Warner E, Plewes DB, Hill KA, Causer PA, Zubovits JT, Jong RA, Cutrara MR, DeBoer G, Yaffe MJ, Messner SJ, Meschino WS, Piron CA, Narod SA. Surveillance of BRCA1 and BRCA2 Mutation Carriers With Magnetic Resonance Imaging, Ultrasound, Mammography, and Clinical Breast Examination. JAMA. 2004;292(11):1317-1325. doi:10.1001/jama.292.11.1317
Author Affiliations: Division of Medical Oncology, Department of Medicine (Dr Warner and Mss Hill and Cutrara), Imaging Research and Department of Medical Biophysics (Drs Plewes, DeBoer, and Yaffe, and Mr Piron), Department of Medical Imaging (Drs Causer and Jong), Department of Pathology (Dr Zubovits), Division of Clinical Trials and Epidemiology (Dr DeBoer), Departments of Preventive Oncology and Family and Community Medicine (Dr Messner), and Center for Research in Women's Health (Dr Narod), Sunnybrook and Women's College Health Sciences Centre and University of Toronto; and Department of Genetics, North York General Hospital (Dr Meschino), Toronto, Ontario.
Context Current recommendations for women who have a BRCA1 or BRCA2 mutation are to undergo breast surveillance
from age 25 years onward with mammography annually and clinical breast examination
(CBE) every 6 months; however, many tumors are detected at a relatively advanced
stage. Magnetic resonance imaging (MRI) and ultrasound may improve the ability
to detect breast cancer at an early stage.
Objective To compare the sensitivity and specificity of 4 methods of breast cancer
surveillance (mammography, ultrasound, MRI, and CBE) in women with hereditary
susceptibility to breast cancer due to a BRCA1 or BRCA2 mutation.
Design, Setting, and Participants A surveillance study of 236 Canadian women aged 25 to 65 years with BRCA1 or BRCA2 mutations who underwent
1 to 3 annual screening examinations, consisting of MRI, mammography, and
ultrasound at a single tertiary care teaching hospital between November 3,
1997, and March 31, 2003. On the day of imaging and at 6-month intervals,
CBE was performed.
Main Outcome Measures Sensitivity and specificity of each of the 4 surveillance modalities,
and sensitivity of all 4 screening modalities vs mammography and CBE.
Results Each imaging modality was read independently by a radiologist and scored
on a 5-point Breast Imaging Reporting and Data System scale. All lesions with
a score of 4 or 5 (suspicious or highly suspicious for malignancy) were biopsied.
There were 22 cancers detected (16 invasive and 6 ductal carcinoma in situ).
Of these, 17 (77%) were detected by MRI vs 8 (36%) by mammography, 7 (33%)
by ultrasound, and 2 (9.1%) by CBE. The sensitivity and specificity (based
on biopsy rates) were 77% and 95.4% for MRI, 36% and 99.8% for mammography,
33% and 96% for ultrasound, and 9.1% and 99.3% for CBE, respectively. There
was 1 interval cancer. All 4 screening modalities combined had a sensitivity
of 95% vs 45% for mammography and CBE combined.
Conclusions In BRCA1 and BRCA2 mutation
carriers, MRI is more sensitive for detecting breast cancers than mammography,
ultrasound, or CBE alone. Whether surveillance regimens that include MRI will
reduce mortality from breast cancer in high-risk women requires further investigation.
Women with BRCA1 and BRCA2 mutations who do not undergo prophylactic surgery have a lifetime
risk of breast cancer of up to 85%, with a significantly higher risk of breast
cancer than the general population from age 25 years onward.1 It
has been recommended that breast cancer surveillance for these women include
monthly breast self-examination beginning at 18 to 20 years, semiannual clinical
breast examination (CBE) by a health care professional beginning between 20
and 35 years, and annual mammography beginning from 25 to 35 years.2,3 These recommendations are based on
expert opinion; there has been no prospective study demonstrating a benefit
of surveillance on cancer-specific mortality in this population. Eight randomized
clinical trials of mammography have been conducted in the general population4 but it is expected that fewer than 1% of the study
participants in these trials would have a BRCA mutation.
