Risk–benefit analysis of preoperative breast MRI in patients with primary breast cancer

Results The findings of the BMRI examinations changed the clinical management in 48 patients (40.3%). Seventeen women underwent mastectomy instead of breast conservation, eight patients underwent extended excision, 21 additional lesions were clarified by MRI intervention, and two ultrasound-detected lesions were not biopsied because of negative MRI. Histologically malignant additional or extended biopsies ( n = 34) and two cases of spared biopsies resulted in 36 (30.3%) women who benefited from preoperative BMRI. Twelve patients (10.1%) had additional biopsies of MRI-detected benign lesions, and therefore, had an unfavourable outcome due to BMRI. The change in clinical management and patient benefit were independent of BD and TT ( p > 0.05). Conclusion Preoperative BMRI was beneficial for 30.3% of 119 patients with PBC. The percentage of additional biopsies of benign lesions (10.1%) seems acceptable. Introduction To avoid multiple operations in patients with breast cancer, accurate pretherapeutic assessment of both breasts is necessary. Compared with mammography and ultrasound, breast magnetic resonance imaging (MRI) better predicts tumour extent and can detect otherwise occult foci of carcinoma in the same and contralateral breast. 1–10 Furthermore, breast MRI can potentially lower the recurrence rate, however, the influence of preoperative breast MRI on breast cancer recurrence is currently controversial. 11,12 Therefore, breast MRI is superior to mammography and ultrasound for the local staging of breast cancer. 13 The result of preoperative breast MRI can alter therapeutic procedures, but not all alterations are beneficial for the patients, in particular, unnecessary operations on benign lesions 4,14 are disadvantageous. Evidence is needed that the benefit to patients from preoperative breast MRI outweighs the disadvantages due to false-positive MRI findings. Patients with dense breast parenchyma have a higher risk of developing breast cancer than those with fatty breasts 15 ; however, mammographic sensitivity decreases with increasing breast density. 16,17 Recent studies show that MRI can even depict intraductal 6,14,18,19 and invasive lobular carcinoma 6,20,21 with high sensitivity, which is mandatory for accurate preoperative assessment. The influence of breast density and tumour type on therapy changes and the patient benefit from preoperative breast MRI need to be clarified. The purpose of the study was to analyse and compare the risks and benefits of preoperative breast MRI in patients with newly diagnosed breast cancer. Changes to clinical management and whether the patient benefit due to breast MRI varied with mammographic breast density or tumour type were also evaluated. Materials and methods Patients One hundred and forty-four consecutive patients who underwent preoperative bilateral breast MRI for staging of primary breast cancer during a 1-year period from July 2005 to August 2006 were prospectively evaluated. Written, informed consent was given by each patient prior to breast MRI. Institutional review board-approval was obtained. Patients who were operated on at another institution were excluded because of lack of digital data for analysis ( n = 8). Similarly, patients with mammograms from other hospitals or private practices whose images were not available were excluded ( n = 17). Thus, the remaining 119 patients (age 32–83 years, mean 52.7 years) formed the study population. In 25 patients, neoadjuvant chemotherapy was performed prior to surgery. Patient characteristics are listed in Table 1 . Imaging All patients had bilateral, two-view mammograms in craniocaudal and mediolateral-oblique projections, which were carried out at our institution or in external hospitals or private practices. Mammograms from our institution were captured digitally (Senograph 2000D, GE Healthcare, Chalfont St Giles, UK). All patients underwent bilateral breast ultrasound in our institution. Two ultrasound systems with high-resolution, linear, multi-frequency probes (7.5–13.5 MHz) were used (HDI 3000, ATL, Philips Medical Systems, Best, The Netherlands; Voluson 730, GE Healthcare, Chalfont St Giles, UK). Bilateral breast MRI was performed at 1.5 T (Gyroscan Intera, Philips Medical Systems, Best, The Netherlands) by means of a dedicated double breast coil. A routinely used sequence protocol with an unenhanced T2-weighted inversion recovery sequence (STIR) followed by high-resolution, contrast-enhanced, dynamic, T1-weighted, three-dimensional, gradient-echo sequences (fast field echo, FFE) was applied ( Table 2 ). A bolus of gadopentetate-dimeglumine (Magnevist ® , Schering, Berlin, Germany) at a dose of 0.16 mmol/kg of body weight and a bolus of 20 ml 0.9% sodium chloride were injected intravenously immediately after the first (unenhanced) dynamic series. The acquisition time of every dynamic series was 65 s, so that the whole dynamic scan of one unenhanced and seven enhanced series was finished in 520 s. Image subtraction was performed for each of the seven enhanced dynamic series. Image interpretation All mammograms and digital breast MRI images were read in consensus by two radiologists (K. C. S. and A. B.). Mammographic breast density and findings in both breasts were categorized according to BI-RADS ® . 