Wo m e n ’s I m a g i n g • O r i g i n a l R e s e a r c h

Raikhlin et al.
Breast MRI in High-Risk Patients

Women’s Imaging
Original Research

Breast MRI as an Adjunct to

Mammography for Breast Cancer
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved

Screening in High-Risk Patients:

Retrospective Review
Antony Raikhlin1,2 OBJECTIVE. In July 2011, the provincial government of Ontario, Canada, approved
Belinda Curpen 3 funding for the addition of annual breast MRI to mammography screening for all women 30–
Ellen Warner4 69 years old considered to be at high risk for breast cancer. The purpose of this study was to
Carrie Betel 3 evaluate the diagnostic performance of screening breast MRI as compared with mammogra-
Barbara Wright 3 phy in a population-based high-risk screening program.
MATERIALS AND METHODS. A retrospective review identified 650 eligible high-
Roberta Jong 3
risk women who underwent screening breast MRI and mammography between July 2011 and
Raikhlin A, Curpen B, Warner E, Betel C, Wright B, January 2013 at one institution. Results of 806 screening rounds (comprising both MRI and
Jong R mammography) were reviewed.
RESULTS. Malignancy was diagnosed in 13 patients (invasive cancer in nine, ductal car-
cinoma in situ in three [one with microinvasion], and chest wall metastasis in one). Of the 13
cancers, 12 (92.3%) were detected by MRI and four (30.8%) by mammography. In nine of
these patients, the cancer was diagnosed by MRI only, resulting in an incremental cancer de-
tection rate of 10 cancers per 1000 women screened. MRI screening had significantly high-
er sensitivity than mammography (92.3% vs 30.8%) but lower specificity (85.9% vs 96.8%).
MRI also resulted in a higher callback rate for a 6-month follow-up study (BI-RADS category
3 assessment) than mammography (119 [14.8%] vs 13 [1.6%]) and more image-guided biop-
sies than mammography (95 [11.8%] vs 19 [2.4%]).
CONCLUSION. MRI is a useful adjunct to mammography for screening in high-risk
women, resulting in a significantly higher rate of cancer detection. However, this was found
to be at the cost of more imaging and biopsies for lesions that ultimately proved to be benign.
Keywords: BRCA1, BRCA2, breast cancer, breast MRI,
high-risk screening, mammography

DOI:10.2214/AJR.13.12264
Received November 17, 2013; accepted after revision
July 15, 2014.
1
Department of Medical Imaging, St. Michael’s Hospital,
University of Toronto, Toronto, ON, Canada.
2
Present address: Department of Medical Imaging,
Lakeridge Health, 1 Hospital Ct, Oshawa, ON L1G 2B9,
Canada. Address correspondence to A. Raikhlin
(antony.raikhlin@gmail.com)
3
Department of Medical Imaging, Breast Division,
­ unnybrook Health Sciences Centre, University of Toronto,
S [8]. These factors drive the need for an appro-
Toronto, ON, Canada.
4
Department of Medical Oncology, Sunnybrook Health
Sciences Centre, University of Toronto, Toronto, ON, Canada
AJR 2015; 204:889–897
0361–803X/15/2044–889
© American Roentgen Ray Society
W
omen with a predisposing ge-
netic mutation (e.g., BRCA1 or
BRCA2 mutation), those with a
strong family history of breast
cancer, and those who have received radiation
therapy to the chest when younger than 30
years old have a significantly increased cumu-
lative lifetime risk of breast cancer in compar-
ison with the general population (20–65% vs
~11%) [1–7]. Unfortunately, this risk is also
associated with diagnosis at a younger age,
with a substantial proportion of cancers oc-
curring in women younger than 50 years old
priate screening regimen. Although mam-
mography has been shown to be an effective
screening modality in the general population,
particularly in women with mammographi-
cally nondense breasts, it is less effective in
the population of women at higher risk [9, 10].
In part this is due to the increased mammo-
graphic density seen at a younger age [11] and
the inherent biologic differences of BRCA-
linked tumors that may make them mammo-
graphically occult or misread as benign [12,
13]. The faster doubling time of tumors in
younger women and BRCA mutation carriers
(relative to age-matched noncarriers) [14, 15]
means that a small missed tumor will most
likely present as a larger interval cancer be-
fore the next round of screening.
The ability of MRI to detect breast tumors
is largely independent of breast density and
instead relies on the enhancement character-
istics of different tissues. Numerous studies
have shown that breast MRI may be a use-
ful adjunct to mammography for screening of
high-risk patients, with greater cancer detec-
tion rates than mammography alone [5, 16–
31]. In light of these emerging data, begin-
ning in July 2011, the provincial government of
Ontario, Canada began funding annual breast
MRI as an adjunct to mammography for breast
cancer screening of all eligible high-risk On-

