Sentinel lymph node biopsy (SLNB) is the standard of care for axillary staging of patients with breast cancer who have a clinically and radiologically negative axilla [1–6]. Preoperative ultrasound with fine-needle aspiration (FNA) or core biopsy of abnormal lymph nodes is used to identify node-positive patients who can proceed directly to axillary lymph node dissection (ALND) and therefore avoid an additional unnecessary surgical procedure. However, the sensitivity of ultrasound-guided FNA/biopsy for the preoperative detection of axillary nodal metastases is variable with figures ranging between 25 and 97 % [7, 8]. Houssami et al. [9] in a meta-analysis found the sensitivity and specificity of ultrasound-guided FNA/biopsy for axillary staging to be 79.6 and 98.3 % respectively. The inclusion of only patients undergoing ultrasound-guided FNA or biopsy in their calculations as opposed to all patients undergoing ultrasound imaging artificially inflates the sensitivity rates for studies [10, 11] within the systematic review [9]. More realistic figures are described by Britton et al. [12] whose assessment of 139 patients with proven invasive breast cancer demonstrated that the sensitivity of ultrasound-guided core biopsy for the identification of involved axillary lymph nodes was 53.4 %. Core biopsy sensitivity also varied between macro- and micrometastatic lymph node involvement with figures of 60.3 % and 30 % respectively. Of the 73 lymph node-positive patients confirmed on SLNB, only 5 (7 %) exhibited suspicious ultrasound appearances with an irregular outline and absence of a fatty hilum, and all of these yielded a malignant core biopsy result. From the remaining positive nodes, 8 (11 %) had a smooth outline but no fatty hilum and 7 of these produced a malignant core biopsy result; 22 (30 %) were multilobulated nodes with malignant core biopsy results in 17; 11 (15 %) unilobulate nodes of which 7 were positive on core biopsy; 25 (34 %) nodes with a smooth outline and fatty hilum of which 3 were positive on core biopsy. Cho et al. [13] found in their study of ultrasound-guided needle localisation of axillary nodes and subsequent confirmatory surgical SLNB, that a significant correlation between increasing cortical thickness of nodes and the presence of malignancy was present (P < 0.001). They found that a cortical thickness of >3.5 mm and absence of a fatty hilum was associated with 90 % malignancy compared with 2 % for a thickness of <1.5 mm. The work of Garcia-Ortega et al. [14] supports this since the ultrasonographic finding with the highest positive predictive value for malignancy was absence of a fatty hilum (93.1 %). This raises the question of how best to target axillary nodes for core biopsy if 34 % of normal-appearing nodes are positive for malignancy [12]. Certainly more intelligent targeting of the SLN using ultrasound contrast agents such as microbubbles may be of benefit. Sever et al. [15, 16] have demonstrated that contrast-enhanced ultrasound (CEUS) using intradermal injection into the breast of phospholipid-stabilised microbubbles containing sulphur hexafluoride gas with a mean diameter of 2.5 μm, can be successful. In 2 studies consisting of 44 and 80 patients they were able to confirm a sensitivity of 89 % for SLN detection compared with radioisotope and blue dye [15, 16]. However, when Cox et al. [17] consequently used CEUS with microbubbles in a larger cohort of 371 patients, the sensitivity was found to be only 61 % and therefore comparable to non-enhanced ultrasound. This technique has also been applied to a porcine model and the execution of percutaneous excision of the SLN using the Intact Breast Lesion Excision System (BLES, Intact Medical, Framingham, MA, USA) [18]. In this porcine study Sever et al. [18] demonstrated that the excision of the whole SLN was successful in the groins of the three pigs tested and that the nodal architecture of the removed nodes was preserved on microscopy. Further refinement of such a system with improved sensitivity could provide an alternative to surgical SLNB. Other techniques such as ultrasound elastography have also shown potential for preoperative axillary staging in breast cancer. Taylor et al. [19] found that the sensitivity, specificity, positive predictive value and negative predictive value of conventional ultrasound were 76, 78, 70 and 81 % respectively; 90, 86, 83 and 93 % respectively for visual ultrasound elastography; and for strain scoring, 100, 48, 58 and 100 % respectively. Wojcinski et al. [20] in their study using real-time ultrasound elastography found a significantly harder cortex in metastatic lymph nodes (P = 0.006) and a sensitivity and specificity of 40 and 96.8 % respectively for B-mode ultrasound compared to 60 and 79.6 % respectively for ultrasound elastography. They found that the highest sensitivity of 73.3 % was achieved by the combination of conventional ultrasound and elastographic features of cortical thickness >3 mm on B-mode ultrasound or blue cortex on the elastogram. The highest specificity of 99.3 % was achieved by findings of a cortical thickness >3 mm on B-mode and blue cortex on the elastogram. These studies suggest that ultrasound elastography could act as an adjunct to conventional ultrasound to improve the performance of conventional ultrasound alone. However, in the current clinical context by using preoperative ultrasound staging of the axilla routinely, potentially up to 75 % of patients will still have a negative SLNB and undergo an unnecessary invasive procedure with its associated morbidity. Furthermore, ultrasound is of limited value for intraoperative localisation of sentinel nodes in view of limited surgical ultrasound training and the limitations of using portable ultrasound machines to identify departmental radiological findings.

