Development and Validation of Nomograms Predictive of Axillary Nodal Status to Guide Surgical Decision-Making in Early-Stage Breast Cancer

Preoperative MRI Features Associated With Axillary Nodal Burden and Disease-Free Survival in Patients With Early-Stage Breast Cancer

OBJECTIVE

To investigate the potential association among preoperative breast MRI features, axillary nodal burden (ANB), and disease-free survival (DFS) in patients with early-stage breast cancer.

MATERIALS AND METHODS

We retrospectively reviewed 297 patients with early-stage breast cancer (cT1-2N0M0) who underwent preoperative MRI between December 2016 and December 2018. Based on the number of positive axillary lymph nodes (LNs) determined by postoperative pathology, the patients were divided into high nodal burden (HNB; ≥3 positive LNs) and non-HNB (<3 positive LNs) groups. Univariable and multivariable logistic regression analyses were performed to identify independent risk factors associated with ANB. Predictive efficacy was evaluated using the receiver operating characteristic (ROC) curve and area under the curve (AUC). Univariable and multivariable Cox proportional hazards regression analyses were performed to determine preoperative features associated with DFS.

RESULTS

We included 47 and 250 patients in the HNB and non-HNB groups, respectively. Multivariable logistic regression analysis revealed that multifocality/multicentricity (adjusted odds ratio [OR] = 3.905, 95% confidence interval [CI]: 1.685-9.051, P = 0.001) and peritumoral edema (adjusted OR = 3.734, 95% CI: 1.644-8.479, P = 0.002) were independent risk factors for HNB. Combined peritumoral edema and multifocality/multicentricity achieved an AUC of 0.760 (95% CI: 0.707-0.807) for predicting HNB, with a sensitivity and specificity of 83.0% and 63.2%, respectively. During the median follow-up period of 45 months (range, 5-61 months), 26 cases (8.75%) of breast cancer recurrence were observed. Multivariable Cox proportional hazards regression analysis indicated that younger age (adjusted hazard ratio [HR] = 3.166, 95% CI: 1.200-8.352, P = 0.021), larger tumor size (adjusted HR = 4.370, 95% CI: 1.671-11.428, P = 0.002), and multifocality/multicentricity (adjusted HR = 5.059, 95% CI: 2.166-11.818, P < 0.001) were independently associated with DFS.

CONCLUSION

Preoperative breast MRI features may be associated with ANB and DFS in patients with early-stage breast cancer.

A nomogram model for predicting the risk of axillary lymph node metastasis in patients with early breast cancer and cN0 status

Axillary staging is commonly performed via sentinel lymph node biopsy for patients with early breast cancer (EBC) presenting with clinically negative axillary lymph nodes (cN0). The present study aimed to investigate the association between axillary lymph node metastasis (ALNM), clinicopathological characteristics of tumors and results from axillary ultrasound (US) scanning. Moreover, a nomogram model was developed to predict the risk for ALNM based on relevant factors. Data from 998 patients who met the inclusion criteria were retrospectively reviewed. These patients were then randomly divided into a training and validation group in a 7:3 ratio. In the training group, receiver operating characteristic curve analysis was used to identify the cutoff values for continuous measurement data. R software was used to identify independent ALNM risk variables in the training group using univariate and multivariate logistic regression analysis. The selected independent risk factors were incorporated into a nomogram. The model differentiation was assessed using the area under the curve (AUC), while calibration was evaluated through calibration charts and the Hosmer-Lemeshow test. To assess clinical applicability, a decision curve analysis (DCA) was conducted. Internal verification was performed via 1000 rounds of bootstrap resampling. Among the 998 patients with EBC, 228 (22.84%) developed ALNM. Multivariate logistic analysis identified lymphovascular invasion, axillary US findings, maximum diameter and molecular subtype as independent risk factors for ALNM. The Akaike Information Criterion served as the basis for both nomogram development and model selection. Robust differentiation was shown by the AUC values of 0.855 (95% CI, 0.817-0.892) and 0.793 (95% CI, 0.725-0.857) for the training and validation groups, respectively. The Hosmer-Lemeshow test yielded P-values of 0.869 and 0.847 for the training and validation groups, respectively, and the calibration chart aligned closely with the ideal curve, affirming excellent calibration. DCA showed that the net benefit from the nomogram significantly outweighed both the 'no intervention' and the 'full intervention' approaches, falling within the threshold probability interval of 12-97% for the training group and 17-82% for the validation group. This underscores the robust clinical utility of the model. A nomogram model was successfully constructed and validated to predict the risk of ALNM in patients with EBC and cN0 status. The model demonstrated favorable differentiation, calibration and clinical applicability, offering valuable guidance for assessing axillary lymph node status in this population.

Noninvasive Staging of Lymph Node Status in Breast Cancer Using Machine Learning: External Validation and Further Model Development

BACKGROUND

Most patients diagnosed with breast cancer present with a node-negative disease. Sentinel lymph node biopsy (SLNB) is routinely used for axillary staging, leaving patients with healthy axillary lymph nodes without therapeutic effects but at risk of morbidities from the intervention. Numerous studies have developed nodal status prediction models for noninvasive axillary staging using postoperative data or imaging features that are not part of the diagnostic workup. Lymphovascular invasion (LVI) is a top-ranked predictor of nodal metastasis; however, its preoperative assessment is challenging.