AJR:204, April 2015 889


tario women. It was recommended to include
women who received therapeutic chest radia-
tion when younger than 30 years old, in addi-
tion to women with BRCA or other breast can-
cer–predisposing gene mutations and those
with an estimated lifetime risk of breast can-
cer greater than 25% (based predominantly
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved
on personal or family cancer history), who
have constituted the great majority of subjects
in prior MRI screening studies. In contrast to
other high-risk subgroups, this recommenda-
tion was based on expert consensus only, be-
cause the data regarding MRI screening in
these women were at the time insufficient.
In this article, we retrospectively review our
institution’s experience in the first 18 months
with the addition of MRI to mammography in
a population-based high-risk screening pro-
gram. In addition, we compare the diagnostic
performance of the two modalities.
Materials and Methods
Study Population
With research ethics board approval, a retro-
spective review of the radiology department data-
base and electronic patient records was performed to
find high-risk asymptomatic women who underwent
breast cancer screening with MRI and mammog-
raphy at a tertiary center (Sunnybrook Health Sci-
ences Centre, University of Toronto) between July 1,
2011, and January 1, 2013. As per provincial funding
guidelines, eligible women were 30–69 years old and
were assessed as being at high risk for breast cancer
according to the following criteria: was known car-
rier of a deleterious gene mutation (e.g., BRCA1 or
BRCA2); had estimated lifetime risk of breast cancer
greater than 25%, either as an untested first-degree
relative of a mutation carrier or via assessment us-
ing International Breast Cancer Intervention Studies
(IBIS) [32] or Breast and Ovarian Analysis of Dis-
ease Incidence and Carrier Estimation Algorithm
(BOADICEA) [33] breast risk assessment tools; or
received chest radiation when younger than 30 years
old and at least 8 years previously.
Women meeting the aforementioned criteria
for high risk who had a previous breast cancer
(treated with breast conservative surgery or uni-
lateral mastectomy) and were continuing breast
screening were not excluded. Most women had
had prior mammograms, and some had also had
one or more prior diagnostic or screening breast
MRI study in the years preceding initiation of the
screening program. These were available as com-
parison studies for the radiologist interpreting the
screening study. Only screening studies were in-
cluded, and MRI referrals for follow-up of previ-
ously detected findings (e.g., a 6-month or 1-year
follow-up) were not included. Screening rounds
890 AJR:204, April 2015
Raikhlin et al.
in which the interval between mammography and
MRI exceeded 120 days were excluded.
Imaging Studies
Mammography—Conventional four-view full-
field bilateral digital mammograms were ob-
tained. Additional spot magnification views were
acquired when judged to be necessary.
MRI—MRI examinations were performed on a
1.5-T system (Signa, GE Healthcare) using a stan-
dard dedicated bilateral breast coil (Sentinelle Van-
guard, Sentinelle Medical). The MRI protocol was
as follows: localizer axial T1-weighted fast spoiled
gradient-echo imaging (TE/TR, 4.2/150) through
the chest; sagittal T2-weighted fast spin-echo fat-
saturated imaging (TE/TR, 88/3000) through both
breasts; sagittal T1-weighted fast spoiled gradient-
echo imaging (TE/TR, 4.2/9); simultaneous sag-
ittal T1-weighted fat-saturated imaging (TE/TR,
3.2/6.6), including unenhanced and four contrast-
enhanced dynamic runs through both breasts using
Vibrant (volume imaging for breast assessment, GE
Healthcare) (3-mm slices, < 90 s/dynamic run); con-
trast-enhanced axial 3D fast acquisition with mul-
tiphase Efgre3D (enhanced fast gradient-echo 3D)
(FAME) with fat suppression (TE/TR, 1.8/7.8); and
postprocessing imaging, including subtraction, max-
imum-intensity-projection (MIP), and 3D reformat-
ted MIP images. Contrast-enhanced images were
obtained after a bolus injection of 0.1 mmol/kg of
gadodiamide (Omniscan, GE Healthcare). Whenev-
er possible, premenopausal patients were scheduled
in the 2nd week of their menstrual cycle.
Image Interpretation and Workup
All MRI and mammography studies were in-
terpreted by one of seven radiologists in the breast
imaging division, five of whom received fellow-
ship training in breast imaging and had 1–20 years
of experience and the other two of whom had more
than 25 years of experience in breast imaging in-
terpretation (but no fellowship training). Results
were categorized in accordance with the BI-
RADS classification as one of the following [34]:
0, incomplete (additional workup needed); 1, neg-
ative; 2, benign; 3, probably benign; 4, suspicious
(indeterminate but possibly malignant); 5, highly
suggestive of malignancy.
When a lesion was categorized as BI-RADS 3,
a 6-month follow-up was recommended. For some
MRI studies categorized as BI-RADS 3, a targeted
ultrasound was initially recommended to correlate
the MRI finding. If the lesion was seen on ultra-
sound, additional workup depended on ultrasound
characteristics; if it was not seen on ultrasound, fol-
low-up MRI in 6 months was recommended.
Biopsy was recommended for all BI-RADS 4
and 5 lesions, using MRI, ultrasound, or stereo-
tactic guidance. Whenever possible, ultrasound-
guided biopsies were performed for MRI and
mammographically detected findings. MRI-guid-
ed biopsies were performed for sonographically
and mammographically occult lesions.
All ultrasound- and stereotactically guided bi-
opsies were performed using a 14-guage core bi-
opsy system. MRI-guided biopsies were performed
using a 9-guage vacuum-assisted device.
Statistical Analysis
Results from imaging studies that received a fi-
nal assessment of BI-RADS category 4 or 5 were
considered positive findings. All other results
were considered negative findings.
Electronic patient charts were reviewed at least
1 year after a negative result to corroborate the
finding. If there was no interval cancer diagnosed,
results were considered true-negative.
Results that received a final assessment of BI-
RADS category 4 and 5 for which cancer was
confirmed by image-guided biopsy or surgically
excised specimen within 1 year were considered
true-positive. Invasive carcinoma and ductal car-
cinoma in situ (DCIS) were accepted as positive
for malignancy. Biopsy results necessitating sur-
gical excision, such as atypical ductal hyperpla-
sia, were correlated with histopathologic evalua-
tion of the surgical specimen, and if malignancy
was found, were considered true-positive. Other-
wise, all other histopathologic results, including
atypical ductal hyperplasia and lobular carcinoma
in situ, were considered negative for malignancy.
False-negative results were those that received
a final assessment of BI-RADS category 1, 2, or
3—and therefore considered negative—for which
cancer was found within 1 year. False-positive re-
sults were those that received a final assessment of
BI-RADS category 4 or 5 for which no cancer was
found within 1 year.
The diagnostic performance of MRI and mam-
mography was calculated, including sensitivity,
specificity, positive predictive value, negative pre-
dictive value, likelihood ratios, callback rates, and
added cancer yield. Negative studies for which
1-year follow-up did not yet occur were not in-
cluded in the calculation (part of the true-negative
fraction). Comparisons of MRI and mammogra-
phy were performed using the exact binomial test
and Fisher exact test with the aid of a statistician.
Results
A total of 878 screening rounds with MRI
and mammography were identified. Of these,
72 screening rounds were excluded because
the interval between MRI and mammography
exceeded 120 days. (No cancers were found
in the 72 excluded MRI and mammography
 Breast MRI in High-Risk Patients

screening rounds.) Therefore, 806 MRI and
mammography screening pairs in 650 women
were eligible for analysis. The mean number
of screening rounds per patient was 1.24.
The demographic characteristics of the
study participants are summarized in Table 1.
The mean age in our cohort was 45.3 years
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved
MRI), and one patient who did not return for
a biopsy. Pathology revealed seven invasive
ductal carcinomas (three of which were also
seen on mammography), three DCISs, one
chest wall metastatic adenocarcinoma with
features suggestive of a breast primary (in a
BRCA2 mutation–positive patient with a his-
tory of ipsilateral breast and ovarian cancer),
five ADH lesions, and the rest were benign.
The only lesion given a BI-RADS 5 assess-
ment was subsequently biopsied with MRI
guidance and shown to be benign.
A previous MRI was available for com-
parison in 541 of the 806 screening rounds

