Combined conventional ultrasonography with clinicopathological features to predict axillary status after neoadjuvant therapy for breast cancer: A case–control study

Objectives:

This study aimed to evaluate the value of a model combining conventional ultrasonography and clinicopathologic features for predicting axillary status after neoadjuvant therapy in breast cancer.

Methods:

This retrospective study included 329 patients with lymph node-positive who underwent neoadjuvant systemic treatment (NST) from June 2019 to March 2022. Ultrasound and clinicopathological characteristics of breast lesions and axillary lymph nodes were analyzed before and after NST. The diagnostic efficacy of ultrasound, clinicopathological characteristics, and combined model were evaluated using multivariate logistic regression and receiver operator characteristic curve (ROC) analyses.

Results:

The area under ROC (AUC) for the ability of the combined model to predict the axillary pathological complete response (pCR) after NST was 0.882, that diagnostic effectiveness was significantly better than that of the clinicopathological model (AUC of 0.807) and the ultrasound feature model (AUC of 0.795). In addition, eight features were screened as independent predictors of axillary pCR, including clinical N stage, ERBB2 status, Ki-67, and after NST the maximum diameter reduction rate and margins of breast lesions, the short diameter, cortical thickness, and fatty hilum of lymph nodes.

Conclusions:

The combined model constructed from ultrasound and clinicopathological features for predicting axillary pCR has favorable diagnostic results, which allowed more accurate identification of BC patients who had received axillary pCR after NST.

Advances in knowledge:

A combined model incorporated ultrasound and clinicopathological characteristics of breast lesions and axillary lymph nodes demonstrated favorable performance in evaluating axillary pCR preoperatively and non-invasively.

Introduction

Breast cancer (BC) is the most common malignancy worldwide and the leading cause of cancer-related deaths in females.¹ According to the SEER 5-year relative survival rate (2012–2018) data, the five-year survival rate of BC patients with regional lymph node metastases decreased from 99.1 to 86.1% compared to those without metastases.² Neoadjuvant systemic treatment (NST) is the standard of care for BC with lymph node metastasis, which could decrease the clinical stage of the tumor and increase the chances of breast and axillary preservation. Attaining lesion pathological complete response (pCR) after NST is strongly associated with a benefit in disease-free survival and overall survival for BC patients.^3,4 According to reports, approximately 40% of BC patients achieved complete eradication of axillary lymph node (ALN) lesions after NST, and this can be up to 60% in hormone receptor-negative/ Erb-b2 receptor tyrosine kinase 2 (ERBB2; formerly HER2 [human epidermal growth factor receptor 2]) positive patients.⁵ Presently, axillary lymph node dissection (ALND) has still been used as the primary surgical procedure for patients with initially clinical lymph node positive.^4,6 However, for patients who achieve pCR in the axilla after NST, ALND does not increase patient benefit and create risks associated with lymphedema, upper extremity numbness, and nerve injury. For these patients, it is especially vital to avoid unnecessary risks while maintaining safety.

Recently, several less invasive procedures, including sentinel lymph node biopsy (SLNB),⁷ marking ALN with radioactive iodine seed,⁸ and targeted axillary dissection,⁹ have been proposed to substitute ALND. A meta-analysis study that included 20 studies involving 2,217 patients demonstrated that combined resection of pre-NST marker-positive lymph nodes by SLNB is the most precise way to clarify axillary staging after NST. However, SLNB is not accurate enough as an independent procedure for post-NST axillary staging in patients with an initial clinical stage of N+, with an overall false-negative rate (FNR) of 17%.¹⁰ For now, there is no consensus on axillary management after NST in patients with clinical lymph node-positive BC, which may be associated with the difficulty in accurately assessing ALN status after NST preoperatively. Therefore, how to accurately evaluate the status of ALN after NST preoperatively to accurately identify patients with axillary pCR and thus guide clinical decision-making has become an urgent clinical problem to be solved.

