The Role of Magnetic Resonance Imaging in Screening Women at High Risk of Breast Cancer

Abstract

Most women at very high risk of breast cancer because of a mutation in the genes BRCA1 or BRCA2, or a very strong family history of breast cancer, opt for intensive breast screening rather than bilateral prophylactic mastectomy. Annual screening mammography has low sensitivity in this population in part because of the greater breast density and faster tumor growth of younger women, resulting in cancers being detected at a suboptimal stage. In 11 prospective comparative studies, the addition of annual contrast-enhanced magnetic resonance imaging (MRI) of the breast to mammography demonstrated more than 90% sensitivity, more than twice that of mammography alone. False-positive rates were higher with the addition of MRI, but specificity improved on successive rounds of screening. Although survival data are not yet available, the stage distribution of these tumors predicts a significant reduction in breast cancer mortality rate compared with that of screening without MRI. Accordingly, annual MRI plus mammography is now the standard of care for screening women aged 30 years or older who are known or likely to have inherited a strong predisposition to breast cancer (based on the above evidence) and for women who received radiation therapy to the chest before the age of 30 years (based on expert opinion). Further research is necessary to define the optimal screening schedule for different subgroups. Formal studies of other high-risk populations (eg, biopsy showing lobular neoplasia or atypical ductal hyperplasia, dense breasts, and personal history of breast cancer at a young age) should be done before MRI screening is routinely adopted for these women.

Screening mammography has been proven in multiple randomized trials to lower breast cancer mortality rate by 25% to 30% for women in the general population. However, there are groups of women at much higher risk of developing breast cancer for whom screening mammography may not have adequate sensitivity. For many women at very high risk, because of their young age or tumor pathology, mammography is less sensitive than it is for women in the general population. Even for women at very high risk for whom mammographic sensitivity should theoretically be adequate, too many cancers are missed or are detected at a suboptimal stage. This is because, for a screening modality with a given sensitivity, the absolute number of false-negative rates (or cancers detected at a suboptimal stage) will be proportional to the prevalence of cancer in the screened population. In women at very high risk, a more sensitive screening regimen is necessary, even at the expense of reduced specificity. Significantly increased risk of breast cancer may be due to hereditary/familial factors, previous irradiation to the breasts, premalignant breast disease, personal history of breast cancer, and breast density.

HEREDITARY/FAMILIAL RISK

The inherited predisposition to develop breast cancer was first described in 1866 by the French neurosurgeon Dr Paul Broca. Upon observing 14 cases of breast cancer in the 4 generations of his wife's family, he, unfortunately, correctly predicted her eventual demise from breast cancer.¹ After this initial report, similar publications began appearing. These family histories suggested a highly penetrant, autosomal dominant gene. It was not until the Human Genome Project of the 1980s, however, that scientists began to localize the genetic abnormalities responsible for hereditary breast cancer, culminating in the discovery of BRCA1² on chromosome 17 in 1994 and BRCA2 on chromosome 13 in 1995.³ To date, hundreds of mutations have been documented in each of these 2 very large genes, and a mutation in either BRCA1 or BRCA2 is found in approximately 40% of the families with cancer histories suggestive of an inherited predisposition to breast cancer. Women who have been shown by genetic testing to have inherited 1 of these mutations from either parent have a 50% to 85% lifetime risk of breast cancer.^4,5 Moreover, the risk of cancer is substantial from the age of 30 years onward. The prevalence of BRCA mutations is approximately 1 in 500 in the general population but approximately 1 in 50 among women of Jewish ethnicity.⁶ BRCA mutations confer greatly increased risk of other cancers as well, particularly ovarian cancer; however, screening for such cancers is outside the scope of this review.

A variety of additional genes have been identified (eg, p53, and PTEN), which, if inherited in their mutated state, confer a similar high risk of breast cancer, but such mutations are much less common. In most of the families with a strong family history of breast cancer in which a BRCA mutation is not identified, the responsible genetic abnormality cannot be determined, and thus, assuming an autosomal dominant pattern, there is no way of distinguishing the 50% of truly high-risk women from the 50% at average risk. Also, many women decline genetic testing for fear of the psychological effects of a confirmed mutation on themselves or their families, or concern regarding ramifications such as insurance or employment discrimination. (However, with the recent signing of the Genetic Information Nondiscrimination Act, genetic testing may become more widely pursued⁷). For these situations, several models such as BRCAPRO⁸ and BOADICEA⁹ have been developed to estimate the probability of a BRCA mutation and overall breast cancer risk in an individual woman, given a particular personal and family history. Although it is the predisposition to breast cancer that is hereditary in such families, the cancers for which they are at risk are often loosely referred to in the literature as hereditary breast cancer.

