- Email: [email protected]
Preoperative MRI evaluation of lesion–nipple distance in breast cancer patients: thresholds for predicting occult nipple–areola complex involvement
Clinical Radiology xxx (2018) 1e9
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement G. Mariscotti a, *, M. Durando a, N. Houssami b, C.M. Berzovini a, F. Esposito a, M. Fasciano a, P.P. Campanino c, D. Bosco a, R. Bussone d, A. Ala d, I. Castellano e, A. Sapino f, L. Bergamasco g, P. Fonio a, G. Gandini a a
Department of Diagnostic Imaging and Radiotherapy, Radiology Institute, University of Turin, Turin, Italy Screening and Test Evaluation Program, School of Public Health, Sydney Medical School, University of Sydney, Sydney, 2006, NSW, Australia c Breast Imaging Service, Koelliker Hospital, Turin, Italy d della Salute e della Scienza di Torino, Presidio Ospedaliero Breast Surgery, Department of Surgery, A. O. U. Citta Molinette, Turin, Italy e della Salute e della Scienza di Torino, Department of Biomedical Sciences and Human Oncology, A. O. U. Citta Presidio Ospedaliero Molinette, Turin, Italy f Fondazione del Piemonte per l’Oncologia (FPO), Candiolo Cancer Institute (IRCCs), Turin, Italy g della Salute e della Scienza di Torino, Presidio Department of Surgical Sciences, University of Turin, A. O. U. Citta Ospedaliero Molinette, C.so Bramante 88, 10126, Torino, Italy b
art icl e i nformat ion Article history: Received 6 September 2017 Accepted 14 March 2018
AIM: To identify clinically occult nippleeareola complex (NAC) involvement using preoperative magnetic resonance imaging (MRI), to inform selection of patients eligible for nipplesparing mastectomy (NSM) or skin-sparing mastectomy (SSM). MATERIAL AND METHODS: This was a retrospective study of 195 patients, who had preoperative breast MRI (February 2011 to January 2017) before undergoing surgical treatments (NSM or SSM) for newly diagnosed breast cancer. Tumour features at MRI (mass or non-mass lesion, diameter, lesioneNAC distance [LND]) and pathology (lesion diameter, histopathological type, receptor status) were recorded, as well as the type of surgery (NSM/SSM) and presence (NACþ) or absence (NACe) of tumour at intraoperative evaluation of retroareolar tissue. ManneWhitney test, Fisher’s exact test, logistic regression, and receiver operating characteristic (ROC) curve analysis were used for analysis of NACþ versus NACe to assess variables that predict NAC tumoural involvement. RESULTS: Over the study period, NACþ was proven histologically in 71/200 (35.5%) surgical treatments, while there were 129/200 NACe (72 NSM and 128 SSM performed). LND at MRI was statistically (p<0.001) lower in NACþ patients than in NACe patients. The area under the ROC curve (0.82, 95% confidence interval [CI]: 0.76e0.88) indicated 10 mm as the best cut-off,
* Guarantor and correspondent: G. Mariscotti, Department of Diagnostic Imaging and Radiotherapy, Radiology Institute, University of Turin, A. O. U. Citt a della Salute e della Scienza di Torino, Presidio Ospedaliero Molinette, Via Genova 3, 10126, Torino, Italy. Tel.: þ39 0116336095; þ39 0116336622; fax: þ39 0116960310. E-mail address: [email protected] (G. Mariscotti). https://doi.org/10.1016/j.crad.2018.03.008 0009-9260/Ó 2018 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
2
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
with sensitivity of 82%, specificity of 72%, and accuracy of 79%. A 5-mm cut-off enhanced sensitivity, whereas a 15-mm cut-off favoured specificity. CONCLUSIONS: MRI is a useful tool for identifying NACþ patients; a 10-mm cut-off for LND assists selection of patients for NSM, although intraoperative retroareolar tissue examination remains mandatory. Ó 2018 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction Over the last 20 years, advances in breast surgery and reconstruction techniques have led to significant improvements in terms of quality of life and aesthetic results for patients with breast cancer. Despite much progress made also in diagnosing breast cancer at early stages of disease, breast-conserving surgery is not always a viable option, with up to one-third of patients still requiring (or at times requesting) mastectomy as definitive surgical treatment.1e3 In this setting, nipple-sparing mastectomy (NSM) has acquired a progressively increasing role in breast reconstructive approaches, pairing good cosmetic outcomes with generally good oncological outcomes. The main oncological concern with NSM is the possibility of leaving residual tumour cells in the nippleeareola complex (NAC), which may later manifest as local recurrence; this has been reported to range from 0% to 24% in recent studies,4 not statistically different from conventional mastectomy.3,5 Intraoperative histopathological examination of the subareolar tissue is in fact mandatory to exclude the presence of malignancy at this site, and consequently, to avoid a conversion to skin-sparing mastectomy (SSM).3,4 Recent reviews and meta-analyses underline the need for further long-term follow-up studies to determine the best reconstructive techniques, although several data already support the oncological safety of NSM for appropriately selected patients.3 The selection is, however, a crucial issue: the inclusion criteria for NSM treatment eligibility reported in the literature are quite variable, despite many pathological and radiological tumour features having been identified as predictive of NAC involvement. Conventional diagnostic imaging techniques, such as mammography and sonography, have been proposed for the evaluation of tumoural NAC involvement: specifically, it has been reported that the mammographic distance of the tumour to the nipple can be applied as an independent variable in a predictive equation that may assist the clinician in determining appropriate candidates for NSM6,7; however, mammography often fails to correctly evaluate the retroareolar lesion, because it is difficult to differentiate it from the normal nipple structures or from normal retroareolar glandular density (mainly in dense breasts).8,9 Sonography depicts nipples with posterior acoustic shadowing, so when a malignant mass is found, it might be difficult to evaluate the malignant invasion of the NAC.9 More recently, preoperative imaging with breast magnetic resonance imaging (MRI) has been shown to play a role for
predicting NAC involvement, particularly in the evaluation of lesionenipple distance (LND) and in the identification of pathological enhancements between the index lesion and the NAC.10 Currently, well circumscribed single or multifocal lesions with LND >20 mm are considered suitable for NSM3,11; however, a recent study by Ponzone et al.2 showed that a LND of 10 mm or even 5 mm can be a more specific (88.6% and 85%, respectively), though less sensitive (32.2% and 38.7%, respectively) cut-off for prediction of NAC tumoural involvement. Other criteria suggested in the literature for NSM (suggesting low likelihood of NAC tumoural involvement) include peripheral location of tumours, diameter <5 cm, and human epidermal growth factor 2 (HER2) negativity.3,12 This paper reports a retrospective study aimed at evaluating the role of MRI in the preoperative identification of otherwise clinically occult tumoural NAC involvement in breast cancer patients, to potentially inform surgical planning in this clinical setting. MRI findings are compared with tumour histopathological features and intraoperative retroareolar tissue sampling responses.
Materials and methods Study population The breast cancer records of the Breast Imaging Service of the Radiology Institute of the A. O. U. Citta della Salute e della Scienza of Turin (Italy) were reviewed retrospectively to identify all consecutive patients with newly diagnosed breast cancer who underwent preoperative breast MRI between February 2011 and January 2017 and were subsequently treated with mastectomy, either NSM or SSM. This retrospective study was approved by the local institutional ethics committee. The study sample was constructed using the following exclusion criteria: (1) patients whose preoperative MRI was not performed at our Breast Imaging Service, (2) patients with breast implants, (3) patients who underwent neoadjuvant chemotherapy, (4) patients with a previous history of breast cancer surgery (who had local recurrence during follow-up), (5) patients undergoing prophylactic mastectomy. As the aim of the study was to assess the ability of MRI to detect an otherwise occult tumoural involvement that could be recognised only through subareolar tissue histopathological examination, patients who showed presence of nipple involvement at clinical examination (nipple retraction, Paget’s disease of the nipple, or suspicious nipple
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
discharge) were also excluded. Based on these criteria, there were 195 eligible patients who had undergone 200 breast surgical treatments (Fig 1).
MRI examination MRI examinations were conducted according to the EUSOMA working group recommendations,13 using a 1.5 T MRI system (Philips, Achieva, Best, The Netherlands) with a dedicated seven-channel coil (Philips Sense breast 7). All patients signed an informed consent form before undergoing the MRI examination. The dynamic study was performed using a T1-weighted high-resolution isotropic volume examination sequence in the axial plane (Dyn-EThrive, 5.18 ms repetition time/2.52 ms echo time; 352352 matrix; 340340 mm field of view; 1 mm section thickness; 110.9 mm voxel size; acquisition time <70 seconds), acquired before and five times after intravenous contrast medium injection (0.1 mmol/kg dimeglumine gadobenate, Multihance, Bracco Imaging, Milano, Italia) at a
3
rate of 2 ml/s, followed by 20 ml saline flush. Before contrast medium injection, T1-weighted and T2-weighted time inversion recovery sequences were also acquired, as well as diffusion-weighted imaging (DWI), according to current recommendations. Subtracted images and multiplanar reconstructions (MPR) were derived from the dynamic study data-set.
