A Predictive Tool to Estimate the Risk of Axillary Metastases in Breast Cancer Patients with Negative Axillary Ultrasound

Ann Surg Oncol (2014) 21:2229 2236 DOI 10.1245/s10434-014-3617-6 ORIGINAL ARTICLE BREAST ONCOLOGY A Predictive Tool to Estimate the Risk of Axillary Metastases in Breast Cancer Patients with Negative Axillary Ultrasound T. J. Meretoja, MD, PhD 1, P. S. Heikkilä, MD, PhD 2, A. S. Mansfield, MD 1, G. Cserni, MD, DSc, PhD 3,4, E. Ambrozay, MD 5, G. Boross, MD 6, J. Zgajnar, MD, PhD 7, A. Perhavec, MD, PhD 7, B. Gazic, MD 8, R. Arisio, MD, PhD 9, T. F. Tvedskov, MD, PhD 10, M.-B. Jensen, MSc 11, and M. H. K. Leidenius, MD, PhD 1 1 Breast Surgery Unit, Helsinki University Central Hospital, Helsinki, Finland; 2 Department of Pathology, Helsinki University Central Hospital, Helsinki, Finland; 3 Department of Pathology, Bács-Kiskun County Teaching Hospital, Kecskemét, Hungary; 4 Department of Pathology, University of Szeged, Szeged, Hungary; 5 Department of Radiology, Mamma ZRT, Bács-Kiskun County Teaching Hospital, Kecskemét, Hungary; 6 Department of Surgery, Bács-Kiskun County Teaching Hospital, Kecskemét, Hungary; 7 Department of Surgical Oncology, Institute of Oncology, Ljubljana, Slovenia; 8 Department of Pathology, Institute of Oncology, Ljubljana, Slovenia; 9 Department of Pathology, A.O. Citta della salute e della scienza-sant Anna Hospital, Turin, Italy; 10 Department of Breast Surgery, Copenhagen University Hospital, Copenhagen, Denmark; 11 Danish Breast Cancer Cooperative Group, Copenhagen University Hospital, Copenhagen, Denmark ABSTRACT Background. Sentinel node biopsy (SNB) is the gold standard in axillary staging in clinically node-negative breast cancer patients. However, axillary treatment is undergoing a paradigm shift and studies are being conducted on whether SNB may be omitted in low-risk patients. The purpose of this study was to evaluate the risk factors for axillary metastases in breast cancer patients with negative preoperative axillary ultrasound. Methods. A total of 1,395 consecutive patients with invasive breast cancer and SNB formed the original patient series. A univariate analysis was conducted to assess risk factors for axillary metastases. Binary logistic regression analysis was conducted to form a predictive model based on the risk factors. The predictive model was first validated internally in a patient series of 566 further patients and then externally in a patient series of 2,463 patients from four other centers. All statistical tests were two-sided. Electronic supplementary material The online version of this article (doi:10.1245/s10434-014-3617-6) contains supplementary material, which is available to authorized users. Ó Society of Surgical Oncology 2014 First Received: 6 January 2014; Published Online: 25 March 2014 T. J. Meretoja, MD, PhD e-mail: tuomo.meretoja@fimnet.fi Results. A total of 426 of the 1,395 (30.5 %) patients in the original patient series had axillary lymph node metastases. Histological size (P \ 0.001), multifocality (P \ 0.001), lymphovascular invasion (P \ 0.001), and palpability of the primary tumor (P \ 0.001) were included in the predictive model. Internal validation of the model produced an area under the receiver operating characteristics curve (AUC) of 0.731 and external validation an AUC of 0.79. Conclusions. We present a predictive model to assess the patient-specific probability of axillary lymph node metastases in patients with clinically node-negative breast cancer. The model performs well in internal and external validation. The model needs to be validated in each center before application to clinical use. Axillary lymph node dissection (ANLD) was the gold standard in the diagnosis and treatment of axillary lymph node metastases in breast cancer for decades. Because of a significant risk of morbidity, a few randomized trials were conducted, to evaluate whether omission of ALND is safe in low-risk breast cancer patients. 1 3 The results of these studies reported acceptable axillary recurrence rates and similar survival in patients with ALND or without any axillary surgery. 1 3 The encouraging results from these studies have been published after the introduction of sentinel node biopsy (SNB) and have perhaps not gained best possible attention

