An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Karen Wernli, Susan Brandzel, Diana Buist, Mary Bush, Wendy DeMartini, Laura Ichikawa, Cameron Haas, Louise Henderson, Dianne Johnson, Karla Kerlikowske, Dave McColgin, Diana Miglioretti, Larissa Nekhlyudov, Tracy Onega, and Dori Rosenberg.
Author Information and AffiliationsAnnually, more than 250 000 US women are diagnosed with breast cancer and are recommended for yearly surveillance mammography for second cancers after treatment completion. Many also receive breast magnetic resonance imaging (MRI), without evidence of effectiveness.
(1) Assess patient and provider perspectives on surveillance breast imaging; (2) compare effectiveness of breast MRI with or without mammography to mammography only; (3) develop patient decision aids with education on breast imaging and results from imaging tests.
The research was based within 5 breast imaging registries in the US Breast Cancer Surveillance Consortium. Using a structured interview guide we conducted 6 focus groups with 41 women who had a personal history of breast cancer. We conducted a systematic literature review, following all relevant standards. We identified 33 938 surveillance mammograms and 2506 breast MRIs from 13 266 women following treatment for stage 0 to III breast cancer diagnosed in 2003-2012; second cancer events were ascertained within 1 year after imaging. We estimated performance measures using end-of-day radiologic assessment and compared them using multivariable logistic regression to adjust for potential confounders. Finally, we developed a prototype, web-based surveillance decision tool with information about breast imaging and results, incorporating women’s personal characteristics.
Women reported anxiety when anticipating surveillance results and discomfort but high trust in imaging. Our systematic review indicated need for robust analyses, including comparative studies in breast imaging. We observed 397 second breast cancers within 1 year of mammogram (286 invasive, 108 ductal carcinoma in situ [DCIS], 3 unknown) and 44 cancers within 1 year of breast MRI (30 invasive cancers, 13 DCIS, 1 unknown). Unadjusted performance measures comparing breast MRI to mammography only (respectively, per 1000 examinations), were cancer detection rates (within 1 year of exam): 10.8 (95% confidence interval [CI], 9.6-12.0) vs 8.2 (95% CI, 7.98.5); and interval cancer rates: 6.8 (95% CI, 5.8-7.8) vs 3.5 (95% CI, 3.3-3.7). Sensitivity was 61.4% (95% CI, 46.5-76.2) vs 70.3% (95% CI, 65.8-74.8); specificity was 88.2% (95% CI, 86.9-89.5) vs 88.5% (95% CI, 88.1-88.8); and positive predictive value of cancer among biopsied women was 19.5% (95% CI, 12.3-26.7) vs 30.5% (95% CI, 27.0-34.0). By multivariable logistic regression, breast MRI was associated with improved cancer detection rate (OR = 1.68; 95% CI, 1.04-2.69; p = 0.03) and specificity (OR = 1.20; 95% CI, 1.03-1.40; p = 0.02), with no statistically significant difference in sensitivity (OR = 1.10; 95% CI, 0.45-2.72) and a 2.23-fold increased biopsy rate (95% CI, 1.86-2.66).
Women are challenged to understand surveillance breast imaging. Surveillance breast MRI compared with mammography in community practice had 2-fold higher biopsy rates with improved cancer detection, without improvement in sensitivity or a decrease in interval cancer rates. Our promising prototype decision aid provides women and clinicians with information about surveillance imaging. In sum, our results fill a gap about breast imaging for women with a personal history of breast cancer and can facilitate survivorship care planning after a breast cancer diagnosis.
In 2017, an estimated 252 700 women were likely diagnosed with breast cancer,1 and nearly 90% will survive the first 5 years after diagnosis because of early detection and improved treatment.2,3 More than 3 million US women are living with a personal history of breast cancer (PHBC),1 which affects 1 in 43 women older than aged 204 If a woman diagnosed with breast cancer survives the first year after diagnosis, her 5-year survival is estimated at 97% if the tumor was in the breast only and 84% if the tumor had some lymph node involvement.1 Women with PHBC represent one of the largest US cancer survivorship groups, and their prognosis is favorable, warranting additional imaging to detect second breast cancer events.
In the absence of signs and symptoms of breast cancer, women are recommended to receive annual mammography for surveillance imaging, per national clinical oncology guidelines.5,6 The purpose of surveillance imaging is to detect second breast cancer events: either a recurrent breast cancer or a second primary breast cancer.7 Detection of second breast cancer events in women without any symptoms is associated with earlier-stage tumors of smaller size without node metastases and ultimately lower breast cancer mortality, when detected in the contralateral breast.8 A second breast cancer event occurs in 1 in 89 US women, and death from breast cancer occurs in 1 in 909 US women up to 5 years after diagnosis.1,4
In observational studies in women with PHBC, surveillance mammography is associated with a 31% to 69% reduction in breast cancer mortality.9 Nonetheless, mammography is not a perfect test, and it has been shown to be less accurate in women with PHBC than in women without PHBC.10 Using data from the Breast Cancer Surveillance Consortium (BCSC), Houssami et al estimated that the overall sensitivity of mammography in women with PHBC was 65.4% (95% confidence interval [CI], 61.5%-69.0%) compared with 76.5% (95% CI, 71.7%-80.7%) in women without PHBC.10 Reduced sensitivity means that a smaller proportion of breast cancers occurring within 1 year of a mammogram are detected on mammograms in women with vs without PHBC. Sensitivity was lowest in women who were younger than 50 years old at diagnosis (51.0%), had denser breasts (55.3%), or received chemotherapy or other endocrine therapy (54.1%). Unfortunately, women with some of these characteristics also have higher breast cancer recurrence rates.11,12 Identifying imaging surveillance strategies for women with PHBC, including comparing modality and frequency, was at the top of a recent list of comparative effectiveness research (CER) priorities from the Institute of Medicine13 and Friends of Cancer Research.14,15
Use of breast magnetic resonance imaging (MRI) has rapidly increased in clinical practice,16,17 and the modality is now offered at approximately 70% of US radiology practices among members in the Society for Breast Imaging.18 The American Cancer Society, the American Society of Clinical Oncology, and the Society of Breast Imaging do not recommend breast MRI for routine surveillance and follow-up among asymptomatic women with PHBC,19 because of insufficient evidence on its benefits and harms.14,20,21 However, in 2018, the American College of Radiology recommended the use of breast MRI in women with a PHBC if they were diagnosed before aged 50 or have dense breast tissue.22 Prior to this announcement, breast MRI was already used in US clinical practice for surveillance imaging. Among women who received a screening breast MRI in the past 16 years, about 30% to 45% had a PHBC.16,17
Several reports have described the performance of surveillance breast MRI in women with PHBC.23-27 The largest observational study conducted by Lehman et al evaluated breast MRI performance in women with a PHBC compared with women with genetic mutations, in whom adjunct breast MRI is recommended.28 Among 915 women the calculated sensitivity was 80%, with 16 cancers detected overall (both invasive and ductal carcinoma in situ [DCIS]). Specificity was 94%, and the positive predictive value among biopsies recommended and performed was 25%. However, as with other single-institution studies, analyses were not adjusted for patient factors that could influence performance outcomes, and CIs of the unadjusted measures were not calculated.