MRI interpretation Two radiologists dedicated to breast imaging (with 12 and 5 years of experience in breast MRI, respectively) reviewed the set of images for every patient in consensus, and were blinded to clinical and pathological information. Any discrepancy was resolved by a third radiologist (with 7 years of experience in breast MRI). According to the BI-RADS MRI lexicon,14 each breast lesion was classified as focus, mass, or non-mass enhancement. Lesions were also categorised as unifocal, multifocal, and/or multicentric according to lesion number and location. MPR of three-dimensional images from the first postcontrast dynamic study were used to assess the maximum diameter of the lesion. For multifocal or multicentric lesions, the reference measure included the whole area occupied in the breast. On the first high-resolution, postcontrast acquisition (and related subtracted images), the LND was measured by electronic callipers as the shortest distance (in millimetres) between the lesion border and the base of the nipple, using the MPR images, more often on para-sagittal or para-axial planes. The maximum intensity projection (MIP) tool was used as an aid to better identify the lesion and suspicious satellite foci or associated in situ component, but not for the assessment of distance between NAC and the closest margin of pathological enhancement, that could be referred to the lesion.
Technique for intraoperative assessment of the subareolar tissue during NSM
Figure 1 Flow chart of patients’ selection in the present study.
NACesparing mastectomy was performed through an italic S incision, detaching the superior part of the breast by dissecting the Cooper ligaments first, and then removing the mammary gland along the pectoralis major fascia. The NAC was elevated just beneath the level of the deep dermis. Following breast removal, by nipple eversion the central ducts are transected at the base of the nipple. Surgical practice of NAC isolation was performed by hydro-dissection of the areola: 10e20 ml of cold saline solution was injected into the deep subareolar dermis to obtain complete detachment of the skin, then the areola was isolated by dissecting the swollen plane with scissors and the nipple could be cored without increasing the risk of ischaemic complications. The retroareolar tissue specimen was sent for intraoperative frozen section biopsies: if the tissue was positive for invasive or in situ carcinoma, the NAC was removed, thereby converting surgical treatment to SSM.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
4
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
Histopathological assessment of NAC involvement and subareolar tissue examination All NSM candidates underwent intraoperative subareolar tissue examination, according to the procedure described above. The NAC was considered involved by tumour whenever at least one of the samples at histopathological intraoperative examination revealed the presence of invasive ductal or lobular carcinoma cells, or ductal carcinoma in situ. Presence of in situ lobular neoplasia was not considered as criterion for NAC involvement. The subareolar tissue was subsequently reviewed at definitive histology to confirm the intraoperative diagnosis. Patients who were candidate for SSM did not undergo intraoperative subareolar tissue examination: the nipple was removed with the breast and examined at definitive histology. The NAC tumoural involvement criteria at definitive histology were the same as those adopted for intraoperative assessment.
Data collection Age, tumour features at MRI (number and type of lesions, diameter, LND), type of surgery performed (NSM or SSM), and histopathological information (tumour histology, lymph nodal status, intraoperative and definitive subareolar tissue examination results) including prognostic factors (oestrogen receptor [ER], progesterone receptors [PR], Ki67 proliferation index, and HER2B receptor presence] were recorded in a dedicated database.
Statistical analysis Continuous variables were checked for normality using the ShapiroeWilk W-test. Normality was rejected, so variables were reported as median and interquartile range (IQR; representing 25th percentile to 75th percentile). The nonparametric ManneWhitney test for two independent distributions was used for their comparisons. The consensus between the sets of MRI LND measurements of the two readers was estimated using Wilcoxon’s test and the intraclass correlation coefficient (ICC) and the chi2 coefficient. Linear correlations between two continuous variables were tested by Pearson’s linear coefficient r with standard error (SE) and relative probability p (r¼0). Categorical variables were expressed as counts and percentages and compared with the chi-square test (with Yates’s correction for 22 tables) or when more appropriate using Fisher’s exact test. Odds ratios (ORs) were reported with 95% confidence levels. Binary logistic regression was used to measure the relationship between NAC tumoural involvement (NACþ) and the various continuous and categorical dependent variables. The frequency distributions of the LND for the NACþ cases and cases with no tumoural NAC involvement (NACe) were used to compute the stratum (or Level) specific likelihood ratios (SSLR) defined as the ratio of the two frequencies at different levels.15e17
The frequency distributions were also the basis for computing the receiving operating characteristics (ROC) curve for measuring the discriminating ability of LND through the area under the curve (AUC). Sensitivity and Specificity at the various levels of the ROC curve were used to identify the most appropriate cut-off between presence or absence of NAC involvement through: (1) maximisation of the harmonic mean (HM) of sensitivity and specificity and of Youden’s index18 and (2) minimisation of the distance of the curve from the (0,1) upper left vertex. Statistical significance was set at two-tails p<0.05. Statplus: Mac version v6 (AnalystSoft, Walnut, CA, USA) was used for the analysis.
Results Tumour features at preoperative imaging and histopathological characteristics The patients’ age at diagnosis ranged from 27 to 76 years (median 51 years; IQR 46e62 years). Of all included patients, 132 had unifocal disease, 55 multifocal or multicentric disease, eight bilateral disease (five underwent bilateral mastectomy, and three unilateral mastectomy and contralateral wide excision) yielding a total of 200 surgical radical treatments (NSM or SSM) for newly diagnosed breast cancers. Overall, 72 NSM and 128 SSM were performed (36% and 64%, respectively), the latter being numerically prevalent in the first period of observation, with a progressively increasing number of NSM in more recent years. NACþ was histologically proven in 71/200 (35.5%) surgical treatments. There were no false-negative cases of the intraoperative subareolar tissue examination compared to the definitive histology. Table 1 reports the comparison between the two groups of NACþ and NACe patients considering age, LND measurements and tumour diameter at pathology. The consensus of the two readers on the MRI LND measurements was high (no statistical differences and ICC and chi2 close to 1), with the few discrepancies resolved by the third radiologist. The other tumour features at MRI and their histopathological characteristics, subdivided in the two groups NACþ and NACe, are reported and compared in Table 2. Among the continuous variables considered, univariate analysis evidenced significant differences between NACþ and NACe patients for LND (p<0.0001) and maximum tumour diameter at MRI (p¼0.003). The logistic regression analysis confirmed the significant differences for LND: p<0.0001, OR¼0.93, with 95% CI: 0.90e0.95, indicating a large value of LND as a protective factor, but not for the lesion size at MRI (p¼0.48; OR¼1.00 with 95%CI: 0.98e1.03). This variable was found to be inversely correlated with LND (r¼e0.32, SE¼0.05; p<0.0001). Fig 2 shows the box plot for LND.
LND threshold for NAC involvement The strong statistical significance of LND seen in the univariate analysis and logistic regression suggested a
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
5
Table 1 Patients’ age and lesion to nippleeareola complex (NAC) distance and lesion diameter at magnetic resonance imaging subdivided in the two groups with or without NAC tumoural involvement.
Age (years)a LesioneNAC distance (LND), mma Tumour diameter at MRI, mma a b
NACþ (n¼71)
NACe (n¼129)
Univariate p-value
Log regression p-value
53 (45e64) 0 (0e13)
51 (46e61) 20 (15e39)
0.50 <0.00001
<0.00001b
47 (27e60)
33 (23e47)
0.003
0.48
Data are expressed as median (IQR). OR¼0.92 (0.90e0.95).
possible discriminator role for this variable. Fig 3 shows the LND frequency distributions for the NACþ and NACe groups. They intersect between 10 and 15 mm, with NACþ cases prevailing at smaller distances (SSLR¼5.9 for LND¼5 mm, decreasing to 2 for LND¼10 mm level) and being overcome by NACe cases at larger distances (SSLR¼0.5 for LND¼15 mm level decreasing to 0.36 for NLD¼20 mm). The ability of LND to discriminate between NACþ and NACe was graded as good by the ROC curve procedure (Fig 4), with AUC¼ 0.82 (0.76e0.88). The three methods used to choose the most appropriate threshold between NACþ and NACe converged on LNDthr¼10 mm, characterised by SNS¼82%, SPC¼72%, positive predictive value PPV¼84%, negative predictive value NPV¼69%, diagnostic accuracy DA¼79% and likelihood ratio for positive test LRþ¼2.9. Compared to the similar indication given by the frequency distributions, the ROC curve threshold has the added value of being validated by the quantitative estimate of AUC. The ROC curve for the lesion size at MRI confirmed the findings of the logistic regression against a predictor role for this variable: AUC¼0.60, corresponding to a poor discriminating ability. Fig 5 shows that the 79% diagnostic accuracy, which characterises the choice of LNDthr¼10 mm made on the conventional statistical considerations, remains essentially unvaried in the 5 mm interval around LNDthr, because of the opposite variations of sensitivity and specificity. When lowering the threshold from 10 to 5 mm, the sensitivity increases to 92% at the expenses of specificity, which decreases to 57%; when raising the threshold from 10 to 15 mm, sensitivity drops down to 69% favouring the increase in specificity to 77%. Representative images of LND assessment at MRI of NACe and NACþ cases are demonstrated in Figs 6 and 7.
