Combination of ultrasonic elastography and Ki-67 index as a novel predictive modality for the prognosis in clinic stage III breast cancer

Background To investigate the (UE) and Methods Of 112 patients, Multi-parameter of UE were obtained after analyzing elastograms of 112 lesions which were recorded classification of Ki-67 by histopathologic diagnosis. Multivariate Survival analysis was performed by using multivariate Cox model. The nomograms was established to predict the probability of clinic stage Ⅲ breast cancer according to the selected independently significant variables in the multivariate Cox model. The model was internally validated using Harrell’s concordance index. Results ⅢC had significant difference in all ultrasonic elastography (UE) parameters (P < 0.05) and were statistically different from ⅢA and ⅢB (P < 0.01). The Cox model was RR (t)= Exp(1.104* X1+0.901*X4+0.972*X5) (X1: classification of Ki-67, X4: classification of strain ratio, X5: classification of Young's modulus). The nomogram showed that Ki-67 had the greatest influence on survival prediction, followed by Young's modulus and strain ratio. Internal validation revealed a concordance index of 0.76. Conclusions Our study explored a novel modality in predicting the prognosis of clinic stage Ⅲ breast cancer combining a tumor marker with breast UE imaging. A nomogram was developed to prove that the model was helpful to estimate the 5-year survival of the patients. with imaging. A multivariable was established and confirmed that the combination of the UE parameters and Ki–67 index was an efficient protocol for predicting the prognosis of clinic stage Ⅲ breast cancer. A nomogram was developed to prove that the model was helpful to estimate the 5-year survival of the patients. Our ultimate goal was to assist physician to tailor chemotherapeutic modalities and regimens for each individual patient with advanced breast cancer based on risk assessment.

The clinic stage Ⅲ (ⅢA, ⅢB, ⅢC) breast cancer was usually considered as advanced breast cancer with a 5-year survival rate range from 30% to 40% and the treatments mainly rely on systemic chemotherapy, local radiotherapy and surgical excision [1,2]. In recent years, the prognostic stage group was first included into the cancer staging system of American Joint Committee Cancer (AJCC) 8th in 2016 [3] because the effect of treatment often determines the prognosis of patients. Therefore, a reliable predictive modality was crucial for surveillance of the disease and evaluation of prognosis.
Previous studies have reported that ultrasonic elastography (UE) which includes compression elastography (CE) and shear wave elastography (SWE) was useful in differentiating benign from malignant breast lesions because of that qualitative and quantitative information on tissue stiffness could be obtained by it additional to that provided by conventional ultrasound [4][5][6], especially to identify breast cancers with different grades which may appear the same sonographic features of conventional ultrasound. It was a radiation-free and easy-to-operate method compare to mammography and magnetic resonance imaging and its repeatability and low price also made it a rare advantage. Researchers suggested that ultrasonic elastography also played an important role in evaluating the prognosis of breast cancer forwhich with higher clinic stage has more aggressive pathological properties tend to have higher stiffness readings and worse prognosis [7,8]. Thus, UE would be helpful to determine optimal individualised treatments for patients with clinic stage Ⅲ breast cancer if it was confirmed to be a reliable prognostic modality.
The Ki-67 antigen, as an important tumor marker associated with cell proliferation, was thought to affect survival of breast cancer [9,10]. A higher (≥25%) or lower ( 12 %) Ki-67 index is a significant predictive factor for the response or resistance to the treatment and the over expression of Ki-67 is also an important risk factor for local recurrence of breast cancer. A recent research proposed that a predictive modality combined radiological and pathological biomarkers could provide individualised treatment regimens which could improve the pathological complete response (pCR) rates and prognosis for patients with advanced breast cancer [7]. Hence, it remains to be observed whether the combination of elastography and other tumor markers can improve the diagnostic performance of prediction through a single marker [11].
In summary, our purpose of the study was to access predictive diagnostic performance of combining UE parameters and Ki-67 index in clinic stage Ⅲ breast cancer using survival analysis and nomogram in order to develop it a novel predictive modality for the prognosis of breast cancer.

