Errors in Conduct of the CNBSS Trials of Breast Cancer Screening Observed by Research Personnel

Abstract

Objective

To investigate why the Canadian National Breast Screening Study (CNBSS) did not show mortality reduction with mammography. This study explored long-standing concerns related to the validity of the randomization process, methods of recruiting women to participate in the trials, and training of the staff working in the CNBSS.

Methods

Surviving former CNBSS personnel, whose roles involved direct recruitment, enrollment, randomization, clinical examination, image interpretation, and management of patients in the CNBSS were interviewed. Individuals were contacted and consented to provide firsthand accounts of daily operations and adherence to research protocols via standardized questions. Consistency of observational data with quantitative results from the CNBSS trials was evaluated.

Results

Eleven of 28 (39.3%) staff confirmed that women with preexisting symptoms of breast cancer were systematically recruited at some centers; 57.1% (16/28) confirmed that personnel performing screening in CNBSS had very limited training and experience; 39.3% (11/28) verified that imaging equipment was often substandard; 50% (14/28) indicated that mammography image quality was generally poor; and 28.6% (8/28) corroborated that in some cases surgeons were unwilling to perform biopsies or surgeries for women with suspicious abnormalities found only on screening mammography that lacked a palpable correlate.

Conclusion

These firsthand accounts provide new information confirming that the CNBSS did not consistently and rigorously assess the true efficacy of screening mammography. The staff accounts clarify reasons why the CNBSS results were outliers compared with the six other randomized trials of screening mammography and should not be used as credible scientific evidence to inform health policy.

breast cancer, mortality, screening mammography, randomized controlled trials, research personnel

Key Messages

• Firsthand accounts by staff who worked for the Canadian Breast Cancer Screening Trials (CNBSS) in the 1980s confirm that women with preexisting symptoms of breast cancer were recruited and included in the trials, the training and experience of staff performing and interpreting the mammography was very limited, and mammographic-only findings were often not acted upon by the surgeons.

• Information from these accounts help explain the poor sensitivity of cancer detection and the apparent lack of mortality benefit from screening in the CNBSS compared to other randomized controlled trials of screening mammography.

• By recruiting many women with symptoms of breast cancer, the CNBSS were not true trials of screening mammography and their results should not be used as credible evidence to generate health care policy.

Introduction

Eight randomized controlled trials (RCTs) were conducted from the 1960s to the 1980s to evaluate the efficacy of screening mammography with regard to decreasing mortality from breast cancer (1–7). All trials confirmed significant benefits of screening mammography except the two in the Canadian National Breast Screening Study (CNBSS), CNBSS-1 for women aged 40 to 49 years and CNBSS-2 for women aged 50 to 59 years. In fact, a slight mortality increase was noted in the CNBSS-1 trial in the screening arm at 11 to 14 years of follow-up (8). The design of the CNBSS trials have led to them either being included (9,10) or excluded (11–14) in breast cancer screening meta-analyses. In 2002, the International Agency for Research on Cancer of the World Health Organization excluded CNBSS-2 from its breast cancer screening meta-analysis because the design of the RCT included clinical breast examination (CBE) in the screening intervention (13). Other meta-analyses excluded the CNBSS because it recruited and enrolled volunteers rather than inviting and tracking a randomized population sample (14). However, due to the large size of the CNBSS-1 trial and the specifically targeted age range of participants, it continues to heavily influence guidelines for breast cancer screening (15) of women aged 40 to 49 years (9,10). Multiple experts have hypothesized that the results of CNBSS-1, indicating no benefit or even possible harm related to screening mammography, reflect both design flaws and the failure to adhere to the standards of RCTs during day-to-day operations of the CNBSS (16–20).

It remains critically important to investigate whether the results of the CNBSS trials are valid and should continue to serve as the basis for national guidelines, recommendations, and policies for modern breast screening practices. This requires information along two general lines. Were the CNBSS trials valid studies of breast cancer screening in the era in which they were conducted? And, if so, do their findings translate to current screening? The purpose of this article is to share and critically consider new eyewitness accounts of the daily management, operations, methodology regarding inclusion and exclusion criteria, and randomization during the CNBSS trials to determine if they adhered to basic standards for RCTs.

Methods

To gain insight into the quality of the execution of the CNBSS, interviews of former staff members with firsthand knowledge and eyewitness accounts of the daily operations of the CNBSS were conducted. Interviews were designed to provide information about those questions. The need for consent was waived by the Ottawa Hospital Research Ethics Board.

