| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Breast Imaging |
1 From the Department of Radiology, Catharina Hospital, Michelangelolaan 2, 5623 EJ Eindhoven, the Netherlands (L.E.M.D.); Department of Public Health, Erasmus MC-University Medical Center, Rotterdam, the Netherlands (J.H.G., H.J.d.K.); and Department of Radiology, University Medical Center Nijmegen, the Netherlands (J.H.C.L.H.). Received April 27, 2003; revision requested July 8; final revision received September 26; accepted October 21. Address correspondence to L.E.M.D. (e-mail: LEMDuijm@hotmail.com).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: The study population consisted of women who participated in the Dutch Nationwide Breast Cancer Screening Program, in which biennial screening is offered to women aged 50–75 years. An arbitration panel of three radiologists assessed those screening mammograms for which two screening radiologists did not reach a consensus about referral necessity. Women were referred for further analysis if at least one arbitration panel radiologist considered referral to be necessary.
RESULTS: The two screening radiologists agreed on the recommendation for referral of 498 (0.8%) of 65,779 screened women and on the recommendation for no referral of 64,949 (98.7%) women. They initially disagreed about the referral in 332 (0.5%) cases. After a mutual consultation, disagreement persisted regarding 183 (0.3%) mammograms. The arbitration panel referred 89 of these cases for further analysis, which revealed cancer in 20 (22%) cases. In three (3%) of the 94 cases that were not referred by the panel, breast cancer was detected at the site of previously discrepant mammographic findings seen at subsequent screening performed 2 years later. If all 183 discrepant cases had been referred, the referral rate would have increased from 0.8% to 0.9% at subsequent (incident) screenings and from 1.5% to 1.7% at initial screenings. In addition, at subsequent screenings, the number of cancers detected per 1,000 women screened would have increased from 4.4 to 4.5.
CONCLUSION: Mammograms with discrepant findings constitute a very important subset of screening mammograms. All lesions that are subsequently proved to be malignant may not be detected with panel arbitration.
© RSNA, 2004
Index terms: Breast neoplasms, 00.31, 00.32 • Breast radiography, 00.11 • Breast radiography, quality assurance • Cancer screening • Diagnostic radiology, observer performance
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
In the Dutch Nationwide Breast Cancer Screening Program, screening mammograms are read independently by two radiologists, who must reach a consensus as to whether thewoman should be referred for further examination. The readers usually reach a consensus; however, the percentage of screening examinations at which there is persistent reader disagreement regarding referral necessity and how such cases should be dealt with are still not known. Thus, the purpose of our study was to prospectively determine the value of arbitration by a panel of radiologists when two radiologists performing independent reading of screening mammograms do not reach a consensus regarding the referral recommendation.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The study was approved by the national review board (ie, Centrale Commissie voor Medisch-wetenschappelijk Onderzoek). Therefore, additional approval from our local institutional review board (ie, Medisch Ethische Toetsings Commissie, Catharina Ziekenhuis Eindhoven) was not required. Women participating in the nationwide Dutch screening program are asked to give written informed consent for use of their data for scientific purposes. All women included in our study had given this informed consent.
Details of the Dutch Nationwide Breast Cancer Screening Program, which currently offers biennial screening mammography to women aged 50–75 years, are described elsewhere (5). In brief, at initial screening (ie, the first time a woman undergoes mammographic screening in the breast cancer screening program), two-view mammography involving the acquisition of mediolateral-oblique and craniocaudal views of each breast is performed. At subsequent screenings, a one-view mammogram is usually obtained. Two-view mammography is performed at about 20%–30% of subsequent screening examinations. The indications for two-view mammography include any changes in mammographic features, such as new or increased number of microcalcifications or areas of density; a complicated judgment regarding interpretation due to breast surgery or dense fibroglandular tissue; and/or a longer than 2-year interval since the previous screening examination.
All mammographic examinations are performed by specialized screening mammography technicians. The mammograms are obtained by using a commercially available unit (at fixed center: Performa, Oldelft, Tuusula, Finland; at mobile center: Alfa RT, Oldelft) with a focal spot size of 0.3 mm. Dedicated mammography screens (at fixed center: Mamoray MR-R, Agfa, Schrobenhausen, Germany; at mobile center: Mamoray MR-S, Agfa) and film (at both centers: Mamoray HDR; Agfa, Mortsel, Belgium), as well as extended-cycle dedicated processing, are used.
