| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, University of Washington Medical Center, Seattle
Cancer Care Alliance, 825 Eastlake Ave. E, G2-600, PO Box 19023, Seattle, WA
98109-1023.
2 Group Health Center for Health Studies, Seattle, WA.
3 Public Health Sciences Division, Fred Hutchinson Cancer Research Center,
Seattle, WA.
4 Department of Epidemiology, University of Washington School of Public Health
and Community Medicine, Seattle, WA.
5 Applied Research Program, Division of Cancer Control and Population Sciences,
National Cancer Institute, Bethesda, MD.
Received March 1, 2007;
accepted after revision September 4, 2007.
Supported by the National Cancer Institute CA063731.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsdiscussionReferences |
|---|
MATERIALS AND METHODS. The subjects were women 30 years old and older without breast implants or previous breast cancer who received notice of a probably benign finding on a bilateral screening mammogram between January 1, 1996, and June 30, 1999, in a community-based practice. From 82,898 mammograms, we identified 129 breast lesions designated probably benign that progressed to malignancy within 3 years of an index examination (cases) and matched them to 129 lesions designated probably benign that did not progress to malignancy within 3 years (controls). A breast imaging specialist blinded to case–control status interpreted all examinations and recorded detailed lesion descriptors according to the BI-RADS lexicon.
RESULTS. Case lesions were more likely in patients who were older, postmenopausal, or had a strong family history of breast cancer or previous biopsy. The lesions were more likely masses with obscured, indistinct, or spiculated margins compared with control lesions (84.6% vs 66%, p = 0.03). Case lesions were more likely calcifications (29.5% vs 17.8%, p = 0.03). No cases were encountered among calcifications considered typically benign in the BI-RADS lexicon (vascular or coarse), and no controls were encountered among calcifications considered suspicious or highly suggestive of malignancy in the BI-RADS lexicon (amorphous, pleomorphic, branching, and fine linear) (p < 0.0001).
CONCLUSION. In community practice, patient and lesion mammographic characteristics can be predictive of the likelihood of a subsequent cancer diagnosis of mammographic lesions designated as probably benign. Careful evaluation of mass margins and of the morphologic features of calcifications can help distinguish a malignant lesion from a probably benign finding.
Keywords: BI-RADS • breast cancer • mammography • probably benign finding
|
TopAbstractIntroductionMaterials and MethodsResultsdiscussionReferences |
|---|
Category 3, probably benign finding, is used for a group of breast lesions that meet specific imaging criteria but are not definitely benign according to current standards. The likelihood of malignancy among these lesions is so low (< 2%) that biopsy may be safely deferred in almost all instances [3]. For these lesions, the recommendation is repetition of imaging at a short-interval (6 months) follow-up evaluation followed by annual mammography to assess stability. This practice of short-interval follow-up of mammographic lesions that have a 98% probability of being benign is supported by robust scientific evidence and by the ACR [3–7]. The purpose of the probably benign finding and short-interval follow-up is to provide a safe and cost-effective alternative to open surgical or percutaneous needle biopsy while providing imaging surveillance of lesions with a low probability of malignancy.
Results of some studies [8–10] raise specific concerns over the use of the probably benign category. There is high variability in the use of this category by practicing radiologists. Probably benign is used for as few as 1.2% and as many as 14% of screening mammograms interpreted. Furthermore, although the ACR links the probably benign finding to a recommendation for short-term follow-up, in actual community practice, short-interval follow-up frequently is not recommended for a probably benign finding [10, 11]. Finally, the cancer yield of a probably benign finding in community practice appears to be greater than the 2% reported in clinical trials [12]. We undertook this study to identify patient and lesion characteristics associated with a diagnosis of breast malignancy within 3 years of a probably benign finding (BI-RADS category 3) on a mammogram in a community radiology practice.
|
TopAbstractIntroductionMaterials and MethodsResultsdiscussionReferences |
|---|
Cases and Controls
Our study was conducted with women 30 years old and older with no breast
implants and no previous breast cancer who received notice of a probably
benign finding after bilateral screening mammography performed between January
1, 1996, and June 30, 1999. For a lesion to be considered a control, the woman
had to be enrolled in the health system for 3 years after the mammographic
examination. For a lesion to be considered a case, the woman had to have
received the diagnosis of invasive breast cancer or ductal carcinoma in situ
within 3 years of mammography.
