Trial Protocol and Statistical Analysis Plan
eTable 1. Sensitivity Analysis for Breast Thickness by Considering Center as Random Effect
eTable 2. Description of Intervention of the Radiographer for the 275 Patients in the Self-compression Arm
eTable 3. Questionnaire About Self-compression Technique Fulfilled by 22 Radiographers out of the 25 Participating Radiographers
Data Sharing Statement
Customize your JAMA Network experience by selecting one or more topics from the list below.
Henrot P, Boisserie-Lacroix M, Boute V, et al. Self-compression Technique vs Standard Compression in Mammography: A Randomized Clinical Trial. JAMA Intern Med. 2019;179(3):407–414. doi:10.1001/jamainternmed.2018.7169
Could self-compression of the breast be used by women during mammography?
In this randomized clinical trial that included 549 women in France, self-compression mammography was found to be noninferior to the standard compression approach in achieving minimal breast thickness. The compression force was higher in the self-compression group compared with the standard compression group without increasing pain or compromising image quality.
Self-compression may be an effective mammography technique for women who want to take an active role in their own breast examination.
Many women dread undergoing mammography, and some may not attend or reattend breast cancer screening because of the discomfort or pain induced by breast compression.
To evaluate the noninferiority of the self-compression mammography technique for reducing breast thickness compared with standard compression.
Design, Setting, and Participants
This prospective, parallel-group, noninferiority randomized clinical trial was conducted from May 7, 2013, to October 26, 2015, at 6 cancer care centers in France. Participants were women aged 50 to 75 years, without a history of recent breast surgical procedure or treatment, and who could perform self-compression. Analyses were performed on intention-to-treat basis from January 27, 2017, to March 30, 2018.
Patients were randomized 1:1 to the self-compression group or the standard compression group.
Main Outcomes and Measures
Primary end point was breast thickness expressed as the mean of 4 views: right and left craniocaudal and right and left mediolateral oblique. The predefined noninferiority margin was a difference of 3 mm, with a 1-sided 95% CI. Secondary end points included compression force, image quality, requirement for additional views, pain, and patient satisfaction and radiographer assessment questionnaires.
Among the 549 women randomized, 548 (97.3%) completed the trial. Of these, 275 (48.8%) (mean [SD] age, 61.35 [6.34] years) were randomized to the self-compression arm and 273 (48.5%) (mean [SD] age, 60.84 [6.41] years) to the standard compression arm. The difference in the mean thickness between the 2 arms was lower than the noninferiority margin, with an upper 1-sided 95% CI less than 3 mm (–0.17; 95% CI,−∞ to 1.89 mm; P < .05). Compression force was higher in the self-compression group compared with the standard compression arm for the 4 mammographic views. Pain was statistically significantly lower in the self-compression group (n = 274) compared with the standard compression group (n = 269) (median [interquartile range (IQR)] score, 2 [1-5] vs 3 [1-5]; P = .009). No difference was reported in the image quality scores of the 2 groups or in the number of additional views performed (median [IQR] extra views, 2 [2-2] vs 2 [2-3] extra views; P = .64), whatever the indication, including insufficient image quality (29 [16.8%] vs 27 [15.0%] insufficient quality views; P = .65). No adverse effects or pain were reported by the participants after the self-compression mammography.
Conclusions and Relevance
Self-compression does not appear to be inferior to standard compression mammography in achieving minimal breast thickness without increasing pain or compromising image quality; this technique may be an effective option for women who want to be involved in their breast examination.
ClinicalTrials.gov identifier: NCT02866591
Mammography is currently the main imaging modality for breast cancer screening, diagnosis of a breast abnormality, and follow-up. Despite its utility, many women dread undergoing mammography because this examination can be uncomfortable or painful.1-6 This apprehension has been observed in the field of screening; pain or discomfort experienced by women has been reported as a common factor in nonattendance or non-reattendance of screening.7-12 In a review of the literature, Whelehan et al13 found that pain was reported by women as the cause of non-reattendance in 25% to 46% of cases.