2230 T. J. Meretoja et al. in the breast cancer research community. However, SNB also is an invasive procedure associated not only with costs but also with a certain risk of long-time morbidity, although less so than ALND. 4 6 Moreover, preoperative axillary ultrasound is a routine procedure in many breast surgery units nowadays. Approximately 50 % of patients with axillary involvement can be identified preoperatively. 7 For these reasons, a new randomized trial has been launched in Milan to evaluate whether SNB can be safely omitted in patients with negative preoperative axillary ultrasound. 8 The results of this trial are still to come. In any case, omission of axillary surgery in breast cancer patients poses the problem of unknown lymph node status and thus may complicate the planning of adjuvant chemoand/or radiotherapy. For the aforementioned reasons, the purpose of this study was to identify risk factors for axillary metastases in breast cancer patients with negative axillary ultrasound and to construct a predictive model to evaluate the risk of axillary metastases in these patients. We further aim to validate our model internally with patients from our own unit and externally using patient series from other institutions. Such predictive model could prove valuable in weighing pros and cons of SNB in individual patients. METHODS Original Patient Series The original series of this study comprises of a consecutive series of 1,395 patients having SNB for primary, invasive breast cancer at the Breast Surgery Unit of Helsinki University Central Hospital (HUCH) from July 2010 to May 2012 (Table 1). Preoperative axillary ultrasound was performed in all patients and a fine-needle aspiration biopsy was acquired whenever suspicious nodes were encountered by ultrasound. General, while subjective, criteria for suspicious nodes in ultrasound included enlargement, structural abnormality, and cortical thickening. Negative axillary ultrasound thus includes patients with no suspicious nodes or with suspicious nodes but tumor-negative in fine-needle aspiration biopsy. Patients having had neoadjuvant treatment or having recurrent breast cancer were excluded from this study. Information on patients, procedures, tumors, and axillary status were collected from a prospectively gathered sentinel node database of the HUCH. The patients were grouped as axillary node-negative or -positive. Patients with isolated tumor cells (ITC) as their only axillary finding (pn0i?) were included in the node-negative group. Patient and tumor characteristics for these two groups are given in Table 1. Surgical techniques as well as pathological workup of the breast and axillary specimen were conducted according to normal clinical routine as described in detail earlier. 9 Shortly, all patients underwent either wide local excision or mastectomy. Patients with sentinel node metastases larger than 2 mm underwent routine ALND. Patients with sentinel node metastases not larger than 2 mm or those with ITC either underwent or avoided ALND according to risk of residual disease in the axilla, evaluated using our own previously published nomogram. 10 Altogether, 439 patients underwent SNB and completion ALND, whereas 936 underwent SNB only. Twenty patients with unsuccessful SNB underwent direct ALND of level I II axillary lymph nodes. Specialized breast pathologists examined the breast specimen. We also approached molecular subtypes of the tumors according to an immunohistochemistry-based surrogate classification. 11,12 Internal Validation Series A total of 566 further consecutive patients, with similar inclusion criteria to the original patient series, formed the internal validation series. These patients were operated between May 2012 and December 2012 at HUCH. Surgical treatment and pathology methodology were similar to the original patient series. External Validation Series Four European centers participated in this study and delivered patient information on a total of 2,463 patients to form the external validation series. The inclusion criteria were similar to the original patient series. The surgical treatment and pathology methodology were conducted according to each center s daily practice. The participating centers in the external validation were: Sant Anna Hospital, Turin, Italy (1,863 patients operated between 2003 and 2012), Bács-Kiskun County Teaching Hospital, Kecskemét, Hungary (200 patients operated between 2010 and 2012), Copenhagen University Hospital, Denmark (200 patients operated in 2011), Institute of Oncology, Ljubljana, Slovenia (200 patients operated in 2012). Statistical Analyses A univariate analysis was conducted to assess risk factors for axillary metastases. Distribution of continuous variables was analyzed using the Mann Whitney U test, and the v 2 test was used for categorical variables. All statistical tests were two-sided with P values \ 0.05 considered significant.

