Machine Learning Prediction of Lymph Node Metastasis in Breast Cancer: Performance of a Multi-institutional MRI-based 4D Convolutional Neural Network

Purpose To develop a custom deep convolutional neural network (CNN) for noninvasive prediction of breast cancer nodal metastasis. Materials and Methods This retrospective study included patients with newly diagnosed primary invasive breast cancer with known pathologic (pN) and clinical nodal (cN) status who underwent dynamic contrast-enhanced (DCE) breast MRI at the authors' institution between July 2013 and July 2016. Clinicopathologic data (age, estrogen receptor and human epidermal growth factor 2 status, Ki-67 index, and tumor grade) and cN and pN status were collected. A four-dimensional (4D) CNN model integrating temporal information from dynamic image sets was developed. The convolutional layers learned prognostic image features, which were combined with clinicopathologic measures to predict cN0 versus cN+ and pN0 versus pN+ disease. Performance was assessed with the area under the receiver operating characteristic curve (AUC), with fivefold nested cross-validation. Results Data from 350 female patients (mean age, 51.7 years ± 11.9 [SD]) were analyzed. AUC, sensitivity, and specificity values of the 4D hybrid model were 0.87 (95% CI: 0.83, 0.91), 89% (95% CI: 79%, 93%), and 76% (95% CI: 68%, 88%) for differentiating pN0 versus pN+ and 0.79 (95% CI: 0.76, 0.82), 80% (95% CI: 77%, 84%), and 62% (95% CI: 58%, 67%), respectively, for differentiating cN0 versus cN+. Conclusion The proposed deep learning model using tumor DCE MR images demonstrated high sensitivity in identifying breast cancer lymph node metastasis and shows promise for potential use as a clinical decision support tool. Keywords: MR Imaging, Breast, Breast Cancer, Breast MRI, Machine Learning, Metastasis, Prognostic Prediction Supplemental material is available for this article. Published under a CC BY 4.0 license.

Strict versus Liberal Use of Sentinel Node Biopsy in Breast Cancer Surgery: Any Clinical Outcome Differences? A 20-Year Clinical Experience

Introduction

As applied to early breast cancer (BC) patients, sentinel node biopsy (SNB) has undergone major changes over the years, especially concerning the widening of indication criteria or skipping systematic axillary lymph node dissection (ALND) after a positive SN. We aimed to ascertain whether a strict versus a more liberal use of SNB resulted in different clinical outcomes in our clinical experience.

Methods

We studied consecutive BC patients undergoing SNB between January 1, 2000, and March 31, 2020. There were 1,587 patients and 1,634 SNB procedures. Cases were divided into two study groups: the "strict" SNB group (unifocal tumors up to 35 mm in which ALND was always performed for a positive SN, amounting to 1,183 SNBs), and the "liberal" SNB group (extended tumor size up to selected T3 cases, as well as multifocal or bilateral disease, and patients with previous contralateral BC, not always followed by ALND after a positive SN, amounting to 451 SNBs). Patients were closely followed up to the end of the study.

Results

Clinico-pathological variables were strikingly different between study groups, with the liberal group showing a higher risk profile. Cox regression analysis for disease recurrence did not show significant differences in axillary, lymph node, or locoregional recurrence rates or distant relapse. There were no differences in survival between groups.

Conclusion

It seems reasonable to adopt the liberal SNB approach, as the goal of surgical management in early BC patients must be attaining optimal locoregional disease control, no matter the differences in distant metastatic spread rates across different BC risk profiles.

Novel nomograms for predicting survival for immediate breast reconstruction patients diagnosed with invasive breast cancer-a single-center 15-year experience

Background

Immediate breast reconstruction is widely accepted following oncologic mastectomy. This study aimed to build a novel nomogram predicting the survival outcome for Chinese patients undergoing immediate reconstruction following mastectomy for invasive breast cancer.

Methods

A retrospective review of all patients undergoing immediate reconstruction following treatment for invasive breast cancer was performed from May 2001 to March 2016. Eligible patients were assigned to a training set or a validation set. Univariate and multivariate Cox proportional hazard regression models were used to select associate variables. Two nomograms were developed based on the training cohort for breast cancer-specific survival (BCSS) and disease-free survival (DFS). Internal and external validations were performed, and the C-index and calibration plots were generated to evaluate the performance (discrimination and accuracy) of the models.

Results

The 10-year estimated BCSS and DFS were 90.80% (95% CI: 87.30%-94.40%) and 78.40% (95% CI: 72.50%-84.70%), respectively, in the training cohort. In the validation cohort, they were and 85.60% (95% CI, 75.90%-96.50%) and 84.10% (95% CI, 77.80%-90.90%), respectively. Ten independent factors were used to build a nomogram for prediction of 1-, 5- and 10-year BCSS, while nine were used for DFS. The C-index was 0.841 for BCSS and 0.737 for DFS in internal validation, and the C-index was 0.782 for BCSS and 0.700 for DFS in external validation. The calibration curve for both BCSS and DFS demonstrated acceptable agreement between the predicted and actual observation in the training and the validation cohorts.

Conclusion

The nomograms provided valuable visualization of factors predicting BCSS and DFS in invasive breast cancer patients with immediate breast reconstruction. The nomograms may have tremendous potential in guiding individualized decision-making for physicians and patients in choosing the optimized treatment methods.

Reporting of Surgically Removed Lymph Nodes for Breast Tumors: Recommendations From the International Collaboration on Cancer Reporting

CONTEXT.—

The International Collaboration on Cancer Reporting (ICCR), supported by major pathology and cancer organizations, aims at the standardization of evidence-based pathology reporting of different types of cancers, with the inclusion of all parameters deemed to be relevant for best patient care and future data collection. Lymph node metastasis is one of the most important prognostic factors in breast cancer.

OBJECTIVE.—

To produce a histopathology reporting guide by a panel of recognized experts from the fields of pathology and surgery with elements deemed to be core (required) and noncore (recommended) to report when assessing regional lymph nodes of patients with breast cancer.