Nonrandomized observational studies of cohorts of BRCA mutation
carriers undergoing routine mammographic screening have demonstrated that
approximately 50% of the breast tumors are detected at screening and 50% present
as interval cancers between screening mammograms.5,6
Contrast-enhanced magnetic resonance imaging (MRI) of the breast has
been shown to have high sensitivity for detecting early breast cancer, albeit
with lower specificity than mammography, and is not affected by breast density.7 A number of studies have suggested that MRI surveillance
may benefit women at high risk8- 15 but
the sensitivity and specificity of MRI for BRCA mutation
carriers have not been fully evaluated in the screening setting. Ultrasound
of the breast is less practical than mammography for screening the general
population because of its lower specificity and its dependency on operator
experience.16 However, it has been found in
several studies17 that ultrasound is more sensitive
than mammography for screening women with dense breasts, and ultrasound may
be particularly useful for the surveillance of young women at high risk.
To determine the extent to which MRI and ultrasound increase the ability
to detect small breast cancers in BRCA1 and BRCA2 mutation carriers beyond that of mammography and
CBE, and to estimate the sensitivity of this combined screening regimen, we
screened 236 mutation carriers with all 4 modalities on an annual basis for
up to 3 years.
Between November 3, 1997, and March 31, 2003, 236 female BRCA1 and BRCA2 mutation carriers between
25 and 65 years were recruited from familial cancer clinics in southern Ontario
and Montreal, Canada. Women with a past history of unilateral breast cancer
were eligible if the contralateral breast was intact. Pregnant or lactating
women were asked to defer their participation. Women with a history of bilateral
breast cancer, who were currently undergoing chemotherapy, or who were known
to have metastatic disease were excluded. Women who weighed more than 91 kg
were excluded for technical reasons. Participation in the study was offered
to all eligible women in the context of genetic counseling. Women were invited
to contact the study coordinator (K.A.H.) directly if they wished to participate.
Informed consent was obtained from all participants. Preliminary results for
the first round of screening of the first 96 patients were reported in a previous
The study was approved by the institutional review boards of the participating
institutions. Eligible women were invited to begin the screening protocol
after at least 1 year had passed since their last mammogram. The protocol
consisted of an intake questionnaire that included, among other demographic
factors, an open-ended question on maternal and paternal ethnicity (included
because we thought the information might be meaningful), and evaluation by
4 screening modalities: CBE, mammography, screening ultrasound, and MRI. All
4 were performed on the same day at the Sunnybrook campus of the Sunnybrook
and Women's College Health Sciences Centre, Toronto, Ontario. For premenopausal
women, screening was performed during the second week of the menstrual cycle
to minimize the occurrence of breast densities or of enhancing masses related
to cyclical hormonal variation. For women with a past history of breast cancer
who had undergone breast-conserving surgery, bilateral breast screening was
performed. For those women who had undergone unilateral mastectomy, screening
of the contralateral breast was performed. Each imaging study was read and
scored independently by a different radiologist (P.A.C. and R.A.J.) who specialized
in breast imaging. Radiologists were blinded to the results of the CBE. Imaging
was repeated annually and CBE was performed every 6 months.
Physical examination of the breasts and regional lymphatic areas was
performed at 6-month intervals by either a physician or a registered nurse
(E.W., M.R.C., and S.J.M.) experienced in breast examination. Each examination
was coded as normal, suggestive of benign disease, indeterminate, or suspicious
for malignancy. Indeterminate examinations were repeated 3 months later.