We attempted to contact via e-mail, telephone, or social media all known individuals who were employed in various capacities at any of the clinical sites during active recruitment and execution of the CNBSS trials. Those who agreed to participate were included and asked a set of 14 questions about the study (Supplementary Material) by one of three interviewers (J.M.S., P.B.G., and S.A.). The staff were asked the same questions, which were provided by e-mail before the interviews. The questionnaire included questions about key aspects of daily operations, protocol adherence, and performance, including how the trial was conducted at their site; recruitment and admission processes for patient volunteers; who had knowledge of the CBE findings; inclusion of symptomatic women; the randomization process; quality of training of the personnel; quality of the mammograms; and general observations about the trials.

This study was designed to obtain data on the following questions:

1. Were the participants recruited to the trials in a manner to accurately represent the general screening population and to allow testing of the trial hypotheses? Specifically, were the participants asymptomatic women?

2. Was randomization conducted such that the only difference in outcome between participants allocated to the intervention and control groups could be attributed to the effect of the intervention?

3. Did the intervention (mammography and/or CBE) realistically reflect the contemporary quality of screening at the time of the trials?

   • Were the radiologists and technologists who conducted the mammography screening and the physical examiners adequately trained and experienced?

   • Was the technical quality of the mammograms reflective of high-quality screening and state-of-the-art practice circa 1980 to 1985?

4. Were the actions taken by clinical personnel in response to findings at screening appropriate and consistent with standard of care?

Finding the Staff

All staff listed on CNBSS publications (1,2) were invited to participate in telephone interviews performed over a six-month period between January 2021 and July 2021. Attempts to locate and contact former staff took place via social media and included requests for contact information from biostatisticians, epidemiologists, physicists, radiologists, surgeons, and technologists who had been employed by CNBSS. Staff who responded to the invitation were verified to have worked on the trials and cross-validated with CNBSS publications and contacts with other former staff.

An initial 90-minute online virtual (Zoom, San Jose, CA) meeting was held on February 6^th, 2021, with 4 radiologists who participated in the CNBSS, 1 radiologist who reviewed the study, the study lead physicist (M.J.Y.), and the radiologist interviewers (J.M.S., P.B.G., S.A.). The meeting was recorded. The purpose of the meeting was to obtain a broad overview of the CNBSS, to address areas of concern, and to explore the questions that were subsequently included in the focused interviews.

Individual interviews were conducted with other former staff by telephone and/or subsequent e-mail correspondence with one of the radiologist interviewers.

Results

Despite the passing of 41 years since the study was launched, 29 former staff were located and indicated a willingness to be contacted for telephone interviews and to share the information they provided. The average time for each of the interviews was 30 minutes. A summary of the accounts is provided in Table 1 (full report in Supplementary Material). Many staff were deceased, including 6 research coordinators and 4 radiologists, and several radiologists and technologists suffered severe cognitive decline and were unable to be interviewed. The names of 21 research coordinators, 15 course directors, 37 surgeons, and 43 radiologists were listed for each of the 15 centers in the CNBSS studies (1,2). The names of the staff technologists and nurse examiners were not provided, and it is expected that 1 to 2 technologists and nurse examiners were working at each center. Of a total of approximately 160 original staff, we were able to interview 28 individuals from 8 centers. Those interviewed consisted of 4 radiologist reviewers, 9 radiologists, 9 technologists, 2 research coordinators, 1 nurse examiner, 1 surgeon, and 1 physicist. One other staff member indicated that she had been asked by the trial management to sign a nondisclosure agreement and her responses were not recorded. The staff interviewed had, on average, worked on the study for 3.6 years (range, 1 month to 6 years). Two surgeons who were contacted declined to be interviewed. One more research coordinator agreed to be interviewed but declined to allow her responses to be shared due to a concern of potential liability and a lack of interest in uncovering any problems with the CNBSS. Her data were not included. Except for one who agreed to have her responses reported but not her name, all other staff interviewed agreed to provide their names and responses.

Table 1.

