A nomogram for predicting pathologic node negativity after neoadjuvant chemotherapy in breast cancer patients: a nationwide, multicenter retrospective cohort study (CSBrS-012)

Purpose This study aimed to investigate the factors associated with pathologic node-negativity (ypN0) in patients who received neoadjuvant chemotherapy (NAC) to develop and validate an accurate prediction nomogram. Methods The CSBrS-012 study (2010–2020) included female patients with primary breast cancer treated with NAC followed by breast and axillary surgery in 20 hospitals across China. In the present study, 7,711 eligible patients were included, comprising 6,428 patients in the primary cohort from 15 hospitals and 1,283 patients in the external validation cohort from five hospitals. The hospitals were randomly assigned. The primary cohort was randomized at a 3:1 ratio and divided into a training set and an internal validation set. Univariate and multivariate logistic regression analyses were performed on the training set, after which a nomogram was constructed and validated both internally and externally. Results In total, 3,560 patients (46.2%) achieved ypN0, and 1,558 patients (20.3%) achieved pathologic complete response in the breast (bpCR). A nomogram was constructed based on the clinical nodal stage before NAC (cN), ER, PR, HER2, Ki67, NAC treatment cycle, and bpCR, which were independently associated with ypN0. The area under the receiver operating characteristic curve (AUC) for the training set was 0.80. The internal and external validation demonstrated good discrimination, with AUCs of 0.79 and 0.76, respectively. Conclusion We present a real-world study based on nationwide large-sample data that can be used to effectively screen for ypN0 to provide better advice for the management of residual axillary disease in breast cancer patients undergoing NAC.


Introduction
With the recognition of the importance of biology and systematic therapy in local control, we gradually agree that larger surgery does not cure bad biology in breast cancer (1).The adoption of a true multidisciplinary treatment approach, rather than the sequential use of different therapies, decreases the extent of surgery and its associated morbidity (2,3).
Axillary lymph node dissection (ALND) has traditionally been used as routine axillary surgical management for breast cancer patients (4).Multiple prospective, randomized trials led by the American College of Surgeons Oncology Group (ACOSOG) Z0011 trial (5) demonstrated that sentinel lymph node biopsy (SLNB) can replace ALND in patients with low nodal burden disease because of noninferior local control and survival, but with lower surgical morbidity.Neoadjuvant chemotherapy (NAC) results in frequent downstaging of tumors in both the breast and axilla, which can lead to fewer surgeries in patients with larger tumors at diagnosis.The implementation of NAC has enabled selected women to undergo breast-conserving surgery (BCS) in the last two decades; however, for patients who received NAC, the chance of de-escalated axillary surgery has not improved (6).The National Comprehensive Cancer Network (NCCN) breast cancer guidelines recommend SLNB for patients with cN0 to ycN0 disease after NAC, but ALND is still recommended for patients who are converted from cN+ to cN0, and SLNB is usually considered a relative contraindication due to its low identification rate and high false-negative rate (FNR) (7,8).In the SENTinel NeoAdjuvant (SENTINA) study (9), the detection rate of SLNB after NAC in patients with cN+ to cN0 disease was 80.1% (95% CI 76.6-83.2),and the false-negative rate was 14.2% (95% CI 9.9-19.4).However, approximately 74% of breast cancer patients with cN0 disease are sentinel lymph node-negative, and postoperative complications still occur even after SLNB (10,11).
Patients with a low risk of residual axillary involvement after NAC could benefit from omitting axillary surgical intervention if there are accurate tools for nodal response prediction (12).Currently, the commonly used clinical imaging methods for evaluating the axillary region include ultrasound, mammography, magnetic resonance imaging (MRI), and positron emission tomography CT (PET-CT) (13-15).Nevertheless, the accuracy of these techniques remains low, and there are no unified guidelines for axillary imaging evaluation of NAC response (16).The ACOSOG Z1071 (Alliance) trial reported that axillary ultrasound (AUS) after NAC can identify abnormal nodes, guide patient selection for SLN surgery instead of ALND, and reduce the FNR of SLNB to less than 10%.However, the accuracy of AUS after NAC was low; only 43.2% of patients who were negative for AUS were confirmed to have nodal pCR by ALND (16).Investigators also attempted to decrease the FNR by marking positive lymph nodes at diagnosis before NAC.However, clips can be found during surgery in only half of the patients (17).Imaging-guided localization (IGL) of the clipped node was introduced to increase the likelihood of clip removal.The lowest FNR was achieved when IGL was added to SLN biopsy, a procedure called targeted axillary dissection (TAD) (18,19).However, it did not significantly change the performance of tailored axillary surgery, which left ≥2 positive nodes behind in 47.6% of the patients (20,21).In all these explorations, the prediction model based on clinical and pathological factors still has clinical value and application prospects.The present study aimed to identify factors that are predictive of ypN0 and construct a novel nomogram that can effectively predict nodal negativity and thus potentially avoid axillary surgery, which can reduce women's loss of function and lymphedema.

