A nomogram based on clinicopathological features and serological indicators predicting breast pathologic complete response of neoadjuvant chemotherapy in breast cancer

A single tumor marker is not enough to predict the breast pathologic complete response (bpCR) after neoadjuvant chemotherapy (NAC) in breast cancer patients. We aimed to establish a nomogram based on multiple clinicopathological features and routine serological indicators to predict bpCR after NAC in breast cancer patients. Data on clinical factors and laboratory indices of 130 breast cancer patients who underwent NAC and surgery in First Affiliated Hospital of Xi'an Jiaotong University from July 2017 to July 2019 were collected. Multivariable logistic regression analysis identified 11 independent indicators: body mass index, carbohydrate antigen 125, total protein, blood urea nitrogen, cystatin C, serum potassium, serum phosphorus, platelet distribution width, activated partial thromboplastin time, thrombin time, and hepatitis B surface antibodies. The nomogram was established based on these indicators. The 1000 bootstrap resampling internal verification calibration curve and the GiViTI calibration belt showed that the model was well calibrated. The Brier score of 0.095 indicated that the nomogram had a high accuracy. The area under the curve (AUC) of receiver operating characteristic (ROC) curve was 0.941 (95% confidence interval: 0.900–0.982) showed good discrimination of the model. In conclusion, this nomogram showed high accuracy and specificity and did not increase the economic burden of patients, thereby having a high clinical application value.


Association between common serological indicators and bpCR rate after NAC (univariable analysis). The association between bpCR and 65 indices was assessed (Supplementary
online).

Association between clinicopathological characteristics and common serological indicators and bpCR rate after NAC (multivariable logistic regression analysis). Multivariable analysis was
performed on the indicators included in the univariable analysis. Multivariable logistic regression showed that BMI, CA125, TP, BUN, Cys-C, K, POH, PDW, APTT, TT, and hepatitis B surface antibodies were independent predictors (Table 3). Of these factors, TP, BUN, Cys-C, APTT, and hepatitis B surface antibody positivity showed positive association to bpCR, while BMI, CA125, K, POH, PDW, and TT showed negative association. Establishment and verification of the nomogram multi-factor prediction model. According to 11 clinical factors including BMI, CA125, TP, BUN, Cys-C, K, POH, PDW, APTT, TT, and hepatitis B surface antibody, a nomogram ( Fig. 2) was developed to predict the bpCR rate of breast cancer patients after receipt of NAC. The predicted rate of bpCR can be obtained by summing the scores of the 11 factors. The Brier score of 0.095 indicated that the nomogram had a high accuracy. In the 1000 Bootstrap resampling internal verification calibration curve (Fig. 3), the trend of the predicted value and the true value were both consistent, and average absolute error between the predicted value and the true value was 0.041, indicating a good calibration between the predicted and actual observed values. The 95% CI of GiViTI calibration band (Fig. 4) did not cross the diagonal bisector, and the p value was 0.370, such that the nomogram was deemed as well calibrated. In the ROC curve analysis, the AUC of the nomogram was 0.941 (95% confidence interval: 0.900-0.982, Fig. 5). When the prediction probability was 0.560, the maximum value of Youden index was obtained, and at this time, the sensitivity was 90.1%; the specificity was 87.8%; all of which indicated that the nomogram had good discrimination and good prediction ability.