Among 82,898 screening mammograms (55,022 unique women), we identified 1,711 examinations of 1,242 unique women with a BI-RADS assessment of probably benign. We separated women with a probably benign finding into cases and controls. Using the Surveillance Epidemiology and End Results database for western Washington, we identified all instances in which cancer was diagnosed in the ipsilateral breast within 3 years of a probably benign finding (n = 150). All examinations were reviewed to confirm that the laterality and location of the malignant lesion corresponded to the laterality and location of the lesion designated as probably benign.
Eligible controls were identified among women who received an assessment of probably benign but in whom ipsilateral breast cancer did not develop within 3 years. We matched one case to one control on the basis of the month and year of the index screening mammogram to account for secular changes in the use of the probably benign assessment and to ensure similar duration of follow-up. We did not match other characteristics, including age, because we planned analytic comparison of those characteristics.
Among the 150 instances of cancer identified in the database, 21 were ineligible because images were not available in 12 instances, information in the medical record and the automated database was discordant for either the case or the control in seven instances, and radiology reports indicated the presence of a palpable mass at the time of the index examination in two instances. A total of 129 matched case–control sets were used in the analyses.
Most (85%, 218/258) of the women had an assessment of a unilateral probably benign lesion. For patients with assessments of bilateral probably benign, the breast lesion that became malignant was included as a case. For control patients with bilateral lesions, we selected the lesion on the same side as the matched case. We selected the first mammogram for women with more than one eligible mammogram.
Data Collection
The study radiologist, who was blinded to case–control status,
interpreted all study mammograms with the purpose of evaluating detailed
lesion characteristics. All instances were evaluated to confirm that the
probably benign lesion corresponded to the malignancy diagnosed. The study
radiologist interpreted mammograms using the BI-RADS lexicon for assessments
and recommendations, evaluation of breast density, and lesion-specific
descriptors. Lesion descriptors included location, size, type (calcifications,
mass, architectural distortion, global asymmetry, focal asymmetry), and
specific characteristics of calcifications (morphologic features, number,
distribution) and masses (size, density, margins, shape). We separated
calcification type by malignancy potential into low (vascular and coarse),
medium (punctuate), and high (amorphous, pleomorphic, branching, fine linear).
Lesions were categorized as stable or increasing over time for index
examinations in which a comparison was available (63 cases, 50 controls).
We examined patient characteristics, which were obtained from self-report and automated data at the time of interpretation of the study mammogram. These characteristics included age, hormone use, breast biopsy history, menopause status, family history, comparison mammograms, and results from additional mammographic views and sonographic examinations.
Statistical Analysis
We compared cases with controls using chi-square tests for association or
exact tests, as determined by sample size. Comparisons were made for all
characteristics of patients (age, menopause status, breast density, use of
hormone replacement therapy, family history, and biopsy history), specific
lesion descriptors recorded by the study radiologist (i.e., mass margins and
shape or calcification morphologic features and distribution), and the
availability of comparison images. We did not incorporate matching into our
analyses because we did not expect strong correlation between matched pairs
owing to weak matching criteria (time of index mammogram).