Pain results from the positioning requirements of the breast to optimize imaging and make the mammogram reproducible. Compression is one of these requirements, with no other objective than to reduce breast thickness to its minimal achievable value. This compression has beneficial effects on image quality, such as decreased scattered radiation and therefore improved contrast and detection of lesions, shorter acquisition time and thus less motion artifacts and lower radiation dose, better facilitation of breast immobilization, and reduced unwanted magnification.14-16
Despite compression being a key factor in image quality, no guidelines or standards exist, to date, that describe the optimal breast compression technique. The Mammography Quality Standards Act indicates that “compression shall be between 111 and 200 N.”17(p20) European guidelines indicate that compression “should be firm but tolerable,”18
Several methods, other than simply a patient request, have been proposed to limit the discomfort or pain of compression. One such method is a mechanical approach, using devices that analyze the ratio of force variation over thickness variation and are programmed to stop the compression when a predefined threshold is reached. Nevertheless, this method does not take into account the nonhomogeneity of the breast, and the compression could mainly apply to the pectoral muscle in the mediolateral oblique view, without compressing the whole breast evenly.19
A more recent approach is the analysis of the pressure, namely, the ratio of compression force over contact area. This approach could provide uniformity in the amount of compression according to the individual breast morphologic structure and plasticity.20 Improving the design of compression paddles (for example, using flexible devices) could also decrease discomfort. However, a preliminary study indicates that a flexible compression paddle could move breast tissue from the detector area at the chest wall side.21
An alternative method is to give women control of the compression of their breasts. This technique was proposed by Kornguth et al22 in a study in which 1 breast was compressed by the radiographer and the other by the woman. The women reported a less painful experience and a greater overall satisfaction with self-compression. No loss of image quality was reported.
To assess the technique in conditions similar to routine mammographic screening or follow-up, we designed a multicenter randomized clinical trial. Our goal was to evaluate the noninferiority of the self-compression technique for reducing breast thickness compared with standard compression.
This multicenter, prospective, parallel-group, noninferiority randomized clinical trial (Interest of Self-compression Technique on Tolerance of Mammography) was conducted from May 7, 2013, to October 26, 2015, at 6 cancer care centers in France. The trial protocol was approved by the institutional review boards of each participating center (Institut de Cancérologie de Lorraine Alexis Vautrin, Nancy, France; Institut Bergonié, Bordeaux, France; Centre François Baclesse, Caen, France; Centre de Radiologie RX125, Nancy, France; Institut Curie, Paris, France; and Centre d’Imagerie Majorelle. Nancy, France). All patients provided written informed consent before the randomization. A data and safety monitoring board oversaw the trial to completion. The trial protocol is included in Supplement 1.
Patients from the 6 centers (4 hospitals that perform screening, follow-up, and treatment and 2 breast-imaging clinics) coming for a mammography were evaluated according to the inclusion criteria: 50 to 75 years of age, with a World Health Organization Performance Status score lower than 2, and who called for a screening or follow-up after treatment for a breast lesion after a predefined time lapse. A performance status score lower than 2 was retained to ensure the woman’s ability to stand up and perform a self-compression pretest.
The exclusion criteria were as follows: history of breast operation for benign lesions or surgical treatment and/or radiation therapy for breast cancer within the past 3 years, breast macrobiopsy (with >14-gauge needle) within the previous year, breast implants, previous mastectomy, and assessment motivated by a clinical breast abnormality. The time interval between surgical procedure or macrobiopsy and mammography was selected to limit the influence of invasive procedures in the assessment of compression pain.
The workflow constraints and the multicenter design of the trial resulted in limiting the inclusion rate to 1 eligible woman per session with the agreement of the radiologist at the center. After screening, eligible participants were verbally informed by the radiographers at each center about the aim of the study to assess their interest in self-compression during mammography.
Participants were invited to perform a self-compression pretest. For this test, we explained the use of the compression device and verified the ability of the participant to manipulate the foot and hand controllers until the breast thickness achieved a minimal value. The instruction we provided was, “When you find that your breast is sufficiently immobilized, you will stop pressing the foot controller.” The participant’s inability to manipulate the foot controller, despite the explanation, was considered a pretest failure. We informed the participants who successfully completed the pretest of the study details and the rules of biomedical trials. Written informed consent was obtained and recorded.