(range, 30–69 years). A total of 376 women

(57.8%) had had at least one prior breast MRI TABLE 1:  Demographic Characteristics of Study Participants (n = 650)
at our institution that was available for com-
Parameter Result
parison during image interpretation. The av-
erage number of days between mammogra- Age (y) at first screening, mean (range) 45.3 (30–69)
phy and MRI in the same screening round High-risk category
was 16.1 days (range, 0–120 days). In 328 of Deleterious gene mutation 287 (44.2)
the 806 screening rounds (40.7%), the MRI
BRCA1 mutation 141 (21.7)
and mammography were done on the same
day; in 462 (57.3%), mammography preceded BRCA2 mutation 144 (22.2)
MRI; and in 16 (2.0%), MRI preceded mam- PTEN mutation (Cowden syndrome) 2 (0.3)
mography. The interpreting radiologist was > 25% estimated lifetime risk 312 (48.0)
not blinded and had full access to the results
Chest radiation when younger than 30 years olda 51 (7.8)
of any prior imaging investigation—includ-
ing the screening mammogram if it was per- Prior breast cancer 49 (7.5)
formed before MRI, and vice versa. A total of Prior MRI 376 (57.8)
11 cancers were detected on the first round of Screening rounds completed
screening and two on the second round.
First round only 650 (100)
MRI Screening Studies First and second rounds 156 (24)
Of the 806 MRI screening rounds, a fi- Note—Except where otherwise indicated, data are given as no. (%).
aFor treatment of Hodgkin lymphoma in all patients.
nal assessment of BI-RADS category 1 or 2
was made in 608 studies (75.4%), BI-RADS
3 in 119 studies (14.8%), BI-RADS 4 in 78 TABLE 2:  BI-RADS Category Assessments, Callbacks, Biopsies, and Cancers
studies (9.7%), and BI-RADS 5 in one study Detected in MRI and Mammography Screening Rounds (n = 806)
(0.1%) (Table 2). For the 119 BI-RADS 3 as- Parameter MRI Mammography
sessments, an initial targeted ultrasound was
BI-RADS category
recommended in 80 cases (67.2%). This re-
sulted in an ultrasound-guided biopsy in 1 or 2 608 (75.4) 774 (96.0)
eight cases, on all of which the findings 3 119 (14.8)a 13 (1.6)
were benign (five lesions were upgraded to 4 78 (9.7)b 17 (2.1)c
BI-RADS 4 assessment based on ultrasound
5 1 (0.1)d 2 (0.2)e
findings, two were categorized as BI-RADS
3 lesions on ultrasound but the patients opt- No. of biopsies triggered 95 (11.8) 19 (2.4)
ed for biopsy; and in one case, an inciden- MRI guided 68 (71.6)
tal mass was discovered that was unrelated Ultrasound guided 27 (28.4) 13 (68.4)
to the MRI finding). At the 6-month follow-
Stereotactically guided 6 (31.6)
up MRI, seven lesions evolved or were new
(two in one patient) and were biopsied. One Cancers detected
of these revealed an invasive ductal carcino- DCIS or invasive cancer 12 (1.5) 4 (0.5)
ma (counted as a false-negative for the pur- Cancer yieldf 15 5
poses of statistical analysis because the ini-
Note—Except where otherwise indicated, data are shown as no. (%). US = ultrasound, DCIS = ductal
tial MRI finding was “negative”), and the carcinoma in situ.
rest were benign. aInitially 80 patients received callback for ultrasound correlation, eight of whom were upgraded to ultrasound-

The 78 BI-RADS 4 assessments resulted
in 61 MRI-guided biopsies (two-site biopsies
in seven patients), 18 ultrasound-guided bi-
opsies, five attempted MRI-guided biopsies
that were aborted because previously identi-
fied suspicious enhancement was no longer
guided biopsy. At 6-month follow-up, six patients were upgraded to MRI-guided biopsy one to ultrasound-
guided biopsy.
b 61 MRI-guided biopsies (two-site biopsies in seven patients) and 18 ultrasound-guided biopsies triggered. Five
MRI-guided biopsies were aborted because suspicious enhancement no longer seen, and one was canceled
because patient did not return.
cSix stereotactically guided and 11 ultrasound-guided biopsies triggered.
dSingle MRI-guided biopsy triggered.
eTwo ultrasound-guided biopsies triggered.

seen (these were followed up with a 6-month f No. of cancers per 1000 women screened.