|
TopAbstractIntroductionMaterials and MethodsResultsdiscussionReferences |
|---|
The differences in patient characteristics of cases and controls are summarized in Table 1. Cases were more likely among patients who were older (62.7 vs 56.3 years, p < 0.0001) or postmenopausal (78% vs 62%, p = 0.005). Cases also were more common among women with a first- or second-degree family history of breast cancer (43% vs 28%, p = 0.009) or history of breast biopsy (31% vs 16%, p = 0.003). Use of hormone replacement therapy was not significantly different (p = 0.27) between cases (60%) and controls (51%). The proportion of women with heterogeneously or extremely dense breasts was not significantly different between cases (46%) and controls (44%).
|
Masses were the most common type of lesion among both cases (40.3%) and controls (38.8%) (Table 2). The frequency of focal asymmetry was similar for cases (21.7%) and controls (27.1%). There were significantly more calcifications among cases (29.5%) than controls (17.8%) (p = 0.03). However, there were significantly fewer global densities among cases (1.5%) than controls (7.8%) (p = 0.02).
|
The characteristics of 102 masses (52 cases, 50 controls) are summarized in Table 3. Overall, there was no significant difference in shape of masses among cases and controls. Most (67/102, 65.7%) of the masses were round or oval. However, there was a trend toward more irregular masses among cases (13/52, 25.0%) than among controls (5/50, 10.0%). Lesion margins differed significantly for cases and controls (p = 0.04). Cases had a greater likelihood (76.0% vs 66.0%) of having obscured or indistinct margins. Spiculated masses were present in 9.6% (5/52) of the cases, whereas there were no such lesions among controls.
|
Cases were more likely than controls to have calcifications (38/129, 29.5% vs 23/129, 17.8%; p = 0.03) (Table 2). The characteristics of calcifications in cases and controls are summarized in Table 4. There was no difference in the proportion of cases and controls with punctate calcifications (34.2% vs 39.1%). No cases had vascular or coarse calcification types, which are described as typically benign (BI-RADS categories 1 and 2). Neither cases nor controls had milk of calcium calcification types. In addition, no controls had amorphous, pleomorphic, branching, or fine linear calcifications, which are described as a suspicious abnormality or highly suggestive of malignancy according to BI-RADS. These findings resulted in a significant difference in the malignancy potential of calcifications between cases and controls (p < 0.015).
|
Comparison images were available for 63 of 129 (48.8%) of the cases and 50 of 129 (38.8%) of the controls. Of the examinations with comparisons, 97 of 113 (85.8%) showed an increase in the lesion without a significant difference (p = 0.12) between cases (57/63, 90.5%) and controls (40/50, 80%).
|
TopAbstractIntroductionMaterials and MethodsResultsdiscussionReferences |
|---|
Our results indicate that the criteria for a probably benign finding were not strictly followed. Among the 258 examinations in our study, only 52 (20% overall; 25 of 129 [19%] controls; 27 of 129 [21%] cases) met the morphologic criteria for a probably benign finding: noncalcified circumscribed mass, focal asymmetry, or a group of round or punctate calcifications [1]. We found amorphous, pleomorphic, branching, and fine linear calcifications along with masses with indistinct and spiculated margins among the lesions assessed as probably benign.
In an unblinded retrospective study, Rosen et al. [14] reported on 51 lesions that had been assessed as probably benign and later were diagnosed as malignant. The authors found that none of the 51 lesions met the strict morphologic criteria for probably benign at the time of the original recommendation for short-interval follow-up. For example, linear and pleomorphic calcifications and irregular masses with indistinct margins were assessed as probably benign. Our results in a blinded study of probably benign lesions in community practice support the findings from the study Rosen et al. conducted in an academic setting.
The cancer yield of 8.8% (150/1,711) for lesions categorized as probably benign in this study population is higher than the expected yield of less than 2% and supports the observation that failure to adhere to a strict set of probably benign lesion characteristics may result in a stronger association with malignancy. The converse findings are also interesting: 27 of 129 cases did meet the strict morphologic criteria for a probably benign assessment. This figure translates to a cancer yield of 1.6% (27/1,711) for all instances of probably benign findings in our study and falls squarely within the accepted published value of less than 2%.