MANAGEMENT OPTIONS FOR WOMEN AT RISK OF HEREDITARY BREAST CANCER

One extremely effective option for women at very high risk is bilateral mastectomy, which virtually eliminates any future risk of breast cancer. However, this mutilating procedure is unacceptable to most women with proven BRCA1 or BRCA2 gene mutations.¹⁰ It is even less acceptable to women who decline genetic testing or for those with a strong family history but indeterminate mutation status.¹¹ Furthermore, accumulating evidence for the effectiveness of strategies such as chemoprevention^12,13 and premenopausal oophorectomy^14,15 at reducing risk by approximately 50% makes the case for bilateral mastectomy less compelling for many women. However, with any risk reduction strategy other than bilateral mastectomy, the relative risk of breast cancer compared with that of the general population remains sufficiently high for these women to require a more intensive screening regimen. A recommendation for screening as an alternative to mastectomy can only be justified ethically if most of the tumors can be detected either before invasion (ductal carcinoma in situ [DCIS]) or at an early stage of invasion (node-negative cancers <1 cm in diameter) for which the systemic recurrence rate is less than 10%.¹⁶

MAMMOGRAPHY-BASED SCREENING FOR HEREDITARY BREAST CANCER

Until recently, consensus guidelines for screening women with an inherited predisposition to breast cancer have relied chiefly on mammography, the only screening modality proven to reduce breast cancer mortality rate in any population, as there had never been a screening trial directed at the population at risk of hereditary breast cancer. As late as 2004, the American National Comprehensive Cancer Network recommended monthly breast self-examination starting at the age of 18 years and semiannual clinical breast examination plus annual mammography from the age of 25 years onward. Unfortunately, studies of women with BRCA mutations undergoing conventional mammography-based screening have been extremely disappointing. In 3 prospective^17-19 and 1 retrospective²⁰ series, the interval cancer rate ranged from 35% to 50%, very few cases of DCIS were detected, and 40% to 78% of the invasive cancers were greater than 1 cm in size, with 20% to 56% involving axillary lymph nodes.

The relatively poor performance of mammography in these studies can likely be attributed to the relatively young age of these women and the pathophysiology of BRCA1-related breast cancers. Young age is known to be associated with higher average breast density²¹ and faster tumor doubling time.²² Moreover, a recent study has demonstrated that the doubling time of breast cancers of women with BRCA1 or BRCA2 mutations is higher than that of age-matched women with familial breast cancer but no BRCA mutation.²³ Growth rate in BRCA1-related cancers may be particularly rapid as these cancers are very often high grade.^24,25 Three studies^26-28 have demonstrated that, even among women of a given age group, BRCA1-related cancers are more often either mammographically occult or misread as benign. The authors attributed this to the tendency of these tumors to be cellular and fleshy with round pushing margins (rather than scirrhous with irregular infiltrating margins) and to be associated with less DCIS^24,25 and hence fewer microcalcifications.

With the availability of genetic testing for mutations in BRCA1 and BRCA2 from 1995 onward and mounting evidence for the ineffectiveness of mammographic screening in this population, it became imperative to find a more sensitive imaging modality for screening most women with these mutations who declined prophylactic mastectomy. On the basis of its extremely high reported sensitivity in the diagnostic setting, magnetic resonance imaging (MRI) was the most promising candidate for this role. The effectiveness of MRI for imaging dense breasts and the fact that it did not use ionizing radiation were particularly appealing for screening young women.

MAGNETIC RESONANCE IMAGING SCREENING STUDIES FOR HEREDITARY BREAST CANCER

From the mid-1990s onward, a series of prospective, nonrandomized studies were performed in North America and in Europe for screening women known or likely (based on family history) to have inherited a breast cancer predisposition gene mutation. Because mammography was the only screening modality linked to reduced breast cancer mortality rate in any population, the approach was to add annual MRI (with or without additional modalities) to annual mammography. These studies differed significantly with respect to study size, number of participating centers, patient population, MRI technique, study duration, use of additional modalities, and definition of a positive scan result. Yet, surprisingly, all showed substantially higher sensitivity for MRI compared with mammography, albeit with lower specificity.