Risk of Axillary Metastases in Breast Cancer 2231 TABLE 1 Univariate analysis comparing patients with axillary lymph node metastases to those with no metastases in the original patient series Axillary lymph node metastases (n = 426) No axillary lymph node metastases (n = 969) All patients (n = 1,395) P Histological size of the primary tumor (mm) Median (range) 19 (4 80) 13 (1 150) 15 (1 150) Standard deviation 10.8 10.0 10.6 Histological tumor size by T-class pt1 (2 20 mm) 253 (24.5 %) 778 (75.5 %) 1,031 pt1a (1 5 mm) 3 (4.8 %) 60 (95.2 %) 63 pt1b (6 10 mm) 39 (12.0 %) 287 (88.0 %) 326 pt1c (11 20 mm) 211 (32.9 %) 431 (67.1 %) 642 pt2 (21 50 mm) 163 (47.0 %) 184 (53.0 %) 347 pt3 (51 mm or larger) 10 (58.8 %) 7 (41.2 %) 17 Body mass index 0.74 Median (range) 25.5 (16 56) 25.5 (12 55) 25.5 (12 56) Standard deviation 5.1 5.3 5.3 Patient age (year) \50 98 (44.3 %) 123 (55.7 %) 221 C50 328 (27.9 %) 846 (72.1 %) 1,174 Multifocality of the primary tumor No 310 (27.5 %) 816 (72.5 %) 1,126 Yes 116 (43.1 %) 153 (56.9 %) 269 Lymphovascular invasion in the primary tumor No 287 (24.8 %) 870 (75.2 %) 1,157 Yes 137 (61.7 %) 85 (38.3 %) 222 Missing 2 14 16 Estrogen receptor status 0.24 Negative 38 (26.2 %) 107 (73.8 %) 145 Positive 385 (31.0 %) 857 (69.0 %) 1,242 Missing 3 5 8 Progesterone receptor status 0.18 Negative 118 (28.0 %) 303 (72.0 %) 421 Positive 305 (31.6 %) 660 (68.4 %) 965 Missing 3 6 9 HER-2 status 0.22 Negative 370 (30.1 %) 861 (69.9 %) 1,231 Positive 56 (34.8 %) 105 (65.2 %) 161 Missing 0 3 3 Histological grade of the primary tumor Grade I 95 (22.5 %) 327 (77.5 %) 422 Grade II 187 (32.0 %) 398 (68.0 %) 585 Grade III 143 (37.6 %) 237 (62.4 %) 380 Missing 1 7 8 Histology of the primary tumor 0.04 Ductal carcinoma 300 (30.8 %) 673 (69.2 %) 973 Lobular carcinoma 75 (35.2 %) 138 (64.8 %) 213 Other 50 (24.0 %) 158 (76.0 %) 208 Missing 1 0 1 Biological subtype Luminal A 168 (25.3 %) 496 (74.7 %) 664 Luminal B 183 (39.0 %) 286 (61.0 %) 469

2232 T. J. Meretoja et al. TABLE 1 continued Axillary lymph node metastases (n = 426) No axillary lymph node metastases (n = 969) All patients (n = 1,395) P Luminal HER-2 34 (31.8 %) 73 (68.2 %) 107 HER-2 enriched 21 (39.6 %) 32 (60.4 %) 53 Triple negative 15 (18.5 %) 75 (81.5 %) 92 Missing 3 7 10 Palpability of the primary tumor Not palpable 77 (16.9 %) 378 (83.1 %) 455 Palpable 343 (37.6 %) 570 (62.4 %) 913 Missing 6 21 27 Localization of the primary tumor 0.001 Central 70 (40.7 %) 102 (59.3 %) 172 Upper medial 55 (22 %) 195 (78 %) 250 Lower medial 26 (25.5 %) 76 (74.5 %) 102 Upper lateral 202 (31.6 %) 437 (68.4 %) 639 Lower lateral 73 (32.4 %) 152 (67.6 %) 225 Missing 0 7 7 Reporting two-sided P values HER-2 human epidermal growth factor receptor 2 Variables with P values \0.1 in the univariate analysis were included in a binary logistic regression analysis using a backward stepwise likelihood ratio method. The logistic regression analysis was used as a basis for a predictive model, which included all variables with P values \0.05. The predictive model was validated both internally and externally by independent patient series. Area under the receiver operating characteristic curve (AUC) was used to evaluate the discrimination of the model and Hosmer Lemeshow goodness-of-fit test was used to assess the calibration of the model. Calibration was further characterized by plotting quintiles of predicted probability against actual probability of axillary metastases. Sensitivity and specificity of the model were calculated for various cutoff values. SPSS Statistics version 20 (SPSS Inc., Chicago, IL) software was used to conduct the statistical analyses. Ethical Considerations All patient information in this study was gathered anonymously and retrospectively with no influence on patient therapy. Institutional review boards and ethical committees were consulted in each center as required. RESULTS Of the 1,395 (30.5 %) patients in the original patient series, 426 had axillary lymph node metastases; 100 patients had micrometastasis as their largest lymph node metastasis and 326 patients had macrometastases. Of the 969 node-negative patients, 105 