Nomogram for predicting axillary upstaging in clinical node-negative breast cancer patients receiving neoadjuvant chemotherapy

The prediction of axillary lymph node status after neoadjuvant chemotherapy (NAC) becoming critical because of the advocation of the de-escalation of axillary management. We investigate associated factors of axillary upstaging in clinical node-negative (cN0) breast cancer patients receiving NAC to develop and validate an accurate prediction nomogram. We retrospectively analyzed 1892 breast cancer patients with stage of cT1-3N0 treated by NAC and subsequent surgery between 2010 and 2020 in twenty hospitals across China. Patients randomly divided into a training set and validation set (3:1). Univariate and multivariate logistic regression analysis were performed, after which a nomogram was constructed and validated. In total, pathologic node negativity (ypN0) achieved in 1406 (74.3%) patients and another 486 (25.7%) patients upstaged to pathologic node positive (ypN+). Breast pathologic complete response (bpCR) was achieved in 445 (23.5%) patients and non-bpCR in 1447 (76.5%) patients. A nomogram was established by ER, tumor histology, HER2 status, cycle of NAC treatment, and the bpCR, which were confirmed by multivariate logistic analysis as independent predictors of nodal upstaging in the training cohort (n = 1419). The area under the receiver operating characteristic curve (AUC) of the training cohort and validation cohort (n = 473) were 0.73 (95% CI 0.693–0.751) and 0.77 (95% CI 0.723–0.812) respectively. We present a nomogram with a nationwide large sample data which can effectively predict axillary upstaging after neoadjuvant chemotherapy to give better advice for individualized axillary lymph node management of breast cancer.


Introduction
The administration of neoadjuvant chemotherapy (NAC) has evolved from a strategy for patients with inoperable breast cancer to an option for those with operable breast cancer (Montemurro et al. 2020). NAC provide accurate prognostic estimates based on the extent of tumor response, which can also guide adjuvant therapy, and provide frequent tumor down-stage, which can lead to smaller surgeries in patients with larger tumors at diagnosis (Gu et al. 2022). Based on these benefits, more patients with clinical node negative (cN0), not just clinical node positive (cN +) breast cancer receiving NAC (Foldi et al. 2021;Pilewskie and Morrow 2017;Feng et al. 2022). Pathologic complete response (pCR) was more common in axillary lymph nodes than in the primary tumor, thus it has raised questions about the optimal approach to the axilla after the use of NAC (Barron et al. 2018;Tadros et al. 2017; Barrio et al. 2016).
Traditionally, axillary lymph node dissection (ALND) has been the standard management for all breast cancer surgery (Andersson et al. 2018). However, because of the morbidities like lymphedema, restriction of shoulder joint motions, seroma formation in the armpit, numbness, which have deleterious long-term effects on the quality of life, there have been increasing efforts to investigate the feasibility of sentinel lymph node biopsy (SLNB) with fewer complications and did not affect the accuracy of diagnosis or prognosis (Fisher et al. 2020;Krag et al 2007;Giuliano et al. 2016). NCCN (National Comprehensive Cancer Network) guideline (Gradishar et al. 2022) for breast cancer recommended SLNB for cN0 patients before NAC and cN+ patients that turned to cN0 after NAC, however the false negative rate of SLNB is over 10% even though three lymph nodes detected, moreover, for patients who do not have lymph node metastases, SLNB still bring unnecessary complications (Foldi et al. 2021;Pilewskieet al. 2022). Currently, prospective randomized surgical trials investigating the omission of SLNB in patients who are clinically node negative after NAC (ycN0) (Reimer et al. 2020;Hersh and King 2022).
There is a clinical need to identify the subgroup of patients with very low risk of axillary disease in whom SLNB might be omitted (Kong et al. 2022). Thus, recently there were extensive literature exists predicting axillary pathological complete response after NAC in clinically cN+ patients (Weiss et al. 2021;Montagna et al. 2020;Chen et al. 2021). However, for patients with cN0 disease prior to NAC, there is still some patients upstage to positive nodal disease after NAC (ypN+) and there is no good data to advise them of residual potential disease risk at the time of surgery. The purpose of our study was to identifying risk factors that associated with nodal upstaging after NAC and present a novel nomogram which can effectively predict nodal upstaging in lymph nodes.

