Feasibility of using negative ultrasonography results of axillary lymph nodes to predict sentinel lymph node metastasis in breast cancer patients

Abstract Knowledge of the pathology of axillary lymph nodes (ALN) in breast cancer patients is critical for determining their treatment. Ultrasound is the best noninvasive evaluation for the ALN status. However, the correlation between negative ultrasound results and the sentinel lymph nodes (SLN) pathology remains unknown. To test the hypothesis that negative ultrasound results of ALN predict the negative pathology results of SLN in breast cancer patients, we assessed the association between ALN ultrasonography‐negative results and the SLN pathology in 3115 patients with breast cancer recruited between October 2010 and April 2016 from a single cancer center, prospective database. Of these patients who met the inclusion criteria, 2317 (74.4%) had no SLN pathological metastasis. In the univariate analysis, other 798 patient with positive SLN tended to be under age 40 and premenopausal, having large tumor sizes (>2 cm), higher histological grade of primary tumor, positive hormone receptors, and negative HER‐2 status (P < .05 for all). In the multivariate analysis, menstrual status, tumor size, ER status and histological types of primary tumor remained to be independent predictors for SLN pathological metastasis. The area under curve (AUC) was 0.658 (95% CI = 0.637‐0.679), P > .05. In conclusion, only a 74.4% consistency between ALN ultrasonography‐negative results and negative pathological SLN results, although menstrual status, tumor size, histologic subtypes of primary tumor and ER status were found to be statistically independent predictors of positive SLN among patients negative for ALN ultrasonography. Therefore, the present study suggests that negative ultrasound results of ALN do not adequately predict the negative pathology results of SLN in breast cancer patients.


| INTRODUCTION
The pathological status of axillary lymph nodes (ALN) is critical in the treatment of breast cancer patients. Sentinel lymph node (SLN) biopsy is the standard of care for clinical management of breast cancer patients. [1][2][3][4] The SLN pathological result-based treatment may prevent patients from undergoing unnecessary ALN dissection. Ultrasound is currently the best noninvasive method to evaluate ALN. [5][6][7][8] Due to improvements in ultrasonography equipment, advances in technology and refined diagnostic standards, the consistency between ALN ultrasonography-negative results and negative SLN results has been improved. The SOUND (sentinel node vs observation after axillary UltraSouND) trial 9 was established to compare SLN biopsy vs observation, when axillary ultra-sound is negative in patients with small breast cancer sizes, who are candidates for breast conserving surgery. However, to our knowledge, few studies have evaluated the association between ALN ultrasonographynegative results and pathological SLN results. Moreover, the influence of clinicopathological factors on SLN results in ALN ultrasonography-negative patients has not been reported. Based on our single cancer center prospective database, we retrospectively analyzed the possible effects of age, menstrual status, primary tumor size, histologic subtypes of primary tumor, hormone receptor status, and HER-2 status on SLN pathology results in breast cancer patients with ALN ultrasonographynegative results, and evaluated the feasibility of predicting the SLN pathological results for ALN negative ultrasonography in these patients. We hypothesize that ultrasonography together with clinical indexes predict SLN metastasis for breast cancer patient with ALN ultrasonography-negative results.

| Patients
Data were extracted from the prospective database of Peking University Cancer Hospital Breast Cancer Prevention and Treatment Center. All subjects included in the present study consisted of consecutively unselected patients treated between October 2010 and April 2016.

| Inclusion criteria
The primary tumor was diagnosed as an invasive breast cancer by core needle biopsy; ALN were ultrasonographynegative; Patients underwent SLN biopsy in the same department; Patients were untreated before undergoing SLN biopsy.

| Exclusion criteria
The invasive breast cancer was pathologically diagnosed by surgical resection rather than by core needle biopsy; ALN were ultrasonography-positive; Patients underwent treatment prior to SLN biopsy.

| Ultrasonography
Axillary ultrasonography was performed by 3 sonographers who had at least 5 years of experience. Ultrasonography examinations were performed with an Acuson Antare unit (Siemens Medical Solutions, Mountain View, CA), a Mylab90 unit (Esaote, Genova, Italy) and a LOGIQ E9 unit (GE Healthcare, Milwaukee, WI) with transducers of 9 MHz or higher frequencies. ALN were divided into 2 groups: The ultrasonography negative axilla group including no ALN detected or a cortical thickness <3 mm. The definition of ALN-positive ultrasonography results was consistent with the standards of the cancer center. ALN-positive ultrasonography results were determined by one of the following 3 criteria. 10 (1) Regular target annular lymph node with a peripheral cortical thickness ≥3 mm; (2) Eccentric target annular lymph node with a local cortical thickness ≥3 mm; and (3) Hypoechoic lymph node without a hilus structure of the lymph node.