A growing number of studies and trials support the concept that patients with low axillary tumour burden do not benefit from an ALND [21–24]. The recent update by Galimberti et al. [23] of the International Breast Cancer Study Group (IBCSG) trial 23–01 demonstrated that an ALND can be omitted in patients with very low volume metastatic nodal involvement. This randomised, multicentre, phase III clinical trial compared ALND with no ALND in patients with micrometastases alone in the sentinel lymph node (SLN). A total of 934 clinically node-negative patients were randomised, with a primary tumour of less than 5 cm and with less than 2-mm tumour focus in one or two SLNs. Two thirds of patients (67 %) had tumours less than 2 cm, and 89 % were oestrogen-receptor positive (ER+). The 5-year disease-free survival (DFS) and overall survival (OS) rates in the ALND- versus SLNB-only groups were 87.3 versus 88.4 % (P = 0.48) and 97.6 versus 98 % (P = 0.35) respectively. The SOUND trial (Sentinel node vs. Observation after axillary UltrasouND) [25] will assess ultrasound for preoperative staging of the axilla. This prospective, multicentre, randomised controlled study is recruiting patients with breast cancers less than 2 cm in size and suitable for breast-conserving surgery and SLNB, who undergo preoperative axillary ultrasound to rule out suspicious nodal involvement. Patients with a single indeterminate lymph node undergo ultrasound-guided fine-needle aspiration. Patients with either negative cytology of the single indeterminate lymph node or with negative ultrasound of the axilla, are randomised into ether conventional management (SLNB ± ALND) or no axillary surgical staging. In the conventional arm, ALND is only performed in the presence of macrometastases, but not for isolated tumour cells or micrometastases. The primary endpoint is distant disease-free survival, which will act as a proxy for overall survival, allowing results to be acquired in a shorter period of time than overall survival. Secondary endpoints will be cumulative incidence of distant and axillary recurrences, disease-free survival and overall survival.

In patients with nodal metastasis, the ACOSOG Z0011 trial [24] demonstrated that when ALND was omitted there was extremely low local recurrence and excellent overall survival, so long as whole breast radiotherapy and systemic therapy were administered. In this trial 27 % of patients randomised to ALND were found to have further axillary involvement and by inference a similar proportion in the other study arm would have residual axillary disease. However, about 40 % of patients entered into ACOSOG Z0011 [24] had micrometastases and the trial was criticised for not meeting the planned accrual and for lacking statistical power. Therefore, with very low axillary tumour burden and in line with the IBCSG 23–01, axillary surgery can be omitted. This leaves the questions of how to identify patients with macrometastases reliably and what the future of SLNB will be. If we are pursuing a less invasive approach, it is time to consider non-invasive techniques to assess the axilla.