DATA SOURCES.—

Published literature, previous guidelines/recommendations, and current cancer staging principles were the basis of the data set drafted by the expert panel. This was discussed in a series of teleconferences and email communications. The draft data set was then made available for public consultation through the ICCR Web site. After this consultation and ICCR ratification, the data set was finalized.

CONCLUSIONS.—

The ICCR has published a data set for the reporting of surgically removed lymph nodes (including sentinel lymph node biopsy, axillary lymph node dissection, targeted axillary surgery, and lymph node sampling specimens) for breast tumors. This is part of a series of 4 ICCR breast cancer-related data sets. It includes 10 core elements along with 2 noncore elements. This should allow for synoptic reporting, which is more precise, uniform, and complete than nonsynoptic reporting, and leads to improved patient outcomes.

Development and validation of nomograms for predicting axillary non-SLN metastases in breast cancer patients with 1-2 positive sentinel lymph node macro-metastases: a retrospective analysis of two independent cohorts

BACKGROUND

It is reported that appropriately 50% of early breast cancer patients with 1-2 positive sentinel lymph node (SLN) micro-metastases could not benefit from axillary lymph node dissection (ALND) or breast-conserving surgery with whole breast irradiation. However, whether patients with 1-2 positive SLN macro-metastases could benefit from ALND remains unknown. The aim of our study was to develop and validate nomograms for assessing axillary non-SLN metastases in patients with 1-2 positive SLN macro-metastases, using their pathological features alone or in combination with STMs.

METHODS

We retrospectively reviewed pathological features and STMs of 1150 early breast cancer patients from two independent cohorts. Best subset regression was used for feature selection and signature building. The risk score of axillary non-SLN metastases was calculated for each patient as a linear combination of selected predictors that were weighted by their respective coefficients.

RESULTS

The pathology-based nomogram possessed a strong discrimination ability for axillary non-SLN metastases, with an area under the receiver operating characteristic (ROC) curve (AUC) of 0.727 (95% CI: 0.682-0.771) in the primary cohort and 0.722 (95% CI: 0.653-0.792) in the validation cohort. The addition of CA 15-3 and CEA can significantly improve the performance of pathology-based nomogram in the primary cohort (AUC: 0.773 (0.732-0.815) vs. 0.727 (0.682-0.771), P < 0.001) and validation cohort (AUC: (0.777 (0.713-0.840) vs. 0.722 (0.653-0.792), P < 0.001). Decision curve analysis demonstrated that the nomograms were clinically useful.

CONCLUSION

The nomograms based on pathological features can be used to identify axillary non-SLN metastases in breast cancer patients with 1-2 positive SLN. In addition, the combination of STMs and pathological features can identify patients with patients with axillary non-SLN metastases more accurately than pathological characteristics alone.

Associated Features with Non-Sentinel Lymph Node Involvement in Early Stage Breast Cancer Patients who Have Positive Macrometastatic Sentinel Lymph Node

Objective

The main goal of this study is to determine the clinico-pathological factors that correlate non-sentinel lymph nodes (LNs) involvement in clinically node negative breast cancer (BC) patients with positive macrometastatic sentinel lymph node (SLN) in order to derive future evidence to define a subgroup where completion axillary lymph node dissection (cALND) might not be recommended.

Materials and Methods

Total 289 SLN biopsies were performed in clinically node negative BC patients between March 2014 and April 2017. Seventy patients who performed cALND due to positive macrometastatic SLN were retrospectively selected and classified into two groups, according to non-SLN involvement (NSLNI). Clinico-pathological features of patients were examined computerized and documentary archives.

Results

Extracapsular extension (ECE) of SLN, number of harvested SLNs, metastatic rate of SLNs, absence of ductal carcinoma in situ (DCIS) and presence of multilocalization were significantly associated with the likelihood of non-SLN involvement after univariate analysis (p<0,05). Absence of DCIS and presence of multilocalization were found to be significant after multivariate analysis.

Conclusion

Careful examination of clinico-pathological features can help to decide avoiding cALND if enough LNs are removed and the rate of SLN metastases is low, particularly in case DCIS accompanying invasive cancer in patients without multi localized tumour.

Different outcome in node-positive breast cancer patients found by axillary ultrasound or sentinel node procedure

Background

The Z0011 trial initiated a paradigm shift in the axillary treatment of breast cancer patients with a positive sentinel lymph node biopsy (SLNB), disregarding patients with a positive ultrasound-guided lymph node biopsy (UGLNB). We examined whether relevant differences exist between these patients to determine if the conclusions of the ACOSOG Z0011 trial are applicable to UGLNB-positive patients.

Methods

Patients diagnosed with invasive breast cancer in the Netherlands between January 2008 and December 2014 were selected from the Netherlands Cancer Registry.

Results

A total of 11,820 cases were included: 9149 cases in the SLNB group and 2671 in the UGLNB group. Multivariate analyses showed that UGLNB-positive patients were older (p < 0.001), more likely to have a poorly differentiated tumor (p < 0.001), had a negative hormone receptor status (p < 0.001), and more often had extensive nodal involvement (p < 0.001). However, they were less likely to undergo adjuvant radiation (p = 0.004) or systemic therapy (p < 0.001). Even after adjusting for these factors, UGLNB-positive patients had a worse overall survival (HR = 1.38; 95% CI 1.23–1.56) than SLNB-positive patients.

Conclusion

This nationwide retrospective study shows that young patients found positive by UGLNB have less favorable disease characteristics and a worse prognosis compared to patients with a positive SLNB. Selection by ultrasound plays an important role when axillary treatment strategies are considered. Hence, the conclusions of the Z0011 trial cannot unconditionally be applied to patients with a positive UGLNB.

