| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Imaging, Diagnosis, Prognosis |
Authors' Affiliations: Departments of 1 Diagnostic Radiology and 2 Gynaecology and Obstetrics, Medical University of Vienna, Vienna, Austria; 3 Department of Medical Imaging, University of Toronto Princess Margaret Hospital, Toronto, Canada
Requests for reprints: Christopher C. Riedl, Department of Diagnostic Radiology, Medical University of Vienna, Währinger Gürtel 18-20, 1090 Vienna, Austria. Phone: 43-1-40400-4819; Fax: 43-1-40400-4898; E-mail: christopher.riedl{at}meduniwien.ac.at.
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
Experimental Design: In this prospective comparison study, women at high risk for breast cancer were offered annual surveillance examinations, consisting of mammography, ultrasound, and MRI, at a single tertiary care breast center. The sensitivity and specificity of each modality was based on the histopathologic evaluation of suspicious findings from all modalities plus the detected interval cancers.
Results: Three hundred and twenty-seven women underwent 672 complete imaging rounds. Of a total of 28 detected cancers, 14 were detected by mammography, 12 by ultrasound, and 24 by MRI, which resulted in sensitivities of 50%, 42.9%, and 85.7%, respectively (P < 0.01). MRI detected not only significantly more invasive but also significantly more preinvasive cancers (ductal carcinoma in situ). Mammography, ultrasound, and MRI led to 25, 26, and 101 false-positive findings, which resulted in specificities of 98%, 98%, and 92%, respectively (P < 0.05). Thirty-five (35%) of these false-positive findings were atypical ductal hyperplasias, lesions considered to be of premalignant character. Nine (26%) of those were detected by mammography, 2 (6%) with ultrasound, and 32 (91%) with MRI (P < 0.01).
Conclusion: Our results show that MRI of the breast improves the detection of invasive cancers, preinvasive cancers, and premalignant lesions in a high-risk population and should therefore become an integral part of breast cancer surveillance in these patients.
Until recently, mammography represented the only accepted and widely recommended breast cancer surveillance modality. However, mammography offers a low sensitivity in young women due to higher glandular breast density (12). In addition, the ionizing radiation that penetrates the breast tissue during mammography is presumed to carry an increased risk for women with a genetic predisposition for breast cancer due to impaired DNA repair mechanisms. In contrast, magnetic resonance imaging (MRI) of the breast is known to have a high sensitivity for breast cancer, independent of breast density, and bears no radiation risk (13). Although the clinical value of MRI of the breast has already long been established for local staging of breast cancer, detection of occult breast cancer, and assessment of patients with inconclusive conventional imaging findings, its feasibility for surveillance of women at high risk for breast cancer, is still a topic of discussion (14). Until recently, guidelines only suggested breast examination and mammography starting at age 30 years, at the latest. Now, a first guideline recommends additional MRI breast cancer screening for high-risk patients based on initial studies on this topic (15).
The purpose of this study was to elucidate the value of MRI compared with the conventional imaging techniques, mammography and ultrasound, in the surveillance of women at high risk for breast cancer and to compare our results with those reported in the literature.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Patient selection. The recruitment for imaging surveillance of breast cancer for women with a positive test result for a germ-line BRCA1 or BRCA2 mutation was initiated in January 1999. In January 2002, additional recruitment was initiated of women with negative or pending test results but who fulfilled the family history criteria that made them eligible for genetic testing at our institution [based on risk assessment following a Claus model (16), modified by the Department of Gynecology at our institution]: (a) three or more relatives on the same side of the family with breast cancer diagnosed before the age of 61 or one relative diagnosed with ovarian cancer at any age, (b) two or more relatives on the same side of the family with breast cancer diagnosed before the age of 51 or one relative diagnosed with ovarian cancer at any age, and (c) a relative with breast cancer diagnosed before the age of 36. In any case, the women had to be a first-degree relative of one of the affected relatives or be one of the affected herself. Women had to be of age 25 or older but could also be included at a younger age because surveillance was recommended to begin at an age 5 years younger than the youngest incidence of breast cancer in that family (17). Pregnant or lactating women were asked to defer their participation. Women who were proved not to be carriers in a family with a proven mutation were excluded because their breast cancer risk was not assumed to be elevated. Further, women who underwent bilateral mastectomy or women who were known to have metastatic disease, as well as women with pacemakers, were excluded from the study. Women with clinical signs of breast cancer at their first visit were also excluded but became eligible to participate 1 year after treatment. Written informed consent was obtained from all participating women before entering the study.