Axillary MRI provides an alternative to SLNB in determining axillary burden. Diffusion-weighted MRI (DWI) has been applied to the evaluation of primary breast lesions and Kamitani et al. [26] assessed its role in the detection of axillary metastases. However, in their study they found a sensitivity and specificity of only 53.8 and 86.9 % respectively for detecting axillary metastases in the 110 breast cancers that were evaluated. Scaranelo et al. [27] assessed the use of unenhanced MRI versus diffusion weighted (DW) imaging for axillary staging in 61 patients with breast cancer and correlated this with histopathological findings. They demonstrated that the sensitivity, specificity and accuracy for unenhanced axial T1-weighted MRI was 88, 82 and 85 % respectively and 84, 77 and 80 % for DW imaging. They demonstrated the accuracy of preoperative axillary evaluation using unenhanced MRI and the reproducibility of DW imaging, but its unlikely clinical function. Memarsadeghi et al. [28] demonstrated that unenhanced MRI was inferior to ultrasmall superparamagnetic iron oxide (USPIO)-enhanced MRI with sensitivity, specificity and accuracy of 55, 81 and 79 % respectively for unenhanced MRI and 100, 98 and 98 % for the detection of axillary involvement. Fornasa et al. [29] performed their own characterisation of axillary lymph nodes using DW-MRI and found that the apparent diffusion coefficients (ADCs) of lymph nodes with metastases were significantly lower (mean: 0.878 × 10-3 mm2/s; range: 0.3-1.2) than those with benign lymph nodes (mean: 1.494; range: 0.6-2.5) (P < 0.001). By applying this refinement they established a threshold value of 1.09 × 10-3 mm2/s DWI resulting in a sensitivity, specificity and accuracy for metastatic lymph node identification of 94.7, 91.7 and 93 % respectively. Nakai et al. [30] compared the outcomes of DW-MRI and conventional MRI using USPIO contrast enhancement. They demonstrated that whilst more axillary lymph nodes are identified on conventional MRI compared to DW-MRI (83 % versus 76 % respectively), the sensitivity, specificity and overall accuracy was 83, 98 and 95 % respectively for DW-MRI and 70, 98 and 93 % for conventional MRI respectively. Mortorello et al. [31], in their assessment of axillary staging using contrast-enhanced (gadodiamide) MRI, found that the presence of an axillary node with no fatty hilum and the number of nodes with no fatty hilum on MRI significantly correlated with pathologic node positivity (P = 0.04), whilst kinetics, node number and node size did not correlate. Ex vivo characterisation of axillary nodes by Luciani et al. [32] demonstrated that an MR short axis threshold of 4 mm yielded the best predictive value for metastatic nodal involvement with a sensitivity and specificity of 78.6 and 62.3 % respectively. Other factors significantly correlated with metastatic lymph node involvement were irregular contours (sensitivity 35.7 and specificity 96.7 %), central nodal hyperintensity on inversion recovery T2-weighted images (sensitivity 57.1 and specificity 91.4 %) and a cortical thickness of >3 mm (sensitivity 63.6 and specificity 83.2 %).

Data from studies using superparamagnetic iron oxide (SPIO) indicate that these contrast agents have sensitivity and specificity for the detection of axillary involvement of 98 and 96 % respectively [33]. Meng et al. [34] assessed the cost-effectiveness of MRI and PET for the evaluation of axillary lymph nodes in early breast cancer. They found that if MRI could accurately diagnose axillary involvement, the most cost-effective strategy was to replace SLNB with axillary MRI. With this strategy, true-positive patients would undergo a single surgical procedure (ALND) replacing two sequential surgical procedures (SLNB followed by ALND). True-negative patients would not require any surgery. The challenge of applying this to routine clinical practice is the false-positive rate of 6.3 % for MRI versus 0.2 % for SLNB [34]. Harnan et al. [33] in their systematic review considered the use of MRI assessment of axillary lymph node status in early breast cancer and recorded a mean sensitivity and specificity of 90 and 95 % respectively. The highest mean sensitivity and specificity were seen with superparamagnetic iron oxide (SPIO)-enhanced MRI, with values of 98 % and 96 % respectively. Stadnik et al. [35] compared the efficacy of USPIO-enhanced MRI with 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) for axillary staging in ten patients with breast cancer who subsequently underwent axillary clearance. Histopathological axillary staging was negative for nodal malignancy in five patients and positive in the remaining five. There was one false positive for USPIO-enhanced MRI and one false negative for FDG-PET. This provided a sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of 100, 80, 80 and 100 % for USPIO-enhanced MRI and 80, 100, 100 and 80 % respectively for FDG-PET. However, the combination of USPIO-enhanced MRI and FDG-PET provided sensitivity, specificity, PPV and NPV of 100 %, demonstrating its ability to identify patients for direct axillary clearance and avoiding SLNB. To this effect, the future prospect of the combination of radionuclide-based imaging techniques such as PET and single photon emission computed tomography (SPECT) with MRI could represent the next generation of imaging investigations. Torres Martin de Rosales et al. [36–38] demonstrated that dual-modality imaging agents based on the conjugation of radiolabelled bisphosphonates (BP) directly to the surface of SPIO provided excellent stability and allowed for in vivo co-localization of lymph nodes in a murine model. The application of these dual-imaging agents to future human studies provides exciting prospects to improve the sensitivity and specificity of preoperative axillary staging. Peare et al. [36] in their meta-analysis of FDG-PET for axillary staging assessed 25 eligible studies confirming the efficacy of FDG-PET compared to histopathological assessment after ALND/SLNB. They found that from a combined population of 2,460, there were 703 true positives and 1,288 true negatives, 339 false negatives and 130 false positives. Summary receiver-operating characteristic (SROC) curves were plotted for the aggregate data. This analysis found an area under the curve (AUC) of 0.95 (95 % CI 0.91-0.97) and a Q* value of 0.89 (95 % CI 0.85-0.92). For the AUC and Q* (sensitivity = specificity) a value closer to 1 represents perfect performance. Therefore, in radiological terms, the outcome of FDG-PET would represent good performance. However, Peare et al. [36] concluded that the performance of FDG-PET was not sufficient to replace surgical assessment of the axillary lymph nodes. The reason for this was inclusion of studies reporting on false-negative results because of the inability of FDG-PET to detect micrometastatic deposits. Cooper et al. [39] in their systematic review of PET for the assessment of axillary lymph node status in early breast cancer found that across 26 studies evaluating PET or PET/CT, the mean sensitivity and specificity were 63 and 94 % respectively. On this basis replacing SLNB with PET would avoid the adverse effects of SLNB, but lead to more false-negative patients at risk of recurrence and more false-positive patients undergoing unnecessary ALND, meaning that currently it could not be accepted as the standard of care over SLNB.