Effects of a regional guideline for completion axillary lymph node dissection in women with breast cancer to reduce variation in surgical practice: A qualitative study of physicians' views

BACKGROUND

Recently the impact of completion axillary lymph node dissection (cALND) after positive sentinel lymph node biopsy on significant outcomes has been questioned, leading to variation in surgical practice. To address this variation, a multidisciplinary working group created a regional guideline for cALND. We explored the views and experiences of surgeons, medical oncologists (MOs), radiation oncologists (ROs) in a qualitative study that examined guideline implementation in practice.

METHODS

The Pathman framework (awareness, agreement, adoption and adherence) informed the interview guide design and analysis. Semi-structured interviews were conducted with MOs, ROs and surgeons and transcribed. Transcripts were coded independently by 2 members of the study team and analyzed. Disagreements were resolved through consensus.

RESULTS

Twenty-eight physicians (5 MO; 6RO; 17S) of 41 (68% of those approached) were interviewed. Ten of 11 (91%) hospital sites (54% community; 46% academic) and all 4 cancer clinics within the region were represented. Twenty-seven physicians (96%) were aware of the guideline, with all physicians reporting agreement and general adherence to the guideline. Most physicians indicated nodal factors, age and patient preference were key components of cALND decision-making. Physicians from all disciplines perceived that the guideline helped reduce variation in practice across the region. There were concerns that the guideline could be applied rigidly and not permit individual decision-making.

CONCLUSIONS

Physicians identified breast cancer as an increasingly complex and multidisciplinary issue. Facilitators to guideline implementation included perceived flexibility and buy-in from all disciplines, while individual patient factors and controversial supporting evidence may hinder its implementation.

Factors Influencing Non-sentinel Node Involvement in Sentinel Node Positive Patients and Validation of MSKCC Nomogram in Indian Breast Cancer Population

Current guidelines recommend completion axillary lymphnode dissection (ALND) when sentinel lymphnode (SLN) contains metastatic tumor deposit. In consequent ALND sentinel node is the only node involved by tumor in 40-70 % of cases. Recent studies demonstrate the oncologic safety of omitting completion ALND in low risk patients. Several nomograms (MSKCC, Stanford, MD Anderson score, Tenon score) had been developed in predicting the likelihood of additional nodes metastatic involvement. We evaluated accuracy of MSKCC nomogram and other clinicopathologic variables associated with additional lymph node metastasis in our patients. A total of 334 patients with primary breast cancer patients underwent SLN biopsy during the period Jan 2007 to June 2014. Clinicopathologic variables were prospectively collected. Completion ALND was done in 64 patients who had tumor deposit in SLN. The discriminatory accuracy of nomogram was analyzed using Area under Receiver operating characteristic curve (ROC). SLN was the only node involved with tumor in 69 % (44/64) of our patients. Additional lymph node metastasis was seen in 31 % (20/64). On univariate analysis, extracapsular infiltration in sentinel node and multiple sentinel nodes positivity were significantly associated (p < 0.05) with additional lymph node metastasis in the axilla. Area under ROC curve for nomogram was 0.58 suggesting poor performance of the nomogram in predicting NSLN involvement. Sentinel nodes are the only nodes to be involved by tumor in 70 % of the patients. Our findings indicate that multiple sentinel node positivity and extra-capsular invasion in sentinel node significantly predicted the likelihood of additional nodal metastasis. MSKCC nomogram did not reliably predict the involvement of additional nodal metastasis in our study population.

A head to head comparison of nine tools predicting non-sentinel lymph node status in sentinel node positive breast cancer women

BACKGROUND AND OBJECTIVES

This study was conducted to evaluate the performance of available tools predicting non-sentinel lymph node (non-SLN) status in women with SLN positive breast cancer and to see if they can be safely used in everyday clinical practice.

METHODS

Data of 220 women with breast cancer who underwent a SLN biopsy at the Máxima Medical Centre between 2000-2008 were analysed. Tools evaluated were: the models from Memorial Sloan Kettering Cancer Centre, Stanford, Mayo, Cambridge, Gur, and MOU, and the scores from Saidi, Tenon, and MDA. Model performance was assessed using calibration, discrimination and Nagelkerke's explained variation.

RESULTS

The MSKCC nomogram showed best overall performance with best discrimination (AUC 0.69), second best calibration, and highest explained variation (31%). The 10% low risk threshold led to defining only 22% (38/176) of the women as being low risk while in fact 66% (116/176) were non-SLN negative. The false negative rate was 13% (5/38).

CONCLUSIONS

Current models for predicting non-SLN metastases in SLN positive breast cancer are not yet ready for implementation in general practice. Further research efforts should improve model performance in selecting patients or perhaps find a role in support in the paradigm shift to a "treat none unless" approach.

Local relapse after breast-conserving therapy for ductal carcinoma in situ: a European single-center experience and external validation of the Memorial Sloan-Kettering Cancer Center DCIS nomogram

PURPOSE

Adjuvant treatments after breast-conserving surgery (BCS) for ductal carcinoma in situ to prevent local relapse are considered standard of care. However, patient selection to prevent increased morbidity without proven survival benefit remains a challenge. To predict the risk of ipsilateral breast tumor relapse (IBTR) after BCS, the Memorial Sloan-Kettering Cancer Center (MSKCC) developed a nomogram. The aim of this study was to develop our own prediction model for IBTR and to provide an external validation of the MSKCC nomogram.

METHODS

From 1973 to 2010, 467 patients were treated with BCS for ductal carcinoma in situ at the University Hospital Leuven. Clinicopathologic and treatment parameters of all patients were used to create a multivariable model. The predictive value of the model was evaluated using the concordance index (C-index) and concordance probability estimate (CPE). Multiple imputation was used to account for missing data to allow the MSKCC model to be tested on 467 patients.

RESULTS

Median follow-up was 7.2 years, with 48 women who developed an IBTR. Omission of adjuvant endocrine therapy, younger age, and positive or close surgical margins were significantly associated with an increased risk of IBTR. The bootstrap-corrected C-index for 10-year prediction by our own model was 0.63 and the CPE was 0.61. The C-index and CPE for the 10-year relapse probabilities predicted by the MSKCC nomogram were 0.66 and 0.61, respectively.