Image Interpretation and Case Referral
At the time of a subsequent screening, the previously obtained screening mammograms are always available for comparison with those currently being obtained. To be certified as a screening radiologist, a radiologist must attend an intensive 3-week mammography course at the National Expert and Training Centre for Breast Cancer Screening in Nijmegen, the Netherlands. Every 3 years, screening radiologists at each screening center get feedback on their screening results from a team at that center. All mammograms are read independently by two radiologists, who are required to reach a consensus regarding referral of the woman for further examination. A woman will not be referred if her mammographic findings are normal (ie, no apparent abnormalities) or benign (eg, calcified fibroadenoma, lipoma, lymph nodes, and vascular calcifications) or if her mammographic examination reveals a nonspecific minimal sign (eg, vague area of density with an incomplete sharp border and a diameter between 5 and 30 mm [density comparable to that of glandular tissue], fewer than six clustered nonspecific microcalcifications, or subtle architectural distortions that include asymmetric glandular tissue) (6). In such cases, the woman is invited to undergo a subsequent screening examination after 2 years.
In cases of a suspicious lesion—that is, a mass with irregular or non–well-defined borders—or a malignant lesion—that is, a spiculated mass or microcalcifications of the ductal type—the woman is referred by her general practitioner to the surgical oncologist of a hospital in her region. After the surgeon performs a physical examination of the woman, a two-view mammogram (craniocaudal and mediolateral-oblique views) of both of her breasts is obtained, and additional local-compression or magnification mammograms are obtained if the radiologist considers these views to be necessary. These additional local-compression or magnification mammographic views are obtained according to the radiologist’s recommendation, regardless of whether there was initial consensus between the two screening radiologists regarding the referral.
Depending on the findings at physical examination and mammography, further evaluation may include breast ultrasonography (US), cytologic or histologic biopsy (either US- or stereotactically guided), or excisional biopsy. The radiologist who assesses these mammographic results makes the decision of whether US and/or biopsy should be performed. After having been referred for further evaluation, all patients with benign breast imaging or biopsy results undergo routine follow-up mammography at 6 months and 1 year to exclude malignancy.
In the Bevolkings Onderzoek Borstkanker Zuid screening region, the two readers interpret the mammograms at the same center. In the cases of discrepant referral recommendations, the second reader takes the mammograms from the central unit of the screening center to the hospital. The second reader classifies the mammographic finding of a discrepant reading into one of the following categories: category 1, which includes areas of vague density; category 2, which includes microcalcifications; category 3, which includes breast parenchyma asymmetries; and category 4, which includes subtle architectural distortions. The two screening radiologists then try to reach consensus. Only when consensus about referral necessity is not reached between these two radiologists is the screening mammogram subjected to arbitration by a panel of three other randomly selected screening radiologists. The three panel radiologists are informed that the two radiologists disagreed about the necessity for referral of the case. Each panel radiologist is blinded to the reading results of the other arbitration panelists. Women are referred for further analysis when at least one of the three arbitration panel radiologists recommends referral.
Screening Follow-up
The follow-up period for screened women includes the time through the next screening round, with a screening interval of about 2 years. We collected data on any diagnostic procedures, breast cancer diagnoses, histopathologic analyses, and TNM classifications. Most screening-detected and interval breast cancers were identified by linking the screening records to the regional cancer registries (5,7). Interval cancers are breast cancers that are diagnosed either after the screening examination yields negative results—with no recommendation for referral in these cases—or after the screening examination yields positive results but the evaluation does not lead to a diagnosis of breast cancer and before an eventually performed screening examination that yields positive results and leads to a diagnosis of cancer.
To trace as many interval cancers as possible, we used all available information in addition to the usual follow-up data, including (a) all radiation therapy reports from the regional radiation therapy institute on women who underwent radiation treatment for breast malignancy and had participated in the screening program; (b) inquiries about pathology specimens at the various regional pathology laboratories, some months after any hospital had requested the screening mammograms of a participant who had not been referred for further analysis; (c) the pathology records that were obtained if a woman cancelled an appointment for subsequent screening because breast cancer had been diagnosed after a previously performed screening examination yielded negative results; and (d) the occasional reports on interval cancers from general practitioners or medical specialists to the screening center.
Finally, to verify the absence of breast malignancy, one of the authors (L.E.M.D.) made telephone calls to all women who had not undergone subsequent screening 2 years after the screening examination that yielded discrepant findings. These inquiries were also directed at the women who had not undergone a subsequent screening examination after a false-positive mammogram was obtained. A false-positive screening mammogram was defined as a mammogram with findings that led to a referral for further assessment that did not result in a diagnosis of cancer. Lobular carcinomata in situ were considered to be benign lesions. Follow-up of at least 2 years was conducted for all women with discrepant mammographic readings.