AJR:204, April 2015 891


TABLE 3:  Diagnostic Performance of MRI and Mammography in Detecting
Breast Cancer
Parameter MRI
Sensitivity 92.3 (66.7–99.6)
Specificity 85.9 (82.5–88.7)
Positive predictive value 15.2 (8.9–24.7)
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved
Negative predictive value 99.8 (98.6–100)
Positive likelihood ratio 6.54 (4.99–8.59)
Negative likelihood ratio 0.09 (0.01–0.59)
Note—Except where otherwise indicated, data are shown as % (95% CI).
ap < 0.05 was considered threshold of statistical significance.
(67.1%) and unavailable for 265 (32.9%).
This greatly influenced the rate of callbacks
for follow-up imaging or biopsy. The call-
back rate for a 6-month follow-up MRI (BI-
RADS 3) or an image-guided biopsy (BI-
RADS 4 or 5) was significantly lower for
studies with a previous MRI for comparison
(10.2% for BI-RADS 3, 7.4% for BI-RADS 4
and 5 [17.6% cumulative for BI-RADS 3–5]),
as opposed to studies where no such compar-
ison was available (24.2% for BI-RADS 3,
14.7% for BI-RADS 4 and 5 [38.9% cumula-
tive for BI-RADS 3–5]).
Mammography Screening Studies
Of the 806 mammography screening stud-
ies, 707 (87.7%) received an initial assess-
ment of BI-RADS category 1 or 2, and 99
(12.3%) received an initial assessment of
BI-RADS category 0 and therefore required
A B
892 AJR:204, April 2015
Raikhlin et al.
Mammography pa
30.8 (12.7–57.6) 0.0215
96.8 (94.9–98.1) < 0.0001
21.1 (8.5–43.3)
98.1 (96.4–99)
9.74 (3.75–25.33)
0.71 (0.5–1.03)
further workup. Of the latter, 67 were subse-
quently categorized as BI-RADS 1 or 2, 13
as BI-RADS 3, 17 as BI-RADS 4, and two
as BI-RADS 5. The final BI-RADS assess-
ments are given in Table 2.
The 17 BI-RADS 4 assessments resulted
in 11 ultrasound-guided biopsies and six ste-
reotactically guided biopsies. Of these, his-
tology revealed two invasive carcinomas
(one of which was also seen on MRI), and
the rest of the lesions were benign.
Only two BI-RADS 5 assessments were
given. Both resulted in a diagnosis of inva-
sive carcinoma after an ultrasound-guided
biopsy. (Both were also seen on MRI.)
MRI Versus Mammography
MRI resulted in a higher callback rate for a
6-month follow-up study (BI-RADS 3 assess-
ment) than did mammography (119 [14.8%]
vs 13 [1.6%] of studies). Similarly, there was
a much higher rate of MRI-triggered biop-
sies (from pooled BI-RADS 3–5 assessments)
than mammography-triggered biopsies. A to-
tal of 95 biopsies (11.8% of all MRI screening
examinations) were performed as a result of
findings on MRI screening, as compared with
only 19 (2.4%) after mammography screen-
ing examinations. A diagnosis of malignancy
was made in 12 of all MRI biopsies (12.6%),
as compared with four (21.1%) of the mam-
mography-triggered biopsies. The cancer
yield from MRI was 1.5% versus only 0.5%
for mammography (Table 2), resulting in an
incremental cancer detection rate of 10 can-
cers per 1000 women screened for MRI in
comparison with mammography.
Of the 806 screening rounds, 488 were
included in the calculation of sensitivity,
specificity, positive predictive value, nega-
tive predictive value, and likelihood ratios.
The studies that were excluded from analysis
were studies with negative findings for which
1-year follow-up did not yet occur (i.e., such
that the true-negative or false-negative sta-
tus could not be confirmed). These results
are shown in Table 3. The sensitivity of MRI
was 92.3% (95% CI, 66.7–99.6%), and this
was statistically significantly higher than the
30.8% (95% CI, 12.7–57.6%) sensitivity of
mammography. However, the specificity of
MRI (85.9% [95% CI, 82.5–88.7%]) was sig-
nificantly lower than the specificity of mam-
mography (96.8% [95% CI, 94.9–98.1%]).
Fig. 1—41-year-old woman with BRCA1 mutation.
Screening mammogram findings were negative.
A, Mediolateral oblique view from digital
mammography shows only peripherally calcified oil
cyst.
B, Sagittal subtraction maximum-intensity-projection
breast MR image obtained from first dynamic series
shows linear nonmass enhancement (circle) in right
upper outer quadrant. This was biopsied with MRI
guidance and was diagnosed as ductal carcinoma in
situ. Tiny (< 1 mm) focus of microinvasion was found
on subsequent lumpectomy.
 Breast Cancers

Note—US = ultrasound, Stereo = stereotactic; CL = contralateral, IL = ipsilateral, NME = nonmass enhancement, Ca2+ = calcifications, NP = not performed, NA = not applicable, IDC = invasive ductal carcinoma, ILC = invasive
Breast MRI in High-Risk Patients

eSentinel node sampling was attempted, but no axillary lymph nodes were identified; any lymph nodes were likely removed during prior sentinel node sampling as part of workup and treatment of ipsilateral breast cancer 3
A total of 13 cancers (nine invasive cancers,

Status
Lymph
Node

NA

NA
2/4

cSuspicious chest wall enhancement at initial screening MRI, so correlation with chest CT and consideration for biopsy was suggested. CT-guided biopsy of the chest wall was eventually done at 6-month follow-up, and
–
–

–
–

–
–

–
–
three DCISs [one with microinvasion], and one
chest wall metastasis) were found in 13 women.
The characteristics of the patients, imaging find-
Tumor
Grade
(I–III)
ings, and tumor stage are summarized in Table 4.

NA

IIIf
III
III
III

III
II
II

II
II

II
I

I
The mean age of the women was 45.3 years (range,
30–60 years). Only three of these women (23.1%)
Histopathology Size (cm)

had a history of breast cancer. Six women (46.2%)

Tumor

< 0.1
1.5f
NA
2.2

0.3
0.4
0.6
0.5
1.4

1.0
1.6
1.0

1.1
were BRCA1 mutation carriers, three (23.1%)
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved

were BRCA2 mutation carriers, two (15.4%) had

an estimated lifetime risk of breast cancer great-

ACA, breast
er than 25%, and one (7.7%) had a history of pri-

primary
or chest radiation at age 29 years for treatment of

DCISf
DCIS

DCIS
Hodgkin lymphoma.
IDC

IDC

IDC
IDC

IDC
IDC
IDC

IDC
ILC

Overall, MRI diagnosed 12 of the 13 can-

cers (92.3%) detected at screening, whereas
Image-
Guided
Biopsy
Stereo
MRI

MRI

MRI
MRI
MRI
MRI

only four of the 13 cancers (30.8%) were diag-

US

US
US
US

US

CT

dInitial MRI-guided biopsy sample showed only atypical ductal hyperplasia at pathology. At subsequent lumpectomy, microscopic focus of DCIS was diagnosed.
nosed by mammography. Nine of these cancers
(69.2%) were seen on MRI alone (Figs. 1 and 2),
Correlate

one (7.7%) on mammography alone, and three

NP
NP

NP

NP

NP
US

–
+
+
+
–

(23.1%) on both MRI and mammography (Fig.