CONCLUSIONS

Despite the small number of events, the need for multiple imputation, and few patients without radiation, the MSKCC nomogram performance was somewhat better than our model. This shows that the MSKCC nomogram is externally valid. The MSKCC nomogram allows users to integrate the information from 10 different variables to provide a more precise risk stratification than the use of conventional single variables or hazard ratios.

Axillary nodal metastases in Italian early breast cancer patients with positive sentinel lymph node: can axillary node dissection be avoided by using predictive nomograms?

AIMS AND BACKGROUND

Clinical guidelines recommend axillary lymph node dissection (ALND) in cases of metastatic sentinel lymph node (SNL) in patients with clinically node-negative early breast cancer. However, a relevant number of ALND could be avoided in a subset of patients in whom the risk of non-SNL metastases is low. In order to define this population, several authors have proposed mathematical models, which have been validated in many studies. These studies reached different conclusions regarding which model demonstrated the best statistical discrimination power, mainly due to differences in clinical and pathologic variables used, and particularly differences in the number of dissected SLNs.

METHODS

We retrospectively reviewed clinically node-negative patients who underwent ALND in our surgical ward after the diagnosis of breast cancer metastases on SLN biopsy from January 2000 to December 2012. The predictive accuracy of the widely used nomograms to predict the risk of additional nodal disease in our patients with SLN breast cancer metastases was measured by receiver operating characteristic curve. We then attempted to develop a new nomogram by analyzing the dataset.

RESULTS

A total of 105 patients were included in this study, with ratio of metastatic lymph node/removed lymph node of about 0.89; we found axillary nodal metastases on ALND in only 31 patients (29.5%). Applied to our dataset, Mayo nomogram showed the best area under the receiving operator characteristic curve (0.74) followed by our model (0.71). Instead, the Memorial Sloan-Kettering model showed poor discrimination, as did Tenon (0.56).

CONCLUSIONS

Based on our data, we cannot recommend the clinical use of validated predictive nomograms in order to avoid ALND. We suggest setting up a multicenter Italian study to build a model specific to our setting and based on larger series.

Validation of the Memorial Sloan Kettering Cancer Center nomogram for predicting non-sentinel lymph node metastasis in sentinel lymph node-positive breast-cancer patients

BACKGROUND

The main purpose of the study reported here was to validate the clinical value of the Memorial Sloan Kettering Cancer Center (MSKCC) nomogram that predicts non-sentinel lymph node (SLN) metastasis in SLN-positive patients with breast cancer.

METHODS

Data on 1,576 patients who received sentinel lymph node biopsy (SLNB) at the Shandong Cancer Hospital from December 2001 to March 2014 were collected in this study, and data on 509 patients with positive SLN were analyzed to evaluate the risk factors for non-SLN metastasis. The MSKCC nomogram was used to estimate the probability of non-SLN metastasis and was compared with actual probability after grouping into deciles. A receiver-operating characteristic (ROC) curve was drawn and predictive accuracy was assessed by calculating the area under the ROC curve.

RESULTS

Tumor size, histological grade, lymphovascular invasion, multifocality, number of positive SLNs, and number of negative SLNs were correlated with non-SLN metastasis (P<0.05) by univariate analysis. However, multivariate analysis showed that tumor size (P=0.039), histological grade (P=0.043), lymphovascular invasion (P=0.001), number of positive SLNs (P=0.001), and number of negative SLNs (P=0.000) were identified as independent predictors for non-SLN metastasis. The trend of actual probability in various decile groups was comparable to the predicted probability. The area under the ROC curve was 0.722. Patients with predictive values lower than 10% (97/492, 19.7%) had a frequency of non-SLN metastasis of 17.5% (17/97).

CONCLUSION

The MSKCC nomogram can provide an accurate prediction of the probability of non-SLN metastasis, and offers a reference basis about axillary lymph node dissection. Axillary lymph node dissection could be avoided in patients with predictive values lower than 10%.

A new and simple predictive formula for non-sentinel lymph node metastasis in breast cancer patients with positive sentinel lymph nodes, and validation of 3 different nomograms in Turkish breast cancer patients

BACKGROUND

Nomogram accuracies for predicting non-sentinel lymph node (SLN) involvement vary between different patient populations. Our aim is to put these nomograms to test on our patient population and determine our individual predictive parameters affecting SLN and non-SLN involvement.

PATIENTS AND METHODS

Data from 932 patients was analyzed. Nomogram values were calculated for each patient utilizing MSKCC, Tenon, and MHDF models. Moreover, using our own patient- and tumor-depended parameters, we established a unique predictivity formula for SLN and non-SLN involvement.

RESULTS

The calculated area under the curve (AUC) values for MSKCC, Tenon, and MHDF models were 0.727 (95% confidence interval (CI) 0.64-0.8), 0.665 (95% CI 0.59-0.73), and 0.696 (95% CI 0.59-0.79), respectively. Cerb-2 positivity (p = 0.004) and size of the metastasis in the lymph node (p = 0.006) were found to correlate with non-SLN involvement in our study group. The AUC value of the predictivity formula established using these parameters was 0.722 (95% CI 0.63-0.81).

CONCLUSION

The most accurate nomogram for our patient group was the MSKCC nomogram. Our unique predictivity formula proved to be as equally effective and competent as the MSKCC nomogram. However, similar to other nomograms, our predictivity formula requires future validation studies.

Sentinel lymph node biopsy in breast cancer: predictors of axillary and non-sentinel lymph node involvement

BACKGROUND

Sentinel lymph node biopsy is a standard method for the evaluation of axillary status in patients with T1-2N0M0 breast cancers.

AIMS

To determine the prognostic significance of primary tumour-related clinico-histopathological factors on axillary and non-sentinel lymph node involvement of patients who underwent sentinel lymph node biopsy.

STUDY DESIGN

Retrospective clinical study.