A key clinical knowledge gap is CER on how breast MRI performs in addition to mammography compared with mammography alone in surveillance of women with PHBC; there are no previous systematic reviews on this topic. For our Surveillance Imaging Modalities for Breast Cancer Assessment (SIMBA) project, we conducted a mixed methods, observational CER study on breast MRI and mammography in women with PHBC breast cancer, with the following aims: (1) Assess and understand the patient and physician experience in the decision for surveillance breast imaging; (2) generate quantitative evidence on the difference in test performance and breast cancer–specific mortality between breast MRI compared with mammography overall, by subgroups of women, and by patterns of surveillance imaging used; and (3) develop prototype patient decision aids that detail outcomes of importance to patients and physicians for incorporation into clinical practice.
Patient engagement in the SIMBA study occurred by 3 distinct yet intersecting groups: patient partners, a patient advisory board, and a stakeholder panel. These groups provided perspectives throughout our research, from initial proposal development through data collection and discussions about dissemination of results.
Patient partners: Prior to submission, engagement with patients began with community focus groups. We invited women with a recent breast cancer to discuss imaging received after cancer treatment. Announcements were made through an employee newsletter, a young survivor’s group listserv, and flyers at a local breast cancer walk. Women were from the greater Puget Sound region of Washington State. Initial discussions centered on patient experiences of imaging after breast cancer treatment was completed. Listening to women’s experiences brought to light the variance in screening protocols after treatment in different care settings and demonstrated women’s strong desire to share their experience and insights. All participants were excited about the proposed study, and all indicated a willingness to stay involved as the research progressed. From these early meetings, Dr. Wernli invited 2 participants to be patient partners (Ms. Bush and Ms. Johnson) and invited remaining attendees to join a patient advisory board. Ms. Bush and Ms. Johnson were invited to be patient partners before the first submission for funding. They participated in all aspects of the study from inception to completion: reviewing and editing the proposal and attending scientific committee calls and meetings, stakeholder meetings, patient advisory board meetings, and team planning meetings with the principal investigator (PI, Wernli) and project manager (Brandzel). Their role evolved to serve as “translators” between patients and professionals involved in SIMBA. For example, participating in patient focus groups (described in aim 1) broadened their perspectives as they learned from the other women; this experience gave them more confidence to represent the patient point of view to other stakeholders. Patient partners thus collated women’s shared stories as well as relating their own.
The PI attempted to build strong patient involvement in SIMBA, by showing respect for all participants at each meeting and by giving patient partners meaningful roles and tasks. Over the 3-year study, the patient partners noted that they felt accepted and appreciated by the larger research team. Ms. Bush’s and Ms. Johnson’s research involvement included co-authoring 5 manuscripts and abstract presentations,29-32 being invited speakers on numerous presentations about patient-centered research to both local Seattle and national conference audiences,33-38 going to PCORI annual meetings,34,37,39 attending all patient focus groups (aim 1), writing and contributing to blogs about SIMBA,40,41 participating in an invited PCORI tweet chat for breast cancer awareness month in 2016,42 being invited speakers at a PCORI webinar on engagement,43 participating in the development of the patient decision aid (aim 3), and participating in 2 SIMBA videos.42,44 Further, they wrote primary content for and reviewed this final research report. The patient partners took the initiative for this work, beyond being co-authorship on manuscripts and scientific presentations. None of the other dissemination activities were anticipated at the start of SIMBA.
Having patients involved in this study gave a human face to the research process. Patient partners emphasized that patient decisions are complex, and these individuals provided inspiration for ways to distribute the information to have the most impact on women with PHBC.
Patient advisory board: The patient advisory board was formed with members of the initial focus groups. After funding was awarded, Dr. Wernli, Ms. Brandzel, Ms. Bush, and Ms. Johnson reached out to women of color with previous breast cancer, through advertising to communities of color, attending at a local play on breast cancer experiences of African American woman, and visiting prayer circles for women with cancer. Hence, patient advisory board members represented diverse ages, cancer stage at diagnosis, socioeconomic status, and cultures, and they received cancer treatment in a range of cancer treatment settings. A total of 11 women participated in the patient advisory board, which met every 6 months over 3 years of SIMBA. Strong relationships developed over the study period, and conversations were rich with experience. The patient partners attended patient advisory board meetings, helped create agendas, and gave presentations. In between meetings, the SIMBA team distributed newsletters to all members, with features often written by patient partners. Board members were informed about ongoing research activities across all aims, including detailed discussions and explanations regarding activities for all aims. Board members participated in creating the prototype decision aids (aim 3) for women to learn more about surveillance imaging examinations and the type of imaging they might prefer. Board members participated in user-centered design experiences hosted by the design firm that generated prototype decision aids, offering feedback in their development. We held our final board meeting jointly with the stakeholder panel and research team so all could share their experiences and reflect on SIMBA final results.
Stakeholder panel: The SIMBA team, including the patient partners, hosted biannual stakeholder advisory board meetings with the following contributors: (1) Cary Kauffman, MD, breast surgeon, Bellingham, Washington; (2) Sara Javid, MD, breast surgeon, Seattle Washington; (3) Jennifer Specht, MD, medical oncologist, Seattle, Washington; (4) Nicole Grous, MD, medical oncologist, Olympia, Washington; (5) Janet Chestnut, MD, medical oncologist, Spokane, Washington; (6) Janie Lee, MD, breast radiologist, Seattle, Washington; (7) Tim Gleason, MD, breast radiologist, Seattle, Washington; (8) Robyn Sneeringer, MPH, Susan G. Komen Puget Sound, Seattle, Washington; (9) Nici Feldhammer, MEd, senior manager, Hospital Systems of American Cancer Society, Seattle, Washington; (10) Sean Stitham, MD, physician in the Clinical Review Unit, Kaiser Permanente, Seattle, Washington; and (11) Larissa Nekhlyudov, MD, primary care physician, Boston, Masssachusetts.
Each meeting was a 5- to 6-hour discussion of SIMBA research activities and progress. Guest attendees at meetings included (1) Caprice Greenberg, MD, MPH, University of Wisconsin, who discussed her work on breast imaging after cancer diagnosis, and (2) members of the Artefact design firm, who talked about developing the prototype decision aid (aim 3). From our discussions, we wrote a manuscript on the systematic review of breast MRI in women with PHBC, described in aim 2. As mentioned above, the final meeting was with the patient advisory board and research team.