3). With the exclusion of the chest wall metastatic
Spiculated mass

adenocarcinoma, of the other 11 MRI diagnosed

Mammographic

Distortion; Ca2+

cancers, eight (72.7%) were seen as an enhanc-

Findings

aScored from 1 to 4 as follows: 1, predominantly fatty; 2, scattered fibroglandular densities; 3, heterogeneously dense; 4, very dense.

ing mass on MRI, with an average size of 0.9 cm

–
–

–
–
–
–
–
–

(range, 0.5–2.8 cm). An initial targeted ultra-

Mass
Ca2+
TABLE 4:  Characteristics of Patients With Cancer, Imaging Findings, and Tumor Stage

sound for correlation was performed in seven of

the MRI-detected cancers and was able to identi-
2.8-cm clumped NME

fy the lesion in five of the seven patients (71.4%),

2.8-cm linear NME

1.3-cm linear NME

f Microinvasion (< 0.1 cm) was detected at pathology. However, microinvasive focus was not assigned tumor grade.
MRI Findings

all of whom had an enhancing mass on MRI. The

lobular carcinoma, DCIS = ductal carcinoma in situ, ACA = adenocarcinoma. Dashes denote negative findings.
Chest wall en-

two patients in whom ultrasound correlation was

2.8-cm mass
0.6-cm mass

0.5-cm mass
0.6-cm mass
0.5-cm mass

0.6-cm mass
1.0-cm mass

1.1-cm mass
hancement
–

bInitial screening MRI was categorized as BI-RADS 3; at follow-up MRI at 6 months, cancer was diagnosed.

negative had nonmass enhancement on MRI.

The one cancer that was only diagnosed on mam-
mography was seen as microcalcifications (Fig.
4). All but two of the cancers (84.6%) were diag-
MRI?
Prior

Yes

Yes

Yes
Yes
Yes
Yes
No
No

No

No
No

No
No

nosed in the first round of screening.

The mean size of the invasive cancers was 0.9
Mammographic
Breast Densitya

cm (range, 0.3–1.6 cm). Only one patient had docu-

mented lymph node metastases. In patient 2, the ini-
tial MRI examination was categorized as BI-RADS
3
3
3

3
1

3
1
2
2
3
3
2
3

3 and a 6-month follow-up MRI was recommend-

ed. At this follow-up MRI, a new rapidly enhanc-
ing 0.6-cm mass with delayed washout was found,
Cancer?

Yes (CL)

Yes (IL)

Yes (IL)
Breast
Prior

which proved on pathology to be an invasive can-

No
No
No

No

No

No
No

No
No
No

cer. This was therefore scored as an MRI-detected

cancer but considered a false-negative screening
Screen

result (interval cancer) for the purposes of statis-

1

1
1
2

1
1
1
1
2
1
1
1

tical analysis because the initial MRI assessment

was BI-RADS 3 (considered “negative” in our
Prior chest radiation
High-Risk Category

study). Patient 10 had a complicated history of

> 25% lifetime risk

> 25% lifetime risk

> 25% lifetime risk

BRCA2 mutation

BRCA2 mutation
BRCA2 mutation

BRCA1 mutation
BRCA1 mutation
BRCA1 mutation
BRCA1 mutation

BRCA1 mutation
BRCA1 mutation

both ovarian and breast cancer. The initial screen-

ing MRI showed a suspicious area of enhancement
in the ipsilateral breast and abnormal enhancement
cancer was diagnosed.

in the ipsilateral chest wall. An MRI-guided biop-

sy was recommended for the breast lesion (which
years previously.

was negative), and correlation with chest CT, along

Age
(y)

46

44
43
50

30

38

49
35

58

36
60

59

41

with possible CT-guided biopsy, was suggested for

the chest wall enhancement. This was not initial-
Patient

10c

12e
11d

13

ly performed, but the patient returned for a fol-

2b
1

9
4
5
6
7
8

low-up MRI in 6 months, at which time the chest

AJR:204, April 2015 893

 Raikhlin et al.
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved

A B C
Fig. 2—44-year-old woman with BRCA1 mutation. Screening mammogram findings were negative.
A, Mediolateral oblique view shows heterogeneously dense breast and metallic clip from previous biopsy.
B, Sagittal subtraction maximum-intensity-projection breast MRI obtained from first dynamic series shows large region of mass and clumped nonmass enhancement
(arrows) in right upper breast.
C, ROI (circle) placed on central enhancing component shows washout kinetics. Subsequent MRI-guided biopsy and excision showed high-grade in situ and invasive
ductal carcinoma.