METHODS

In the present study, 157 sentinel lymph node biopsies were performed in 151 consecutive patients with early stage breast cancer between June 2008 and December 2011.

RESULTS

Successful lymphatic mapping was obtained in 157 of 158 procedures (99.4%). The incidence of larger tumour size (2.543±1.21 vs. 1.974±1.04), lymphatic vessel invasion (70.6% vs. 29.4%), blood vessel invasion (84.2% vs. 15.8%), and invasive lobular carcinoma subtype (72.7% vs. 27.3%) were statistically significantly higher in patients with positive SLNs. Logistic stepwise regression analysis disclosed tumour size (odds ratio: 1.51, p=0.0021) and lymphatic vessel invasion (odds ratio: 4.68, p=0.001) as significant primary tumour-related prognostic determinants of SLN metastasis.

CONCLUSION

A close relationship was identified between tumour size and lymphatic vessel invasion of the primary tumour and axillary lymph node involvement. However, the positive predictive value of these two independent variables is low and there is no compelling evidence to recommend their use in routine clinical practice.

Predicting the likelihood of additional lymph node metastasis in sentinel lymph node positive breast cancer: validation of the Memorial Sloan-Kettering Cancer Centre (MSKCC) nomogram

AIM

To identify important clinicopathological parameters that are most helpful in predicting additional non-sentinel lymph node (SLN) metastasis among patients with a positive SLN biopsy in the Singapore breast cancer population.

METHODS

A total of 1409 patients who underwent SLN biopsy were reviewed over a 5 year period from July 2004 to October 2009. A Singapore General Hospital (SGH) nomogram was developed from predictors in the Memorial Sloan-Kettering Cancer Centre (MSKCC) nomogram using 266 patients with primary invasive breast cancer and a positive SLN biopsy who subsequently had an axillary lymph node dissection. The SGH nomogram was calibrated using bootstrapped data, while the MSKCC nomogram was calibrated using SGH data. The performance of these two nomograms was compared with the calculation of the area under the receiver-operator characteristics curve and adequacy indices.

RESULTS

The MSKCC nomogram achieved an area under the curve (AUC) of 0.716 (range 0.653-0.779) in our study population, while the SGH nomogram, which used only three pathological parameters, lymphovascular invasion, number of positive and negative SLN biopsies, achieved an AUC of 0.750 (range 0.691-0.808). The SGH nomogram with a higher adequacy index (0.969) provided better estimates compared with the MSKCC nomogram (0.689).

CONCLUSIONS

The use of the MSKCC nomogram was validated in our local patient population. The SGH nomogram showed promise to be equally, if not, more predictive as a model in our own population, while using only three pathological parameters.

Evolution of axillary nodal staging in breast cancer: clinical implications of the ACOSOG Z0011 trial

BACKGROUND

Management of the axilla in breast cancer patients has evolved from routine axillary lymph node dissection (ALND) for all patients to a highly selective approach based on the assessment of the sentinel lymph nodes (SLNs) as well as tumor and patient characteristics. Although ALND continues to have an important role in staging and regional control for many breast cancer patients, recent trial results question the need for routine ALND in patients who have positive SLNs.

METHODS

Not all axillary disease becomes clinically detectable or relevant with respect to recurrence and survival. Therefore, recent trends indicate that many surgeons have omitted ALND in subgroups of patients, particularly those with clinically node-negative, SLN-positive, early-stage breast cancer undergoing breast-conserving therapy with postoperative irradiation. This review explores trends in axillary management, focusing primarily on the clinical implications of the results from the American College of Surgeons Oncology Group (ACOSOG) Z0011 randomized controlled trial.

RESULTS

According to the results of the ACOSOG Z0011 trial, the use of SLN dissection alone did not result in inferior survival compared with ALND in patients with limited SLN disease treated with breast-conserving therapy. This subgroup of women was spared the morbidity associated with ALND. However, several points of debate, including the smaller than anticipated sample size, the older study population, and the length of follow-up, suggest caution when applying these findings to all women with breast cancer.

CONCLUSIONS

Although the findings of ACOSOG Z0011 are impressive, in clinical practice they are applicable to a limited number of women with breast cancer: those with T1-2 primary tumors with clinically negative axilla and 1 to 2 positive SLNs undergoing breast-conserving surgery and adjuvant whole-breast irradiation. The next generation of clinical trials may answer some of the remaining questions regarding how best to manage the axilla in additional subsets of patients undergoing treatment of breast cancer.

The efficacy of axillary ultrasound in the detection of nodal metastasis in breast cancer

OBJECTIVE

Recent reports indicate a lack of survival benefit for axillary lymph node dissection (ALND) versus sentinel lymph node biopsy in early breast cancer. To study this issue further, we assessed the accuracy and effectiveness of ultrasound examination in detecting axillary nodal involvement in breast cancer patients with the aim of refining our current clinical pathways.

MATERIALS AND METHODS

Ultrasound data were collected from breast cancer cases over 3 years. Images were reviewed by experienced radiologists and the following characteristics were assessed: size, morphology, hyperechoic hilum, and cortical thickness of the ipsilateral axillary nodes. The findings were correlated with histologic outcomes after ALND.

RESULTS

Two hundred twenty-four cases were included in the analysis, 113 (50.4%) of which had evidence of metastatic nodal involvement at final histology. Of these 113 cases, ultrasound findings for 59 (52.2%) were positive. The overall positive predictive value of ultrasound for detecting metastatic nodal involvement measured 0.81. The negative predictive value was 0.60. The sensitivity was 53.7%; specificity, 85.1%; and accuracy, 67.9%. The ultrasound morphologic lymph node features with the greatest correlation with malignancy were the absence of a hyperechoic hilum (p = 0.003) and increased cortical thickness (p = 0.03). Patients with a metastatic nodal burden density of at least 20% were more likely to have abnormal findings on axillary ultrasound examination (p = 0.009).

