Based on the logistic regression, we screened for predictors and constructed a concise and well fitted nomogram containing the variables of tumor size, hormone receptor status, regimens of neoadjuvant chemotherapy and cycles of neoadjuvant chemotherapy to predict the pCR rate of HER2 negative patients. This would be a convenient application for clinicians. Using the method of calibration plot with bootstrap sampling, as well as internal and external validation by AUC and ROC, the nomogram proved to be of good fitness.
In this study, we first screened variables that could predict the response to neoadjuvant chemotherapy by univariate logistic regression. Tumor size, hormone receptor status, and neoadjuvant chemotherapy regimens and cycles were included in the construction of the nomogram. Next, we intended to identify several independent predictors of the pCR rate. In the multivariate logistic regression analysis, we found that T4 status (
P = 0.015, odds ratio: 0.281, 95 % CI: 0.101–0.779), hormone receptor positivity (
P < 0.001, odds ratio: 0.224, 95 % CI: 0.125–0.400) and receiving the paclitaxel with carboplatin/paclitaxel with cisplatin regimen (
P = 0.003, odds ratio: 27.696, 95 % CI: 3.131–245.030) were the most important predictors of pCR in this model. Compared with T1 patients, T4 patients had worse responses to chemotherapy, which is consistent with previous research [
20]. Hormone receptor status was another independent predictor, and hormone receptor positive patients had lower pCR rates than hormone receptor negative patients. Our findings are concordant with previous studies [
20‐
22] that show that hormone receptor positive tumor cells are less sensitive to chemotherapy compared with hormone receptor negative cells. Patients treated with paclitaxel with carboplatin/paclitaxel with cisplatin had better neoadjuvant chemotherapy responses compared with those treated with cyclophosphamide, epirubicin and 5-fluorouracil. Anthracyclines such as epirubicin and doxorubicin were once considered to be the most effective agents in the treatment of breast cancer, but the use of them has been declining recently [
10]. In our current study, the anthracycline-based regimens included cyclophosphamide, epirubicin and 5-fluorouracil, epirubicin and 5-fluorouracil followed by paclitaxel or docetaxel with epirubicin and navelbine with epirubicin. Cyclophosphamide, epirubicin and 5-fluorouracil was the standard anthracycline-based regimen, and the pCR rate after 6 cycles of cyclophosphamide, epirubicin and 5-fluorouracil was reported to be 14–15 % [
23,
24]. However, only 1.9 % of patients who received cyclophosphamide, epirubicin and 5-fluorouracil in our study reached pCR, which may be partially due to the relatively higher proportion of larger tumor size (T3: 50.5 %; T4: 13.1 %) and fewer neoadjuvant chemotherapy cycles received (1–2 cycles: 49.5 %) in the cyclophosphamide, epirubicin and 5-fluorouracil cohort. The total pCR rate for epirubicin and 5-fluorouracil followed by paclitaxel or docetaxel and epirubicin patients was low (7.8 %) which may due to the relatively higher proportion of hormone receptor positive patients (87.1 %). The cumulative cardiac toxicity of anthracyclines has also limited its use, especially in older patients or in those with cardiovascular comorbidities. Therefore, non-anthracycline based regimens are required. Paclitaxel, a mitotic inhibitor and anti-microtubule agent, results in a G2-M phase arrest [
25]. Carboplatin and cisplatin share similar anti-cancer mechanisms, as they are both DNA alkylating agents [
26]. The combination of paclitaxel and platinum is now widely used in breast cancer patients, and the agents have no overlapping toxicities [
27]. Previous research has already assessed the efficacy and the toxicity of the paclitaxel with carboplatin/paclitaxel with cisplatin regimen in adjuvant therapy and in neoadjuvant chemotherapy. The pCR rate of patients who received paclitaxel with carboplatin/paclitaxel with cisplatin as neoadjuvant chemotherapy ranged from 9.5 to 19.4 % [
28,
29]. The data from our center is 19.4 %, similar to previous studies. The paclitaxel with carboplatin/paclitaxel with cisplatin regimen achieved greater therapeutic effect than any anthracycline-based regimens, especially in triple negative breast cancer patients. Triple negative breast cancer patients have higher rate of BRCA1/2 (Breast Cancer 1/2) mutation and are sensitive to platinum (because of the deficiencies in the DNA repair mechanism) [
30,
31]. In aggregate, these results suggested that platinum contained therapy is recommend for triple negative breast cancer patients.
The nomogram provides a simple graphical representation of sophisticated statistical prediction models and has been accepted as a reliable tool for predicting clinical events. It is especially widely used in oncology [
11]. Previously, several studies constructed nomograms to predict the pCR rate of neoadjuvant chemotherapy. The first of these studies appeared in 2005 [
12]. Rouzier et al. constructed two nomograms to predict the responses to anthracycline-based neoadjuvant chemotherapy and to combined anthracycline and paclitaxel neoadjuvant chemotherapy. The nomograms were validated externally. Colleoni et al. constructed a nomogram to predict pCR probability based on a population of 783 patients [
13]. The nomogram proved to be well fitted after external validation by 101 patients. However, the HER2 status was not mentioned in these two studies. Keam et al. constructed another nomogram to predict pCR and predict which patients would not relapse [
14]. Overall, 370 patients who received 3 cycles of neoadjuvant docetaxel or doxorubicin were included in this study. However, the HER2 status was not stratified and the validation of the nomogram was only performed internally. The advantage of our research is that we first constructed a nomogram for predicting the pCR rate among HER2 negative patients, and the nomogram was proven to be well fitted by internal and external validation. We selected HER2 negative patients as our target population for two reasons. First, the pCR rates of these patients were relatively low, so individualized therapy for each patient was required. Second, confounding variables such as HER2 blockade treatment were limited in our cohort. Additionally, we discovered that paclitaxel with carboplatin/paclitaxel with cisplatin was the more favored neoadjuvant chemotherapy regimen compared with cyclophosphamide, epirubicin and 5-fluorouracil in HER2 negative patients.
One limitation of our study was that the design was a single center analysis. Applying the nomogram in another database will greatly improve the power of our current result, and we have carefully searched through existing public databases. Unfortunately, we were unable to find a proper database containing all of the variables analyzed in our current study (age, menopause status, tumor size, nodal status, hormone receptor status, neoadjuvant chemotherapy regimens, neoadjuvant chemotherapy cycles and response to neoadjuvant chemotherapy). We expect to assess the nomogram with large-scale randomized prospective clinical trials. The efficacy and safety of the paclitaxel with carboplatin/paclitaxel with cisplatin regimen used in neoadjuvant chemotherapy also needs to be assessed. Another limitation was that the molecular mechanisms of the paclitaxel with carboplatin/paclitaxel with cisplatin regimen (more so than the cyclophosphamide, epirubicin and 5-fluorouracil regimen) were unclear so further research is required in the future to study these mechanisms.