Effect of neoadjuvant chemotherapy regimen on relapse-free survival among patients with breast cancer achieving a pathologic complete response: an early step in the de-escalation of neoadjuvant chemotherapy

Patients with breast cancer who have a pathologic complete response (pCR) to neoadjuvant chemotherapy (NACT) have improved survival. We hypothesize that once pCR has been achieved, there is no difference in subsequent postsurgical recurrence-free survival (RFS), whichever NACT regimen is used. Data from patients with breast cancer who achieved pCR after NACT between 1996 and 2011 were reviewed. RFS was estimated by the Kaplan-Meier method, and differences between groups were assessed using log-rank testing. Cox proportional hazards regression analysis adjusted for age, menopausal status, stage, grade, tumor subtype, and adjuvant endocrine HER2-targeted radiation treatment. Among 721 patients who achieved pCR after NACT, 157 (21.8%) were hormone receptor-positive (HR), 310 (43.3%) were HER2-amplified, and 236 (32.7%) were triple-negative; 292 (40.5%) were stage IIA, 153 (21.2%) were stage IIB, 78 (10.8%) were stage IIIA, 66 (9.2%) were stage IIIB, and 132 (18.3%) were stage IIIC. Most patients (367 [50.9%]) had been treated with adriamycin-based chemotherapy plus taxane (A + T), 56 (7.8%) without taxane (A no T), 227 (31.5%) with HER2-targeted therapy, and 71 (9.8%) provider choice. Median follow-up was 7.1 years. Adjuvant chemotherapy was employed in 196 (27%) patients, adjuvant endocrine in 261 (36%), and adjuvant radiation in the majority (559 [77.5%]). There was no statistically significant difference in RFS by NACT group. Adjusted RFS hazard ratios, comparing each treatment with the reference group A + T, were 1.25 (95% CI 0.47–3.35) for A no T, 0.90 (95% CI 0.37–2.20) for HER2-targeted therapy, and 1.28 (95% CI 0.55–2.98) for provider choice. These data suggest that postsurgical RFS is not significantly influenced by the choice of NACT or cancer subtype among patients achieving pCR.


Background
Neoadjuvant chemotherapy (NACT), originally used to downstage breast cancer, may achieve a pathologic complete response (pCR) in a proportion of patients, with no residual invasive disease in the breast or nodes [1]. The residual cancer burden (RCB) present after NACT reflects response to chemotherapy and predicts survival. Consequently, low (RCB I) or no residual disease (RCB 0, defined as pCR) is associated with improved survival compared with more residual disease (RCB II or III), a relationship that holds true across breast cancer subtypes [1][2][3][4]. Large randomized clinical trials have shown that more recent NACT regimens have provided increased pCR rates [5,6], and some trials have shown resultant improved survival [1,7]. Although the timing of chemotherapy and use of NACT have not historically impacted overall survival (OS) [8][9][10][11][12][13], important therapeutic information may be gained by evaluating tumor response to chemotherapy [14,15]. Although an imperfect tool [1], pCR has been approved by the U.S. Food and Drug Administration as an acceptable surrogate endpoint for key clinical trials [16][17][18][19].
The relationship between complete pathologic response to NACT and survival in triple-negative breast cancer has been used for statistical modeling and clinical trial power calculations [20]. This modeling assumes that all patients with a pCR have similar survival, but this assumption of NACT regimen equivalence has not been examined. The objectives of this study were to determine if a specific treatment regimen may be associated with preferential survival in patients who achieve a pCR and to identify risk factors for recurrence among patients with a pCR.

