Original Research  n  Breast

Imaging

Note:  This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

Undiagnosed Breast Cancer at MR Imaging: Analysis of Causes1

Emmanuelle Bouic Pages, MD Ingrid Millet, MD Denis Hoa, MD Fernanda Curros Doyon, MD Patrice Taourel, MD, PhD

1

 From the Department of Medical Imaging, CHU Lapeyronie, 371 avenue du Doyen Gaston Giraud, 34295 Montpellier, France. Received September 7, 2011; revision requested October 14; revision received December 19; accepted December 29; final version accepted January 13, 2012. Address correspondence to P.T. (e-mail: [email protected]).

Purpose:

To retrospectively review the causes of false-negative results on prior magnetic resonance (MR) imaging studies in patients who developed breast cancer as revealed on a follow-up MR imaging study and to determine the presumptive causes of these false-negative findings.

Materials and Methods:

Fifty-eight pairs of MR imaging studies from one institution were assessed, consisting of a prior study without a diagnosis of cancer and a diagnostic study with subsequent findings of 60 cancers in 58 women at MR imaging (mean interval between prior and diagnostic MR examinations, 13.8 months). Two radiologists reviewed in consensus, in a nonblinded fashion, each pair of MR studies, comparing the diagnostic and the prior MR imaging studies to evaluate the rate of false-negative findings. The prospective reports were then analyzed to classify false-negatives findings in breast enhancement of breast cancers not identified at the time of imaging, potentially misinterpreted, and mismanaged. False-negative results on prior MR studies were retrospectively reassessed to identify possibly reasons why cancers had been not recognized, potentially misinterpreted, or mismanaged.

Results:

Twenty-eight (47% [95% confidence interval {CI}: 34%, 59%]) of the 60 cancers were retrospectively diagnosed as Breast Imaging Reporting and Data System grade 3, 4, or 5 lesions. Analysis of the prospective reports showed that six lesions (10% [95% CI: 2%, 18%]) had been not identified at the time of diagnosis, 15 lesions (25% [95% CI: 14%, 36%]) were potentially misinterpreted, and seven lesions (12% [95% CI: 3%, 20%]) were mismanaged. The main causes of misinterpretation were smooth margins of a mass (n = 4), stability in size (n = 3), and location of a nonmass in a postsurgical area (n = 5). Mismanagement was mainly due to inadequate correlations between MR imaging and ultrasonographic (US) features, with inaccurate sampling with US guidance in five cases.

Conclusion:

In patients with breast cancer seen at MR imaging, retrospective evaluation of the prior MR imaging studies showed potential observer error in 47% of cases, resulting more from misinterpretation than from nonrecognition or mismanagement of cancers.  RSNA, 2012

q

 RSNA, 2012

q

40

radiology.rsna.org  n  Radiology: Volume 264: Number 1—July 2012

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

O

f all currently available breast im­ aging techniques, magnetic resonance (MR) offers the highest sensitivity for both invasive and intraductal cancers. In the literature data (1–7), MR imaging sensitivity was higher than that of mammography, regardless of the mammographic density or breast cancer type. Overlooking breast cancers, particularly invasive forms, with MR imaging is uncommon, but false-negative diagnoses may occur with a reported sensitivity of 90.9% for invasive cancer and 73% for ductal carcinoma in situ (DCIS) (8). Undiagnosed breast cancer on MR images may be owing to a lack of enhancement or to missed or misinterpreted enhancement. Lack of enhancement is due to parameters intrinsic to tumors that have not elicited neovascularity for their vascular supply, as mainly encountered with DCIS or with very small invasive carcinomas (4,9–11). Missed enhancement is generally due to the small size of enhancement or to the presence of a surrounding strongly enhanced glandular parenchyma, whereas in cases of misinterpreted enhancement, the enhancement has been identified but interpreted as a benign structure or as normal parenchyma. Retrospective unblinded analyses of mammograms prospectively considered as normal in women with breast cancer detected at follow-up screening mammography have revealed that twothirds of patients had lesions visible in retrospect on a mammogram considered normal, which included diagnostic errors with missed and misinterpreted cancers and subtle nonspecific mammographic findings (12,13). These analyses enabled identification of the mammographic

characteristics of missed cancer and analysis of the possible reasons for diagnostic errors. To our knowledge, although breast MR imaging has been increasingly used for clinical practice, to date no analysis has focused on this topic with breast MR imaging. The aim of our study was twofold: (a) to retrospectively determine the rates of nonenhancement, not identified enhancement, and misinterpreted or mismanaged enhancement in patients with breast cancer diagnosed by means of MR imaging and undiagnosed on the prior study, and (b) to assess the MR imaging characteristics and the possible reasons for not identified, misinterpreted, or mismanaged malignant enhancement.

Materials and Methods Case Selection This Health Insurance Portability and Accountability Act–compliant retrospective study was approved by our institutional review board; the informed consent requirement was waived. A cross-computer search of our pathology records and our MR imaging records identified a total of 520 patients with breast cancer between January 2005 and December 2010, with an interval

Advance in Knowledge nn In patients with breast cancer seen at MR imaging, retrospective evaluation of prior MR imaging studies revealed potential observer error in 28 (47%) of 60 cancers, resulting more from misinterpretation (n = 15) than from mismanagement (n = 7) or nondetection (n = 6). Radiology: Volume 264: Number 1—July 2012  n  radiology.rsna.org

Implication for Patient Care nn Some guidelines must be followed to decrease the rate of misinterpretation or mismanagement in breast MR imaging: (a) The stability of an enhancement does not confirm benignancy; (b) even if encountered in areas of scarring, postsurgical enhancement in a treated breast must be considered as suspicious; (c) nonenhancing septa are not sufficient to confirm the benignancy of a mass; (d) at biopsy guided by targeted US for lesions detected initially at MR imaging, falsenegative findings may occur in benign concordant results, making clip placement and follow-up imaging recommended.

