Prediction of positive resection margins in patients with non-palpable breast cancer

Available online at www.sciencedirect.com

ScienceDirect EJSO 41 (2015) 106e112

www.ejso.com

Prediction of positive resection margins in patients with non-palpable breast cancer M.W. Barentsz a,*, E.L. Postma b, T. van Dalen c, M.A.A.J. van den Bosch a, H. Miao d, P.D. Gobardhan e, L.E. van den Hout a, R.M. Pijnappel a, A.J. Witkamp b, P.J. van Diest f, R. van Hillegersberg b, H.M. Verkooijen g a

Department of Radiology, University Medical Center Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands b Department of Surgery, University Medical Center Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands c Department of Surgery, Diakonessenhuis Utrecht, PO Box 80250, 3508 TG Utrecht, The Netherlands d Saw Swee Hock School of Public Health, National University of Singapore and National University Health System, MD 3, 16 Medical Drive, Singapore 117597, Singapore e Department of Surgery, Amphia Hospital Breda, Molengracht 21, 4818 CK Breda, The Netherlands f Department of Pathology, University Medical Center Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands g Imaging Division, University Medical Center Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands Accepted 24 August 2014 Available online 2 September 2014

Abstract Background: In patients undergoing breast conserving surgery for non-palpable breast cancer, obtaining tumour free resection margins is important to prevent reexcision and local recurrence. We developed a model to predict positive resection margins in patients undergoing breast conserving surgery for non-palpable invasive breast cancer. Methods: A total of 576 patients with non-palpable invasive breast cancer underwent breast conserving surgery in five hospitals in the Netherlands. A prediction model for positive resection margins was developed using multivariate logistic regression. Calibration and discrimination of the model were assessed and the model was internally validated by bootstrapping. Results: Positive resection margins were present in 69/576 (12%) patients. Factors independently associated with positive resection margins included mammographic microcalcifications (OR 2.14, 1.22e3.77), tumour size (OR 1.75, 1.20e2.56), presence of DCIS (OR 2.61, 1.41e4.82), Bloom and Richardson grade 2/3 (OR 1.82, 1.05e3.14), and caudal location of the lesion (OR 2.4, 1.35e4.27). The model was well calibrated and moderately able to discriminate between patients with positive versus negative resection margins (AUC 0.70, 95% CI, 0.63e0.77, and 0.69 after internal validation). Conclusion: The presented prediction model is moderately able to differentiate between women with high versus low risk of positive margins, and may be useful for surgical planning and preoperative patient counselling. Ó 2014 Elsevier Ltd. All rights reserved.

Keywords: Breast cancer; Non-palpable lesions; Tumour margins; Prediction model

Introduction * Corresponding author. Department of Radiology, Room E.01.132, University Medical Center Utrecht, Heidelberglaan 100, 3584 CX Utrecht, The Netherlands. Tel.: þ31 88 7556689; fax: þ31 30 2581098. E-mail address: [email protected] (M.W. Barentsz). http://dx.doi.org/10.1016/j.ejso.2014.08.474 0748-7983/Ó 2014 Elsevier Ltd. All rights reserved.

In breast conserving surgery, obtaining tumour free resection margins is essential for local control.1 Patients with tumour positive margins often need to undergo subsequent surgery. In the literature, reoperation rates range

M.W. Barentsz et al. / EJSO 41 (2015) 106e112

from 10.6 to 48%.1e4 Risk factors for positive margins include lobular histology, larger tumour size, presence of ductal carcinoma in situ (DCIS), lymphovascular invasion, non-palpability, multifocality, and presence of mammographic microcalcifications.5e9 These risk factors can be used to create a prediction model or nomogram (i.e. graphical presentation of a prediction model)10 which calculates individual probabilities for positive resection margins. Several prediction models have been developed for T1e2 breast cancer.7,11 These models include different prognostic variables and have moderate to good discriminative abilities (area under the ROC curve of 0.707 and 0.823.11 Since the introduction of population-based screening programs, early stage invasive breast cancer frequently presents as small and non-palpable lesions, which are particularly amenable to breast conserving surgery combined with radiotherapy. Non-palpable tumours are more likely to be resected with positive margins than palpable tumours (OR 1.51)7 and incomplete excisions are seen in up to 60% of patients.12e15 Until now, no model has been developed specifically for women with non-palpable invasive breast cancer. The aim of this study was to develop a prediction model for predicting positive resection margins after breast conserving surgery in patients with non-palpable invasive breast cancer and provide a short overview of the literature.