Study protocol. Mammography, ultrasound, and MRI of the breast were offered every 12 months. Eligible women were invited to begin the surveillance protocol at least half a year after their last mammography (although most of the eligible women had been under mammography surveillance before enrollment to the study, none of the women had been under MRI surveillance previously). BRCA-positive women were offered an additional ultrasound examination every 6 months. In case of an incomplete annual surveillance round (i.e., not all three imaging modalities were done), or if a woman did not show up at all, the round was excluded from the comparison analysis of the different modalities, but the woman was reinvited the following year. The various examinations with different modalities were preferably done on the same day and had to be done within a month.
Mammography. Oblique and craniocaudal mammograms (Senographe 2000 until October 2000, thereafter Senographe 2000D, both GE Medical Systems) were reviewed by a single radiologist with 10 years of reading experience. The radiologist was aware of the women's risk factors but was blinded to the results of the other imaging modalities. The diagnosis for each breast was scored on a five-point scale using the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) categories (0 = needs further work-up; 1 = negative; 2 = benign finding; 3 = probably benign finding, short follow-up interval suggested; 4 = suspicious abnormality, biopsy recommended; 5 = highly suggestive of malignancy; ref. 18).
Ultrasound. High-frequency breast ultrasound was done by one of two experienced radiologists of the breast imaging group who were blinded to the results of the other imaging studies, with a high-frequency 12-MHz transducer HDI 5000 (Advanced Technology Laboratories). The reports for each breast were scored on a five-point scale equivalent to the mammographic BI-RADS categories and, after 2004, according to the ultrasound edition of the American College of Radiology BI-RADS atlas (19).
Magnetic resonance imaging. Premenopausal women had their MRI studies scheduled in the 2nd week of their menstrual cycle to minimize nonspecific, hormone-related enhancement of benign breast tissue (20). MRI of the breast was done on a 1-T machine with a regular breast coil (Gyroscan T10-NT, Philips). The MRI protocol consisted of a sagittal T2 sequence (TR/TE/FA 3 ms/0.12 ms/90°) and T1-weighted dynamic sequences (three-dimensional gradient-echo technique): 36 axial images (TR/TE/FA 6.2 ms/3.1 ms/35°) with a 256 x 256 matrix and 3 mm slice thickness were done once before and six times after i.v. injection of 0.1 mmol/kg body weight gadopentetate dimeglumine (Dotarem, Guerbet) at 70-s intervals for 8 min after contrast medium administration.
Breast MRI studies were read by one radiologist with 10 years of reading experience who was blinded to the other surveillance modalities. Criteria for distinguishing between benign and malignant contrast-enhancing lesions were based on both morphology and enhancement kinetics (19, 21–24). Again, the reports for each breast were coded in a pattern equivalent to the mammographic BI-RADS categories and, after 2004, according to the American College of Radiology BI-RADS atlas (19).
Final diagnosis and follow-up. If one or more of the imaging modalities revealed a suspicious finding (BI-RADS categories 4 or 5), diagnostic, needle-localized, open breast biopsy was done (25). This procedure was chosen because, at our institution, referring physicians prefer open biopsy over needle core biopsy for patients at high risk for breast cancer. Imaging guidance for needle localization was usually done with the modality that allowed the best delineation of the lesion. In case of benign findings (BI-RADS categories 1 and 2), results were validated either by the absence or presence of an interval cancer or, in the case of the last surveillance round, based on the findings of all the imaging modalities combined (6). In case of a probably benign finding (BI-RADS category 3), repeated short-term follow-up every 6 months was suggested until the lesion was either recategorized as not suspicious (BI-RADS 1 or 2) or upgraded to BI-RADS 4 or 5 and biopsy was done.
Literature search. Using the MEDLINE database and other available search algorithms, a comprehensive literature search of studies that involved MRI surveillance of women at high risk for breast cancer was done. The search included articles published between January 1995 and April 2007; only articles with original results were included. Articles without enough available information to assess sensitivity and specificity were excluded. Preliminary reports of studies on which subsequent follow-up reports were published were also excluded.
The following data were recorded for each article, as well as for all articles combined: (a) country, author, and year of publication; (b) number of women; (c) mean age; (d) number with prior breast cancer (and percent of total population); (e) number of mutation carriers (and percent of total population); (f) total number of done screens and average number of screens per woman; (g) number of cancers detected; (h) number of interval cancers; (i) sensitivities and specificities for mammography, ultrasound, and MRI, as well as the absolute numbers on which these diagnostic variables were based; (j) number of cancers detected only by MRI versus the total number of detected cancers with respective percentage; (k) number of cancers missed at MRI; (l) number of cancers detected only by mammography versus the total number of detected cancers with respective percentage; (m) number of ductal carcinoma in situ (DCIS) detected only by MRI versus the total number of detected DCIS with the respective percentage; and (n) DCIS missed at MRI.