Conventional 4-view film/screen mammograms were reviewed by a single
radiologist (P.A.C. or R.A.J.). Further views were performed when judged to
be necessary. Mammograms were scored on a 5-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 should be considered; 5 = highly suggestive of malignancy).19
Simultaneous bilateral MRI was performed by using a 1.5-T magnet (Signa,
General Electric Medical Systems, Milwaukee, Wis). The first 38 patients were
imaged in the first year of the study with a single turn elliptical coil after
a bolus injection of 0.1 mmol/kg of gadolinium-diethylenetriamine
penta-acetic acid (Omniscan, GE Healthcare, Oakville, Ontario). After appropriate
imaging to localize the breast, bilateral 3-dimensional spoiled gradient recalled
images (SPGR) were collected in the coronal plane (repetition time [TR]/echo
time [TE]/flip angle, 12.9 ms/43 ms/20° with 28 slices of 4-6 mm thickness),
preinjection, and for a period of 10 minutes postinjection. The scan time
for each 3-dimensional data set was 90 seconds. For all subsequent scans,
a phased-array coil arrangement was used,20 which
provided sagittal images with a 2.5-fold greater signal-to-noise ratio. The
protocol includes a localizing sequence, sagittal, fat-suppressed T2-weighted
Fast Spin Echo (TR/TE, 4000 ms/102 ms), followed by simultaneous sagittal
imaging of both breasts using dual 3-dimensional sagittal TR-interleaved SPGR
sequences (TR/TE/flip angle, 18.4 ms/4.3 ms/40° from 28 partitions per
breast) as proposed by Greenman et al.21 Imaging
was performed both before and after a rapid intravenous injection of 0.1 mmol/kg
of gadolinium-diethylenetriamine penta-acetic acid.
Each volumetric bilateral acquisition was obtained in 2 minutes 49 seconds.
Slice thickness was 2 to 3 mm, without a gap, using a matrix of 256 ×
256 and a field of view of 18 to 20 cm. Frequency was in the anteroposterior
direction. Preconstrast images were subtracted from postcontrast images to
suppress the fat signal.
In cases in which a potentially suspicious area of enhancement was detected,
a diagnostic MRI scan was performed, which comprised a set of high temporal
and spatial resolution unilateral MRI scans of the suspicious breast. During
the first 2 minutes postinjection, dynamic images were acquired in the form
of 9 adjacent, 2-dimensional SPGR with fat saturation (TR/TE/flip angle, 150
ms/4.2 ms/50°) that allowed dynamic monitoring of tissue enhancement with
a temporal resolution of 20 seconds. This was followed by a single high spatial
resolution 3-dimensional SPGR scan with fat saturation (TR/TE/flip angle/matrix/slice
thickness, 50 ms/4.2 ms/50°/256 × 512/1 mm) that took 7 minutes.
This was then followed up by an additional series of dynamic images to monitor
contrast media washout. The scanning time for the entire set of images was
10 minutes. These images were used to help further characterize the lesion
morphology and to provide kinetic enhancement characteristics for clinical
The MRI results were initially scored according to the preliminary BI-RADS
classification.22 Assessment was based primarily
on morphology, using enhancement kinetics for indeterminate and presumed benign
lesions.23 The criteria included mass vs nonmass
enhancement and symmetry of enhancement. Enhancement patterns were assessed
both qualitatively and quantitatively, using time-signal intensity curves
(including the degree of enhancement and the delayed enhancement pattern).
Cases with a detected abnormality were classified as follows. Cases for which
the diagnosis of a specific benign condition was presumed (the presence of
nonenhancing internal septations in a circumscribed or lobulated mass with
no washout on delayed imaging for fibroadenoma, rim enhancement associated
with an inflamed cyst, or a reniform shaped intramammary lymph node with a
feeding vessel to the hilum) were classified as BI-RADS 2.22,24 Cases
with a focus or a symmetric diffuse enhancement pattern were also classified
as BI-RADS 2.22 Nonmass lesions with features
that were predictive of malignancy (asymmetric clumped enhancement in a ductal,
linear, or segmental distribution) were classified as BI-RADS 4.22,24- 27 Masses
that possessed features that were predictive of malignancy (spiculated or
irregular margins, rim enhancement, irregular shape associated with early
enhancement with washout during delayed phase) were classified as BI-RADS
5. If a mass did not possess features whereby a specific benign condition
could be diagnosed, it was considered to be BI-RADS 4 or 5. Masses that were
believed to be due to fibroadenomas or to intramammary lymph nodes but could
not be confidently classified as such, or asymmetric nonmass enhancements
that did not fall into the previously mentioned categories, were classified
as BI-RADS 3. Women with BI-RADS 3 lesions did not routinely undergo biopsy
but were followed up at 6 months, 1 year, and 2 years after the initial imaging
study. If a lesion resolved, decreased, or remained stable during 2 years,
it was reclassified as BI-RADS 2.