Patients selection
This study was approved by our institutional ethics committee and informed consent was gained from each patient before the UE examination. From January 2013 to February 2019, 122 lesions in 118 patients of breast cancer in clinic stage Ⅲ underwent conventional ultrasound and UE before surgical excision or ultrasound guided biopsy (UGB). Inclusion criteria: ①Primary breast cancer, not metastatic cancer and distant metastasis.② Single lesion.③ No radiotherapy or chemotherapy.④ No severe cardiovascular disease. Finally, 112 lesions in 112 patients of breast cancer in clinic stage Ⅲ (mean age 46.92 ± 8.45 years old, range 23-58 years old) were included in the final analysis. 10 lesions in 6 patients were excluded later for the following reasons: 6 lesions in 4 patients were received radiotherapy or chemotherapy before the study. 2 lesions in 2 patients suffered from severe cardiovascular disease. 2 patients were found to have more than one lump.

Instruments and methods
US imaging and UE examinations were performed using an Voluson E10 US scanner (GE, Connecticut, USA) equipped with a 9-15 MHz ML6-15-D linear probe which had a function of real-time CE imaging and an Aprio 500 US scanner (TOSHIBA Tokyo, Japan) equipped with a 7-14 MHz PLT-805AT linear probe which had a function of real-time SWE imaging were used in this research.
All the lesions were first examined in the model of two-dimensional gray-scale to locate the position and measure the size. Then the function of CE were activated and the images were displayed in a mode of split-screen with the US images in the left and the same images with color-coded elasticity features superimposed in the right. Using a free-hand technique, the breast gland was vertically compressed by the probe under slight and constant pressure. The real-time pressure elastic indicator bar, displayed on the right screen, was used to help the operator to obtain accurate elastograms. One reproducible elastogram was selected and digitally recorded by the operator when the real-time pressure elastic indicator bar turned green and lasted for at least 3 or 5 consecutive frames [12]. SWE images were obtained during standard gray-scale US imaging. The SWE probe was applied as slightly as possible to the lesions and kept for a few seconds to allow the highquality SWE images to be frozen and saved. An optionally sized region of interest (ROI) box was set in the lesion area, including the lesion, surrounding normal breast gland and adjacent fatty tissue [13]. The combined gray-scale US and SWE examination was between 5-10 min, of which 1-2 min were for SWE image acquisition. The elasticity of lesion was depicted by a color-coded map representing Young's modules (kPa) and shear wave speed (m/s) at each pixel with a color range from blue (soft) to red (hard). The average of the three measurements from three images was used for further analysis.
The blue area percentage was obtained by analyzing the elastograms of lesions, from which the elastic score was determined by the Tsukuba score system. The strain ratio was obtained after the strain rate of the lesion and the soft tissue adjacent to it at the same depth was measured. Young's modulus and shear wave speed were automatically calculated from SWE images by the machine's software system [12,13]. All the elastic imaging operations were performed by two performers who had more than 3 years' experience in breast ultrasonic examination and evaluated by two observers who had more than 6 years' experience in breast ultrasonic diagnosis.

Immunohistochemical analysis
All the specimens collected after surgical excision and UGB were evaluated by the same pathologist. Positive status for ER (estrogen receptor) and PR (progesterone receptor) was defined as nuclear staining in≥1% of the tumour cells. Immunohistochemical staining for HER2 (human epidermal growth factor receptor-2) was scored according to standard criteria as 0, 1+, 2+, or 3+. Scores of 0 and 1+ were considered as negative. Cells with Ki-67 nuclear staining were considered to be positive (cancerous). The proliferation index

Statistical analysis
Statistical analyses were performed using SPSS version 22.0 software (IBM Inc, Chicago, USA). Numerical variables were expressed as means±standard deviations. Variance of analysis (ANOVA) was used to analyse the differences in clinicopathological characteristics among the three subgroups (ⅢA, ⅢB, ⅢC) and (Ki-67Ⅱ, Ⅲ and Ⅳ). Kaplan-Meier method was performed to analyze the survival difference in all cases. Log-rank method was used to compare the survival rates between groups. Survival analysis was performed by using multivariate Cox model. The nomograms were established with the selected independently significant variables in the multivariate Cox model. The predictive accuracy of the model was quantified by calculating the concordance index (c-index). A c-index of 0.5 indicates that outcomes are completely random, whereas close to 1 indicates that the accuracy of the model is better.