107

Variables The following variables were prospectively recorded: demographics, medical history, tumour location, palpability, type of surgery, data on perioperative procedures (e.g. method of localization, lymphoscintigraphy), tumour characteristics (i.e. modified Bloom and Richardson grade,17 hormonal receptor status, HER2 status since 2005 and histological subtype) and margin status. Additional data on radiological characteristics and missing histological information was collected through review of the original radiology and pathology reports. Breast density was scored visually by a dedicated breast radiologist. The main outcome measure of the study, i.e. resection margins after first breast conserving surgery, was defined as positive when invasive or in situ tumour cells were touching the inked surface of the resected specimen (e.g. focally positive margins were considered positive). Statistical analysis External validation existing nomogram Before model building, a recently published nomogram for predicting positive surgical margins after BCS was validated.7 All patients were entered in the model and the predicted probabilities of positive margins were calculated. A receiver operating characteristic (ROC) curve was fitted and the area under the curve (AUC) was obtained.

Patients and methods Study population Two prospectively acquired cohorts were used for analysis. The ethical principles of the Helsinki Declaration were followed and approval was obtained from the local ethics committees. The first cohort consisted of patients recruited in the context of a multicentre trial (ROLL trial) between December 2007 and April 2011 at four sites in the Netherlands; one university medical centre and three large community hospitals.16 Some 318 women (>18 years) with biopsy proven non-palpable breast cancer were randomized to either guidewire localization (GWL) or radioguided occult lesion localization (ROLL).16 In this randomized controlled multicentre trial, positive resection margins (in situ or invasive) were seen in 12.6% of patients. The second cohort was a hospital-based cohort from a large community hospital in the Netherlands and consisted of 1430 patients diagnosed with invasive breast cancer between 1999 and 2010. Some 258 patients had non-palpable histologically proven invasive breast cancer and were treated with breast conserving surgery.12 In this hospital based cohort, 29/258 (11.2%) patients had an incomplete tumour resection with invasive or in situ carcinoma present in the resection margins.

Model building Missing data analysis was performed to evaluate the amount and pattern of missingness (Supplement 1). Single conditional mean imputation was used for imputing missing values if values were missing at random (MAR) or missing not at random (MNAR) (instead of missing completely at random; MCAR).18 The association between the variables of interest and the outcome (positive resection margins) was univariately evaluated. Categorical variables were compared with the Chi square test. Normally distributed continuous variables were compared with independent t-tests, not-normally continuous variables with the ManneWhitney U test. All diagnostic variables with a p-value <0.20 in univariate analyses were entered in a logistic regression model. Odds ratios (OR) with 95% confidence intervals (CI) were calculated. Manual stepwise backward elimination of variables was performed and the optimal model fit was based on Akaike’s Information Criterion. Calibration of the final model was tested with the HosmereLemeshow test. Predicted probabilities were calculated and a corresponding ROC curve was derived. Discrimination of the model was calculated by the AUC. Model validation A bootstrap resampling procedure using 100 iterations was used to correct overfitting of the model (i.e. internal validation). Corrected coefficients were calculated by

108

M.W. Barentsz et al. / EJSO 41 (2015) 106e112

Table 1 Clinical and histological characteristics in relation to margin status of 576 patients undergoing breast conserving surgery for non-palpable invasive breast cancer. Characteristics