had ITC in their axillary lymph nodes. Factors associating with axillary metastases in the univariate analysis were patient age (P \ 0.001), histological size of the primary tumor (P \ 0.001), multifocality (P \ 0.001), lymphovascular invasion (P \ 0.001), histological grade (P \ 0.001), molecular subtype (P \ 0.001), histology of the tumor (P = 0.04), palpable tumor (P \ 0.001), and tumor localization (P = 0.001; Table 1). Histological size (P \ 0.001), multifocality (P \ 0.001), lymphovascular invasion (P \ 0.001), and palpability of the primary tumor (P \ 0.001) remained statistically significant in the multivariate logistic regression analysis and were included in the predictive model (Table 2). The model produced a P value of 0.41 for the Hosmer Lemeshow goodness-of-fit test indicating a good fit and calibration of the model. The AUC for the original patient series was 0.734 (95 % confidence interval [CI] 0.705 0.763), which suggests a good discrimination. A mathematical model was generated from the logistic regression analysis to predict the risk of axillary metastases, with p denoting the probability of axillary metastases: logitðpþ ¼ 2:260 þ 0:564 a þ 1:404 b þ 0:595 c þ 0:036 d: The letters in the equation denote the variables: a = multifocality of the primary tumor (1 if multifocal, 0 if unifocal); b = lymphovascular invasion in the primary

Risk of Axillary Metastases in Breast Cancer 2233 TABLE 2 Binary logistic regression analysis using backward stepwise likelihood ratio method in the original patient series Coefficient Wald P Odds ratio 95 % CI for odds ratio Lower Upper Multifocality of the primary tumor 0.564 13.564 1.758 1.302 2.374 Lymphovascular invasion in the primary tumor 1.404 74.866 4.070 2.962 5.594 Palpability of the primary tumor 0.595 14.106 1.812 1.329 2.472 Histological size of the primary tumor 0.036 29.025 1.037 1.023 1.051 Constant -2.26 204.537 0.104 CI confidence interval Reporting two-sided P values TABLE 3 Performance of the predictive model in internal and external validation N Preoperative axillary ultrasound performed Patients with axillary metastases AUC (95 % CI) Sensitivity Specificity Patients with predicted risk \30 % Original patient series 1,395 100 426 (30.5 %) With all information available 1,351 100 418 (30.9 %) 0.734 (0.705 0.763) 62.4 71 819 (60.6 %) Internal validation series 566 100 147 (26 %) 0.731 (0.682 0.78) 59.9 76.6 380 (67.1 %) External validation series 2,463 895 (36.3 %) 0.79 (0.772 0.809) 80.8 64 1175 (47.7 %) Center A (Turin) 1,863 75 724 (38.9 %) 0.801 (0.781 0.822) 84.8 62.1 817 (43.9 %) Center B (Kecskemét) 200 100 49 (24.5 %) 0.751 (0.672 0.83) 75.5 63.6 108 (54 %) Center C (Copenhagen) 200 100 59 (29.5 %) 0.717 (0.641 0.793) 52.5 78 138 (69 %) Center D (Ljubljana) 200 84 63 (31.5 %) 0.697 (0.623 0.772) 65.1 65.7 112 (56 %) AUC area under the receiver operating characteristics curve, CI confidence interval Sensitivity and specificity calculated for a cutoff value of \30 % risk estimate score tumor (1 if present, 0 if not); c = palpability of the primary tumor (1 if palpable, 0 if not); d = histological size of the primary tumor in mm. The predictive model is given as a supplementary Excel-file calculator (Supplementary data 1) and at the website of the breast surgery unit of HUCH http://www.hus.fi/breastsurgery/ predictivemodel. The model was validated by entering patient data from the internal and external validation series into the equation. 147 of the 566 (26.0 %) patients in the internal validation series and 895 of the In the external validation series, 2,463 (36.3 %) patients had axillary metastases. AUC were calculated for each patient series (Table 3). Calibration of the model for the original and the external validation series is plotted in Fig. 1. Sensitivity and specificity for each patient series are given for a cutoff value of less than 30 % risk estimate in Table 3. Sensitivity and specificity of the model for other cutoff values in the external validation center with the largest number of patients (N = 1863) are given in Table 4. Only one patient in the original series, none in the internal validation series, and three patients in the external validation series had a risk estimate of less than 10 % for axillary metastases. Actual Probability of Axillary Lymph Node Metastases 80% 70% 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% Predicted Probability of Axillary Lymph Node Metastases DISCUSSION Reference Line Original Patient Series External Validation Series FIG. 1 Calibration of the model in the original patient series and the external validation series Present Predictive Model The current study presents a predictive model for probability of axillary lymph node metastases in patients