Reader Experience
In our study, eight radiologists (including L.E.M.D.) were responsible for reading the screening mammograms. Each of these radiologists, including the three arbitration panelists, evaluates about 3,000–3,500 screening mammograms yearly. This number includes screening mammograms that have been read before by another screener according to our double-reading system. At the start of the study, the radiologists’ experience in reading screening mammograms varied from 15 to 36 months (mean, 31 months). In addition to performing screening activities, all radiologists each year evaluate at least 550 diagnostic mammograms, which are obtained on an outpatient basis in a teaching hospital and a diagnostic center dedicated to general practice. All screening radiologists frequently attend accredited breast radiology courses. Every 3 months, the status of all cancers recently detected at screening, as well as the status of all traced interval cancers, is evaluated.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
|
|
Biopsy was performed in 43 (48%) of the 89 women referred by the panel and revealed breast cancer in 20 (46%) cases. These 20 malignancies were seven cases of ductal carcinoma in situ and 13 invasive cancers. Four of the 13 invasive cancers were smaller than 10 mm in diameter (stages T1a and T1b). Three women were proved to have lymph node metastases. Of the 69 women who did not receive a diagnosis of breast cancer, 21 (30%) preferred to undergo outpatient mammographic surveillance rather than continue participating in the breast cancer screening program.
In three (3%) of the 94 women who were not referred by the arbitration panel (63 women at subsequent screening, 31 women at initial screening), breast cancer at the site of previously discrepant mammographic findings was detected at subsequent screening mammography performed 2 years later. Pathologic examination revealed one ductal carcinoma in situ and two invasive ductal cancers (17 and 21 mm in diameter, both with axillary lymph node metastases).
Overall Findings
A total of 619 of the 65,779 screened women, or approximately 9.4 women per 1,000 screened, were referred for further evaluation. Supplementary breast imaging techniques—specifically, local-compression or magnification mammography and/or breast US—were performed without biopsy in 189 (30.5%) of the 619 referred women, or 6.5 women per 1,000 screened. The remaining 430 (69.5%) women underwent additional biopsy, which revealed breast cancer in 320 (74.4%) of these patients, for a resulting overall biopsy detection rate of approximately 4.9 cases per 1,000 women screened.
In addition to the 320 screening-detected cancers, 100 interval cancers were observed. The calculated breast cancer risk was 588 cases per 1,000 women screened for the women with whom there was initial agreement regarding referral, and 1.5 cases per 1,000 women screened for the women with whom there was initial agreement regarding nonreferral. The calculated breast cancer risk for the women with discrepant readings was 93 cases per 1,000 women screened.
If radiologists in the screening program had used a strategy in which none of the 183 cases for which there was persistent disagreement between the two readers regarding referral had been referred for further evaluation (strategy 1), the cancer detection rates would have been 5%–7% lower (Table). If the radiologists had used a strategy in which they referred all 183 cases (strategy 2) or one in which they referred all 332 initially discrepant cases (strategy 3), the cancer detection rates at subsequent screening examinations would have been, respectively, 1.4% or 1.8% higher. Strategy 3 would have resulted in a 74.2% increase in the number of false-positive cases compared with the observed number of false-positive cases.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
In our study, 9.4 women per 1,000 screened were referred for further assessment. This result is consistent with national screening results: Of all women screened in the Dutch Nationwide Breast Cancer Screening Program between 1990 and 1999, 1% were referred for further assessment (8,9). We found that initial disagreement regarding referral of patients occurred with only a small proportion of screening examinations. If all of the women in these cases had been referred for diagnostic assessment without the readers having reached a consensus, the referral rate would have been almost doubled. This doubled referral rate, however, is still very low compared with the referral rates in many other screening programs: In some studies (10–12), the referral of 5%–12% of screened women for diagnostic examination has been reported.
Yankaskas et al (13) calculated that practices with recall rates of between 4.9% and 5.5% achieved the best trade-offs between sensitivity and positive predictive value. This does not mean, however, that every increase in referral rate will lead to better results. Therefore, the Dutch program seeks to find an optimal balance between the number of women who are referred for further examination but ultimately do not receive a diagnosis of breast cancer (ie, false-positive cases) and the number of cancers detected (ie, true-positive cases). In our series, the biopsy rate was 6.5 biopsies per 1,000 women screened, while Yankaskas et al (13) and Burnside et al (14) reported biopsy rates of, respectively, 10.5 and 12.3 biopsies per 1,000 women screened. In the Burnside et al (14) study, the recall rate was 4.0% and the cancer detection rate was 5.3 cases per 1,000 women screened; biopsy was performed in about one-third of the recalled cases and revealed breast cancer in 43.0% of them. In our study, there was a comparable detection rate (4.9 cases per 1,000 women screened) but a substantially smaller number of biopsies performed per 1,000 screened women.
