Neoadjuvant chemotherapy (NCT) has emerged as the preferred initial treatment for patients diagnosed with locally advanced breast cancer. NCT is administered to those with locally advanced breast cancer to make tumors more operable, to improve outcomes, and to predict treatment outcomes based on patient responses [1]. The indications for NCT have expanded to include patients with early breast cancer and prospective trials showing no survival disadvantage associated with the use of NCT [2,3]. The rate of the overall response to NCT is high, which allows more patients to undergo breast conservation surgery (BCS) with better cosmetic outcomes [3,4]. Furthermore, BCS after NCT does not increase ipsilateral breast tumor recurrence [5].

In a significant number of patients, however, NCT does not result in sufficient tumor reduction to allow BCS. Moreover, a small number of patients experience disease progression during NCT. The relative ineffectiveness of NCT in these patients may be associated with poor chemosensitivity of the tumor and to their clinicopathologic characteristics. Accurate prediction of each patient’s likelihood of achieving breast conservation after NCT is important in establishing a treatment plan for patients with operable breast cancers. Although many models and nomograms have been developed to predict the response of NCT, these models were designed to predict the ability of NCT in achieving a pathologic complete response (pCR) and not the ability of NCT to achieve BCS [6-8]. We hypothesized that the efficacy of the current models could be improved by including clinical, imaging, and pathologic factors. We, therefore, formulated models that increased the accuracy in predicting the likelihood of achieving successful breast conservation and residual tumor size ≤ 3 cm in patients who receive NCT.

The baseline characteristics of 513 patients are shown in Table 1. The median age was 42.4 years (range, 24 to 67 years), and 54 patients (10.5%) achieved pCR in response to NCT. NCT consisted of a mean 4.8 cycles of taxane and anthracycline-based preoperative chemotherapy. Thirtyseven out of 513 patients (7.21%) were observed to have a positive resection margin initially. BCS after NCT was performed in 257 patients (50.1%).

We have described here the development of nomograms predicting BCS and residual tumor size ≤ 3 cm in breast cancer patients receiving NCT. After matching each factor to the appropriate point, the probability which responds to the total points is figured out. These models were validated in an independent patient set, and showed good performance in terms of discrimination and calibration, with some new factors predicting surgery and tumor response.

These nomograms provided probability estimates that may be useful for individual patients, predicting tumor response to NCT and BCS after NCT. Neoadjuvant therapy is administered to increase the likelihood of performing successful BCS on patients who are not or are borderline eligible for BCS prior to NCT. Our finding, in which 50.1% of patients underwent BCS after NCT, was in agreement with the rates of 38%-68% reported in earlier studies [2,10], suggesting that predictive tools are necessary to assess the likelihood of BCS for patients treated with NCT. Furthermore, as management of breast cancer patients become more multidisciplinary, objective models in clinical settings are going to be necessary for both surgeons and oncologists, when predicting treatment outcomes. Prior to the administration of NCT for breast conservation, it is important to determine whether a patient has a low probability of becoming eligible for BCS. These patients can be offered mastectomy, with or without immediate reconstruction. Alternatively, NCT could be offered after a patient is informed about the low probability of achieving BCS and provides informed consent.

The MDACC nomograms were developed to determine the eligibility for BCS and to predict small residual tumor (≤ 3 cm) for patients receiving NCT [11]. Clinicopathologic factors predicting BCS included the following: ER status, tumor size and grade, multicentricity, and histological type; whereas factors predicting tumor size ≤ 3 cm included ER status, initial tumor diameter, histologic grade and type, and number of chemotherapy cycles. These nomograms were validated in an independent test set of 109 patients, with a c-index of 0.71 for predicting BCS eligibility, although decisions regarding mastectomy versus BCS were reported to be influenced more by subjective factors, such as regional and cultural attitudes than by clinical factors [11]. Furthermore, tumor biology may differ between Asian and Western breast cancer patients [12,13]. Therefore, new tools are needed that can predict which patient in clinical practice are eligible for NCT for breast conserving purposes.

We did not include a histologic grade as a predictive marker for NCT, since the histologic grade in the core biopsy tends to underestimate the grade in the surgical specimen, due in part to the underestimation of the mitotic count [14,15]. The agreement between core needle biopsy and surgical biopsy specimens for grade, mitosis, tubules, and pleomorphism was reported to be 61% [14]. In practice, histologic grade of a presurgical biopsy specimen as a determinant for histologic grade of a surgical specimen may be subjective, dependent on individual pathologists, and our pathologist did not report information from needle biopsy specimens.

The predictive factors of our two nomograms differed, which suggest that any decision of the type of surgery should depend on the surgeon's judgment, not only on objective tumor size after NCT. Indeed, the nomogram for residual tumor size ≤ 3 cm includes greater biologic factors, including ER status, Ki-67 level, DCIS component, and initial size, in agreement with previous models [11]. Furthermore, we found that achieving breast conservation and tumor size ≤ 3 cm were independent of the number of cycles of chemotherapy (> 6 cycles vs. ≤ 6 cycles) or regimen of chemotherapy (including taxane or not).

Although we thought that the initial breast volume would be a predictive factor for breast conservation, we discovered that it was not significant. Rather, the ability to conserve the nipple areola complex was more important for surgical decision making, since the distance between the nipple and the lesion was a significant factor for BCS.

Recently, Mathieu et al. [16] reported that breast cancer index, which is a combination of the two-gene ratio HOXB13: IL17BR(H:1) and the five-gene molecular grade index [17,18], was found to be clinically useful in identifying patients who are not candidates for BCS. In addition, they showed it was the only biologic predictor of BCS in their study. In the near future, some genetic factors predicting chemotherapeutic response or these risk stratifying genes will help our decision making in addition to our model combining clinical, pathological, and imaging variables.

This study had several limitations. The study subjects did not receive homogeneous NCT, and most of HER2 positive patients did not received HER2 targeted therapy; thus, chemotherapy may be a confounding factor. We also did not consider whether BCS was possible before NCT, and we did not distinguish between NCT designed to reduce tumor size or preserve the breast. Moreover, the retrospective nature of the study design made it difficult to obtain information concerning those subjective factors.

In conclusion, we have established a new model to predict BCS and residual tumor size ≤ 3 cm after NCT. This model had a better predictive accuracy than the previously similar models, including novel factors that can affect surgical decision making and tumor response. Use of our nomograms can assist in surgical decision making for patients who are candidates for NCT.