| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Breast Imaging |
1 From the Department of Radiology, Breast Imaging Division, Duke University Medical Center, Hospital South, Rm 24254, Box 3808, Durham, NC 27710. From the 2000 RSNA scientific assembly. Received August 13, 2001; revision requested September 19; revision received October 15; accepted October 31. Address correspondence to E.L.R. (e-mail: rosen017@mc.duke.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: A search of the mammography database yielded 295 cases that were initially followed up with short-term interval mammography but eventually received a biopsy recommendation for the same breast. Of the 83 malignancies (81 patients) for which mammograms and pathology reports were available for review, 51 malignancies corresponded to the lesions for which short-term follow-up was recommended. Each case was retrospectively reviewed to determine whether the lesion followed up represented the subsequently diagnosed malignancy. Each lesion was characterized with appropriate Breast Imaging Reporting and Data System descriptors, based on the mammographic imaging available when short-term follow-up was first recommended. These characteristics were then used to determine if, in retrospect, the mammographic appearance met previously published criteria for probably benign lesions.
RESULTS: Of the 51 malignancies, 23 (45%) appeared mammographically as microcalcifications, 12 (24%) as masses, four (8%) as architectural distortion, and 12 (24%) as developing densities. None fulfilled strict criteria for a probably benign lesion when reviewed in retrospect. Forty-seven (92%) of 51 lesions had already demonstrated progression at the time of follow-up recommendation.
CONCLUSION: Short-term mammographic follow-up is often recommended for lesions that, in retrospect, do not fulfill established diagnostic criteria for probably benign lesions.
© RSNA, 2002
Index terms: Breast, 00.13 • Breast, biopsy, 00.1261 • Breast neoplasms, diagnosis, 00.31 • Breast neoplasms, radiography, 00.113, 00.114, • Breast radiography, utilization, 00.113, 00.114
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The previously published low frequency of malignancy and associated favorable prognosis of malignant lesions ultimately observed among probably benign lesions are predicated on the stringent application of specific diagnostic criteria used to both identify probably benign lesions and assess interval stability (1,2,5,12–14). We have observed, however, that in clinical practice, some lesions that do not meet the strict published criteria for probably benign lesions are nonetheless presumed to have a low suspicion for malignancy. Many of these are placed into a short-term mammographic follow-up category (Breast Imaging Reporting and Data System [BI-RADS] assessment category 3) rather than a benign or suspicious category, at the discretion of the radiologist. In fact, the American College of Radiology BI-RADS illustrated manual concedes that currently "most approaches [to the probably benign lesion] are intuitive" (15). Thus, it is possible that the types of lesions routinely assessed as probably benign in clinical practice are actually different than the "probably benign" lesions that have been systematically evaluated and published in the literature.
The purpose of this study was to determine whether systematically evaluated criteria for probably benign lesions were actually applied to lesions placed into that category.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Initially, all the examinations included in this study were prospectively performed and interpreted by one of six dedicated mammographers working at a large academic medical center. All six mammographers were full-time breast imaging radiologists and met all state and federal regulations during the study period. Existing prior studies and all pertinent clinical information were typically available to the interpreting radiologist at the time of the examination. As is our current practice, either another attending radiologist, breast imaging fellow, or resident double read many of these cases. During the study period, no standard diagnostic mammography protocol was used to evaluate lesions being considered for short-interval follow-up. However, none of the lesions were placed directly from a screening examination into the probably benign category. Instead, each case first underwent a diagnostic examination, which was managed at the discretion of the attending radiologist interpreting the case. Short-interval follow-up was recommended based on the radiologist’s assessment that the imaging characteristics of the lesion, as demonstrated on their diagnostic evaluation, were of very low suspicion.