Results
Patient population and baseline characteristics 20 patients underwent conservative breast surgery and 92 patients underwent mastectomy. Histopathology and molecular subtypes of the 112 lesions were confirmed by UGB, which revealed that 89 were invasive ductal carcinomas (79.5%), 14 were ductal carcinoma in situ (12.5%) and 9 were invasive lobular carcinomas (8.0%) (Table1). Other immunohistochemical and molecular factors such as ER positive status, and PR positive status were significantly different among the three clinic stage groups (P < 0.01), suggesting that these factors may have predictive value. In addition, the Ki-67 index significantly differed for pairwise comparisons among the three subgroups, which was consistent with previous works (Figure1) [10].

UE evaluation
Acquisition of elastography parameters and process of UGB are shown in Figure 2. UE stiffness readings at three subgroup of clinic stage Ⅲ breast cancer were summarised in Table 2. The elastic score of all elastograms evaluated by all the observers were 3 points and above. ANOVA indicted that ⅢC was statistically different to other two groups in all the parameters (P<0.05) and ⅢB had no statistical difference except elastic score and Young's modulus compared to the ⅢA (Table 2)

Prognosis nomogram and c-index
A nomogram was developed to predict 5-years survival of clinic stage Ⅲ breast cancer using the above three covariates identified in the multivariate Cox model [14]. Each point of independent covariates could be determined according to the intersection of a vertical line drawn from the covariate to the point axis and then the total risk scores were calculated by adding each covariate point (Figure 4). 5-years survival probability of clinic stage Ⅲ breast cancer could be read on the total point axis. The predictive accuracy of the nomogram relative to our data, using Harrell's c-index, was 0.764 before calibration and 0.762 after calibration. A calibration curve illustrated the comparison between the prediction from the nomogram and the actual outcomes of the 112 patients ( Figure 5).

Discussions
Our study showed that the all the UE parameters in subgroup of ⅢC were significantly different compare to the other two subgroups. Only three parameters (classification of Ki-67, strain ratio and Young's modulus) entered into the final equation which represented tumor markers, semi-quantitative parameter and quantitative parameter of UE respectively. The nomogram was successfully established for predicting the prognosis of clinic stage Ⅲ breast cancer and showed that classification of Ki-67 has the greatest influence among the three parameters. The combination of Ki-67 and UE multi-parameter enabled accurate assessment and prediction of clinic stage Ⅲ breast cancer and the nomogram has proved that the combination of the above two made the model most effective. Therefore, it was a novel predictive modality for the prognosis of clinic stage Ⅲ breast cancer. In present, the nomogram, as a weighted model, is widely used in oncology to improve decision making and accurately predict the potential interaction because of its visual advantages among predictors and enable an evaluation of the extent of the impact of each predictor on the probability of survival[23]. Ki-67 was highly significant in our analysis which has been correlated with survival reproducibly among patients in several large studies, followed by Young's modulus and strain ratio. Thus, for example, a patient's Ki-67 index was 68% which belonged to Ki-67 Ⅲ (almost 70 points), value of strain ratio was 3.56 (almost 41 points), value of Young's modulus was 100kPa (almost 49 points) and a total points was 160 that converts to a 5-year survival probability of 15%. We believe that our nomogram could be a simple and easy tool for both the physician and patients for estimating the disease outcome and could contribute to decision making regarding adjuvant chemotherapy[24].
There were still some shortcomings in this study. The potential of selection bias can't be excluded because it was a retrospective design. It was necessary to have a further validation by using other institutional data although our nomogram showed well-validated results. The biological characteristics and biomechanical characteristics of clinic stage Ⅲ breast cancer were significantly related to the prognosis of patients. They were not only an important reference for the clinical monitoring it, but also a reliable prediction for the life and death risk of the patients. The authors will collect more samples to improve the Cox model and strive for more accurate prediction.

Conclusions
Our study explored a novel modality in predicting the prognosis of clinic stage Ⅲ breast cancer combining a tumor marker with breast UE imaging. A multivariable Cox model was established and confirmed that the combination of the UE parameters and Ki-67 index was an efficient protocol for predicting the prognosis of clinic stage Ⅲ breast cancer. A nomogram was developed to prove that the model was helpful to estimate the 5-year survival of the patients. Our ultimate goal was to assist physician to tailor chemotherapeutic modalities and regimens for each individual patient with advanced breast cancer based on risk assessment.  Tables   Table 1 Clinical and pathological characteristics of three subgroups at baseline c. More than one cells have expected count less than 5.      Nomogram predicting the probability of 5-years survival of clinic stage Ⅲ breast