Methods
Patients diagnosed with primary breast cancer on the basis of cytology or histopathology needle biopsy, treated between 1996 and 2011 with various regimens of NACT at a single high-volume cancer center, were reviewed using a prospectively maintained electronic database. Patients were considered eligible for review if, after NACT, they achieved pCR on the basis of final surgical pathology. pCR in this study was defined as eradication of invasive disease in both the breast and lymph nodes (T0/Tis, N0), the definition used at our institution, consistent with the large pooled analysis performed by Cortazar et al. [1]. Patients were excluded if any part of their chemotherapy regimen was received at an outside institution or after definitive breast surgery. This singleinstitution study was approved by the University of Texas MD Anderson Cancer Center Institutional Review Board.
Patients were divided into the following four major chemotherapy regimens: adriamycin-based chemotherapy alone (A no T), adriamycin-based chemotherapy plus taxane (A + T), HER2-targeted therapy (AT + HER2), or provider choice (PC). Provider choice regimens were all others; that is, all included a taxane base, but with the addition of clinical trial agents such as capecitabine or bevacizumab. Duration of NACT was calculated as the date of first chemotherapy treatment to the date of surgery, including a typical 4-to 6-week window between completion of NACT and surgery. There were three NACT duration cohorts: ≤ 4 months, > 4 and ≤ 7 months, and > 7 months. The group treated for ≤ 4 months was treated early in the study time frame according to a departmental algorithm that included four cycles only (most commonly doxorubicin, cyclophosphamide, and 5-fluorouracil; or fluorouracil, adriamycin, and cyclophosphamide [FAC]). Recurrence-free survival (RFS) of pCR patients was compared between these four regimens, agnostic to the details of drugs received. Recurrence and death data were collected by in-depth chart review. The RFS was estimated by the Kaplan-Meier method; calculations were made at 5 years; and the differences between groups were assessed using log-rank testing. Values were statistically significant if P < 0.05. RFS was next compared by breast cancer tumor subtype (hormone receptor [HR]-positive, HER2-amplified, and triple-negative [TN]), and then within each of these subtypes the RFS was again calculated and compared for the four specific NACT regimens. RFS was next calculated by age group (21-40, 41-60, and > 60 years old), stage (IIA, IIB, IIIA, IIIB, and IIIC), and grade (1-3). Age groups were determined by quartiles, with patients aged 21-40 years constituting the youngest quartile and those aged > 60 years the oldest quartile [21]. Last, the RFS was compared by length of neoadjuvant chemotherapy.
Adjuvant treatments were analyzed via landmark analysis [22] with time point zero set to 6 months following surgery to allow for completion of adjuvant treatments (adjuvant chemotherapy, radiation). The landmark cutoff did not include completion of endocrine or anti-HER2 therapy. Patients who were censored or failed (recurrence or death within 6 months of surgery) were omitted. This approach allowed adjuvant treatments to be included as baseline factors in the analysis. Patients omitted were those who failed as described (n = 4) or if clinicopathologic details (age, grade, stage, HR or HER2 status, adjuvant treatments received) were missing (n = 75). Too few patients had neoadjuvant radiation treatment (n = 1) or neoadjuvant endocrine treatment (n = 6) for landmark analysis. Cox proportional hazards regression analysis adjusted for NACT (using treatment arm A + T as the reference group), age (reference group 41-to 60-year-olds), menopausal status (reference group premenopausal), stage (reference group stage IIA), grade (reference group grade 1), tumor subtype (reference group TN breast cancer), and adjuvant treatments (reference group for each therapy was no treatment received).

Patient characteristics
A total of 921 patients were identified over the 15year period, 200 of whom received part of their chemotherapy treatment at an outside institution and were excluded, leaving 721 patients in the survival analyses ( Fig. 1 were HER2-amplified, 236 (32.7%) were TN, and 18 (2.5%) were of unknown type. Of the 310 HER2amplified patients, 128 were HR+, 178 were HR−, and 4 had unknown HER2 status. Patients were grouped into four chemotherapy regimens: 367 (50.9%) patients treated with A + T, 56 (7.8%) treated with A no T, 227 (31.5%) treated with AT + HER2, and 71 (9.8%) treated with PC ( Fig. 1). HER2+ patients were treated both before and after the standard use of HER2-targeted therapies, 247 received HER2-targeted therapy (20 of these in the PC group), and 60 did not. Survival was also examined within this subset. The median length of NACT was 6.0 months, ranging from < 1 month to 12 months. The majority of patients with shorter duration of NACT (< 4 months) received four cycles of FAC or fluorouracil, epirubicin, and cyclophosphamide (43 of     without issue. Two patients stopped NACT after three cycles because of neutropenia, and one patient proceeded to surgery after only two cycles per the patient's request. Definitive surgical treatments included total mastectomy and axillary lymph node dissection (ALND) in 256 (35.5%), total mastectomy and sentinel lymph node biopsy (SLNB) in 46 (6.4%), skin-sparing mastectomy and ALND in 44 (6.1%), skin-sparing mastectomy and SLNB in 51 (7.1%), partial mastectomy and ALND in 176 (24.4%), partial mastectomy and SLNB in 130 (18%), 1 radical mastectomy (0.1%), 2 wide local excisions after breast conservation
Boldface type represents those statistically significant values

