Shear Wave Elastography Combined with Molecular Subtype in Early Predicting Response to Neoadjuvant Chemotherapy for Breast Cancer: A Prospective Case-control Study


 Background: This study was designed to investigate the performance of quantitative shear wave elastography (SWE) and the accuracy of SWE with the molecular subtype for early prediction of pathological response of breast cancer to neoadjuvant chemotherapy (NAC).Methods: In this prospective case-control study, 102 patients were screened from September 2016 to August 2020. Characteristics of conventional ultrasonography (US), SWE and contrast-enhanced magnetic resonance imaging (CE-MRI), were recorded, and the changes were compared to the pre-NAC baseline data. The pathological response was classified according to the Miller Payne grading system. Multivariate logistic regression was used to develop a predictive model for the response to NAC. Results: Significant differences related to changes in SWE characteristics of breast lesions between between pathological response groups were observed earlier than size on the conventional US and MRI images. According to the multivariate predictive model, the best parameter for predicting the pathological response after the first cycle of NAC was the molecular subtype of the tumor [area under the curve (AUC) = 0.83] with low sensitivity (66.04%). Better predictive performance was achieved when △AE/B and the molecular subtype were applied in combination after the second cycle of NAC (AUC = 0.92) with a higher sensitivity (86.79%). The predictive performance of molecular subtype combined with △AE/B and △SWVmean after the third and fourth cycles of NAC were improved (AUC = 0.94).Conclusion: The SWE can be an early predictor of the pathological response to NAC for breast cancer. The combination of SWE and the molecular subtype may be a preferred method for the clinical evaluation of NAC to facilitate the personalization of treatment regimens for breast cancer.


Introduction
Breast cancer is the most common malignant tumor and the most frequent cause of tumor-related mortality in women worldwide. 1 Neoadjuvant chemotherapy (NAC) followed by mastectomy is one of the most important parts of breast cancer therapy. NAC can be used as an in vivo chemosensitivity test, which enables the evaluation of chemotherapy. Moreover, NAC has other advantages such as reducing the tumor burden and downstaging cancer to increase the chances of surgery or breast-conserving surgery. NAC has also been reported to reduce the need for axillary lymph node dissection (ALND) and prevent micro-metastasis during surgery in a previous study. 2 3 However, patients with chemotherapy resistance cannot bene t from NAC and may suffer from adverse reactions due to unnecessary chemotherapy. Furthermore, the timing of the operation may be delayed. Although a large number of clinical trials have shown that NAC is effective for treating breast cancer, studies have also indicated that its e cacy is variable. Previous investigations demonstrated that the overall response rates for NAC range from 69-100%, with pathological complete response (pCR) from 10-15% of patients. 4 Therefore, a practical approach for the early evaluation and prediction of the patient response to NAC is of great importance for the optimization of therapy in clinical practice.
In clinical practice, traditional imaging, including mammography, ultrasound (US), magnetic resonance imaging (MRI), and positron emission tomography (PET), coupled with physical examination is the primary method to evaluate the NAC response. However, the performance of mammography alone for predicting the pathological response to NAC is limited. [5][6][7] MRI, PET, and other emerging techniques are time-consuming, expensive, and contraindicated for some patients. [8][9][10][11][12][13] Therefore, there is no consensus approach for evaluating the response to NAC in patients with breast cancer during chemotherapy.
As a convenient, real-time, low-cost, and non-invasive imaging approach, US is superior for wide utilization for the evaluation of NAC for breast cancer. [5][6][7] Patients with breast cancer were recommended to be re-evaluated for NAC response regularly using US after each two cycles of NAC according to the China Anti-Cancer Association breast cancer guideline. 14 And several studies have shown that US and MRI display similar performance in evaluating the response to NAC. 15 16 However, the accuracy of conventional US is limited for monitoring changes in tumors during NAC. 5-7 13 Ultrasound Elastography (UE) is an emerging imaging method based on biomechanical properties that can be used to diagnose breast lesions and assess breast cancer therapy. 17 Elastosonography, including strain elastography (SE) and shear wave elastography (SWE), was performed in this study according to the guidelines and recommendations for elastography presented by the World Federation of Ultrasound in Medicine and Biology (WFUMB). 18 NAC can cause microscopic pathological changes such as tumor cell apoptosis and brosis, altering the bio-mechanical properties of tumor tissues. 19  SWE is a quantitative and objective modality for assessing changes in the biomechanical characteristics of a tumor. 22 Therefore, it is expected to achieve clinical application for the evaluation of NAC for breast cancer. There has been relatively little research on the differences between SWE and conventional imaging methods or investigations on the performance of SWE with clinicopathological factors for predicting the early response to NAC for breast cancer. The present study was an endeavor to evaluate SWE's performance in predicting the pathological response to NAC in patients with breast cancer compared to conventional US and MRI based on tumor morphological changes. Furthermore, the study was aimed at generating a multi-factor prediction model using imaging and clinicopathological factors in a new predictive modality to obtain additional information for the development of individualized therapy regimens.

