Fax: (713) 563-2366
Breast conservation after neoadjuvant chemotherapy†
A prognostic index for clinical decision-making
Article first published online: 7 JAN 2005
Copyright © 2005 American Cancer Society
Volume 103, Issue 4, pages 689–695, 15 February 2005
How to Cite
Chen, A. M., Meric-Bernstam, F., Hunt, K. K., Thames, H. D., Outlaw, E. D., Strom, E. A., McNeese, M. D., Kuerer, H. M., Ross, M. I., Singletary, S. E., Ames, F. C., Feig, B. W., Sahin, A. A., Perkins, G. H., Babiera, G., Hortobagyi, G. N. and Buchholz, T. A. (2005), Breast conservation after neoadjuvant chemotherapy. Cancer, 103: 689–695. doi: 10.1002/cncr.20815
Presented at the 86th Annual Meeting of the American Radium Society, Napa Valley, California, May 1–5, 2004, where Allen M. Chen was the recipient of the Young Oncology Essay Award.
- Issue published online: 3 FEB 2005
- Article first published online: 7 JAN 2005
- Manuscript Accepted: 20 OCT 2004
- Manuscript Received: 12 OCT 2004
- Nellie B. Connally Breast Cancer Research Fund
- Arlette and William Coleman Foundation
- Stanford and Joan Alexander Foundation
- breast conservation;
- neoadjuvant chemotherapy;
- prognostic index;
- ipsilateral breast tumor
The appropriate selection criteria for breast-conserving therapy (BCT) after neoadjuvant chemotherapy are poorly defined. The purpose of the current report was to develop a prognostic index to help refine selection criteria and to serve as a general framework for clinical decision-making for patients treated by this multimodality approach.
From a group of 340 patients treated with BCT after neoadjuvant chemotherapy, the authors previously determined 4 statistically significant predictors of ipsilateral breast tumor recurrence (IBTR) and locoregional recurrence (LRR): clinical N2 or N3 disease, residual pathologic tumor size > than 2 cm, a multifocal pattern of residual disease, and lymphovascular space invasion in the specimen. The M. D. Anderson Prognostic Index (MDAPI) was developed by assigning scores of 0 (favorable) or 1 (unfavorable) for each of these 4 variables and using the total to give an overall MDAPI score of 0–4.
The MDAPI stratified the 340 patients into 3 subsets with statistically different levels of risk for IBTR and LRR after neoadjuvant chemotherapy and BCT. Actuarial 5-year IBTR-free survival rates were 97%, 88%, and 82% for patients in the low (MDAPI overall score 0 or 1, n = 276), intermediate (MDAPI score 2, n = 43), and high (MDAPI score 3 or 4, n = 12) risk groups, respectively (P < 0.001). Corresponding actuarial 5-year LRR-free survival rates were 94%, 83%, and 58%, respectively (P < 0.001).
Patients with an MDAPI score of 0 or 1, which made up 81% of the study population, had very low rates of IBTR and LRR. The MDAPI enabled the identification of a small group (4%) of patients who are at high risk for IBTR and LRR and who may benefit from alternative locoregional treatment strategies. Cancer 2005. © 2005 American Cancer Society.
Neoadjuvant chemotherapy has become a widely accepted in the multimodality treatment of both operable and inoperable breast tumors.1–3 The advantages of this sequencing strategy include permitting the in vivo assessment of disease response to a particular chemotherapy regimen and allowing selected patients in whom mastectomy was recommended initially the possibility of being treated with breast conservation.4–9 However, the use of breast-conserving therapy (BCT) after neoadjuvant chemotherapy remains controversial because of concerns that the rates of ipsilateral breast tumor recurrence (IBTR) and locoregional tumor recurrence (LRR) may be higher than those reported for BCT when surgery is used first. These concerns have arisen because conservative surgery directed at the postchemotherapy residual tumor nidus may, theoretically, risk leaving an increased microscopic burden of disease in the tumor bed region of the breast. The clinical data regarding IBTR and LRR after neoadjuvant chemotherapy and BCT have been inconsistent. Some series have reported rates ≤ 30% and others have reported rates roughly equivalent to those of patients treated with BCT after surgery.10–15 As a result, considerable debate exists regarding how to optimally select appropriate patients who can be treated safely with this approach.