For this study, two dedicated breast imaging radiologists (E.L.R., J.A.B.) independently and retrospectively reviewed all available breast images for 81 patients (83 malignancies) who developed ipsilateral breast malignancy after a short-interval follow-up recommendation. Each case was initially reviewed to determine whether the microcalcifications, densities, masses, or architectural distortions initially interpreted as probably benign actually represented the subsequently diagnosed malignancy. Fifty-one malignancies in 49 patients corresponded to lesions previously labeled "probably benign"; the other 32 malignancies in 32 patients were not related to the probably benign lesion. Each reviewer further characterized these 51 malignant probably benign mammographic lesions by using the American College of Radiology BI-RADS. Lesions were categorized by the mammographic abnormality as microcalcifications, mass, architectural distortion, or asymmetric density and were evaluated with use of BI-RADS descriptors and assessment categories for each class of lesion. These data were used to determine how many of the 51 malignancies, in retrospect, fulfilled the published criteria for a probably benign breast lesion (2,10,12–14).
Because all of the lesions were focal, diffuse probably benign categories were not considered. Thus, the lesions considered appropriate for the probably benign category were (a) noncalcified solid masses with an oval or lobular shape and with primarily circumscribed margins, (b) clustered calcifications that were round or oval at magnification mammography, and (c) focal asymmetric densities that resembled normal fibroglandular tissue at diagnostic evaluation. If prior mammograms and/or ultrasonographic (US) images were currently available, and had been available at the time of the short-term follow-up recommendation, they were also reviewed and incorporated into this decision analysis. Technical adequacy of each examination was also assessed. The independently compiled data from each reviewer were then reviewed together, and a final evaluation based on consensus was achieved.
Available data from audits of our mammography database during the study period were assessed by one author (E.L.R.) to determine the number of mammographic examinations performed during the study period. Complete audit data from 1998 were reviewed to determine the frequency of breast carcinoma diagnosed, recall rates, and positive predictive value for biopsy recommendations in our department.
Prognostic features of the invasive malignancies including size and stage at the time of surgery were evaluated. The time between initial mammographic detection and tissue diagnosis was determined by calculating the time from the initial follow-up recommendation to the time of biopsy recommendation.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Of the 51 malignancies, 23 (45%) appeared mammographically as clustered microcalcifications, 12 (24%) as asymmetric densities, 12 (24%) as masses, and four (8%) as foci of architectural distortion (Table 1).
|
The mean time interval between the initial follow-up recommendation and the subsequent biopsy recommendation for the 51 malignancies was 13.9 months (range, 0–36 months; 8.8 months for calcified masses, 12.1 months for microcalcifications, 14.2 months for developing densities, 14.5 months for noncalcified masses, and 31.5 months for architectural distortion).
On the basis of our retrospective review of these malignancies, none fulfilled the published strict criteria for a probably benign finding at the time of detection. The most common reason for exclusion from the probably benign category was that the lesion had already demonstrated interval progression when it was placed into the probably benign category. In fact, in the original report, the lesion was described as new or progressing in 41 (80%) of the 51 cases. According to our retrospective review, six additional lesions were determined to be new or increasing so that 47 (92%) of the 51 malignancies were new or increasing compared with those in prior studies in which they were classified as probably benign.
Twenty-two (96%) of 23 cases of clustered microcalcifications were new or increasing compared with those in all available prior studies. Although magnification mammography was performed for 21 (91%) of the 23 cases, motion blur, which substantially reduced image quality, was present in 13 (62%) of the 21 cases. However, all were focal, clustered microcalcifications, none of which consisted of the uniform homogeneous round or oval calcifications described as probably benign (2,5,10,12) (Table 2) (Figs 1–3).
|
|
|
|
|
|
|
|
|
Two (50%) of four cases of architectural distortion were new at the time of short-interval follow-up recommendation. One lesion was present but unchanged compared with that on a prior mammogram obtained 12 months previously. The other lesion was identified on a baseline study, but could not be localized on two orthogonal images (Fig 5). Although spot compression mammography was performed in all four cases, none were evaluated with either spot compression magnification mammography or directed US (Fig 6).
|
|
|
|
Of the 28 lesions consisting of a mass, focal density, or distortion, only five (18%) were interrogated with directed US. All five were masses and were correctly interpreted as solid with US. None of the findings were incorrectly attributed to cysts.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The probably benign category is based on three principles established in two large prospective trials, from the University of California at San Francisco (UCSF) and the Hospital Pereira Rossell in Montevideo, Uruguay (1,2,10,22). First, for patients and referring clinicians to forgo immediate biopsy, lesions assigned to the probably benign assessment category must have a very low likelihood of malignancy. Second, the small percentage of lesions assigned to the probably benign group that are actually malignant are rapidly identified—usually within 6–12 months—by a change in appearance at subsequent imaging. Finally, the small numbers of cases ultimately identified as malignant are of early stage and maintain prognosis similar to that of other cancers identified at routine screening mammography.