| SLN biopsy
The 99m Tc-rituximab was used as the tracer for SLN biopsy. The tracer was injected into 2 of the following sites: the peritumoral breast parenchyma, subcutaneous and subareolar tissues. The injection dose was 18.5 MBq, injected in the morning of the surgery day, or 37 MBq on the day before surgery. SLN biopsy was performed under local anesthesia. SLN were identified by a handled gamma detection probe (Neoprobe, USA or Crystal, Germany). All radioactive nodes with a counting rate ≥10% of the hottest node were removed.
These data were used to correlate the results of ALN ultrasonography-negative and SLN biopsy pathology and to be analyzed for possible influence of clinicopathological factors on SLN biopsy pathological results.

| Statistical methods
In the univariate analysis, the patients' age, menstrual status, primary tumor size, histological subtypes of primary tumor, hormone receptor status, and HER-2 status were evaluated for the association with the SLN pathologic results. A Chisquare test was used in the univariate analysis to evaluate the factors that influenced the SLN pathologic results, and a value of P < .05 was defined as statistically significant. In the multivariate analysis, factors that showed to be relevant to the SLN results in the univariate analysis were included in the multivariate logistic regression model. The receiver operating characteristic (ROC) curve employed to compare among variables for the ability to predict diagnostic results.
Area under the ROC curve (AUC) was measured to test the prediction performance. AUC > 0.75 was defined as statistically significant.

| RESULTS
Data were extracted from the prospective database, and all subjects included in the present study consisted of consecutively unselected patients treated between October 2010 and April 2016, resulting in a total of 5978 cases. According to the inclusion and exclusion criteria, 3115 patients with ALN ultrasonography-negative results who underwent SLN biopsy were included in the comparison of the ALN ultrasonography-negative results and SLN pathological results, including 2317 (74.4%) patients without SLN metastasis and 798 (25.6%) patients with SLN metastasis ( Table 1).
Subsequently, multivariate analyses were carried out for those variables found to be statistically significant on univariate analyses. As a result, menstrual status, tumor size, ER status, and histologic subtypes of primary tumor remained to be independent predictors in the multivariate analyses ( univariate, but not in multivariate, analyses were correlated with those in the multivariate analyses. For example, age was highly correlated with menstrual status (r 2 = .146, P < .001), and both PR and HER-2 were highly correlated with ER (r 2 = .444, P < .001 and r 2 = .062, P < .001, respectively). However, the AUC for the prediction model was 0.658 (95% CI = 0.637-0.679), P > .05 (Table 5 and Figure 3).