wall enhancement had progressed and CT-
guided biopsy was again recommended. This
was subsequently performed, and histopathol-
ogy revealed metastatic adenocarcinoma with
features most supportive of a primary breast
cancer; however, no definitive breast prima-
ry tumor was detected on follow-up imaging
within 1 year. This was considered a true-pos-
itive MRI-detected malignancy because the
initial screening MRI correctly characterized
the chest wall enhancement as suspicious. For
patient 11, initial MRI-guided biopsy showed
atypical ductal hyperplasia only; however, sub-
sequent lumpectomy specimen revealed a fo-
cus of DCIS. This was therefore scored as an
MRI-detected malignancy.
Discussion
Our retrospective study of 650 women
who underwent 806 MRI and mammography
screening rounds has shown the sensitivity of
MRI (92.0%) to be statistically higher than the
sensitivity of mammography (30.8%); howev-
er, this was at the cost of decreased specificity
(85.9% for MRI vs 96.8% for mammography).
The addition of MRI resulted in an incremen-
tal cancer yield of 10 cancers per 1000 wom-
en screened as compared with mammography.
Our findings are in accordance with mul-
tiple other studies that have investigated the
use of MRI screening in women at high risk
for breast cancer. In 2008, Warner et al. [29]
published a meta-analysis of 11 such prospec-
tive, nonrandomized studies. The sensitivity
of MRI was higher than mammography in all
studies, ranging from 64% to 100%, whereas
the sensitivity of mammography ranged from
32% to 40%. In all but one study, by Kuhl et
al. [20], the specificity of MRI was lower than
that of mammography.
We considered only BI-RADS category
4 and 5 assessments to be positive findings.
However, some studies have also considered
BI-RADS category 3 assessments to be posi-
tive findings [17, 21, 23].
Although limited by its retrospective de-
sign, our study is notably different from most
of the aforementioned major studies in that it
reflects the results of high-risk screening in
an everyday clinical setting, rather than with-
in the confines of a clinical trial, where fac-
tors such as patient recruitment and selection
of test interpreters are carefully manipulated
and optimized. All of the radiologists in our
breast imaging division participated in im-
age interpretation, some of whom were only
in their 1st year of practice; this is in contrast
to the usual practice of employing only expe-
rienced readers in clinical trials [19, 30, 31].
Because this study was part of a provin-
cially funded population-based program, no
payer restraints limited access to the offered
screening. Furthermore, because Canada has
a universal health care system, no financial
barriers existed to receiving appropriate diag-
nostic investigation or treatment as a result of
the screening.
The prevalence of pure DCIS among the 12
primary cancers in our study was only 17%
(25% if the case with microinvasion is includ-
ed), similar to other studies, where its preva-
lence ranged from 8% to 28% [29]. This may
in part be due to the inherently low prevalence
of DCIS in BRCA1 mutations carriers, where
invasion is thought to occur at an early stage

894 AJR:204, April 2015

 Breast MRI in High-Risk Patients
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved

A B C
Fig. 3—46-year-old woman with history of Hodgkin lymphoma treated with chest radiation when 29 years old.
A, Mediolateral oblique view from digital mammography shows heterogeneously dense breast and new spiculated mass (circle) in posterior third.
B, Sagittal subtraction maximum-intensity-projection breast MRI obtained from first dynamic series shows corresponding enhancing mass (circle) in upper breast.
C, Targeted ultrasound image of left breast shows corresponding irregular solid hypoechoic mass (calipers) at 11 o’clock 5 cm from nipple. Mass was subsequently
biopsied with ultrasound guidance and diagnosed as invasive ductal carcinoma.

of cancer development [35, 36]. Regardless,
MRI proved to be much better than mammog-
raphy at detecting in situ disease in our study,
given that all three of the noninvasive cancers
(DCISs) were detected by MRI alone. Of the
eight invasive cancers detected by MRI, 75%
were 1 cm in size or smaller (mean, 0.8 cm
[range, 0.3–1.6 cm]), and only one (12.5%)
was node positive. This is comparable to the
data reported in the meta-analysis by Warner
et al. [29], where more than 50% of the can-
cers detected on MRI were in situ or no larger
than 1 cm, and 12–21% were node positive. In
contrast, 50% of the invasive cancers detect-
ed by mammography in our study were larg-
er than 1 cm, with an average size of 1.2 cm
(range, 0.6–1.6 cm). Prior studies examining
conventional mammography-based screen-
ing in high-risk women with BRCA mutations
showed similarly disappointing results, with
very few cases of DCIS detected, 40–78% of
the invasive tumors being greater than 1 cm in
size, and 20–56% being node positive [9, 10,
27, 37, 38]. The relatively poor performance
of mammography has been attributed to the
younger age of these high-risk women and the
inherent biology of BRCA mutation–positive
tumors. Younger age is associated with higher
average breast density and faster tumor dou-
bling time, whereas BRCA mutation–positive
tumors exhibit a faster growth rate and may
have round and pushing margins that may
lead to their being misread as benign [11–15,
27]. Although there is no proof that the ad-
dition of breast MRI screening improves the
survival rate of high-risk women, the assump-
tion is that earlier detection of in situ disease
and smaller tumors will likely allow bet-
ter outcomes. However, long-term follow-up
studies tracking distant disease-free survival
in high-risk women are needed to corroborate
this assumption.
The majority of the breast cancers in our
study were detected in women with BRCA
mutations. This was not surprising, given
that they are at the highest risk, with a re-
ported lifetime cumulative risk of breast can-
cer as high as 84% in some patients [8].
The majority of MRI-detected cancers in
our study were seen as small enhancing mass-
es, with an average size of 0.9 cm. About half
of these could be correlated with a finding on
a targeted ultrasound and consequently could
be biopsied with ultrasound guidance—a much
faster and cheaper procedure, as well as more
comfortable for the patient. However, about
half the cases necessitated MRI-guided biop-
sy; therefore, the capability to perform MRI-
guided biopsies is essential for any center that
offers MRI breast cancer screening. Nonmass
enhancement was seen less commonly and was
associated with DCIS. As expected, these le-
sions could not be detected on ultrasound.
In addition to women with BRCA mutations
and elevated risk owing to various family his-
tory factors, our study included a subgroup of
women not well studied previously—those
who had received chest radiation when young-
er than 30 years old. The most recent (2007) re-
vision of the American Cancer Society guide-
lines [39] includes breast MRI in the annual
screening of high-risk women; nevertheless,
the recommendation for MRI screening in this
subgroup is based on expert consensus only,
because the data on the results of breast MRI
screening in these women are insufficient. In
our study, 51 such patients were included, and
one of the 13 cancers detected was in a 46-year-
old woman who had received chest radiation
when 29 years old for treatment of Hodgkin
lymphoma. However, the cancer was detected

AJR:204, April 2015 895

 Raikhlin et al.
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved

A B C
Fig. 4—35-year-old woman with BRCA2 mutation. Screening MRI findings were negative.
A and B, Mediolateral oblique view (A) and corresponding magnification view (B) from digital mammography show small new cluster of suspicious pleomorphic
calcifications (circle) in left breast. Subsequent stereotactically guided biopsy of calcifications showed in situ ductal carcinoma, with invasive lobular carcinoma found
on histopathology of excised sample.
C, Sagittal subtraction maximum-intensity-projection breast MR image obtained from first dynamic series shows no suspicious enhancement in region of calcifications
on mammogram.