Sickles (2,10) at UCSF and Varas et al (1) at the Hospital Pereira Rossell have provided empirical evidence to support the hypothesis that specific breast lesions fulfill all three principles of the probably benign lesion. Both studies concluded that a variety of mammographic findings are appropriate for short-term follow-up. The most common of these are (a) noncalcified solid masses with an oval or lobular shape and primarily circumscribed margins, (b) clustered calcifications that are round or oval on magnification mammograms, and (c) focal asymmetric densities that resemble normal fibroglandular tissue at diagnostic evaluation. Of particular importance in both studies, lesions assigned to the probably benign category could not be new or increasing.
In fact, interval change in the mammographic appearance was the main feature that prompted a biopsy recommendation for these lesions (2,10,12,13,23). In a study of 3,184 lesions labeled "probably benign," biopsy was performed on 161 lesions (5%) for change in mammographic appearance (2). Among these 161 biopsies, 17 cancers were identified, which represents only 0.5% of the 3,172 cases. However, the fact that those 17 cancers represent 10.6% of the 161 biopsies performed underscores the diagnostic importance of a change in the mammographic appearance for rapidly identifying the very small number of malignancies among lesions initially meeting the mammographic criteria for the probably benign category.
The follow-up protocol in the Sickles study consisted of a unilateral mammogram obtained at 6 months and bilateral mammograms obtained at 12, 24, and 36 months. The few malignant lesions that initially fulfilled the criteria for probably benign lesions were rapidly identified with this follow-up protocol. Sixty-seven percent of malignancies were identified at the 6-month examination, and 94% were identified at the 12-month follow-up examination. These malignancies maintained the same good prognosis as that of other cancers identified at screening mammography. All lesions were either stage 0 or stage I, with a mean size of 1.4 cm.
In our study, lesions assigned to the probably benign category did not maintain the very low risk of malignancy, the short time to definitive biopsy, or the excellent prognosis demonstrated in previous studies (1,2,8,10). The positive predictive value of biopsy in our study was 28.7% (51 cancers in 178 biopsies) compared with only 10.6% in the Sickles study. In fact, the 28.7% positive predictive value parallels the overall positive predictive value in our practice, whereas the positive predictive value in this group should reflect a lower value, as in the UCSF study. Although the 51 cancers represent only 0.7% of the original 6,908 cases labeled probably benign in our study, we believe this number may be misleading. One of the limitations of this retrospective study is that the examinations were initially interpreted by six breast imaging radiologists. Although it is not possible to retrospectively review all of the almost 7,000 cases, it has been our experience that a substantial number of the lesions included in the probably benign category met imaging criteria for definitively benign lesions such as milk-of-calcium calcifications, superimposition, and fat-containing masses. In addition, several types of examinations, such as postcore biopsy and lumpectomy examinations, were routinely placed into the probably benign category. This practice would underestimate the proportion of malignant lesions by diluting the pool of probably benign lesions.
To achieve outcomes similar to those published in the literature likely requires both careful selection and surveillance of cases placed into the probably benign category. According to Sickles, this requires that the lesion is nonpalpable, undergoes full diagnostic imaging evaluation, and is compared with prior study findings. Only those nonpalpable lesions that have not demonstrated progression and that fall into one of the variety of mammographic findings considered probably benign should be placed into the short-term follow-up category. Comparison with prior mammograms, according to Sickles, "almost always will result in a different management recommendation than short-interval follow-up. For example, a newly apparent or progressing lesion will undergo tissue diagnosis because the very demonstration of progression during surveillance is what prompts biopsy instead of continued follow-up, whereas a finding already shown to exhibit long-term stability will not require surveillance at all" (10). In our series, however, 41 (80%) of 51 cancers were reported prospectively as new or increasing, and this should have precluded a probably benign diagnosis.