Statistical analysis. Data were analyzed with the Statistical Package for the Social Sciences statistical package (SPSS Windows, version 13.0; SPSS) and EGRET for Windows (version 2.0.3; Cytel Software Corp.). Nominal data, such as sensitivity and specificity, are presented using percentages. At imaging, BI-RADS categories 1 to 3 were considered negative imaging findings and BI-RADS categories 4 and 5 were considered positive findings. At pathology, lesions were grouped into malignant (in situ, invasive, and metastatic cancer), atypical ductal hyperplasia (ADH), and benign lesions (all other histopathologic findings; ref. 26). ADH was defined as proliferative epithelial lesions that possessed some, but not all, of the features of DCIS (27). In accordance with the traditional literature, and to be able to compare our results with the results of previous articles, in an initial analysis of the diagnostic variables, ADH was considered a false-positive finding at imaging. Further, in consideration of our high-risk study population and recent articles, which state that, similar to DCIS, ADH could be considered a nonobligatory direct precursor of invasive ductal cancer (28–30), a second analysis was done, where ADH was considered an important finding at imaging and, hence, a true-positive result. To compare the performance of the different modalities in terms of detection of DCIS or ADH, the Cochran test was used.
To assess the four diagnostic variables, sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV), for each imaging modality, we used only data from complete surveillance rounds. The sensitivity was defined as the number of cancers detected by a given modality divided by the total number of cancers detected by all modalities plus interval cancers (detected between surveillance rounds) during the entire study. Specificity was defined as the number of true-negative results divided by the sum of true-negative results and false-positive results (i.e., all findings leading to a negative biopsy). The PPV for each modality was defined as the number of biopsy-proven cancers as a proportion of the number of suspicious findings (BI-RADS: 4 or 5 or equivalent) that resulted in a biopsy. The NPV was defined as the number of true-negative results as a fraction of the total number of true-negative and false-negative studies. Because some bilateral findings resulted in different pathologic findings, the analysis was based on the total number of screened breasts (two per examination), not just on the number of examinations. To compare the diagnostic variables between the three different modalities, logistic regression with repeated measures (generalized estimating equation) was used. To compare the diagnostic variables between initial and follow-up rounds for each modality, a 
2 test (linear-by-linear association) was used. A P value of 0.05 was considered to indicate a significant result.
Our analysis was based on the total number of screened breasts. In all previously published studies on surveillance of high-risk patients, accuracies were based on the number of patients screened (3, 4, 6–8). To allow for a comparison of our data with data from the literature, we did an additional statistical analysis that combined the bilateral exams into a single examination, where, in case of bilateral suspicious lesions, the finding with the worse prognosis was used for analysis.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
. Fifteen of the 335 women presented at the first visit with palpable lesions suggestive of breast cancer, for which they were offered appropriate treatment but were temporarily exempted from the study. Seven of these 15 women accepted our invitation to participate in the study after they were treated accordingly; eight did not. Thus, the total study population consisted of 327 women, 22 to 80 years of age (median, 41). During the first phase of the study, from January 1999 to January 2001, 46 women who were positive for a germ-line BRCA1 or BRCA2 mutation were included. After 2001, an additional 281 women who fulfilled the family history criteria were included. At the time of this analysis, 47 of these 281 women proved to be carriers of mutations in the BRCA genes (constituting a total of 93 BRCA mutation carriers); 3 were carriers of mutations in other gene loci, 53 women had negative test results, and 178 women had test results still pending for one or both BRCA genes (see Table 1
). Twenty-seven (8%) of the 327 women were excluded during the course of the study due to bilateral mastectomy. Ten women died during the course of the study, 5 of them due to breast cancer and 5 due to ovarian cancer.
|
|
Suspicious findings. A total of 136 suspicious findings (BI-RADS category 4 or 5) were detected, with at least one of the three modalities in 672 complete surveillance rounds. No suspicious findings were detected at 95 additional 6-month surveillance examinations with ultrasound only. One interval cancer was found between surveillance rounds and one was found at a biopsy of a lesion classified as probably benign. The comparison for the three modalities in this study was therefore based on 1,344 screened breasts in 672 complete annual surveillance rounds, with 136 suspicious imaging findings plus 2 cancers not suspicious at imaging. The 136 suspicious imaging findings were detected in 116 of 672 (16%) rounds in 104 women. In 20 cases, lesions were found in both breasts at the same surveillance round. Multifocal and multicentric lesions in the same breast were considered as one finding. In 12 cases, lesions were found in the same woman at different surveillance rounds.