Pages et al

between MR imaging and cancer biopsy of less than 1 month. In 58 of these 520 patients, a prior MR imaging examination had been performed 6–24 months before the one performed at the time of diagnosis (hereafter called the “diagnostic MR imaging” examination), and the mean interval between prior and diagnostic MR imaging was 13.8 months. Sixty tumors that constituted the study group were present in these 58 patients; two patients had a bilateral tumor. All patients were women between the ages of 40 and 78 years (mean, 58 years) at diagnosis. Diagnosis of menopause was made by means of menstrual and medical history after 12 months of amenorrhea, or having no menstrual periods. The tumors were infiltrative carcinomas in 48 cases and DCIS in 12 cases. Infiltrative carcinomas included 42 ductal carcinomas (15 with ScarffBloom-Richardson [SBR] grade 1, 20 with SBR grade 2, and seven with SBR grade 3), five lobular carcinomas (all with SBR grade 2), and one mucinous carcinoma. The pathologic size of invasive carcinoma ranged from 2 to 27 mm (mean, 9.9 mm). Thirty-seven patients had been included in a prior study assessing growth of breast cancer recurrences by means of consecutive MR imaging examinations (14).

Published online before print 10.1148/radiol.12111917  Content code: Radiology 2012; 264:40–50 Abbreviations: BI-RADS = Breast Imaging Reporting and Data System DCIS = ductal carcinoma in situ GRE = gradient echo IDC = invasive ductal carcinoma SBR = Scarff-Bloom-Richardson Author contributions: Guarantors of integrity of entire study, E.B.P., I.M., P.T.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, E.B.P., D.H., P.T.; clinical studies, E.B.P., I.M., F.C.D., P.T.; statistical analysis, I.M., D.H.; and manuscript editing, E.B.P., D.H., P.T. Potential conflicts of interest are listed at the end of this article.

41

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

Indications for performing the prior breast MR examination included highrisk screening due to prior breast cancer (24 patients), positive family history of breast cancer (14 patients, with four patients with BRCA mutation) or “problem-solving” in 20 patients. In 38 patients, the prior breast MR examination was the first one performed, whereas in 22 patients more than one MR imaging examination had been previously performed (hereafter called “previous MR imaging”) and these were available for analysis. Indications for performing the diagnostic MR imaging examination were as follows: in seven patients, for the staging of a breast cancer diagnosed with other modalities; in 35 patients, for personal risk factors; in 13 patients, for solving problems after conventional evaluation; and in three patients, for unexplained reasons.

MR Imaging Technique All contrast material–enhanced MR mammographic examinations were performed with a 1.5-T magnet (Vision Plus; Siemens, Erlangen, Germany) by using a bilateral breast surface coil. Patients were placed in the prone position on the table. A transverse threedimensional T1-weighted gradient-echo (GRE) dynamic sequence was performed before contrast agent administration, followed by repeat performance of this same sequence 1, 2, 3, 4, 5, and 6 minutes after contrast agent administration. The precontrast and postcontrast three-dimensional T1-weighted GRE dynamic MR images were acquired with the following parameters: repetition time msec/echo time msec, 8.4/4.5; flip angle, 25°; matrix, 448 3 309 pixels; field of view, 320 3 320 mm; and section thickness, 2 mm, with no intersection gap. The total acquisition time for the three-dimensional T1-weighted GRE dynamic sequence was 360 seconds. An additional T2weighted turbo spin echo sequence was performed before contrast agent administration: 12 240/98; matrix, 448 3 403 pixels; echo train length, 18; field of view, 320 mm; and section thickness, 2 mm. The acquisition time for the 42

Pages et al

T2-weighted turbo spin echo sequence was 220 seconds. Postcontrast three-dimensional T1weighted GRE dynamic MR images were acquired after administration of 0.1 mmol gadobenate dimeglumine (MultiHance; Bracco Imaging, Milan, Italy) per kilogram of body weight through an 18-gauge cannula positioned in an antecubital vein. Gadobenate dimeglumine was administered by using an automatic injector (Spectris; Medrad, Indianola, Pa) at a rate of 3 mL/sec, followed by administration of 10 mL of saline at the same rate.

MR Interpretation MR studies were retrospectively evaluated by two breast imagers with expertise in breast MR imaging (P.T. and E.B.P., with 15 and 5 years experience, respectively) on a dedicated picture archiving and communication system unit (GE Healthcare). This consensus retrospective review involved a nonblinded comparison of the diagnostic and the prior MR imaging studies. The review was performed with all clinical and diagnostic information about the location, size, and histologic findings of the malignancy available, and with the prospective report of the prior MR imaging examination. Review of previous MR imaging studies, when available, was performed in the same way in consensus by the two same readers. The enhancing lesion corresponding to a known cancer was morphologically classified on the diagnostic MR images by using the Breast Imaging Reporting and Data System (BI-RADS) MR lexicon (15) and measured in its greatest diameter on the axial plane. For the prior MR imaging examination, the prospectively performed interpretation (clinical MR studies are reported on by a staff radiologist with at least 2 year experience in breast MR imaging, from a pool of six breast radiologists) was considered as correct in retrospect when there was not any enhancement at the site where an enhancing lesion subsequently developed, or where there was only a focus within multiple and bilateral foci, or where there were multiple and bilateral