We conducted this work within the BCSC.45 The BCSC is a collaborative network of breast imaging registries in community-based settings with linkages to tumor and/or pathology registries. It began in 1994 in response to the Mammography Quality Standards Act.46 The BCSC is supported by a central Statistical Coordinating Center (SCC). Five registries contributed data: Carolina Mammography Registry, Kaiser Permanente Washington, New Hampshire Mammography Network, San Francisco Mammography Registry, and Vermont Breast Cancer Surveillance System.5 BCSC data are prospectively collected from women and radiologists as part of routine clinical care at the time of breast imaging. Each registry site sends its data to the SCC for pooling and statistical analysis. All data undergo rigorous quality control checks. The BCSC meets all key definitions of a registry from the PCORI Methodology Report,47 including measurement of outcomes in people of interest, descriptive data linkages, follow-up for outcomes with cancer registries, rigorous standards for data safety and security, consistent data collection across registries, and use of appropriate statistical analysis.48 All registries follow similar data collection methods and adhere to data quality and integrity.45,46
Each registry and the SCC received Institutional Review Board approval for either active or passive consent or a waiver of consent to enroll participants, link study data, and perform analytic studies. All procedures are Health Insurance Portability and Accountability Act compliant. All registries and the SCC have received a Federal Certificate of Confidentiality and other protection for the identities of women, physicians, and facilities that are subjects of this research. Our research project is registered at clincialtrials.gov (NCT02212834).49
We conducted 6 focus groups in 3 registry locations (New Hampshire, North Carolina, and San Francisco) to better understand the experiences and preferences for surveillance breast imaging of women with a PHBC. Women were recruited for the focus groups by the registry staff, either through direct recontact with women from the BCSC registry or through community-based efforts. Eligible women were 18-75 years old with a diagnosis of American Joint Committee on Cancer (AJCC) stage 0 to III breast cancer within 5 years of the focus group date. Eligible women must have completed initial treatment for breast cancer to begin surveillance imaging; however, women could be taking adjuvant hormone therapy. Women’s surveillance imaging history was not a specific selection criterion. Women needed to be able to speak English and travel to the focus group location. Focus group discussions lasted 2 hours, and women received $60 for their time and travel. The SIMBA team developed a semistructured interview guide with input from patient partners. Specific questions on experiences with surveillance imaging asked about (1) information received about surveillance examinations after breast cancer treatment, (2) type of imaging received, (3) patient satisfaction with imaging received, (4) impact of imaging costs, (5) experience with recall from a positive surveillance examination that was not cancer, and (6) sources of information for breast health. The facilitator (SDB) used the semistructured interview guide to ensure that the same core questions were asked at all 6 focus groups. She also probed with follow-up questions based on the discussion content. All focus groups were recorded by a professional stenographer, and a deidentified, verbatim transcript was made for analysis. Each focus group was attended by at least 1 patient partner, who contributed as an expert for the research team. All research activities were approved by the local Institutional Review Boards of the focus group site.
Two members of the study team (SDB and DER) were responsible for the main analysis of the focus groups. A third team member (KJW) provided initial input on the coding list and was available to review identified codes. A combined deductive and inductive thematic analysis was conducted. All coders reviewed an initial transcript to develop a codebook that was iteratively refined and updated as additional transcripts were coded. We then grouped codes into themes that described experiences and preference patterns. We used qualitative software Atlas.ti version 7.5.250 to organize the codes.
Each of the 3 registry sites (New Hampshire, North Carolina, and San Francisco) identified clinicians in primary care, oncology, radiology, and surgery as potential interviewees. Site investigators initially contacted eligible clinicians by email. The project PI (KJW) followed up by scheduling a 20- to 30-minute interview on clinical practice and use of surveillance breast imaging in women with PHBC. Clinicians provided oral consent and permission to audiorecord the conversations. They were asked to discuss (1) their medical background and training, (2) whether they created post-treatment care plans for women with breast cancer and the guidelines they followed, (3) who made recommendations for surveillance imaging, (4) why they might recommend a woman receive a breast MRI after treatment for breast cancer, and (5) women’s involvement in decision making about breast imaging after breast cancer treatment. Clinicians received $100 for their time. Conversations were audiorecorded and transcribed. The PI summarized clinical practice and reasons for use of breast MRI. A total of 7 physicians completed the interviews for an average of 2 per registry site. We did not conduct formal analysis of clinician interviews. All research activities were approved by the local Institutional Review Boards of recruitment.
We conducted a systematic review assessing the test performance characteristics of breast MRI among women with PHBC, using standards for systematic reviews that include an emphasis on diagnostic accuracy.51-53 The review was registered with PROSPERO on February 29, 2016 (CRD42016035823). We searched PubMed and EMBASE for English-language publications from January 1, 2000, to September 6, 2016, using prespecified Medical Subject Headings (MeSH) terms. We excluded studies from before 2000 and publications that were limited to abstracts or that provided insufficient details. We used Covidence online software for data management.54 Three investigators (CH, LN, KJW) selected studies in 3 phases. First, 2 investigators independently made inclusion/exclusion judgments based on titles and abstracts of all identified articles based on criteria specified a priori. Excluded articles were justified by the investigator according to the specified criteria, with the option to add reasons not listed. Disagreements were resolved by a third investigator as an arbitrator. Articles with insufficient information in the title or abstract were included in the next phase. Second, studies underwent full text review, again by 2 investigators with the third as arbitrator. The primary reason for exclusion was “wrong patient population.” These studies mostly failed to isolate women with a PHBC as a population for breast MRI and frequently included women with other high-risk factors, such as BRCA mutations. Finally, we included full-text articles that met our eligibility criteria for data abstraction and quality assessment. From each included study, we abstracted key data and findings (even if not originally reported as an outcome) to calculate cancer rate, cancer detection rate, percentage invasive cancer, sensitivity, specificity, recall, positive predictive value (PPV)1 for abnormal findings at screening, and PPV3 for biopsies performed.55,56 Data were insufficient for a meta-analysis because there were too few studies to analyze and heterogeneity across studies.52,57-59 Hence, we qualitatively summarized the results.
We used registry data from the BCSC, pooled at the SCC, for study analyses. We describe the sources of data and the methods used for analyses. All research activities for this work were approved by the Kaiser Permanente Washington Health Research Institute IRB.
We identified women with a primary incident breast cancer diagnosis of DCIS (stage 0) or AJCC60 stage I to III invasive cancer at aged 18 or older between January 2003 and June 2012, using state or Surveillance, Epidemiology, and End Results (SEER) cancer registries and pathology databases. We excluded women treated with bilateral mastectomy.