In the Dutch screening program, it is the standard that all mammograms are double read, and women are referred only after the two readers reach a consensus. There are no guidelines, however, regarding how to reach the consensus or the protocol to follow when no consensus can be reached. To avoid a situation in which one of the initial readers is overruled by the other, we introduced the use of an arbitration panel of three radiologists. The institution of such a panel was feasible in our screening setting because screening mammography is usually performed by a team of five to 10 radiologists at a central location. However, we understand that in many other screening practices, this tight infrastructure might not exist, and, thus, instituting such a panel could be inconvenient, expensive, or even impossible.
In our study, the arbitration panel facilitated a decrease in the number of referrals of discrepant cases, which under other circumstances would have been referred. This reduced number of referrals is consistent with other published results: Kopans (15) observed that with a double-reading system, a third arbiter could facilitate a reduced number of recalls. However, the interpretation of discrepant mammograms by an arbitration panel did not result in an accurate diagnosis for all of the women with suspicious lesions in our study. Breast cancer was diagnosed in one of every four women who were referred by the arbitration panel, and three additional cancers were detected at the next screening round in the 94 women who had not been referred. If the latter group had been referred as well, the cancer detection rate would have increased slightly (particularly at subsequent screenings) but at the expense of an increased number of false-positive cases.
The positive predictive values of referral would have decreased still further if all women with mammographic findings on which the two readers initially disagreed had been referred. In this scenario, for every additional cancer detected, nine additional women would have been referred without cancer having been diagnosed. In the Netherlands, the cost of diagnostic assessments, including diagnostic mammography and magnetic resonance imaging, is about $700; the prevention of late-stage disease would save about $20,000 in costs for palliative breast cancer treatment. Thus, the referral of all women in cases of initial reader disagreement may still be acceptable in terms of cost effectiveness. Therefore, we advise the further examination of women with mammograms for which two readers do not reach a consensus.
There are conflicting reports about the effects of false-positive mammograms on subsequent screening attendance. In a British study, McCann et al (16) found that women who had false-positive mammography results at first screening were less likely to attend subsequent screenings than were women with true-negative (ie, no mammographic abnormalities) screening mammogram results. However, in two North American studies—those of Burman et al (17) and Pinckney et al (18)—the opposite was found: Women with false-positive mammograms, especially those who had not undergone mammography previously, were more likely to return for the next scheduled screening. In our study, nearly a third of the women in whom further diagnostic assessment did not reveal breast cancer preferred to be followed up with outpatient mammographic surveillance, although the Dutch screening program guidelines recommend continued participation in the breast cancer screening program in such cases. To our knowledge, no data on the reattendance rate in the Dutch Nationwide Breast Cancer Screening Program are available.
The detection of breast cancer depends on many other characteristics of the screening program, which vary between and even within countries (19). In the Dutch screening program, double reading is a standard. In a recent study in the United States, Beam et al (20) reported that double reading was significantly associated with better detection accuracy. Thus, screening programs that involve single reading may benefit from the adoption of a double-reading system, although an increased number of false-positive recalls may be a concomitant result (21).
The number of mammographic views obtained may be another important determinant in breast cancer detection. In the Dutch Nationwide Breast Cancer Screening Program, two-view mammography is performed at initial screening, whereas subsequent screenings usually involve one-view mammography. Specialized radiographers have been trained to compare the mediolateral-oblique views with those obtained at the previous screening and to immediately obtain additional views if any changes are seen. In the Netherlands, craniocaudal views are obtained at about 25% of subsequent screenings. Radiographers in the United Kingdom National Health Service Breast Cancer Screening Programme began performing two-view mammography at all screening examinations at the end of March 2003. The results of studies from the United Kingdom National Health Service Breast Cancer Screening Programme (22,23) indicate that the detection and interpretation of cancers have been improved owing to an increased number of mammographic views, although a policy of performing two-view mammography may be efficient only if double reading is performed in the program.
The availability of previous mammograms also may influence referral and detection rates. Frankel et al (24) observed that substantially fewer screening findings were interpreted as abnormal when the prior images were available. This finding was confirmed by Burnside et al (14), who also reported that comparing previous and subsequent screening mammograms facilitated the detection of cancers at an earlier stage.