patchy enhancements without any particular features in terms of morphology, intensity, or kinetics at the site where enhancing lesions subsequently developed or where a focus was accurately graded as BI-RADS 3, with a 6-month follow-up recommendation. In other cases, where an enhancement could be retrospectively indentified, it was classified according to the BI-RADS MR lexicon, measured in its greatest diameter in the axial plane on the 3 minutes postcontrast phase image and scored as not recognized, potentially misinterpreted, or mismanaged. All measurements were performed by the same operator who was one of the two reviewers (E.B.P.). The differentiation between detection, interpretation, and management factors was based on the prospective report. A malignant enhancement was considered as not recognized when it was not described on the prospective MR imaging report, as potentially misinterpreted when it was described as benign, and as mismanaged when it was described as malignant or possibly malignant (BI-RADS 4 or 5) without pathologic confirmation of malignancy. To explain why the findings were originally misdiagnosed on the prior MR imaging study, we assessed (a) the diagnostic quality of the study by evaluating motion artifacts and inadequate infusion of contrast material if no contrast enhancement was present within the cardiac cavities and the internal mammary vessels, (b) the background enhancement by using a grading scale described by Kuhl (16), with the absence of background enhancement indicating a homogeneously dark background, mild and moderate background enhancement, and severe background enhancement referring to strong, early-phase, multifocal, or diffuse fibroglandular enhancement that may mask enhancing breast cancer, (c) the lesion size, (d) the tumor location, which could be near or beyond the outer boundary of the field of view, (e) the presence of benign lesions in contact with the misdiagnosed enhancement, which may be a confounding factor, (f) an eventual stability of the enhancement

radiology.rsna.org  n Radiology: Volume 264: Number 1—July 2012

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

size in women with available previous MR imaging studies, and (g) a clinical history of homolateral breast conservative treatment, since lumpectomy site enhancement is common and may be a cause of pitfalls (17). For assessing stability in size as a potential cause of misinterpretation of an enhancement at the prior MR imaging examination we analyzed the growth rate between previous MR studies when available and the prior MR imaging study. Before the prior MR imaging study, there was one additional previous MR study available in 13 patients, two in six patients, and three in three patients, making 34 time interval segments of at least 6 months between consecutive MR imaging examinations for analysis of tumoral growth. The interval between consecutive MR examinations ranged from 4.3 months to 38.4 months (mean, 14.6 months). A growth rate analysis was performed by measuring the greatest lesion diameter in the axial plane. Size of the lesions was measured by using calipers on the 3-minute-postcontrast phase image on the picture archiving and communication system unit. A tumor was considered as stable on a time interval segment when the greatest tumor diameter did not change by more than 10% during this interval, as decreasing when its greatest diameter decreased by more than 10%, and as increasing when its greatest diameter increased by more than 10% or when the tumor appeared for the first time. For the percentages of lesions not recognized, potentially misinterpreted, or mismanaged, 95% confidence intervals were calculated. The sizes of the lesions at the time of diagnosis were compared according to the type of diagnostic error by using the Student t test.

Results MR images at diagnosis showed 37 enhancing masses in 36 patients, 16 cases of non-masslike enhancement in 16 patients, and seven foci in seven patients. Two patients had bilateral cancer, with a mass in each side for one patient and a mass in one breast and a

Pages et al

Table 1 Thirty-Two Cases of Malignancy in Which No Error Was Found Parameter Mean age (y) Hormonal status  Premenopausal  Menopausal   Hormone replacement therapy Risk factors  BRCA1   Personal history  Ipsilateral  Controlateral   Familial history Pathologic finding   DCIS intermediary grade   DCIS high grade   IDC SBR grade 1   IDC SBR grade 2   IDC SBR grade 3   ILC SBR grade 2 BI-RADS feature (diagnostic MR imaging)  Mass   Non-masslike enhancement  Focus Mean size (mm) (diagnostic MR imaging)  Mass   Non-masslike enhancement  Focus Time interval from prior to diagnostic MR imaging (mo) BI-RADS feature (prior MR imaging)   BI-RADS 1 (no enhancement)   BI-RADS 3 (isolated focus)   Patchy enhancement or multiple bilateral foci

Finding 61 (40–78) 4 24 4 1 14 3 11 12 2 5 10 10 3 2 16 12 4 10.2 (6–20) 29 (10–75) 5 (5–5) 16.4 (6–24) 11 6 15

Note.—Except where indicated, data are numbers of patients or lesions (32 malignancies in 32 women). Data in parentheses are the ranges. ILC = invasive lobular carcinoma.

non-masslike enhancement in the controlateral breast for the other patient. Enhancing masses corresponded to invasive carcinomas in 33 lesions and to DCIS in four lesions, non-masslike enhancements were due to invasive carcinomas in nine lesions and to DCIS in seven lesions, and focus corresponded to invasive carcinomas in all cases. On the prior MR imaging study, even in retrospect, there was no mistake in 32 of 60 cancers (53% [95% confidence interval: 41%, 66%]). There was no enhancement in 11 cases and there were bilateral and multiple foci without any predominant or patchy