Witness Summary of CNBSS Trials Across Provinces

Witness	Witness Work Location (No. Years of Involvement)	Women With Symptomsa Included	Poor Mammogram Quality	Minimal Training of Research Staff, Radiologists, and/or Technologists	Surgeons Did Not Operate on Non-palpable Mammographic Findings	Use of Outdated Equipment
1. Radiologist reviewer	Toronto, ON (3)	-	Yes	-	-	-
2. Radiologist reviewer	Toronto, ON (1)	Yes	Yes	Yes	-	Yes
3. Radiologist reviewer	Toronto, ON (2)	-	Yes	-	-	Yes
4. Radiologist reviewer	Toronto, ON (1)	-	Yes	-	-	-
5. Physicist	Toronto, ON (6)	-	Yes	Yes	Yes	Yes
6. RT	Toronto, ON (0.08)	Yes	-	Yes	-	-
7. Radiologist	Toronto, ON (6)	Yes	Yes	Yes	Yes	Yes
8. Research coordinator, RN	Toronto, ON (5)	-	No	-	-	-
9. Radiologist	London, ON (3.5)	-	Yes	Yes	Yes	Yes
10. Research coordinator, RN	Hamilton, ON (4)	Yes	-	Yes	Yes	-
11. Radiologist	Hamilton, ON (3)	-	-	Yes	Yes	-
12. RT	Hamilton, ON (1.5)	Yes	-	Yes	-	-
13. Radiologist	Ottawa, ON (5)	Yes	Yes	Yes	Yes	-
14. RT	Ottawa, ON (5)	Yes	Yes	Yes	-	Yes
15. RT	Ottawa, ON (5)	Yes	Yes	Yes	-	Yes
16. RT	Ottawa, ON (5)	Yes	No	No	-	Yes
17. Radiologist	Ottawa, ON (5)	Yes	Yes	Yes	-	-
18. RT	Ottawa, ON (5)	Yes	Yes	Yes	-	Yes
19. RT	Edmonton, AB (5)	-	Yes	Yes	-	-
20. Radiologist	Red Deer, AB (5)	-	-	-	-	-
21. Radiologist	Red Deer, AB (5)	Yes	No	Yes	Yes	-
22. RT	Vancouver, BC (2.5)	-	Yes	-	-	Yes
23. RT	Vancouver, BC (3.5)	Yes	No	-	-	-
24. Withheldb	Vancouver, BC (NA)	-	-	-	-	-
25. Nurse examiner, RN	Vancouver, BC (2)	Yes	-	-	-	-
26. Radiologist	Halifax, NS (1)	-	-	-	-	Yes
27. Radiologist	Halifax, NS (5)	Yes	-	-	-	-
28. Surgeon	Halifax, NS (5)	-	-	Yes	Yes	-
Total (N = 28) n (%)		11 (39.3%)	14 (50%)	16 (57.1%)	8 (28.6%)	11 (39.3%)

“Yes” indicates that the witness directly observed and verbally confirmed the described inadequacy of the CNBSS protocol. “No” indicates that the witness directly observed the contrary. “-” indicates that the witness did not directly observe or confirm the described inadequacy of the CNBSS protocol.

Abbreviations: AB, Alberta, BC, British Columbia; CNBSS, Canadian National Breast Screening Study; NS; Nova Scotia; ON, Ontario; RN, registered nurse; RT, radiology technologist (medical radiation technologist in Canada).

^aBreast abnormalities noted by the participant or her health care provider prior to CNBSS clinical breast examination.

^bWithheld due to nondisclosure agreement.

The responses were recorded whether the staff could respond to the questions or not. Some questions could not be answered by the respondents given their lack of direct involvement in that aspect of the study, or if they preferred not to answer.

The CNBSS protocol states that the inclusion criteria were women aged 40 to 59 years who had no previous history of breast cancer, had no mammograms in the past 12 months, and were not pregnant (21). Although the intention of the trials was to evaluate detection of preclinical breast cancer, the protocol did not explicitly exclude women with symptoms and signs of breast cancer. It was confirmed by 41.4% (12/29) of the interviewees that women with palpable lumps or other symptoms, disclosed prior to performance of CBE, were included in the CNBSS (Table 1). One staff member (witness 6 in Table 1) directly witnessed protocol violation of the randomization process, in that women initially randomized to be in the control group were assigned to the mammography group during the registration process (19). This demonstrates that the requirement for consistent randomization was not met. In addition to multiple external review reports of poor quality mammography there were also internal reports suggesting that, in many cases, the quality of care in the intervention was inadequate (16,17,22–26). Very poor mammography quality was attested to by 50% (14/28) of the staff, while 14.3% (4/28) attested to the contrary and 39.3% (11/28) were unable to comment. Confirmation of a lack of training of the technologists and radiologists was provided by 57.1% (16/28) and 39.3% (11/28) attested to the poor quality or inadequacy of the mammography equipment. On the other hand, 53.6% (15/28) attested to the excellent training of the nurses for the CBE. This was consistent with a published report describing the training process(27).