We recently performed a meta-analysis of all such studies published up to September 2007.²⁹ Eleven studies were identified.^30-40 All but 2 of the studies, which were restricted to patients with known BRCA mutations, also included patients with various family history criteria but no known mutation. Although median age ranged from 40 to 47 years, the range varied more widely. For example, the 22-center UK study³² included only women between the ages of 35 and 49 years, whereas the single-center Italian study³⁵ included women between the ages of 23 and 81 years. All the included studies used dynamic contrast-enhanced MRI with T1-weighted imaging based on spoiled gradient-recalled MRI, with gadolinium-diethylenetriamine pentaacetic acid as the contrast agent, and obtained multiple sets of postinjection images to provide information about tumor enhancement kinetics. Where provided, image quality varied somewhat because of differences in both hardware (eg, magnets and coils) and software (eg, pulse sequences). In the European studies, which account for most of the data, the imaging was performed in either the axial or the coronal plane, whereas in the North American studies, sagittal imaging was used.

In all the studies, mammography and MRI were conducted within a period of no more than 90 days and were usually on the same day. In several studies, screening ultrasound and/or clinical breast examination were also performed. All but 2 single-center studies specifically reported that each screening modality was assessed independently of the others. In all studies, biopsy-confirmed cancer was considered the definitive positive result for sensitivity calculations. However, some studies considered American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) scores of 3, 4, and 5 to be positive, whereas others considered only BI-RADS 4 and 5 as positive.

The sensitivity of mammography ranged from 32% to 40%, with the exception of the study by Hagen et al⁴⁰ in which it was 59%. Magnetic resonance imaging sensitivity was higher than that of mammography in all studies, ranging from 64% in the large Dutch study³¹ to 100% in the smaller studies. The sensitivity of the combination of MRI and mammography ranged from 80% to 100% in the multicenter studies. In all but the study by Kuhl et al,³⁰ MRI specificity was lower than that of mammography, and specificity of the combined modalities ranged from 73% to 93%. As one would expect, specificity was substantially lower in the studies that scored BI-RADS 3 lesions as positive. In 2 of the 3 studies that reported MRI specificity by round of screening, specificity was considerably higher on the subsequent screens likely because of the fact that false-positive lesions found on the first screen and determined to be benign would no longer be rated as positive on subsequent screens. For example, in the first round of screening in our Toronto study,³³ 16% of participants were called back for a second MRI scan because of the indeterminate results or results suggestive of a lesion, and a further 8% required a 6-month MRI. By year 3, these numbers had dropped to 7% and 2%, respectively. Similarly, the false-positive MRI biopsy rates dropped from 7% in the first year to 2% to 3% in subsequent years. In those studies in which additional screening modalities were included, very few cancers missed by both MRI and mammography were picked up by any of the other modalities.

The results of the meta-analysis are summarized in Table 1. The diagnostic odds ratio is the ratio of the odds of a positive result among those with the disease to the odds of a positive result among those without. Higher diagnostic odds ratios (indicating higher overall test accuracy and discriminatory power) were generally associated with BI-RADS 4+ as the criterion for positivity compared with BI-RADS 3+ and with the combination of MRI and mammography compared with MRI alone or mammography alone.

TABLE 1: Meta-analysis of 11 Prospective Studies of Annual Mammography Plus Breast MRI for Screening High-Risk Women

With 1 exception⁴⁰ in which the results were generally poorer than in those of the other studies, in the studies with more than 1 round of screening, the interval cancer rate was less than 10%. Overall, more than 50% of the cancers detected were either in situ or no larger than 1 cm in diameter; however, 22% of tumors in the Dutch study, the largest of the studies, were more than 2 cm in diameter. Of the invasive cancers, 12% to 21% were node positive. In the few studies for which data were available by year of screening, tumor stage was similar for prevalent and incident screens. No study provided data on relapse-free or overall survival.