High-resolution ultrasound was performed using a 7.5-MHz transducer
by a technologist supervised by an experienced physician (P.A.C. or R.A.J.)
and blinded to the other imaging studies. The reports were coded using a preliminary
BI-RADS ultrasound lexicon.28 Any solid lesion,
unless benign by criteria established by Stavros et al,29 was
considered to be suspicious enough for cancer to warrant a biopsy. The first
7 patients did not receive ultrasound.
A biopsy was recommended if 1 of CBE, mammogram, MRI examination, or
ultrasound was judged to be suspicious for cancer (BI-RADS categories: 4 or
5 or equivalent). If the MRI screening test was abnormal (BI-RADS: 3, 4, or
5) but no other modality was abnormal, a diagnostic MRI procedure was performed
approximately 4 weeks later. Cases that were suspicious for malignancy on
diagnostic MRI examination (BI-RADS: 4 or 5) proceeded to biopsy. If a lesion
was detected only by MRI, the mammograms were reviewed and a targeted second-look
ultrasound was performed to help guide the biopsy. Any lesion detected by
the targeted ultrasound was correlated with the MRI by lesion morphology,
size, and location to ensure that the ultrasound visible lesion corresponded
with the MRI lesion prior to biopsy.
Core and excisional biopsies were performed under ultrasound or stereotactic
guidance, with the exception of 10 women for whom the abnormality was visualized
by MRI but was not observed with directed ultrasound or mammography. In these
cases an excisional biopsy was performed using an MRI-guided wire localization
technique30,31 similar to that
proposed by Orel et al.32 This device consisted
of 2 fenestrated plates that provide medial-lateral compression of the breast.
Fine needle positioning within these fenestrations (a set of 6 × 8 square,
5/8" apertures) was performed using guide plugs accommodating needles of various
gauge sizes (20, 14, and 9 gauges). These drilled guide holes provided a means
of delivering the needle into the breast through a finite number of positions.
Attached to these compression plates was a pair of fiducial markers positioned
at known locations relative to the fenestrations. This configuration enabled
stereotactic needle delivery with the option of medial or lateral approach,
using computer program assistance to calculate the appropriate fenestration,
guide hole, and needle insertion depth into the tumor for biopsy of wire localization.
Using this apparatus, we have subsequently performed percutaneous MRI-guided
biopsies using 14-gauge and 9-gauge vacuum assisted devices for MRI-only visualized
BI-RADS 4 and 5 lesions as described previously.33- 35 These
cases are not included in this patient cohort.
Prior to biopsy for suspected interval cancers (presenting between scheduled
rounds of imaging and CBE), repeat imaging with mammography, ultrasound, and
MRI was performed.
Patients were followed up annually by questionnaire to determine whether
any cancers had been diagnosed since the last screening interval. All study
participants were followed up for 1 year from the date of the last screening
examination and all new cancers and prophylactic mastectomies were recorded.
Study participants who completed at least 1 screening examination but who
left the study for any reason before completing 3 screening rounds were also
Sensitivity was defined as the number of cancers detected by a given
modality (or combination of modalities) divided by the total number of cancers
detected by all 4 modalities plus interval cancers during the entire 3-year
study period (from the date of first screen to 1 year following the last screen).
Differences in the relative sensitivity of each modality were compared with
Fisher exact test. As sensitivities of overlapping combinations of screening
modalities are not independent when applied to the same patients without blinded
repetition, statistical testing of comparisons between them was not performed.
The positive predictive value (PPV) for each modality was defined as the number
of biopsy-proven cancers as a proportion of the number of suspicious studies
(BI-RADS: 4 or 5 or equivalent) that resulted in a biopsy. Specificity was
defined as the number of true-negative results divided by the sum of true-negative
results and false-positive results (ie, examinations leading to a negative
biopsy). These definitions of PPV and specificity did not include in the denominators
women who had additional diagnostic studies that did not result in a biopsy.
The negative predictive value was defined as the number of true-negative results
as a fraction of the total number of true-negative and false-negative studies.
The change in the benign biopsy rate with successive rounds of screening was
tested for statistical significance using a χ2 test for trend.