It was confirmed by 28.6% (8/28) of staff that despite preclinical detection of suspicious findings on mammography, which represents the essential purpose of the test and the fundamental reason for conducting this nationally funded trial, some surgeons often declined to perform the recommended surgical biopsies given either their lack of confidence in the mammograms or their inability to localize the lesions.

Discussion

Excess of Advanced Cancers in the Mammography and Physical Exam Group

A key requirement for RCTs is that participants in the intervention and control arms must have equal characteristics at the outset (28). The published CNBSS reports emphasized how well matched the participants were for marital status, education, reproductive status, socioeconomic status, family history of breast cancer, smoking history, ethnic origin, and type of occupation (1,2). However, a plausible explanation for a large imbalance in the number of advanced breast cancers in the screening arm versus the usual care (UC) arm of the CNBSS-1 has not been provided.

The stages of cancers in the CNBSS trials were not reported, however, the authors did provide the number of cancers for which there were four or more positive accompanying axillary lymph nodes. These poor prognosis cancers have been used here, and by other authors (16,20,21), as surrogates for advanced cancers. Table 2 recapitulates the CNBSS-1 data for women with advanced cancers. In screening year 1, 8.3% (5/60) of the invasive cancers in the UC arm had four or more positive nodes. In the screening mammography and physical examination (MP) arm, 22.1% (19/86) of screen-detected invasive cancers had four or more nodes, 17 of which were detected by CBE. The excess of 14 advanced cancers in the MP group is highly significant. Not only the number of screen-detected advanced cancers, but the total number (including interval and incident cancers) was also greater in the MP arm (24.5%, 25/102) compared to 8.3% (7/84) (P = 0.0036) in the UC arm, as previously reported (2). After 6 years, the total number of advanced cancers in the screening MP arm (16.4%, 47/286) was significantly higher than in the UC arm (8.5%, 23/272) (P = 0.0066) (20). It is, therefore, not surprising that the breast cancer mortality rate in CNBSS-1 was 1.36 times higher in the screening MP arm than the UC arm at 7 years follow-up.

Table 2.

Numbers of Patients With ≥4 Nodes Relative to Invasive Cancers in CNBSS1

	MP Arm, n/N (%)					UC Arm, n/N (%)
Screening Year	Diagnosed by MG Alone	Diagnosed by CBE With or Without MG	Interval Cancer	Incident Cancer	Total	Patients With Nodes	Interval Cancer	Incident Cancer	Total
1	2/21 (9.5%)	17/65 (26.2%)	6/16 (37.5%)	-	25/102 (24.5%)*	5/60 (8.3%)	2/24 (8.3%)	-	7/84 (8.3%)
2–5	1/46 (2.2%)	8/73(11%)	5/37 (13.5%)	5/28 (17.9%)	19/184 (10.3%)	-	-	15/148 (10.1%)	15/148 (10.1%)
6 (follow-up year)	-	-	-	3/45 (6.7%)	3/45 (6.7%)	-	-	1/40 (2.5%)	1/40 (2.5%)
Total	3/67 (4.5%)	25/138 (18.1%)	11/53 (20.8%)	8/73(11%)	47/286 (16.4%)*	5/60 (8.3%)	2/24(8.3%)	16/188 (8.5%)	23/272 (8.5%)

The numerators are the number of patients with ≥4 nodes and the denominators are the number of patients with invasive cancers, with proportions in parentheses. Adapted from calculations based on information provided in Table 7 (2). Differences denoted by asterisks are highly significant. P ≤ 0.01.

Year 6 was not a screening year but the year post-screening to determine the number of interval and incident cancers.

Interval cancer was defined as a cancer diagnosed less than 12 months after a normal screen.

Incident cancer was defined as a cancer diagnosed after 12 months of a normal screen.

Abbreviations: CBE, clinical breast examination; MG, mammography; MP, screening mammography and physical examination arm; UC, usual care arm.