CONCLUSION

Axillary ultrasound has a low negative predictive value and negative ultrasound results do not exclude axillary node metastases with sufficient sensitivity to justify its routine clinical use. Clinical pathways need to consider an evidence-based approach, focusing on the criteria by which we select breast cancer patients for ALND.

Risk factors for sentinel lymph node metastasis and validation study of the MSKCC nomogram in breast cancer patients

OBJECTIVE

To evaluate the risk factors for sentinel lymph node metastasis and validate the value of the Memorial Sloan-Kettering Cancer Center nomogram for the prediction of sentinel lymph node metastasis in breast cancer patients.

METHODS

A sentinel lymph node biopsy database containing 1227 consecutive breast cancer patients (416 patients with at least one positive sentinel lymph node) was retrospectively analyzed. The predictive value of the Memorial Sloan-Kettering Cancer Center nomogram was calculated by the trend line and the area under the receiver-operator characteristic curve. Meanwhile, predictors for sentinel lymph node metastasis were also evaluated.

RESULTS

Tumor size, histological grade, lymphovascular invasion, mulifocality, estrogen receptor and progesterone receptor status were significant independent predictors for sentinel lymph node metastasis (all P<0.01). The Memorial Sloan-Kettering Cancer Center nomogram presented an area under the receiver-operator characteristic curve value of 0.730. Patients with predictive value<16% had a frequency of sentinel lymph node metastasis of 0.9%. Those with values larger than 70% had a frequency of 96.2%.

CONCLUSIONS

The risk factors for sentinel lymph node metastasis in our study were consistent with those in the Memorial Sloan-Kettering Cancer Center nomogram. The Memorial Sloan-Kettering Cancer Center nomogram is a useful tool that could accurately predict the probability of sentinel lymph node metastasis in our breast cancer patients. Axillary surgical staging might be avoided in patients with a predictive value of <16% and axillary lymph node dissection might be done directly in those with a predictive value >70%, while other patients should still accept sentinel lymph node biopsy.

Which nomogram is best for predicting non-sentinel lymph node metastasis in breast cancer patients? A meta-analysis

To present a systematic [corrected] review and meta-analysis to evaluate the nomograms developed to predict non-sentinel lymph node (NSLN) metastasis in breast cancer patients. We focused on the six nomograms (Cambridge, MSKCC, Mayo, MDA, Tenon, and Stanford) that are the most widely validated. The AUCs were converted to odds ratios for the meta-analysis. In total, the Cambridge, Mayo, MDA, MSKCC, Stanford, and Tenon models were validated in 2,156, 2,431, 843, 8,143, 3,700, and 3,648 patients, respectively. The pooled AUCs for the Cambridge, MDA, MSKCC, Mayo, Tenon, and Stanford models were 0.721, 0.706, 0.715, 0.728, 0.720, and 0.688, respectively. Subgroup analysis revealed that in populations with a higher micrometastasis rate in the SLNs, the Tenon and Stanford models had a significantly higher predictive accuracy. A meta-regression analysis revealed that the SLN micrometastasis rate, but not the NSLN-positivity rate, was associated with improved predictive accuracy in the Tenon and Stanford models. The performance of the MSKCC and Cambridge models was not influenced by these two factors. All of these prediction models perform better than random chance. The Stanford model seems to be relatively inferior to the other models. The accuracy of the Tenon and Stanford models is influenced by the tumor burden in the SLNs.

Axillary recurrence after a tumor-positive sentinel lymph node biopsy without axillary treatment: a review of the literature

Background

Sentinel lymph node biopsy (SLNB) has become standard of care as a staging procedure in patients with invasive breast cancer. A positive SLNB allows completion axillary lymph node dissection (cALND) to be performed. The axillary recurrence rate (ARR) after cALND in patients with positive SLNB is low. Recently, several studies have reported a similar low ARR when cALND is not performed. This review aims to determine the ARR when cALND is omitted in SLNB-positive patients.

Methods

A literature search was performed in the PubMed database with the search terms “breast cancer,” “sentinel lymph node biopsy,” “axillary” and “recurrence.” Articles with data regarding follow-up of patients with SLNB-positive breast cancer were identified. To be eligible, patients should not have received cALND and ARR should be reported.

Results

Thirty articles were analyzed. This resulted in 7,151 patients with SLNB-positive breast cancer in whom a cALND was omitted (median follow-up of 45 months, range 1–142 months). Overall, 41 patients developed an axillary recurrence. 27 studies described 3,468 patients with micrometastases in the SLNB, of whom 10 (0.3 %) developed an axillary recurrence. ARR varied between 0 and 3.7 %. Sixteen studies described 3,268 patients with macrometastases, 24 (0.7 %) axillary recurrences were seen. ARR varied between 0 and 7.1 %. Details regarding type of surgery and adjuvant treatment were lacking in the majority of studies.

Conclusions

ARR appears to be low in SLNB-positive patients even when a cALND is not performed. Withholding cALND may be safe in breast cancer selected patients such as those with isolated tumor cells or micrometastatic disease.

Validation of the Louisville Breast Sentinel Node Prediction Models and a Proposed Modification to Guide Management of the Node Positive Axilla

The ACOSOG Z11 trial is rapidly changing use of axillary dissection, but it is not known how generalizable the Z11 results are. This study compares characteristics of the Z11 patients with the larger group of sentinel node-positive patients and evaluates two previously described Louisville algorithms to determine whether they might still be useful to predict extent of axillary node involvement and guide management of the axilla. The Yale Breast Center database was queried to calculate the Louisville prediction points for patients with a positive sentinel node and to compare the predicted with actual results. Of 1215 sentinel node biopsies performed between 2004 and 2010, 282 (23%) had at least one positive node. Thirty-one per cent of these patients would have been eligible for Z11. This group had much less axillary node involvement than the 69 per cent who were ineligible. The Yale data confirmed the accuracy of the two Louisville models and showed that tumor size, number of positive sentinel nodes, and proportion of positive sentinel nodes were all significant predictors. However, these results were much more robust if at least three sentinel nodes had been removed. The Z11 patients were clearly a good risk group. The data validate the two Louisville models and suggest that the models may be useful to select patients to avoid axillary dissection, both among the currently Z11-eligible and -ineligible populations. A modified algorithm is proposed in which all patients with a positive sentinel node have at least three total nodes removed.