Age, years (mean  SD) Invasive tumour diameter, cm (mean  SD) Referral method National screening program Routine clinical examination Other Missing Presentation on mammography Mass lesion Mass lesion with microcalcifications Microcalcifications only Other Missing Breast density ACR 1 ACR 2 ACR 3 ACR 4 Missing Visible on ultrasound Yes No Missing Location within the breast Under Other Missing MRI performed Yes No Missing BIRADS classification II/III IV V Missing Histological type Ductal Lobular Other Missing DCIS component present on biopsy Yes No Missing Bloom and Richardson grade Grade 1 Grade 2 Grade 3 Missing Oestrogen receptor Positive Negative Missing

Negative margins (N ¼ 507)

Positive margins (N ¼ 69)

Values (%)

Values (%)

61.0  8.0 1.22  0.5

59.7  7.4 1.44  1.0

51 (73.9) 11 (15.9) 6 (8.7) 1 (1.4)

336 (66.3) 73 (14.4)

31 (44.9) 12 (17.4)

27 (5.3) 31 (6.1) 40 (7.9)

12 (17.4) 7 (10.1) 7 (10.1)

180 (35.5) 176 (34.7) 71 (14.0) 14 (2.8) 66 (13.0)

21 (30.4) 29 (42.0) 11 (15.9) 2 (2.9) 6 (8.7)

451 (89.0) 43 (8.5) 13 (2.6)

54 (78.3) 9 (13.0) 6 (8.7)

94 (18.5) 398 (78.5) 15 (3.0)

21 (30.4) 44 (63.8) 4 (5.8)

74 (14.6) 418 (82.4) 15 (3.0)

10 (14.5) 58 (84.1) 1 (1.4)

42 (8.3) 172 (33.9) 162 (32.0) 131 (25.8)

7 (10.1) 24 (34.8) 23 (33.3) 15 (21.7)

438 (86.4) 43 (8.5) 24 (4.7) 2 (0.4)

55 (79.7) 5 (7.2) 3 (4.3) 6 (8.7)

Characteristics

P-valuea

0.213 0.010 0.300

328 (75.3) 54 (10.7) 42 (8.3) 29 (5.7)

Table 1 (continued )

<0.001

0.719

0.152

0.013

0.942

0.751

Progesterone receptor Positive Negative Missing Her-2-Neu Positive Negative Missing Localization technique Guidewire Radioguided occult lesion localization Intraoperative ultrasound Total resected volume, cm3 (median; IQR)b

Negative margins (N ¼ 507)

Positive margins (N ¼ 69)

Values (%)

Values (%)

378 (74.6) 123 (24.3) 6 (1.2)

49 (71.0) 13 (18.8) 7 (10.1)

33 (6.5) 451 (89.0) 23 (4.5)

11 (15.9) 51 (73.9) 7 (10.1)

254 (50.1) 143 (28.2)

37 (53.6) 22 (31.9)

110 (21.7) 64.5 (42.3e94.2)

10 (14.5) 57.7 (43.0e90.8)

P-valuea

0.534

0.003

0.377

0.770

SD: standard deviation; DCIS: ductal carcinoma in situ; IQR: inter quartile range. This table is based on patient information before imputation was performed. a P-values are based on valid proportions and analysed by Chi square, independent t-test or ManneWhitney U where appropriate. b The total resected volume was calculated by the formula: 4/3p (½a∙½b∙½c), with a, b and c representing the three specimen dimensions, assuming an ellipsoidal shape of the excised specimen.

averaging all 100 bootstrap samples. Calibration of the shrunk model was evaluated by making a calibration plot. Discrimination of the model was assessed by an ROC curve. Finally, the corrected AUC was calculated (i.e. the average of all 100 AUCs). P-values of 0.05 or less were considered statistically significant. Imputation and data analyses were performed with SPSS (IBM SPSS Statistics, version 20.0) and R (version 2.11.1; R Foundation for Statistical Computing, Vienna, Austria) was used for bootstrapping and internal validation.19 Results

0.988

Patient characteristics

0.002 82 (16.2) 406 (80.1) 19 (3.7)

