Synopsis

This study evaluated a radiological nodal staging metric for accurate identification of diseased axillary lymph nodes (aLNs) on MRI associated with breast cancer. Comparison was made with pathological nodal staging as the ground truth. Radiological scoring of aLNs identifies diseased nodes as suspicious more correctly than it identifies normal nodes as normal. Overall, radiological scoring of the axillary lymph nodes based on morphological characteristics showed moderate accuracy in distinguishing disease or no-disease in the nodes based on standard clinical breast MRI data. Improvement in image signal-noise-ratio (SNR), spatial resolution and contrast of the axillary lymph nodes is needed.

Introduction

Post-neoadjuvant chemotherapy (NAC), pathological staging of the axillary lymph nodes (aLNs) using sentinel biopsy or dissection is critically important for prognosis and subsequent management of breast cancer patients. However, this approach has several limitations; it is invasive, has significant negative side-effects and exposes patients to systemic radioactivity (for sentinel biopsy), questioning the generic use of this procedure. A non-invasive imaging approach that enables staging of disease of all aLNs in situ may prove useful.

Although MRI is used for radiological staging of the breast tumor, similar radiological scoring of aLNs based on MRI is not currently used in clinical practice because aLNs are small in size and sometime outside of field of view.

The goal of this study was to evaluate a radiological nodal staging (RNS) metric for accurate identification of diseased aLNs. Comparison was made with pathological nodal staging (PNS) as the ground truth. We also evaluated improvement of prediction accuracy with addition of nodal and tumor volume to RNS receiver operating curve (ROC) analysis.

Methods

Ninety-four patients with stage 2 or 3 breast cancer receiving NAC from ISPY-1 [1] clinical trial with acceptable image quality of the axilla were analyzed. Standard unilateral breast MRIs were acquired at 1.5T. High resolution (0.7x0.7x2.0-mm) post-contrast MRI images were used for analysis. Post-NAC MRI was scored based on RNS. Comparison was made with post-surgical PNS. Patients were scored for suspicion of disease in the axilla based on morphological features of the cortex and hilum. A higher suspicion was assigned to patients with thickened cortex and absence of fatty hilum. Figure 1 shows visual examples of these categories.

A generalized linear model was generated for the single and multi-predictor combinations. The prediction performance of this model was assessed by ROC analysis.

Results

Of our 94 patient cohort, 46 patients had nodal pathological complete response (PCR). To parallel the definition of PCR, we separated the scores into two groups: no-disease in the axilla (PNS=0 and RNS=0) versus any disease in axilla (PNS 1,2,3 and RNS 1,2,3). RNS was evaluated taking all aLNs per patient similar to pathological analysis (Table 1). For residual nodal disease (PNS 1,2,3), there was 83% agreement in which the MRI showed suspicion. For no residual nodal disease (PNS 0), there was 11% agreement in which the MRI showed no suspicion.

Similar scoring was performed for the most suspicious node for each patient (Table 2). There was a slight increase in the scoring with no suspicion with PCR of 1 compared to Table 1. This suggests that it is more accurate to detect suspicion by looking at each individual node instead the axilla as a whole.

We evaluated whether RNS can accurately detect diseased aLNs. The area under the curve (AUC) based on RNS of the node alone was 0.55. Addition of nodal and tumor volume yielded an AUC of 0.66. (Table 3). The sensitivity increased to 87% showing that tumor and aLN volume added confidence in identifying axilla to be disease-free.

Discussion

Comparison of RNS and PNS (Table 1) suggests that it is easier to identify a diseased node as suspicious. This reflects the reason why many aLN biopsies/dissections are negative, underscoring the notion that invasive biopsies could be avoided if a more reliable method becomes available [2].

From our analysis, 13-26% of patients had no suspicion in the axilla on the MRI. We hypothesize the reason for this is that the ISPY data analyzed did not have the contralateral axilla for comparison of that patient “normal”. It has been shown by Baltzer et al that symmetry of the axilla is shown to increase the negative predictive value of excluding metastasis [3].

Imaging has potential to minimize or obviate unnecessary surgical procedures in the case of patients that have a pathological complete response correctly reflected as radiological complete response [4]. However, our findings indicate that RNS based on morphological characteristics had moderate accuracy for detection of residual disease. Future studies should improve image signal noise ratio (SNR), spatial resolution, and contrast of the aLNs. In particular, high resolution of the axilla, especially of small nodes, can aid on the identification of the fatty hilum which is challenging for many aLN at current spatial resolution.

Conclusion

Acknowledgements

References

[1] Hylton NM et al. Neoadjuvant Chemotherapy for Breast Cancer: Functional Tumor Volume by MR Imaging Predicts Recurrence-free Survival-Results from the ACRIN 6657/CALGB 150007 I-SPY 1 TRIAL. Radiology. 2016 Apr;279(1):44-55. doi: 10.1148/radiol.2015150013. Epub 2015 Dec 1.

[2] Zahoor S et al. Sentinel Lymph Node Biopsy in Breast Cancer: A Clinical Review and Update. J Breast Cancer. 2017 Sep;20(3):217-227. doi: 10.4048/jbc.2017.20.3.217. Epub 2017 Sep 22.

[3] Baltzer PA et al. Application of MR mammography beyond local staging: is there a potential to accurately assess axillary lymph nodes? evaluation of an extended protocol in an initial prospective study. AJR Am J Roentgenol. 2011 May;196(5):W641-7. doi: 10.2214/AJR.10.4889.

[4] van la Parra RF et al. Selective elimination of breast cancer surgery in exceptional responders: historical perspective and current trials. Breast Cancer Res. 2016 Mar 8;18(1):28. doi: 10.1186/s13058-016-0684-6.