Radiology: Volume 264: Number 1—July 2012  n  radiology.rsna.org

enhancement in 15 cases, whereas in six cases an isolated focus was accurately graded as BI-RADS 3 (Table 1). In these 32 cancers, the studies were considered technically adequate in all cases. In 28 of the 60 reviewed lesions (47% [95% confidence interval: 34%, 59%]), an abnormality was observed in retrospect, which changed the BIRADS classification. In these 28 cancers, all studies were also considered technically adequate. In six of these 60 cancers (10% [95% confidence interval: 2%, 18%]), a lesion was missed (Fig 1), with a 43

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

Pages et al

Figure 1

Figure 1:  Cancer not recognized at the time of imaging, related to a preaxillary location, in a 65-year-old menopausal woman with a history of left breast cancer treated by means of mastectomy. Transverse (a) T2-weighted turbo spin-echo and (b) contrast-enhanced T1-weighted GRE substracted MR images show small mass (arrow) in contact with vessels, not described on the prospective report. (c) On MR images obtained 2 years later, an 8-mm spiculated mass (arrow) is seen, corresponding to an SBR grade 2 invasive ductal carcinoma (IDC).

diagnosis of invasive cancer in all cases at follow-up (Table 2). In these six cancers, the background enhancement was given a score of 1 in two cases, 2 in two cases, 3 in one case, and 4 in one case. In 15 of the 60 cancers (25% [95% confidence interval: 14%, 36%]), the lesion was correctly identified but was incorrectly interpreted as a benign lesion or benign enhancing tissue. At the retrospective analysis as well as in the prospective report, the lesion appeared as a mass in eight cases, as a nonmasslike enhancement in five cases, and as a focus in two cases (Table 3). In three of these lesions, the enhancement was considered stable in the prior MR imaging prospective report (Fig 2). Retrospective analysis of growth rates showed cancers stable or decreasing in size in 23.5% of the time interval segments analyzed, with at least one segment during which the tumoral size was stable or decreasing in eight of the 23 lesions for which previous MR images were available before the prior MR study and a mean time of stable or decreasing size of 12.7 months (range, 5.4–27.5 months). In the prospective report, stability in size was reported as suggesting benignancy in three lesions. Four masses in four patients were prospectively considered as fibroadeno 44

mas because they had smooth margins in three cases or because they had nonenhancing internal septas prospectively considered as specific of benignity in one case (Fig 3) and/or because they occurred in patients with known fibroadenomas in three cases. The mass was considered to be an intramammary lymph node because of its location and its hyperintense signal on T2-weighted images in one case. The non-masslike enhancements were considered benign because they occurred in a postsurgical area and were interpreted as scars in five cases (Fig 4). One focus was considered benign because it was stable in one case. In seven of 60 cancers (12% [95% confidence interval: 3%, 20%]), the lesion was prospectively diagnosed as suspicious for malignancy, with the identification of a mass prospectively graded as BI-RADS 4 or 5 (Table 4). In five cases, a correlate was identified on the second-look US study and a USguided biopsy was performed, with a pathologic diagnosis of papilloma in one case, fat necrosis in one case, and benign mastopathy in three cases. In two cases (including one case in which a biopsy was also performed with US guidance), a biopsy was performed by using MR imaging guidance, with a diagnosis

of fibro­ adenoma in one case and fat necrosis in one case. In two cases, the patient refused the biopsy: in one case after bleeding, which interrupted a first attempt, and in one case because of claustrophobia in a patient for whom an MR-guided biopsy was required. In patients in whom there were diagnostic errors, the mean size of the lesion was 9.5 mm (3–18) on the prior MR imaging study and 12.3 mm (3–30) on the diagnostic MR imaging study, with a mean interval between prior and diagnostic MR imaging of 15 months (6–24). Data according to the reasons for misdiagnosed cancer are given in Table 5. At the time of diagnosis, the size of the lesion not recognized on the prior MR imaging study was smaller than that of the lesion potentially misinterpreted or mismanaged (mean size difference, 5 mm [range, 21.8 to 75]; P , .001).

Discussion Classically, nondetection of breast carcinoma at MR imaging is due to histomorphologic characteristics of breast cancers (11): Small (4–5 mm), invasive cancers diffusely growing in a cellby-cell fashion, as encountered in lobular carcinomas, intraductal carcinomas,

radiology.rsna.org  n Radiology: Volume 264: Number 1—July 2012

Radiology: Volume 264: Number 1—July 2012  n  radiology.rsna.org

Personal history  (homolateral)

IDC SBR 2

Familial history ILC SBR 2* BRCA1 IDC SBR 3 Personal history IDC SBR 2  (contralateral) IDC SBR 1 Familial history IDC SBR 1

Risk Factor

23

24 19

15 11 24

Familial history

Personal history (ipsilateral)

Personal history (ipsilateral) Familial history DCIS high grade Mucinous carcinoma IDC SBR 2 DCIS intermediary grade DCIS low grade

IDC SBR 2 IDC SBR 3 IDC SBR 3 IDC SBR 1 IDC SBR 2 6 12 15 12 12

12 6 12 15 12

23 12 12 24 13

Interval from Prior MR to Diagnostic MR Study (mo)

Mass

Focus Focus

Mass Focus Mass

BI-RADS Feature

14 6 19

5

7

3 13 7 7 8

Size at Previous MR Imaging Study (mm)

†

Woman with bilateral cancer.