22 (31.9) 46 (66.7) 1 (1.4) 0.022

274 (54.0) 178 (35.1) 47 (9.3) 8 (1.6)

25 (36.2) 24 (34.8) 12 (17.4) 8 (11.6)

471 (92.9) 30 (5.9) 6 (1.2)

54 (78.3) 8 (11.6) 7 (10.1)

0.041

Of the 576 patients included, the mean age at diagnosis was 60.8 years (Table 1). The majority of patients (75.2%) were referred through the national screening program and 94.3% of the lesions were visible on mammography, whereas 87.7% were visible on ultrasound. Invasive ductal cancer was found in 493/576 (85.6%) patients and 48/576 (8.3%) patients had invasive lobular cancer. DCIS component was present in 104/576 (18.1%) patients. Final histological examination showed positive resection margins of either the invasive or in situ component in 69/576 (12%) patients. The majority of patients underwent GWL (291/576, 50.5%), 120/576 (20.8%) lesions were localized with IOUS, and 165/576 (28.7%) by means of ROLL. In the

M.W. Barentsz et al. / EJSO 41 (2015) 106e112

guidewire group, 30/291 (10.3%) patients presented with microcalcifications only and 44/291 (15.1%) presented with a mass lesion with associated microcalcifications. A total of 4/120 (3.3%) patients undergoing IOUS localization presented with microcalcifications only and 17/120 (14.2%) with a mass lesion with microcalcifications. In the ROLL group, 4/165 (2.4%) presented with microcalcifications only and 24/165 (14.5%) with a mass lesion with microcalcifications. External validation of existing nomogram The following variables were entered in the recently published nomogram: presence of microcalcifications, performance of MRI, clinical nodal status, clinical tumour size, breast density, ER status, presence of DCIS on biopsy, histological type, tumour grade and non-palpability. The AUC of the model was 0.617 (95% CI 0.542e0.693) (Fig. 1). Prediction model Univariate analyses showed an association ( p-value <0.20) between eight variables and positive resection margins. Presence of microcalcifications on mammography (P ¼ 0.004), non-visibility of the tumour on ultrasound (P ¼ 0.152), caudal location of the lesion within the breast (P ¼ 0.013), increasing tumour size (P ¼ 0.010), presence of DCIS (P ¼ 0.002), Bloom and Richardson grade 1 vs. 2/ 3 (P ¼ 0.040), oestrogen receptor negative (P ¼ 0.041) and HER2 positive tumours (P ¼ 0.003) were associated with an increased risk of positive resection margins. All variables were entered in the multivariate logistic regression model. After stepwise backward model reduction, the final model included five variables which were independently

109

and statistically associated with positive resection margins: microcalcifications on mammography (OR 2.14, 95% CI 1.22e3.77), invasive tumour size (per 1 cm increase) (OR 1.75, 95% CI 1.20e2.56), presence of DCIS component on biopsy (OR 2.61, 95% CI 1.41e4.82), Bloom and Richardson grade 2/3 (OR 1.82, 95% CI 1.05e3.14), and caudal location of the lesion within the breast (OR 2.40, 95% CI 1.35e4.27) (Table 2). The AUC of the final model was 0.70 (95% CI, 0.63e0.77). Discrimination of the model after bootstrap shrinking (i.e. internal validation) was slightly lower with an AUC of 0.69. When grouping patients in quintiles, the prevalence of positive margins was 29.6% in the highest quintile versus 4.2% in the lowest quintile (Table 3). Example of prediction In the current model, if all variables associated with positive resection margins (i.e. a tumour caudal in the breast, presenting with microcalcifications on mammography with DCIS present, B&R grade 2 or 3 and 3.0 cm in diameter) are present, the predicted probability of positive resection margins would be 75%. A patient with the most favourable profile (i.e. a tumour, not caudally located, no microcalcifications on mammography without DCIS, B&R grade 1 and 0.5 cm in size would have a 3.0% chance of having positive resection margins. In our study, the highest predicted probability of incomplete excision was 75% and the lowest predicted probability was 2.7%. Discussion We developed a novel prediction model for positive resection margins in patients with non-palpable breast cancer undergoing breast conserving surgery. This model is

Figure 1. Receiver operating characteristic curve showing the discriminative powers of this studies’ prediction model and the prediction model by Pleijhuis et al. (2013).