Because of the high costs and lower specificity of MRI screening, it would be beneficial to determine whether there are subgroups of women at very high risk for whom mammography alone might be adequate. However, there is no evidence that mammography is adequate for screening for hereditary breast cancer in women older than 50 years, women with BRCA2 mutations⁴¹ (whose tumor pathology is more similar to that of sporadic cancers^24,25), or women with fatty breasts. In our own study, as expected, there was a trend toward greater mammographic sensitivity for invasive cancer in women with fattier breasts compared with that for women with greater breast density (37%-43% vs 8%-12%, P = 0.1). However, even in the subgroup of women with fatty breasts, mammographic sensitivity was still inadequate, as illustrated in Figure 1. The sensitivity of mammography for detecting DCIS was much lower than that of MRI, regardless of breast density.⁴¹

FIGURE 1: Magnetic resonance imaging only detected lesion in the fatty breast of a 63-year-old woman with a BRCA2 mutation. A, A gadolinium contrast-enhanced sagittal T1-weighted 3D fast spoiled gradient recalled fat saturated (50/4.2; flip angle, 50 degrees) image shows a 7-mm heterogeneous lesion with marked early enhancement (arrow). A 7-mm infiltrating duct carcinoma of the left breast was diagnosed at the MRI-guided vacuum-assisted biopsy. B and C, Mammogram (mediolateral oblique and craniocaudal views of the left breast) done on the same day, showing a mostly fatty breast with no evidence of malignancy.

The lower sensitivity of mammography for women in these studies compared with the approximately 50% sensitivity reported when screening women with BRCA mutations using mammography without MRI is likely due to the much smaller size of the invasive cancers detected by MRI. Some of the cancers detected by MRI, although still mammographically occult, would likely have been detected by mammography on a subsequent screen but at a larger size.

MAGNETIC RESONANCE IMAGING SCREENING LEARNING CURVE

There is evidence of a learning curve for radiologists conducting MRI breast screening, with the sensitivity of MRI rising with experience⁴² and the number of benign lesions investigated falling with experience.⁴³ This learning curve appears to be particularly steep for DCIS. In the studies reviewed in the above meta-analysis, the proportion of in situ cancers (DCIS) in the larger single-center studies varied from 8% to 27%, whereas in the 2 largest multicenter studies, the UK³² and Dutch³¹ studies, sensitivity for DCIS was at the lower end of this range, at 17% and 13%, respectively. In the studies with higher overall detection rates for DCIS, most of these noninvasive cases were detected by MRI only, whereas in the studies with lower rates of DCIS, most noninvasive cases were detected by mammography only. In our study, we were initially unable to detect any cases of DCIS with MRI,³⁴ but in our more recent cohort of patients,³³ the sensitivity of MRI for detecting DCIS greatly exceeded that of mammography. Similar results have been reported by others.⁴⁴ A case of a large DCIS that we initially misread as benign is illustrated in Figure 2. The results of the Dutch and UK studies reflect the likely initial effectiveness of MRI screening in a population context, and it is expected that, with training and advances in technology, sensitivity would increase further and approach that reported by experienced single centers.

FIGURE 2: Large mammographically occult DCIS with microinvasion in a 47-year-old woman with no known mutation. A, MRI right breast showing clumped segmental enhancement, which was originally thought to be benign parenchymal enhancement. Breast examination and targeted ultrasound were normal. Ductal carcinoma in situ was found on MRI-guided biopsy. At mastectomy, a 6-cm grade II DCIS with a 2-mm focus of microinvasion was found. B, Normal right mammogram.

DIAGNOSTIC ISSUES

In our study, approximately one third of mammographically occult enhancing breast lesions were visualized with targeted ultrasound. Higher BI-RADS score (4 or 5 vs 3) and mass (vs nonmass) enhancing lesions were predictive of greater likelihood of sonographic visibility. However, sonographic visibility of MRI-detected lesions was not helpful in predicting benign versus malignant histopathologic lesions (Fig. 3). It is therefore essential that centers embarking on breast screening for high-risk women have the expertise to perform MRI-guided biopsies.

FIGURE 3: Magnetic resonance imaging-detected invasive duct carcinoma not seen on targeted ultrasound in a 51-year-old woman with a BRCA1 mutation. Contrast-enhanced sagittal T1-weighted images (A) taken 90 seconds after contrast injection show a tiny early enhancing mass and, (B) taken 5 minutes later, show early central washout. Cancer was found on MRI-guided biopsy. Lumpectomy revealed a 6-mm grade III, estrogen receptor (ER)⁻/progesterone receptor (PR)⁻, sentinel node negative (−ve), invasive duct carcinoma. C, Normal mammogram of the left breast.