22 For analysis of breast MRI a dedicated workstation (View forum, Philips Medical Systems, Best, the Netherlands) was used. All findings were assessed by region of interest (ROI) analysis. Morphology and enhancement kinetics were combined according to a published score system, which can be translated into corresponding BI-RADS ® categories. 23 Change in clinical management and risk–benefit analysis Breast side, localization, and maximum size of the index tumour were noted using mammography, ultrasound, and MRI. The detected lesions were counted separately for the right and the left breasts. The BI-RADS ® category for mammographic and ultrasound-detected lesions, as well as the breast MRI score and BI-RADS ® category for MRI-detected lesions were noted. All breast MRI studies with findings that did not concur with the mammography and ultrasound findings were thoroughly analysed. Therapy changes as a consequence of additional suspicious (BI-RADS ® 4 and 5) MRI findings were extended excision, additional biopsy, or mastectomy instead of breast conservation. Larger excision was performed if the lesion's extent on MRI was at least 10 mm larger than on conventional imaging or if additional suspicious foci were depicted within 30 mm distance to the index cancer. Correlation with histopathological results was performed. Preoperative breast MRI was defined as beneficial to the patient if: (1) the extension of the index cancer was larger than on mammography and ultrasound and if malignancy was confirmed by histology; (2) additional malignant lesions were depicted on MRI; (3) inconspicuous MRI findings in regions of doubtful ultrasound-detected lesions prevented biopsy. All benign lesions that were biopsied due to suspicious (BI-RADS ® 4 and 5) MRI findings were defined as disadvantageous for the patient. Mammographic breast density and histopathologic tumour type were correlated with change of clinical management and patient benefit. Surgery and biopsy All 119 patients were operated on at the breast centre at our institution. According to international guidelines 24 either breast conserving treatment (BCT) or mastectomy was performed. Preoperatively image-guided hook-wire localization was carried out in the case of non-palpable tumours. If suspicious, enhancing lesions were only seen on MRI, the patients' mammograms were reviewed and breast ultrasound was repeated. MRI-guided intervention was only considered if additional detected lesions were once again occult to conventional imaging. All additionally suspicious MRI-detected lesions were histologically proven by MRI-guided localization and subsequent excisional biopsy or MRI-guided vacuum-assisted biopsy (Vacora ® , BARD, Karlsruhe, Germany) by means of a commercially available target system (Philips Medical Systems, Best, the Netherlands). For MRI-guided localization, a metal coil (MREYE ® Breast Localization Coil, Bjaeverskov, Denmark) was used. The details of the MRI-guided lesion localization procedure have been previously described. 25 Histopathologic evaluation Using formalin-fixed and paraffin-embedded tissue, 5 μm thick sections were cut and stained with haematoxylin and eosin according to standardized histopathological techniques. Sections were analysed by experienced breast pathologists from our breast centre. Breast lesions were classified according to the current classification of the World Health Organization. 26 Malignant tumours were staged according to the TNM system of the Union Internationale Contre le Cancer (UICC). 