Although a woman's average lifetime risk for developing breast cancer in the United States is one in eight, some women are at higher risk [1]. Risk factors for breast cancer include previous breast cancer, a family history of breast cancer, genetic mutations of BRCA1 or BRCA2 (breast cancer oncogenes), a biopsy-proven diagnosis of atypia, lobular carcinoma in situ or radial scar, and prior mantle radiation for Hodgkin's disease [2-7]. Management options for high-risk women range from close mammographic surveillance to prophylactic bilateral mastectomy [1]. Recently, chemopreventive agents have become available that may reduce the likelihood of breast cancer in women at high risk [8]. However, these interventions do not completely eliminate the risk of death from breast cancer.

Breast MRI has high sensitivity in breast cancer detection, reported to be as high as 94-100%, but lower specificity, 37-97% [9]. The high sensitivity of breast MRI is suitable for a screening test, but the lower specificity is problematic because it results in a large number of benign biopsies with their attendant costs in time, anxiety, and deformity, as well as their economic burden [10]. Few data address the frequency and positive predictive value of biopsy in breast MRI screening [11-16]. Furthermore, previous investigations have been few and most have included women with abnormal mammographic findings in the population undergoing MRI [11, 13, 14, 16]. Our study was undertaken in a larger population to determine the rate and results of biopsy in the first round of breast MRI screening in asymptomatic women with normal mammographic findings who were at a high risk of developing breast cancer.

The standard of care for breast cancer screening of asymptomatic women is mammography. Annual mammographic screening, which begins at 40 years old for most women, should begin earlier in women with specific risk factors [2]. Women with BRCA1 or BRCA2 gene mutations should begin annual mammographic screening at 25-35 years old [3]. Women with a family history of premenopausal breast cancer in a first-degree relative should begin annual mammographic screening at an age that is 10 years younger than her relative was when she developed breast cancer, but not before 25 years old. Women who have had prior breast cancer, atypia, or lobular carcinoma in situ should begin annual mammographic screening after diagnosis. Women who have received mantle radiation for Hodgkin's disease should begin annual mammographic screening 8 years after completion of radiation [2].

Screening mammography can detect breast cancers in high-risk women but has limitations, particularly in women with dense breasts that can obscure a cancerous lesion [21, 22]. Kopans [23] has estimated that if a group of 100 women with breast cancer are screened with mammography and clinical breast examination, 80 of the cancers will be detected, by mammography in 68 women (85%) and by physical examination in 12 (15%). Another 20 cancers will become palpable during the next year and will be diagnosed as interval cancers, which tend to be aggressive. Ancillary screening modalities can detect at least some of these interval cancers early, thereby decreasing breast cancer mortality [23], but such screening also generates additional biopsies of benign lesions, with the associated impact of false-positive results [10].

We present the largest study to date evaluating the use of breast MRI to screen women at high risk of developing breast cancer. Breast MRI can detect cancers that are mammographically and clinically occult [11-15] (Table 5). Of our asymptomatic high-risk women with normal mammographic findings who had breast MRI screening, biopsy was recommended in 17%. Biopsy revealed cancer that was occult to mammography and physical examination in 24% of women who had biopsy and in 4% of women who had breast MRI screening. The 4% prevalence of MRI-detected cancer in high-risk women in our study is comparable to the 2-7% prevalence of cancer detected only at MRI in prior reports [11-16] (Table 5). The 24% positive predictive value of biopsy is within the 18-88% range of positive predictive values for biopsy based on MRI findings in high-risk women [11-16] (Table 5) and within the 20-40% range of positive predictive values for mammographically guided needle localization and surgical excision in the general population [24].

More than half (57%) of our MRI-detected, mammographically occult cancers were DCIS. In prior reports, DCIS has accounted for 0-43% of cancers detected by MRI screening in high-risk women (Table 5). The higher proportion of DCIS among MRI-detected cancers in our study may reflect our interpretation criteria, which emphasize lesion morphology, and our imaging parameters, which emphasize high resolution [18]. Although Ernster et al. [25] have suggested that detection of DCIS by screening may lead to overtreatment of an innocuous disease, published data do not support their contention. Previous investigators have shown that approximately 30% of women with untreated DCIS develop ipsilateral invasive breast cancer at long-term follow-up; among these, more than half develop distant metastases [26-28]. These studies show the potential for untreated DCIS to progress to invasive cancer with its associated morbidity and mortality, and support the usefulness of detecting and treating DCIS.