Ultrasonography (US) is the preferred modality to evaluate ALN before and after NST, which can routinely display level I and II axillary lymph nodes.^11–13 Nevertheless, using conventional US alone is insufficient to accurately assess axillary lymph nodes' status after NST.^14,15 A prospective multicenter study investigated the performance of US in evaluating axillary LN in patients with biopsy-proven positive lymph nodes in the axilla after neoadjuvant chemotherapy (NAC). The SN FNAC trial reported a sensitivity of 52.8%, specificity of 78.3%, and accuracy of 61.5% for diagnosing residual lymph nodes in the axillary US after NAC.¹⁵

Previous studies indicated that axillary pCR was significantly associated with clinical stage, clinicopathological features such as ERBB2 and breast pCR.^5,16 Thus, we hypothesized that the inclusion of clinical and pathological factors in addition to US characteristics of breast lesions and axillary LN may enhance the predictive performance of axillary pCR. The purpose of this research was to investigate the diagnostic efficacy of pre-NST and post-NST breast lesions and axillary ultrasound features combined with clinicopathological features for predicting axillary pCR after NST in BC patients with clinical lymph node positive.

Material and methods

Patients

This case–control study was conducted, from June 2019 to March 2022, on 709 female BC patients with clinical T_1-4N_1-3M₀ stage who underwent NST. The inclusion criteria were as follows: (1) biopsy-proven ipsilateral metastatic ALNs with locally advanced BC; (2) patients who met the neoadjuvant therapy population criteria according to the National Comprehensive Cancer Network (NCCN) guidelines; (3) complete clinicopathological and imaging data were available before and after neoadjuvant therapy; and (4) without a history of axillary disease treatment or radiation. The exclusion criteria were as follows: (1) multiple or bilateral breast lesions (n = 128); (2) no surgery after NST at our institution (n = 89); (3) disease progression during treatment (n = 96); and (4) inability to tolerate neoadjuvant therapy (n = 67). In the end, 329 patients were included in this retrospective study (Figure 1). Ethical approval was obtained from the Institutional Review Board (IRB-2023–125) with a waiver for informed consent.

Six or eight cycles of NST (taxane-based, anthracycline-based, or taxane- and anthracycline-based NAC for six or eight cycles) according to NCCN and China Anti-Cancer Association guidelines, either alone or combined with neoadjuvant anti-HER2-targeted therapy and/or endocrine therapy according to molecular subtypes.

Ultrasound examinations and character analysis

US examinations of breast and ALN were performed by five experienced radiologists (>5 years of experience) at pre-NST and post-NST that were performed using the Philips iU22 (L12-5 linear array probe at 7.5–14 MHz) or GE Logiq E9 (L11 linear array probe at 5–13 MHz).

Routine ultrasound features of breast lesions include maximum diameter (mm), margins (Circumscribed, Indistinct, Spiculated, Angular), elasticity score, color Doppler flow imaging (CDFI) score, calcification (none, macrocalcifications, or microcalcifications), and maximum diameter reduction rate. (1) Tumor size: according to the eighth edition of the AJCC TNM staging criteria, it was divided into T1-T4 groups¹⁷; (2) Adler’s semi-quantitative blood flow classification: 0 ~ 3 levels according to the blood flow of the mass¹⁸; (3) Elastography score: according to the different colors of the elasticity imaging map, it is divided into 0 ~ 5 levels¹⁹; and (4) The maximum diameter reduction rate is defined as (maximum diameter pre-NST - maximum diameter post-NST)/ maximum diameter pre-NST*100%.

ALN ultrasound features include long diameter (mm), short diameter (mm), the ratio of long/short diameter (≥2 or <2), boundary (clear or blurred), shape (oval, round, or irregular), maximum cortical thickness (≤3 mm or >3 mm), and presence of lymphatic gates.²⁰ In particular, the maximum cortical thickness should be measured in the largest lymph node cross-section, and if it is unevenly thickened, the thickness at the widest point should be taken.

Histopathologic assessment and character analysis

Pathological diagnosis of breast lesions and axillary lymph nodes in all patients, confirmed by coarse or fine needle aspiration biopsy, before NST. The expression of estrogen receptor (ER), progesterone receptor (PR), ERBB2, and Ki-67 index was detected by immunohistochemistry (IHC).^21–23 The molecular subtype was classified into luminal A, luminal B, HER2-enriched, and triple-negative according to the 2017 St. Gallen guidelines.²⁴ The pathological type was classified into invasive ductal carcinoma (IDC), invasive lobular carcinoma (ILC), ductal carcinoma in situ (DCIS), and other types.