110

M.W. Barentsz et al. / EJSO 41 (2015) 106e112

Table 2 Multivariate logistic regression model for prediction of positive resection margins. Variables

Odds ratio

Microcalcifications on mammogram No 1.0 (ref) Yes 2.14 Presence of DCIS component on biopsy No 1.0 (ref) Yes 2.61 Bloom and Richardson grade 1 1.0 (ref) 2/3 1.82 Caudal location within breast No 1.0 (ref) Yes 2.40 Invasive tumour size (cm) 1.75

95% CI

P-value

1.22e3.77

0.008

1.41e4.82

0.002

1.05e3.14

0.032

1.35e4.27 1.20e2.56

0.003 0.004

moderately able to differentiate between women with high versus low risk of positive margins and may be useful for surgical planning and preoperative informing of patients. We found five variables to be significantly and independently associated with positive margins: microcalcifications on mammography, size of the invasive tumour, presence of DCIS component, B&R grade, and location of the lesion within the breast. These factors partially overlap with other prediction models for margin positivity, which were developed for both palpable and non-palpable breast cancer.7,11 We externally validated one of these models. The performance of this model in our dataset was lower (AUC 0.617, 95% CI 0.542e0.693) than the reported discrimination in the original derivation and validation sets to (AUC 0.70, 95% CI 0.66e0.74) (Table 4).7 The difference in discriminative power might be related to differences in patient populations, as reflected by the difference in a priori probability of having involved margins. In the study by Pleijhuis et al., 21.4% of patients with non-palpable lesion in the modelling group had positive margins and 28.7% in the validation cohort compared to 12% in our study. One other difference was that both palpable and non-palpable lesions were included in this study and the AUC was fitted for the total group of patients. The prediction model by Shin et al. was better able to discriminate between negative and positive margins than our model, with an AUC of 0.823 (95% CI 0.785e0.862). This model was based on 1034 patients with palpable and Table 3 Quintiles of predicted probability versus observed surgical margin status. Quintiles

Observed positive margins n/n (%)

Predicted positive margins % (range)

1 2 3 4 5

5/118 (4.2) 5/112 (4.5) 17/116 (14.7) 8/115 (7.0) 34/115 (29.6)

3.9 6.4 9.1 14.0 26.6

(2.7e4.6) (4.6e7.7) (7.7e10.4) (10.4e17.5) (17.7e74.7)

non-palpable invasive and in situ breast cancer undergoing breast conserving surgery in a single centre. Factors associated with incomplete resection were microcalcifications on mammography, breast density, >0.5 cm difference of tumour size between MRI and ultrasound, presence of DCIS and lobular histology. Despite the high discriminative abilities, the model was derived from and validated with single-centre data, which may impair generalizability. Also, the model included MRI data, which is not routinely available for all breast cancer patients. Some of the variables identified in our model and that of the other groups, are recognized as risk factors for resection margin positivity. Microcalcifications, lobular histology, lymphovascular invasion, large tumour size, presence of DCIS and multifocality have been identified as risk factors for incomplete resections.3,6,9,20e25 In the clinical management of T1e2 lobular breast cancer, for example, routine MRI is recommended according to the Dutch Breast Cancer Guidelines.26 When DCIS or microcalcifications are present, mammograms are more extensively investigated for signs of more extensive disease and multiple biopsies can be performed. Nevertheless, despite the known association between these factors and positive resection margins, patients with these factors still had a higher probability of having postoperative positive resection margins. Lovrics et al. reported that negative surgical margins are associated with larger excision volume, smaller tumour size, and palpability.27 Several other studies also found an association between the amount of removed tissue and lower rates of positive resection margins.28,29 However, more excised volume may lead to less favourable cosmetic results.30 In the current management, there seems to be a lack of correlation between the invasive tumour size and excised volume31 and in non-palpable breast cancer, this lack of association is even more present.32 In another study, unnecessarily large volumes of healthy breast tissue were excised while clear margins were not assured.33 Pleijhuis et al. found that non-palpability was associated with positive resection margins. In our model, increasing size of the invasive tumour was correlated with an increased probability of positive margins. This might be due to the fact that the non-palpable tumours are difficult to localize, so therefore harder to excise with free tumour margins. Increasing tumour size might be correlated to incomplete excision because of possible underestimation of the true tumour size and location within the breast, as reflected by our results. Finally, prognostic studies aiming to predict margin positivity may be useful in clinical practice, whilst determining the diagnostic and surgical strategy. For example, when preoperative findings indicate a high probability of incomplete excision a wider excision could performed. At the same time, applying models may lead to overtreatment, as in some patients more extensive tissue resection or even mastectomy may be unnecessary performed. A decision to perform a preoperative MRI could also be made