Discussion
For breast cancer patients receiving NAC, predicting the probability of attaining bpCR before receiving NAC will help clinicians to formulate the most accurate treatment plan, thereby improving the prognosis. Laboratorybased markers have begun to play an increasingly important role in predicting pCR after NAC [18][19][20] . Laboratory indicators can be easily assessed and comprehensively reflect the patient's whole-body multi-system condition. However, previous studies only selected single laboratory indices, thereby resulting in relatively low prediction efficiency. In order to make the prediction more accurate, we need multi-system serum markers to fully predict the bpCR rate after NAC to achieve accurate individualized treatment for breast cancer patients. www.nature.com/scientificreports/ A previous study found that a lower BMI before breast cancer diagnosis was associated with attaining pCR in NAC 22 ; our results also confirmed this. BMI and breast cancer prognosis likely affect each other through metabolic factors 23,24 . The tumor marker CA125 is a prognostic marker for various tumors, especially gynecological tumors [25][26][27] . Among the three common breast cancer serum tumor markers-carcino-embryonic antigen(CEA), CA153, and CA125-we found that only CA125 was an independent predictor. Therefore, CA125 might be more valuable than CEA and CA153 in predicting the possibility of attaining bpCR after NAC in breast cancer.
Of the 13 tested indicators that reflected liver function, we found that only TB was an independent predictor. In patients with cancer, malnutrition can lead to many undesirable consequences such as decreased quality of life, decreased treatment response, and increased treatment-related toxicity 28 . TB is the main indicator of the nutritional status of the body. This study found that breast cancer patients with TB > 74.15 g/L were more likely to attain bpCR. In some reports, ALB levels could affect the long-term prognosis of breast cancer patients 29 and high DBIL reduced the risk of breast cancer 30 ; however in our study, while ALB and DBIL levels were related to the rate of obtaining bpCR, they are not independent predictors.
Of the five indicators reflecting renal function, only BUN and Cys-C were independent predictors. Studies have shown that elevated plasma Cys-C levels are observed in 40% of breast cancer patients and are positively correlated with tumor volume. Therefore, Cys-C can be used as a marker of breast cancer occurrence and progression 31 . Interestingly, in our study, patients with high levels of Cys-C were more likely to reach bpCR. The mechanism may be that Cys-C as a cysteine protease inhibitor can reduce cancer invasion and metastasis. This phenomenon was confirmed in ovarian cancer 32 , whether breast cancer has the same effect needs further www.nature.com/scientificreports/ research. In addition, we found that patients with low levels of BUN were more difficult to achieve bpCR. Urea is produced by the liver and excreted by the kidneys. Studies have shown that low levels of BUN suggested early damage to liver function 33 , and liver dysfunction is a risk factor for multiple tumor prognosis 34 .
In serum electrolyte testing, we found that low K, low POH, and high Mg were positively related factors of bpCR, while both K and POH levels were independent risk factors. According to research, increased serum potassium levels can be found in many types of tumors 35,36 . The likely mechanism of high potassium in serum promoting tumorigenesis and progression is as follows: (1) Higher levels of serum potassium itself can promote tumor growth through immune mechanisms 37 . For example, a cervical cancer study has shown that potassium inhibits the activity of T cells, leading to a decrease in the body's ability to prevent tumor progression 38 . (2) There may be some inherent individual factors that help regulate serum potassium, which also helps tumor progression. For example, genetic variations in ion channels and ion pumps involved in potassium homeostasis are associated with genes that regulate cell proliferation and differentiation 39,40 . Serum phosphate is an essential nutrient for the synthesis of nucleic acids, phospholipids, and high-energy metabolites such as adenosine triphosphate (ATP); thus, rapidly dividing cells require a continuous supply of phosphate 41,42 . High phosphorus in the serum PDW is an independent predictor in the complete blood count test. Various pro-inflammatory cytokines such as tumor necrosis factor-α, interleukin-1, and interleukin-6 are up-regulated as tumors develop and progress 43 . These cytokines promote the maturation of heterologous megakaryocytes, leading to the production and release of immature platelets of various characteristics and sizes into the circulatory system 44 , thereby increasing PDW values. This explains why breast cancer patients with high PDW have a worse prognosis 45 . Similar results were found in this study, and it is therefore more difficult for patients with high PDW to reach bpCR.
Tumor biology is closely related to coagulation 46 . Existing research showed that patients with malignant tumors often suffered from coagulation dysfunction, which manifested as activation of the coagulation system and fibrinolytic system 47,48 . In this study, two coagulation indicators-APTT and TT-were independently related to bpCR after NAC in breast cancer patients. Patients with low APTT and high TT have lower bpCR rates. Abnormal coagulation function not only increases the possibility of thromboembolism, but also promotes tumor growth and the spread of cancer cells 49 , thus affecting the tumor's response to treatment.
In this study, we found that breast cancer patients with hepatitis B surface antibody positivity were more likely to reach bpCR. The mechanism of hepatitis B virus (HBV) and breast cancer progression and prognosis is unclear, but reports have shown that HBV infection is a risk factor for breast cancer 50 . The mechanism may be related to the interference of hepatitis B virus X protein (HBx) on cell repair mechanism, which leads to cell carcinogenesis 51 . In addition, HBV infection causes liver dysfunction, which affects estrogen regulation, and estrogen is a known risk factor for breast cancer. Therefore, HBV infection may indirectly induce breast epithelial cells to become cancerous 50 .
The human body is a multi-system coordinated and integrated organic entity. Tumors not only affect the local organs and tissues but also the functions of various systems throughout the body. Further, the function of each system in the whole body also reflects the existing degree of tumor influence and the future prognosis of the disease to some extent. Therefore, laboratory indicators can reveal disease prognosis to a certain extent. We analyzed several clinical factors and 65 common laboratory indicators, including tumor markers, liver function indicators, renal function indicators, electrolyte test indicators, complete blood count test indicators, coagulation  www.nature.com/scientificreports/ function tests indicators, and hepatitis B surface antibody to analyze the association between these indicators and bpCR. The selected indicators comprehensively covered the clinical characteristics of tumors, and considered the impact of patients' multiple system functions on bpCR. Finally, 11 independent predictors were selected, and a nomogram model for predicting bpCR after NAC in breast cancer patients was successfully established. The AUC of the ROC and the calibration curve indicated that nomogram had good discrimination and calibration. The Brier score showed that the nomogram had high prediction accuracy. These results confirmed that our nomogram had high predictive value. We believe our model can help clinicians to accurately predict the response of breast cancer patients to NAC and the possibility of attaining bpCR, and select and formulate the most beneficial individualized treatment plan for patients in a timely manner. In addition, our model is also economically advantageous, because we choose conventional clinical pathological characteristics and common serological laboratory indicators, which will not increase the financial burden of patients. Our study has some limitations. First, this is a retrospective analysis performed in a single -center, and the number of patients involved is relatively small. In this study, many factors were tested, so there existed the risk of overfitting. Therefore, large-scale, multicenter studies are necessary to further validate our findings and the accuracy of the nomogram. Second, there are unknown factors related to bpCR that we have not yet evaluated.   www.nature.com/scientificreports/ do not recommend NAC for Luminal A patients unless the tumor is too large to be resected, which is also the reason why there is no Luminal A in our cases. Whether our findings are also applicable to Luminal A patients needs further study and discussion. In Conclusion, we screened 11 independent predictors related to bpCR after NAC from routine clinicopathological features and laboratory serum markers of breast cancer, and successfully constructed a nomogram prediction model with accurate and specific predictions, which do not increase the economic burden of patients, and have high clinical application value.