The screening interval is yet another important factor. Screening intervals vary from 1 to 3 years (18). In the Netherlands, biennial screening seems to be very cost effective (25). In the United States, the American Cancer Society, the American Medical Association, and the American College of Radiology recommend annual screening. Performing mammographic screening biennially in randomized trials, Nystrom et al (26) observed a significant reduction (21%) in breast cancer mortality in association with participation in screening. There are likely to be greater mortality reductions with shorter screening intervals. Hunt et al (27) observed that compared with biennial screening, annual screening resulted in the detection of smaller tumors that had a more favorable prognosis. In our study, the higher rate of findings positive for lymph node metastasis and the lower percentage of invasive cancers smaller than 10 mm in diameter at subsequent screenings may have been attributed to the biennial screening interval (24).
We acknowledge that our study had several other limitations in addition to the biennial screening interval. In the Dutch Nationwide Breast Cancer Screening Program, women with nonspecific minimal signs at mammography are not referred for further diagnostic assessment. In the Maes et al study (6), nonspecific minimal signs were seen on 11% of the screening mammograms. In their study, although the additional risk of cancer in women with these signs was 0.5% compared with the 0.37% cancer risk in all screened women, the favorable stage of invasive breast cancers in women with previously detected nonspecific minimal signs warrants inviting these patients to the next screening. Because of this policy, the referral rate in the Dutch screening program is much lower than the referral rates in other programs. However, a critical reconsideration of nonspecific minimal signs may lead to the detection of subgroups of subtle mammographic abnormalities that warrant further assessment.
In general, the screening results observed in this study are consistent with those observed in the Dutch Nationwide Breast Cancer Screening Program (5). However, the frequency of lymph node–positive breast cancers in our study was higher (28.6% vs 24.0% at subsequent screening, 35.5% vs 28.0% at initial screening) despite comparable numbers of ductal cancers in situ, invasive cancers, and invasive cancers smaller than 10 mm in diameter. The higher frequency of lymph nodes positive for cancer was probably due to the recent popularity of sentinel node biopsy and the increased use of immunohistochemistry. Although the frequency of cancer-positive nodes can still be used as an indicator of the success of a screening program, it should be interpreted carefully, taking into account whether immunohistochemistry was used in determining the value and how cases of nodal micrometastases were classified (28).
We conclude that mammograms with discrepant readings are an important subset of screening mammograms, particularly for women with a higher risk for breast cancer. Arbitration by additional radiologists in cases in which two readers cannot reach a consensus regarding referral does not seem useful. Therefore, women should be referred for further diagnostic assessment whenever two independent readers do not reach a consensus.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. M. Shaw, F. L. Flanagan, H. M. Fenlon, and M. M. McNicholas Consensus Review of Discordant Findings Maximizes Cancer Detection Rate in Double-Reader Screening Mammography: Irish National Breast Screening Program Experience Radiology, February 1, 2009; 250(2): 354 - 362. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. J. Gilbert, S. M. Astley, M. G.C. Gillan, O. F. Agbaje, M. G. Wallis, J. James, C. R.M. Boggis, S. W. Duffy, and the CADET II Group Single Reading with Computer-Aided Detection for Screening Mammography N. Engl. J. Med., October 16, 2008; 359(16): 1675 - 1684. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Taplin, L. Abraham, W. E. Barlow, J. J. Fenton, E. A. Berns, P. A. Carney, G. R. Cutter, E. A. Sickles, D. Carl, and J. G. Elmore Mammography Facility Characteristics Associated With Interpretive Accuracy of Screening Mammography J Natl Cancer Inst, June 18, 2008; 100(12): 876 - 887. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Skaane, S. Hofvind, and A. Skjennald Randomized Trial of Screen-Film versus Full-Field Digital Mammography with Soft-Copy Reading in Population-based Screening Program: Follow-up and Final Results of Oslo II Study Radiology, September 1, 2007; 244(3): 708 - 717. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. E. M. Duijm, J. H. Groenewoud, J. Fracheboud, and H. J. de Koning Additional Double Reading of Screening Mammograms by Radiologic Technologists: Impact on Screening Performance Parameters J Natl Cancer Inst, August 1, 2007; 99(15): 1162 - 1170. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Skaane, A. Kshirsagar, S. Stapleton, K. Young, and R. A. Castellino Effect of Computer-Aided Detection on Independent Double Reading of Paired Screen-Film and Full-Field Digital Screening Mammograms Am. J. Roentgenol., February 1, 2007; 188(2): 377 - 384. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