27 Histopathological tumour type and maximum tumour size in millimetres were noted in the pathology report. Statistics Statistical analyses were performed by use of statistics software (JUMP 5.0; SAS Institute Inc., USA). To test the independence of variables, Pearson's Chi-square test was used to assess the change of clinical management and patient benefit for significant correlation with histological tumour type and mammographic breast density. Significant correlation was defined as p < 0.05. Results Tumours As shown in Fig. 1 , the index tumour of most patients was invasive ductal carcinoma (IDC; 61/119; 51.3%), followed by invasive lobular carcinomas (ILC; 31/119; 26.1%), mixed invasive ductal–lobular carcinomas (15/119; 12.6%), and four invasive medullar cancers (3.4%). In eight patients, ductal in situ carcinomas were proven (6.7%). Twenty-five patients with larger carcinomas of more than 20 mm in diameter (21%) underwent neo-adjuvant chemotherapy. Because of therapy-associated tissue alteration and tumour shrinkage, the final histopathological tumour size is not always accurately measurable. The tumour size of these patients measured on MRI varied from 21–98 mm, mean 41.4 mm. The corresponding clinical TNM stages were 17 cases of cT2 (>20 mm) and eight cases of cT3 (>50 mm). The histopathological tumour size and corresponding TNM stages were evaluated for patients without neoadjuvant treatment ( n = 94). The maximum tumour diameter of these patients varied from 3–80 mm, mean value 20.4 mm. The majority of 70.2% (66/94) patients had either ductal carcinoma in situ (DCIS; n = 7) or small invasive cancers of a maximum diameter of 20 mm (pT1; n = 59). The remaining patients had larger tumours with diameters of >20 mm (pT2; n = 25) or >50 mm (pT3; n = 3). Change in clinical management The results of the preoperative breast MRI changed the initial clinical management of 48 patients (40.3%). The most frequent change in clinical management was additional MRI-guided intervention ( Fig. 2 ) in order to clarify suspicious enhancing lesions that were occult on mammography and ultrasound (21/48; 43.7%) consisting of either MRI-guided lesion localization and subsequent excisional biopsy ( n = 15) or MRI-guided vacuum-assisted biopsy ( n = 6). MRI-guided intervention revealed malignancy in 52.4% (11/21) of cases, whereby nine synchronous contralateral cancers (7.5% of all patients) were found. The size of the additional malignant foci ( n = 11) varied from 7–25 mm (mean 14.5 mm). Characteristics of MRI detected contralateral cancers are listed in Table 3 . The second most frequent therapy change (17/48; 35.4%) was mastectomy instead of BCT due to extended or multifocal tumour involvement of the breast, which was not depicted on mammography and ultrasound ( Fig. 3 ). Ten of these patients (10/17; 58.8%) underwent mastectomy based upon MRI findings only either because of the patient's request( n = 3) or due to multiple BI-RADS ® 5 lesions with a high probability of malignancy ( n = 3) or in case of large carcinomas (>4 cm) that were only partially depicted using mammography and ultrasound ( n = 4). The remaining patients ( n = 7) underwent mastectomy due to biopsy-proven (ultrasound-guided large-core needle biopsy) areas of further malignancy. The final histological results of all patients who underwent mastectomy instead of BCT ( n = 17) showed either more lesions or a larger extent of the known cancer compared with mammography and ultrasound examinations. Eight patients (16.6%) underwent a larger excision than initially intended ( Fig. 4 ). Seven of them underwent a larger excision based upon MRI findings only. One patient underwent a larger excision due to biopsy-proven (ultrasound-guided large-core needle biopsy) malignancy close to the known cancer. In six of these cases, histology showed an additional DCIS in the surrounding area of the known invasive cancer or additional cancer foci. In two cases, suspicious MRI findings were caused by benign lesions (one fibroadenoma, one adenosis). In two patients (4.2%) with doubtful ultrasound-detected lesions, benignity could be proven by a reassuring appearance on MRI without any enhancement or mass in the corresponding breast region. In these cases, biopsy was avoided. To prove the stability of these lesions ultrasound follow-up was performed at 6 monthly intervals. No additional cancers have been discovered to date. Risk–benefit analysis The percentage of MRI-detected additional suspicious findings was 73.9% (34/46). Together with two patients who avoided biopsy due to negative MRI findings, 75% (36/48) of all therapy changes were beneficial for the patients. In 12 out of 48 patients (25%) the change in clinical management due to preoperative breast MRI was disadvantageous because of additional biopsies or extended excision of benign lesions. With regard to the entire patient population, 30.3% (36/119) had a beneficial outcome due to preoperative breast MRI, whereas 10.1% (12/119) had an unfavourable outcome because of unnecessary benign biopsies. The remaining 59.7% (71/119) of patients had no change in management due to preoperative breast MRI. Influence