Discussion
It is well established that pCR differs by therapeutic regimen and breast cancer subtype [3,6,14,23], with patients who achieve a pCR demonstrating improved survival over those who do not. However, any influence of the NACT regimen employed on subsequent survival  in patients who have already achieved a pCR has not been explored previously. Data presented in the present study suggest that subsequent RFS among patients achieving pCR is not influenced by or associated with the regimen of NACT used to achieve that pCR. The very high recurrence-free survival in this population supports the value of achieving a pCR, but it also suggests that there may be limited value in continued systemic treatments in the postoperative setting for the pCR patient [15]. Because reaching a pCR, rather than the specific therapies used to achieve it, appears to be paramount, the early determination of treatment response will be an important research focus. Imaging alone [24][25][26] or targeted biopsy techniques [27,28] may be effective tools in determining pCR after NACT. Imaging modalities that show promise in the assessment of chemotherapy response include ultrasound of both breast and axilla [24], dynamic contrast-enhanced magnetic resonance imaging (MRI) [29], MRI texture analysis [25], and trimodal imaging (ultrasound, mammography, and MRI) [30,31]. Shear-wave elastography of tumor stiffness before the start of chemotherapy has been associated with tumor response [26], and 18 F-fluorodeoxyglucose positron emission tomography/computed tomography (PET/ CT) has also been studied [32]. The latter work correlated metabolic response by PET/CT during NACT after as few as two cycles of chemotherapy to improved survival at 71-month follow-up. Such imaging may allow oncologists to determine who is responding to chemotherapy earlier and tailor treatment accordingly or even avoid surgery altogether [33,34]. Such concepts are being explored through the NRG BR005 [31] trial and a National Institutes of Health Breast Cancer Steering Committee initiative [35].
The generalizability of a single-institution cohort to NACT patients treated with current regimens or outside the institution requires confirmation. First, these data lack molecular phenotypes. It is known that luminal A tumors are less sensitive to chemotherapy [1], but the relationship of residual tumor and survival holds true across constructed subtype [3], so the presented work should be negligibly disadvantaged by this. Second, although there was no statistical difference in RFS between NACT regimens, the wide CIs do not demonstrate true equivalence between regimens. This may reflect the relatively low event rate of recurrences after pCR (85 of 721 patients), similar to recent trial reports [7,9,15], or it may be due to sample size. Third, it is difficult to interpret the effect of chemotherapy duration, owing to the retrospective nature of the analysis. NACT durations > 7 months are likely due to NACT complications and unexpected delays of surgical treatment, but the shorterduration NACT patients present an interesting observation. Although it is promising that patients who achieved a pCR after only four cycles had survival similar to those receiving more conventional regimens, many patients at our institution are treated on study protocols or clinical trials, so this may be attributed to observational bias and, as such, should be interpreted cautiously. Regimen durations should be compared in future de-escalation trials. Last, these data were collected before taxotere/cyclophosphamide (TC) or paclitaxel+trastuzumab +/-pertuzumab (TH+/-P) were widely recognized as accepted neoadjuvant regimens, as indicated by the very few patients receiving these regimens, and highlighting the need now for further prospective trials to investigate the safe de-escalation of NACT.
Another consideration in this body of literature is that the relationship between NACT, pCR, and improved survival is complex. For example, the addition of taxanes led to increased pCR in NACT trials and also increased survival in adjuvant trials, suggesting an association between increased pCR and improved survival [36,37]. However, the largest prospective cohort, NSABP B27, failed to show OS or DFS benefit despite a doubling of pCR. This influenced the pooled CTNeoBC analysis [1], which did not validate pCR as a surrogate endpoint for survival, suggesting that survival data after pCR should be interpreted with care.
Despite these limitations, this study does point to "atrisk" patient populations. Patients aged 21-40 years and stage IIIC patients had significantly increased risk of death on multivariate analysis. These higher-risk populations should be considered when making treatment decisions regarding additional adjuvant chemotherapy, and they merit future studies. Also, this study highlights that regardless of how you achieve pCR or RCB 0 (1), the outcome is the same. If a reliable imaging modality can be developed to predict treatment response after only a few cycles of chemotherapy, we may be able to administer taxane and HER2-targeted therapies only, and avoid adriamycin-based therapy, for example. Prospective trials minimizing neoadjuvant cytotoxic agents based on interim findings suggestive of pCR may be warranted.

Conclusions
These data suggest that postsurgical RFS among patients with pCR is not significantly influenced by the type of NACT prior to surgery. Because this work provides indirect data only, meta-analysis of randomized trial data should be explored to evaluate wider populations and larger sample sizes to minimize bias. However, the present study provides some reassurance for designing prospective trials aimed at personalization and de-escalation of neoadjuvant chemotherapy.

Additional file
Additional file 1: Figure S1.