Women’s characteristics included age at diagnosis, race/ethnicity, and education. We geocoded women’s addresses based on US Census block group and linked to median household income. Primary breast cancer characteristics included AJCC stage, treatment with surgery, chemotherapy and radiation, histology (invasive/DCIS), estrogen receptor (ER) and progesterone receptor (PR) status combined as both negative (ER–/PR–) or either positive (ER+ or PR+), and grade accounting for cancer type (ie, DCIS vs invasive). Primary breast cancer mode of detection was based on history of mammogram, ultrasound, and breast MRI examinations and categorized as screen detected, interval, or clinically detected.61,62 We defined receipt of preoperative breast MRI as any MRI examination within 2 months prior or 6 months after primary diagnosis date, based on prior work.63 We did not have information on breast cancer mutation status (ie, BRCA mutations) or prior chest irradiation as a child, which are 2 reasons for breast MRI referral.
We included mammography and breast MRI examinations conducted from 2005 to 2012. Examination characteristics included indication for exam, American College of Radiology Breast Imaging Reporting and Data System (BI-RADS)3,55,56 assessment, and BI-RADS breast density (mammogram only). For breast MRI examinations and mammograms with missing breast density, we ascertained breast density from the closest mammogram occurring at least 6 months after the primary cancer diagnosis. We included examinations with a screening indication and excluded those (1) within 6 months of the primary diagnosis date; (2) within 60 days after an examination of the same modality; (3) missing BI-RADS assessment or assessment 6 (known cancer diagnosis); or (4) from facilities with incomplete capture of second breast cancer events, where the biopsy rate after a BI-RADS category 4 or 5 assessment on a screening mammogram without prior breast cancer was less than 65%, to minimize loss to follow-up.64 We calculated years since primary breast cancer diagnosis and years since prior surveillance examination at time of imaging exam.
We used cancer registry and pathology files to identify second breast cancer events, either diagnosis of DCIS or invasive cancer, within 12 months of surveillance exam. We considered either a new primary or cancer recurrence as a second breast cancer event. We did not ascertain distant recurrences outside of the breast as second breast cancer events. BCSC registries follow women for cancer events based on systems for data ascertainment with facilities, minimizing missing second breast cancer events and loss to follow-up.
We calculated performance measures for breast imaging surveillance examinations using standard BCSC definitions based on BI-RADS end-of-day assessment and second breast cancer diagnoses within 1 year of a breast imaging surveillance examination (Table 1).61 To further clarify, the preceding surveillance examination was the index test, and we used the subsequent occurrence of second breast cancer diagnosis to calculate performance measures (eg, sensitivity and specificity), as is the gold standard in evaluating breast imaging. We used BI-RADS assessment at the end of the day (ie, after any additional imaging work-up) for all measures,61 except PPV (ie, the proportion of women with a second breast cancer event among those who screen positive from the index exam) of cancer among women who received a biopsy (PPV3; ie, the proportion of women with a second breast cancer event among those who received a breast biopsy). BI-RADS end-of-the-day assessment allowed for comparison between surveillance index mammography with any additional imaging resolved within the same day, with surveillance breast MRI, which also resolves to end-of-day assessment (Figure 1). We considered a BI-RADS end-of-day assessment 1 or 2 as negative, and 0, 3, 4, or 5 as positive.55,56,61
Definition of Surveillance Examination Performance Measures.
Description of BI-RADS Assessment and Categorization Based on Calculation of All Performance (Except PPV3) and PPV3, per American College of Radiology BI-RADS Atlas.
PPV3 used the final BI-RADS assessment, which differed from end-of-day assessment only for BI-RADS 0 examinations. These examinations are coded as 0 because the examination is an incomplete assessment and indicates additional imaging evaluation is needed for the patient. These examinations were followed up to 90 days for the first BI-RADS assessment that was not 0.55,61,For PPV3 we considered a subsequent BI-RADS assessment of 1, 2, or 3 as negative, and 4 or 5 as positive. We excluded examinations that could not be resolved to a nonzero assessment (N = 196 mammograms, N = 11 MRI) from PPV3 calculations.
We considered all-cause and breast cancer–specific mortality secondary outcomes in this analysis. We based this decision largely on input from our patient partners, who did not prioritize mortality as a planned outcome. Therefore, because our project focused on patient-centered outcomes, we prioritized other results. Other stakeholders also supported our approach of focusing on imaging performance rather than mortality, given that the mortality rate in women with stage 0 to III cancer would be low overall. We do not present results from mortality analyses.
Patterns of breast imaging: We describe the patterns of both mammography and breast MRI beginning 6 months after a breast cancer diagnosis. We examined indications for the imaging examination (ie, screening, short-interval follow-up, diagnostic evaluation, and staging/treatment) and intervals of surveillance imaging (ie, timing between tests).
Performance measures: We compared the performance of surveillance breast MRI with surveillance mammography only. We restricted the mammography group to examinations in women who did not receive a surveillance breast MRI at any time during the study. The breast MRI group included breast MRI examinations from women with and without surveillance mammography. For follow-up, among examinations without a second cancer, a date of death was recorded less than 1 year after the examination for 1% of mammograms and 0.5% of MRI examinations, and likely minimally affects internal validity (ie, loss to follow-up) required by PCORI Methodology Standards. From our proposal, we anticipated a surveillance sample of 2000 breast MRI examinations and 40 000 mammograms. Our final sample included more than 2500 breast MRI examinations and more than 33 000 mammograms. From the original power calculations, we accounted for clustering of examinations within women with an intraclass correlation coefficient of 0.5 and expected an effective sample size of 80%. We had 80% power to measure statistical differences with sensitivity (minimum detectable difference effect of 29.7% and odds ratio comparing breast MRI with mammography of 1844); specificity (1.9% and OR = 1.2); recall (2.2% and OR = 1.2); cancer detection rate (7.7/1000 and OR = 1.9); PPV1 (7.7% and OR = 2.3); and biopsy rate (1.2% and OR = 1.3). We did not have sufficient power to detect differences for measures of sensitivity, and we did not power on PPV3 or by subgroups.
Frequencies of characteristics for women, tumors, and treatments were described for both imaging modalities. We did not account for clustering by reading radiologist. We calculated unadjusted performance measures with 95% CIs for each group. We used logistic regression to compare performance measures of breast MRI with mammography only, adjusting for potential confounding factors. Our preliminary model adjusted only for age at diagnosis. We then used 2 approaches for more complete adjustment: Model 1, our primary model, included all possible potential confounders as model covariates; and model 2, our secondary model, used a 1:1 matched sample based on the propensity for MRI to compare imaging examinations under similar conditions.
Models 1 and 2 used Imputation and Variance Estimate Software (IVEware) Version 0.2 to impute missing covariate data.65 We created 5 imputation data sets, we fit models to each data set. Covariate data were missing mostly on primary tumor characteristics; hence, imputation was at the woman level and merged with exam-level data. The imputation model included BCSC registry, age at diagnosis, year of diagnosis, AJCC stage, histology, mode of detection, preoperative MRI, surgery, radiation, chemotherapy, ER/PR status (invasive only), grade, and breast density prior to diagnosis.