of mammographic breast density According to BI-RADS ® almost half of the patients (56/119; 47.1%) had heterogeneously dense breasts of category ACR 3. A fattier breast composition of ACR 2 existed in 35 patients (29.4%). Twenty-two patients (18.5%) had very dense breast tissue (ACR 4) and six patients (5%) had fatty breasts (ACR 1). Pearson's chi-square test shows no dependency between breast density and MRI-related therapy changes ( p = 0.94) as illustrated in Table 4 . The same is true for patient benefit due to preoperative breast MRI. Pearson's chi-square test reveals no relationship between patient benefit and mammographic breast density ( p = 0.69). Therapy changes of patients with a fatty breast composition (ACR 1) were all favourable (3/3; 100%). Most of the patients with denser breast tissue of category ACR 2 (10/13; 76.9%) and ACR 3 (16/23; 69.6%) had favourable therapy changes. Patients with very dense parenchyma (ACR 4) showed a slightly higher benefit due to change in clinical management (7/9; 77.8%), which was not statistically significant compared with the other categories. Influence of tumour type Statistical testing revealed no relationship between histopathological tumour type of the index tumour and the likelihood of therapy changes ( p = 0.68). Preoperative breast MRI changed the therapeutic procedure in 36.1% (22/61) of patients with IDC and in 41.9% (13/31) of patients with ILC, as well as in 50% (2/4) of patients with mixed invasive tumours. In the case of DCIS and medullary cancers, clinical management was changed in 62.5% (5/8) and 40% (6/15), respectively. Both the change in therapy and patient benefit due to preoperative breast MRI were independent of the histopathological type of the index tumour ( p = 0.47). Discussion Change of clinical management and risk–benefit analysis In the present study, the malignancy rate of MRI-detected additional suspicious findings was high (73.9%). MRI-detected synchronous contralateral breast cancer was present in 7.5%. A similar proportion of true-positive additional MRI findings and rates of synchronous contralateral breast cancers (6–9%) in breast cancer patients have been reported in the literature. 1,3,6,8,28 In the present study the findings of breast MRI altered the initial clinical management in 40.3% of patients with primary breast cancer. Thirty-six of the 119 (30.3%) patients experienced a benefit because of preoperative breast MRI, and most of the therapy changes were favourable for these patients (75%). Many studies reporting change in clinical management as a result of breast MRI reveal similar results with therapy change in 20–50% of the patients. 4,14,20,29 Fewer studies examine patient benefit from preoperative breast MRI. Two authors 20,29 showed that 71 and 81% of therapy changes due to MRI findings were beneficial for the patients. A favourable effect of breast MRI was reported in 11–23% of early-stage breast cancer patients eligible for breast-conserving therapy. 5,14 A higher percentage (30.3%) was found in the present study, which was probably caused by the high number of large tumours. In the present study population, 45.4% of all tumours were >20 mm on final histopathology or pretherapeutic MRI if neoadjuvant treatment was given. The patient benefit as a result of MRI is potentially higher in these cases because they present more often with additional cancer foci or larger cancers that are completely or partially occult on conventional imaging. Only few authors report false-positive MRI findings in breast cancer patients. Twelve patients of the present study population (10.1%) had unnecessary biopsies or extended excisions of benign lesions due to suspicious MRI findings and, as a result, a non-favourable outcome. Bedrosian et al. 29 reported similar results with 7.5% false-positive MRI-detected lesions. Unfavourable effects of preoperative breast MRI were seen in 6.6% of 212 cases in the study by Tillman et al. 14 Those patients had either additional benign biopsies or unnecessary mastectomies due to MRI findings. In the present study unnecessary mastectomies did not occur. Biopsy is advisable prior to a change of clinical management, especially if mastectomy is planned. Influence of mammographic breast density and tumour type In the present study, changes in clinical management due to preoperative breast MRI were independent of mammographic breast density. However, the subgroup of patients with fatty breasts of category ACR 1 was small ( n = 6), which limits the validity of this result. Reports about the influence of breast density on the result of breast MRI in breast cancer patients are contradictory. Some authors presume that breast MRI probably leads to more therapy changes in the case of dense breasts. 