Study population
We used data from CSBrS-012 database of Chinese Society of Breast Surgery which covers twenty hospitals across China. It included breast cancer patients who completed standard breast and axillary surgery after NAC between January 2010 and December 2020. This study was approved by the Ethics Committee of First Hospital of Jilin University. The need for informed consent was waived. The inclusion criteria of our study were (1) cT1-3N0 breast cancer patients (2) NAC prior to surgery (3) complete clinical and pathological information, and (4) successful mastectomy or breast-conserving surgery and ALND or SLNB after NAC. Patients who changed NAC regimen midway, patients received neoadjuvant endocrine therapy were excluded.

Clinical/pathological parameters
Clinical/pathological parameters analyzed including age, tumor diameter before NAC measured by ultrasound, histology, ER positivity (defined as ≥ 1% positive cells by immunohistochemistry(IHC)), PR positivity (defined as ≥ 1% positive cells), HR positivity (ER and/or PR positivity), HER2 status (3+ by IHC or 2+ by IHC and positive by in-situ hybridization defined as positive; 0 or 1+ by IHC or 2+ and negative by in-situ hybridization defined as negative), Ki-67 index (detected by IHC), biologic subtypes (performed by referring to the 2011 St.Gallen Consensus (Goldhirsch et al. 2011): HR+/HER2−, HR+/ HER2+, HR−/HER2+, TNBC). cN0 was defined as no suspicious lymph nodes on axillary ultrasound or suspicious lymph nodes on axillary ultrasound but negative on either fine needle aspiration cytology or core needle biopsy or SLNB. Suspicious lymph nodes were considered in case of hypoechoic round shape, focally thickened cortex, or absent fatty hilum. The clinical tumor response was measured according to Response Evaluation Criteria in Solid Tumors (RECIST) (Eisenhauer et al. 2009). pCR in the breast was defined as the absence of any residual both invasive and in situ cancer in breast after NAC. ypN0 was defined as the absence of any tumor cells in the axillary lymph node after NAC. NAC was provided in accordance with National Comprehensive Cancer Network breast cancer guidelines. In our study, we divided NAC regimens into 3 category: (1) Anthracyclines combined with Taxanes regimen; (2) Taxanes combined with Platinums regimen; (3) Other regimens.

Statistical analysis
All the included patients were assigned into a training cohort and a validation cohort randomly according to a ratio of 3:1. Pearson's χ 2 test was applied to compare baseline differences in clinical/pathological parameters between training and validation cohort. We performed the univariate logistic regression analysis to test the characteristics that related to axillary upstaging in patients with cN0 prior to NAC, and used multivariable logistic regression analysis to identify the independent predictors. A predictive nomogram of axillary lymph node upstaging after NAC was established based on the independent predictors determined by multivariate analysis. The ROC (receiver operating characteristic) curve and calibration curve were drawn and AUC (area under the ROC curve) was calculated to verify the performance of the nomogram. Statistical calculations were performed in SPSS version 21.0 (Inc., Chicago, IL, USA) and R 4.2.2 (R Project for Statistical Computing) software.

Patient characteristics
Clinicopathological characteristics of patients in the training and validation cohorts are described in Table 1 The pathological results of operation after NAC showed that bpCR was achieved in 445 (23.5%) patients and non-bpCR in 1447 (76.5%) patients, ypN0 in 1406 (74.3%) patients and axillary upstaging (ypN+) in another 486 (25.7%) patients. Among patients who achieved bpCR the rate of ypN0 was 91.9%, while in the non-bpCR subgroup it was 68.9%. HR negative subtype patients achieved more ypN0 rate than HR positive subtypes (87.5% vs. 67.3%, p < 0.01), conversely, HR+/HER2− patients have more risk of axillary upstaging than other subtypes (p < 0.01) (Fig. 1). Among 243 bpCR patients with HR−/HER2+ or TNBC subtypes, there was only 12 patients upstaged to ypN+ after NAC, the rate of nodal negativity was 95.1%.