A major reason for the differences in reported outcomes between series is the varying selection criteria that were used to determine BCT eligibility after neoadjuvant chemotherapy. Although some general guidelines for patient selection exist, the group of patients who meet these criteria remain rather inhomogenous with respect to their subsequent risk of IBTR and LRR. In a previous analysis of patients treated with BCT after neoadjuvant chemotherapy at The University of Texas M. D. Anderson Cancer Center (MDACC; Houston, TX), we identified four risk factors that predicted IBTR and LRR.16 These factors were advanced lymph node disease (N2 or N3) at initial clinical presentation, pathologic tumor size > 2 cm, multifocal pattern of residual disease, and lymphovascular space invasion. The presence of any 1 of these factors was associated with a 5-year actuarial IBTR-free rate of 87–91% and a 5-year LRR-free rate of 74–84%. However, it remains unclear how these factors interact with one another and how to best incorporate these data in the context of clinical decision-making. In the current study, we developed the M. D. Anderson prognostic index (MDAPI) based on these four factors with the goal of stratifying patients into subgroups with distinct risks for IBTR and LRR after treatment with BCT after neoadjuvant chemotherapy.
MATERIALS AND METHODS
The data from 340 consecutive patients with histologically confirmed, noninflammatory breast carcinoma treated with BCT after neoadjuvant chemotherapy at the MDACC between 1987 and 2000 were utilized to develop the MDAPI. These patients represent the same population reported in our previous study.16 Demographic, clinicopathologic, and treatment variables were abstracted retrospectively from the medical records of each patient. Patients were staged in accordance with the 2002 American Joint Committee on Cancer (AJCC) guidelines. Disease status was assessed at presentation using physical examination, mammography, and ultrasound of the breast and lymph node basin. All patients underwent staging evaluations to exclude the presence of metastatic disease. Table 1 shows the clinical and tumor characteristics of the study population. Nearly all patients (96%) had Stage II or Stage III disease and only 12 patients (4%) had Stage I disease. The median age of the population was 47 years (range, 22–84 years).
|Characteristics||No. of patients (%)|
|< 40||97 (29)|
|> 60||51 (15)|
|Residual tumor size (cm)|
|> 2||46 (14)|
|Solitary mass||182 (54)|
|Multifocal residual disease||78 (23)|
|No residual disease||80 (24)|
|Surgical margin status|
|Negative (no tumor)||266 (78)|
|Close (< 0.2 cm)||53 (16)|
|Involved (tumor present)||15 (4)|
|Lymphovascular space invasion|
|No. of positive lymph nodes|
|> 10||15 (4)|
The neoadjuvant chemotherapy regimen followed those outlined in prospective institutional protocols that were open during the study period and were generally doxorubicin or taxane based. Full details concerning treatment have been documented in previous reports.16–18 All patients were evaluated in a multidisciplinary setting after completion of neoadjuvant chemotherapy to determine eligibility for BCT. The conservative surgical procedure involved excision of the residual primary tumor with a margin of normal tissue. In most patients, no attempt was made to resect the prechemotherapy tumor volume. When final pathologic examination indicated positive or unknown margins, patients typically underwent reexcision to obtain negative margins or, alternatively, they were converted to mastectomy. The choice of axillary procedure was determined by patient and physician preference. Standard Level I and II axillary lymph node dissection, with or without sentinel lymph node biopsy, was performed in 276 patients (81%). Forty-one patients (12%) had a sentinel lymph node biopsy alone, and no axillary surgery was performed in the remaining 23 patients (7%). All patients were treated with adjuvant external-beam radiotherapy to the intact breast with tangential fields. The median breast dose was 50 gray (Gy) delivered in 25 fractions over 5 weeks, with most patients receiving a tumor bed boost (median dose, 10 Gy) using electrons. Regional lymph node radiotherapy was delivered at the discretion of the practicing radiation oncologist. All patients received the entire planned course of radiotherapy.