concurrently on both MRI and mammography,
which in this case does not inform as to the po-
tential efficacy of one modality over the other.
A recent retrospective review by Freitas et al.
[16] reported on 98 women with a history of
chest radiation who were screened with MRI
and mammography and found that the sensi-
tivity of MRI was 92% and the sensitivity of
mammography was 69%, but this difference
was not statistically significant. Further stud-
ies are therefore needed to clarify the utility of
breast MRI screening in this subgroup.
Apart from its higher cost, the main draw-
back of MRI screening is the reduced spec-
ificity, which results in a much higher pro-
portion of callbacks for additional imaging
or follow-up, as well as a higher rate of be-
nign biopsies. However, as was evident in our
study, the callback rate significantly decreas-
es if there is a previous MRI available for
comparison. Logically, this is because any
finding that was suspicious on a prior scan
and determined to be benign would no lon-
ger be considered suspicious on the subse-
quent study. Similar findings were observed
in other studies where the rate of callbacks
from MRI decreased after the first round of
screening [19, 29, 30].
The main limitation of our study is its retro-
spective design. This resulted in irregular time
intervals between the MRI and mammography
examinations performed in the same screening
round and led to the exclusion of some cases
from analysis. Furthermore, we were limited
by our sample number and length of follow-up,
with an average of only 1.24 rounds of screen-
ing per patient. The rates of prevalent versus in-
cident cancers therefore could not be evaluated.
Limited 1-year follow-up for negative studies
led to the exclusion of a significant number of
cases from statistical analysis. Furthermore,
unlike a clinical trial, the radiologists inter-
preted the MRI and mammography studies in
a clinical setting and thus were not blinded to
the results of any prior investigations—includ-
ing the screening mammogram if it was per-
formed before MRI, and vice versa. It could
therefore be argued that for the three cases in
which a cancer was detected on both studies,
the MRI was not truly a “screening” examina-
tion because the radiologist was already look-
ing for an abnormality on the basis of the mam-
mographic findings. However, there could not
have been any bias for the other nine cancers
diagnosed on MRI alone (because the mam-
mographic findings were reported as normal)
or for the lone cancer that was diagnosed on
mammography alone (because the MRI find-
ings were reported as normal).
Finally, there was some heterogeneity in
the interpretation of screening studies, given
that some patients had a prior MRI for com-
parison (performed either in the years be-
fore initiation of screening or during the first
round of screening and available for compar-
ison in the second round) whereas others did
not. As expected, callback rates for addition-
al investigations were significantly lower for
studies that had a previous MRI for compari-
son than for those that did not. This may have
also influenced the overall calculated sensi-
tivity of MRI in our study; however, the sen-
sitivity of 92.0% is in a range that was also
seen in other clinical trials [29].
In addition to the aforementioned uncertain-
ties, the most important of which is the yet-un-
known impact of MRI screening on breast can-
cer mortality, further research will be needed
to evaluate the utility of MRI in other high-risk
groups (e.g., women with a history of a high-
risk lesion on biopsy, those with very dense
breasts, or those with breast cancer at a young
age), as well as to determine the optimal age at
which to begin and end screening and the opti-