Prediction of lymph node upstaging
Univariate analysis revealed that ER status, tumor histology, HER2 status, biologic subtype, cycle of NAC treatment, and bpCR or not had statistical significance for ypN+ (p < 0.05), and all those factors except from biologic subtype were confirmed as independent predictors of axillary upstaging in multivariate logistic regression analysis (Table 2).

Construction and validation of the nomogram
Based on independent predictors determined by multivariate logistic regression analysis, a predictive nomogram of axillary lymph node upstaging after NAC was established. We can assign a point value to each variable by draw a vertical line between each variable value and the first row. By summing up the scores of each variable we can calculate the total nomogram score. Then, a probability of axillary upstaging after NAC in cN0 patients can be determined by drawing a vertical line from the total score to the bottom row (Fig. 2).
We used ROC curve and calibration curve to quantify the performance of the nomogram. The ROC curve was constructed using the ER, tumor histology, HER2 status, cycle of NAC treatment, and bpCR. The AUC was 0.73 in the training cohort and 0.77 in the validation cohort (Fig. 3). The calibration curve showed a good and satisfactory agreement between the predicted and actual probabilities both in the training and validation cohorts (Fig. 4).

Discussion
NAC aims to preoperatively downstage breast as well as axillary nodal in breast cancer patients (Samiei et al. 2021). However, with the more recognization to clinical value of the response of the breast tumor or lymph node to NAC, NAC is increasingly used for early-stage breast cancer with cN0 (Feng et al. 2022). Risk of axillary nodal upstaging in primary cN0 breast cancer patients receiving NAC are unknown. In the current study, we focused on the pathological nodal status of cN0 patients who were underwent surgery followed by NAC, and analyzed parameters which could be a prediction factors of axillary upstaging in this kind of patients. Among 1892 patients who were cN0 prior to NAC, there was approximately 74.3% of patients still maintain negative axillary lymph node and another 25.7% patient upstaged to ypN+ . We included 13 clinical/pathological features as potential predictors. ER status, tumor histology, HER2 status, cycle of NAC treatment, and bpCR associated with axillary upstaging according to the multivariate analysis.
Previous studies showed that the rate of axillary response is significantly associated with biological subtypes and bpCR (Tadros et al. 2017). TNBC and HER2 positive breast cancers can achieve axillary pCR rates greater than other types (Asaoka et al. 2020;Samiei et al. 2021;Samiei et al. 2020;Haque et al. 2018). Is that parallel in upstaging? Researchers from Mayo Clinic have investigated nodal upstaging in 228 cN0 patients receiving NAC and neoadjuvant endocrine therapy and found that ER+/HER2− subtype carries higher risk for nodal upstaging rather than other subtypes (Hammond et al. 2022). In our study, although it showed in histogram ( Fig. 1) that HR+/HER2− subtype appears more rate of axillary upstaging than TNBC and HER2 positive subtypes, however, biological tumor subtype is not an independent predictor but ER and HER2 status is. Moreover, it had been confirmed before that the response of patients with HR positive disease to NAC was relatively low (Lopez-Tarruella Among patients who achieved bpCR the rate of axillary upstaging was 8.1%, while in the non-bpCR group, it was 31.1%. In reverse, it consistent with studies that evaluating downstaging. Among 243 bpCR patients with ER−/HER2+ and TNBC subtypes there was only 12 patients upstaged to ypN+ after NAC, the rate of nodal negativity was approximately 95.1%. Unfortunately, surgical axillary management is a routine procedure for all these patients, even it is a SLNB, patients suffer unnecessary complications like Lymphedema, paraes-thesia, arm and shoulder impairment, and pain (Verbelen et al. 2019;Gebruers et al. 2015).