Participants were randomly assigned at a 1:1 ratio to either the self-compression group or the standard compression group (Figure). Randomization was centralized by computer-generated random numbers in blocks of 4, with stratification according to the center. This study was not blinded owing to the practical barriers to masking. All of the information required by the protocol was supplied in an Electronic Case Report Form.
In the standard compression arm, the radiographer controlled the positioning and compression as usual. In the self-compression arm, after standard positioning, the radiographer set the compression to the amount of 40 N and then gave the participant control of the ongoing compression process.
Mammographic views were obtained in the same way for each group: (1) right craniocaudal, (2) left craniocaudal, (3) right mediolateral oblique, and (4) left mediolateral oblique. After performing the 4 views, the participant was asked to rate her pain level, which was then recorded, and was given a satisfaction questionnaire to fill out.
Without knowledge of which technique was used, the radiologist first rated the image quality of the 4 views in terms of motion artifacts. Then, the radiologist asked the participant for additional views or the re-performance of poor-quality views, owing to positioning, contrast, or noise, as is the routine practice in mammography. Any additional radiologic investigations required were conducted afterward outside of the study. The women were invited to communicate with us by telephone all of the adverse effects they experienced after their participation in the trial. At the end of the trial, radiographers were invited to respond to a questionnaire.
The primary outcome was the breast thickness in millimeters, measured by the radiographer using the mammograph control panel. The secondary outcomes were compression force in N, measured by the radiographer using the control panel; image-quality score; overall pain; number of and indications for additional views; patient satisfaction scores; number and type of radiographer interventions; and a radiographer assessment questionnaire.
The image-quality score for each view was rated from 1 to 4 by the radiologist on an interpretation screen. Score 1 indicated absence of blurring; score 2, minor blurring visible after magnification; score 3, blurring visible without magnification that could be responsible for loss of information; and score 4, major blurring that compromised interpretation.
Participants rated their overall pain with a 10-point visual analog scale (VAS) (score range: 0 [no pain] to 10 [maximum pain]). Using the nongraduated side of the VAS, the women placed the cursor between the left limit, corresponding to an absence of discomfort, and the right limit, corresponding to the maximum thinkable pain. The radiographer read the values of the graduated side of the VAS and rounded the measurement to the nearest integer.
Levels of patient satisfaction were measured using a published Mammography Questionnaire.23 Items pertaining to participant attitudes (such as knowledge about the procedure, nervousness about mammography, expected pain, and trust in the diagnostic value of mammography) were rated on a 5-point Likert scale (5, strongly agree; 4, agree; 3, neutral; 2, disagree; 1, strongly disagree). Six psychometric scales comprising 14 items were also completed that evaluated physical pain, psychological distress, staff punctuality and technical skills, information provided, and physical surroundings.24 Response scores in each scale were transformed with a potential range of 0 to 100, in which 100 represents the best possible experience.24
The number and types of and the indications for additional views were recorded by the radiographer. In the self-compression group, the radiographer recorded any interventions required during the compression process.
The radiographers from the 6 centers were asked to complete a questionnaire about the self-compression technique at the end of the trial. The questionnaire included 6 questions: (1) Is self-compression difficult to explain? (2) Is your experience with the patient improved? (3) Is the examination more complex? (4) Is the examination time substantially increased? (5) Is it difficult for the patient to perform self-compression? (6) Do you think self-compression could be routinely used? Four responses were available: not at all, a little, most of the time, almost always.
This noninferiority trial tested the null hypothesis that the breast thickness is at least 3 mm higher in the self-compression arm than in the standard compression arm, with a 1-sided 95% CI. The true difference between the 2 arms was assumed to be 0. A margin of 3 mm was chosen on the basis of the study by Chida et al.25 In their work, the authors demonstrated that reducing the compression force from a standard of 120 N to 90 N corresponded to a mean increase in the breast thickness of 3 mm without compromising image quality. The number of participants in each group was 275, with an alpha-level risk of 5% and a power of 90%.