M.W. Barentsz et al. / EJSO 41 (2015) 106e112

111

Table 4 Literature overview of prediction models focussing on margin status after breast conserving surgery. Author Shin et al.

2012

Population

No. patients with positive margins (%)

Predictors of positive margins

OR (95% CI)

AUCa

Patients with invasive or in situ palpable and non-palpable breast cancer undergoing BCS

151/1034 (14.6%)

Microcalcifications on mammogram Breast density on mammogram Type 2 Type 3 Type 4 >0.5 cm difference MRI e ultrasound DCIS present on needle biopsy Lobular component on needle biopsy Suspicion of multifocal disease Preoperative MRI-scan absent Positive preoperative N-stage Non-palpable tumour Microcalcifications on mammogram Preoperative T2-stage Breast density on mammogram Presence of DCIS component Lobular histology ER positive Elston III grade Microcalcifications on mammogram Invasive tumour size Presence of DCIS component on biopsy Bloom and Richardson grade 2/3 Caudal location within breast

1.57 (1.04e2.39)

0.823m 0.846v

Pleijhuis et al.

2013

Patients with T1e2 palpable and non-palpable breast cancer undergoing BCS

233/1185 (19.7%)

Present study

2014

Patients with non-palpable breast cancer undergoing BCS

69/576 (12.0%)

1.59 1.61 4.52 10.0 1.58 3.99 2.81 1.80 1.73 1.51 1.37 1.33 1.22 3.11 2.90 1.80 1.44 2.14 1.75 2.61 1.82 2.40

(0.53e4.81) (0.56e4.62) (1.57e12.95) (6.50e15.39) (1.01e2.45) (1.31e12.12) (1.30e6.06) (1.02e3.18) (0.97e3.07) (1.07e2.13) (0.95e2.00) (0.87e2.02) (1.00e1.49) (2.19e4.42) (1.71e4.91) (1.04e3.13) (0.96e2.16) (1.22e3.77) (1.20e2.56) (1.41e4.82) (1.05e3.14) (1.35e4.27)

0.70m 0.69v

0.70m 0.69v

CI confidence interval, AUC area under the curve, BCS breast conserving surgery, DCIS ductal carcinoma in situ, ER oestrogen receptor. a The first presented AUC is from the modelling (m) cohort, the AUCs after validation (v) are presented secondly.

based on the preoperative knowledge on the chance of margin positivity. However, the randomized MONET-trial showed an increased re-excision rate in patients undergoing a preoperative MRI compared to patients not undergoing a preoperative MRI.34 Even though this study also included patients with DCIS only, care must be taken before applying changes to the management of breast cancer. Therefore, prospective studies are needed to measure the benefits and downsides of implementing new prediction rules for planning surgery of patient non-palpable breast lesions. In conclusion, a prediction model including five routinely available factors was built to predict positive resection margins following breast conserving surgery of non-palpable breast cancer. Our model correctly classifies almost 70% of all patients in terms of positive resection margins. Prospective studies and external validation of this model should be performed to evaluate the clinical usefulness. Acknowledgements We thank the ROLL study group and breast surgeons, pathologists and nurse practitioners from the Diakonessen Hospital Utrecht for their help in collecting the data. Conflict of interest statement None declared.

Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.ejso.2014.08.474. References 1. Park CC, Mitsumori M, Nixon A, et al. Outcome at 8 years after breast-conserving surgery and radiation therapy for invasive breast cancer: influence of margin status and systemic therapy on local recurrence. J Clin Oncol 2000;18:1668–75. 2. Jeevan R, Cromwell DA, Trivella M, et al. Reoperation rates after breast conserving surgery for breast cancer among women in England: retrospective study of hospital episode statistics. BMJ 2012;345: e4505. 3. Aziz D, Rawlinson E, Narod SA, et al. The role of reexcision for positive margins in optimizing local disease control after breastconserving surgery for cancer. Breast J 2006;12:331–7. 4. Talsma AK, Reedijk AM, Damhuis RA, Westenend PJ, Vles WJ. Reresection rates after breast-conserving surgery as a performance indicator: introduction of a case-mix model to allow comparison between Dutch hospitals. Eur J Surg Oncol 2011;37:357–63. 5. Keskek M, Kothari M, Ardehali B, Betambeau N, Nasiri N, Gui GP. Factors predisposing to cavity margin positivity following conservation surgery for breast cancer. Eur J Surg Oncol 2004;30:1058–64. 6. Kurniawan ED, Wong MH, Windle I, et al. Predictors of surgical margin status in breast-conserving surgery within a breast screening program. Ann Surg Oncol 2008;15:2542–9. 7. Pleijhuis RG, Kwast AB, Jansen L, et al. A validated web-based nomogram for predicting positive surgical margins following breastconserving surgery as a preoperative tool for clinical decision-making. Breast 2013;22:773–9.

112

M.W. Barentsz et al. / EJSO 41 (2015) 106e112

8. van der Heiden-van der Loo M, de Munck L, Visser O, et al. Variation between hospitals in surgical margins after first breast-conserving surgery in the Netherlands. Breast Cancer Res Treat 2012;131:691–8. 9. Chagpar AB, Martin 2nd RC, Hagendoorn LJ, Chao C, McMasters KM. Lumpectomy margins are affected by tumor size and histologic subtype but not by biopsy technique. Am J Surg 2004;188:399–402. 10. Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol 2008;26:1364–70. 11. Shin HC, Han W, Moon HG, et al. Nomogram for predicting positive resection margins after breast-conserving surgery. Breast Cancer Res Treat 2012;134:1115–23. 12. Barentsz MW, van Dalen T, Gobardhan PD, et al. Intraoperative ultrasound guidance for excision of non-palpable invasive breast cancer: a hospital-based series and an overview of the literature. Breast Cancer Res Treat 2012;135:209–19. 13. Barentsz MW, van den Bosch MA, Veldhuis WB, et al. Radioactive seed localization for non-palpable breast cancer. Br J Surg 2013; 100:582–8. 14. Tafra L, Fine R, Whitworth P, et al. Prospective randomized study comparing cryo-assisted and needle-wire localization of ultrasoundvisible breast tumors. Am J Surg 2006;192:462–70. 15. Mariscal Martinez A, Sola M, de Tudela AP, et al. Radioguided localization of nonpalpable breast cancer lesions: randomized comparison with wire localization in patients undergoing conservative surgery and sentinel node biopsy. AJR Am J Roentgenol 2009;193:1001–9. 16. Postma EL, Verkooijen HM, van Esser S, et al. Efficacy of ‘radioguided occult lesion localisation’ (ROLL) versus ‘wire-guided localisation’ (WGL) in breast conserving surgery for non-palpable breast cancer: a randomised controlled multicentre trial. Breast Cancer Res Treat 2012;136:469–78. 17. Elston CW, Ellis IO. Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up. Histopathology 1991;19:403–10. 18. van der Heijden GJ, Donders AR, Stijnen T, Moons KG. Imputation of missing values is superior to complete case analysis and the missingindicator method in multivariable diagnostic research: a clinical example. J Clin Epidemiol 2006;59:1102–9. 19. R Development Core Team. A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2010. http://www.lsw.uni-heidelberg.de/users/christlieb/ teaching/UKStaSS10/Rrefman.pdf. 20. Smitt MC, Horst K. Association of clinical and pathologic variables with lumpectomy surgical margin status after preoperative diagnosis or excisional biopsy of invasive breast cancer. Ann Surg Oncol 2007; 14:1040–4. 21. Dillon MF, Hill AD, Quinn CM, McDermott EW, O’Higgins N. A pathologic assessment of adequate margin status in breast-conserving therapy. Ann Surg Oncol 2006;13:333–9.