How did the unbalanced distribution of advanced cancers occur? Eyewitness accounts from staff, provided in this article, shed light on the long-suspected deviations from rigorous randomization. In all but one center, the CBE was performed by nurse examiners (except in Quebec, where it was performed by physicians) before the randomization allocation was given. Depending on the center, the coordinator or a clerk could be responsible for the allocation, and allocation was supposed to be done by writing the patient’s name in the next line on the list and then telling the nurse and patient what the assignment was. Simply skipping a line or two could allow preferential allocation to the screening MP arm. Although the coordinators were told that this was an important part of the study, one coordinator noted that it was difficult to do correctly and very easy to make errors. One technologist reported that for three months, she was asked daily to alter the list of patients allocated to screening mammography, adding one to two patients per day onto the mammography list because they had palpable findings, while removing one to two patients who had not yet been examined. This maintained equal numbers of women allocated to the two trial arms but preferentially included those with palpable findings in the MP arm (witness 6, Table 1).

The fact that protocol violation of randomization has long been suspected is well documented (9,16,17). In 1996, a group in Winnipeg reviewed previous health claims submitted for subjects who entered the CNBSS center in Manitoba. They found that although prior breast cancer was an exclusion criterion, 9 prior claims for breast cancer were found among the CNBSS participants. Of these individuals, 8 were assigned to the screening MP arm and 1 to the UC arm (P = 0.05). Of the women 40 to 49 years old, all 4 who had prior claims for breast cancer were assigned to the screening MP arm (P = 0.12) (18).

The CNBSS authors theorized that the higher incidence of advanced cancers in the screening arm was due to self-selection bias. Women who volunteered to participate were much less likely to refuse the allocation to screening compared with those women invited by letter, as is done in population-based RCTs. They also suggested that the higher rate of the advanced cancers in the screening arm was due to the improved detection of cancers with mammography. But this argument is belied by the fact that the vast majority of the excess cancers in the screening arm were palpable, whereas the expected effect of mammography would be to increase the number of non-palpable lower stage cancers detected.

The fraction of cancers detected by mammography alone in CNBSS was relatively low, consistent with the concerns expressed by 50% (14/28) of the interviewees over the quality of the mammography. This was confirmed by independent review (26) and is further discussed in a related article (18).

The higher incidence of advanced cancers in the screening MP arm is much more likely to be due to protocol violation of randomization, as described above. There were additional anecdotal reports by the staff who were interviewed suggesting that this occurred at other sites as well. Using data from the first report of CNBSS-1, one author (M.J.Y.) calculated previously that if as few as 7 women with advanced cancers had been systematically moved from the intended allocation in the UC arm to the screening MP arm, the mortality outcome results of the study would shift dramatically from a finding of benefit to one of no benefit (29).

Problems With Diagnostic Assessment of Screen-detected Abnormalities

Another major deviation from the protocol was observed for participants at some centers in whom there were suspicious findings on the original CBE (Table 1). The protocol required that patients with findings at CBE or mammography screening be subsequently assessed at the study review center by the review site surgeon (15), and these assessments were convened weekly. In cases where the surgeon confirmed suspicious findings, the patient would be referred to their family physicians, who were instructed to arrange diagnostic imaging, work-up, and treatment, if necessary, in the community.

One radiologist who worked both at CNBSS and a diagnostic center (witness 27, Table 1) noted that the CNBSS nurse examiners at their site frequently requested mammograms for CNBSS patients in whom they had detected a suspicious area on CBE. It is unknown whether this occurred in the UC or the MP arm. This indicated a protocol violation if in the MP arm or a deviation if in the UC arm. The protocol clearly specified that participants with a CBE finding in the UC arm should be seen first at the review center and examined by the review site surgeon (30), after which diagnostic mammography might be recommended. Bypassing the review and directly arranging a diagnostic mammogram constituted a protocol deviation. The extent to which this occurred for participants in the UC arm, the screening arm, or both is not known. The effect of performing mammography on the UC patients before the study review would have been an increase in the amount of diagnostic mammography being performed in the UC arm. This would have helped to lead to an earlier detection of a breast cancer that might otherwise have been dismissed by the study surgeon that year; 45.5% (1072/2358) of cases referred for study surgeon review in CNBSS-2 were considered to be false positive CBE exams and were not sent for diagnostic imaging (27). Because many of the CBE findings would be deemed by the surgeon as being false positives, waiting for the review clinic surgeon to examine the women would have resulted in fewer women in the UC arm being sent for mammograms. Therefore, by increasing the number of women in the UC arm undergoing prompt mammography, this protocol deviation could have further diluted the measured effect of the screening intervention. While this would not technically constitute screening, it certainly would increase the positive predictive value (PPV) in the UC arm. This represents an inconsistency in how CBE-positive results were managed and could have affected mortality outcomes. We do not have data on how frequently this may have occurred; however, 28% (659/2358) of patients within the UC arm underwent diagnostic mammography examinations (27).

