Meta-analysis of tumour burden in pre-operative axillary ultrasound positive and negative breast cancer patients

The management of the axilla in breast cancer has seen the emergence of surgical conservatism away from ANC in view of its lack of disease-free and overall survival benefit [4, 22, 23]. This has resulted in the increasing importance of the quantification of axillary disease in order to differentiate between ‘high’ and ‘low’ axillary burden. The current criteria for this is based upon those established within the ACOSOG Z0011 trial [4] and then formally supported by the American Society of Clinical Oncologists (ASCO) [24]. During the ACOSOG Z0011 [4] trial pre-operative axillary ultrasound was not performed. This is directly in conflict with mainstream practice, which directs those patients with suspicious axillary findings on ultrasound to undergo tissue sampling by either core biopsy or fine needle aspiration. Those patients identified to have metastatic disease pre-operatively, are streamlined directly to ANC. This means that out of those patients who have metastatic axillary disease, one-third (Table 2) of them would have undergone SNB rather than ‘fast track’ ANC if they were part of the ACOSOG Z0011 trial and half of these would have fit criteria of ‘low axillary burden’ and been omitted ANC. This certainly is suggestive that the use of pre-operative ultrasound is directing a greater number of patients to ANC and denying them axillary preservation than is otherwise necessary.

It has been established that axillary burden in those patients with pre-operative positive ultrasound findings is greater than in those that are negative. Van Wely et al. [13] in their meta-analysis of ultrasound-guided biopsy for determining axillary burden demonstrated that the number of involved nodes was significantly higher in patients in whom axillary metastases were identified by ultrasound-guided biopsy (P < 0.001). Unfortunately, due to the limitations of the data available at the time, the authors were only able to categorize axillary burden in the US positive and negative groups to one to three and greater than three nodes and were unable to segregate ITC, micro and macrometastatic disease. This prevented the relevance of their data to modern guidelines and the criteria for ‘low and high axillary burden’ [24]. This meta-analysis supports the finding of van Wely et al. [13] as we too found that there was significantly higher axillary burden in the pre-operative positive US group compared to the negative one (OR 5.95, 95% CI 5.80–6.11). This was at the clinically relevant threshold of two nodes or fewer and greater than two nodes, in addition to only considering macrometastases. Our meta-analysis also demonstrated that the mean number of nodes expected to be retrieved in patients with a positive US is twice that of those with a negative US [3.0 vs. 1.6 (SE 0.2)]. It has also been shown that the mode distribution of involved nodes is one and two to three for US negative and positive patients, respectively. This clearly demonstrates that axillary burden is higher in the US positive group. It is logical, as the sensitivity of US has been consistently shown to be inferior to that of SNB for the identification of metastatic disease [11, 12]. Therefore, for disease to be visible to the radiologist one would expect that it might be of greater quantity.

However, in an era of increasingly targeted and personalized therapy, it is not adequate to consider these data purely within the context of ‘greater axillary burden’ present in the US positive group. This meta-analysis identified 43% of US positive patients had two or fewer nodes consistent with low axillary burden, which would allow them to avoid ANC. This raises the question of how do we prevent the overtreatment of this group? For the 43% of patients receiving an unnecessary ANC, a possibility could exist in avoiding the performance of pre-operative axillary ultrasound. Such a technique would increase the patients directed to SNB rather than ‘fast track’ ANC by approximately one-third and of this one-third, half would avoid cANC on the basis of subsequent SNB histology. The only patients fast tracked to ANC would be those patients with palpable axillary lymphadenopathy—consistent with the protocols of RCTs of SNB [5‐10]. This approach would make a contribution, but may give the appearance of a regressive step and could be detrimental in patients with a high body mass index where axillary ultrasound has been demonstrated to possess great accuracy [25]. It is known that this approach to omitting pre-operative axillary ultrasound has already been adopted in certain academic breast units within the USA. However, published data on the outcomes of this practice are not currently available and the American Society of Breast Surgeons [26] still advocates pre-operative axillary ultrasound in the presence of diagnosed breast cancer. Another approach would be to improve quantification of axillary US findings—which is difficult with the inter-personal variation associated with the performance of US. However, the impetus should be placed upon the radiologist to quantify the number of suspicious nodes visible. The suitable threshold to be used for this would have to be ascertained through further clinical evaluation, as it cannot be assumed to extrapolate to two or fewer visible nodes on US. Only the study by Caudle et al. [19] demonstrated that by using a threshold of two suspicious nodes or fewer that the number of patients with two or fewer involved nodes at ANC was 55%. If a threshold of one suspicious node only was applied this percentage would inevitably increase. Therefore, it is likely that the cut off on axillary ultrasound should be one to two suspicious nodes but confirmation of this should form the basis of future research—as such data are minimal within the literature. In addition to quantification, the suspicious nodes could also be marked pre-operatively at US using clips, which could then allow targeted intra-operative excision. This technique has successfully reduced the false negative rate at SNB within the neo-adjuvant chemotherapy setting to under 2% [27]. It has to be stressed that unless ultrasound evolves and is approached in this manner, it is likely to become disadvantageous and its role increasingly questionable.