Two hundred sixty-two patients (77%) received postoperative chemotherapy with indications varying depending on patient and physician biases as well as the protocol open at the time of treatment. In general, tamoxifen was recommended to postmenopausal patients with estrogen receptor-positive tumors after the completion of adjuvant chemotherapy or after surgery in patients who did not receive adjuvant chemotherapy. One hundred thirty-one patients (39%) received tamoxifen.
Pathologists specializing in oncology analyzed all specimens and reported residual tumor size, margin status, presence of lymphovascular space invasion, and pattern of residual morphology. Only 15 patients, representing 4% of the total patient population, had positive margins. For the purpose of the current study, a multifocal pattern of residual tumor was defined histologically by noncontiguous foci of disease. In such cases, pathology reports typically described nests of tumor visible on multiple slides and interspersed among fibrosis, necrosis, granulomas, and giant cells. Pathology slides were not re-reviewed for the current study.
In our previous analysis of the outcomes of these patients, we used univariate analysis to identify 4 statistically significant predictors of IBTR and LRR (P < 0.05, all).16 Table 2 summarizes the results of these analyses. These four factors were initial clinical lymph node status (N0–N1 vs. N2–3), pathologic tumor size (≤ 2 cm of invasive disease vs. > 2 cm), pattern of tumor morphology after pathologic analysis (solitary vs. multifocal residual disease), and lymphovascular space invasion in the tumor specimen (present vs. absent).
|Variables||5-Yr IBTR-free rate (%)||P value||5-Yr LRR-free rate (%)||P value|
|Clinical N classification|
|Pathologic tumor size (cm)|
|Lymphovascular space invasion|
The MDAPI was derived by assigning a score of 0 (favorable) or 1 (unfavorable) based on the presence or absence of each of these factors. The final MDAPI score (0, 1, 2, 3, or 4) for any patient was determined by totaling the scores from the 4 individual variables in question. The objective of the MDAPI was to create statistically different subgroups based on risk for IBTR and LRR using these four predictors. Because 9 patients had either incomplete or unknown values, the data from 331 patients with known values for all 4 variables were used to perform our final analysis. IBTR-free and LRR-free survival rates based on MDAPI scores were estimated by the Kaplan–Meier method. All events were measured from the date of histologic diagnosis. The median follow-up period for surviving patients was 63 months (range, 10–180 months). The statistical significance between survival curves was determined by two-sided log-rank test.19 All tests were 2 tailed, and P < 0.05 was significant.
Of the 340 patients, 16 (5%) experienced an IBTR, which resulted in an actuarial 5-year IBTR-free rate of 95%. Twenty-nine patients (9%) developed LRR, which resulted in an actuarial 5-year LRR-free rate of 91%. Sites of LRR included the ipsilateral breast (n = 16 [55%]), supraclavicular fossa (n = 7 [24%]), infraclavicular fossa (n = 2 [7%]), axilla (n = 2 [7%]), and internal mammary lymph nodes (n = 2 [7%]). Distant metastases developed in 45 patients (13%), yielding a 5-year distant metastases-free rate of 87%.
The distribution of MDAPI scores for the 331 patients with known values for all 4 variables was as follows: 157 patients (47%) had an MDAPI score of 0; 119 patients (37%) had an MDAPI score of 1; 43 patients (13%) had an MDAPI score of 2; 12 patients (4%) had an MDAPI score of 3; and no patients had an MDAPI score of 4.