A nomogram based on patients information can identify patients with very low risk of axillary disease in whom SLNB might be omitted (Moorman et al. 2022). Resent years, several nomograms have been developed to predict the axillary pathological complete response of patients who underwent NAC (Gu et al. 2022;Guo et al. 2020;Jin et al. 2016;Kang et al.2022;Hwang et al. 2019), but there was few research which discussed axillary lymph node upstaging during the NAC. To the best of our knowledge, the present nomogram is the first to predict axillary lymph node upstaging in cN0 patients who received NAC that based on real world data from a large number of multicenter patients. Apart from the ER and HER2 status and breast pCR, tumor histology and NAC treatment cycle are involved in our predicted nomogram. As we can see (Fig. 2), IDC got more score than other histological subtypes in the possibilities of axillary upstaging. In a previous nomogram (Corsi et al. 2021) to predict the probability of nodal pCR after NAC, histological tumor type "Others" (OR 2.06, 95% CI 1.22-3.47, p = 0.007) achieve more nodal pCR than IDC, which was consistent with our result. However, Chi-Chang Yu et al. (2022) investigate the influencing factors of axillary lymph node status after NAC in cN0 patients and found that compared with the patients in the ypN0 group, patients in the ypN+ group were more likely to have a non-ductal histologic type (17.6% vs. 3.4%, p = 0.032). Whether IDC is a protective factor or a risk factor for lymph node metastasis has not been concluded. In the present study, we classified NAC regimens (1) Anthracyclines combined with Taxanes regimen; (2) Taxanes combined with Platinums regimen; (3) Other regimens. Different chemotherapy regimen can not predict nodal upstaging but NAC treatment cycle can. As  we can see from the nomogram (Fig. 2), if a patient receive 6 or 4 cycle of NAC, the likelihood of axillary upstaging is higher than those of 8 cycle, and > 8 cycle. It may because that 4 or 6 cycle chemotherapy was not enough for tumor response. More than 8 cycle was the best choice in this article. Our nomogram has additional value for the selection of cN0 patients who are not good candidates for axillary deescalation. But it still has some limitations: (1) There were a few patients who changed regimen during NAC, and we excluded them because their chemotherapy regimen was too individual and specific and they didn't have very detailed information about the replacement plan that there was no way to categorize. (2) Some studies including Chi-Chang Yu et al. (2022) found that lymphovascular invasion was the strongest (OR 29.37, 95% CI 7.15-120.68) independent risk predictor of axillary metastasis in cN0 patients undergoing NAC. Unfortunately, in our study we couldn't analyze this Only variables with p values < 0.05 were included in the multivariate analysis cT clinical tumor stage before NAC, ER estrogen receptor, PR progesterone receptor; HR hormone receptor, HER2 human epidermal growth factor receptor 2, H trastuzumab, P pertuzumab, TNBC triple-negative breast cancer, IDC invasive ductal carcinoma, bpCR breast pathologic complete response, CR complete response, PD progressive disease, PR partial response, SD stable disease, OR odds ratio, 95% CI 95% confidence interval Fig. 2 Nomogram to predict the probability of ypN+ in cN0 patients before NAC. ER estrogen receptor, HER2 human epidermal growth factor receptor 2, H trastuzumab, P pertuzumab, IDC invasive ductal carcinoma, Cycle cycle of neoadjuvant chemotherapy, bpCR breast pathologic complete response Fig. 3 The ROC (receiver operating characteristic curves curve) for prediction model of ypN+ in patients with cN0 prior to NAC. A is the training cohort and B is the validation cohort. AUC area under the curve, 95% CI 95% confidence interval 1 3 factor because it was not included in the initial database.
(3) This study used a retrospective method, which makes it prone to potential bias compared with a prospective study.
(4) No external validation was set up in this study.

Conclusions
The ER status, tumor histology, HER2 status, cycle of NAC treatment, and bpCR were associated with axillary upstaging in initial cN0 breast cancer patients receiving NAC. Furthermore, we constructed a nomogram model that could effectively predict the risk of axillary upstaging.