Sentinel lymph node micrometastasis in human breast cancer: an update

INTRODUCTION

The advent of sentinel lymph node biopsy (SLNB) and advances in histopathological and molecular analysis techniques have been associated with an increase in micrometastasis (MM) detection rate. However, the clinical significance of sentinel lymph node micrometastasis (SLN MM) continues to be a subject of much debate. In this article we review the literature concerning SLN MM, with particular emphasis on the prognostic significance of SLN MM. The controversies regarding histopathological assessment, clinical relevance and management implications are also discussed.

METHODS

Literature review facilitated by Medline and PubMed databases. Cross referencing of the obtained articles was used to identify other relevant studies.

RESULTS

Published studies have reported divergent and rather conflicting results regarding the clinical significance and implications of axillary lymph node (ALN) MM in general and SLN MM in particular. Some earlier studies demonstrated no associations, however most recent studies have found SLN MM to be an indicator of poorer prognosis and to be associated with non-SLN involvement. The use of adjuvant chemotherapy and/or hormonal manipulation therapy is associated with an improved survival in patients with SLN MM. Complete ALND may be safely omitted provided that adjuvant systemic therapy recommendations are equal to patients with node-positive disease. However, optimal management of SLN MM is yet to conclude. Furthermore, the identification of MM remains largely dependent on the analytical technique employed and the use of immunohistochemistry (IHC) increases the detection rate of SLN MM. Discrepancies in the histopathological interpretation of TNM classification of SLN tumour burden do exist. Published studies were non-randomized and have significant limitations including a small sample size, limited follow-up period, and lack of standardization and reproducibility of pathological examination of the SLN.

CONCLUSION

Patients with SLN MM have a poorer prognosis than those who are SLN negative. Therapeutic recommendations regarding patients with SLN MM should be taken in the context of multidisciplinary team setting and in selected cases of SLN MM, complete ALND may be safely omitted. A better reproducibility of pathological interpretation of the TNM classification is required so that future therapeutic guidelines can be applied without confusion.

Diagnostic accuracy and receiver-operating characteristics curve analysis in surgical research and decision making

In surgical research, the ability to correctly classify one type of condition or specific outcome from another is of great importance for variables influencing clinical decision making. Receiver-operating characteristic (ROC) curve analysis is a useful tool in assessing the diagnostic accuracy of any variable with a continuous spectrum of results. In order to rule a disease state in or out with a given test, the test results are usually binary, with arbitrarily chosen cut-offs for defining disease versus health, or for grading of disease severity. In the postgenomic era, the translation from bench-to-bedside of biomarkers in various tissues and body fluids requires appropriate tools for analysis. In contrast to predetermining a cut-off value to define disease, the advantages of applying ROC analysis include the ability to test diagnostic accuracy across the entire range of variable scores and test outcomes. In addition, ROC analysis can easily examine visual and statistical comparisons across tests or scores. ROC is also favored because it is thought to be independent from the prevalence of the condition under investigation. ROC analysis is used in various surgical settings and across disciplines, including cancer research, biomarker assessment, imaging evaluation, and assessment of risk scores.With appropriate use, ROC curves may help identify the most appropriate cutoff value for clinical and surgical decision making and avoid confounding effects seen with subjective ratings. ROC curve results should always be put in perspective, because a good classifier does not guarantee the expected clinical outcome. In this review, we discuss the fundamental roles, suggested presentation, potential biases, and interpretation of ROC analysis in surgical research.

Meta-analysis of predictive factors for non-sentinel lymph node metastases in breast cancer patients with a positive SLN

AIMS

A meta-analysis was performed to identify the clinicopathological variables most predictive of non-sentinel node (NSN) metastases when the sentinel node is positive.

METHODS

A Medline search was conducted that ultimately identified 56 candidate studies. Original data were abstracted from each study and used to calculate odds ratios. The random-effects model was used to combine odds ratios to determine the strength of the associations.

FINDINGS

The 8 individual characteristics found to be significantly associated with the highest likelihood (odds ratio >2) of NSN metastases are SLN metastases >2mm in size, extracapsular extension in the SLN, >1 positive SLN, ≤1 negative SLN, tumour size >2cm, ratio of positive sentinel nodes >50% and lymphovascular invasion in the primary tumour. The histological method of detection, which is associated with the size of metastases, had a correspondingly high odds ratio.

CONCLUSIONS

We identified 8 factors predictive of NSN metastases that should be recorded and evaluated routinely in SLN databases. These factors should be included in a predictive model that is generally applicable among different populations.

Axillary dissection vs no axillary dissection in women with invasive breast cancer and sentinel node metastasis: a randomized clinical trial

CONTEXT

Sentinel lymph node dissection (SLND) accurately identifies nodal metastasis of early breast cancer, but it is not clear whether further nodal dissection affects survival.

OBJECTIVE

To determine the effects of complete axillary lymph node dissection (ALND) on survival of patients with sentinel lymph node (SLN) metastasis of breast cancer.

DESIGN, SETTING, AND PATIENTS

The American College of Surgeons Oncology Group Z0011 trial, a phase 3 noninferiority trial conducted at 115 sites and enrolling patients from May 1999 to December 2004. Patients were women with clinical T1-T2 invasive breast cancer, no palpable adenopathy, and 1 to 2 SLNs containing metastases identified by frozen section, touch preparation, or hematoxylin-eosin staining on permanent section. Targeted enrollment was 1900 women with final analysis after 500 deaths, but the trial closed early because mortality rate was lower than expected.