Our data allow us to suggest subgroups of high-risk women most likely to benefit from breast MRI screening. The positive predictive value of biopsy based on MRI findings was significantly greater in women with a family history of breast cancer (positive predictive value, 32%), particularly among those women who had both a family history and a personal history of breast cancer (positive predictive value, 50%). In the first round of screening, MRI detected an otherwise occult cancer in 4% of all high-risk women in this study, in 6% of women with a family history of breast cancer, in 7% of women who had undergone prior breast-conserving surgery for breast cancer, and in 8% of women who had both a personal history and a family history of breast cancer. Breast MRI screening is likely to have the highest yield in women with both a family history and a personal history of breast cancer, particularly those previously treated with breast conservation.

Women with mutations in the BRCA1 or BRCA2 genes, who constituted 5% of our population, account for approximately 5-10% of women with breast cancer in the general population [3]. Lifetime breast cancer risks of up to 85% have been calculated for women with germline mutations in either BRCA1 or BRCA2, although other studies of less-selected groups have suggested that risks may be lower [3, 29]. Mammography may have lower sensitivity in detecting breast cancer in women with a genetic predisposition to breast cancer than in the general population [11, 30, 31]. Previous studies of women with genetic mutations or strong family histories of breast cancer have shown that MRI is more sensitive than mammography in detecting breast cancer in these women, with reported sensitivities of 33-50% for mammography and 86-100% for MRI [11, 13, 14]. Further study of MRI screening in women with genetic predisposition to breast cancer is needed.

Breast sonography has been suggested as a screening test for high-risk women with dense breasts [32-35]. MRI, however, has several advantages over sonography. The prevalence of cancer at MRI screening in high-risk women is 2-7% [11-15], significantly greater than the four to six per 1000 prevalence of cancer in high-risk women reported in prior studies of screening breast sonography [32, 33, 35]. In studies of high-risk women who have had sonography and MRI in addition to mammography, the 86-100% sensitivity of MRI was greater than the 13-43% sensitivity of sonography [11, 13, 16]. MRI is more sensitive than sonography in the detection of DCIS [36]: DCIS, which accounted for more than half the MRI-screening detected cancers in our study, accounted for only 0-17% of cancers detected at screening sonography in prior reports [11, 32-35]. The 18-88% positive predictive value of biopsy in studies of MRI screening [11-15] is significantly greater than the 7-14% positive predictive value of biopsy in studies of screening sonography [11, 32-35]. Breast sonography does retain some advantages, however, including widespread availability, ready access for biopsy procedures, speed, and lower cost.

Although MRI screening can detect otherwise occult breast cancers in high-risk women, several caveats should be remembered. MRI is expensive, and neither the technique nor the interpretive criteria are standardized; variability may occur in performance and interpretation [9]. MRI may not be feasible in some women, such as those with pacemakers, aneurysm clips, or claustrophobia [37]. Breast MRI screening should be performed in a setting in which it is possible to perform biopsy of lesions detected at MRI only; biopsy systems are commercially available but not yet widely used [17]. Breast MRI may result in benign biopsy findings, with the attendant cost and anxiety [10]: among high-risk women who have breast MRI screening, 3-15% have a subsequent biopsy yielding benign results [11-15]. Our data represent results from screening a high-risk population; MRI screening done on women with lower risk than those reported here might have a lower detection rate. Finally, the impact of MRI detection of cancer on survival has not yet been determined.

In conclusion, in our high-risk population that, to our knowledge, is the largest series to date, breast MRI led to a biopsy recommendation in 17% of women. Cancer was found in 24% of women who had biopsy and in 4% of women who had breast MRI screening. More than half the MRI-detected cancers were DCIS. Kopans [23] has stated that the only definitive proof of benefit of breast cancer screening comes from randomized controlled trials, with death as the end point [23]. Neither our study nor previous reports provide such proof, but trials of breast MRI screening are in progress. Further work, including refinements in technique and interpretation; long-term follow-up; and assessments of sensitivity, specificity, and cost effectiveness, is needed to define the role of breast MRI in screening women at high risk.

We thank Alicia Parlanti, Cathleen Cooper, Jeffrey Radeckje, and Flavia Facio for their work in performing this study. Important technical assistance came from Richard Fischer and Charles Nyman.

Address correspondence to E. A. Morris ([email protected]).