Axillary radiotherapy could play a role in management but it must be remembered that the AMAROS trial [3] did not use pre-operative axillary ultrasound and excluded patients on the basis of palpable lymphadenopathy—consistent with the original trials of SNB [5‐10] and ACOSOG Z0011 [23]. Whilst there was no threshold limit of axillary burden to exclude patients from randomization, only 5% of patients in the radiotherapy and ANC arms had a burden of three or more involved nodes at SNB. It is also important to note that only 60% of involved nodal burden in each arm consisted of macrometastases. Axillary radiotherapy certainly could be considered a therapeutic option for the 43% of patients who fit the low axillary burden criteria (two or fewer involved nodes) despite a histologically involved axilla on pre-operative US. However, as a blanket treatment of patients with burden greater than two nodes, it would be unproven.

The issue of patients undergoing mastectomy and their axillary burden was not addressed by the ACOSOG Z0011 trial [23]. Within the included trials of this meta-analysis, mastectomy rates varied between 32 and 65% [19, 21]. Whilst it may be assumed that the ACOSOG Z0011 criteria could be extrapolated to mastectomy patients, it would mean that patients who would have avoided chest wall radiotherapy in the event of ANC may become committed to it if no further intervention is performed to the axilla. This would be the only safe option in the absence of further clinical trial data.

It is also acknowledged that whilst attempting to avoid ANC in the identified cohort of 43%, it means that the remaining 57% of patients, who do not fit the criteria of ‘low axillary burden’ will have to undergo two operations with the associated additional exposure to a general anaesthetic and potential financial implications—although the latter could be offset by the saving in the ‘low axillary burden’ group avoiding ANC. In other aspects such as the administration of primary chemotherapy, this approach would not detrimentally influence patient management and would allow accurate axillary staging with SNB post-chemotherapy, especially if ‘clipping of suspicious nodes’ is performed [27, 28].

The use of other imaging modalities should be considered to complement physical examination, particularly in obese patients where such clinical assessment is very challenging. Magnetic resonance imaging (MRI) has failed in mainstream uptake for axillary staging in breast cancer due to its high false positive rate when compared to SNB (6.3 vs. 0.2%, respectively) [29]. This is a significant limitation when a key aim of modern axillary management is to avoid overtreatment of the axilla or if MRI is being used to identify ‘high axillary burden’ and ‘fast track’ to ANC—avoiding SNB altogether. A systematic review of positron emission tomography (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 was 63 and 94%, respectively [30]. Clearly, these areas of research should continue to be explored but will also face challenges within the financial constraints of public healthcare systems.

It must be remembered that such advancements are not only targeted at reducing the 43% of patients with positive US who have an unnecessary ANC but also the 79% of US negative patients with low axillary burden found on SNB who could potentially have avoided SNB. The case of the latter may well be addressed by ongoing current research [31] but the 43% provide a greater long-term area of concern. Only the article by Caudle et al. [19] attempted to identify discriminating factors for high axillary burden within the ultrasound positive group. On multivariate analysis they identified lobular histology as a risk factor (OR 1.77; 95% CI 1.06–2.95). Urgent analysis of further risk factors within this group, are warranted to prevent overtreatment and could potentially involve the use of molecular profiling. In the void of advanced imaging modalities, quantification of axillary ultrasound is essential in non-palpable axillary disease. Within this meta-analysis, none of the included articles [16‐21] provided any details on the quantification of axillary disease on US, highlighting that this is clearly an area, which needs to be improved upon. It would be suggested that the number of abnormal axillary lymph nodes should be clearly stated by the conducting radiologist. Those with inconsistent ultrasound findings or a single abnormal node should be offered SNB without axillary ultrasound directed biopsy. Those patients with more than one abnormal node or those with an abnormal node who are about to commence primary chemotherapy, should be considered for biopsy and clipping of the node(s). Further studies are needed to refine the technique of clipping the node and SNB with removal of the clipped node. The only other current option would be to avoid axillary ultrasound altogether to avoid overtreatment.