial using a “pre-randomization” technique. After assessing women for eligibility this study, patients were pre-assigned to one of the treatment groups. Patients were then approached for participation and disclosure of their pre-randomization arm was included in the consent process. Only patients who accepted their pre-assigned therapy and provided informed consent were enrolled in the trial. A total of 2,024 women were randomized to the three treatment arms. Despite these measures, the assigned treatment was still refused by 175 patients, with 78 refusing mastectomy, 55 refusing segmental mastectomy plus breast irradiation, and 41 refusing segmental mastectomy alone [12]. Overall, however, this “pre-randomization” design allowed patients to consider their personal preferences for treatment when deciding whether to participate in the trial and allowed successful completion of the trial, which defines our current standard of care for breast cancer surgery.

In addition to representing a challenge to trial accrual, patient preference limits to some degree what types of questions can be answered in efficacy trials, as randomization for many questions may be either unethical or not feasible. One pertinent example is examining outcomes of risk-reducing surgery for women at elevated risk of developing breast cancer (i.e. BRCA1 and BRCA2 mutation carriers). To date, there is no level 1 evidence supporting a survival benefit to risk-reducing surgery, and the necessary study is unlikely to ever occur given the ethics associated with randomizing high-risk women in such a preference–laden scenario. Other examples of clinical questions unlikely to be answered through efficacy RCTs include consideration of the role of contralateral mastectomy for women with a new cancer diagnosis and outcomes after different types of post-mastectomy reconstructive procedures.

Studying Patient–Centered Outcomes: With the improvement in recurrence and survival observed in breast cancer, increasing emphasis has been placed on alternative patient-centered outcomes such as range of motion, quality of life, sexual function, cosmesis, and patient satisfaction. These outcomes can be negatively impacted by the breast cancer treatment administered, and are highly valued by breast cancer survivors. Unfortunately, it can be challenging to measure these outcomes as appropriate and sensitive tools are not always available. Additionally, the relative importance of many of these outcomes may vary based on the individual patient, and objective and subjective measures of the same outcome may not correlate. For example, in a prospective study of lymphedema in women who underwent SLN biopsy or an ALND, significant differences were observed in the rates of patient-reported (subjective) and objectively measured lymphedema [13]. Further, many survivors with objectively measured changes in limb volume did not report having clinical symptoms of lymphedema. This example highlights some of the challenges associated with evaluating patient-centered outcomes and even defining what outcomes should be measured.

Further, understanding patient satisfaction in relation to surgical decision-making, reconstruction, and cosmesis is in its infancy. Although reconstruction plays a key role in patient satisfaction for their surgical breast cancer treatment, it has no effect on cancer recurrence or survival and is therefore rarely reported in RCTs. Assessment of cosmetic outcome is extremely challenging and difficult to quantify, although robustly validated measurement tools are becoming more readily available [14]. Patients’ assessment of their cosmetic outcome is heavily impacted by pre-operative expectations and their overall satisfaction with their cancer care, resulting in possible discordance between objective and subjective measures. Finally, there are notably wide regional, socio-economic, and demographic variations in rates of breast reconstruction after mastectomy that are poorly understood. Although these variations are likely multi-factorial, patient preferences and values play an important role in patient decision-making for reconstruction and the differences in priorities between different populations represent an additional challenge in assessing these and similar outcomes.

In day-to-day practice, physicians routinely encounter clinical questions about breast cancer care which have not been satisfactorily addressed with RCT data and are unlikely to be ever addressed in this manner. However, many of these questions may be appropriate to examine using CER methodologies. Summarized below are several key clinical questions in breast cancer which do not currently have RCT-based evidence available to guide decision-making. We will discuss why these particular questions are important to address as we strive to improve the quality of care we provide breast cancer patients, examine why they are not amenable to study using typical efficacy trials, and explore how these questions may be approached from a CER perspective.