MAGNETIC RESONANCE IMAGING SCREENING FOR OTHER HIGH-RISK GROUPS

The abovementioned 11 studies were almost exclusively restricted to women at high risk of breast cancer because of either a known genetic abnormality or a strong family history of breast and/or ovarian cancer. The relative effectiveness of MRI compared with that of mammography has not been studied systematically in other high-risk groups such as women with premalignant breast disease (atypical hyperplasia or lobular carcinoma in situ), personal history of breast cancer (particularly women younger than 50 years), women with very dense breasts, or women who have received previous therapeutic irradiation to the breasts. Because of the very high risk of breast cancer in survivors of Hodgkin disease who received chest irradiation before the age of 30 years,⁴⁵ MRI screening for these women is recommended by many based on expert consensus opinion. Screening with MRI may also be indicated for women with a combination of moderate risk factors that together create a high risk, and models to estimate the aggregate risk of multiple factors have been developed.^46,47

MAGNETIC RESONANCE IMAGING SCREENING RECOMMENDATIONS

Annual MRI, in addition to mammography, is now recommended by the American Cancer Society for screening women with a 20% to 25% or greater lifetime risk of breast cancer from the age of 30 years onward for as long as the woman is in good health.⁴⁸ Similar recommendations, albeit with a slightly higher lifetime risk threshold, have been made by experts in many other countries including Canada, United Kingdom, France, the Netherlands, Germany, and Australia. Because risk of early breast cancer is increased by a family history of early breast cancer,⁴⁹ screening from the age of 25 years onward should be considered for members of families with multiple cases of breast cancer, at least 1 of which has been diagnosed to have breast cancer before the age of 30 years.

MAGNETIC RESONANCE IMAGING SCREENING AND MORTALITY RATE REDUCTION

Although the significantly greater sensitivity of MRI is unquestionable, like any screening intervention, its ultimate clinical effectiveness is dependent on its ability to reduce mortality rate, as opposed to simply increasing lead time. The lack of any published or ongoing randomized trials is unfortunate and likely resulted from the fact that, once preliminary evidence from comparative pilot studies of MRI and mammography was available, randomized studies were no longer considered to be feasible or even ethical.

Because the tumor stage distribution in the 11 studies discussed previously generally seems more favorable than that from the reports for mammography without MRI, it is highly likely that MRI screening does decrease breast cancer mortality rate. This argument is based on the "proof of principle" that the tumor stage shift observed with screening mammography in the general population has resulted in a corresponding reduction in breast cancer mortality rate. It would follow that any intervention that significantly lowers breast cancer stage distribution in the high-risk population should also decrease mortality rate.

A more rigorous test of the effectiveness of MRI screening, however, will be a comparison of the long-term outcomes (relapse and survival rates) of well-matched cohorts of high-risk women who differ only with respect to whether they received MRI screening.

AREAS FOR FURTHER RESEARCH

In addition to the uncertainties already mentioned, there are many unanswered questions about screening high-risk women with breast MRI. The optimal age at which to begin and end screening MRI and the optimal screening interval for various subgroups based on age, risk, and breast density is unknown. Some experts recommend that screening with MRI and mammography be staggered at 6-month intervals to minimize the rate of interval cancers, whereas others advocate doing both screening tests at the same time or within a very short period so that both tests can be interpreted together. It has been suggested that mammography be omitted completely for women younger than 40 years given its lack of known efficacy in this age group in the general population and the theoretical risk of carcinogenesis in younger women, particularly those with mutations in the DNA repair genes BRCA1 or BRCA2. However, there is no evidence to date that early mammography increases breast cancer risk in women with BRCA mutations,⁵⁰ and given the variation between centers in the ability to detect DCIS with MRI, such an approach should be taken with extreme caution.

CONCLUSIONS

Annual MRI plus mammography is now the standard of care for screening women aged 30 years or older who are known or likely to have inherited a strong predisposition to breast cancer and for women who received radiation therapy to the chest before the age of 30 years. Further research is necessary to define the optimal screening schedule for different subgroups, as well as long-term survival rates. Although it is tempting to extrapolate results from hereditary cancer studies to other populations of women at high risk of breast cancer (eg, biopsy showing lobular neoplasia or atypical ductal hyperplasia, dense breasts, and personal history of breast cancer at a young age), studies of these specific populations should be done.