Fax: (416) 586-8695
Added cancer yield of breast magnetic resonance imaging screening in women with a prior history of chest radiation therapy
Article first published online: 5 SEP 2012
Copyright © 2012 American Cancer Society
Volume 119, Issue 3, pages 495–503, 1 February 2013
How to Cite
Freitas, V., Scaranelo, A., Menezes, R., Kulkarni, S., Hodgson, D. and Crystal, P. (2013), Added cancer yield of breast magnetic resonance imaging screening in women with a prior history of chest radiation therapy. Cancer, 119: 495–503. doi: 10.1002/cncr.27771
- Issue published online: 22 JAN 2013
- Article first published online: 5 SEP 2012
- Manuscript Revised: 17 JUL 2012
- Manuscript Accepted: 17 JUL 2012
- Manuscript Received: 29 MAY 2012
- breast cancer;
- magnetic resonance imaging;
- high risk;
- chest radiation
Recommendation for breast magnetic resonance imaging (MRI) screening for women with a prior history of chest radiation is currently based on expert opinion, because existing data are very scant. The objective of this study was to evaluate added cancer yield of screening breast MRI in this population.
A retrospective review identified 98 women with a prior history of chest radiation therapy who had screening mammography and screening MRI performed at the authors' institution between January 2004 and July 2010. Medical records of these patients and results of 558 screening studies (296 mammograms and 262 MRI) were reviewed. Sensitivity, specificity, positive predictive value, negative predictive value, and added cancer yield were calculated.
Malignancy was diagnosed in 13 patients, invasive cancer was diagnosed in 10 patients, and ductal carcinomas in situ was diagnosed in 3 patients. The median latency from completion of radiation to detection of the breast cancer was 18 years (range, 8-37 years). Of the 13 cancers, 12 (92%) were detected by MRI, and 9 (69%) by mammography. For mammography, the sensitivity, specificity, positive predictive value, and negative predictive value were 69%, 98%, 82%, and 95%, respectively; and, for MRI, these values were 92%, 94%, 71%, and 99%, respectively. In 4 of 98 patients, cancer was diagnosed on MRI only, resulting in an incremental cancer detection rate of 4.1% (95% confidence interval, 1.6%-10%).
The current results indicated that MRI is a useful adjunct modality for screening high-risk women who have a prior history of chest radiation therapy, resulting in a 4.1% (4 of 98 women) added cancer detection rate. The authors concluded that both MRI and mammography should be used to screen women in this high-risk group. Cancer 2013. © 2012 American Cancer Society.
An estimated 50,000 to 55,000 women in the United States have received moderate to high-dose chest radiation for pediatric or young adult cancer, and these women are at clinically significant increased risk of breast cancer and breast cancer mortality after they achieve a cure of their primary cancer.1 In this population, the cumulative incidence of breast cancer by ages 40 to 45 years ranges from 12% to 20%.2, 3 This incidence is similar to that reported among women who have a breast cancer susceptibility (BRCA) gene mutation, in whom, by age 40 years, the cumulative incidence ranges from 10% to 19%4 and is significantly higher than in women at the same age in the general population, in whom the cumulative incidence of breast cancer is only 1%.5, 6 Like in the general population, breast cancer outcomes among childhood cancer survivors are strongly associated with disease stage at diagnosis.7, 8
Mammography allows for the early diagnosis of breast carcinoma, and mammographic screening effectively decreases breast cancer mortality in the general population.9 Numerous studies have demonstrated that magnetic resonance imaging (MRI) is superior to mammography for breast cancer screening in high-risk women with a strong family history of breast carcinoma, especially among women with cancer-predisposing mutations in either breast cancer susceptibility gene (BRCA1 or BRCA2).10-21
The American Cancer Society, the European Society of Breast Cancer Specialists, the American College of Radiology, and other groups22-24 recommend annual breast screening MRI supplementary to mammography for women who have a history of radiation to the chest between ages 10 years and 30 years. However, contrary to the MRI screening recommendation for women with BRCA mutations, the recommendation for MRI screening in patients with a previous history of chest radiation therapy is based on expert consensus, because there are insufficient data on the results from breast MRI screening in these women. The objective of the current study was to evaluate the added cancer yield of screening breast MRI over mammography in women with a prior history of chest radiation therapy.