INTERVENTIONS

All patients underwent lumpectomy and tangential whole-breast irradiation. Those with SLN metastases identified by SLND were randomized to undergo ALND or no further axillary treatment. Those randomized to ALND underwent dissection of 10 or more nodes. Systemic therapy was at the discretion of the treating physician.

MAIN OUTCOME MEASURES

Overall survival was the primary end point, with a noninferiority margin of a 1-sided hazard ratio of less than 1.3 indicating that SLND alone is noninferior to ALND. Disease-free survival was a secondary end point.

RESULTS

Clinical and tumor characteristics were similar between 445 patients randomized to ALND and 446 randomized to SLND alone. However, the median number of nodes removed was 17 with ALND and 2 with SLND alone. At a median follow-up of 6.3 years (last follow-up, March 4, 2010), 5-year overall survival was 91.8% (95% confidence interval [CI], 89.1%-94.5%) with ALND and 92.5% (95% CI, 90.0%-95.1%) with SLND alone; 5-year disease-free survival was 82.2% (95% CI, 78.3%-86.3%) with ALND and 83.9% (95% CI, 80.2%-87.9%) with SLND alone. The hazard ratio for treatment-related overall survival was 0.79 (90% CI, 0.56-1.11) without adjustment and 0.87 (90% CI, 0.62-1.23) after adjusting for age and adjuvant therapy.

CONCLUSION

Among patients with limited SLN metastatic breast cancer treated with breast conservation and systemic therapy, the use of SLND alone compared with ALND did not result in inferior survival.

TRIAL REGISTRATION

clinicaltrials.gov Identifier: NCT00003855.

Impact of the sentinel node frozen section result on the probability of additional nodal metastases as predicted by the MSKCC nomogram in breast cancer

OBJECTIVE

Sentinel lymph node frozen section is used to obviate the need for a second operation in breast cancer patients with involved nodes. However, the overall sensitivity, specificity and accuracy of sentinel lymph node frozen section are debated, and the impact of sentinel lymph node frozen section positivity on the risk of additional nodal metastases is not known and was the focus of this investigation.

METHODS

We used our hospital record system to identify 176 sentinel lymph node biopsies done out of 354 cases of Stage T1-3N0 breast cancers managed from 2005 to 2007 and evaluated the sentinel lymph node frozen section results against the predictions of additional nodal metastases based on the Memorial Sloan-Kettering Breast Cancer Nomogram which is a validated tool for this purpose.

RESULTS

Sentinel lymph node metastases size was an independent predictor of sentinel lymph node frozen section sensitivity and those with macrometastases had 15 times the odds (odds ratio, 15.4; 95% confidence interval, 3.4-69.1) of having a true-positive frozen section when compared with those with micrometastases. The breast cancer nomogram predicted that the latter patients have a very low probability of additional nodal metastases with a median probability at 10% (inter-quartile range, 7-14%).

CONCLUSIONS

A negative sentinel lymph node frozen section is also associated with a low probability of additional nodal metastases. Additional prognostic factors in the breast cancer nomogram are of little clinical impact because the most predictive factor in the nomogram is the method of detection.

Value of Memorial Sloan-Kettering Cancer Center nomogram in clinical decision making for sentinel lymph node-positive breast cancer

Background

The aim of this study was to determine the value of the Memorial Sloan-Kettering Cancer Center (MSKCC) nomogram for individual decision making in a Dutch cohort of women with breast cancer with a positive sentinel lymph node (SLN) but at low risk of additional nodal metastases.

Methods

Data were collected on 168 patients with a positive SLN who underwent completion axillary lymph node dissection. The predicted probability of non-SLN metastases was calculated for each patient, using the MSKCC nomogram. Specificity and false-negative rates were calculated for subgroups with a predicted risk of no more than 5, 10 or 15 per cent. A receiver operating characteristic (ROC) curve was constructed and the area under the curve (AUC) calculated.

Results

The discrimination of the MSKCC nomogram, measured by the AUC, was 0·68. For low predicted probability cut-off values of no more than 5, 10 and 15 per cent, the false-negative rates were 20, 14 and 19 per cent, and the specificities were 4, 27 and 32 per cent, respectively. The low-risk category (5 per cent or less) consisted of only 3·0 per cent of the study population.

Conclusion

The performance of the MSKCC nomogram was insufficient to make it a useful tool for individual decision making in this cohort of women with SLN-positive breast cancer.

Nomogram for predicting the risk of local recurrence after breast-conserving surgery for ductal carcinoma in situ

PURPOSE

While the mortality associated with ductal carcinoma in situ (DCIS) is minimal, the risk of ipsilateral breast tumor recurrence (IBTR) after breast-conserving surgery (BCS) is relatively high. Radiation therapy (RT) and antiestrogen agents reduce the risk of IBTR and are considered standard treatment options after BCS. However, they have never been proven to improve survival, and in themselves carry rare but serious risks. Individualized estimation of IBTR risk would assist in decision making regarding the various treatment options for women with DCIS.

PATIENTS AND METHODS

From 1991 to 2006, 1,868 consecutive patients treated with BCS for DCIS were identified. A multivariate Cox proportional hazards model was constructed using the 1,681 in whom data were complete. Ten clinical, pathologic, and treatment variables were built into a nomogram estimating probability of IBTR at 5 and 10 years after BCS. The model was validated for discrimination and calibration using bootstrap resampling.

RESULTS

The DCIS nomogram for prediction of 5- and 10-year IBTR probabilities demonstrated good calibration and discrimination, with a concordance index of 0.704 (bootstrap corrected, 0.688) and a concordance probability estimate of 0.686. Factors with the greatest influence on risk of IBTR in the model included adjuvant RT or endocrine therapy, age, margin status, number of excisions, and treatment time period.

CONCLUSION

The DCIS nomogram integrates 10 clinicopathologic variables to provide an individualized risk estimate of IBTR in a woman with DCIS treated with BCS. This tool may assist in individual decision making regarding various treatment options and help avoid over- and undertreatment of noninvasive breast cancer.