Page 4/27
The present study was a prospective case-control study. The study's design and protocol were approved by the ethics committee of the institutional review board at the Cancer Center of Sun Yat-Sen University (approval GZR2017-047). Written informed consent for study participation and data collection were obtained from all patients.

Patients and Study Procedure
Patients were consecutively recruited at the Cancer Center of Sun Yat-Sen University (Guangdong, China) between October 2016 and August 2020. The inclusion criteria were: (i) the patients had invasive breast cancer con rmed with US-guided core needle biopsy, without distant metastasis; (ii) the patients received eight cycles of a standardized NAC regimen; (iii) the patients had at least one targeted measurable lesion according to RECIST 1.1, 23 and (iv) surgery was conducted after completion of NAC. The exclusion criteria were: (i) patients who did not complete full course of NAC as intolerance to chemotherapy, (ii) tumor progression, leading to palliative chemoradiotherapy instead of surgery, (iii) lesion size exceeding the scope of a color-coded map in the assessment view of SWE, and (iv) low-quality data in quantitative SWE with a large area of elasticity defects on the SWE velocity image.
Each included patient was continuously followed up with US and MRI before the surgery. US examinations, including conventional US and SWE, were performed one day before the biopsy (time point t0) and one day before the second (t1), third (t2), fourth (t3), and fth cycles (t4) of NAC (time point t1-t4). MRI examinations were performed before the biopsy and within 3 days before the fth cycle of NAC (time point t4). Surgical excision was performed approximately 2 weeks after eight cycles of NAC. The ow chart depicting the study design is shown in Fig. 1.

NAC regimen
NAC strategies were performed according to the standard protocol at our institution. All patients received eight cycles of chemotherapy based on anthracycline/taxane. Moreover, patients with positive HER-2 expression received trastuzumab, starting from the fth cycle of NAC.
Related examinations B-mode US and SWE images were obtained with the Siemens S2000 ultrasound system (Siemens Medical Solutions, Mountain View, CA, USA) equipped with a 9L4 linear transducer. Conventional US and SWE images of breast lesions were acquired with patients in the supine position according to the breast US examination guideline from the American Institute of Ultrasound in Medicine. 24 Firstly, conventional US scans were performed to locate the breast lesions and obtain greyscale and color Doppler ow images. The longest diameter (D US ) and area (A B ) of the breast lesion on the B-mode image were recorded. Secondly, SWE was performed at the same position, depth, focus position, and gain setting used for conventional US scanning. Furthermore, a probe was held still and applied perpendicular to the skin with the minimum amount of pressure possible. SWE was performed by setting the region of interest (ROI) for stiffness assessment to include the breast tumors and surrounding normal tissue. SWE was performed with the patients holding their breath for approximately 5 seconds. A quality map displayed in green-yellow-red indicating high-intermediate-low quality was obtained rst to evaluate the Shear Wave Velocity (SWV). Next, the image was switched to the SWV map, which was displayed in color mode to indicate the stiffness ranging from soft (blue) to intermediate (green or yellow) and hard (red). The part corresponding to high SW quality (green areas on the quality map) was selected for the measurement of the SWV to ensure SWE reliability. Within the ROI, SWV values ranging from 0.5 to 10 m/s on the velocity map were obtained by placing three SWV-ROIs (2×2 mm) over the lesion's stiffest and softest parts, respectively. Finally, the image was switched to a Shear Wave Time (SWT) map, and the pro le of the breast lesion was delineated to measure the area of the lesion (A E ) according to the color-coded differences between the tumor and the surrounding tissues.
The US examinations were conducted by two board-certi ed radiologists (Jia-Xin Huang and Shi-Yang Lin) with at least 2 years of experience in the performance and interpretation of breast US, as well as at least 6 months of experience in performing elastography before the onset of this study. SWE images were reviewed and con rmed to be eligible by a third radiologist (Xiao-Qing Pei) who had 20 years of experience in ultrasonography and 5 years of experience in breast elastography imaging. The US examination and measurement procedures are shown in Fig. 2.
All breast MRI examinations were performed on an eight-channel 3.0-T system (Discovery MR750, GE Medical Systems, Milwaukee, WI, USA). Patients were imaged in the head-rst prone position, and images were obtained with bilateral axial views. Images were interpreted by radiologists with more than 5 years of experience with MRI of the breast. Measurements of the maximal diameter (D MRI ) of breast cancer were obtained on post-contrast subtracted T1-weighted images.