MATERIALS AND METHODS:
With Research Ethics Board approval, a retrospective review of the radiology department database and cancer registry was performed. Imaging studies were performed between January 2004 and July 2010 at 2 tertiary centers (Mount Sinai Hospital and Princess Margaret Hospital) affiliated with the University of Toronto.
We identified 120 asymptomatic women who received a radiation dose ≥15 grays (Gy) for pediatric or young adult cancer and were referred for screening mammography and screening MRI of the breast. Of those women, 22 were excluded because the interval between mammography and MRI studies exceeded 4 months.
Therefore, 98 patients were eligible for an assessment of diagnostic accuracy and cancer yield. The mean age in our cohort of 98 women was 37 ± 9.1 years (range, 19-65 years). Of 98 included women, 65 received chemotherapy in addition to radiation therapy. In total, 558 screening rounds were performed, including 296 mammography screening rounds and 262 MRI screening rounds. The mean number of MRI screening rounds per patient was 2.67 (range, 1-8 MRI screening rounds per patient), and the mean number of mammography screening rounds per patient was 3.02 (range, 1-7 mammography screening rounds per patient).
Mammography examinations were performed using full-field digital mammography (Senographe 2000D; GE Medical Systems, Milwaukee, Wis). In addition to the standard mediolateral oblique and craniocaudal projections, additional and spot magnification views were performed when needed.
MRI examinations were performed on a 1.5-T system (Signa Excite [GE Medical Systems] or Espree or Avanto [Siemens Healthcare, Erlangen, Germany]) and a 3.0-T system (Verio; Siemens Healthcare) with a standard, bilateral, dedicated breast coil (Sentinelle Vanguard; Sentinelle Medical, Inc., Toronto, Ontario, Canada). In total, 30 breast MRI studies (9.7%) were performed on the 3.0-T MRI. The remaining 280 MRI studies (90.3%) were performed on the 1.5-T MRI system.
The MRI scan protocol parameters included precontrast, axial, T1, fast-spin-echo images without fat saturation; T2-weighted images with fat suppression; and dynamic contrast-enhanced (DCE), T1-weighted imaging sequences. The DCE sequence consisted of a precontrast scan and at least 3 postcontrast scans. MRI examinations for premenopausal patients were scheduled in the second week of the menstrual cycle to minimize enhancement of benign breast parenchyma. Breast MRI protocols were compliant with quality standards of the American College of Radiology.
Results from all 613 imaging studies, including 310 MRI examinations and 303 mammography examinations, were categorized in accordance with the Breast Imaging Reporting and Data System (BI-RADS) lexicon25 as 1 of the following: BI-RADS 0, additional workup is recommended; BI-RADS 1, negative study; BI-RADS 2, benign lesion; BI-RADS 3, probably benign; BI-RADS 4, suspicious for malignancy; and BI-RADS 5, highly suspicious. The patient workflow is presented in Figure 1. Examinations that had an initial assessment of BI-RADS 0 received a final assessment of BI-RADS 1 to 5 based on the results from workup imaging tests.
If a lesion was categorized as probably benign (BI-RADS 3), then a short-term imaging follow-up was recommended. Image-guided biopsies were performed for all suspicious lesions (BI-RADS 4 and 5). Image-guided biopsies were performed under ultrasound, mammography or MRI guidance. Ultrasound-guided biopsies were performed for sonographically visible lesions irrespective to the modality that detected the abnormality. MRI-guided biopsies were performed for mammographically and sonographically occult lesions. A retrospective review of imaging studies from all patients who had malignant results also was performed to correlate the imaging findings with pathology results.
For the statistical analyses, we considered results from imaging studies that had a final assessment category of BI-RADS 4 and BI-RADS 5 as positive. All other results were considered negative. All patients who had negative results had at least 1 year of follow-up to validate negative findings.
Histologic results of invasive carcinoma or ductal carcinoma in situ (DCIS) were accepted as positive for malignancy. All other histologic results, including atypical ductal hyperplasia and lobular neoplasia, were categorized as negative for malignancy.
The diagnostic performance (sensitivity, specificity, positive predictive value [PPV], negative predictive value [NPV], and added cancer yield [the additional number of cancers detected by single modality per patient]) of imaging tests was calculated. Comparisons between mammography and MRI were performed using the McNemar test, the Fisher exact test, and the methods described by Moscovittz and Pepe26 and Altman and Bland.27 A P value < .05 was considered an indicator of a statistically significant difference. Analyses were performed using statistical software (SPSS version 17.0, SPSS for Windows 2008; SPSS Inc., Chicago, Ill).