It is interesting that 50 of 129 (38.8%) of the controls and 63 of 129 (48.8%) of the cases had previous images available at the time of the index examination, because an assessment of probably benign is typically not recommended for mammograms with comparisons [1]. Among the cases in which there were comparison images, most (57 of 63, 90.5%) of the lesions were already increasing in size (for masses) or number (for calcifications) at the time of the recommendation for short-interval follow-up. This finding is consistent with that reported by Rosen et al. [14], that is, that 92% of malignant lesions assessed as probably benign had already progressed (in size for masses and in number for calcifications) compared with findings at previous examinations.
In our study, 40 of 50 (80.0%) lesions designated as probably benign that did not progress to cancer were also increasing compared with findings at the previous examination. How can we explain the apparent increase in the mammographic findings that did not progress to malignancy? The morphologic criteria for probably benign lesions were not strictly observed in 80% of the study examinations. It may be that many of the control lesions had typically benign morphologic features but were assessed as probably benign because they were changing, such as increasing coarse or diffuse punctate calcifications. This theory may explain why some lesions with typically benign characteristics were nevertheless classified BI-RADS category 3 (on the basis of the increase since the previous examination). Another possibility is that some of these lesions were assessed as potentially evolving dystrophic calcifications. New calcifications suggestive of early fat necrosis may be recommended for short-interval follow-up rather than biopsy in appropriate clinical settings [15].
Taken together, the strict criteria for the probably benign assessment were not followed in at least 80% of the examinations in our study owing to morphologic features, availability of previous images, or changes in the findings over time. However, we cannot determine the true frequency of misapplication of the probably benign assessment because we included only the 258 matched probably benign cases and controls in the final analysis. A review of all 1,711 examinations with a probably benign assessment would be needed to ascertain the true frequency of misapplication of BI-RADS category 3.
We found that patient characteristics may influence the risk of malignancy in lesions designated probably benign. In the current BI-RADS lexicon, the only patient characteristic recognized for assessments and recommendations in the management of breast lesions is the presence or absence of a palpable finding. Nevertheless, previous reports have shown that radiologists assessing mammograms are influenced by patient history. Elmore et al. [16] provided false patient histories to mammogram readers and found that diagnostic evaluation was recommended to more patients when the history indicated increased risk. Kerlikowske et al. [17] reported that the positive predictive value of screening mammography was higher among women with a family history of breast cancer. Our results also suggest that as with palpable findings, factors such as patient age, family history, biopsy history, and menopause status can be helpful in predicting likelihood of malignancy of mammographic lesions and thus influencing recommendations.
The results of our study support guidelines developed by the ACR for use of BI-RADS category 3. Lesions that do not meet the strict morphologic criteria of BI-RADS category 3 have a likelihood of malignancy that is higher than 2%. According to the recommendations of the BI-RADS committee, "probably benign" is not intended for lesions that are indeterminate but for lesions that are almost certainly benign. In the most recent version of BI-RADS [1], intuitive approaches to this category are not supported, and clearly defined lesion features are further emphasized. Careful evaluation of the morphologic features of calcifications and mass margins can help exclude a malignant lesion from a probably benign finding. In addition, individual patient history and clinical information, as with palpable findings, may increase the risk of malignancy above the threshold of 2% even for lesions that satisfy the strict criteria of BI-RADS category 3. In the most recent version of BI-RADS [1], it is suggested that biopsy be performed for a probably benign lesion on the basis of patient or clinician preference. Our findings support this suggestion and indicate that patient characteristics such as age or family history may support use of biopsy rather than follow-up in individual cases.
Acknowledgments
We thank Sue Peacock of the University of Washington for her assistance in
completing this manuscript. We also thank Alice Park, Deb Seger, and Letitia
Hodgkinson from Group Health Cooperative for their help with data collection
and project management.
|
TopAbstractIntroductionMaterials and MethodsResultsdiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