Imaging Parameters Calculation
The maximal and minimal values of SWV (SWV max and SWV min ) were recorded by averaging three measured values, respectively. The mean SWV values (SWV mean ) for the six SWV-ROIs were calculated. 25 Additionally, the area ratio (A E/B ) was calculated as the ratio of the area of the breast tumor in the SWE time image to that in the B-mode screen. The relative changes of the breast lesion [Δ values (%)] in the Bmode US, SWE, and MRI images were calculated at each time point. The equations used to calculate these parameters can be found in Appendix A.

Pathological evaluation
All pathological results were determined by two board-certi ed pathologists with consensus who were blinded to the imaging information for the patients.
Before the onset of NAC, the diagnosis was determined with a US-guided core needle biopsy. Samples from the core needle biopsy were examined to obtain the biological characteristics of the breast tumors.
The expression levels of estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and Ki-67 in core needle biopsy specimens were evaluated with immunohistochemistry (IHC) and uorescence in situ hybridisation (FISH). The criteria for the IHC, FISH and molecular subtype results were determined according to the St. Gallen consensus. [26][27][28][29][30] After eight cycles of NAC, all patients received mastectomy along with ALND or sentinel lymph node biopsy (SLNB). Assessment of the NAC response was performed using the Miller-Payne system as de ned in Appendix B. A grade of 4 or 5 was categorized as a Major Histological Response (MHR) and a grade of 1, 2, or 3 was de ned as a Non-major Histological Response (NMHR).

Data collection
Data on the demographics, IHC characteristics of breast cancer, breast tumor size/stage, lymph node stage, clinical stage, and pathological response grade of NAC were collected. During the chemotherapy follow-up, the maximum diameter of the breast tumor on the B-mode US, the maximum and average values of SWV, the ratio of tumor area on the elastic image to that on the B-mode US, and the maximum diameter of the breast tumor on the enhanced T1 MRI image were recorded.

Statistical analysis
Continuous data were presented as the mean and standard deviation (SD), and categorical variables were presented as counts. Univariate analysis for the pathological response was performed by using a t-test or Mann-Whitney U test to compare continuous quantitative variables between the responsive and nonresponsive groups. The Kruskal-Wallis test, X 2 test, or Fisher exact test were used to compare the categorical variables in the two pathological response groups.
When both the imaging parameters and clinicopathological characteristics showed favorable e ciency in univariate analysis, we generated a multivariate regression model by using the factors with statistical signi cance to create a new predictive modality. A p-value of < 0.05 was entry probability for stepwise of multivariate logistic regression while a p-value of more than 0.10 was de ned as removal probability. The receiver operating characteristic (ROC) curve was drawn to determine the performance of the imaging ndings and combined parameters (CP) in predicting the pathological response. 32 An area under the ROC (AUC) value of > 0.9 indicated a great diagnostic value, 0.9 > AUC > 0.7 suggested a moderate diagnostic value, and AUC < 0.7 was considered to indicate a poor diagnostic value. 32 The cut-off point for these features was determined by maximizing the Youden coe cient. 33