Of 98 patients, 88 (90%) received chest radiation for the treatment of Hodgkin lymphoma; 9 (9%) had received chest radiation for the treatment of non-Hodgkin lymphoma; and, in 1 patient, no relevant history was mentioned in the medical records. Among those who received a known radiation dose (61%), the total chest radiation dose ranged from 15 Gy to 35 Gy, and 47% of patients received a moderate to high dose (≥20 Gy), which is associated with greater potential impact to the breast.23 The mean age at the time of screening rounds was 37 years (range, 19-65 years). The median latency from completion of radiation to screening breast MRI was 13 years (range, 2-34 years).
A BI-RADS category of 1 or 2 was used to assess 242 MRI studies (92.3%) and 282 mammograms (95.3%). Six-month follow-up was recommended in 51 (16.5%) of MRI studies and in 10 (3.3%) of mammograms. Suspicious lesions were reported in 17 MRI studies (6.5%) and in 11 mammograms (3.7%). There were patients who had diagnostic discrepancy between mammography and MRI.
Specifically, in 7 women, mammography was negative (BI-RADS 1-2) and MRI revealed a suspicious lesion (BI-RADS 4-5). All of these patients underwent a biopsy. Four cancers were detected (MRI true-positive), and 3 women had benign biopsies (MRI false-positive).
In 2 women, MRI studies were normal (BI-RADS 1-2), and mammography was positive (BI-RADS 4-5); both women underwent a biopsy. Pathology revealed DCIS in 1 woman and benign results in other woman. Fifty-one MRI studies and 10 mammograms were assessed with a BI-RADS category of 3. Of 61 lesions, 4 evolved on short-term follow-up and were biopsied. Malignancy was diagnosed in of those 2 lesions. Imaging findings led to 19 percutaneous biopsies, including 12 biopsies performed under ultrasound guidance, 5 stereotactic biopsies, and 2 MRI-guided biopsies. Detailed results of MRI and mammography imaging studies for the entire cohort of patients are presented in Figure 2.
Breast malignancy was diagnosed in 13 of 98 patients (14%). Of 13 cancers, there were 3 diagnoses of DCIS and 10 invasive breast cancers. There were no interval cancers; ie, all cancers were detected by screening imaging tests. Two patients had multifocal disease, but none had multicentric or bilateral tumors.
Although a detailed dosimetric analysis was not within the scope of this study, the radiation therapy records were reviewed for the 13 patients who had screen-detected cancers. In 6 women, radiation therapy was delivered at other hospitals, and detailed records could not be obtained. Four patients received 35 Gy to mantle fields plus 20 to 25 Gy to the spleen and para-aortic lymph nodes. One patient received 35 Gy to a mantle field only, 1 patient received 35 Gy to a tangent field encompassing an extranodal mass in the right chest wall, and 1 patient received 20-Gy whole-abdominal radiation therapy with a 15-Gy boost to para-aortic lymph nodes. Radiotherapy simulation and/or treatment films were available for 5 patients. In 4 of these women, the breast cancer arose within the treatment field; whereas, in 1 women, it was unclear whether the cancer arose within or at the edge of the field. It is noteworthy that 2 cancers appeared to arise within the field that was used to treat the spleen (ie, the inferior left breast).
Six of 13 patients who were diagnosed with breast carcinoma had received chemotherapy in addition to radiation therapy when they were treated for lymphoma. The median latency from completion of radiation to detection of breast cancers was 18 years (range, 8-37 years). The median age at cancer diagnosis was 39 years (range, 29-65 years).
Of the 13 cancers, 12 (92%) were detected by MRI, and 9 (69%) were detected by mammography. MRI alone was able to detect 4 (31%) of the breast cancer that were diagnosed in our cohort; 3 were invasive tumors (median size, 16 mm; range, 1.5-2.1cm), and 1 was DCIS. MRI missed only 1 cancer in a woman aged 39 years who had DCIS that manifested as microcalcifications on screening mammography.
Ten cancers (77%) that were diagnosed in our cohort were small cancers (DCIS and invasive cancers <20 mm). In 3 women who had cancers >20 mm, the interval between screening rounds was >12 months. The largest T3 cancer initially was assessed as BI-RADS 3 by MRI; however, unfortunately, the patient did not comply with follow-up recommendations and attended the next screening round 2 years after the date of her BI-RADS 3 assessment.