Discussion In the present study, the distance between the tumour and the NAC evaluated at preoperative breast MRI was found to be the best parameter for discriminating NACþ from NACe patients, and therefore, for planning the most appropriate surgical treatment in patients needing nonconservative surgery. Differences were highly significant both at univariate analysis and at logistic regression analysis (p<0.0001); the odds ratio OR¼0.93, with 95% CI:
0.90e0.95, indicated a large value of LND as a protective factor. The discriminating ability of LND was rated good also by ROC curve analysis with an estimated AUC of 0.82 (0.76e0.88). The possible cut-offs for identifying patients with a higher likelihood of NAC tumoural involvement were found to be in the interval 5LND15 mm. A LND of 10 mm is the threshold to be chosen based on the statistical requirements of balancing sensitivity and specificity; however, in a clinical setting, it may often be preferable, or even necessary, to advantage one parameter over the other, as long as a good overall accuracy is achieved: so, lower cutoffs increase sensitivity (but reduce specificity), whereas higher cut-offs increase specificity (reducing sensitivity). In the present data, the absence of suspicious pathological enhancement at distances 10 mm from the nipple would have correctly identified 81% of patients without NAC tumoural involvement; the percentage rises to 93% assuming as cut-off 5 mm. These results are consistent with what was recently proposed by D’Alonzo et al.1 (retrospective series) and Ponzone et al.2 (prospective analysis): this agreement in findings across studies raises a strong argument in favour of a MRI LND cut-off to identify patients with a higher likelihood of NAC tumoural involvement lower than the previously proposed 20 mm distance.11 The measurement method used was not timeconsuming, and is routinely performed in the breast MRI reporting procedure at our Breast Imaging Service. Different LND measurement methods are reported in literature. In the study of Byon et al.,19 using a 3 T MRI system, the authors proposed to use the distances between nipple and chest wall and lesion and chest wall on axial images for an algebraic calculation, which also considered the distance between different axial sections on which the nipple and lesion were found. D’Alonzo et al.1 found the highest ROC curve values for measurements obtained on axial MIP images, using a 1.5 T scanner with a section thickness acquisition of 2.6 mm. In the study from Cho et al.20 the measurements were made on sagittal, non-MIP, images and they used a section thickness of 1 mm, as in the present study. In the present series, different methods for measuring LND were not compared as this was not within the scope of the present study, but, in the authors’ opinion, three-dimensional data sets from the first post-contrast dynamic study are best used with MPR images, from which different planes of the breast anatomy and structure can be obtained.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
6
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
Table 2 Tumour features at magnetic resonance imaging (MRI) and histopathological characteristics, subdivided in the two groups with or without nippleeareola complex (NAC) tumoural involvement. NACþ (%) MRI features Lesion type Mass 34/71 (47.9%) Non-mass enhancement 28/71 (39.4%) Mass/non-mass 9/71 (12.7%) Focality Unifocal diseases 46/71 (64.8%) Multifocal/multicentric 19/71 (26.8%) diseases Bilateral diseases 6/71 (8.5%) Background parenchymal enhancement Minimal 10/71 (14.1%) Mild 25/71 (35.2%) Moderate 25/71 (35.2%) Marked 11/71 (15.5%) Breast density Almost fatty 9/71 (12.7%) Scattered fibroglandular 26/71 (36.6%) Heterogeneous dense 25/71 (35.2%) Extremely dense 11/71 (15.5%) Pathological features Histology Ductal invasive 27/71 (38.0%) carcinomas Lobular histotype 25/71 (35.2%) carcinomas Ductal in situ 10/71 (14.1%) carcinomas (DCIS) Other invasive 9/71 (12.7%) Lymph node status Negative 43/71 (60.6%) Positive 28/71 (39.4%) ERa Negative 6/70 (8.6%) Positive 64/70 (91.4%) PRa Negative 27/70 (38.6%) Positive 43/70 (61.4%) HER2b Negative 50/67 (74.6%) Positive 17/67 (25.4%) Ki67 proliferation indexc Low proliferation 29/63 (46%) High proliferation 34/63 (54%)
NACe (%)
p-Value
66/129 (51.2%) 44/129 (34.1%) 19/129 (14.7%)
0.74
86/129 (66.7%) 36/129 (27.9%)
0.71
Figure 2 Box plot showing LND between NACþ and NACepatients (p<0.0001).
7/129 (5.4%) 18/129 45/129 51/129 15/129
(14.0%) (34.9%) (39.5%) (11.6%)
0.86
14/129 46/129 56/129 13/129
(10.9%) (35.7%) (43.4%) (10.1%)
0.82
59/129 (45.7%)
0.37
32/129 (24.8%)
0.16
23/129 (17.8%)
0.36
15/129 (11.6%)
0.99
89/129 (69.0%) 40/129 (31.0%)
0.30
11/123 (8.9%) 112/123 (91.1%)
0.99
38/123 (30.9%) 85/123 (69.1%)
0.35
84/112 (75%) 28/112 (25%)
0.90
56/109 (51.4%) 53/109 (48.6%)
0.60
Data are n/N (%). NACþ, nippleeareola complex tumoural involvement; NACe, no nippleeareola complex tumoural involvement; ER, oestrogen receptor; PR, progesterone receptor. a In NACe group for six in situ carcinomas ER and PgR are not available; in NACþ group for one in situ carcinoma ER and PgR are not available. b In NACe group for 17 carcinomas HER 2 results are not available; in NACþ group for four carcinomas HER 2 results are not available. c In NACe group for 20 carcinomas Ki67 results are not available; in NACþ group for eight carcinomas Ki67 results are not available.
The MRI-measured tumour diameter was also examined for a possible association with NAC tumoural involvement; this parameter was used in the past to select suitable patients for NSM, as an increase in the diameter of the primary tumour is associated with an increased risk for nipple involvement.3,4 Tumour size as assessed at MRI showed a much lower discriminating ability than LND, both at the logistic regression and at the ROC curve analysis.
Figure 3 Frequency distribution of LND for the two samples of patients with NACþ and NACe.
Figure 4 ROC curve: AUC ¼ 0.82 (0.76e0.88).
Interestingly, it was inversely correlated with the lesioneNAC distance. Breast dimensions could influence this finding and behave as confounding factor. As suggested for breast-conserving oncoplastic surgery, the surgical approach must be based not only on the location and size of the tumour, but also on the size and shape of the breast.21 In fact, for a non-conservative approach, the relative volume of a tumour compared to the breast volume could be considered a boundary value in planning the specific method of surgical treatment22: for this reason, a large lesion in a relatively small breast could more likely extend close to the NAC (and consequentially more likely involve it), while a
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
7
Figure 5 Sensitivity, specificity and diagnostic accuracy as a function of the choice for the LND threshold. The 79% diagnostic accuracy remains essentially unvaried in the 5 mm interval around the LND threshold (10 mm), which best satisfies the statistical requirements.
lesion of the same diameter in a larger breast has more space to develop without encroaching on the NAC. In future studies, the evaluation of breast dimensions and tumour/ breast volume ratios could provide interesting information on the actual relevance of tumour diameter among the factors that might be associated with NAC involvement. The currently available data appear to support that NSM could be performed safely for any tumour size, if there is no clinical or imaging evidence of NAC or skin involvement.23 More studies with long-term follow-up and with larger populations are required to ensure that the risk of local recurrence is minimised. Compared with previous papers,2,3,12,24 no associations between tumoural NAC involvement and other reported
risk factors, such as receptor status, DCIS component, or positive lymph node, were evidenced, re-enforcing the role of preoperative imaging in treatment planning; however, the study differs from other works10,24 reported in literature as it examined only patients without a previous identification or suspicion of clinical NAC abnormalities. Apart from imaging and histopathological findings, other aspects must be considered in the choice of the best surgical treatment: in fact, complications such as skin flap necrosis and partial or total NAC necrosis may occur after surgery, and different publications identified risk factors that must be carefully evaluated, such as smoking,25,26 age, radiotherapy after surgery, patient body mass index, or prosthesis weight.27 For this reason, multidisciplinary discussion
Figure 6 A 48-year-old woman who underwent preoperative breast MRI: (a) axial T1-weighted contrast enhanced and (b) subtracted images showing fibroglandular scattered breast tissue and minimal background parenchymal enhancement. In the left breast, outer quadrants, close to para-areolar region, an irregular 19-mm mass, biopsy-proven invasive lobular carcinoma, is noted; the LND (13 mm; green line) was measured using electronic callipers between the lesion border and the base of the nipple. Intraoperative subareolar tissue examination was negative for cancer and nipple-sparing mastectomy was performed. Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
8
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
Figure 7 A 43-year-old woman who underwent preoperative breast MRI: (a) axial and (b) sagittal T1-weighted contrast images showing heterogeneous dense breast tissue and moderate background parenchymal enhancement. In the right breast, inner lower quadrant, a 21-mm non-mass enhancement, biopsy-proven invasive ductal carcinoma with extensive intraductal component, is noted; the LND (9 mm; green line) was measured using electronic callipers between the lesion border and the base of the nipple. Intraoperative subareolar tissue examination was positive for cancer and a skin-sparing mastectomy was performed.