7

3 4

13 4 8

Size (mm)

1

1 4

2 2 2

Mass

Focus Mass

Mass Focus Mass

BI-RADS Feature

Nonmass Mass Nonmass Mass Mass

Mass Mass Nonmass Nonmass Focus

Focus Mass Nonmass Mass Mass

BI-RADS Feature

17 8 14 6 8

8 12 Regional 9 4

4 15 11 7 10

Size (mm)

1

1 4

2 3 2

Background Enhancement

Nonmass Mass Nonmass Mass Mass

Mass Mass Nonmass Mass Focus

Mass Mass Nonmass Mass Mass

BI-RADS Feature

30 18 17 7 8

10 14 Regional 10 4

20 17 15 8 12

Size (mm)

Diagnotic MR Imaging Study

8

3 8

17 4 9

Size (mm)

Diagnotic MR Imaging Study

Prior MR Imaging Study

Background Enhancement

Prior MR Imaging Study

* Right-sided cancer in a patient with bilateral cancer and personal history of right breast surgery (benign histologic finding).

Note.—ILC = invasive lobular carcinoma.

48/Menopausal 63/Menopausal 54/Menopausal* 63/Menopausal† 63/Menopausal†

55/Menopausal 46/Premenopausal 57/Menopausal 59/Menopausal 64/Menopausal

IDC SBR 2 IDC SBR 3 ILC SBR 2 IDC SBR 2 DCIS high grade

64/Menopausal 53/Premenopausal 57/Menopausal 72/HRT 55/Menopausal

Personal history (ipsilateral) Personal history (ipsilateral) Familial history BRCA2, personal   history (ipsilateral) Personal history (ipsilateral) BRCA1

Pathologic Finding

Patient Age (y)/Hormonal Status Risk Factors

Lesions Potentially Misinterpreted

Table 3

*Left-sided cancer in patient with bilateral cancer. ILC = invasive lobular carcinoma.

60/Menopausal

65/Menopausal 44/Premenopausal

54/Menopausal 44/Premenopausal 65/Menopausal

Patient Age (y)/Hormonal Status

Interval from Prior MR to Diagnostic Pathologic Finding MR Study (mo)

Lesions Not Recognized at the Time of Imaging

Table 2

Stability in size Fibroadenomatosis Benign feature Treated breast Contiguous benign lesion,   stability in size Treated breast Benign feature Treated breast Fibroadenomatosis Fibroadenomatosis

Stability in size Treated breast Treated breast Stability in size Benign feature

Reasons

Small size Small size, background   enhancement marked Preaxillary seat

Small size Preaxillary seat

Reasons

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging Pages et al

45

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

Pages et al

Figure 2

Figure 2:  Lesion potentially misinterpreted, related to stability in size, in a 64-year-old menopausal woman. (a) Transverse contrast-enhanced T1-weighted GRE substracted MR image with retroareolar deep focus (arrow). (b) Transverse contrast-enhanced T1-weighted GRE substracted MR image obtained 6 months later. Right focus is stable in size, which was considered benign on the propective report. (c) Transverse contrast-enhanced T1-weighted GRE substracted image obtained 2 years later shows 20-mm spiculated mass corresponding to an SBR grade 2 IDC.

or even inflammatory breast cancer (11), may not enhance. Other falsenegative results may be due to technical reasons, such as motion artifacts, troubles related to contrast agent injection, or lesion positioned outside the field of view of MR imaging. Furthermore, the relationship between MR imaging sensitivity and parameters related to the patient’s status, including age, BRCA mutation, menopausal status, and breast density, have been investigated (7), with a decrease of MR imaging sensitivity in women with high breast density at mammography. Our study highlights missed breast cancer may result, in part, from pitfalls in MR imaging interpretation. One recent study identified possible sources of diagnostic error in a multicentric trial of MR imaging screening in women with high familial risks (18). In that study, in cases in which malignancy could be retrospectively identified, it was scored it as missed or misinterpreted. We added a third category, “mismanaged,” in patients for whom the lesion was adequately considered as suspicious at MR imaging but for whom biopsy did not succeed in identifying the malignancy. This third category is in accordance with recent results showing failure in biopsy with MR imaging guidance (19) 46

or with targeted US (20). To our knowledge, this is the largest study involving detailed analysis of causes of undiagnosed cancer at MR imaging diagnosis in a relatively large number of breast carcinomas. On the prior MR images, we retrospectively observed 11 foci, of which six were accurately graded as BI-RADS 3 on the prospective report, three were not recognized, and two were classified as definitively benign; 18 masses, of which three were not recognized on the prospective report, eight were classified as definitively benign, and seven were recognized as potentially malignant (BI-RADS 4 or 5) but mismanaged; and five nonmasses that were classified on the prospective report as definitively benign mainly because they were attributed to benign change after conservative therapy. In a multicentric retrospective study analyzing 427 pairs of consecutive mammograms consisting of initial normal screening mammogram and a subsequent finding of cancer at screening, findings were judged visible on the prior mammograms in 67% of patients, including in women with very subtle or benign lesions in the appearance of findings, and missed breast cancer (12,13). Contrary to mammographic studies, we did not try

to differentiate cancer with nonspecific mammographic findings from true diagnostic errors as judged by a majority of reviewers, but instead we attempted to analyze the mechanisms underlying the lack of breast cancer identification. Consequently, because our review was made with full knowledge of subsequent finding and pathologic result, some lesions not recognized or interpreted as benign lesion are not diagnostic errors. However, the analysis of the distribution of the potential reason for misdiagnosed cancer at MR imaging remains valid; of importance, as shown in the prospective report of the prior MR imaging study, enhancement was not identified in only 10% of the lesions, whereas it was potentially misinterpreted or mismanaged in 37% of the lesions. Potential misinterpretation was found to be the leading cause of nondiagnosis of breast cancer. Similar results have been shown in studies focused on the radiologic detection of pulmonary nodules (21,22), which categorized the three major kinds of error as search error (failing to look at the abnormality), recognition error (looking at the abnormality but not identifying it), and decision-making error (identifying the abnormality and then ignoring it), with recognition failure as the leading