Model 1 adjusted for the variables in the imputation model plus age at examination, years since prior surveillance examination, and race/ethnicity, and included year and breast density at current examination instead of at primary diagnosis. Due to difficulties with model convergence for interval cancer rate and sensitivity, we used a reduced model that did not adjust for surgery, radiation, year of examination, or race/ethnicity. We fit the age-adjusted model and model 1 using generalized estimating equations (GEE) with an independent working correlation matrix to account for multiple examinations per woman.66 To ascertain differences in patient subgroups, we restricted analyses to 2 groups of women with PHBC: (1) with dense breast tissue; (2) diagnosed younger than 50 years old with primary breast cancer. We selected these categories a priori based on preliminary discussions with physicians to inform referral for breast MRI examinations for surveillance imaging.
For model 2, we calculated propensity scores for each of the 5 imputation data sets. Using logistic regression, we estimated the probability of breast MRI vs mammography only based on women’s characteristics at time of examination (BCSC registry, age at examination [quadratic], year of examination, time since prior examination, race/ethnicity, and breast density) and primary cancer characteristics (age at diagnosis [quadratic], AJCC stage, histology, preoperative MRI, and mode of detection). We used a SAS macro67 for 1:1 matching of mammograms and breast MRI examinations without replacement, using the logit of the propensity score with a caliper width of 0.2 standard deviations. We ran model 2 on the reduced sample of matched examinations using a 3-step define method68 to account for correlation among multiple examinations per woman and matched pairs.
All analyses except for multiple imputation used SAS Version 9.2 (Cary, NC: 2016, SAS Institute). We used the MIANALYZE procedure in SAS to combine results across the 5 imputation data sets. We considered 2-sided p < 0.05 statistically significant.
Cumulative probability of false positive and biopsy: We calculated false-positive (FP) rates for recall using initial end-of-day assessment and biopsy recommendation using final assessment for 3 groups: (1) mammography only; (2) breast MRI with no prior mammography; and (3) breast MRI with prior mammography. A woman could be in both breast MRI groups if she had a mammogram after her first breast MRI. We restricted data to surveillance screening rounds 1 to 3, using the SIMBA performance data sets (described above). We calculated the following outcomes at each screening round:
For breast MRI examinations, we used any mammogram occurring prior to the breast MRI to calculate first FP rate and cumulative FP rate. We did not adjust outcomes for patient covariates. Surveillance examinations could be missing due to the inclusion criteria for performance measures. Therefore, we also excluded examinations for any of the following:
The SIMBA study team worked with the design group Artefact69 to design and develop a prototype of a digital decision aid for women with PHBC to learn about surveillance imaging with mammography and breast MRI. Our collaboration used a human-centered design approach to develop a web-based app. Artefact investigated the requirements and principles that would provide women with the most effective understanding of SIMBA study results and give women confidence in their ability to have productive conversations about surveillance imaging with their clinicians. This work included input from (1) the SIMBA Stakeholder Panel (see section above), (2) our SIMBA Patient Advisory Board members, and (3) additional patients and clinicians identified by Artefact outside of SIMBA stakeholders.
Interviews conducted by Artefact with women with PHBC revealed preferences about the depth and breadth of clinical information, tone of the content, and initial predispositions (eg, women expressed interest in breast MRI if their initial breast cancer was not detected on mammography). These findings clarified communication needs. In later sessions, Artefact collaboratively designed with the SIMBA team multiple refinements to the prototypes, improving usability, clarity, and user comprehension.
Artefact also conducted a landscape analysis to discover and compare existing decision aids from any clinical domain. A literature review provided evidence-based best practices for surveillance imaging and criteria that were used to evaluate existing decision aids. The review identified principles from cognitive psychology for guiding more accurate risk perceptions, countering existing biases, and helping women better imagine the experience and outcomes of different surveillance imaging choices. We developed the SIMBA decision tool in adherence to the International Patient Decision Aid Standards, a research-based set of metrics and guidelines.70
The result of our design project was a functional prototype that gives women with PHBC information and experiential attributes about surveillance imaging with mammography and breast MRI (Figures 2-4). The prototype walks women through the experience of each imaging type, provides stories from women who have been through the process of choosing an imaging modality, and compares the risks and benefits of the 2 modalities. Women are encouraged to reflect on what matters most to them, with questions linked to a Likert scale to mark what is most important to them (eg, physical discomfort, health risks, cost) but without ordering or restricting responses. Other sections provide information on average out-of-pocket costs and allow for open-ended responses. Women can note questions and concerns to share with their clinicians. In theory, women with PHBC could receive individually calculated reports on comparative outcomes, based on their prior clinical history, and guidance on understanding which option is more aligned with their values and preferences. This personal report generated by the decision aid shows the probability of 4 different test outcomes: accurate detection of second breast cancer, accurate negative finding of breast cancer, FP result, and false-negative result. The report also highlights the benefits and risks that the woman might find most relevant to her understanding. We did not test the effects of the decision aid in this study.
Opening Site to the SIMBA Decision Tool Developed by Artefact.
Description of Key Characteristics of Mammography and Breast MRI.
Comparison of Potential Outcomes Between Mammography and Breast MRI.
We present the results for each of the specific aims.
A total of 41 women participated in the focus groups, discussing experiences and preferences regarding surveillance breast imaging. Women were aged 38-75, and 43.9% were diagnosed with DCIS or stage I breast cancer. The most common themes assessed from the focus groups are described below, as noted in our publication of our work.30
“I dread it for a couple of weeks before. My anxiety builds as I get closer to the date.”
“I had to find the right dose for Ativan just to get me comfortable enough to get my boobies in there.”
“The worst point is between when you do your initial images and then you’re waiting to see if you need more images. Because then as soon as they ask for more images, the mind just goes crazy about what they found and what they see.”
“I would rather deal with the false positives than miss something—or think I might miss something.”
“I want something to be so definite right away and not to have any issues, no false positives, nothing. I want to know, and I want to know now, and I realize that’s not possible.”
“You have to trust your doctor. That’s it. You have to trust that they know what they’re doing.”
“I’ve learned over time that I have to ask her questions, and my doctor, she’s well known for the type I have, but she is so smart that I sometimes keep quiet and don’t say anything. And I realize now, no, no, no, I don’t care how stupid I look, I’m going to ask her.”
“I’ve begged my oncologist, I’m like, ‘I want an MRI.’ And my oncologist says, ‘The mammogram did its job. Your cancer was found on the mammogram. We will continue with surveillance with mammography unless you have a problem.’”
“To me it seems like there is a logistical handoff problem—more knowledge is better. I would have, of course, liked a handout about that and everything else that I would need to know about.”
“I was told from the beginning that I would have to have yearly mammograms, just on the right side . . . they really make it very easy to do the whole follow-up.”
“My oncologist let me change to just mammograms every 6 months without the MRI. I like her manner. She says what she believes is the best, but then she doesn’t argue or make you feel badly if you want to do something else, but the thing is, of course, she is the expert.”