22. Dzierzanowski M, Melville KA, Barnes PJ, MacIntosh RF, Caines JS, Porter GA. Ductal carcinoma in situ in core biopsies containing invasive breast cancer: correlation with extensive intraductal component and lumpectomy margins. J Surg Oncol 2005;90:71–6. 23. Miller AR, Brandao G, Prihoda TJ, Hill C, Cruz Jr AB, Yeh IT. Positive margins following surgical resection of breast carcinoma: analysis of pathologic correlates. J Surg Oncol 2004;86:134–40. 24. Mai KT, Chaudhuri M, Perkins DG, Mirsky D. Resection margin status in lumpectomy specimens for duct carcinoma of the breast: correlation with core biopsy and mammographic findings. J Surg Oncol 2001;78:189–93. 25. Moore MM, Borossa G, Imbrie JZ, et al. Association of infiltrating lobular carcinoma with positive surgical margins after breastconservation therapy. Ann Surg 2000;231:877–82. 26. V. Dutch breast cancer guidelines. [accessed 11.04.13]. http://www. oncoline.nl/mammacarcinoom. 27. Lovrics PJ, Cornacchi SD, Farrokhyar F, et al. The relationship between surgical factors and margin status after breast-conservation surgery for early stage breast cancer. Am J Surg 2009;197:740–6. 28. Saarela AO, Rissanen TJ, Lahteenmaki KM, et al. Wire-guided excision of non-palpable breast cancer: determinants and correlations between radiologic and histologic margins and residual disease in reexcisions. Breast 2001;10:28–34. 29. Noh WC, Paik NS, Kim MS, et al. Ipsilateral breast tumor recurrence after breast-conserving therapy: a comparison of quadrantectomy versus lumpectomy at a single institution. World J Surg 2005;29: 1001–6. 30. Vrieling C, Collette L, Fourquet A, et al. The influence of patient, tumor and treatment factors on the cosmetic results after breastconserving therapy in the EORTC ‘boost vs. no boost’ trial. EORTC Radiotherapy and Breast Cancer Cooperative Groups. Radiother Oncol 2000;55:219–32. 31. Christiaens MR, Cataliotti L, Fentiman I, et al. Comparison of the surgical procedures for breast conserving treatment of early breast cancer in seven EORTC centres. Eur J Cancer 1996;32A:1866–75. 32. den Hartogh MD, van Asselen B, Monninkhof EM, et al. Excised and irradiated volumes in relation to the tumor size in breast-conserving therapy. Breast Cancer Res Treat 2011;129:857–65. 33. Krekel N, Zonderhuis B, Muller S, et al. Excessive resections in breast-conserving surgery: a retrospective multicentre study. Breast J 2011;17:602–9. 34. Peters NH, van Esser S, van den Bosch MA, et al. Preoperative MRI and surgical management in patients with nonpalpable breast cancer: the MONET e randomized controlled trial. Eur J Cancer 2011;47: 879–86.