Superparamagnetic iron oxide injected interstitially into the periareolar region of the breast has also been successfully used to guide SLNB [40, 41]. Johnson et al. [41] showed that the 13 nodes containing metastases had variable quantities of iron within them, but the iron was not present in the areas of the node containing the metastasis. This study therefore demonstrated that heterogeneous enhancement of the SLNs and non-SLNs on contrast-enhanced MRI indicated a metastatic focus. This technique could potentially identify patients with significant axillary involvement preoperatively, who therefore require ALND. The current SentiMAG Multicentre Trial [42], a prospective phase II non-randomised clinical trial comparing SLNB using a magnetic tracer and handheld magnetometer versus the standard technique, will determine if the novel technique is viable in a larger patient cohort. This study has an MRI sub-protocol to evaluate preoperative axillary MRI (with a magnetic tracer) for SLN imaging and characterisation. The findings of this preoperative imaging will then be correlated with the outcome of SLNB to determine the ability of SPIO-enhanced MRI to characterise the sentinel lymph nodes preoperatively. However, this will be of limited clinical value since the CE marked magnetic dye is not a contrast agent and can only be used to localise sentinel nodes. Future contrast agents should further evaluate these findings.

Selective axillary surgery has been evaluated in trials assessing feasibility of SLNB in patients presenting with node-positive disease and who undergo primary chemotherapy. The ACOSOG Z1071 Trial [43] evaluated the utility of SLNB after primary systemic treatment (PST) in 756 patients presenting with node-positive disease. At the time of surgery, all patients underwent SLNB followed by ALND. The SLNB identification rate was 92.5 % and SLNB correctly identified nodal status in 84 % of cases. Of those patients in whom an SLN was identified there was a complete pathological response in 40.3 %. Of the patients with a positive SLNB, the SLN was the only site of disease in 40 %. The false-negative rate for SLNB was 12.8 %. These findings are also supported by the SENTINA Trial [44], which considered the timing of SLNB in PST. In this four-arm, prospective, multicentre cohort study, patients who were clinically node negative before PST underwent SLNB and those who were found to be negative underwent no further axillary surgery (group A); those who were positive underwent post-PST SLNB and ALND (group B). Patients who wer clinically node positive before PST did not undergo any axillary intervention until completion of PST, when they were divided into those who became clinically node negative and underwent SLNB and ALND (Group C) and those who remained node positive and underwent ALND (group D). They interestingly found that 70.8 % of patients in group B and 52.3 % in group C converted to node-negative status. The false-negative rate was 14.2 % in group C but as high as 51.6 % in group B. Irrespective of the high false-negative rates in group B, both SENTINA [44] and ACOSOG Z1071 [43] suggest that there might be patients who would benefit from a more targeted surgical approach to the axilla rather than a routine, one-size-fits-all ALND. With preoperative imaging it could be possible to identify the PST responders in ACOSOG Z1071 [43] and SENTINA [44] and those who only have low axillary tumour burden (40 % of patients with only the SLN involved in ACOSOG Z1071 [43]) and surgically only excise those involved nodes, reducing the morbidity associated with ALND.

Axillary tumour burden reflects the biology of the tumour. Ideally, the extent of axillary surgery should be guided by the actual rather than the likely axillary tumour burden. To evaluate this concept a robust axillary imaging technique is required. The concept of selective axillary surgery could be used to identify and remove only involved axillary nodes. This would rely on improved preoperative imaging combined with novel techniques to guide surgeons intraoperatively to involved nodes. Patients with a high clinical index of suspicion for axillary involvement could then be offered further imaging, and those with a low index of suspicion, conservative management. The future for SPIO contrast-enhanced MRI is promising and it is essential that funding is prioritised to include imaging trials and sub-protocols to evaluate preoperative axillary MRI. These trials will allow us to move forwards to a truly minimally invasive approach and individualised axillary management in breast cancer.