Analyses were performed on the intention-to-treat sample (all randomized participants) and those who completed the compression intervention as initially allocated. Both analyses yielded similar results; thus, only the intention-to-treat results are reported. No imputation for missing values was performed.
Continuous data are presented as means and 2-sided 95% CI or as medians and interquartile ranges (IQRs) according to the Kolmogorov-Smirnov test. Categorical data are presented as counts and percentages.
The reproducibility of breast thickness between the 4 mammographic views was investigated using the intraclass correlation coefficient according to the Fleiss method.26 A value greater than 0.8 was considered to be a good agreement. Then, for each patient, the mean of the breast thickness measurements from the 4 views was computed. The primary outcome analysis evaluated noninferiority by calculating the 1-sided 95% CI for the difference in breast thickness (ie, standard compression minus self-compression). If the upper bound of the 1-sided 95% CI for this difference was less than the inferiority margin (ie, >3 mm), inferiority could be rejected.
Two-sided superiority analyses were conducted on an intention-to-treat basis for the breast thickness of each view and for all secondary outcomes. For continuous data, unpaired, 2-tailed t tests or Mann-Whitney tests were used; for categorical data, 2-sided χ2 tests or Fisher exact tests were used when any expected counts are fewer than 5. A sensitivity analysis including a random center effect was performed with a linear mixed model for breast thickness.
The Statistical Analysis Plan is included in Supplement 1. Statistical analyses were performed with SAS, version 9.4 (SAS Institute Inc). Analyses were performed from January 27, 2017, to March 30, 2018.
Inclusion into the study was effective from May 7, 2013, to October 26, 2015. Of the 563 participants who were assessed for eligibility and agreed to perform the pretest, 11 (1.9%) did not pass the pretest, 1 (0.2%) did not meet the inclusion criteria (breast implant was discovered after the pretest), and 2 (0.4%) declined participation after a successful pretest. Randomization was obtained in 549 women (97.5%), and 1 woman (0.2%) withdrew from the trial after randomization. Data were available in the intention-to-treat analysis for the 548 women (97.3%) who completed the trial, of whom 275 (48.8%) (mean [SD] age, 61.35 [6.34] years) were randomized to the self-compression arm and 273 (48.5%) (mean [SD] age, 60.84 [6.41] years) to the standard compression arm (Figure). Table 1 details demographic and clinical characteristics.
The reproducibility of the breast thickness for the 4 views was excellent, with an intraclass correlation coefficient of 0.917 (95% CI, 0.902-0.929). Table 2 summarizes the breast thickness results. The results are consistent with the noninferiority of self-compression for the primary outcome, with the upper bound of the 1-sided 95% CI for the difference in the mean breast thickness being less than the prespecified noninferiority margin of 3 mm (–0.17; 95% CI, −∞ to 1.89; P < .006). No statistical difference was found between the 2 groups for any of the 4 views. Similar results were found by taking into account the center effect (eTable 1 in Supplement 2).
Table 2 summarizes the compression force results. Compression force was statistically significantly higher in the self-compression group for the 4 views. Table 3 summarizes the results of rated pain. Pain was statistically significantly lower in the self-compression group (n = 274) compared with the standard compression group (n = 269) (median [IQR] score, 2 [1-5] vs 3 [1-5]; P = .009). No differences were observed between the standard compression and self-compression groups in the Mammography Questionnaire response scores in any of the 6 scales (Table 3).
Table 4 details the results of the image-quality scores. No differences were found concerning motion artifacts between the self-compression and standard compression groups in any of the 4 views.
The major indication for performing extra views was breast density and this resulted in the addition of lateromedial views. No differences between the self-compression group and standard compression group were found in the number of additional views performed (median [IQR] extra views, 2 [2-2] vs 2 [2-3]; P = .64), whatever the indication, including insufficient image quality (29 views [16.8%] vs 27 views [15.0%]; P = .65) (Table 4).