radiology.rsna.org  n Radiology: Volume 264: Number 1—July 2012

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

Figure 4

Figure 3

Figure 3:  Lesion potentially misinterpreted, related to benign appareance of enhancement, in 55-year-old menopausal woman with BRCA2 mutation and history of left breast cancer treated by means of surgery and radiation therapy. (a) Transverse contrast-enhanced T1-weighted GRE substracted MR image shows left ovoid mass with smooth margins and nonenhanced internal septations. (b) On transverse contrast-enhanced T1-weighted GRE substracted MR image obtained 1 year later, the mass slightly increased in size and a discrete ductal enhancement appeared. Biopsy was performed with ultrasonographic (US) guidance, and a diagnosis of DCIS without invasive component was made.

type of error, followed by search and decision-making errors. In our study, the three main causes of misinterpretation of a malignant enhancement were the presence of criteria of benignancy,

Pages et al

Figure 4:  Lesion potentially misinterpreted in a treated breast in a 48-year-old menopausal woman with a history of left breast cancer treated by means of surgery and radiation therapy (last course of radiation therapy 10 months earlier). (a, b) Transverse contrast-enhanced T1-weighted GRE substracted image depicts fat necrosis annular enhancement and focal non-masslike enhancement. (c) Transverse contrast-enhanced T1-weighted GRE substracted image obtained 6 months later obtained to control the non-masslike enhancement shows persistent non-masslike enhancement that was considered benign posttherapeutic enhancement on the prospective report. (d) On transverse contrast-enhanced T1-weighted GRE substracted image obtained 1 year later, non-masslike enhancement had increased in size. MR imaging–guided biopsy revealed DCIS.

stability in size, and a past lumpectomy. Although the smooth margin of a mass is the most predictive feature of benignancy (9), carcinomas have been found in 17% of smoothy margin masses (23). This is out of line with the very low likelihood of cancer with smooth margins at US or mammography (24). As noted by Liberman et al (23), this difference could be explained by differences in the spatial resolution of the different examinations but also by differences in histologic correlates at mammography

Radiology: Volume 264: Number 1—July 2012  n  radiology.rsna.org

or US (in which the margin represents the interface between the lesion and the adjacent parenchyma) as compared with MR imaging (in which the margin represents the interface between the area of vascularity and the surrounding tissue). We thus assumed that smooth margins shown on MR images need to be correlated with smooth margins at mammography or US before concluding that a lesion is benign. The presence of nonenhancing internal septations, initially considered as highly predictive 47

48

Pages et al

17 20 Mass Mass 17 18 Mass Mass 17 24 Note.—ILC = invasive lobular carcinoma.

IDC SBR 2 DCIS low grade Familial history 54/Menopausal 52/Premenopausal

Familial history Personal history  (ipsilateral)

Inadequate target biopsy Inadequate target biopsy

Fat necrosis/US and   MR imaging Papilloma/US Fibroadenoma/MR imaging

Benign mastopathy/US No material obtained/US

Inadequate target biopsy Bleeding per biopsy Biopsy denied Inadequate target biopsy 11 10 11 10 Mass Mass Mass Mass 9 8 10 9 Mass Mass Mass Mass 12 6 18 17 IDC SBR 1 IDC SBR 2 IDC SBR 1 ILC SBR 2 54/Menopausal 48/Premenopausal 51/Premenopausal 53/Menopausal

Begnin mastopathy/US Inadequate biopsy target 15 Mass 13 Mass 12 IDC SBR 2 Personal history  (ipsilateral) 52/Menopausal

Size (mm) BI-RADS Feature Size (mm) BI-RADS Feature Patient Age (y)/Hormonal Status

Lesions Mismanaged

Table 4

Risk Factors

Pathologic Finding/Diagnosis

Interval from Prior MR to Diagnostic MR Study (mo)

Prior MR Imaging Study

Diagnosis MR imaging

Reasons

Pathologic Finding/Biopsy Guidance

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

of fibroadenomas (25) may be encountered in breast cancer, as shown in an international breast MR consortium study that included patients with a suspicious clinical or conventional imaging finding, in which eight of 17 lesions rated as having nonenhancing internal septations were cancers (9). The stability of proved breast cancer is a known paradigm. It is mainly encountered in older patients, whereas high-risk younger patients have higherdoubling-time breast cancer (26). However, even in high-risk patients known to have more aggressive tumors, the lesion may remain stable—for example, in the dismissed breast cancers in the MARIBS study (18), two of the 12 misinterpreted breast cancers had a stable size at two consecutive screenings. Stability of a malignant enhancement has been shown by means of MR imaging follow-up of breast lesions initially considered to be benign (20). In a prior study (14), we demonstrated that the breast cancer growth rate was highly variable, with periods of stability. Here we show that the stability of the lesion for at least 6 months of follow-up (range, 6–38.4 months) at MR imaging was the main reason for error in three of the 15 misinterpretation cases. These results confirm the rule that lesions with suspicious findings of malignancy should not be ignored because of their stability. Potentially mismanaged enhancement constituted 25% of the diagnostic errors in our study, with the majority of mismanagement due to inadequate correlations between enhancement at MR imaging and the image used for biopsy at the targeted US. As MR imaging–guided biopsy is expensive, time consuming, and not always available, biopsy guidance with US is preferable if the target lesion can be seen with US, which is the case in about 50% of patients (27,28). However, it is essential to determine whether a finding identified with targeted US definitely corresponds to the concerned MR imaging lesion. A recent study has shown on follow-up images in 80 benign concordant USguided biopsies, the sonographic lesion did not correspond to the MR imaging radiology.rsna.org  n Radiology: Volume 264: Number 1—July 2012