In two of the 136 cases, a lesion diagnosed as BI-RADS category 4 at MRI could no longer be identified at the time of MRI-guided needle localization. Because these two cases had adequate imaging follow-up over 3 years that remained unsuspicious, these cases did not have to be excluded from comparison of the different modalities but were considered false-positive findings at MRI. The histopathologic results of the 134 remaining cases, plus the one interval cancer and the cancer that was probably benign at imaging, are summarized in Table 1.
Seventy one (52%) of the 136 suspicious findings proved to be benign, 39 (29%) revealed ADH, and 26 (19%) revealed cancer, including 11 (8%) DCIS and 15 (11%) invasive cancers. In addition, one woman without suspicious findings at a complete surveillance round returned after 3 months, reporting she "felt a nonpalpable irritation in the right breast." Open surgery revealed an invasive grade 3 medullary carcinoma of 15 mm in size. No suspicious findings could be found in retrospect on the mammograms or MRI images from 3 months earlier. This case was classified as interval cancer. Another lesion that was operated after it was diagnosed as BI-RADS 3 (probably benign) on MRI and revealed a 5-mm invasive adenocarcinoma was also classified as "false negative" (Table 2 ).
|
Of the 11 DCIS, 5 (45%) were detected with all three modalities, 1 only by mammography (9%), and 5 (45%) only by MRI. This means 5 were detected by ultrasound (45%), 6 by mammography (55%), and 10 (91%) with MRI (P < 0.05).
Mammography, ultrasound, and MRI led to 25, 26, and 101 false-positive findings, which resulted in false-positive rates of 64%, 68%, and 81%, respectively, and in specificities of 98%, 98%, and 92%. The sensitivities and specificities, as well as the NPV and PPV for the three modalities, when ADH was considered a false-positive finding, are listed in Table 3A . Of the 39 lesions with ADH, 10 were detected (26%) with mammography, 4 (10%) with ultrasound, and 36 (92%; P < 0.01) with MRI. Twenty-seven (69%) were detected only by MRI. Thus, MRI did significantly better (P < 0.01) in the diagnosis of ADH, lesions considered to be of premalignant character. The sensitivities and specificities, as well as the NPV and PPV for the three modalities when ADH was considered a worthwhile finding and, therefore, a "true positive," are listed in Table 3B. Under such premises, the sensitivity of mammography and ultrasound dropped from 50% and 43% to 36% and 24%, respectively, and increased for MRI from 86% to 90%.
|
|
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
The biggest concern about using MRI as a surveillance tool is a potentially higher number of false-positive findings, which result in additional stress for already anxious patients as well as in further costs at subsequent biopsy (in addition to the already costly MRI examination). Although, in most studies, the specificity of MRI was indeed lower than that of mammography, in one report, it was higher (Table 4; ref. 4). In our study, the specificity of 92% for MRI was also lower than for mammography and ultrasound, with specificities of 98% each. It should be mentioned, though, that the specificity of MRI increased significantly between the first round and follow-up rounds from 90% to 95%. It is the 95% specificity of the follow-up rounds that should be compared with the specificity of mammography and ultrasound of all rounds because only MRI was added as a new surveillance modality and specificities are usually lowest at the first round of surveillance with a specific modality (31).
In addition, the high false-positive rate in this study was partially caused by an unusually high number of detected lesions classified as ADH. These lesions constituted 35% of all benign findings, and the majority of the lesions were detected by MRI. For an initial statistical analysis of the diagnostic variables, we considered ADH as a benign result because this is the widely accepted approach when analyzing breast surveillance studies. However, recent studies that compared benign tissue, ADH, DCIS, and invasive cancer, using mRNA expression profiling, quantitative real-time PCR analysis, and analysis of allelic imbalance, showed that ADH is a genetically advanced precancerous lesion and that ADH, similar to DCIS, is a nonobligatory direct precursor of invasive ductal cancer (28–30). Further, the behavior of ADH is not predetermined, and its development could still be influenced by preventive measures (i.e., hormone therapy). This emphasizes the need for early detection of precancerous lesions, particularly in this patient population. Standard treatments of ADH, such as surgical resection without chemotherapy or radiotherapy, have been extremely effective in reducing the risk of recurrence (30). Gynecologists at our institution consider the removal of a lesion containing ADH, therefore, as a means of secondary prevention. In addition to the benefit of removing tissue with high potential for malignant transformation, the diagnosis of ADH also aids in further individual risk assessment and might be used by patients as additional information for their decision whether to undergo prophylactic measures, such as hormone therapy, or prophylactic mastectomy. Considering these premises, we did a second statistical analysis of the diagnostic variables, where ADH was considered a true-positive finding. Under such premises, the diagnostic variables of MRI improved substantially compared with mammography and ultrasound (Table 3A and B).