Study population
The Chinese Society of Breast Surgery (CSBrS-012) is a nationwide, multicenter, 10-year retrospective clinical epidemiological study conducted across 20 hospitals in China.The CSBrS-012 study included female primary breast cancer patients who received NAC and underwent standard breast and axillary surgery after NAC between January 2010 and December 2020.The 20 hospitals are located in central, northern, eastern, northwestern, northeastern, and southwestern China, and represent different levels of breast cancer burden.After excluding patients with incomplete data, 7,711 patients were enrolled in the study.Hospitals were randomly assigned to two groups comprising 6,428 patients in the primary cohort from 15 hospitals and 1,283 patients in the external validation cohort from five hospitals.We then randomized the patients in the primary cohort at a 3:1 ratio into the training and internal validation sets (Figure 1).The study was performed in accordance with the Declaration of Helsinki and approved by the Ethical Review Committee of the First Hospital of Jilin University (No. 2021-066).As this was a retrospective study and all data analyses were performed anonymously, the requirement for informed consent from the patients was waived.

Patient characteristics
Variables included age, clinical tumor (cT) and clinical nodal (cN) stages before NAC, tumor histology, ER, PR, HR, HER2, Ki-67, biological subtypes, NAC regimen, NAC treatment cycle, and pCR status.Immunohistochemistry (IHC) was used to detect the expression of ER, PR, HER2, and Ki-67.ER and PR were defined as positive if ≥1% of cells were positive.HR was defined as positive if the ER and/or PR were positive.HER2 expression was defined as

Statistical analysis
Statistical analyses were performed using SPSS version 21.0 (Inc., Chicago, IL, USA) and R 4.2.2 (R Project for Statistical Computing) software.The differences in clinicopathological parameters between the training and internal validation sets were evaluated using Pearson's c 2 test.Univariate logistic regression and backward stepwise selection were used for the final multivariate model.A predictive nomogram for ypN0 was established based on independent risk factors identified via multivariate analysis.The predictive value of the model was appraised using receiver operating characteristic (ROC) curves and calibration curves.The AUC (area under the receiver operating characteristic curve) was calculated.

Patient characteristics and NAC response
A total of 7,711 female breast cancer patients, with a median age of 49 years, were enrolled.The baseline characteristics of the patients are summarized in Table 1.The proportion of patients with initial stage cT1-2 tumors was 79.8%, and that with initial stage cT3-4 tumors was 20.2%.The proportion of patients with cN0-1 stage disease in the study population was greater than that of patients with cN2-3 stage disease (81.5% vs. 18.5%).Most patients had invasive ductal cancer (6,971 [90.4%]).Anthracyclines and taxanes were the most common NAC treatments (76.3%).Approximately half of the HER2+patients received targeted The pathological responses of the breast and axillary lymph nodes according to the biological subtype are summarized in Table 3. Responses to NAC in the different subgroups were generally consistent between the breast and axillary regions.In both the breast and axilla, HR-negative patients showed a better response to NAC than HR-positive patients (p <0.001).In both the breast and axilla, HR+/HER2− subtypes exhibited relatively poor responses to NAC compared to the other subtypes (p <0.001).The ypN0 rate for all subtypes was significantly higher than the bpCR rate (p <0.05).

Associations between ypN0 and clinicopathologic parameters
According to the univariate logistic regression analyses of the training set, cN stage, ER, PR, HER2, Ki67, NAC treatment cycle, and bpCR were associated with ypN0.All of the above parameters were subjected to multivariate logistic regression using backward selection analysis, and a lower cN stage, ER-negative status, PR- negative status, HER2-positive status with targeted therapy, Ki67 level ≥20, more NAC treatment cycles, and bpCR were confirmed to be independent predictors of ypN0 (Table 4).