meetings, involving radiologists, pathologists, oncologists, and surgeons remain the most useful tool for the proper assessment of optimal treatment for every patient. Even though preoperative MRI proved to be an informative tool, intraoperative subareolar tissue frozen section still stands as the reference standard to confirm the absence of tumoural cells in patients who are candidates for NSM.28 In the literature, studies evaluating the accuracy of intraoperative subareolar frozen section are limited and falsenegative rates vary from 0.7%29 to 16.2%.30 Although some institutions favour the use of intraoperative subareolar frozen sections,30e32 others routinely rely on final subareolar histopathology results only.33,34 In a recent publication, Alperovich et al.28 demonstrated that subareolar intraoperative frozen sections offer reliable and relatively accurate results without any unwarranted NAC resections in their large series of patients, helping intraoperative decision making and guiding reconstructive planning. On the contrary, Chan et al.35 proposed that intraoperative NAC examination could even be avoided, as it is not cost-effective in patients with low risk of NAC involvement at preoperative imaging (defined by tumour-NAC distance >20 mm and tumour diameter 20 mm), relying on the high negative predictive value of MRI (98.4%) shown in their series. Even though this proposal may be of interest, to the authors’ knowledge and considering the results of the present study, intraoperative examination is still indicated and is mandatory for all patients undergoing NSM to confirm the absence of tumoural cells in the NAC. The present study includes the largest sample of patients without clinical evidence of NAC involvement; previous retrospective studies on the role of MRI in the evaluation of NAC from Cho et al.20 and Byon et al.19 had larger samples, but they also included patients with clinically evident NAC involvement. The present study has some limitations, namely its retrospective nature, which is in part addressed by having
all MRI examinations reviewed blinded to surgical and histopathological data; nonetheless, to the authors’ knowledge, the only prospective study published on this topic is by Ponzone et al.2 who reported similar specificity values for LND cut-offs. In addition, the present study is a single institution experience at an academic medical centre. An important limitation is that MRI findings were not compared with other conventional imaging (such as mammography). Although a previous study7 proposed an image-based model providing a NAC involvement score based on LND (based on mammography), histopathological stage, and tumour size with valuable sensitivity and specificity (92% and 77%, respectively), the role of breast MRI has been confirmed more recently to be important prior to consideration of NSM36 for identifying retroareolar tissue tumours with or without nipple involvement and predicting NAC involvement with higher sensitivity and specificity (up to 99.5% and 100%, respectively37). In conclusion, the present findings are encouraging and consistent with other available literature in supporting the role of preoperative breast MRI for the identification of NAC tumoural involvement.
Acknowledgements The authors thank Dr Alessandro De Leucio, Dr Giulia Negro, Dr Giovanni Cappello, and Dr Cristina Casella for their clinical and technical support to this work. N.H. receives support via a National Breast Cancer Foundation (NBCF, Australia) Breast Cancer Research Leadership Fellowship.
References 1. D’Alonzo M, Martincich L, Biglia N, et al. Clinical and radiological predictors of nippleeareola complex involvement in breast cancer patients. Eur J Cancer 2012;48:2311e8.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9 2. Ponzone R, Maggiorotto F, Carabalona S, et al. MRI and intraoperative pathology to predict nippleeareola complex (NAC) involvement in patients undergoing NAC-sparing mastectomy. Eur J Cancer 2015;51:1882e9. 3. Headon HL, Kasem A, Mokbel K. The oncological safety of nipple-sparing mastectomy: a systematic review of the literature with a pooled analysis of 12,358 procedures. Arch Plast Surg 2016;43:328e38. 4. Huang NS, Wu J. Nipple-sparing mastectomy in breast cancer: from an oncologic safety perspective. Chin Med J (Engl) 2015;128:2256e61. 5. Sakurai T, Zhang N, Suzuma T, et al. Long-term follow-up of nipplesparing mastectomy without radiotherapy: a single center study at a Japanese institution. Med Oncol 2013;30:481. 6. Loewen MJ, Jennings JA, Sherman SR, et al. Mammographic distance as a predictor of nippleeareola complex involvement in breast cancer. Am J Surg 2008;195:391e5. 7. Schecter AK, Freeman MB, Giri D, et al. Applicability of the nippleeareola complex-sparing mastectomy: a prediction model using mammography to estimate risk of nippleeareola complex involvement in breast cancer patients. Ann Plast Surg 2006;56:498e504. 8. Dent BL, Miller JA, Eden DJ, et al. Tumour-to-nipple distance as a predictor of nipple involvement: expanding the inclusion criteria for nipple-sparing mastectomy. Plast Reconstr Surg 2017;140:1ee8e. 9. Liao CY, Wu YT, Wu WP, et al. Role of breast magnetic resonance imaging in predicting malignant invasion of the nippleeareolar complex: potential predictors and reliability between inter-observers. Medicine (Baltimore) 2017;96:e7170. 10. Piato JR, de Andrade RD, Chala LF, et al. MRI to predict nipple involvement in breast cancer patients. AJR Am J Roentgenol 2016;206:1124e30. 11. Steen ST, Chung AP, Han SH, et al. Predicting nippleeareolar involvement using preoperative breast MRI and primary tumour characteristics. Ann Surg Oncol 2013;20:633e9. 12. Mallon P, Feron JG, Couturaud B, et al. The role of nipple-sparing mastectomy in breast cancer: a comprehensive review of the literature. Plast Reconstr Surg 2013;131:969e84. 13. Sardanelli F, Boetes C, Borisch B, et al. Magnetic resonance imaging of the breast: recommendations from the EUSOMA working group. Eur J Cancer 2010;46:1296e316. 14. Morris EA, Comstock CE, Lee CH, et al. ACR BI-RADS magnetic resonance imaging. ACR BI-RADSÒ atlas, breast imaging reporting and data system. Reston, VA: American College of Radiology; 2013. 15. Harrell F, Califf R, Pryor D, et al. Evaluating the yield of medical tests. JAMA 1982;247:2543e6. 16. Reid MC, Lane DA, Feinstein AR. Academic calculations versus clinical judgments: practicing physicians’ use of quantitative measures of test accuracy. Am J Med 1998;104:374e80. 17. Steurer J, Fischer JE, Bachmann LM, et al. Communicating accuracy of tests to general practitioners: a controlled study. BMJ 2002;324:824e6. 18. Youden WJ. Index for rating diagnostic tests. Cancer 1950;3:32e5. 19. Byon W, Kim E, Kwon J, et al. Magnetic resonance imaging and clinicopathological factors for the detection of occult nipple involvement in breast cancer patients. J Breast Cancer 2014;17:386e92. 20. Cho J, Chung J, Cha ES, et al. Can preoperative 3-T MRI predict nippleeareolar complex involvement in patients with breast cancer? Clin Imaging 2016;40:119e24.
9
21. Weber WP, Soysal SD, Fulco I, et al. Standardization of oncoplastic breast conserving surgery. Eur J Surg Oncol 2017;43:1236e43. 22. Lesar M, Stanec M, Banovi c M. Significance of the relative size of a breast tumour in deciding the method of surgical treatment. Tumouri 2006;92:18e20. 23. Coopey SB, Tang R, Lei L, et al. Increasing eligibility for nipple-sparing mastectomy. Ann Surg Oncol 2013;20. 3218e3122. 24. Billar JA, Dueck AC, Gray RJ, et al. Preoperative predictors of nippleeareola complex involvement for patients undergoing mastectomy for breast cancer. Ann Surg Oncol 2011;18:3123e8. 25. Frey JD, Alperovich M, Levine JP, et al. Does smoking history confer a higher risk for reconstructive complications in nipple-sparing mastectomy? Breast J 2017;23:415e20. 26. Orzalesi L, Casella D, Santi C, et al. Nipple sparing mastectomy: surgical and oncological outcomes from a national multicentric registry with 913 patients (1,006 cases) over a six-year period. Breast 2016;25:75e81. 27. Laporta R, Longo B, Sorotos M, et al. Breast reconstruction following nipple-sparing mastectomy: clinical outcomes and risk factors related complications. J Plast Surg Hand Surg 2017;20:1e9. 28. Alperovich M, Choi M, Karp NS, et al. Nipple-sparing mastectomy and sub-areolar biopsy: to freeze or not to freeze? Evaluating the role of sub-areolar intraoperative frozen section. Breast J 2016;22:18e23. 29. Jensen JA, Orringer JS, Giuliano AE. Nipple-sparing mastectomy in 99 patients with a mean follow-up of 5 years. Ann Surg Oncol 2011;18:1665e70. 30. Crowe JP, Patrick RJ, Yetman RJ, et al. Nipple-sparing mastectomy update: one hundred forty-nine procedures and clinical outcomes. Arch Surg 2008;143:1106e10. 31. De Alcantara Filho P, Capko D, Barry JM, et al. Nipple-sparing mastectomy for breast cancer and risk-reducing surgery: the Memorial SloanKettering Cancer Center experience. Ann Surg Oncol 2011;18:3117e22. 32. Petit JY, Veronesi U, Orecchia R, et al. Nipple sparing mastectomy with nipple areola intraoperative radiotherapy: one thousand and one cases of a five years’ experience at the European institute of oncology of Milan (EIO). Breast Cancer Res Treat 2009;117:333e8. 33. Peled AW, Foster RD, Stover AC, et al. Outcomes after total skin-sparing mastectomy and immediate reconstruction in 657 breasts. Ann Surg Oncol 2012;19:3402e9. 34. Spear SL, Rottman SJ, Seiboth LA, et al. Breast reconstruction using a staged nipple-sparing mastectomy following mastopexy or reduction. Plast Reconstr Surg 2012;129:572e81. 35. Chan SE, Liao CY, Wang TY, et al. The diagnostic utility of preoperative breast magnetic resonance imaging (MRI) and/or intraoperative subnipple biopsy in nipple-sparing mastectomy. Eur J Surg Oncol 2017;43:76e84. 36. Murthy V, Chamberlain RS. Defining a place for nipple sparing mastectomy in modern breast care: an evidence based review. Breast J 2013;19:571e81. 37. Friedman EP, Hall-Craggs MA, Mumtaz H, et al. Breast MR and the appearance of the normal and abnormal nipple. Clin Radiol 1997;52:854e61.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement G. Mariscotti a, *, M. Durando a, N. Houssami b, C.M. Berzovini a, F. Esposito a, M. Fasciano a, P.P. Campanino c, D. Bosco a, R. Bussone d, A. Ala d, I. Castellano e, A. Sapino f, L. Bergamasco g, P. Fonio a, G. Gandini a a