2234 T. J. Meretoja et al. TABLE 4 Sensitivity and specificity of our model for various cutoff values for the largest external validation center Center A (Turin) N = 1,863 Sensitivity with breast cancer who are clinically node-negative with axillary ultrasound. The model performs well in internal validation and external validation in four European centers, in terms of AUC. Furthermore, the model is well calibrated in terms of Hosmer Lemeshow goodness-of-fit test. The proportion of patients with less than 10 % risk of axillary metastases was negligible in all patient series of this study. This finding is in agreement with a study evaluating risk factors for sentinel node metastases in a large patient series. 13 The proportion of node-positive patients in the ALND arm was 23 and 28 % in the two previous reports comparing ALND and observation, and it was approximately 30 % in most of the series entered into this study. 2,3 Therefore, we have reported the sensitivity and specificity of our model for a 30 % cutoff value in Table 3. However, our model provides a continuous patient-specific risk estimate and therefore is not tied to specific cutoff thresholds. This is an advantage of the model, because the clinical decision-making thresholds may well change in the future. Risk Factors in our Model Specificity Number of patients in the low-risk group \20 % risk 89.4 47.1 613 (32.9 %) \30 % risk 84.8 62.1 817 (43.9 %) \40 % risk 77.9 71.6 975 (52.3 %) \50 % risk 61.2 82 1,215 (65.2 %) Histological tumor size, multifocality, lymphovascular invasion, and palpability of the primary tumor were included in the predictive model, because they were found to be risk factors for axillary metastases in logistic regression analysis. All of these have been reported as risk factors for axillary metastases in previous studies. 13 15 We gave no instructions to the external validation centers for how to define or record the different variables in order to retain the external patient series as close to daily practice as possible. Multifocality of the primary tumor is defined in our original series and internal validation series as any tumor with histologically two or more separate foci irrespective of the distance between them. Similar definition of multifocality is used in the external validation centers B and D. External validation center A defines multifocality as two foci separated by 5 mm or more and center C as more than 20 mm of nonneoplastic parenchyma. Nonetheless, the model performed best in the external validation center A with a different definition of multifocality but the largest number of patients. Additionally, palpability of the primary tumor is a crude subjective measure, which may differ substantially according to examiner. Furthermore, core-needle biopsy may produce a palpable hematoma, which may incorrectly be interpreted as a palpable tumor. Therefore, the assessment of tumor palpability before core-needle biopsy is essential, and the tumors were generally defined as not palpable if there was any doubt of the finding. Preoperative Axillary Ultrasound Preoperative axillary ultrasound was performed in all patients of the original series and the internal validation series. Axillary ultrasound with needle biopsy of any suspicious nodes reveals preoperatively a substantial proportion of patients with axillary metastases and therefore decreases the probability of finding axillary metastases in axillary surgery. 