2,7,28,30 In the study by Sardanelli and co-workers, 17 MRI was more sensitive than mammography for the detection of malignant foci in fibroglandular and dense breasts compared with fatty breasts. Tillman et al. 14 showed that age was not a predictor for the likelihood of MRI to affect clinical management in patients with breast cancer. Because younger women usually have denser breast parenchyma than older women, the result can be presumably transferred to breast density even though the authors did not comment separately on the influence of breast density on clinical management. Deurloo et al. 5 showed that mammographic breast density was not correlated with the complementary value of MRI to assess tumour extent. However, larger studies are required to determine the influence of breast density on patient benefit from breast MRI. There are several reports about the usefulness of breast MRI in patients with invasive breast cancers and invasive lobular carcinomas, 20,21 but none compares the influence of tumour type with breast MRI findings and patient benefit. Even though the present results show that patients who present with invasive lobular primary cancer had a slight increase in change in clinical management (41.9%), the proportion was not significantly different from other cancer types, in particular, IDC (36.1%). The same applies for DCIS. In the present study the results of breast MRI changed the clinical management more often, but not significantly, in patients with DCIS (62.5%) compared with patients with invasive ductal carcinomas. Similar results are reported by Tillman et al., 14 who compared women with DCIS and those with invasive carcinomas. No significant difference in the likelihood of having a change in management based on MRI findings was found. Similarly, the favourability of MRI-affected changes among patients with intraductal and invasive carcinoma was not significantly different. Therefore, preoperative breast MRI provides important additional information in case of invasive carcinoma and DCIS. Limitations One limitation is the bias of a non-randomized patient selection. The inclusion criteria were newly diagnosed breast cancer and breast MRI performed at our institution. Randomization of patients with breast cancer to preoperative breast MRI or no preoperative breast MRI could not be performed because breast MRI is at the surgeon's request at our institution. In our breast centre, breast surgeons request preoperative breast MRI in approximately 75% of all breast cancer patients. If the results of the mammography and ultrasound were doubtful or inconclusive, or if the mammographic sensitivity was limited by dense breast tissue, the probability of having a preoperative breast MRI was higher. A probable consequence is the small subgroup of patients with almost entirely fatty breasts of category ACR 1 ( n = 6). Conversely, most patients ( n = 91) had moderately dense breasts with categories of ACR 2 and ACR 3, and therefore, represent a typical patient population with newly diagnosed breast cancer. Thus, we suppose that this limitation does not fundamentally affect the study results. Furthermore, there were only a few patients with DCIS ( n = 8; 6.7%) compared with other tumour types. Nevertheless, other similar studies have comparable rates of patients with intraductal tumours of 2–19.3%, 4,5,14 so these rates of DCIS are not unusual in patients with newly diagnosed breast cancer. Conclusion In conclusion, the results of the present study demonstrate that 30.3% of all patients with newly diagnosed breast cancer benefited from preoperative breast MRI. The percentage of patients with non-favourable outcomes as a result of additional benign biopsies of MRI-detected suspicious lesions was acceptably low (10.1%). To avoid over-treatment, minimal-invasive histological confirmation of MRI-detected suspicious lesions should be obtained before surgery, especially if mastectomy is planned. Mammographic breast density and histological tumour type seem to have no significant influence on patient benefit. However, larger, randomized studies with long-term follow-up are needed to identify patient groups that will benefit from preoperative breast MRI with regard to recurrence and outcome. References 1 S.G. Lee S.G. Orel I.J. Woo MR imaging screening of the contralateral breast in patients with newly diagnosed breast cancer: preliminary results Radiology 226 2003 773 778 2 L. Liberman E.A. Morris C.M. Kim MR imaging findings in the contralateral breast of women with recently diagnosed breast cancer AJR Am J Roentgenol 180 2003 333 341 3 P. Viehweg