Department of Diagnostic Imaging and Radiotherapy, Radiology Institute, University of Turin, Turin, Italy Screening and Test Evaluation Program, School of Public Health, Sydney Medical School, University of Sydney, Sydney, 2006, NSW, Australia c Breast Imaging Service, Koelliker Hospital, Turin, Italy d della Salute e della Scienza di Torino, Presidio Ospedaliero Breast Surgery, Department of Surgery, A. O. U. Citta Molinette, Turin, Italy e della Salute e della Scienza di Torino, Department of Biomedical Sciences and Human Oncology, A. O. U. Citta Presidio Ospedaliero Molinette, Turin, Italy f Fondazione del Piemonte per l’Oncologia (FPO), Candiolo Cancer Institute (IRCCs), Turin, Italy g della Salute e della Scienza di Torino, Presidio Department of Surgical Sciences, University of Turin, A. O. U. Citta Ospedaliero Molinette, C.so Bramante 88, 10126, Torino, Italy b
art icl e i nformat ion Article history: Received 6 September 2017 Accepted 14 March 2018
AIM: To identify clinically occult nippleeareola complex (NAC) involvement using preoperative magnetic resonance imaging (MRI), to inform selection of patients eligible for nipplesparing mastectomy (NSM) or skin-sparing mastectomy (SSM). MATERIAL AND METHODS: This was a retrospective study of 195 patients, who had preoperative breast MRI (February 2011 to January 2017) before undergoing surgical treatments (NSM or SSM) for newly diagnosed breast cancer. Tumour features at MRI (mass or non-mass lesion, diameter, lesioneNAC distance [LND]) and pathology (lesion diameter, histopathological type, receptor status) were recorded, as well as the type of surgery (NSM/SSM) and presence (NACþ) or absence (NACe) of tumour at intraoperative evaluation of retroareolar tissue. ManneWhitney test, Fisher’s exact test, logistic regression, and receiver operating characteristic (ROC) curve analysis were used for analysis of NACþ versus NACe to assess variables that predict NAC tumoural involvement. RESULTS: Over the study period, NACþ was proven histologically in 71/200 (35.5%) surgical treatments, while there were 129/200 NACe (72 NSM and 128 SSM performed). LND at MRI was statistically (p<0.001) lower in NACþ patients than in NACe patients. The area under the ROC curve (0.82, 95% confidence interval [CI]: 0.76e0.88) indicated 10 mm as the best cut-off,
* Guarantor and correspondent: G. Mariscotti, Department of Diagnostic Imaging and Radiotherapy, Radiology Institute, University of Turin, A. O. U. Citt a della Salute e della Scienza di Torino, Presidio Ospedaliero Molinette, Via Genova 3, 10126, Torino, Italy. Tel.: þ39 0116336095; þ39 0116336622; fax: þ39 0116960310. E-mail address: [email protected] (G. Mariscotti). https://doi.org/10.1016/j.crad.2018.03.008 0009-9260/Ó 2018 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
2
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
with sensitivity of 82%, specificity of 72%, and accuracy of 79%. A 5-mm cut-off enhanced sensitivity, whereas a 15-mm cut-off favoured specificity. CONCLUSIONS: MRI is a useful tool for identifying NACþ patients; a 10-mm cut-off for LND assists selection of patients for NSM, although intraoperative retroareolar tissue examination remains mandatory. Ó 2018 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction Over the last 20 years, advances in breast surgery and reconstruction techniques have led to significant improvements in terms of quality of life and aesthetic results for patients with breast cancer. Despite much progress made also in diagnosing breast cancer at early stages of disease, breast-conserving surgery is not always a viable option, with up to one-third of patients still requiring (or at times requesting) mastectomy as definitive surgical treatment.1e3 In this setting, nipple-sparing mastectomy (NSM) has acquired a progressively increasing role in breast reconstructive approaches, pairing good cosmetic outcomes with generally good oncological outcomes. The main oncological concern with NSM is the possibility of leaving residual tumour cells in the nippleeareola complex (NAC), which may later manifest as local recurrence; this has been reported to range from 0% to 24% in recent studies,4 not statistically different from conventional mastectomy.3,5 Intraoperative histopathological examination of the subareolar tissue is in fact mandatory to exclude the presence of malignancy at this site, and consequently, to avoid a conversion to skin-sparing mastectomy (SSM).3,4 Recent reviews and meta-analyses underline the need for further long-term follow-up studies to determine the best reconstructive techniques, although several data already support the oncological safety of NSM for appropriately selected patients.3 The selection is, however, a crucial issue: the inclusion criteria for NSM treatment eligibility reported in the literature are quite variable, despite many pathological and radiological tumour features having been identified as predictive of NAC involvement. Conventional diagnostic imaging techniques, such as mammography and sonography, have been proposed for the evaluation of tumoural NAC involvement: specifically, it has been reported that the mammographic distance of the tumour to the nipple can be applied as an independent variable in a predictive equation that may assist the clinician in determining appropriate candidates for NSM6,7; however, mammography often fails to correctly evaluate the retroareolar lesion, because it is difficult to differentiate it from the normal nipple structures or from normal retroareolar glandular density (mainly in dense breasts).8,9 Sonography depicts nipples with posterior acoustic shadowing, so when a malignant mass is found, it might be difficult to evaluate the malignant invasion of the NAC.9 More recently, preoperative imaging with breast magnetic resonance imaging (MRI) has been shown to play a role for
predicting NAC involvement, particularly in the evaluation of lesionenipple distance (LND) and in the identification of pathological enhancements between the index lesion and the NAC.10 Currently, well circumscribed single or multifocal lesions with LND >20 mm are considered suitable for NSM3,11; however, a recent study by Ponzone et al.2 showed that a LND of 10 mm or even 5 mm can be a more specific (88.6% and 85%, respectively), though less sensitive (32.2% and 38.7%, respectively) cut-off for prediction of NAC tumoural involvement. Other criteria suggested in the literature for NSM (suggesting low likelihood of NAC tumoural involvement) include peripheral location of tumours, diameter <5 cm, and human epidermal growth factor 2 (HER2) negativity.3,12 This paper reports a retrospective study aimed at evaluating the role of MRI in the preoperative identification of otherwise clinically occult tumoural NAC involvement in breast cancer patients, to potentially inform surgical planning in this clinical setting. MRI findings are compared with tumour histopathological features and intraoperative retroareolar tissue sampling responses.
Materials and methods Study population The breast cancer records of the Breast Imaging Service of the Radiology Institute of the A. O. U. Citta della Salute e della Scienza of Turin (Italy) were reviewed retrospectively to identify all consecutive patients with newly diagnosed breast cancer who underwent preoperative breast MRI between February 2011 and January 2017 and were subsequently treated with mastectomy, either NSM or SSM. This retrospective study was approved by the local institutional ethics committee. The study sample was constructed using the following exclusion criteria: (1) patients whose preoperative MRI was not performed at our Breast Imaging Service, (2) patients with breast implants, (3) patients who underwent neoadjuvant chemotherapy, (4) patients with a previous history of breast cancer surgery (who had local recurrence during follow-up), (5) patients undergoing prophylactic mastectomy. As the aim of the study was to assess the ability of MRI to detect an otherwise occult tumoural involvement that could be recognised only through subareolar tissue histopathological examination, patients who showed presence of nipple involvement at clinical examination (nipple retraction, Paget’s disease of the nipple, or suspicious nipple
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
discharge) were also excluded. Based on these criteria, there were 195 eligible patients who had undergone 200 breast surgical treatments (Fig 1).
MRI examination MRI examinations were conducted according to the EUSOMA working group recommendations,13 using a 1.5 T MRI system (Philips, Achieva, Best, The Netherlands) with a dedicated seven-channel coil (Philips Sense breast 7). All patients signed an informed consent form before undergoing the MRI examination. The dynamic study was performed using a T1-weighted high-resolution isotropic volume examination sequence in the axial plane (Dyn-EThrive, 5.18 ms repetition time/2.52 ms echo time; 352352 matrix; 340340 mm field of view; 1 mm section thickness; 110.9 mm voxel size; acquisition time <70 seconds), acquired before and five times after intravenous contrast medium injection (0.1 mmol/kg dimeglumine gadobenate, Multihance, Bracco Imaging, Milano, Italia) at a
3
rate of 2 ml/s, followed by 20 ml saline flush. Before contrast medium injection, T1-weighted and T2-weighted time inversion recovery sequences were also acquired, as well as diffusion-weighted imaging (DWI), according to current recommendations. Subtracted images and multiplanar reconstructions (MPR) were derived from the dynamic study data-set.
MRI interpretation Two radiologists dedicated to breast imaging (with 12 and 5 years of experience in breast MRI, respectively) reviewed the set of images for every patient in consensus, and were blinded to clinical and pathological information. Any discrepancy was resolved by a third radiologist (with 7 years of experience in breast MRI). According to the BI-RADS MRI lexicon,14 each breast lesion was classified as focus, mass, or non-mass enhancement. Lesions were also categorised as unifocal, multifocal, and/or multicentric according to lesion number and location. MPR of three-dimensional images from the first postcontrast dynamic study were used to assess the maximum diameter of the lesion. For multifocal or multicentric lesions, the reference measure included the whole area occupied in the breast. On the first high-resolution, postcontrast acquisition (and related subtracted images), the LND was measured by electronic callipers as the shortest distance (in millimetres) between the lesion border and the base of the nipple, using the MPR images, more often on para-sagittal or para-axial planes. The maximum intensity projection (MIP) tool was used as an aid to better identify the lesion and suspicious satellite foci or associated in situ component, but not for the assessment of distance between NAC and the closest margin of pathological enhancement, that could be referred to the lesion.