Nomogram for predicting ypN0
A nomogram to predict ypN0 was developed based on multivariate logistic regression results.Points were assigned to each variable and summed to obtain the total number of points.Finally, the probability of ypN0 was determined by drawing a vertical line from the total score to the bottom row (Figure 2A).For example, a patient with HER2-amplified breast cancer with cN1 and Ki67 >20 who received eight cycles of NAC with single targeted therapy and achieved bpCR had a total of 188 points, so the possibility of ypN0 after NAC for this patient was 88% (Figure 2B), and a patient with triple-negative breast cancer with cN1 and Ki67 >20 who received eight cycles of NAC and did not achieve bpCR had a total of 88 points, so the possibility of ypN0 after NAC for this patient was 40% (Figure 2C).
The discriminatory ability of the nomogram to predict ypN0 status was investigated using receiver operating characteristic (ROC) curve analysis.The AUCs of the training, internal validation, and external validation sets were 0.80, 0.79, and 0.76, respectively, indicating that the nomogram had potentially promising predictive power.The calibration plots presented excellent agreement between the training and validation sets and showed no significant difference between the predicted and actual probabilities of ypN0 (P = 1.000) (Figures 3-5).

Discussion
Among patients with cN0 breast cancer, approximately 74% do not have axillary lymph node metastasis (24).This means that even SLNB represents overtreatment and causes unnecessary complications, with few advantages for many patients.However, the St. Gallen Consensus Panel in 2017 (25) and the German AGO recommendation in 2022 (26) recommend SLNB as the standard surgical procedure for patients who present with cN0 before and after NAC.In patients who are cN+ and achieved nodal pCR after NAC, ALND is still performed in clinical practice in some cases because of the unacceptable identification rate and FNR of SLNB (7)(8)(9)27).
Recently, the 5-year survival results of the SOUND trial were published (28).This was a prospective non-inferiority phase III randomized clinical trial that enrolled 1,463 patients with small breast tumors (<2 cm) and a cN0 stage.Patients were randomized in a 1:1 ratio to either the SLNB group or the no axillary surgery group.Interestingly, omission of axillary surgery was not inferior to SLNB in terms of the 5-year DFS and OS.This was a study of patients who underwent upfront surgery.For patients who receive NAC, multiple prospective trials investigating whether axillary surgery can be safely abandoned in selected patients are underway.The European Breast Cancer Research Association of Surgical Trialists (EUBREAST)-01 is a prospective clinical trial in which axillary surgery will be eliminated completely (no SLNB) for initially cN0 patients with radiological complete remission and breast pCR in the lumpectomy specimen (29).The ASICS trial is a non-inferiority, single-arm trial open to both breast-conserving and mastectomy patients in which no SLNB is performed in cN0, triple-negative, or HER2-positive breast cancer patients with a radiological complete response on MRI (30).The results of these trials are expected in to continue for the next 5 years.Meanwhile, the prediction model for axillary nodal burden based on clinical and pathological factors has clinical value and application prospects.In the present study, we presented and validated a model based on nationwide multicenter data of breast cancer patients to predict the possibility of ypN0 disease after NAC.Moreover, to prove its universality, we externally validated the nomogram using patient information from different hospitals.
Researchers at the MD Anderson Cancer Center first proposed that breast pCR is strongly correlated with nodal status after NCT (31).In the present study, 46.2% of patients achieved ypN0 and 20.3% of patients achieved bpCR, and the rate of ypN0 was greater on patients who achieved bpCR than in the nonbpCR group (75.3% vs. 38.7%).Tumor response to NAC was significantly related to tumor subtype.Barron et al. (32) reported 30,821 patients with cT1/ cT2 cN0/cN1 breast cancer treated with NAC from the American National Cancer Database and reported breast pCR rates of 37.2%, 58.2%, 37.2%, and 13.1%, respectively.The ypN0 rates were 78.6%,   84.5%, 75.3%, and 47.0% for TNBC, HR−/HER2+, HR+/HER2+, and HR+/HER2− subtypes, respectively.In our study, the distribution of tumor subtypes was consistent with that in the above study; however, the rates of bpCR and ypN0 were low because we included cT3-4 and cN2-3 patients.In addition, the pCR and ypN0 rates of HER2+ patients were not significantly high in the current study, possibly because only approximately half of the patients with HER2+ status (50.5%, 1,335/2,642) received  molecular-targeted therapy because the targeted drugs were not covered by medical insurance in the early years.