Tables 3 and 4, respectively, show the crude failure and actuarial 5- and 10-year actuarial rates of IBTR-free and LRR-free survival stratified by MDAPI scores for the study population. Of the 157 patients with an MDAPI score of 0, only 2 developed an IBTR and 4 developed an LRR, resulting in actuarial 5-year IBTR-free and LRR-free survival rates of 99% and 97%, respectively. Of the 119 patients with an MDAPI score of 1, 6 developed an IBTR and 9 developed an LRR, yielding actuarial 5-year IBTR-free and LRR-free survival rates of 94% and 91%, respectively. Of the 43 patients with an MDAPI score of 2, 4 developed an IBTR and 8 developed an LRR, resulting in actuarial 5-year IBTR-free and LRR-free survival rates of 88% and 83%, respectively. Of the 12 patients with an MDAPI score of 3, 3 developed an IBTR and 6 developed an LRR, yielding actuarial 5-year IBTR-free and LRR-free survival rates of 82% and 58%, respectively.
|MDAPI score||No. of patients||No. of patients with IBTR (%)||No. expected||RR||5-yr IBTR-free rate (%)|
|MDAPI score||No. of patients||No. of patients with LRR (%)||No. of expected||RR||5-Yr LRR-free rate (%)|
Because patients with MDAPI scores of 0 and 1 had clinically acceptable rates of IBTR, we elected to combine these patients into 1 group. This left 3 subsets with distinctly different outcomes: a low-risk group (MDAPI score = 0 or 1), an intermediate-risk group (MDAPI score = 2), and a high-risk group (MDAPI score = 3 or 4). Figures 1 and 2 illustrate IBTR-free survival and LRR-free survival, respectively, for these three groups. Five-year IBTR-free survival rates for patients in the low-risk (n = 276), intermediate-risk (n = 43), and high-risk (n = 12) groups were 97%, 88%, and 82%, respectively. The P value of the difference among the 3 groups overall was 0.0001. The difference between the low and intermediate-risk curves was significant at a value of P = 0.05. The P value for the difference between the intermediate-risk and the high-risk was 0.11. The corresponding 5-year LRR-free survival rates for these groups were 94%, 83%, and 58%, respectively. The difference among the 3 groups overall was significant (P < 0.0001). The difference between the low and intermediate-risk curves was significant (P = 0.001) and the difference between the intermediate-risk and the high-risk was significant (P = 0.009). Relative risk ratios for IBTR were 0.7, 2.0, and 7.0 for the low-risk, intermediate-risk, and high-risk groups, respectively, when compared with the population as a whole. For LRR, the corresponding risk ratios were 0.7, 2.2, and 8.2, respectively.
The current study identified and defined specific subgroups based on the risk for IBTR and LRR in patients with breast carcinoma treated by BCT after neoadjuvant chemotherapy. By quantifying pretreatment and pathologic prognostic variables for IBTR and LRR, we devised a classification system that can be used in clinical decision-making and to counsel patients treated with this multimodality approach. Using the MDAPI score, patients who undergo BCT after neoadjuvant chemotherapy can be stratified into a low, intermediate, or high-risk group for IBTR and LRR. It is critical to note that this prognostic index assumes some predefined selection criteria. For example, all patients analyzed in our series had surgery after neoadjuvant chemotherapy and all patients were treated with breast radiotherapy followed by a tumor bed boost. Furthermore, the percentage of patients with positive surgical margins was so low that it could not be analyzed as a prognostic factor. Finally, all patients were selected for BCT after neoadjuvant chemotherapy only if they had resolution of any skin changes during the chemotherapy treatment and had no evidence of macroscopic residual disease or any mammographic abnormalities in the breast after surgery. All of these criteria need to be satisfied for the MDAPI to have validity.