K. Rotter M. Laniado MR imaging of the contralateral breast in patients after breast-conserving therapy Eur Radiol 14 2004 402 408 4 M. Van Goethem K. Schelfout L. Dijckmans MR mammography in the pre-operative staging of breast cancer in patients with dense breast tissue: comparison with mammography and ultrasound Eur Radiol 14 2004 809 816 5 E.E. Deurloo W.F.A. Klein Zeggelink H.J. Teertstra Contrast-enhanced MRI in breast cancer patients eligible for breast-conserving therapy: complementary value for subgroups of patients Eur Radiol 16 2006 692 701 6 W.A. Berg L. Gutierrez M.S. NessAiver Diagnostic accuracy of mammography, clinical examination, US and MR imaging in preoperative assessment of breast cancer Radiology 233 2004 830 849 7 L. Liberman E.A. Morris D.D. Dershaw MR imaging of the ipsilateral breast in women with percutaneously proven breast cancer AJR Am J Roentgenol 180 2003 901 910 8 L. Libermann Breast MR imaging in assessing extent of disease Magn Reson Imaging Clin N Am 14 2006 339 349 9 M. Van Goethem W. Tjalma K. Schelfout Magnetic resonance imaging in breast cancer Eur J Surg Oncol 32 2006 901 910 10 S. Ozaki M. Tozaki E. Fukuma Bilateral breast MR imaging: is it superior to conventional methods for the detection of contralateral breast cancer? Breast Cancer 15 2008 169 174 11 U. Fischer O. Zachariae F. Baum The influence of preoperative MRI of the breasts on recurrence rate in patients with breast cancer Eur Radiol 14 2004 1725 1731 12 L.J. Solin S.G. Orel W.-T. Hwang Relationship of breast magnetic resonance imaging to outcome after breast-conservation treatment with radiation for women with early-stage invasive breast carcinoma or ducal carcinoma in situ J Clin Oncol 26 2008 386 391 13 C. Kuhl W. Kuhn M. Braun Pre-operative staging of breast cancer with breast MRI: one step forward, two Steps back? Breast 16 Suppl. 2 2007 S34 S44 14 G.F. Tillman S.G. Orel M.D. Schnall Effect of breast magnetic resonance imaging on the clinical management of women with early-stage breast carcinoma J Clin Oncol 20 2002 3413 3423 15 K. Kerlikowske J. Shepherd J. Creasman Are breast density and bone mineral density independent risk factors for breast cancer? J Natl Cancer Inst 97 2005 368 374 16 T.M. Kolb J. Lichy J.H. Newhouse Comparison of the performance of screening mammography, physical examination, and breast US and evaluation of factors that influence them: an analysis of 27,825 patient evaluations Radiology 225 2002 165 175 17 F. Sardanelli G.M. Giuseppetti P. Panizza Sensitivity of MRI versus mammography for detecting foci of multifocal, multicentric breast cancer in fatty and dense breasts using the whole-breast pathologic examination as a gold standard AJR Am J Roentgenol 183 2004 1149 1157 18 J.H. Menell E.A. Morris D.D. Dershaw Determination of the presence and extent of pure ductal carcinoma in situ by mammography and magnetic resonance imaging Breast J 11 2005 382 390 19 C.K. Kuhl S. Schrading H.B. Bieling MRI for diagnosis of pure carcinoma in situ: a prospective observational study Lancet 370 2007 485 492 20 K. Schelfout M. Van Goethem E. Kersschot Preoperative breast MRI in patients with invasive lobular breast cancer Eur Radiol 14 2004 1209 1216 21 S.P. Weinstein S.G. Orel R. Heller MR imaging of the breast in patients with invasive lobular carcinoma AJR Am J Roentgenol 176 2001 399 406 22 American College of Radiology (ACR) Breast imaging reporting and data system ® (BI-RADS ® ) atlas 4th edn 2003 American College of Radiology Reston 23 F. Baum U. Fischer R. Vosshenrich Classification of hypervascularized lesions in CE MR imaging of the breast Eur Radiol 12 2002 1087 1092 24 EJTh Ruthgers Position Paper EUSOMA guidelines. Quality control in the locoregional treatment of breast cancer Eur J Cancer 37 2001 447 453 25 K.C. Siegmann C. Gorriz T. Xydeas Preoperative MR imaging-guided localization of 131 breast lesions with modified embolization coils Invest Radiol 40 2005 368 377 26 F.A. Tavassoli P. Devillee World Health Organization Classification of tumors. pathology and genetics. Tumours of the breast and female genital organs 2003 IARC Press Lyon 27 L.H. Sobin C.H. Wittekind UICC TNM classification of malignant tumours 6th edn. 2002 Wiley-Liss New York 28 C.S. Mameri C. Kemp S.M. Goldman Impact of breast MRI on surgical treatment, axillary approach, and systemic therapy for breast cancer Breast J 14 2008 236 244 29 I. Bedrosian R. Mick S.G. Orel Changes in the surgical management of patients with breast carcinoma based on preoperative magnetic resonance imaging Cancer 98 2003 468 473 30 D.A. Bluemke C.A. Gatsonis M.H. Chen Magnetic resonance imaging of the breast prior to biopsy JAMA 292 2004 2735 2742