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

Pages et al

Table 5 Size of Lesions and Interval between Prior and Diagnotic MR Imaging Depending on the Type of Error

Parameter Diagnostic MR mean   size (mm) Prior MR mean size (mm) Mean interval –diagnostic   MR (mo)

Cancer Potentially Misinterpreted on the Prospective Report

Cancer Mismanaged on the Prospective Report

8 (3–17)

13.6 (4–30)

13.4 (10–20)

6.5 (3–13) 19.3

9.5 (4–17) 13.2

12 (8–18) 15.1

Cancer Not Recognized on the Prospective Report

Note.—Data in parentheses are the range.

In summary, our study showed that in patients with breast cancer seen at MR imaging, retrospective evaluation of prior MR imaging studies revealed presence of enhancement in almost half the cases, which were more often potentially misdiagnosed or mismanaged than not detected. Disclosures of Potential Conflicts of Interest: E.B.P. No potential conflicts of interest to disclose. I.M. No potential conflicts of interest to disclose. D.H. No potential conflicts of interest to disclose. F.C.D. No potential conflicts of interest to disclose. P.T. No potential conflicts of interest to disclose.

References finding in 10 cases (20). Guidelines must therefore be carefully followed in biopsy guided by means of targeted US for lesions detected initially on MR images: clip placement and follow-up imaging of benign concordant results are recommended to detect cases in which the presumed US correlate is inaccurate to diagnose unsuspected falsenegative biopsies (20). Our study had several limitations. Our data come from a single institution and might not be exportable to other centers; it could be argued that these results are due to weaknesses of our physicians in breast MR image interpretation. However, all of the interpreting radiologists had experience in breast MR imaging and were cautious not to consider all enhancements as suspicious, providing diagnostic processes able to achieve an acceptable positive predictive value. In agreement with Kuhl (16), considering all breast enhancement as potentially malignant is the main impediment to transferring breast MR imaging into clinical practice, just as performing biopsies of all microcalcifications seen at mammography would be deleterious. Our study was a retrospective analysis of prospective interpretation of breast MR imaging. This is both a shortcoming and a strength. The shortcoming was due to the impossibility of prospectively interpreting the different criteria of the MR imaging BI-RADS classification because all of the criteria were not described in the prospective reports. The strength is

that this study is representative of the real daily clinical practice conditions. In a study (29) including a group of patients with bowel obstruction, it has already been demonstrated that the CT accuracy from prospective reports evaluation was lower than that of a centralized retrospective evaluation. Our results show that enhancement may be undergraded in clinical practice. As noted by Weinstein et al (30), the lack of specific descriptors or guidelines for BI-RADS category 3 could explain why some lesions are considered as likely benign despite irregular margins in masses or ductal enhancement in nonmasses, as shown even in the findings of the experienced readers in the ACRIN study (30). Furthermore, the use of BIRADS lesion characteristics is not accurate to predict likehood of malignancy in non-masslike enhancement (31). We did not adopt the methodology used in studies focused on mammographic pairs and we did not mix previous MR imaging studiess with normal MR imaging studies. It is possible that we considered nonspecific findings as diagnostic errors that theoretically do not require patient recall. That is why we have used the terms of “lesions not identified at the time of diagnosis” and “potentially misinterpreted lesions.” Similarly, two of the mismanaged lesions in our study were due to patient refusal of MR imaging–guided biopsy. However, among the 28 tumors with enhancement present on the previous MR images, only five (18%) were foci.

Radiology: Volume 264: Number 1—July 2012  n  radiology.rsna.org

1. Warner E, Messersmith H, Causer P, Eisen A, Shumak R, Plewes D. Systematic review: using magnetic resonance imaging to screen women at high risk for breast cancer. Ann Intern Med 2008;148(9):671–679. 2. Kuhl CK, Schrading S, Leutner CC, et al. Mammography, breast ultrasound, and magnetic resonance imaging for surveillance of women at high familial risk for breast cancer. J Clin Oncol 2005;23(33):8469–8476. 3. Berg WA, Gutierrez L, NessAiver MS, et al. Diagnostic accuracy of mammography, clinical examination, US, and MR imaging in preoperative assessment of breast cancer. Radiology 2004;233(3):830–849. 4. Kuhl CK, Schrading S, Bieling HB, et al. MRI for diagnosis of pure ductal carcinoma in situ: a prospective observational study. Lancet 2007;370(9586):485–492. 5. Mann RM, Hoogeveen YL, Blickman JG, Boetes C. MRI compared to conventional diagnostic work-up in the detection and evaluation of invasive lobular carcinoma of the breast: a review of existing literature. Breast Cancer Res Treat 2008;107(1): 1–14. 6. Warner E, Plewes DB, Hill KA, et al. Surveillance of BRCA1 and BRCA2 mutation carriers with magnetic resonance imaging, ultrasound, mammography, and clinical breast examination. JAMA 2004;292(11): 1317–1325. 7. Kriege M, Brekelmans CT, Obdeijn IM, et al. Factors affecting sensitivity and specificity of screening mammography and MRI in women with an inherited risk for breast cancer. Breast Cancer Res Treat 2006;100(1):109–119. 8. Bluemke DA, Gatsonis CA, Chen MH, et al. Magnetic resonance imaging of the breast prior to biopsy. JAMA 2004;292(22): 2735–2742.