Imaging parameters evaluation
There were no statistically signi cant differences in pre-treatment imaging measurements, including the observed values in conventional US, SWE, and MRI between the two pathological response groups, as summarized in Table 2.  Table 3.  Fig. 3.
Representative dynamic changes in the breast tumor on MR, B-model US, and SWE images for one patient are presented in Fig. 5. Predictive diagnostic performance of combined parameters According to the results of univariate analysis, variables with statistical signi cance for predicting the chemotherapy response after each cycle of NAC (after the rst NAC cycle: molecular subtype and △A E/B ; after the second, third and fourth NAC cycles: molecular subtype, △A E/B, △SWV mean and △SWV max ) were involved in the multivariate logistic regression predictive model.
After completing the screening of variables for the multivariate logistic regression model, only the molecular subtype was identi ed as a variable capable of predicting the chemotherapy response after the rst cycle of NAC (CP1: molecular subtype alone; AUC = 0.83, p < 0.01). After the second cycle of NAC, the combination of the molecular subtype and △A E/B was statistically signi cant for predicting the response to NAC (CP2: molecular subtype and △A E/B ; AUC = 0.92, p < 0.01). Finally, the combination of the molecular subtype, △A E/B, and △SWV mean after the third and fourth cycles of NAC was regarded as a new statistically signi cant predictor for NAC response (CP3 and CP4: molecular subtype, △A E/B and △SWV mean ; AUC = 0.94, p < 0.01).
The performance of the combined parameters for predicting the response to NAC after each chemotherapy cycle is summarized in Table 5. The sensitivity for predicting the chemotherapy response was only 66.04% after the rst NAC cycle, which only depends on the molecular subtype. Furthermore, by adding △A E/B , the sensitivity of the combined parameters increased to 86.79% with a speci city of 85.19% after the second cycle of NAC. The combination of the molecular subtype and △A E/B after the second cycle of NAC had better performance for an association of the pathological response to NAC than the variable (molecular subtype) selected after the rst cycle of NAC. The molecular subtype combined with △A E/B and △SWV mean had statistically signi cantly better predictive performance for a response to NAC after the third and fourth cycles of chemotherapy than the variable (molecular subtype) selected after the rst cycle of NAC (△AUC = 0.11, p < 0.01). ROC curves for the combined parameters after each NAC cycle for predicting the pathological response to chemotherapy are presented in Fig. 4. The AUCs of the combined parameters after each cycle were compared as shown in Table 6.