7 Centers that do not perform preoperative axillary ultrasound thus have a larger proportion of clinically node-negative patients with axillary metastases found in SNB. Preoperative axillary ultrasound was performed at all of the external validation centers but not on all patients. However, the proportion of patients having had preoperative axillary ultrasound did not seem to affect the performance of the model. Surprisingly, the external validation center with the lowest proportion of preoperative axillary ultrasounds had the best performance of the model in terms of AUC (Center A in Table 3). Comparison to Previous Models To our knowledge, the study by Bevilacqua et al. from Memorial Sloan-Kettering Cancer Center (MSKCC) reports the most widely used prediction tool that provides a continuous risk estimate. 14 The patients in the MSKCC model have been operated in the beginning of sentinel node era between 1996 and 2002. Internal validation of the MSKCC model produced a similar AUC of 0.754 to our present model. A few other models use a nomogram or table of categories to estimate a risk of axillary metastases. 13,15 The MSKCC model included the same risk factors of tumor size, multifocality, and lymphovascular invasion as our model. In addition, the MSKCC model included patient age, tumor location, and estrogen and progesterone receptor status. The MSKCC modeling group of patients included tumor palpability, but the information was not gathered in the validation series and palpability was not included in the predictive model.

Risk of Axillary Metastases in Breast Cancer 2235 Implications of Our Model Our predictive model may have numerous clinical implications. The triple diagnosis, that is clinical examination, breast imaging, and image-guided needle biopsy may fail in some patients, especially in those with small tumors and therefore a low risk of axillary metastasis. 16 In these patients, breast cancer diagnosis is achieved by surgical biopsy and possible SNB would be performed as a second operation after the final histological analyses of the surgical biopsy specimen. Therefore, all the variables of our predictive model would be available before axillary surgery. Thus, the value of axillary surgery in these patients could be contemplated on the basis of an individual risk prediction, in a multidisciplinary breast cancer meeting, as well as with the patient herself. Both the recently launched Milan trial and the two aforementioned previous studies had or will have certain inclusion criteria, such as age older than 60 or 65 years, clinically T1N0 tumor, or breast conservation and radiotherapy. 2,3,8 The results of these studies cannot be directly applied to patients not fulfilling the inclusion criteria. Our model may offer a useful tool to estimate the possibility of omitting SNB, although the trial criteria were not met. Similarly, a patient with breast cancer might meet trial criteria preoperatively, but final histology may reveal risk factors that could warrant second stage axillary surgery. Finally, our model could even be used as a surrogate to surgical axillary staging. Staging the axilla by mathematically predicting the risk of axillary lymph node metastases may not be as accurate as SNB but may, nonetheless, be used in the planning of adjuvant chemo- and/or radiotherapy in patients without axillary surgery and hence an unknown lymph node status. Limitations of Our Model The value of our model in the preoperative decisionmaking is limited, because many risk factors, such as tumor size and multifocality as well as lymphovascular invasion, cannot be accurately assessed before surgery. Moreover, undetected axillary disease may have different influence on breast cancer recurrence and survival according to various patient and tumor related factors, such as age, tumor grade, or hormone-receptor status. In addition, systemic adjuvant treatment and radiotherapy both influence the clinical significance of undetected axillary disease. Our model is only able to predict the presence of axillary metastases but not their clinical significance, that is, whether they will remain occult or develop into clinical axillary recurrence or lead to