The average size of the invasive, screening-detected breast cancer, excluding the T3 tumor, was 15 mm (median, 16 mm; range, 0.4-2.4 cm). Four of 13 patients (31%) had positive lymph nodes diagnosed at sentinel lymph node biopsy. The features of all breast cancers detected are listed in Table 1.
|Basis Disease||Chest RT Date||Age of Chest RT, y||Total Radiation Dose, Gy||Age at Breast Cancer Diagnosis, y||Interval Between Chest RT and Breast Cancer, y||Palpability at the Time of Diagnosis||MG||MRI||Histology||Tumor Size, cm||Lymph Node Status|
Sensitivity, specificity, PPV, NPV, and cancer yield of mammography and MRI are reported in Table 2. The sensitivity of MRI was 92%, but this was not statistically different (P > .3) from the sensitivity of mammography (69%). Similarly, specificity, PPV, and NPV of MRI and mammography were not statistically different. In 4 of 98 patients, cancer was diagnosed only by MRI, resulting in a 4.1% incremental cancer detection rate, which was not significantly greater than the 1% (1 of 98 cancers detected by mammography only) added cancer yield of mammography (P > .05).
|Percentage (95% CI)|
|Sensitivity||69 (60-78)||92 (86-97)||.375|
|Specificity||98 (93-99)||94 (87-97)||.375|
|Positive predictive value||82 (74-89)||71 (62-79)||.945|
|Negative predictive value||95 (89-98)||99 (94-99)||.950|
|Added cancer yield per patient||1 (0.2-5.6)||4.1 (1.6-10)||.175|
MRI was associated with a higher benign biopsy rate compared with mammography. Six of 98 women (6.1%) had benign biopsy results, and 4 of those women (4.1%) underwent MRI-guided biopsy; 1 of 98 women (1.02%) underwent a benign mammographically guided biopsy, and 1 of 95 women (1.02%) underwent a benign biopsy based on results of both mammography and MRI. Pathology results from benign biopsies included fibrocystic changes without atypia in 3 women, flat epithelial atypia in 2 women, and a complex sclerosing lesion in 1 woman.
To date, data are scarce regarding the optimal breast screening regimen for women who previously received moderate to high-dose chest radiation, although it has been demonstrated that these women are at very high risk for developing breast cancer.28-33 In our cohort of 98 patients, 4 cancers were diagnosed by MRI alone. The MRI cancer yield of 4.1% in our study is comparable to the results from previous studies (Table 3) that evaluated breast MRI screening in high-risk populations.
|No. of Cancers Detected|
|Reference||No. of Eligible Patients||Total No.||No. Detected by MRI||No. Detected by MRI Only||MRI Cancer Yield per No. of Patients, %||MG Sensitivity, %||MRI Sensitivity, %||P for MRI×MG Sensitivity||Size of IC Detected by MRI: Median (Mean), mm||Size of IC Detected Only by MRI: Median (Mean), mm||Rate of DCIS×IC Detected by MRI||No. of Lymph Node- Positive Patients|
|Hereditary risk of breast cancer|
|Kuhl 201012||687||43||25||14||2.0||32.6||90.7||< .001||NA||NA||NA||NA|
|Radiology, 2007/ Lehman et al||171||6||6||4||2.3||33||100||NA||NA||NA||0×6||1|
|Sardanelli 200718||278||18||15||6||2.2||58.8||93.8||NA||18 (17)||12 (16)||4×11||3|
|Lehman 200515||367||4||4||3||0.8||25||100||NA||10 (9)||9 (9)||1×3||0|
|Kuhl 200511||529||27||39||19||3.6||33.3||92.6||< .0001||11 (12)||7.5 (9)||8×31||5|
|Leach 200516||649||35||27||19||2.9||40||77||.01||15 (16)||13 (15)||2×25||5|
|JAMA, 200414||236||22||17||7||3.0||36.4||77.3||.02||10 (11)||6 (10)||4×13||2|
|Kriege 200414||1909||45||35||22||1.2||40||71.1||< .05||NA||NA||NA||NA|
|Morris 200319||367||14||14||14||3.8||NA||NA||NA||4 (NA)||4 (NA)||8×6||2|
|Kuhl 200010||192||9||9||6||3.1||33||100||< .003||13 (12)||13 (13)||2×7||0|
|Postchest radiation therapy|
|Sung 201121||91||10||7||4||4.4||66.7||66.7||NA||0.7 (0.9)||0.5 (0.9)||2×5||NA|
|Current study||98||13||12||4||4.1||69||92||.375||16 (21)||16 (17)||2×10||1|
Although our study did not demonstrated that the sensitivity of MRI (92%) was statistically higher than the sensitivity of mammography (69%), we believe that our results justify the current recommendation for screening MRI in women who have a prior history of chest radiation. Indeed, 4 of 13 cancers (31%) in our study were diagnosed by MRI alone. Larger studies will be necessary to demonstrate whether this is statistically significant.