Methods
Patients and factors. In this retrospective study, patients who were treated in the breast surgery department of the First Affiliated Hospital of Xi'an Jiaotong University from July 2017 to July 2019 were enrolled. The inclusion criteria were as follows: (1) unilateral primary invasive breast cancer diagnosed by biopsy; (2)   All patients received chemotherapy based on taxane and/or anthracyclines in the NAC. Trastuzumab was added to Her2-positive patients' treatment regimen, while some patients received pertuzumab. The tumor size was evaluated by physical examination and B-ultrasound in each cycle, and was evaluated by MRI in every two cycles. According to the NCCN guideline evaluation criteria, those who were effective should complete the neoadjuvant chemotherapy according to the established plan and cycle. When the tumor was not relieved, the treatment plan and cycle should be adjusted in time, and those who were still ineffective after adjustment should be operated. The formulation of the treatment plan for all patients was determined in accordance with the 2019 version of the NCCN Breast Cancer Clinical Practice Guidelines 56 and discussed by general practitioners.
The clinical stage was determined according to the 8th edition of the clinical stage of breast cancer recommended by the American Joint Committee on Cancer (AJCC) 57 . By immunohistochemical staining, ER and PR expression levels < 1% were considered negative. HER-2 status was determined according to the American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines 58 . Molecular typing was determined by St Gallen guidelines 59 . According to the NCCN guidelines 56 , the complete response of the primary pathological site of the breast was defined as histological evidence that no invasive tumor was found in the primary breast lesions after NAC, regardless of the presence of residual ductal carcinoma in situ, namely bpCR(ypT0/is).
BMI is calculated by dividing the patient's weight (kg) by the square of height (m 2 ), that is, BMI = weight (kg)/ height 2 (m 2 ). Hepatitis B surface antibody < 10 mIU/mL as negative and ≥ 10 mIU/mL was positive.
Each patient provided informed consent before treatment. The Study was conducted in accordance with the Declaration of Helsinki and the research code of the First Affiliated Hospital of Xi'an Jiaotong University. And the study was approved by the Ethics Committee of the First Affiliated Hospital of Xi'an Jiaotong University.
Complete blood count was conducted by blood analyzer BC5390 (Mindray, Shenzhen, China); coagulation function test was conducted by automatic hemagglutination apparatus CA7000 (Sysmex, Kobe, Japan); liver function, renal function, and electrolyte tests were conducted by automatic biochemical immune analyzer VITROS5600 (JNJ, New Jersey, US) ; Hepatitis B surface antibody test was conducted by automatic immune analyzer I2000SR (Abbott, Illinois, US).

Statistical analysis.
Of all clinical factors and laboratory indicators, only age was analyzed as a continuous variable, while all other indicators were analyzed as categorical variables. Continuous variables other than age were divided into two groups according to the optimal cut-off value, which was determined by calculating the maximum Youden index (sensitivity + specificity − 1) from the receiver operating characteristic (ROC) curve 60 . The chi-square test or Mann-Whitney U test was used to analyze the association between bpCR and the index. If the expected frequency was < 5, Fisher's exact test was used. All p values were two-sided. Indexes with p ≤ 0.05 in the chi-square test, Mann-Whitney U test, or Fisher's exact test were included in forward stepwise logistic regression (likelihood ratio). Forest plots were drawn based on the results of the univariable analysis, and multivariable binary logistic regression was used to determine independent predictors of bpCR after NAC.  www.nature.com/scientificreports/ Then, based on the multivariable logistic regression model, the nomogram was established. Calibration of the nomogram was carried out by the 1000 bootstrap resampling internal verification and was displayed by the calibration curve. The agreement between the predicted and observed probability was shown by the GiViTI calibration band. The Brier score was calculated to measure the prediction accuracy. The ROC curve was used to display the nomogram discrimination, and the discrimination was quantified by area under the curve (AUC) and the concordance index (C-index). Statistical analysis was performed the IBM SPSS Statistics version 22.0 (IBM Corporation, Armonk, NY, USA) and R software (version 3.6.2; The R Foundation for Statistical Computing, Austria, Vienna). For all analyses, p ≤ 0.05 was considered to indicate statistical significance.

Data availability
The datasets analysed during the current study are available from the corresponding author on reasonable request.