Although assessment of stability may be somewhat hindered by differences in technique between the two examinations, two strategies can facilitate comparison. First, the current examination should be tailored so that mammograms are obtained with a technique similar to that of the previous study. If additional images, such as fine-detail magnification views, were not obtained previously, then they should not be used to determine stability. Such fine-detail images, however, may be obtained to aid in further characterization of the mammographic lesion. Second, if the lesion is not visible on the previous study because of its location (or is only partially included), or if the previous study is limited owing to technical factors such as motion blur, then the current interpretation should simply be made without consideration of stability.
Malignant lesions in our study were not as rapidly identified at follow-up mammography as lesions were in other studies (2). The mean time to biopsy of malignant lesions in our study was 13.9 months (range, 0–36 months). In the Sickles study, almost all cancers were identified by change in the mammographic appearance at 6- or 12-month follow-up. We also observed a larger range of both tumor size and disease stage among cancers initially reported as probably benign. Although the median tumor size was similar for the two studies, the size range of invasive cancers in our study was 0.2–3.8 cm compared with 0.4–1.8 cm in the Sickles study. Further, all cancers in the Sickles study were stage 0 or I, whereas eight (16%) of the 51 malignancies in our study were stage II (three were T2N1M0 [one, four, and five positive nodes] and five were T2N0M0), carrying a substantially worse prognosis.
The higher rate of malignancy, longer time to diagnosis, and more advanced stage at diagnosis in our study reflect the differences in criteria used to assign cases to the probably benign category. Unlike the UCSF study in which only cases initially interpreted by Sickles were included, our study includes cases initially interpreted by six breast imaging radiologists. Furthermore, rather than relying on strict criteria established through rigorous scientific review to assign cases to the probably benign category, each radiologist applied standards based on their individual training and experience. Although palpable lesions were universally excluded, enlarging or new lesions, areas of subtle architectural distortion, and clusters of amorphous or pleomorphic microcalcifications were, in retrospect, common among lesions followed up in our study. Moreover, there was substantial variation in the type of diagnostic imaging evaluation that preceded placement of a lesion into the probably benign category. The American College of Radiology BI-RADS recognizes this individual approach as the current state of contemporary mammography practice, conceding that most approaches to establishing a probably benign assessment (BI-RADS assessment category 3) are "intuitive" (15). This intuitive approach was obviously used in our practice to some extent, based on the findings of our review.
Our data suggest that an intuitive approach to selecting probably benign lesions is not only different than one based on strict criteria but also may fail to meet the overall mission of the probably benign assessment category. The feature most predictive of malignancy among these lesions is not a morphologic characteristic but rather temporal instability. Most malignancies found in our series were lesions with appearances that could be considered probably benign except for the fact that they had already demonstrated interval change, which should have precluded their placement into the probably benign category.
Although the prevalence of malignancy among probably benign-appearing lesions that have progressed is unknown, previous studies have consistently demonstrated a higher positive predictive value for malignancy among these lesions (1,2,8,24). In fact, in the UCSF study, interval progression accounted for 80% of lesions for which biopsy was performed and was the only imaging manifestation among 15 of 17 malignancies. Another study, in which false-negative cancer cases were evaluated, found that 27 (84%) of 32 interval cancers that were prospectively seen on mammograms had been classified as probably benign, and in 17 (53%) of the 32 cases, 6-month follow-up was recommended (25). The authors state that nearly one-third of the cancers were asymmetric or developing densities, suggesting that a substantial percentage of these lesions had already demonstrated progression when they were assessed as probably benign. If anything, the results of this study underscore the importance of interval progression, no matter how subtle, in identifying malignancies among otherwise benign-appearing lesions.
Our data also substantiate the assertion that placement of a lesion into the probably benign category is a poor substitute for incomplete or suboptimal diagnostic evaluation. Many of the malignancies in our study, unfortunately, were either incompletely or improperly evaluated prior to their placement into the probably benign category. Common failures encountered in reviewing the diagnostic evaluations performed on the malignancies in our study included motion blur degrading magnification views, inaccurate or incomplete localization of lesions, and failure to assess interval progression. Had the interpreting radiologist closely followed the standards established by Sickles and Varas et al, immediate biopsy would have been recommended for each of these lesions because of the interval progression.