Several randomized controlled trials have demonstrated that mammographic screening is effective for reducing breast cancer mortality in the general population.9 Currently, to our knowledge, there are no data regarding the impact of breast MRI screening on breast cancer mortality in high-risk populations, but we assume that the detection of DCIS and small breast cancers may be a surrogate endpoint for MRI screening studies, because it is unlikely that a randomized controlled trial would be performed in high-risk populations.
Our study demonstrated that MRI screening allows the detection of otherwise occult DCIS and early stage invasive breast cancers and, as such, is likely to allow for better outcomes. The relatively low sensitivity of mammography in our study may be explained by the young age of our patients and the presence of mammographically dense breast tissue in 61% of women.
Only 2 of 10 invasive tumors (20%) in our study were manifested as calcifications on mammography. This may be another reason for the lower mammographic sensitivity, because tumors that manifest mammographically as microcalcifications only are detected more reliably by mammography.35
An additional argument in support of breast MRI screening is potentially low compliance with mammographic screening in this cohort of women secondary to concerns of added radiation from mammography leading to radiation-induced breast cancer. Berrington de Gonzalez and Reeves presented a mathematical model that demonstrated a substantial risk of mammography radiation-induced breast cancer.36 This concern was rebutted for mammography screening regimens that begin at age 40 years,37 but it still may be relevant for women before age 40 years, especially among young women who have a prior history of chest radiation therapy.
Improving MRI technologies may increase the accuracy of breast MRI in the near future and may validate the use of MRI alone as a screening modality. Larger studies of MRI screening in high-risk women will be required to explore this question. Arguments against use of breast MRI include high cost and low specificity, which may result in large numbers of benign biopsies. However, Warner et al13 reported that the numbers of false-positive results tend to be related inversely to the radiologist's experience.
In our study, the latent period from chest radiation to breast cancer diagnosis was at least 8 years. Our results support current recommendations to begin annual screening 8 years after radiation.23
In our cohort, almost all patients (12 of 13) who were diagnosed with breast cancers had received chest radiation therapy when they were between ages 10 years and 30 years (median age, 24 years; range, 12-49 years). This is concordant with previously published results.28-33 Our study has several limitations, and the most important is its retrospective design. A retrospective design is the main reason for random time intervals between mammography and MRI examinations and between screening rounds performed with the same modality. The retrospective design and small sample size limited our ability to identify reasons other than chest radiation therapy that may have increased the breast cancer incidence in our cohort. Another limitation is that MRI-guided biopsy became available at our institution only in 2006, and this may have decreased the sensitivity of MRI.
In conclusion, despite the limitations mentioned above, our retrospective study demonstrated that breast MRI is a useful adjunctive modality in screening for women with a history of chest radiation therapy. Our results support the current guidelines,22-24 which recommend the follow-up of women who received mantle radiation between ages 10 years and 30 years and should include annual mammography and MRI beginning at age 25 years or 8 years after irradiation, whichever occurs last.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.
- 5Statistical Research and Applications Branch, National Cancer Institute. DevCan: Probability of Developing or Dying of Cancer Software, version 6.3.1. Bethesda, MD: Statistical Research and Applications Branch, National Cancer Institute; 2008. Available at: http://srab.cancer.gov/devcan. Accessed December 22, 2011.
- 9US Preventive Services Task Force. Screening for breast cancer: an update for the US Preventive Services Task Force. Ann Intern Med. 2009; 151: 727-737, W237-W242., , , , , ;
- 23Children's Oncology Group. Long-Term Follow-Up Guidelines for Survivors of Childhood, Adolescent, and Young Adult Cancer. Available at: http://www.survivorshipguidelines.org/. Accessed February 5, 2010.
- 25American College of Radiology. ACR Breast Imaging Reporting and Data System: Breast Imaging Atlas. Reston, VA: American College of Radiology; 2003.
- 27How to obtain the confidence interval from a P value [serial online]. BMJ. 2011; 343: d2090., .