From interviews with 7 practicing clinicians within 3 US regions, we summarize responses to key informant interviews on breast imaging in women with PHBC.
Online library searches yielded 1289 unique citations. Based on title and abstract review, 1235 were excluded, and 52 full texts were reviewed. After review of full articles, we included 8 unique studies/articles in the data summary. Most studies were conducted in the United States. Patients and examinations were collected from 199926 through 2012.71 The largest studies were Lehman et al, with 915 participants, and Giess et al, with 691.27,72 The age range of women included in the studies was fairly homogeneous, with mean and median ages from 46 to 52 years. Cancer detection rates ranged from 9.9 to 19.3 per 1000 examinations; sensitivity ranged from 75% to 100%. PPV3 ranged from 17.9% to 43.5%. These findings supported the critical need for data analysis on the performance of breast MRI within women with a PHBC, as we had proposed, including comparative studies to mammography.
Our quantitative analysis included 19 900 women who underwent 69 516 mammograms and 3766 breast MRI examinations between 2005 and 2012 for any indication.73 Most women with PHBC received mammography alone (88.8%) or a combination of mammography and 1 MRI (6.1%). Less than 1% of women received a breast MRI without mammography in the first 5 years of surveillance imaging. Approximately 70% of surveillance examinations had an indication of “screening.” The proportion of indications varied by time since diagnosis and geographic location. These data were the background for more in-depth analysis and illustrate overall changes in breast MRI use over time.
Our final sample included 33 938 surveillance mammograms and 2506 surveillance breast MRI examinations in 13 266 women with a PHBC. There were 441 women with second cancers within 1 year of an exam, including 316 with invasive cancer, 121 with DCIS, and 4 of unknown type.
Patient and examination characteristics differed for women who received mammography only vs breast MRI (Table 2). Characteristics of women associated with receipt of MRI included younger age (<50 years), college degree or more education, residence in an area with higher income, interval-detected primary breast cancer, dense breast tissue, more recent diagnosis of breast cancer (>2008), preoperative breast MRI for primary breast cancer diagnosis, primary cancer diagnosis of invasive carcinoma, stage IIb or higher primary breast cancer diagnosis, and chemotherapy for first breast cancer. Among surveillance breast MRI examinations, 48% occurred less than 1 year after the prior surveillance examination compared with 12% for surveillance mammograms (Table 3).
Patient and Tumor Characteristics Associated with Mammography Alone or Receipt of Breast MRI, Breast Cancer Surveillance Consortium, 2005-2012.
Examination Characteristics at the Time of Breast Imaging, Breast Cancer Surveillance Consortium, 2005-2012.
Among the 33 938 mammograms were 397 second breast cancer events within 1 year, with 279 detected on surveillance mammogram. Among 2506 breast MRI examinations were 44 second breast cancer events, with 27 detected on surveillance breast MRI. In calculation of unadjusted performance measures, cancer rates were 11.7/1000 mammography-only examinations and 17.6/1000 breast MRI examinations, suggesting women undergoing breast MRI were at higher underlying risk for second breast cancer events (Table 4).
Performance of Mammography and Breast MRI Among Women with a Personal History of Breast Cancer, Breast Cancer Surveillance Consortium, 2005-2012.
Overall, unadjusted performance measures in women who received breast MRI compared with mammogram only were cancer detection rates of 10.8 per 1000 examinations (95% CI, 9.6-12.0) vs 8.2 per 1000 examinations (95% CI, 7.9-8.5); interval cancer rate of 6.8 per 1000 examinations (95% CI, 5.8-7.8) vs 3.5 per 1000 examinations (95% CI, 3.3-3.7); sensitivity of 61.4% (95% CI, 46.5-76.2) vs 70.3% (95% CI, 65.8-74.8); specificity of 88.2% (95% CI, 86.9-89.5) vs 88.5% (95% CI, 88.1-88.8); and PPV3 of cancer among women who received a biopsy of 19.5% (95% CI, 12.3-26.7) vs 30.5% (95% CI, 27.0-34.0). Differences were observed in unadjusted performance measures but less apparent in adjusted models.
We calculated adjusted odds ratios comparing performance of breast MRI with mammography, adjusting for age alone, multivariate logistic regression, and propensity match models. In age-adjusted models, odds ratios (ORs, with 95% CI) of performance measures comparing breast MRI with mammography only showed statistically significantly higher biopsy rate within 1 year for MRI (OR = 2.44; 95% CI, 2.10-2.84) but lower PPV3 (OR = 0.51; 95% CI, 0.30-0.88). In model 1, breast MRI was associated with improved cancer detection rate (OR = 1.68; 95% CI, 1.04-2.69) and specificity (OR = 1.20; 95% CI, 1.03-1.40). However, breast MRI was also associated with a more than 2-fold increased biopsy rate (OR = 2.23; 95% CI, 1.86-2.66), with no difference in sensitivity or PPV3 (Table 5). We detected no other statistically significant differences based on cancer rate, interval cancer rate, or PPV1.
Outcome Models Comparing Performance of Breast MRI to Mammography in Age-adjusted and Multivariable Logistic Regression and Propensity Score Matching.
The primary analysis revealed that MRI increased the cancer detection rate by 68% compared with mammography. Because the cohort study observed only 27 cancers among the MRI-imaged women, the odds ratio was imprecise. The 95% CI was compatible with an increased detection rate as small as 4% and as large as 169%. Women imaged with breast MRI experienced a 123% increase in the rate of biopsy, which was measured precisely because the cohort study observed 253 biopsies among the MRI-imaged women (95% CI, 1.86-2.66). Sensitivity of MRI increased by only 10%, and the estimate was imprecise because of the small number of cancers among MRI-imaged women. The specificity of MRI increased by 20% and was estimated precisely (95% CI, 1.03-1.40). PPV1 increased by 61% in MRI-imaged women, with a wide confidence interval that included a 4% reduction in PPV1 as well as a 170% increase in PPV1. The 27 detected cancers among MRI-imaged women limited the precision. PPV3 was 37% lower, with a wide confidence interval that included a reduction as low as 67%. The upper 95% confidence limit could not rule out a clinically important 20% increase in PPV3 because the OR was calculated with only 23 events in MRI-imaged women.
Model 2 included 82% to 83% of women in the breast MRI group and provides an OR estimate if women with a breast MRI received mammography only. The median propensity score in the 5 imputed sets of matched samples was 0.2 for both the breast MRI and mammography-only groups compared with a median of 0.01 in the total mammography-only group. The magnitude of effect for each of the performance measures remained similar between model 1 and model 2 in a reduced sample with less precision. Biopsy within 1 year was consistent with model 1, with biopsy rates higher among women who received surveillance breast MRI (OR = 2.19; 95% CI, 1.65-2.90). No other statistically significant differences in performance measures remained. Thus, our analyses consistently showed higher biopsy rates for surveillance breast MRI compared with mammography within both models.