| DISCUSSION
Axillary lymph nodes pathological status is important in the treatment strategy for breast cancer, and ALN evaluation includes several noninvasive and invasive methods, of which ultrasound has received much attention in modern breast cancer management, because it is noninvasive, highly sensitive to lymph nodes, indicates and guides further possible invasive examinations, and has a high pathological predictive value. 11,12 SLN biopsy is the gold standard for accurately evaluating ALN pathological status in breast cancer patients. The possibility of exempting routine SLN biopsy for patients with ALN ultrasonography-negative results has become a focus of research.
In the present study of comparing patients of ALN-negative ultrasound results with pathological SLN biopsy results, we found that among all subjects with ALN-negative ultrasound results, 2317 cases without SLN pathological metastasis (74.4%) and 798 cases with SLN pathological metastasis (25.6%). This finding did not provide supporting evidence for our hypothesis, but rather indicated that if the negative SLN biopsy result was replaced by the ALN-negative ultrasound results as the treatment basis, up to 25.6% of patients would be at risk of undertreatment due to underestimated SLN pathological metastasis. Thus, an ALN-negative ultrasound result cannot replace SLN biopsy in accurately evaluating ALN status in breast cancer. Undertreatment may also affect the long-term survival of patients with ALN-negative ultrasound results. To our knowledge, this is the first report to compare data between ALN-negative ultrasound results and negative pathological SLN biopsy results using a large sample in a Chinese population.
In addition to directly comparing the ALN-negative ultrasound results and negative pathological SLN biopsy results, we also selected clinicopathological factors that are widely used to reflect malignancy in breast cancer. These clinicopathological factors included age, menstrual status, primary tumor size, pathological classification, hormone receptor status, and HER-2 status. We attempted to determine whether combining these factors with ultrasound results would help select patients with ALN-negative ultrasound results, who would possibly be exempted from SLN biopsy. Multivariate analyses showed that menstrual status, tumor size, ER status, and pathological types were found to be independently significant predictors. To our knowledge, no publication presented the relationship between menstrual status and SLN metastasis.
In the present study, premenopausal patients were more prone to SLN metastasis than postmenopausal patients, being 28.2% and 22.6%, respectively, (P < .05). Since menstrual status is closely correlated with patients' age, both an agerelated time-dependent exposure as well as hormonal effect could be a reasonable explanation. Kenney et al 13 shows that the ALN metastasis drops in elder patients. Our data show that compared to patients over age 40, patients under age 40 were more likely to have SLN metastasis. Tumor size has long been believed to be an independent predictor of ALN metastasis. [14][15][16][17][18] In the present study, patients with larger tumors (>2 cm) were more likely to have SLN metastasis than those with smaller tumors (≤2 cm). Furthermore, among the patients whose tumor was ≤2 cm (T1), the SLN metastasis rate is 0 (0 of 6) for T1a (≤0.5 cm) patients, 12.6% (28 of 222) for T1b (0.5-1 cm) patients, and 23.17% (314 of 1356) for T1c (1.1-2 cm) patients, which is in consistent with the results of other studies. [14][15][16][17][18] In terms of pathological types, the present study showed that the higher the histological grade was, the greater SLN metastasis patients had. This is also in consistent with the results of other studies. [17][18][19] As for the ER status, our study shows that patients with positive ER were more likely to have SLN metastasis, and patients with positive ER were more likely to be PR positive and HER-2 positive as well. Tan et al 18,20 reported similar results. In the present study, we identified 4 statistically independent SLN status predictors with an AUC of 0.658, indicating that for patients with ultrasonography negative axilla, even combining these clinical and pathological factors cannot adequately predict the SLN status.
Since the prediction accuracy of SLN pathology for ALN status is as high as 97%, SLN biopsy avoids some of the unnecessary ALN dissection with huge trauma, but currently none of the minimally invasive procedures has reach this prediction accuracy alone or in combination. How to further reduce, and even avoid traumatic examination is the eternal pursuit in the strategy of breast cancer diagnosis and treatment. The present study had a goal to replace SLN biopsy by the noninvasive method, ultrasound, to evaluate ALN, but the results showed that the prediction accuracy of SLN pathology by ultrasound negative ALNs was only 74.4%, with an inaccuracy rate of 25.6%, in other words, which is unacceptable in clinical practice. However, we also observed that some clinical factors were predictive of SLN pathology, such as tumor size and ER status, among those patients with ultrasound negative ALNs. In future studies, we will collect additional detailed pathological data, such as sub-classification of tumor size (T1a, T1b, and T1c), to increase and improve the prediction value of ultrasound ALN negative results for SLN pathology. Furthermore, we will consider incorporating some newly discovered predictive-relevant molecules in oncology research.
Some of the characteristic of this present study are unique: (1) the time span for patient collection were relatively short so that the material, equipment, and technology were consistent; (2) the patients were treated by an experienced multidisciplinary breast cancer team, including ultrasound experts and SLN biopsy breast cancer surgeons with homogeneous specialized training; and (3) data were drawn from a large number of homogeneous patients with ALN-negative ultrasound results.
However, the present study also has some limitations. Although the data were collected from the prospective database of a single cancer center, the nature of the present study is retrospective, selection bias is inherent. Further validation by larger, multicenter cohort as well as randomized controlled clinical trials is required.
In conclusion, the present study found a 74.4% consistency between ALN ultrasonography-negative results and negative pathological SLN results. Although menstrual status, tumor size, histologic subtypes of primary tumor, and ER status were found to be statistically independent predictors of SLN pathological metastasis, the nonstatistically significant AUC indicates that the combination of ultrasonography, even F I G U R E 3 The ROC curves of the combination of each independent factors with the clinic-pathologic factors, did not generate an acceptable prediction accuracy.