OBJECTIVE

This paper aimed to externally validate a multilayer perceptron (MLP) model for noninvasive lymph node staging (NILS) in a large population-based cohort (n=18,633) and develop a new MLP in the same cohort. Data were extracted from the Swedish National Quality Register for Breast Cancer (NKBC, 2014-2017), comprising only routinely and preoperatively available documented clinicopathological variables. A secondary aim was to develop and validate an LVI MLP for imputation of missing LVI status to increase the preoperative feasibility of the original NILS model.

METHODS

Three nonoverlapping cohorts were used for model development and validation. A total of 4 MLPs for nodal status and 1 LVI MLP were developed using 11 to 12 routinely available predictors. Three nodal status models were used to account for the different availabilities of LVI status in the cohorts and external validation in NKBC. The fourth nodal status model was developed for 80% (14,906/18,663) of NKBC cases and validated in the remaining 20% (3727/18,663). Three alternatives for imputation of LVI status were compared. The discriminatory capacity was evaluated using the validation area under the receiver operating characteristics curve (AUC) in 3 of the nodal status models. The clinical feasibility of the models was evaluated using calibration and decision curve analyses.

RESULTS

External validation of the original NILS model was performed in NKBC (AUC 0.699, 95% CI 0.690-0.708) with good calibration and the potential of sparing 16% of patients with node-negative disease from SLNB. The LVI model was externally validated (AUC 0.747, 95% CI 0.694-0.799) with good calibration but did not improve the discriminatory performance of the nodal status models. A new nodal status model was developed in NKBC without information on LVI (AUC 0.709, 95% CI: 0.688-0.729), with excellent calibration in the holdout internal validation cohort, resulting in the potential omission of 24% of patients from unnecessary SLNBs.

CONCLUSIONS

The NILS model was externally validated in NKBC, where the imputation of LVI status did not improve the model's discriminatory performance. A new nodal status model demonstrated the feasibility of using register data comprising only the variables available in the preoperative setting for NILS using machine learning. Future steps include ongoing preoperative validation of the NILS model and extending the model with, for example, mammography images.

The value of whole-lesion histogram analysis based on field‑of‑view optimized and constrained undistorted single shot (FOCUS) DWI for predicting axillary lymph node status in early-stage breast cancer

Background

This study aims to estimate the amount of axillary lymph node (ALN) involvement in early-stage breast cancer utilizing a field of view (FOV) optimized and constrained undistorted single-shot (FOCUS) diffusion-weighted imaging (DWI) approach, as well as a whole-lesion histogram analysis.

Methods

This retrospective analysis involved 81 individuals with invasive breast cancer. The patients were divided into three groups: N0 (negative ALN metastasis), N1–2 (low metastatic burden with 1–2 ALNs), and N≥3 (heavy metastatic burden with ≥ 3 ALNs) based on their sentinel lymph node biopsy (SLNB) or axillary lymph node dissection (ALND). Histogram parameters of apparent diffusion coefficient (ADC) depending basically on FOCUS DWI were performed using 3D-Slicer software for whole lesions. The typical histogram characteristics for N0, N1–2, and N≥ 3 were compared to identify the significantly different parameters. To determine the diagnostic efficacy of significantly different factors, the area under their receiver operating characteristic (ROC) curves was examined.

Results

There were significant differences in the energy, maximum, 90 percentile, range, and lesion size among N0, N1–2, and N≥ 3 groups (P < 0.05). The energy differed significantly between N0 and N1–2 groups (P < 0.05), and some certain ADC histogram parameters and lesion sizes differed significantly between N0 and N≥3, or N1–2 and N≥3 groups. For ROC analysis, the energy yielded the best diagnostic performance in distinguishing N0 and N1–2 groups from N≥3 group with an AUC value of0.853. All parameters revealed excellent inter-observer agreement with inter-reader consistencies data ranging from0.919 to 0.982.

Conclusion

By employing FOCUS DWI method, the analysis of whole-lesion ADC histogram quantitatively provides a non-invasive way to evaluate the degree of ALN metastatic spread in early-stage breast cancer.

"Real-world" radiomics from multi-vendor MRI: an original retrospective study on the prediction of nodal status and disease survival in breast cancer, as an exemplar to promote discussion of the wider issues

BACKGROUND

Most MRI radiomics studies to date, even multi-centre ones, have used "pure" datasets deliberately accrued from single-vendor, single-field-strength scanners. This does not reflect aspirations for the ultimate generalisability of AI models. We therefore investigated the development of a radiomics signature from heterogeneous data originating on six different imaging platforms, for a breast cancer exemplar, in order to provide input into future discussions of the viability of radiomics in "real-world" scenarios where image data are not controlled by specific trial protocols but reflective of routine clinical practice.