We evaluated differences in performance measures among certain subgroups by stratifying on women who had dense breast tissue at the time of the examination and women less than 50 years of age at breast cancer diagnosis (Table 6). In all subgroup analysis, only the rate of biopsy within 1 year remained significantly elevated for MRI compared with mammography, with more than a 2-fold increased risk, similar to our full model results.
Outcome Models from Logistic Regression Models with Data Restricted to Patient Subgroups.
A description of the study population included in the FP results is in (Table 7a). The mammography-only group included older women and more women with multiple screening rounds, compared with the 2 MRI groups. The MRI group with prior mammography had a shorter screening interval than that of the other 2 groups.
Characteristics of Study Population Included in Calculation of Cumulative Falsepositive Rates.
FP rates are in (Table 7b). FP recall rates for a current examination and first FP generally decreased by screening round in all groups. The exception was the MRI group without prior mammography, which had a small sample size at screening round 3. Compared with mammography only, the MRI group had higher FP recall for current and first examination without prior mammography but lower with prior mammography. This result was likely due to women having additional screening and a shorter screening interval; however, this strategy also led to a higher cumulative FP recall for MRI with prior mammography compared to recall for MRI without prior mammography; both were higher than mammography only. These results did not consider other factors such as age.
False-positive (FP) Recall by Imaging Type.
The results for FP biopsy recommendation were similar for MRI and mammography only (Table 7c). However, both MRI groups had higher FP rates for the current examination and first examination compared with mammography only. Our performance assessment showed that biopsies occurred more often following MRI than mammography only, with no difference in PPV3.
False‐positive (FP) Biopsy Recommendation by Imaging Type.
There are no specific results reported for this aim.
Our funded project used a comprehensive, mixed methodology approach to understanding the impact of surveillance imaging among patients and clinicians. We evaluated the effectiveness of breast MRI compared with mammography alone in women with a PHBC and developed a prototype patient decision aid to meet the needs of most breast cancer patients. Our systematic review indicated a clear gap in the need for robust analyses of breast MRI performance and comparative studies with mammography. Hence, our current study is the most comprehensive observational study to date of the accuracy of breast surveillance imaging for women with PHBC diagnosed with stage 0 to III breast cancer, including several performance measures, reporting of precision of our estimates, and data from several clinical sites across the country.
We found the use of breast MRI resulted in a 2-fold higher biopsy rate compared with mammography, with improved cancer detection but no significant increase in sensitivity or decrease in interval cancer rates. In addition, we found few differences in accuracy between the 2 imaging tests after accounting for confounding factors. If breast MRI was used even once among all 235 020 US women diagnosed with stage 0 to III breast cancer in 2017, an estimated 14 340 additional women annually would have undergone biopsy than if they had undergone mammography only. We lacked precision for some of the primary outcomes of interest and cannot rule out a benefit of breast MRI; however, with the lack of larger observational studies in this field and inefficiency of a randomized controlled trial, these results fill an important clinical gap to inform our understanding of breast MRI in women with PHBC. The American College of Radiology22 recently recommended breast MRI among women with PHBC diagnosed younger than aged 50 or with dense breast tissue; our subgroup analysis does not support this recommendation. Further, we found that women want personalized information about surveillance breast imaging to feel empowered in their clinical care, even if the recommendation for type and frequency of imaging is universal.
We identified that women with PHBC experience heightened anxiety at the time of surveillance breast imaging. Increased anxiety is demonstrated in other cancer survivors.74 However, better communication from clinicians regarding recurrence risk might be helpful for women to understand their own personal risk. Further, women expressed a need for clear guidance on next steps after treatment, which highlights the importance of survivorship care planning.75,76 Women received surveillance mammography more often than was recommended by national clinical guidelines. Clinicians interviewed suggested that the type and frequency of imaging is often determined by their practice’s clinical standards, and in some practices more frequent imaging with diagnostic mammograms is recommended for patients. Our SIMBA decision tool, currently still a prototype, could be important in facilitating communication between patients and clinicians in survivorship care planning.
Our systematic review demonstrated variation in performance measures among single institutional studies, with little information presented on clinical or demographic covariates.
We determined that women who received a surveillance breast MRI examination differed from women who received mammography alone. Women who received breast MRI were diagnosed at a higher stage and were more likely to receive chemotherapy for treatment, suggesting women who received a breast MRI might have been at increased risk of cancer recurrence within the first 5 years after diagnosis. We also determined that women who were more educated and lived in wealthier neighborhoods were more likely to receive breast MRI. Known disparities exist in access to breast MRI for detection of a first breast cancer.77 However, ours is among the first studies to also articulate differences among women with PHBC.
Prior studies of breast MRI used for surveillance document unadjusted sensitivity ranging from 75% to 100%24,27,78 However, most prior studies of the performance of breast MRI for women with PHBC were in single-institution academic settings, with small sample sizes (<1000 women), limited follow-up of cancer outcomes, and no adjustment for potential confounding patient characteristics.23,24,27,78 In 2016, Lehman et al published a single-institution comparison of breast MRI performance measures among women with PHBC compared with women with genetic mutations.27 The study included examinations from 915 women, of whom about 42% were less than 50 years old. The study had a calculated cancer rate of 22.4 per 1000 examinations, about 5 more cancers detected per 1000 examinations than our results. Cancer detection rate was also higher in the Lehman et al27 study population, at 19.7 per 1000 examinations compared with our result of 10.8 per 1000. Sensitivity in the Lehman et al study was 80% (16 of 20 true cancers detected), and specificity was 94% (841 of 895 negative results confirmed). In comparison, we measured breast MRI sensitivity of 61.4% and specificity 88.2%. Differences in the performance measures could be attributed to different features of the study population (eg, age) and overall prevalence of second breast cancer events. However, the Lehman et al had a calculated 95% CI for sensitivity of 56% to 94%, suggesting that our sensitivity estimates were not statistically different.