After NST, all patients underwent mastectomy or breast-conserving surgery accompanied by ALND. Breast tissue and LNs were stained with hematoxylin and eosin to observe malignant cells and to confirm the presence or absence of metastasis. ALN pCR was defined as free of invasive cancer burden in axillary lymph nodes or only an isolated population of tumor cells (ypN0). The pathologic responses of breast lesions to NST were confirmed using the Miller-Payne grading criteria with surgical specimens and reached Grade 5 was defined as breast pCR.²⁵

Statistical analysis

All statistical analyses were performed using SPSS (version 25.0) and R (version 4.2.1). Continuous variables were expressed as the mean  ±  standard deviation (SD) or medians with interquartile range (IQR), t-test, or Mann-Whitney U-test were used for comparison between two groups, as appropriate. Categorical information was expressed as examples (%), and the χ2 test or Fisher’s exact test was used for comparison between the two groups. Statistical significance was defined as two-tailed p < 0.05.

At first, all clinicopathological and conventional US features related to axillary pCR were assessed using univariate analysis. Then, the variables with p < 0.20 in the univariate analysis were entered into multivariate logistic regression analysis to determine the independent association between the above features and axillary pCR. Subsequently, by integrating the conventional US and clinicopathological characteristics using multiple logistic regression analysis led to the development of a combined model. The diagnostic efficacy of the conventional US model, clinicopathology model, and the combined model for axillary status was compared by the area under the curve (AUC) of the receiver-operating characteristic (ROC) curve. An AUC value of >0.8 is considered acceptable.²⁶

Results

Patient characteristics

A total of 329 patients with confirmed invasive BC with axillary lymph node metastasis (mean age ±standard deviation, 51.1 years ± 10.1; range, 24–75 years) were enrolled. 47 patients underwent breast-conserving surgery, 282 patients underwent mastectomy (14.29 and 85.71% respectively), and all patients underwent ALND. By postoperative pathological analysis, the breast pCR rate was 30.67% (101/329), the axillary pCR rate was 37.69% (124/329), and the total pCR rate was 25.23% (83/329). The patient baseline characteristics according to ALN status are listed in Table 1.

Univariate analyses

For the univariate analysis of clinicopathologic features (Table 1) and compared to patients with residual axillary LN metastases, patients with axillary pCR exhibited a lower clinical N stage (p = 0.025), a higher proportion of ER negative (p < 0.001), PR negative (p < 0.001), ERBB2 positive (p < 0.001), HER2 hyper-expressive subtypes (33.9%, p < 0.001), pCR of breast lesions (66.9%, p < 0.001), and higher median Ki-67 (40.0% compared to 30.0% in patients with residual lymph node disease, p < 0.001). There were no statistically significant differences between the two groups in terms of age, menopausal status, and histological type.

For the univariate analysis of breast US features (Table 2), post-NST maximum tumor diameter (p < 0.001), maximum diameter reduction rate (p < 0.001), elasticity score (p = 0.001), and margin (p = 0.002) significantly correlated with axillary pCR. There was no significant difference between the two groups in terms of CDFI score and calcification

For the univariate analysis of ALN US features (Table 3), post-NST short diameter (p = 0.009), long-to-short diameter ratio ≥2 (p = 0.016), cortical thickness ≤3 mm (p < 0.001), fatty hilum visible (p < 0.001), and oval shape (p = 0.048) were significantly associated with axillary pCR.

Representative pre-NST and post-NST axillary US images of patients in both axillary pCR and non-pCR groups are shown in Figure 2 and Figure 3, respectively.