Technique for intraoperative assessment of the subareolar tissue during NSM
Figure 1 Flow chart of patients’ selection in the present study.
NACesparing mastectomy was performed through an italic S incision, detaching the superior part of the breast by dissecting the Cooper ligaments first, and then removing the mammary gland along the pectoralis major fascia. The NAC was elevated just beneath the level of the deep dermis. Following breast removal, by nipple eversion the central ducts are transected at the base of the nipple. Surgical practice of NAC isolation was performed by hydro-dissection of the areola: 10e20 ml of cold saline solution was injected into the deep subareolar dermis to obtain complete detachment of the skin, then the areola was isolated by dissecting the swollen plane with scissors and the nipple could be cored without increasing the risk of ischaemic complications. The retroareolar tissue specimen was sent for intraoperative frozen section biopsies: if the tissue was positive for invasive or in situ carcinoma, the NAC was removed, thereby converting surgical treatment to SSM.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
4
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
Histopathological assessment of NAC involvement and subareolar tissue examination All NSM candidates underwent intraoperative subareolar tissue examination, according to the procedure described above. The NAC was considered involved by tumour whenever at least one of the samples at histopathological intraoperative examination revealed the presence of invasive ductal or lobular carcinoma cells, or ductal carcinoma in situ. Presence of in situ lobular neoplasia was not considered as criterion for NAC involvement. The subareolar tissue was subsequently reviewed at definitive histology to confirm the intraoperative diagnosis. Patients who were candidate for SSM did not undergo intraoperative subareolar tissue examination: the nipple was removed with the breast and examined at definitive histology. The NAC tumoural involvement criteria at definitive histology were the same as those adopted for intraoperative assessment.
Data collection Age, tumour features at MRI (number and type of lesions, diameter, LND), type of surgery performed (NSM or SSM), and histopathological information (tumour histology, lymph nodal status, intraoperative and definitive subareolar tissue examination results) including prognostic factors (oestrogen receptor [ER], progesterone receptors [PR], Ki67 proliferation index, and HER2B receptor presence] were recorded in a dedicated database.
Statistical analysis Continuous variables were checked for normality using the ShapiroeWilk W-test. Normality was rejected, so variables were reported as median and interquartile range (IQR; representing 25th percentile to 75th percentile). The nonparametric ManneWhitney test for two independent distributions was used for their comparisons. The consensus between the sets of MRI LND measurements of the two readers was estimated using Wilcoxon’s test and the intraclass correlation coefficient (ICC) and the chi2 coefficient. Linear correlations between two continuous variables were tested by Pearson’s linear coefficient r with standard error (SE) and relative probability p (r¼0). Categorical variables were expressed as counts and percentages and compared with the chi-square test (with Yates’s correction for 22 tables) or when more appropriate using Fisher’s exact test. Odds ratios (ORs) were reported with 95% confidence levels. Binary logistic regression was used to measure the relationship between NAC tumoural involvement (NACþ) and the various continuous and categorical dependent variables. The frequency distributions of the LND for the NACþ cases and cases with no tumoural NAC involvement (NACe) were used to compute the stratum (or Level) specific likelihood ratios (SSLR) defined as the ratio of the two frequencies at different levels.15e17
The frequency distributions were also the basis for computing the receiving operating characteristics (ROC) curve for measuring the discriminating ability of LND through the area under the curve (AUC). Sensitivity and Specificity at the various levels of the ROC curve were used to identify the most appropriate cut-off between presence or absence of NAC involvement through: (1) maximisation of the harmonic mean (HM) of sensitivity and specificity and of Youden’s index18 and (2) minimisation of the distance of the curve from the (0,1) upper left vertex. Statistical significance was set at two-tails p<0.05. Statplus: Mac version v6 (AnalystSoft, Walnut, CA, USA) was used for the analysis.
Results Tumour features at preoperative imaging and histopathological characteristics The patients’ age at diagnosis ranged from 27 to 76 years (median 51 years; IQR 46e62 years). Of all included patients, 132 had unifocal disease, 55 multifocal or multicentric disease, eight bilateral disease (five underwent bilateral mastectomy, and three unilateral mastectomy and contralateral wide excision) yielding a total of 200 surgical radical treatments (NSM or SSM) for newly diagnosed breast cancers. Overall, 72 NSM and 128 SSM were performed (36% and 64%, respectively), the latter being numerically prevalent in the first period of observation, with a progressively increasing number of NSM in more recent years. NACþ was histologically proven in 71/200 (35.5%) surgical treatments. There were no false-negative cases of the intraoperative subareolar tissue examination compared to the definitive histology. Table 1 reports the comparison between the two groups of NACþ and NACe patients considering age, LND measurements and tumour diameter at pathology. The consensus of the two readers on the MRI LND measurements was high (no statistical differences and ICC and chi2 close to 1), with the few discrepancies resolved by the third radiologist. The other tumour features at MRI and their histopathological characteristics, subdivided in the two groups NACþ and NACe, are reported and compared in Table 2. Among the continuous variables considered, univariate analysis evidenced significant differences between NACþ and NACe patients for LND (p<0.0001) and maximum tumour diameter at MRI (p¼0.003). The logistic regression analysis confirmed the significant differences for LND: p<0.0001, OR¼0.93, with 95% CI: 0.90e0.95, indicating a large value of LND as a protective factor, but not for the lesion size at MRI (p¼0.48; OR¼1.00 with 95%CI: 0.98e1.03). This variable was found to be inversely correlated with LND (r¼e0.32, SE¼0.05; p<0.0001). Fig 2 shows the box plot for LND.
LND threshold for NAC involvement The strong statistical significance of LND seen in the univariate analysis and logistic regression suggested a
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
5
Table 1 Patients’ age and lesion to nippleeareola complex (NAC) distance and lesion diameter at magnetic resonance imaging subdivided in the two groups with or without NAC tumoural involvement.
Age (years)a LesioneNAC distance (LND), mma Tumour diameter at MRI, mma a b
NACþ (n¼71)
NACe (n¼129)
Univariate p-value
Log regression p-value
53 (45e64) 0 (0e13)
51 (46e61) 20 (15e39)
0.50 <0.00001
<0.00001b
47 (27e60)
33 (23e47)
0.003
0.48
Data are expressed as median (IQR). OR¼0.92 (0.90e0.95).
possible discriminator role for this variable. Fig 3 shows the LND frequency distributions for the NACþ and NACe groups. They intersect between 10 and 15 mm, with NACþ cases prevailing at smaller distances (SSLR¼5.9 for LND¼5 mm, decreasing to 2 for LND¼10 mm level) and being overcome by NACe cases at larger distances (SSLR¼0.5 for LND¼15 mm level decreasing to 0.36 for NLD¼20 mm). The ability of LND to discriminate between NACþ and NACe was graded as good by the ROC curve procedure (Fig 4), with AUC¼ 0.82 (0.76e0.88). The three methods used to choose the most appropriate threshold between NACþ and NACe converged on LNDthr¼10 mm, characterised by SNS¼82%, SPC¼72%, positive predictive value PPV¼84%, negative predictive value NPV¼69%, diagnostic accuracy DA¼79% and likelihood ratio for positive test LRþ¼2.9. Compared to the similar indication given by the frequency distributions, the ROC curve threshold has the added value of being validated by the quantitative estimate of AUC. The ROC curve for the lesion size at MRI confirmed the findings of the logistic regression against a predictor role for this variable: AUC¼0.60, corresponding to a poor discriminating ability. Fig 5 shows that the 79% diagnostic accuracy, which characterises the choice of LNDthr¼10 mm made on the conventional statistical considerations, remains essentially unvaried in the 5 mm interval around LNDthr, because of the opposite variations of sensitivity and specificity. When lowering the threshold from 10 to 5 mm, the sensitivity increases to 92% at the expenses of specificity, which decreases to 57%; when raising the threshold from 10 to 15 mm, sensitivity drops down to 69% favouring the increase in specificity to 77%. Representative images of LND assessment at MRI of NACe and NACþ cases are demonstrated in Figs 6 and 7.