Prior studies of breast MRI in women at high risk for breast cancer due to strong family history, known mutation carrier status, or measures of lifetime cancer risk calculate the sensitivity of breast MRI as 77% to 79%,79,80 indicating improved performance for breast MRI over mammography. However, in these studies, mammography sensitivity is consistently reported at 30% to 40%, which is lower than standards of quality.55,56 Our calculation of breast MRI sensitivity was lower than expected, based on prior studies.10 Our measure of breast MRI sensitivity may differ from prior studies for several reasons. First, our results represent diverse, community-based breast imaging settings, where radiologists’ readings might vary more than in single institution or academic environments.81 BCSC radiologists achieve quality standards in breast imaging, as evidenced by their ability to achieve sensitivity benchmarks for screening breast MRI in a population of women with and without breast cancer.64 A second possible explanation for differences in sensitivity is that previous studies among women who had breast cancer may not have identified all women with false-negative results, and previously reported sensitivity measures could be overestimates.24,27,72,78,82,83 A challenge for studying women who had breast cancer is that cancer registries typically do not report ipsilateral second cancer events. Our study used pathology databases maintained by BCSC registries to ascertain second breast cancers, including recurrences. We excluded facilities with low biopsy capture to avoid missing breast cancer diagnoses. Given that breast MRI is an adjunct to mammography, multimodality breast imaging could have increased the false-negative rate of breast MRI in our sample. Most women receiving breast MRI also received mammography. Hence, a subsequent mammogram might have detected a second breast cancer that was missed on breast MRI. Of note, prior studies also did not account for use of mammography in their calculation of sensitivity. Hence, understanding the impact of new imaging tests for surveillance of second breast cancer events requires understanding the cancer rate in the population evaluated, women’s characteristics, and estimates of accuracy and precision.
In the absence of a large randomized controlled trial in surveillance breast imaging, the analysis and interpretation of surveillance breast imaging is complicated. Confounding directly impacts the performance measures and requires high-quality data to ascertain the important confounding factors. Using data from the high-quality BCSC registries provides the first community-based imaging results to answer important questions on effectiveness. Larger samples of women could potentially identify subgroups of women who might benefit from adjunct breast MRI with mammography. We did not observe differences in performance outcomes other than for breast biopsy by the 2 categories of women most recently recommended by the American College of Radiology to receive breast MRI. However, our data lacked power for measures of sensitivity and subgroup analyses.22 To accrue larger sample sizes, we would need to wait for additional data collection within the BCSC.
The SIMBA decision tool, developed in collaboration with our technology industry partner Artefact, represented a new approach in the design and development of patient-facing tools for clinical research results. Women from our patient advisory board were integral in usability development. They expressed a strong need for additional information regarding breast imaging and the pros and cons of imaging modalities, and a desire to understand data in ways that supported their understanding of their clinical care. Our existing relationships with patient and stakeholders, as required by PCORI, facilitated the collaboration with Artefact to have timely feedback when working on short timelines. The incorporation of user-centered design principles in the SIMBA decision tool improved our patient-centered outcomes and strengthened the value of our patient clinical tool. Because of this work, we believe that interactive experiences are effective ways to disseminate our results. We look forward to communicating the results and the prototype decision aid in ways that reach a broad audience and enhance shared decision making between women and clinicians.
A valuable contribution of our overall work has been the development of methods in patient-centered outcomes research (PCOR). As described in the Patient and Stakeholder Engagement section above, patient partners contributed to several key components of our research. Two important products were developed from this collaboration. First, we developed a patient-centered model of inclusion for PCOR that includes trust, flexibility, and inclusion, among other core values.29 Patient researchers are often included as part of PCOR teams without being provided with much information about how their role, responsibility, and perspective will be used as they participate. Our patient partners, Ms. Johnson and Ms. Bush, led the manuscript on this topic. Second, we developed a testimonial video for future PCOR teams to describe the direct impact to studies and teams of including patients in research. One of our patient partners summarized the impact of patients in research teams most eloquently by stating, “We can make medicine better by working together.”44 The tools we developed help satisfy a growing need for PCOR resources.
Our discussions with focus group participants might not be generalizable to the experience of all US women with PHBC. We attempted to increase variation in perspectives by including several geographic regions; however, we might not have achieved diversity in voice based on regional practice variation or race/ethnicity. Similarly, results from our quantitative analysis might also be limited in generalizability.
Not applicable to our observational study.
We evaluated performance measures among women who had dense breast tissue and in women who were diagnosed before 50 years of age. In both restrictions to the sample population, we observed similar magnitude of results for all performance measures, compared with our overall results. We did not identify any differences in patient outcomes when restricted to these 2 categories. The 2 groups of women were recently recommended by the American College of Radiology to receive breast MRI in addition to mammography.22 Our results indicate a continued 2-fold increased risk of biopsy among this group without other improvements in performance.
Important limitations to consider within our research include the following: (1) We were unable to adjust for all identified confounders, including 5-year breast cancer risk prior to breast cancer diagnosis, due to missing data. This factor was identified by clinicians as influencing their decision to recommend a referral for breast MRI. Our descriptive analysis of available data did not show differences between groups, and we did not further pursue imputing the missing data for an overall imputed risk variable. We might have overlooked missed breast cancer cases, but we suspect that few were missed. Given that other studies in similar populations have estimated higher sensitivity measures, we believe that we might have done better at finding all cancers within 12 months of surveillance exam, given our ability to find false-negative cancers. (2) We lacked sufficient power to detect differences in all performance measures; however, no previous study of breast MRI performance, including sensitivity, has had sufficient power to detect clinically meaningful differences, as indicated by the wide confidence intervals for this measure. Finally, we did not have data on BRCA genetic mutation carriers or prior chest irradiation. A proportion of women who received breast MRI in our sample might have had these rare conditions. (3) Our SIMBA decision aid remains a prototype. We were not able to develop the algorithm to generate personalized information for women or to test differences based on this personalization. Future work could further develop the tool for clinical decision making in survivorship. To put these limitations into context, we cite several notable strengths, including ascertainment of patient primary cancer breast cancer and treatment factors prior to second breast cancer event; linkage of breast imaging to second breast cancer events through both cancer registries and pathology databases; and robust statistical analyses that used multiple approaches to adjust for important clinical confounders.
Our qualitative results showed that women and clinicians are challenged by the transition from treatment to surveillance breast imaging. We identified knowledge gaps regarding the comparative effectiveness of surveillance breast MRI and mammography, and how to present information about these options in patient-centered ways. Surveillance breast MRI compared with mammography in community practice had 2-fold higher biopsy rates with improved cancer detection, without improvement in sensitivity or a decrease in interval cancer rates. These comparative effectiveness findings highlight benefits and harms from surveillance breast imaging with breast MRI. Our promising prototype decision aid will provide women and clinicians with information about surveillance imaging. In sum, our results fill a gap about breast imaging for women with PHBC and can facilitate survivorship care planning after a breast cancer diagnosis.
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
Research reported in this report was [partially] funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (#CE-1304-6656). Further information available at: https://www.pcori.org/research-results/2013/breast-mri-better-finding-second-breast-cancers-mammograms-alone-breast-cancer
Wernli K, Brandzel S, Buist D, et al. 2019 Is Breast MRI Better at Finding Second Breast Cancers than Mammograms Alone for Breast Cancer Survivors? Washington, DC: Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/5.2019.CE.13046656
This Final Research Report is distributed under the terms of use as outlined by the Patient-Centered Outcomes Research Institute (PCORI). For information detailing the use of PCORI content please go to: (See https://www.pcori.org/about-us/terms-use)
Your browsing activity is empty.
Activity recording is turned off.
See more...