Multivariate analyses

Variables with p values less than 0.20 for univariate analysis were selectively included in the multifactor logistic regression analysis. The results found (Figure 4) that ERBB2 positive had the strongest independent association with residual axillary LN metastasis (odds ratio [OR], 9.55; 95% Confidence Interval [CI], 5.10 ~ 17.89). Before NST, Ki-67 index (OR, 1.03; 95% CI, 1.01 ~ 1.04) and after NST, maximum diameter reduction rate (OR, 1.02; 95% CI, 1.01 ~ 1.04), fatty hilum visible (OR, 2.59; 95% CI, 1.13 ~ 5.97), and cortical thickness of less than 3 mm (OR, 2.26; 95% CI, 1.12 ~ 4.57) were protective factors of axillary pCR. Before NST, clinical N stage (OR, 0.57; 95% CI, 0.38 ~ 0.84) and after NST, angular margin of breast lesions (OR, 0.56; 95% CI, 0.38 ~ 0.84) and lymph node diameter (OR, 0.86; 95% CI, 0.75 ~ 0.98) were risk factors of axillary pCR.

Diagnostic performance

Clinicopathological features independently associated with axillary pCR, including clinical N stage, ERBB2 status, and Ki-67 index, were integrated to construct a clinicopathological model, which had an AUC of 0.807 with a sensitivity, a specificity, and an accuracy of 80.65%, 71.71%, and 75.08%, respectively. Ultrasound features, including maximum diameter reduction rate, and after NST margin of breast lesions, short diameter of lymph nodes, cortical thickness, and fatty hilum visible, were integrated to construct an ultrasound model, which had an AUC of 0.795 with a sensitivity, a specificity, and an accuracy of 68.55%, 78.54%, and 74.77%, respectively. A combined model was constructed by integrating ultrasound and clinicopathological variables independently associated with axillary pCR. Compared with the clinicopathological and ultrasound models, the combined model achieved better diagnostic performance. The combined model offered an AUC of 0.882 with improved sensitivity, specificity, and accuracy of 83.87%, 78.54%, and 80.55%, respectively. It had significantly higher AUC than clinicopathological and ultrasound models (p < 0.001, DeLong’s test) (Figure 5). The detailed statistical results of all models are presented in Table 4.

Discussion

This research aimed to develop a non-invasive and convenient model to identify which patients with clinical lymph node-positive BC will reach axillary pCR. Based on the recommendations of the American College of Radiology appropriateness criteria¹¹ and previous studies, we chose ultrasound for the evaluation of breast lesions and axillary lymph nodes and included clinicopathologic features of the lesions as model variables selectively. We demonstrated that a predictive model using ultrasound features of breast lesions and axillary lymph nodes combined with clinicopathologic features could distinguish axillary pCR from residual ALN metastases. The combined model had an AUC of 0.882, and its diagnostic performance was higher than that of the ultrasound-only model (AUC of 0.795) and the clinicopathological model (AUC of 0.807). In addition, all eight independent predictors of axillary pCR are included in the routine pre- and post-NST examinations, so the results are closer to the real clinical experience for clinical reference.

In terms of clinicopathological factors, independent predictors of axillary pCR included clinical N stage, ERBB2 status, Ki-67 index, and breast pCR. This study aimed to accurately predict axillary pCR using the clinicopathological factors available before surgery, but breast pCR or non-pCR requires postoperative pathological clarification, so it was excluded from the multiple logistic regression analysis. Several studies have pointed out a strong correlation between breast pCR and axillary pCR,^27,28 so we chose imaging indices of the breast as a proxy for breast pCR included in this study. This was also verified in our study, reaching an axillary pCR rate of 66.9% in patients with breast pCR (p < 0.001). In agreement with previous studies,²⁹ our results showed that ERBB2 positive was an independent predictor of axillary pCR. Furthermore, according to the American Society of Clinical Oncology/College of American Pathologists guidelines, HER2 hypo-positive status is defined as IHC1+or IHC2+/in situ hybridization negative.²² Various studies have pointed out a correlation between HER2 expression profile and breast pCR. A study that included 1111 BC patients (41% HER2-low-positive) receiving NAC showed that HER2 status did not affect pCR in hormone receptor (HR)-negative patients, whereas low HER2 was associated with lower pCR rates in HR-positive patients.³⁰ Another large cohort study of combined clinical trials showed that in hormone receptor-positive tumors, pCR rates were significantly lower in HER2-low-positive tumors than in tumors with HER2-zero expression.³¹ To further investigate the relationship between HER2-zero, HER2-low, and HER2-positive, with axillary pCR, we grouped them according to HER2 expression, and the results showed that there was no statistical difference between HER2-zero and HER2-low groups, but the difference between the two groups and HER2-positive was significant. However, the correlation between high and low HER2 expression and axillary pCR needs to be investigated in more studies, because of our single-center study and limited sample size. Our results showed that patients with axillary pCR had higher Ki-67 (median 40.0% vs 30.0%, p < 0.001), which is consistent with previous studies.³² Some authors have explored the division into high and low-expression groups according to the different value-added of the Ki-67 index, but at present, there is no uniform standard for Ki-67 group cut-off values.³³ We elected to include specific values in the study, and the results showed that Ki-67 index expression was independently associated with axillary pCR with an optimal cutoff value of 37.5%. According to our study, the AUC for evaluation using only clinical case characteristics was 0.81, suggesting moderate diagnostic efficacy.