Discussion In the present study, the distance between the tumour and the NAC evaluated at preoperative breast MRI was found to be the best parameter for discriminating NACþ from NACe patients, and therefore, for planning the most appropriate surgical treatment in patients needing nonconservative surgery. Differences were highly significant both at univariate analysis and at logistic regression analysis (p<0.0001); the odds ratio OR¼0.93, with 95% CI:
0.90e0.95, indicated a large value of LND as a protective factor. The discriminating ability of LND was rated good also by ROC curve analysis with an estimated AUC of 0.82 (0.76e0.88). The possible cut-offs for identifying patients with a higher likelihood of NAC tumoural involvement were found to be in the interval 5LND15 mm. A LND of 10 mm is the threshold to be chosen based on the statistical requirements of balancing sensitivity and specificity; however, in a clinical setting, it may often be preferable, or even necessary, to advantage one parameter over the other, as long as a good overall accuracy is achieved: so, lower cutoffs increase sensitivity (but reduce specificity), whereas higher cut-offs increase specificity (reducing sensitivity). In the present data, the absence of suspicious pathological enhancement at distances 10 mm from the nipple would have correctly identified 81% of patients without NAC tumoural involvement; the percentage rises to 93% assuming as cut-off 5 mm. These results are consistent with what was recently proposed by D’Alonzo et al.1 (retrospective series) and Ponzone et al.2 (prospective analysis): this agreement in findings across studies raises a strong argument in favour of a MRI LND cut-off to identify patients with a higher likelihood of NAC tumoural involvement lower than the previously proposed 20 mm distance.11 The measurement method used was not timeconsuming, and is routinely performed in the breast MRI reporting procedure at our Breast Imaging Service. Different LND measurement methods are reported in literature. In the study of Byon et al.,19 using a 3 T MRI system, the authors proposed to use the distances between nipple and chest wall and lesion and chest wall on axial images for an algebraic calculation, which also considered the distance between different axial sections on which the nipple and lesion were found. D’Alonzo et al.1 found the highest ROC curve values for measurements obtained on axial MIP images, using a 1.5 T scanner with a section thickness acquisition of 2.6 mm. In the study from Cho et al.20 the measurements were made on sagittal, non-MIP, images and they used a section thickness of 1 mm, as in the present study. In the present series, different methods for measuring LND were not compared as this was not within the scope of the present study, but, in the authors’ opinion, three-dimensional data sets from the first post-contrast dynamic study are best used with MPR images, from which different planes of the breast anatomy and structure can be obtained.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
6
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
Table 2 Tumour features at magnetic resonance imaging (MRI) and histopathological characteristics, subdivided in the two groups with or without nippleeareola complex (NAC) tumoural involvement. NACþ (%) MRI features Lesion type Mass 34/71 (47.9%) Non-mass enhancement 28/71 (39.4%) Mass/non-mass 9/71 (12.7%) Focality Unifocal diseases 46/71 (64.8%) Multifocal/multicentric 19/71 (26.8%) diseases Bilateral diseases 6/71 (8.5%) Background parenchymal enhancement Minimal 10/71 (14.1%) Mild 25/71 (35.2%) Moderate 25/71 (35.2%) Marked 11/71 (15.5%) Breast density Almost fatty 9/71 (12.7%) Scattered fibroglandular 26/71 (36.6%) Heterogeneous dense 25/71 (35.2%) Extremely dense 11/71 (15.5%) Pathological features Histology Ductal invasive 27/71 (38.0%) carcinomas Lobular histotype 25/71 (35.2%) carcinomas Ductal in situ 10/71 (14.1%) carcinomas (DCIS) Other invasive 9/71 (12.7%) Lymph node status Negative 43/71 (60.6%) Positive 28/71 (39.4%) ERa Negative 6/70 (8.6%) Positive 64/70 (91.4%) PRa Negative 27/70 (38.6%) Positive 43/70 (61.4%) HER2b Negative 50/67 (74.6%) Positive 17/67 (25.4%) Ki67 proliferation indexc Low proliferation 29/63 (46%) High proliferation 34/63 (54%)
NACe (%)
p-Value
66/129 (51.2%) 44/129 (34.1%) 19/129 (14.7%)
0.74
86/129 (66.7%) 36/129 (27.9%)
0.71
Figure 2 Box plot showing LND between NACþ and NACepatients (p<0.0001).
7/129 (5.4%) 18/129 45/129 51/129 15/129
(14.0%) (34.9%) (39.5%) (11.6%)
0.86
14/129 46/129 56/129 13/129
(10.9%) (35.7%) (43.4%) (10.1%)
0.82
59/129 (45.7%)
0.37
32/129 (24.8%)
0.16
23/129 (17.8%)
0.36
15/129 (11.6%)
0.99
89/129 (69.0%) 40/129 (31.0%)
0.30
11/123 (8.9%) 112/123 (91.1%)
0.99
38/123 (30.9%) 85/123 (69.1%)
0.35
84/112 (75%) 28/112 (25%)
0.90
56/109 (51.4%) 53/109 (48.6%)
0.60
Data are n/N (%). NACþ, nippleeareola complex tumoural involvement; NACe, no nippleeareola complex tumoural involvement; ER, oestrogen receptor; PR, progesterone receptor. a In NACe group for six in situ carcinomas ER and PgR are not available; in NACþ group for one in situ carcinoma ER and PgR are not available. b In NACe group for 17 carcinomas HER 2 results are not available; in NACþ group for four carcinomas HER 2 results are not available. c In NACe group for 20 carcinomas Ki67 results are not available; in NACþ group for eight carcinomas Ki67 results are not available.
The MRI-measured tumour diameter was also examined for a possible association with NAC tumoural involvement; this parameter was used in the past to select suitable patients for NSM, as an increase in the diameter of the primary tumour is associated with an increased risk for nipple involvement.3,4 Tumour size as assessed at MRI showed a much lower discriminating ability than LND, both at the logistic regression and at the ROC curve analysis.
Figure 3 Frequency distribution of LND for the two samples of patients with NACþ and NACe.
Figure 4 ROC curve: AUC ¼ 0.82 (0.76e0.88).
Interestingly, it was inversely correlated with the lesioneNAC distance. Breast dimensions could influence this finding and behave as confounding factor. As suggested for breast-conserving oncoplastic surgery, the surgical approach must be based not only on the location and size of the tumour, but also on the size and shape of the breast.21 In fact, for a non-conservative approach, the relative volume of a tumour compared to the breast volume could be considered a boundary value in planning the specific method of surgical treatment22: for this reason, a large lesion in a relatively small breast could more likely extend close to the NAC (and consequentially more likely involve it), while a
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
7
Figure 5 Sensitivity, specificity and diagnostic accuracy as a function of the choice for the LND threshold. The 79% diagnostic accuracy remains essentially unvaried in the 5 mm interval around the LND threshold (10 mm), which best satisfies the statistical requirements.
lesion of the same diameter in a larger breast has more space to develop without encroaching on the NAC. In future studies, the evaluation of breast dimensions and tumour/ breast volume ratios could provide interesting information on the actual relevance of tumour diameter among the factors that might be associated with NAC involvement. The currently available data appear to support that NSM could be performed safely for any tumour size, if there is no clinical or imaging evidence of NAC or skin involvement.23 More studies with long-term follow-up and with larger populations are required to ensure that the risk of local recurrence is minimised. Compared with previous papers,2,3,12,24 no associations between tumoural NAC involvement and other reported
risk factors, such as receptor status, DCIS component, or positive lymph node, were evidenced, re-enforcing the role of preoperative imaging in treatment planning; however, the study differs from other works10,24 reported in literature as it examined only patients without a previous identification or suspicion of clinical NAC abnormalities. Apart from imaging and histopathological findings, other aspects must be considered in the choice of the best surgical treatment: in fact, complications such as skin flap necrosis and partial or total NAC necrosis may occur after surgery, and different publications identified risk factors that must be carefully evaluated, such as smoking,25,26 age, radiotherapy after surgery, patient body mass index, or prosthesis weight.27 For this reason, multidisciplinary discussion
Figure 6 A 48-year-old woman who underwent preoperative breast MRI: (a) axial T1-weighted contrast enhanced and (b) subtracted images showing fibroglandular scattered breast tissue and minimal background parenchymal enhancement. In the left breast, outer quadrants, close to para-areolar region, an irregular 19-mm mass, biopsy-proven invasive lobular carcinoma, is noted; the LND (13 mm; green line) was measured using electronic callipers between the lesion border and the base of the nipple. Intraoperative subareolar tissue examination was negative for cancer and nipple-sparing mastectomy was performed. Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
8
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9
Figure 7 A 43-year-old woman who underwent preoperative breast MRI: (a) axial and (b) sagittal T1-weighted contrast images showing heterogeneous dense breast tissue and moderate background parenchymal enhancement. In the right breast, inner lower quadrant, a 21-mm non-mass enhancement, biopsy-proven invasive ductal carcinoma with extensive intraductal component, is noted; the LND (9 mm; green line) was measured using electronic callipers between the lesion border and the base of the nipple. Intraoperative subareolar tissue examination was positive for cancer and a skin-sparing mastectomy was performed.
meetings, involving radiologists, pathologists, oncologists, and surgeons remain the most useful tool for the proper assessment of optimal treatment for every patient. Even though preoperative MRI proved to be an informative tool, intraoperative subareolar tissue frozen section still stands as the reference standard to confirm the absence of tumoural cells in patients who are candidates for NSM.28 In the literature, studies evaluating the accuracy of intraoperative subareolar frozen section are limited and falsenegative rates vary from 0.7%29 to 16.2%.30 Although some institutions favour the use of intraoperative subareolar frozen sections,30e32 others routinely rely on final subareolar histopathology results only.33,34 In a recent publication, Alperovich et al.28 demonstrated that subareolar intraoperative frozen sections offer reliable and relatively accurate results without any unwarranted NAC resections in their large series of patients, helping intraoperative decision making and guiding reconstructive planning. On the contrary, Chan et al.35 proposed that intraoperative NAC examination could even be avoided, as it is not cost-effective in patients with low risk of NAC involvement at preoperative imaging (defined by tumour-NAC distance >20 mm and tumour diameter 20 mm), relying on the high negative predictive value of MRI (98.4%) shown in their series. Even though this proposal may be of interest, to the authors’ knowledge and considering the results of the present study, intraoperative examination is still indicated and is mandatory for all patients undergoing NSM to confirm the absence of tumoural cells in the NAC. The present study includes the largest sample of patients without clinical evidence of NAC involvement; previous retrospective studies on the role of MRI in the evaluation of NAC from Cho et al.20 and Byon et al.19 had larger samples, but they also included patients with clinically evident NAC involvement. The present study has some limitations, namely its retrospective nature, which is in part addressed by having
all MRI examinations reviewed blinded to surgical and histopathological data; nonetheless, to the authors’ knowledge, the only prospective study published on this topic is by Ponzone et al.2 who reported similar specificity values for LND cut-offs. In addition, the present study is a single institution experience at an academic medical centre. An important limitation is that MRI findings were not compared with other conventional imaging (such as mammography). Although a previous study7 proposed an image-based model providing a NAC involvement score based on LND (based on mammography), histopathological stage, and tumour size with valuable sensitivity and specificity (92% and 77%, respectively), the role of breast MRI has been confirmed more recently to be important prior to consideration of NSM36 for identifying retroareolar tissue tumours with or without nipple involvement and predicting NAC involvement with higher sensitivity and specificity (up to 99.5% and 100%, respectively37). In conclusion, the present findings are encouraging and consistent with other available literature in supporting the role of preoperative breast MRI for the identification of NAC tumoural involvement.