Interval Cancer Rates

In the 1992 publication, the interval cancer rates in the CNBSS-1 screening arm (invasive plus in situ) were highest in the first and second years, 0.75 per 1000 (19 cancers) and 0.71 per 1000 (16 cancers), respectively, compared with the subsequent years (0.36 per 1000, 8 cancers; 0.46 per 1000, 10 cancers; 0.64 per 1000, 9 cancers) (2). In CNBSS, the surgeons were responsible for managing patients with abnormal mammograms because image-guided needle biopsy techniques for radiologists were not yet available. Surgeons decided if women with abnormal study mammograms would proceed to a surgical biopsy for diagnosis. Given that the PPV of the study was only 8.2% (31), the surgeons may have been reluctant to perform breast surgery when there was such a high probability of a benign finding, particularly where needle localization was not available. The fact that the interval cancers were highest in the first two years supports the assertion by several staff radiologists that many mammographic findings were dismissed and that the care plan was inconsistently followed for patients in the screening MP arm.

Limitations

The CNBSS was performed 35 to 40 years ago and the recollections of former staff may have been affected by the passage of time. The psychological literature suggests that memory errors occur more often in older individuals and with the passage of time (32,33). However, eyewitness studies show that free recall and recognition of memory with high confidence remain overwhelmingly accurate. A recent study at the University of Toronto found that high output memory accuracy (>94%) was achieved months to years later and was much higher than predicted (34). The authors of that study noted that the fact that memory is subject to contamination does not make memory inherently unreliable. In our study, the eyewitnesses only answered those questions that they could, using free recall and subsequent correspondence. They were validated by multiple other eyewitnesses.

Conclusion

In summary, staff accounts regarding the day-to-day operations of the CNBSS trials confirm many long-suspected errors in the conduct of the studies, some as a direct consequence of the trial design. Many symptomatic women were included, causing recruitment to favor women with a higher risk of breast cancer. The staff performing and reading the mammograms were poorly trained, while, conversely, the nurse clinical breast examiners underwent rigorous training. Of those interviewed, 50% (14/28) confirmed significant concerns regarding image quality and equipment that was already outdated and not the standard of care (even for the 1980s). The surgeons were often unwilling or unable to operate on the preclinical mammography-only findings the CNBSS was designed to detect and treat at an early stage. Because many women with symptoms of breast cancer were included, the CNBSS trials were not a true evaluation of screening for preclinical cancers in the general population. These problems and compelling evidence of violation of the randomization protocol (19) caused significantly more advanced cancers to be included in the screening MP arm of the trials and account for the mortality rate of 1.36 in the screening MP arm of CNBSS-1. This new information regarding the day-to-day operations of the trial confirms previously published suspicions that patient selection, randomization, intervention, and surgical treatment deviated so far from the standards of rigorous RCT design and conduct to explain at outlier results. These results cast serious concerns about the value of these trials for informing policies regarding breast cancer screening.

Acknowledgments

The authors would like to thank Jennie Dale for assistance in compiling the accounts.

Funding

None declared.

Conflict of Interest Statement

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Research collaboration between M.J.Y.’s institution, Sunnybrook Research Institute, and GE Healthcare on breast tomosynthesis and contrast-enhanced mammography. M.J.Y. has received consulting fees from BHR (pharma) and IGAN for unrelated activities. M.J.Y. holds shares in Volpara Health Technologies and is the co-principal of Mammographic Physics Inc., a company that provides consulting on image quality and radiation safety issues in breast cancer imaging. P.B.G. is a volunteer advisor to www.densebreast-info.org and www.densebreastscanada.ca, holds shares in Volpara Health Technologies, and is on the Medical Advisory Board of Besins Healthcare. J.M.S., P.R.E, and M.J.Y. serve on the Editorial Board of the Journal of Breast Imaging. As such, they were not involved in the editorial process. J.M.S., S.A., and P.R.E. do not have any conflicts of interest to declare.