Pathological results
Of the 135 breast lesions, 84 (62.2%, 84/135) were benign, including 32 fibroadenomas, 40 adenosis, 3 intraductal papillomas, 8 inflammatory lesions, 1 benign phyllodes tumor, and 51 (37.8%, 51/135) were malignant, among them, there were 46 invasive carcinomas, 3 intraductal carcinomas and 2 mucinous carcinomas.
Analysis of elastic characteristics of breast lesions
The elastic characteristics of benign and malignant breast lesions were seen in Table 1, Figure 1A and Figure 2A. The Emax, Emean, Esd and Eratio of benign breast lesions was significantly lower than those of malignant lesions (P < 0.001), but there was no significant difference in Emin between benign and malignant breast lesions (P=0.202). Besides, the detection rate of “stiff rim sign” in malignant lesions was significantly higher than that of benign lesions (P < 0.001).
Expression levels of TGF- β 1 and other factors in breast lesions
The expression levels of TGF- β 1 and other factors in breast lesions were seen in Table 2, Figure 1B~H and Figure 2B~H. Immunohistochemical staining showed that TGF- β 1 and PI3K were mainly expressed in the cytoplasm, while Smad2/3, Erk1/2, p38 MAPK, JNK2 and AKT were mainly expressed in the cytoplasm and nucleus. And the expression levels of TGF- β 1 and other factors in malignant breast lesions were significantly higher than those in benign lesions (P < 0.001).
When using the expression levels of TGF-β1 and other factors as an index respectively for the differential diagnosis of benign and malignant breast lesions, the ROC curve was shown in Figure 3, the AUC were 0.931 (0.874-0.967), 0.953 (0.901-0.982), 0.920 (0.861-0.960), 0.957 (0.902-0.986), 0.934 (0.874-0.971), 0.960 (0.901-0.989) and 0.939 (0.883-0.974), respectively. Z test showed that there was no significant difference among the seven groups (P > 0.05). When the cutoff value of the average optical density of TGF-β1, Smad2/3, Erk1/2, p38 MAPK, JNK2, PI3K and AKT was > 0.2175, > 0.1496, > 0.2036, > 0.1447, > 0.1583, > 0.2624, > 0.1230 respectively, the sensitivity and specificity for differential diagnosis of benign and malignant breast lesions were 90.0% and 88.2%, 93.9% and 90.2%, 80.0% and 89.4%, 89.1% and 95.5%, 90.2% and 85.0%, 93.6% and 94.4%, 80.0% and 89.4%, respectively.
Correlation analysis
The correlation between the expression levels of TGF-β1 and other factors and elastic characteristics in breast lesions was shown in Figure 4~7. Spearman test showed that the expression levels of TGF-β1 and other factors were positively correlated with Emax, Emean, Esd and Eratio in breast lesions (correlation coefficient=0.879, 0.595, 0.571, 0.562, 0.516, 0.552, 0.619 for Emax, 0.841, 0.549, 0.503, 0.504, 0.516, 0.542, 0.578 for Emean, 0.865, 0.580, 0.566, 0.593, 0.514, 0.559, 0.649 for Esd and 0.746, 0.510, 0.507, 0.506, 0.536, 0.511, 0.550 for Eratio, P<0.0001).
Expression levels of TGF-β1 and other factors in breast lesions with and without “stiff rim sign” could be seen in Table 3. Immunohistochemical staining showed that the expression levels of TGF-β1 and other factors in breast lesions with “stiff rim sign” were significantly higher than those in breast lesions without “stiff rim sign” (P<0.001).
The correlation between the expression levels of TGF-β1 and other factors in breast lesions was shown in Figure 8. Spearman test showed that the expression levels of Smad2/3, Erk1/2, p38 MAPK, JNK2, PI3K and AKT were positively correlated with TGF-β1 in breast lesions (correlation coefficient=0.678, 0.633, 0.645, 0.611, 0.589, 0.663, P<0.0001).
Axillary lymph node metastasis
Expression levels of TGF-β1 and other factors in malignant breast lesions with and without axillary lymph node metastasis could be seen in Table 4. The difference between malignant breast lesions with axillary lymph node metastasis and malignant breast lesions without axillary lymph node metastasis lied in that the expression levels of TGF- β 1 and other factors of former were significantly higher than those of the latter (P < 0.05).
Basing on the expression levels of TGF- β 1 and other factors in malignant breast lesions, ROC curves of which for differential diagnosis of malignant breast lesions with and without axillary lymph node metastasis were shown in Figure 9. The AUC were 0.853 (0.703-0.946), 0.697 (0.529-0.834), 0.694 (0.527-0.832), 0.706 (0.531-0.845), 0.654 (0.466-0.813), 0.667 (0.493-0.813) and 0.689 (0.516-0.831), respectively. When the expression levels of TGF- β 1, Smad2/3, Erk1/2, p38 MAPK, JNK2, PI3K and AKT were cut off by average optical density > 0.3138, > 0.2679, > 0.2457, > 0.2812, > 0.2760, > 0.3075, > 0.2348 respectively, the sensitivity and specificity for differential diagnosis of malignant breast lesions with or without axillary lymph node metastasis were 86.7% and 83.3%, 80.0% and 62.5%, 80.0% and 58.3%, 69.2% and 65.2%, 50.0% and 80.0%, 86.7% and 40.9%, 42.9% and 91.3%, respectively.
Quantile regression analysis
As shown in Table 5~8, the percentile regression analysis was performed with the expression levels of TGF- β 1, Smad2/3, Erk1/2, p38 MAPK, JNK2, PI3K and AKT in breast lesions as independent variables and Emax, Emean, Esd and Eratio as target variables respectively. At different quantiles, only the expression level of TGF-β1 always had a significant positive effect on Emax, Emean, Esd and Eratio, while Smad2/3 only had a certain effect on Emean at the point of 0.75th quartile, a negative effect on Eratio at the point of 0.45th quartile. Erk1/2 only had a certain effect on Emean at the 0.75th quartile. That is, the expression level of TGF-β1 is the most important factor to determine the Emax, Emean, Esd and Eratio of breast lesions.
Logistic regression analysis
Taking the expression levels of TGF-β1, Smad2/3, Erk1/2, p38MAPK, JNK2, PI3K and AKT in breast lesions as independent variables and "stiff rim sign" as dependent variable, logistic regression analysis was performed. The logistic regression equation was established as follows:
Logit (P) =-5.977+17.049 X1
X1 represented the expression level of TGF-β1, that is, the expression level of TGF-β1 was the main factor determining the presence or absence of "stiff rim sign", as shown in Table 9.