A B C
As expected, clinical nodal stage and breast tumor response strongly predicted ypN0.To avoid the influence of receptors on molecular subtypes, we did not include subtypes in the analysis.We found that patients with ER-negative, PR-negative, and HER2-positive disease had a higher ypN0.Ki67 is a proliferation marker, and patients with higher Ki67 levels showed greater sensitivity to chemotherapy in previous studies (33, 34), which was consistent with our study.In the present study, multivariate analysis revealed that more treatment cycles were associated with ypN0, independent of the tumor histology and treatment regimen.Clinical tumor size has been shown to be a predictor of lymph node status in operable breast cancer patients in several previous studies (35)(36)(37); however, in the context of NAC, the relationship between cT and ypN0 was not significant in our study.
We developed a nomogram based on multivariate logistic regression results.In contrast to previous nomograms that predicted axillary pCR in initially cN+ patients (38)(39)(40) or in specific subtypes (41,42), the current nomogram predicted ypN0 in all patients with stage cT1-4N0-3 disease.The AUC of the nomogram in the ROC curve analysis was 0.80, 0.79, and 0.76 in the training, internal, and external validation cohorts, respectively, and showed good discrimination in the prediction of ypN0.The advantage of our prediction nomogram is that most breast cancer patients who receive NAC can be assessed, and the indicators for building the nomogram can be easily acquired by surgeons.Moreover, as mentioned above, our findings are consistent with those of previous studies in a global context in terms of pCR for different subtypes, which indicates that our nomogram can also be applied to patients in different countries.
There are several limitations in our study.First, histological grade was found to be an independent prognostic factor for pCR in patients with breast cancer in previous studies (43,44).In our study, we could not analyze this factor because it was not included in the initial database.Second, if we put this nomogram into practice for the omission of any axillary surgery, it should be determined before surgery, but bpCR is available after surgery.However, multiple studies have explored methods to detect bpCR without surgery (45-48).A prospective trial showed that image-guided vacuum-assisted core biopsy (VACB) of the primary breast tumor bed following NAC can identify patients who are very likely to have a bpCR with an FNR of <5% (49).Another potential limitation of this study was its retrospective nature.Our study, which reflects the current clinical practices across the country, will facilitate the design of prospective clinical trials in the future.
Future Directions: The developed nomogram may help clinicians weigh the lymph node tumor burden after NAC more appropriately.However, if our research conclusions are extended to

Conclusions
We present a real-world study based on nationwide large sample data and construct a nomogram model that can effectively screen ypN0 to provide better advice for the management of residual axillary disease in breast cancer patients receiving NAC.
(A) A nomogram to predict the probability of ypN0 in breast cancer patients receiving neoadjuvant chemotherapy.PNoT, HER2 positive without targeted therapy; PST, HER2 positive with single agent HER2 blockade; PDT, HER2 positive with dual HER2 blockade.(B) The blue triangle demonstrates usage of the model: a patient with HER2-amplified breast cancer with cN1 and Ki67 >20 who received eight cycles of NAC with single-targeted therapy and achieved bpCR had a total of 188 points, and the possibility of ypN0 after NAC for this patient was 88%; (C) a patient with triple-negative breast cancer with cN1 and Ki67 >20 who received eight cycles of NAC and did not achieve bpCR had a total of 88 points.The possibility of ypN0 after NAC was 40% for this patient.

3 FIGURE 4
FIGURE 3 ROC curve (A) and calibration curve (B) are shown for the prediction model of ypN0 in the training cohort.The ROC curve for the training set indicated an AUC of 0.80.ROC, receiver operating characteristic curve; AUC, area under the curve.
cases of a hypoechoic round shape, focally thickened cortex, or absent fatty hilum.pCR was defined as the absence of residual invasive or in situ carcinoma in the breast or axillary lymph nodes (ypT0/ypN0).NAC and surgery were performed in accordance with the Chinese Society of Clinical Oncology (CSCO) Breast Cancer Guidelines and the National Comprehensive Cancer Network (NCCN) Breast Cancer Guidelines.In our study, we divided NAC regimens into three categories: (1) anthracycline combined with taxane, (2) taxane combined with platinum, and (3) other regimens.

TABLE 1
Baseline patient characteristics.

TABLE 2
Pathologic response of breast and axillary to NAC in whole study population.

TABLE 3
Pathological response of breast and axillary lymph node according to biologic subtype.

TABLE 4
Univariate and multivariate logistic analysis of factors predict the lymph node positivity after NAC in the training set.