METHODS

One hundred fifty-six patients with pathologically proven breast cancer underwent multi-contrast MRI prior to neoadjuvant chemotherapy and/or surgery. From these, 92 patients were identified for whom T2-weighted, diffusion-weighted and contrast-enhanced T1-weighted sequences were available, as well as key clinicopathological variables. Regions-of-interest were drawn on the above image types and, from these, semantic and calculated radiomics features were derived. Classification models using a variety of methods, both with and without recursive feature elimination, were developed to predict pathological nodal status. Separately, we applied the same methods to analyse the information carried by the radiomic features regarding the originating scanner type and field strength. Repeated, ten-fold cross-validation was employed to verify the results. In parallel work, survival modelling was performed using random survival forests.

RESULTS

Prediction of nodal status yielded mean cross-validated AUC values of 0.735 ± 0.15 (SD) for clinical variables alone, 0.673 ± 0.16 (SD) for radiomic features only, and 0.764 ± 0.16 (SD) for radiomics and clinical features together. Prediction of scanner platform from the radiomics features yielded extremely high values of AUC between 0.91 and 1 for the different classes examined indicating the presence of confounding features for the nodal status classification task. Survival analysis, gave out-of-bag prediction errors of 19.3% (clinical features only), 36.9-51.8% (radiomic features from different combinations of image contrasts), and 26.7-35.6% (clinical plus radiomics features).

CONCLUSIONS

Radiomic classification models whose predictive ability was consistent with previous single-vendor, single-field strength studies have been obtained from multi-vendor, multi-field-strength data, despite clear confounding information being present. However, our sample size was too small to obtain useful survival modelling results.

Magnetic resonance imaging features for predicting axillary lymph node metastasis in patients with breast cancer

PURPOSE

The purpose of this study was to assess the clinical value of conventional magnetic resonance imaging (MRI) and intravoxel incoherent motion (IVIM) features for predicting the risk of axillary lymph node (ALN) metastasis in patients with breast cancer.

METHODS

This retrospective study involved 265 patients with breast cancer who underwent 3.0 T breast magnetic resonance imaging examinations prior to surgery and other treatment. Of these, 119 underwent IVIM examination. The features of MRI and IVIM and postoperative pathologic results were collected. The association of MRI features of breast cancer with ALN metastasis were determined by univariate and multivariate analyses. Comparison of IVIM parameters between breast cancer patients with and without ALN metastasis was performed using the Mann-Whitney U test.

RESULTS

Among the 265 patients, 144 (54.3%) had ALN metastasis, and 121 (45.7%) did not. The size and shape of the tumours, T2WI signal, inhomogeneous enhancement, washout intensity-time curves and the values of slow ADC, fast ADC and fraction of fast ADC parameters were significantly associated with ALN metastasis. The AUC of conventional MRI for diagnosing axillary lymph node metastasis was 0.722. The AUC of MRI combined with slow ADC, fast ADC and fraction of fast ADC parameters that were used to diagnose breast cancer with ALN metastasis were 0.814, 0.803 and 0.900, respectively.

CONCLUSIONS

The features of IVIM parameters and conventional MRI can be used to predict the ALN metastasis in patients with breast cancer. MRI combined with fraction of fast ADC showed higher diagnostic efficiency for ALN metastasis in breast cancer than MRI did.

Establishment of Simple Nomograms for Predicting Axillary Lymph Node Involvement in Early Breast Cancer

Purpose

Axillary lymph node (ALN) involvement is an important prognostic factor of early invasive breast cancer. The objective of this study was to establish simple nomograms for predicting ALN involvement based on ultrasound (US) characteristics and evaluate the predictive value of US in the detection of ALN involvement.

Patients and Methods

A total of 1328 patients with cT1-2N0 breast cancer by physical exam were retrospectively analyzed. Univariate analysis was used for the comparison of variables, and multivariate analysis was performed by binary logistic regression analysis. The R software was used to establish simple nomograms based on the US characteristics alone. The receiver operating characteristic (ROC) curves of the prediction model and the verification group were drawn, and the area under the curve (AUC) was calculated to evaluate the discrimination of the prediction model. A calibration curve was plotted to assess the nomogram predictions vs the actual observations of the ALN metastasis rate and axillary tumor burden rate.

Results

The ALN metastasis rates of the training group and the validation group were 35.1% and 34.1%, respectively. Multivariate analysis showed that molecular subtype, lymphovascular invasion, mass descriptors (size, margin, microcalcification and blood flow signal) and LN descriptors (shape, cortical thickness and long-to-short ratio) were independent impact factors in early breast cancer. The AUC of ALN metastasis rate of prediction model based on US features was 0.802, the AUC of high tumor burden rate was 0.873, and the AUC of external validation group was 0.731 and 0.802, respectively. The calibration curve of the nomogram showed that the nomogram predictions are consistent with the actual metastasis rate and the high tumor burden rate. The results showed that preoperative US had a sensitivity of 59.4% and a specificity of 88.9% for predicting the ALN metastasis rate.

Conclusion

The successfully established nomograms based on US characteristics to predict ALN metastasis rate and high axillary tumor burden rate in early breast cancer can achieve individual prediction. Compared with other nomogram predictions, it is more intuitive, and can help clinical decision-making; thus, it should be promoted. However, at this time US features alone are insufficient to replace sentinel lymph node biopsy.