The findings of this study underscore the need for standardized and universally accepted criteria for the BI-RADS probably benign assessment category. A recent study, in which mammography performance in a community setting was evaluated, found that "discordant recommendations were especially evident for mammograms classified as probably benign," and that 27 (46%) of 59 interval cancers had been assessed as probably benign (26). Studies have established the efficacy of the probably benign category, but only when strict diagnostic criteria are applied. Radiologists who choose to use this assessment category in their practice should strive to adhere to these established guidelines so that the desired outcomes of the probably benign category are realized.
Given the high levels of inter- and intraobserver variability repeatedly documented for mammogram interpretations, including use of the BI-RADS assessment categories, it might appear that simple observer variability could account for the differences noted during retrospective review of these cases (27–33). However, an important finding of the current study is that the original interpreting radiologists not only recognized but also reported the key findings necessary to permit the appropriate assessment to be made in almost all of these cases. In 80% (41 of 51 cases) of these misclassified malignancies, the original interpreting radiologist reported that the lesions were either new or increasing or they used a suspicious morphology (such as pleomorphic) or a suspicious distribution (such as linear) to describe the lesion. However, rather than apply previously researched and published objective criteria, which would have led to biopsy, recommendations were based on an intuitive assessment of the imaging findings. Given the relatively high malignancy rate and worse prognosis among this cohort of miscategorized breast lesions, the conclusion of this study is that use of the probably benign category should be based solely on well-studied, documented criteria rather than on the individual radiologist’s instinct. There may be other, not yet validated, lesions that truly are probably benign, but we believe that outcomes of these lesions should be documented prior to their inclusion among the already validated probably benign lesion types.
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, E.L.R.; study concepts, E.L.R.; study design, E.L.R., J.A.B.; literature research, E.L.R.; clinical studies, E.L.R., J.A.B.; data acquisition and analysis/interpretation, E.L.R., J.A.B.; manuscript preparation and definition of intellectual content, E.L.R., J.A.B.; manuscript editing and revision/review, E.L.R., J.A.B., M.S.S.; manuscript final version approval, E.L.R.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. A. Harvey, B. T. Nicholson, and M. A. Cohen Finding Early Invasive Breast Cancers: A Practical Approach Radiology, July 1, 2008; 248(1): 61 - 76. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. D. Lehman, C. M. Rutter, P. R. Eby, E. White, D. S. M. Buist, and S. H. Taplin Lesion and Patient Characteristics Associated with Malignancy After a Probably Benign Finding on Community Practice Mammography Am. J. Roentgenol., February 1, 2008; 190(2): 511 - 515. [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]
|
|
|
|
 |
|
 J. MINIGH Mammographic Film Artifacts. Radiol. Technol., May 1, 2006; 77(5): 389M - 402M. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. B. Mainiero, A. Goldkamp, E. Lazarus, L. Livingston, S. L. Koelliker, B. Schepps, and W. W. Mayo-Smith Characterization of Breast Masses With Sonography: Can Biopsy of Some Solid Masses Be Deferred? J. Ultrasound Med., February 1, 2005; 24(2): 161 - 167. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
O. Graf, T. H. Helbich, M. H. Fuchsjaeger, G. Hopf, M. Morgun, C. Graf, R. Mallek, and E. A. Sickles Follow-up of Palpable Circumscribed Noncalcified Solid Breast Masses at Mammography and US: Can Biopsy Be Averted? Radiology, December 1, 2004; 233(3): 850 - 856. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. S. Majid, E. S. de Paredes, R. D. Doherty, N. R. Sharma, and X. Salvador Missed Breast Carcinoma: Pitfalls and Pearls RadioGraphics, July 1, 2003; 23(4): 881 - 895. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. M. Hall and E. L. Rosen Malignancy in BI-RADS Category 3 Mammographic Lesions * Dr Rosen responds: Radiology, December 1, 2002; 225(3): 918 - 920. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