Although the high sensitivity of MRI for invasive cancers is widely accepted, its value for the detection of preinvasive cancers, such as DCIS, is still a matter of debate (13, 14, 21–24). In many studies, MRI revealed a much lower sensitivity in the detection of DCIS lesions than mammography. In our study, 91% of all DCIS (n = 11) were detected by MRI, significantly more than by mammography (55%) or ultrasound (45%). Similar results were recently reported by Kuhl et al. (32) who found a 90% detection rate of DCIS with MRI versus 58% with mammography in 101 cases of DCIS (80% of which were detected at surveillance scans of women at high risk for breast cancer).
In our study, only 2 of 28 (7%) cancers were detected only by mammography. In both cases, they were detected due to microcalcification. The same observation was made by two other studies (4, 5). According to our literature review, 22 (11%) of a total of 203 cancers were detected only by mammography. This means that mammography has a low rate of detecting additional lesions that are not visible by any other modality. Although the remaining studies did not specify the mammographic features of lesions seen only on mammography, it can be assumed that, in most cases, this was due to microcalcifications. Given that for most cases of microcalcifications a single-view mammogram is sufficient for detection, and given the assumed increased vulnerability of the breast tissue of high-risk patients to radiation, it is questionable whether annual two-view mammography screening is appropriate for all women at high risk. It seems more appropriate to decide for each patient individually, depending on breast parenchyma density, age, and concern about radiation sensitivity, whether a single or a double-view annual mammography, only biannual mammography, or even no mammography at all should be done.
In our study, no additional cancers were detected only by ultrasound. According to our literature review, 4 (3%) of a total of 118 cancers were detected only by ultrasound (the total number is less than in the previous paragraph because not all studies included ultrasound for surveillance), 2 of them at the 6-month interval rounds (where only ultrasound was done). Thus, the benefit of ultrasound seems lower than that of MRI and mammography. Thus, the benefit of ultrasound, particularly every 6 months, is questionable.
It could be hypothesized that the higher sensitivity of MRI is due to the fact that, because most of the participating women had already undergone surveillance by mammography and ultrasound before entering the study, and only MRI was added as a new surveillance modality, the detection of mammographically occult cancers by MRI in the first round exaggerates the actual difference in sensitivities between modalities. Nevertheless, similar to previous studies, in this study, MRI was significantly more sensitive not only in the first surveillance round but also in subsequent rounds (3, 7, 33).
A limitation of this study is the lack of appropriate follow-up. Similar to other studies, we have not investigated the absolute sensitivity but rather the relative sensitivity of one imaging modality compared with the others (6). This seemed acceptable because we wanted to assess the benefits or disadvantages of MRI of the breast over mammography or ultrasound. In addition, the absolute sensitivities can be assumed to be only marginally lower because the overall reported incidents of interval cancers in patients screened by MRI are two per thousand (Table 5).
In conclusion, the results of this study suggest that MRI of the breast improves detection not only of invasive malignant lesions but also of preinvasive lesions (DCIS) and premalignant lesions (e.g. ADH) at surveillance of women with a high risk for breast cancer. It is the detection of such early-stage cancers that screening programs aim for because the rationale of an alleged long-term benefit of breast cancer surveillance is that lesions are detected in stages that still allow curative therapy. Further, the detection of preinvasive and premalignant lesions could have major implications in an individual's risk assessment and further patient management. MRI should therefore become an integral part of breast cancer surveillance in these patients. That such measures will have a positive effect, in terms of saved years of life, seems reasonable to extrapolate but remains unproven.
|
Footnotes |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Received 5/23/07; revised 7/18/07; accepted 7/20/07.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. Elsamaloty, M. S. Elzawawi, S. Mohammad, and N. Herial Increasing Accuracy of Detection of Breast Cancer with 3-T MRI Am. J. Roentgenol., April 1, 2009; 192(4): 1142 - 1148. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Meeting Abstracts Online |