The selection criteria for BCT after neoadjuvant chemotherapy are critically important. Various studies investigating BCT after neoadjuvant chemotherapy differ significantly with respect to their reported rates of IBTR, with rates ranging from those similar to the expected rates after BCT when an initial surgery is performed to rates exceeding 20%.20–23 It is clear that some patients (e.g., patients with inflammatory cancer) included in series with higher IBTR rates would not meet the selection criteria of other series. Similarly, differences in therapeutic approach also may have contributed to the observed variability in outcome across institutions. For instance, some series used radiotherapy as the sole locoregional treatment in patients with a clinical complete response after neoadjuvant chemotherapy.13, 14, 23 Others included a relatively large percentage of patients who were treated with positive margins, with lower radiotherapy dosages, and without postoperative systemic therapy.11, 23 For example, in the Rouzier et al. study,11 11% of patients had positive margins and 47% did not receive tumor bed boosts as a component of radiotherapy compared with 4% and 0%in the current analysis. Both of these factors have been identified as independent predictors of IBTR in patients treated with BCT in the traditional setting utilizing surgery first and again highlight the importance that these factors be considered in addition to those comprising the MDAPI.24–27 The value of the MDAPI is that it further refines the risk of IBTR for patients who meet eligibility criteria currently considered as standard (i.e., noninflammatory cancer, use of surgery in all patients, achievement of negative margins, and no clinical or radiographic evidence of residual disease after surgical resection). The MDAPI is based on 4 factors (clinical N2 or N3 disease, residual pathologic tumor > 2 cm, lymphovascular space invasion, or multifocal pattern of residual disease), which correlated with both IBTR and LRR in univariate analyses. We previously reported that the 5-year rate of IBTR associated with the presence of any one of these factors ranged from 9% to 13%.16 However, the clinical implication of these IBTR rates was not clear because some patients with a given risk factor had additional factors that also contributed to the risk of IBTR. We developed the MDAPI to account for these contributions and to provide a more accurate assessment of how the interaction of these four factors influences the risk of IBTR and LRR.
The MDAPI enabled the identification of a relatively small cohort of patients with a high risk of developing IBTR and LRR after the BCT/neoadjuvant chemotherapy regimen has been received. Specifically, patients with an MDAPI score of 3 or 4 had a 5-year IBTR rate of 18% and a 5-year LRR rate of 42%. This subgroup of patients, which represented only 4% of the entire population, may benefit from alternative treatment strategies. One approach may be to adapt a lower threshold for reexcisions or completion mastectomy in this patient population. However, further data are needed to define the risk of LRR after mastectomy and postmastectomy radiotherapy for patients considered to be at high risk by the MDAPI, because removing the breast may not necessarily lower the risk for LRR in patients with more aggressive tumors. For example, our group has demonstrated that clinical N2–N3 disease is also a risk factor for LRR after neoadjuvant chemotherapy, mastectomy, and postmastectomy radiotherapy.28 In addition, decisions regarding locoregional therapy for patients at high risk for IBTR and LRR should also consider the corresponding risk of developing distant metastasis, in that the factors used in the MDAPI also affect this risk. This high-risk population may benefit from novel biologic therapies that may be able decrease this risk of LRR as well as systemic disease recurrence.
In contrast to these high-risk patients, the MDAPI also defined a favorable subgroup of patients (MDAPI of 0 or 1) who had a 5-year IBTR rate of only 3% and a 5-year LRR rate of only 6%. Importantly, this subgroup represented 84% of the study population. Clearly, this favorable outcome suggests that BCT after neoadjuvant chemotherapy is an excellent treatment option for such patients. As previously indicated, the MDAPI is only applicable to patients in whom conventional selection criteria, such as those mentioned above, are met. In addition, the MDAPI would be strengthened if validated in an independent data set.
The MDAPI is a tool that allows physicians and patients to better predict the risks of IBTR and LRR after the BCT/neoadjuvant chemotherapy regimen has been received. By considering clinicopathologic findings, patients can be stratified into three distinct prognostic groups. The development of a prognostic index is of particular relevance because the number of patients being treated with BCT after neoadjuvant chemotherapy is increasing and an understanding of appropriate selection criteria for such patients is necessary to minimize the risk of disease recurrence for such patients.
- 19Analysis of survival data. New York: Chapman and Hall, 1988., .
- 28Radiation treatment improves local-regional control and survival in patients with locally advanced breast cancer treated with neoadjuvant chemotherapy and mastectomy [abstract]. Int J Radiat Oncol Biol Phys. 2003; 57: S238., , , et al.