Radiographers recorded an intervention in 90 cases (32.7%) of the 275 women who performed the self-compression technique (eTable 2 in Supplement 2 summarizes the reasons for the interventions). Most interventions were encouragements (57 cases [20.7%]) or reexplanations about the manipulation of the foot and hand controllers (22 cases [8.0%]). No case necessitated the radiographer having to perform the breast compression.
Of the 25 radiographers from the 6 centers, 22 (88.0%) responded to a questionnaire (eTable 3 in Supplement 2). Self-compression was not considered to be a complex technique by the radiographers or participants, and some radiographers reported that the technique could improve interactions with patients. The main concern among 14 (63.3%) of 22 radiographers was the time-consuming feature of self-compression, and 12 (54.5%) responded that they were undecided on whether to routinely use the technique. No pain or adverse effects were reported by the participants after the mammography.
The aim of this study was to evaluate noninferiority in a randomized trial (and in this way, to offer an option to women who fear mammography-associated pain). We did not seek to evaluate the generalizability of the self-compression technique.
The findings indicate that the difference in the mean thickness between the standard compression and self-compression arms was lower than the prespecified margin of 3 mm, with an upper 1-sided 95% CI, which is consistent with the criterion for noninferiority. No difference was reported in the quality scores of the 2 groups, and a median number of additional views per woman were equivalent. The compression force was higher, and median pain was statistically significantly lower when the women themselves controlled the compression process.
This lesser pain observation is similar to others reported in the field of pain. Previous works have reported that when a painful stimulus was self-inflicted, it resulted in lower pain level and a higher ability to tolerate the pain, compared with when the painful stimulus was applied by another person.27 Within the mammography setting, the explanation for this phenomenon could be that the woman is aware that she can stop the compression at any time, resulting in a greater ability to tolerate the ongoing crushing phase.
Breast compression remains a major point for image quality, and its benefits are well known by radiologists and radiographers. However, carrying out breast compression often relies on empirical or subjective criteria. Detailed standards have not been established, and existing standards merely propose a range of values between 110 N and 200 N. Otherwise, they indicate methods of testing that are as vague as obtaining elastic resistance or blanching of the breast skin or stopping at the edge of pain.28 In addition, radiologists are not always aware of the compression values when interpreting mammograms.
Those limitations could explain the large variabilities in compression values observed between different sites and even within a site.29 In a study undertaken by Mercer et al30 in 3 sites, the mean percentage change between minimum and maximum compression force per women attending a screening 3 times was 27%, 55%, and 66% for the mediolateral oblique view, and 26%, 57%, and 60% for the craniocaudal view. Those variations can lead to painful experiences for some women and can result in the development of a negative opinion of mammography.11,12,30
Instead of a focus on what value to apply, practice could be reoriented toward the real purpose of breast compression: to reduce thickness to a minimal achievable value. Self-compression could be a way to refocus this process, from the radiographer’s point of view as well as the woman’s.
This study has several limitations. First, pain was significantly lower in the self-compression group. However, the pain difference from the standard compression group was less than the threshold reported in the literature to establish a substantial decrease in pain severity. Todd et al31 indicated that, for a range of VAS scores less than 34 mm, a mean (SD) threshold of 13 (11) mm was necessary to demonstrate a substantial change in pain severity. Pain induced by mammography could be insufficiently high to be rated using the VAS.
Most of the trial participants were patients at the 6 centers. Therefore, these women were not exclusively examined for breast cancer screening, and more than half reported a personal or familial history of breast cancer. This population was more used to mammographic follow-up and often accustomed to the examination process. Thus, they may be able to tolerate breast compression better than the screening population.
Second, we did not evaluate radiation dose, even though a difference in breast thickness is likely to have been associated with dose variation. This study was a multicenter, multimanufacturer trial; thus, attributing dose differences to either the thickness reduction or the mammography device would have been difficult. Comparisons would only have been possible in subgroups for each center, and the statistical power would have been compromised.