Because sensitivity and specificity vary with the chosen operating point (ie,
the minimum BI-RADS score at which a report is considered to be positive),
receiver operating characteristic (ROC) curves were plotted by using Microsoft
Excel (Microsoft Corp, Redmond, Wash) and the corresponding areas under the
ROC curve determined. The CBEs coded as suspicious for malignancy were recoded
as BI-RADS 4 for the purpose of ROC curve construction.
The characteristics of the 236 study participants are listed in Table 1. The mean (range) age of participants
at first screening was 46.6 years (26.4-64.8 years). A total of 205 women
(87%) had a mammogram in the 15 months before starting the study. All women
(100%) completed at least 1 round of screening, 136 (58%) completed at least
2 rounds, and 85 (36%) completed all 3 rounds. A total of 120 women are still
undergoing annual screening. Thirty-one women left the study before completing
all 3 rounds; 16 underwent bilateral mastectomy, 3 were too large to fit into
the MRI machine, 3 stopped their participation due to pregnancy, 4 developed
metastatic cancer, 4 were lost to follow-up, and 1 no longer wished to participate.
A total of 22 cancers (16 invasive and 6 ductal carcinoma in situ [DCIS])
were found in 21 women (1 woman had bilateral cancer). Multicentric cancer
in 1 breast was defined as a single cancer. The characteristics of the patients
with cancer, screening results, and tumor stage are listed in Table 2. The mean (range) age of the 21 women with cancer was 47.4
years (33.4-63.0 years). Seven women (33%) had previous breast cancer. Of
the 22 cancers, 2 (9.1%) were detected by CBE, 8 (36%) by mammography, 7 (33%)
of 21 by ultrasound, and 17 (77%) by MRI. Magnetic resonance imaging was significantly
more sensitive than either mammography (P = .02)
or ultrasound (P = .006).
All patients were followed up for a minimum of 1 year from the date
of the last imaging examination. There was only 1 interval cancer, detected
in a 40-year-old BRCA1 mutation carrier 7 months
after her third screen (patient 59). At the time of her diagnosis, the tumor
was visible with all 3 imaging modalities. In retrospect, it could be observed
on previous screening MRI and mammogram. It is not possible to determine retrospectively
if it was evident on ultrasound. Another woman (patient 323), who elected
to have bilateral mastectomy after her breast cancer was found, had a 2-mm
focus of DCIS in the contralateral breast, which had not been detected 2 months
earlier by any screening modality.
In combination, all 4 screening modalities had a sensitivity of 95%
compared with 45% for mammography and CBE combined. By omitting ultrasound
from the screening regimen, the overall sensitivity decreased from 95% to
86%. The sensitivity of all modalities other than MRI was 64%. The ROC curves
for the entire BI-RADS range are shown in Figure 1. Seven cancers (32%) were detected by MRI but missed by
all other modalities. Two cancers were detected by mammography alone (9.1%)
and 2 detected by ultrasound alone (9.5%). Magnetic resonance imaging detected
9 (75%) of 12 cancers missed by conventional surveillance (mammography and
CBE). Several of these cancers were in patients with relatively fatty breasts
(Figure 2). Ductal carcinoma in
situ accounted for 6 (55%) of 11 cancers in BRCA2 mutation
carriers but there were no cases of DCIS alone (without associated invasion)
among the BRCA1 mutation carriers.
The mean sizes of the invasive cancers on the first and second rounds
of screening were 1.1 cm and 1.3 cm, respectively. For the 3 patients who
had tumor sizes of more than 1 cm during the second round of screening, the
mammogram and MRI study from the previous year were reviewed to determine
whether the tumors could be observed in retrospect. The tumor of patient 87
could not be observed on the mammogram or MRI of the previous year. The tumor
of patient 210 could not be observed on the MRI of the previous year but a
few calcifications were present on the previous mammogram. The invasive tumor
of patient 225 could be observed in retrospect on the MRI of the previous
year, at which time it measured 8 mm.
Seventy of 236 study participants had previous breast cancer. Of the
70 women with previous breast cancer, there were 131 examinations performed
and 6 cancers (4.6%) were detected (3 invasive and 3 DCIS). Of the 166 women
with no previous breast cancer, there were 326 examinations performed and
16 cancers (4.9%) detected (12 invasive and 4 DCIS). Only 2 women had node-positive
breast cancer (9% of all cancers). Both were found on the first round of screening.