In terms of ultrasound features, our results showed that the short diameter of ALN, fatty hilum visible, cortical thickness<3 mm, and the maximum diameter reduction rate and margins of the breast lesion were identified as independent factors associated with axillary pCR after NST. Kim found that the presence of a fatty hilum was the strongest predictor of axillary pCR.³⁴ However, the diagnostic performance of ALN assessment using only conventional ultrasound features is poor. A study that included 13 studies involving 2380 patients showed that the sensitivity, specificity, PPV, and NPV of ALN assessment using axillary ultrasound were 65%, 69%, 77%, and 50%, respectively,³⁵ which is consistent with our findings. To enhance the diagnostic efficacy, it has been proposed to incorporate elastography and contrast-enhanced ultrasound indices into the model. Huang reported that breast shear wave elastography combined with conventional US had a good diagnostic performance for axillary lymph node disease after NST with a combined AUC of 0.90.³⁶ But the study was a single-center observational study with a limited sample size, so the conclusions need to be supported by evidence from more studies with larger samples. Han used lymph node medullary border, lymph node aspect ratio, and contrast-enhanced ultrasonography modality to construct a regression model to assess axillary lymph node status with an AUC of 0.882, however, the specificity was 55.6% and the prediction model had a high level of false positives.³⁷ Others have used magnetic resonance imaging (MRI) for predicting axillary response after NST in BC patients with clinical lymph node-positive.^38–41 Eun used tumor size and lymphatic gate thickness measured by MRI as predictors.³⁹ Huang identified that MRI suggestive of no peri nodular infiltration before and after NAC could be a predictor of axillary pCR.⁴⁰ The AUC of the model in the previous study was 0.662–0835, and its clinical application value needs to be proved by more studies. In conclusion, since the interpretation of qualitative ultrasound features is subjective to the physician, the use of imaging histology to extract quantitative features to predict axillary response after NST has been proposed. Gu proposed a deep-learning imaging histology column line graph model based on NAC and preoperative ultrasound images for independent prediction of axillary lymph node status after NAC.⁴² The AUCs of 0.853 and 0.863 in the validation and test cohorts had the potential to accurately predict LNM status after NAC, but their findings still need to be validated in additional large-sample multicenter studies.

There are some limitations of our research. First, no clips were placed or positioned within the biopsied lymph nodes, making it challenging to determine which lymph nodes were biopsied before NST treatment, which may have contributed to the high FNR in our study. Second, the interpretation results of qualitative ultrasound features such as the margins of breast lesions and the presence or absence of lymphatic gates are largely influenced by the individual experience of the imaging physician. Finally, we are a single-center retrospective study and there may be selection bias. Thus, prospective multicenter studies are required to evaluate the general applicability and clinical translation potential of our proposed model before its practical application.

Conclusion

Our results indicated that ultrasound features combined with clinicopathological indices have higher diagnostic performance than conventional US or pathological indices alone for the evaluation of axillary lymph nodes after NST. The combined model had both excellent sensitivity and specificity, respectively, 83.87 and 78.54%. The AUC of the combined model was 0.882, which indicates that our model can be used to stratify the patients to predict axillary pCR and thus provide decision support for the clinical development of axillary treatment strategies to avoid over- or under-treatment.

REFERENCES