Third, the extra time associated with self-compression was not measured. Undoubtedly, the use of this technique increased the examination time, as reported by 14 of 22 radiographers. However, this technique was novel, not only for radiographers but also for the women. This increased examination time could be reduced after a learning period in screening or annual follow-up practice. Patient-assisted compression was evaluated in 100 women by Balleyguier et al.32 For 90 of 100 patients, the technologist needed less than 1 minute to explain self-compression, and in 80 of 100 cases the entire examination took between 12 and 15 minutes. The generalizability of self-compression and the recording of time taken could be further evaluated in a larger population.
Fourth, the Mammography Questionnaire measured patient satisfaction regarding the overall mammographic experience, but it did not allow for a specific evaluation of self-compression and in comparison to standard compression. Kornguth et al22 reported high levels of patient satisfaction on this specific point in their study, in which 1 breast was compressed by the radiographer and the other breast by the woman herself. In the study by Balleyguier et al,32 90 of 100 women found self-compression useful, and 74% reported that the technique made them more willing to return for their next mammography.
Self-compression could be complemented by other innovations to limit its discomfort, such as newly designed compression paddles; compression devices that analyze the pressure on the breast; and ergonomic tools controlled by the patient, such as a remote control for breast compression.33 In September 2017, the US Food and Drug Administration approved a mammography device with an option for patient-assisted compression.34
Self-compression does not appear to be inferior to standard compression in achieving minimal breast thickness. With this technique, a higher compression force is achieved without increasing pain or compromising image quality. Self-compression mammography may be an effective screening approach for women who want to take an active role in their breast examination.
Accepted for Publication: October 1, 2018.
Corresponding Author: Philippe Henrot, MD, Department of Radiology, Institut de Cancérologie de Lorraine Alexis Vautrin, 6 avenue de Bourgogne, Vandoeuvre-les-Nancy 54519, France (email@example.com).
Published Online: February 4, 2019. doi:10.1001/jamainternmed.2018.7169
Author Contributions: Dr Henrot and Ms Salleron had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Henrot, Gillon, Desandes.
Acquisition, analysis, or interpretation of data: Henrot, Boisserie-Lacroix, Boute, Troufléau, Boyer, Lesanne, Netter, Saadate, Tardivon, Grentzinger, Salleron, Oldrini.
Drafting of the manuscript: Henrot, Gillon, Desandes, Salleron.
Critical revision of the manuscript for important intellectual content: Henrot, Boisserie-Lacroix, Boute, Troufléau, Boyer, Lesanne, Netter, Saadate, Tardivon, Grentzinger, Salleron, Oldrini.
Statistical analysis: Henrot, Salleron.
Obtained funding: Henrot, Gillon, Desandes.
Administrative, technical, or material support: Henrot, Gillon, Oldrini.
Supervision: Henrot, Tardivon.
Conflict of Interest Disclosures: None reported.
Funding/Support: The Interest of Self-compression Technique on Tolerance of Mammography Trial was funded by the French Hospital Research Program PHRC 2012.
Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Meeting Presentation: The results of this study were presented at the 2017 Meeting of the Radiological Society of North America; November 29, 2017; Chicago, Illinois.
Additional Contributions: We acknowledge L. Fernandes, Department of Clinical Research and Biostatistics, Institut de Cancérologie de Lorraine Alexis Vautrin, for administrative, technical, and material support. We also acknowledge the radiographers from the Department of Radiology, Institut de Cancérologie de Lorraine Alexis Vautrin: El Oudghiri, Galus, Kinzelin, Marchand, Rouhard, and Sacré; the Department of Radiology, Institut Bergonié: Florczyk, Labarthe, Le rest, Muckensturm, and Ouillie; the Department of Radiology, Centre François Baclesse: Aime, Babin, and Bajout; the Centre de Radiologie: Buttner and Torzuoli; Department of Radiology, Institut Curie: Benitah, Casalle, David, Girard, and Lalaire; and Centre d’Imagerie Majorelle: Allach, Dine, and Servait. These contributors are paid staffers at these centers.
Data Sharing Statement: See Supplement 3.