All 22 patients who had cancer detected are currently alive and disease-free.
After the first round of screening, 16.5% of participants underwent
a diagnostic MRI scan to clarify the status of an indeterminate or possibly
suspicious lesion. This rate decreased at the second and third rounds of screening
to 9.6% and 7.1%, respectively. For an additional 7.6% of the patients, a
6-month follow-up MRI was recommended for lesions that remained indeterminate.
This rate decreased at the second and third screens to 2.9% and 2.4%, respectively.
In the first round of screening, 5.1% of ultrasound examinations and 0.4%
of mammograms resulted in the recommendation of a 6-month follow-up examination.
A total of 2.1% of CBEs were also thought to be suspicious at the first round
and CBE was repeated 3 months later. The corresponding rates for the second
and third years combined were 3.2% for ultrasound, 0% for mammograms, and
0.4% for CBE.
A total of 34 (14%) of 236 women in the study have had a biopsy for
what proved to be benign disease. Five women (2.1%) have had more than 1 biopsy.
The benign biopsy rate was significantly higher at the first round of screening
(26 [11%] of 236) than at the second (9 [6.6%] of 136) or third round (4 [4.7%]
of 85; P = .05). Only 5.1% of all benign biopsies
were generated by a CBE or a mammography finding; the rest were due to MRI
or ultrasound. The PPV and specificity of MRI and ultrasound by year are given
in Table 3. The PPV of MRI is
consistently higher than that of ultrasound. The specificity of MRI is relatively
low in the first year of screening compared with subsequent years. Of the
32 biopsies necessitated by findings with MRI alone, 22 (69%) could be performed
under directed ultrasound guidance. The most common benign pathological findings
were fibroadenoma (38%), fibrocystic changes (31%), and dense fibrosis (7.7%).
This study of 236 BRCA1 and BRCA2 mutation carriers demonstrates that the addition of annual MRI
and ultrasound to mammography and CBE significantly improves the sensitivity
of surveillance for detecting early breast cancers. The combination of MRI,
ultrasound, and mammography had a sensitivity of 95% compared with 45% for
mammography and CBE. Magnetic resonance imaging alone had a sensitivity of
77%. This study extends our study sample from our previous 96 to 236 BRCA carriers.18 Previously,
we reported the results of the first screening examination only; however,
now we report an average of 1.9 screening examinations per patient.
Only 36% of the breast cancers were detected by mammography. The sensitivity
of mammography is inversely related to breast density and breast density is
much higher, on average, in younger women than in older women.36 Furthermore, BRCA1-related cancers tend to be cellular with round pushing
margins rather than scirrhous with irregular infiltrating margins, resulting
in a more benign mammographic appearance.37BRCA1-associated tumors are also less likely to be associated
with DCIS, which often develops microcalcifications that lead to detection
by mammography, than are nonhereditary cancers or breast cancers in BRCA2 carriers.38,39
The sensitivity of mammography in mutation carriers has been measured
in several retrospective studies. In a small study40 of
Asian patients with palpable breast cancers, only 4 (44%) of 9 tumors detected
in BRCA1 mutation carriers could be observed on the
preoperative mammogram (mean size, 4.1 cm) compared with 18 (95%) of 19 tumors
in age-matched noncarriers (P = .03). Similarly,
in a study of Ashkenazi Jewish women diagnosed with breast cancer younger
than 50 years in Montreal,41 with breast cancers
less than or equal to 2 cm in size, only 2 (25%) of 8 breast cancers could
be observed on the preoperative mammogram of the mutation carriers compared
with 27 (77%) of 35 in noncarriers (P = .009).
Prospective studies of cohorts undergoing mammographic surveillance
have also been discouraging. In a series of 128 BRCA1 and BRCA2 mutation carriers reported by Brekelmans et al,5 9 invasive breast cancers were detected after a median
follow-up of 3 years. Five (56%) of the cancers were node-positive, 7 (78%)
were more than 1 cm in size, and 4 (44%) appeared between rounds of screening
mammography. Scheuer et al6 followed 164 mutation
carriers for a mean of 2 years. Of the 10 cancers detected in women undergoing
conventional surveillance only (mammography and breast self-examination),
2 (20%) were node-positive, 4 (40%) were more than 1 cm in size, and 5 (50%)
were detected between screening mammograms. In contrast with these 2 studies,
the proportion of interval cancers in our study was only 5%.