Discussion
The present study was performed to develop a method to reduce the application of unnecessary chemotherapies in poor responders to avoid severe adverse effects from chemotherapy and guide the selection of an appropriate alternative regimen. The results suggested that SWE can be an early predictor of the pathological response to NAC for breast cancer. The combination of SWE and molecular subtype may be a preferred method for the clinical evaluation of chemotherapy to facilitate the personalization of treatment regimens for breast cancer.
Tumor stiffness obtained with UE, an important biomechanical characteristic, is associated with tumorigenesis and disease progression. Previous experimental studies have revealed that tissue stiffness can be modulated by the extracellular matrix, regulating the proliferation of tumors. 19 20 The composition of the extracellular microenvironment of cancer cells plays a vital role in the breast cancer response to NAC. Higher matrix hardness can increase the risk of chemotherapy resistance. 19 Therefore, as an emerging modality for predominantly representing extracellular matrix features, UE is believed to re ect the microstructural organization of tissues and has the potential to help differentiate a good response to NAC from a poor response.
Compared to SE, SWE provides a more reproducible and quantitative approach for assessing the stiffness of breast lesions. 34 In recent years, the applications and research for SWE in the breast have focused on the diagnosis of breast lesions by assessing the stiffness of tissue. 18 As an emerging and rapidly developing approach to assessing stiffness, SWE has been studied recently for the identi cation of clinical and pathological responses to NAC and has shown great promise in assessing the chemotherapy response in patients with breast cancer. Our study also showed that SWE has an excellent performance in predicting the NAC pathological response in breast cancer patients, similar to those in previous publications 21 35-41 . Furthermore, our study demonstrated that changes in the number of tumor cells and stromal components re ected by SWE can indicate the breast cancer response to NAC earlier than morphological changes re ected by conventional imaging means.
Some scholars believe that NAC's effectiveness for breast cancer is related to various factors, and a single indicator is inadequate to judge the response to NAC. 38 42 43 Although SWE provides a bene t for the identi cation of non-responsive cases, but does not signi cantly improve performance for predicting responders if it is solely used. 44 It has been demonstrated that various breast cancer subtypes differ signi cantly in the chemotherapy response. 45 Therefore, biological features are usually combined with the clinical characteristics of patients with breast cancer to predict the response to NAC. 46 47 Our study also indicated that the molecular subtype is the most signi cant factor related to the NAC response among the clinical and pathological characteristics of the patient at the baseline. After the second cycle of NAC, a combination of the molecular subtype and △A E/B was regarded as statistically signi cant for the prediction of the response to NAC, with signi cantly higher sensitivity. And previous study also proved that SWE is superior in sensitivity compared to standard clinical assessment or dynamic optical breast imaging, highlighting the potential of SWE to protect patients with chemotherapy resistance from unnecessary NAC. 44 48 Moreover, the combination of molecular subtype, △A E/B and △SWV mean , achieved larger AUCs after the third and the fourth cycles of NAC compared to the molecular subtype alone. In summary, our results showed that elastic and molecular characteristics can be combined to improve the performance for predicting the response to NAC.
Our study revealed that SWE has better potential to predict the pathological response to NAC in breast cancer patients compared to conventional imaging modalities. The combination of the molecular subtype and △A E/B could predict the NAC response with excellent performance after the second cycle of chemotherapy. Early evaluation of the response after two cycles of NAC could provide reliable information about chemotherapy resistance. Moreover, a previous study was performed in an attempt to overcome chemotherapy resistance. 49 The results of our study agreed with previous investigations on the same topic, indicating the potential of SWE for predicting the pathological response to NAC.
Furthermore, a decrease in the longest diameter of tumor evaluated routinely in clinical application using conventional imaging means has not achieved satisfactory outcomes in predicting response to NAC.
After NAC, tumor cells become hypoxic and fragment leaving brotic and collagenous tissues, which may keep the diameter of tumor unchanged while the volume reduce considerably. When tumor respond to NAC, decrease in tumour cells and changes in extracellular matrix will appear, which could be re ected by UE. 50 Thus, this study focused on early assessment of the response to NAC based on relative changes in the both morphological and biomechanical characteristics of the tumor. Most of previous works exclusively investigated UE features at baseline or before surgery. 21 35 36 39 44 To our knowledge, this study is the rst work to assess and compare the performance of SWE, conventional US and MRI for differentiating the response to NAC in the early course of chemotherapy. Further, it is the rst attempt to assess the performance of the combination of SWE and molecular subtype in predicting the NAC response, which is less-cost, non-invasive and more accessible with enormous potential clinical and economic bene ts.
However, several limitations were present in the study. The rst limitation is the unavoidable bias due to the small sample size and the observational study design in a single research center. Large-scale randomized controlled trials are needed to con rm our ndings in the future. Next, we did not have enough data to perform subgroup analysis on the speci c molecular type of breast cancer due to the limited sample size. Finally, SWE should be performed on axillary lymph nodes to provide more information regarding the response to NAC. Further research on SWE is warranted to achieve more practical applications and identify broader prospects for assessing and predicting the responses to NAC for breast cancer.

Conclusion
In summary, SWE could be used for the early assessment and prediction of the pathological response to NAC for breast cancer. Our ndings highlight the potential utility of the combination SWE with molecular subtype to early provide useful information on response to NAC for clinical decision-making.    The ROC of combined parameters after rst, second, third, and fourth cycles of NAC (t1-t4).