Acknowledgements The authors thank Dr Alessandro De Leucio, Dr Giulia Negro, Dr Giovanni Cappello, and Dr Cristina Casella for their clinical and technical support to this work. N.H. receives support via a National Breast Cancer Foundation (NBCF, Australia) Breast Cancer Research Leadership Fellowship.
References 1. D’Alonzo M, Martincich L, Biglia N, et al. Clinical and radiological predictors of nippleeareola complex involvement in breast cancer patients. Eur J Cancer 2012;48:2311e8.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008
G. Mariscotti et al. / Clinical Radiology xxx (2018) 1e9 2. Ponzone R, Maggiorotto F, Carabalona S, et al. MRI and intraoperative pathology to predict nippleeareola complex (NAC) involvement in patients undergoing NAC-sparing mastectomy. Eur J Cancer 2015;51:1882e9. 3. Headon HL, Kasem A, Mokbel K. The oncological safety of nipple-sparing mastectomy: a systematic review of the literature with a pooled analysis of 12,358 procedures. Arch Plast Surg 2016;43:328e38. 4. Huang NS, Wu J. Nipple-sparing mastectomy in breast cancer: from an oncologic safety perspective. Chin Med J (Engl) 2015;128:2256e61. 5. Sakurai T, Zhang N, Suzuma T, et al. Long-term follow-up of nipplesparing mastectomy without radiotherapy: a single center study at a Japanese institution. Med Oncol 2013;30:481. 6. Loewen MJ, Jennings JA, Sherman SR, et al. Mammographic distance as a predictor of nippleeareola complex involvement in breast cancer. Am J Surg 2008;195:391e5. 7. Schecter AK, Freeman MB, Giri D, et al. Applicability of the nippleeareola complex-sparing mastectomy: a prediction model using mammography to estimate risk of nippleeareola complex involvement in breast cancer patients. Ann Plast Surg 2006;56:498e504. 8. Dent BL, Miller JA, Eden DJ, et al. Tumour-to-nipple distance as a predictor of nipple involvement: expanding the inclusion criteria for nipple-sparing mastectomy. Plast Reconstr Surg 2017;140:1ee8e. 9. Liao CY, Wu YT, Wu WP, et al. Role of breast magnetic resonance imaging in predicting malignant invasion of the nippleeareolar complex: potential predictors and reliability between inter-observers. Medicine (Baltimore) 2017;96:e7170. 10. Piato JR, de Andrade RD, Chala LF, et al. MRI to predict nipple involvement in breast cancer patients. AJR Am J Roentgenol 2016;206:1124e30. 11. Steen ST, Chung AP, Han SH, et al. Predicting nippleeareolar involvement using preoperative breast MRI and primary tumour characteristics. Ann Surg Oncol 2013;20:633e9. 12. Mallon P, Feron JG, Couturaud B, et al. The role of nipple-sparing mastectomy in breast cancer: a comprehensive review of the literature. Plast Reconstr Surg 2013;131:969e84. 13. Sardanelli F, Boetes C, Borisch B, et al. Magnetic resonance imaging of the breast: recommendations from the EUSOMA working group. Eur J Cancer 2010;46:1296e316. 14. Morris EA, Comstock CE, Lee CH, et al. ACR BI-RADS magnetic resonance imaging. ACR BI-RADSÒ atlas, breast imaging reporting and data system. Reston, VA: American College of Radiology; 2013. 15. Harrell F, Califf R, Pryor D, et al. Evaluating the yield of medical tests. JAMA 1982;247:2543e6. 16. Reid MC, Lane DA, Feinstein AR. Academic calculations versus clinical judgments: practicing physicians’ use of quantitative measures of test accuracy. Am J Med 1998;104:374e80. 17. Steurer J, Fischer JE, Bachmann LM, et al. Communicating accuracy of tests to general practitioners: a controlled study. BMJ 2002;324:824e6. 18. Youden WJ. Index for rating diagnostic tests. Cancer 1950;3:32e5. 19. Byon W, Kim E, Kwon J, et al. Magnetic resonance imaging and clinicopathological factors for the detection of occult nipple involvement in breast cancer patients. J Breast Cancer 2014;17:386e92. 20. Cho J, Chung J, Cha ES, et al. Can preoperative 3-T MRI predict nippleeareolar complex involvement in patients with breast cancer? Clin Imaging 2016;40:119e24.
9
21. Weber WP, Soysal SD, Fulco I, et al. Standardization of oncoplastic breast conserving surgery. Eur J Surg Oncol 2017;43:1236e43. 22. Lesar M, Stanec M, Banovi c M. Significance of the relative size of a breast tumour in deciding the method of surgical treatment. Tumouri 2006;92:18e20. 23. Coopey SB, Tang R, Lei L, et al. Increasing eligibility for nipple-sparing mastectomy. Ann Surg Oncol 2013;20. 3218e3122. 24. Billar JA, Dueck AC, Gray RJ, et al. Preoperative predictors of nippleeareola complex involvement for patients undergoing mastectomy for breast cancer. Ann Surg Oncol 2011;18:3123e8. 25. Frey JD, Alperovich M, Levine JP, et al. Does smoking history confer a higher risk for reconstructive complications in nipple-sparing mastectomy? Breast J 2017;23:415e20. 26. Orzalesi L, Casella D, Santi C, et al. Nipple sparing mastectomy: surgical and oncological outcomes from a national multicentric registry with 913 patients (1,006 cases) over a six-year period. Breast 2016;25:75e81. 27. Laporta R, Longo B, Sorotos M, et al. Breast reconstruction following nipple-sparing mastectomy: clinical outcomes and risk factors related complications. J Plast Surg Hand Surg 2017;20:1e9. 28. Alperovich M, Choi M, Karp NS, et al. Nipple-sparing mastectomy and sub-areolar biopsy: to freeze or not to freeze? Evaluating the role of sub-areolar intraoperative frozen section. Breast J 2016;22:18e23. 29. Jensen JA, Orringer JS, Giuliano AE. Nipple-sparing mastectomy in 99 patients with a mean follow-up of 5 years. Ann Surg Oncol 2011;18:1665e70. 30. Crowe JP, Patrick RJ, Yetman RJ, et al. Nipple-sparing mastectomy update: one hundred forty-nine procedures and clinical outcomes. Arch Surg 2008;143:1106e10. 31. De Alcantara Filho P, Capko D, Barry JM, et al. Nipple-sparing mastectomy for breast cancer and risk-reducing surgery: the Memorial SloanKettering Cancer Center experience. Ann Surg Oncol 2011;18:3117e22. 32. Petit JY, Veronesi U, Orecchia R, et al. Nipple sparing mastectomy with nipple areola intraoperative radiotherapy: one thousand and one cases of a five years’ experience at the European institute of oncology of Milan (EIO). Breast Cancer Res Treat 2009;117:333e8. 33. Peled AW, Foster RD, Stover AC, et al. Outcomes after total skin-sparing mastectomy and immediate reconstruction in 657 breasts. Ann Surg Oncol 2012;19:3402e9. 34. Spear SL, Rottman SJ, Seiboth LA, et al. Breast reconstruction using a staged nipple-sparing mastectomy following mastopexy or reduction. Plast Reconstr Surg 2012;129:572e81. 35. Chan SE, Liao CY, Wang TY, et al. The diagnostic utility of preoperative breast magnetic resonance imaging (MRI) and/or intraoperative subnipple biopsy in nipple-sparing mastectomy. Eur J Surg Oncol 2017;43:76e84. 36. Murthy V, Chamberlain RS. Defining a place for nipple sparing mastectomy in modern breast care: an evidence based review. Breast J 2013;19:571e81. 37. Friedman EP, Hall-Craggs MA, Mumtaz H, et al. Breast MR and the appearance of the normal and abnormal nipple. Clin Radiol 1997;52:854e61.
Please cite this article in press as: Mariscotti G, et al., Preoperative MRI evaluation of lesionenipple distance in breast cancer patients: thresholds for predicting occult nippleeareola complex involvement, Clinical Radiology (2018), https://doi.org/10.1016/j.crad.2018.03.008