896 AJR:204, April 2015

 mal screening interval. These factors may vary
for various high-risk subgroups—on the ba-
sis of their estimated cumulative risk, age, and
breast density—and have yet to be determined.
In conclusion, despite the limitations men-
tioned, our study supports the addition of
breast MRI to mammography in a popula-
tion-based screening program for high-risk
women. Its use has resulted in a significant-
ly higher rate of cancer detection, although
Downloaded from www.ajronline.org by 36.83.94.96 on 03/06/21 from IP address 36.83.94.96. Copyright ARRS. For personal use only; all rights reserved
at the cost of more imaging and biopsies for
lesions that ultimately proved to be benign.
Acknowledgment
For his help with statistical analysis, we
thank George Tomlinson (Dalla Lana School
of Public Health and Department of Medical
Imaging, University of Toronto).
References
1. Antoniou A, Pharoah PD, Narod S, et al. Average
risks of breast and ovarian cancer associated with
BRCA1 or BRCA2 mutations detected in case series
unselected for family history: a combined analysis
of 22 studies. Am J Hum Genet 2003; 72:1117–1130
[Erratum in Am J Hum Genet 2003; 73:709]
2. Bhatia S, Yasui Y, Robison LL, et al. High risk of
subsequent neoplasms continues with extended
follow-up of childhood Hodgkin’s disease: report
from the Late Effects Study Group. J Clin Oncol
2003; 21:4386–4394
3. Collaborative Group on Hormonal Factors in
Breast Cancer. Familial breast cancer: collabora-
tive reanalysis of individual data from 52 epide-
miological studies including 58,209 women with
breast cancer and 101,986 women without the dis-
ease. Lancet 2001; 358:1389–1399
4. Feuer EJ, Wun LM, Boring CC, Flanders WD,
Timmel MJ, Tong T. The lifetime risk of developing
breast cancer. J Natl Cancer Inst 1993; 85:892–897
5. Lord SJ, Lei W, Craft P, et al. A systematic review of
the effectiveness of magnetic resonance imaging
(MRI) as an addition to mammography and ultra-
sound in screening young women at high risk of
breast cancer. Eur J Cancer 2007; 43:1905–1917
6. Milne RL, Osorio A, Cajal TR, et al. The average
cumulative risks of breast and ovarian cancer for
carriers of mutations in BRCA1 and BRCA2 at-
tending genetic counseling units in Spain. Clin
Cancer Res 2008; 14:2861–2869
7. Narod SA, Foulkes WD. BRCA1 and BRCA2: 1994
and beyond. Nat Rev Cancer 2004; 4:665–676
8. Ford D, Easton DF, Stratton M, et al. Genetic het-
erogeneity and penetrance analysis of the BRCA1
and BRCA2 genes in breast cancer families. The
Breast Cancer Linkage Consortium. Am J Hum
Genet 1998; 62:676–689
9. Brekelmans CT, Seynaeve C, Bartels CC, et al.;
Rotterdam Committee for Medical and Genetic
Counseling. Effectiveness of breast cancer sur-
veillance in BRCA1/2 gene mutation carriers and
AJR:204, April 2015 897
women with high familial risk. J Clin Oncol
Breast MRI in High-Risk Patients
2001; 19:924–930
10. Vasen HF, Tesfay E, Boonstra H, et al. Early detec-
tion of breast and ovarian cancer in families with
BRCA mutations. Eur J Cancer 2005; 41:549–554
11. Carney PA, Miglioretti DL, Yankaskas BC, et al.
Individual and combined effects of age, breast
density, and hormone replacement therapy use on
the accuracy of screening mammography. Ann
Intern Med 2003; 138:168–175
12. Chang J, Yang WT, Choo HF. Mammography in
Asian patients with BRCA1 mutations. Lancet
1999; 353:2070–2071
13. Tilanus-Linthorst M, Verhoog L, Obdeijn IM, et
al. A BRCA1/2 mutation, high breast density and
prominent pushing margins of a tumor indepen-
dently contribute to a frequent false-negative
mammography. Int J Cancer 2002; 102:91–95
14. Peer PG, van Dijck JA, Hendriks JH, Holland R,
Verbeek AL. Age-dependent growth rate of pri-
mary breast cancer. Cancer 1993; 71:3547–3551
15. Tilanus-Linthorst MMA, Obdeijn IM, Hop WCJ, et
al. BRCA1 mutation and young age predict fast breast
cancer growth in the Dutch, United Kingdom, and
Canadian magnetic resonance imaging screening tri-
als. Clin Cancer Res 2007; 13:7357–7362
16. Freitas V, Scaranelo A, Menezes R, Kulkarni S,
Hodgson D, Crystal P. Added cancer yield of
breast magnetic resonance imaging screening in
women with a prior history of chest radiation
therapy. Cancer 2013; 119:495–503
17. Hagen AI, Kvistad KA, Maehle L, et al. Sensitivity
of MRI versus conventional screening in the diag-
nosis of BRCA-associated breast cancer in a na-
tional prospective series. Breast 2007; 16:367–374
18. Hartman AR, Daniel BL, Kurian AW, et al. Breast
magnetic resonance image screening and ductal
lavage in women at high genetic risk for breast
carcinoma. Cancer 2004; 100:479–489
19. Kriege M, Brekelmans CT, Boetes C, et al. Effi-
cacy of MRI and mammography for breast-cancer
screening in women with a familial or genetic pre-
disposition. N Engl J Med 2004; 351:427–437
20. Kuhl CK, Schrading S, Leutner CC, et al. Mam-
mography, breast ultrasound, and magnetic reso-
nance imaging for surveillance of women at high
familial risk for breast cancer. J Clin Oncol 2005;
23:8469–8476
21. Leach MO, Boggis CR, Dixon AK, et al.; MAR-
IBS Study Group. Screening with magnetic reso-
nance imaging and mammography of a UK popu-
lation at high familial risk of breast cancer: a
prospective multicentre cohort study (MARIBS).
Lancet 2005; 365:1769–1778
22. Lehman CD, Blume JD, Weatherall P, et al.
Screening women at high risk for breast cancer
with mammography and magnetic resonance im-
aging. Cancer 2005; 103:1898–1905
23. Lehman CD, Isaacs C, Schnall MD, et al. Cancer
yield of mammography, MR, and US in high-risk
women: prospective multiinstitution breast cancer
screening study. Radiology 2007; 244:381–388
24. Liberman L. Breast cancer screening with MRI:
what are the data for patients at high risk? N Engl
J Med 2004; 351:497–500
25. Sardanelli F, Podo F, D’Agnolo G, et al. Multi-
center comparative multimodality surveillance of
women at genetic-familial high risk for breast
cancer (HIBCRIT study): interim results. Radiol-
ogy 2007; 242:698–715
26. Trecate G, Vergnaghi D, Manoukian S, et al. MRI
in the early detection of breast cancer in women
with high genetic risk. Tumori 2006; 92:517–523
27. Warner E. The role of magnetic resonance imag-
ing in screening women at high risk of breast can-
cer. Top Magn Reson Imaging 2008; 19:163–169
28. Warner E, Hill K, Causer P, et al. Prospective
study of breast cancer incidence in women with a
BRCA1 or BRCA2 mutation under surveillance
with and without magnetic resonance imaging. J
Clin Oncol 2011; 29:1664–1669
29. 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;
148:671–679
30. Warner E, Plewes DB, Hill KA, et al. Surveil-
lance of BRCA1 and BRCA2 mutation carriers
with magnetic resonance imaging, ultrasound,
mammography, and clinical breast examination.
JAMA 2004; 292:1317–1325
31. Warner E, Plewes DB, Shumak RS, et al. Com-
parison of breast magnetic resonance imaging,
mammography, and ultrasound for surveillance of
women at high risk for hereditary breast cancer. J
Clin Oncol 2001; 19:3524–3531
32. Tyrer J, Duffy SW, Cuzick J. A breast cancer pre-
diction model incorporating familial and personal
risk factors. Stat Med 2004; 23:1111–1130
33. Antoniou AC, Pharoah PP, Smith P, Easton DF.
The BOADICEA model of genetic susceptibility
to breast and ovarian cancer. Br J Cancer 2004;
91:1580–1590
34. Liberman L, Menell JH. Breast Imaging Report-
ing and Data System (BI-RADS). Radiol Clin
North Am 2002; 40:409–430
35. Armes JE, Venter DJ. The pathology of inherited
breast cancer. Pathology 2002; 34:309–314
36. Sun CC, Lenoir G, Lynch H, Narod SA. In-situ
breast cancer and BRCA 1. Lancet 1996; 348:408
37. Komenaka IK, Ditkoff BA, Joseph KA, et al. The
development of interval breast malignancies in
patients with BRCA mutations. Cancer 2004;
100:2079–2083
38. Scheuer L, Kauff N, Robson M, et al. Outcome of
preventive surgery and screening for breast and
ovarian cancer in BRCA mutation carriers. J Clin
Oncol 2002; 20:1260–1268
39. Saslow D, Boetes C, Burke W, et al. American
Cancer Society guidelines for breast screening
with MRI as an adjunct to mammography. CA
Cancer J Clin 2007; 57:75–89