systemic disease. CONCLUSIONS We present a predictive model to assess the patientspecific probability of axillary lymph node metastases in patients with clinically node-negative breast cancer. The model performs well in internal and external validation. The model needs to be validated in each center before application to clinical use. ACKNOWLEDGMENT Funding The first author of this study was supported by grants from the Academy of Finland, Orion-Farmos Research Foundation, Emil Aaltonen s Foundation, and Maud Kuistila Memorial Foundation. The funders did not have any involvement in the design of the study, the collection, analysis, and interpretation of the data, the writing of the article, or the decision to submit the article for publication. DISCLOSURE REFERENCES The authors have declared no conflicts of interest. 1. Veronesi U, Orecchia R, Zurrida S, et al. Avoiding axillary dissection in breast cancer surgery: a randomized trial to assess the role of axillary radiotherapy. Ann Oncol. 2005;16(3):383 8. 2. Rudenstam CM, Zahrieh D, Forbes JF, et al. Randomized trial comparing axillary clearance versus no axillary clearance in older patients with breast cancer: first results of International Breast Cancer Study Group Trial 10-93. J Clin Oncol. 2006;24(3): 337 44. 3. Martelli G, Boracchi P, Ardoino I, Lozza L, Bohm S, Vetrella G, Agresti S. Axillary dissection versus no axillary dissection in older patients with T1N0 breast cancer: 15-year results of a randomized controlled trial. Ann Surg. 2012;256(6):920 4. 4. Ashikaga T, Krag DN, Land SR, et al. Morbidity results from the NSABP B-32 trial comparing sentinel lymph node dissection versus axillary dissection. J Surg Oncol. 2010;102(2):111 8. 5. McLaughlin SA, Wright MJ, Morris KT, et al. Prevalence of lymphedema in women with breast cancer 5 years after sentinel lymph node biopsy or axillary dissection: patient perceptions and precautionary behaviors. J Clin Oncol. 2008;26(32):5220 6. 6. Gärtner R, Jensen MB, Nielsen J, Ewertz M, Kroman N, Kehlet H. Prevalence of and factors associated with persistent pain following breast cancer surgery. JAMA. 2009;302(18):1985 92. 7. Diepstraten S, Sever A, Buckens C, et al. Value of preoperative ultrasound-guided axillary lymph node biopsy for preventing completion axillary lymph node dissection in breast cancer: a systematic review and meta-analysis. Ann Surg Oncol. 2014; 21(1):51 9. 8. Gentilini O, Veronesi U. Abandoning sentinel lymph node biopsy in early breast cancer? A new trial in progress at the European Institute of Oncology of Milan (SOUND: Sentinel node vs Observation after axillary UltraSouND). Breast. 2012;21(5): 678 81. 9. Meretoja TJ, Leidenius MH, Heikkila PS, Joensuu H. Sentinel node biopsy in breast cancer patients with large or multifocal tumors. Ann Surg Oncol. 2009;16(5):1148 55. 10. Meretoja TJ, Strien L, Heikkilä PS, Leidenius MH. A simple nomogram to evaluate the risk of nonsentinel node metastases in breast cancer patients with minimal sentinel node involvement. Ann Surg Oncol. 2012;19(2):567 76.

2236 T. J. Meretoja et al. 11. Sørlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci USA. 2001;98(19): 10869 74. 12. Cheang MC, Chia SK, Voduc D, et al. Ki67 index, HER2 status, and prognosis of patients with luminal B breast cancer. J Natl Cancer Inst. 2009;101(10):736 50. 13. Viale G, Zurrida S, Maiorano E, et al. Predicting the status of axillary sentinel lymph nodes in 4351 patients with invasive breast carcinoma treated in a single institution. Cancer. 2005; 103(3):492 500. 14. Bevilacqua JL, Kattan MW, Fey JV, Cody HS, Borgen PI, Van Zee KJ. Doctor, what are my chances of having a positive sentinel node? A validated nomogram for risk estimation. J Clin Oncol. 2007;25(24):3670 9. 15. Reyal F, Rouzier R, Depont-Hazelzet B, et al. The molecular subtype classification is a determinant of sentinel node positivity in early breast carcinoma. PLoS One. 2011;6(5):e20297. 16. Leidenius MH, Vironen JH, von Smitten KA, Heikkilä PS, Joensuu H. The outcome of sentinel node biopsy in breast cancer patients with preoperative surgical biopsy. J Surg Oncol. 2009; 99(7):420 3.