Although from a single center, our study is the largest study published
to date of women with BRCA1 and BRCA2 mutations and the only one to include 4 screening modalities
performed on the same day. Other preliminary reports of surveillance studies
for high-risk women have also demonstrated better sensitivity of MRI compared
with mammography.8- 15 However,
most previous studies included a relatively small number of patients with
documented BRCA mutations and follow-up periods were
short. In a large screening study of 196 high-risk women, Kuhl et al13 found 36 cancers. The sensitivity of MRI was 95%
vs 34% for mammography and 42% for ultrasound. Morris et al10 screened
367 women at high risk for breast cancer using MRI. Biopsies were recommended
for 64 women and 14 cancers were detected (8 DCIS and 6 invasive). These 2
single center studies were not restricted to BRCA mutations
carriers. A total of 210 BRCA1 and BRCA2 mutations carriers have been included in a multicenter study
from the Netherlands but only preliminary results have been published.11
It is expected that cancers detected at the first round of screening
(prevalent cancers) should be larger and more likely to be lymph-node positive
than cancers detected at subsequent screens (incident cancers). Three (23%)
of 13 cancers detected at the first screen were more than 1 cm in size and
2 had axillary node involvement. Three (43%) of 7 cancers detected at the
second screen were more than 1 cm in size but none were node-positive. The
prevalence of cancer on the second screen (5.1%) was not appreciably lower
than the proportion found to have cancer on the first screen (5.5%). There
are 2 possible explanations for these unexpected findings. Subtle MRI changes
on the first screen could have been missed, either because of relative lack
of experience performing or reading MRI at the beginning of the study or because
no previous MRI scan was available for comparison. Alternatively, the cancers
could have grown so rapidly that, although undetectable at the first screen,
within 1 year had grown beyond 1 cm. If rapid tumor growth is typical of BRCA-associated cancers, screening at 6-month intervals
should be considered. Although the numbers are small, this is not supported
by our year 3 results.
Ductal carcinoma in situ without invasion was only found in the BRCA2 mutation carriers. Other investigators have commented
on the relative lack of DCIS in BRCA1 mutation carriers38,39 and suggest that among BRCA1 mutation carriers invasion occurs at an early stage of tumor
development. Mammography appears to be a valuable adjunct to MRI for BRCA2 carriers because of the high incidence of DCIS in
this subgroup. However, because of concerns about the potential carcinogenicity
of ionizing radiation in younger women in general42 and BRCA1 mutation carriers in particular,43 some
authors44 have suggested omitting mammography
from the surveillance regimen for BRCA1 mutation
carriers younger than 35 years.
To date, the reluctance to use breast MRI for surveillance of high-risk
women outside the context of a clinical trial relates, to a large extent,
to its high cost and relatively low specificity compared with mammography.45 It is encouraging that the MRI recall rates in our
study decreased substantially from 26% on the first round of screening to
13% on the second round and 10% on the third round. Our overall specificity
of 95% and PPV of 46% for MRI is comparable with the results of other investigators
who have reported specificity rates ranging from 88% to 98% and PPVs ranging
from 24% to 71%.8,10,13 Moreover,
our results improved with successive rounds of screening. In all these studies,
the definitions of specificity and PPV are based on biopsy rates rather than
the number of screens requiring further work-up. Two of the cancers in our
study were detected by ultrasound alone. However, including ultrasound in
the protocol resulted in more additional biopsies than MRI after the first
year of screening. Whether the benefit of ultrasound is justified in light
of the high false-positive rate remains to be observed. We did not observe
any benefit from CBE over and above the combination of the 3 imaging modalities.
In conclusion, our results support the position that MRI-based screening
is likely to become the cornerstone of breast cancer surveillance for BRCA1 and BRCA2 mutation carriers,
but it is necessary to demonstrate that this surveillance tool lowers breast
cancer mortality before it can be recommended for general use.