49

BREAST IMAGING: Undiagnosed Breast Cancer at MR Imaging

9. Schnall MD, Blume J, Bluemke DA, et al. Diagnostic architectural and dynamic features at breast MR imaging: multicenter study. Radiology 2006;238(1):42–53. 10. Ghai S, Muradali D, Bukhanov K, Kulkarni S. Nonenhancing breast malignancies on MRI: sonographic and pathologic correlation. AJR Am J Roentgenol 2005;185(2):481–487. 11. Teifke A, Hlawatsch A, Beier T, et al. Undetected malignancies of the breast: dynamic contrast-enhanced MR imaging at 1.0 T. Radiology 2002;224(3):881–888. 12. Birdwell RL, Ikeda DM, O’Shaughnessy KF, Sickles EA. Mammographic characteristics of 115 missed cancers later detected with screening mammography and the potential utility of computer-aided detection. Radiology 2001;219(1):192–202. 13. Ikeda DM, Birdwell RL, O’Shaughnessy KF, Brenner RJ, Sickles EA. Analysis of 172 subtle findings on prior normal mammograms in women with breast cancer de­ tected at follow-up screening. Radiology 2003; 226(2):494–503. 14. Millet I, Bouic-Pages E, Hoa D, Azria D, Taourel P. Growth of breast cancer recurrences assessed by consecutive MRI. BMC Cancer 2011;11:155. 15. American College of Radiology. Breast imaging reporting and data system (BI-RADS). 4th ed. Reston, Va: American College of Radiology, 2003. 16. Kuhl C. The current status of breast MR imaging. Part I. Choice of technique, image interpretation, diagnostic accuracy, and transfer to clinical practice. Radiology 2007; 244(2):356–378.

50

Pages et al

17. Li J, Dershaw DD, Lee CH, Joo S, Morris EA. Breast MRI after conservation therapy: usual findings in routine follow-up exam­inations. AJR Am J Roentgenol 2010; 195(3):799–807. 18. Obdeijn IM, Loo CE, Rijnsburger AJ, et al. Assessment of false-negative cases of breast MR imaging in women with a familial or genetic predisposition. Breast Cancer Res Treat 2010;119(2):399–407. 19. Li J, Dershaw DD, Lee CH, Kaplan J, Morris EA. MRI follow-up after concordant, histologically benign diagnosis of breast lesions sampled by MRI-guided biopsy. AJR Am J Roentgenol 2009;193(3):850–855. 20. Meissnitzer M, Dershaw DD, Lee CH, Morris EA. Targeted ultrasound of the breast in women with abnormal MRI findings for whom biopsy has been recommended. AJR Am J Roentgenol 2009;193(4):1025–1029. 21. Shah PK, Austin JH, White CS, et al. Missed non-small cell lung cancer: radiographic findings of potentially resectable lesions evident only in retrospect. Radiology 2003;226(1):235–241. 22. Turkington PM, Kennan N, Greenstone MA. Misinterpretation of the chest x ray as a factor in the delayed diagnosis of lung cancer. Postgrad Med J 2002;78(917):158–160. 23. Liberman L, Morris EA, Lee MJ, et al. Breast lesions detected on MR imaging: features and positive predictive value. AJR Am J Roentgenol 2002;179(1):171–178. 24. Sickles EA. Nonpalpable, circumscribed, noncalcified solid breast masses: likelihood of malignancy based on lesion size and age of patient. Radiology 1994;192(2):439–442.

25. Nunes LW, Schnall MD, Orel SG, et al. Breast MR imaging: interpretation model. Radiology 1997;202(3):833–841. 26. Tilanus-Linthorst MM, Obdeijn IM, Hop WC, et al. BRCA1 mutation and young age predict fast breast cancer growth in the Dutch, United Kingdom, and Canadian magnetic resonance imaging screening trials. Clin Cancer Res 2007;13(24):7357–7362. 27. Abe H, Schmidt RA, Shah RN, et al. MRdirected (“Second-Look”) ultrasound examination for breast lesions detected initially on MRI: MR and sonographic findings. AJR Am J Roentgenol 2010;194(2):370–377. 28. Demartini WB, Eby PR, Peacock S, Lehman CD. Utility of targeted sonography for breast lesions that were suspicious on MRI. AJR Am J Roentgenol 2009;192(4):1128– 1134. 29. Sheedy SP, Earnest F 4th, Fletcher JG, Fidler JL, Hoskin TL. CT of small-bowel ischemia associated with obstruction in emergency department patients: diagnostic performance evaluation. Radiology 2006; 241(3):729–736. 30. Weinstein SP, Hanna LG, Gatsonis C, Schnall MD, Rosen MA, Lehman CD. Frequency of malignancy seen in probably benign lesions at contrast-enhanced breast MR imaging: findings from ACRIN 6667. Radiology 2010;255(3):731–737. 31. Gutierrez RL, DeMartini WB, Eby PR, Kurland BF, Peacock S, Lehman CD. BI-RADS lesion characteristics predict likelihood of malignancy in breast MRI for masses but not for nonmasslike enhancement. AJR Am J Roentgenol 2009;193(4):994–1000.

radiology.rsna.org  n Radiology: Volume 264: Number 1—July 2012