Is sentinel lymph node biopsy alone accurate for breast cancer mastectomy? Results of a cohort study of 2423 patients.

Backgroud : Few patients with mastectomy and only with pN0(i+) or pN1mi sentinel node (SN) were included in randomized trial. To demonstrate SN biopsy accuracy for mastectomy. Methods: We examined results of SN among a multi-institutional cohort of patients, <=cT2-N0, who required total mastectomy, according to SN status and complementary axillary lymph-node dissection (cALND) or not. We have analyzed involved non-sentinel node (NSN) rate at cALND, overall (OS) and disease-free survival (DFS). Results: Among 2423 patients we reported 1307 pN0(i-)SN, 120 pN0(i+)SN, 273 pN1miSN and 723 pN1macro-metastases SN with cALND respectively in 24.5, 73.3, 82.4 and 93.1%. Median follow-up was 42.72 months. Among 320 patients with pN0(i-)SN we observed 35 NSN macro-metastases (10.9%) and among 723 patients with SN macro-metastases, cALND was omitted in 50 patients (6.9%): in multivariate analysis, OS and DFS were not significantly different according to cALND or not. Among 120 patients with pN0(i+)SN and 273 with pN1miSN, cALND were respectively omitted in 32 and 48 patients: age, pT-size and SN-status were predictive of NSN involvement. In multivariate analysis, post-mastectomy radiotherapy, regional nodal irradiation and adjuvant chemotherapy were significantly correlated to cALND and a significant lesser DFS rate was reported for patients without cALND (HR: 3.861, p=0.002). Conclusion: SN biopsy appeared as an accurate procedure for axillary staging of breast cancer mastectomy for pN0 SN status. For pN1-macro-metastases it is not possible to propose to avoid cALND. When SN was involved by ITC or micro-metastases, omission of cALND is still controversial

and SN-status were predictive of NSN involvement. In multivariate analysis, postmastectomy radiotherapy, regional nodal irradiation and adjuvant chemotherapy were significantly correlated to cALND and a significant lesser DFS rate was reported for patients without cALND (HR: 3.861, p=0.002). Conclusion: SN biopsy appeared as an accurate procedure for axillary staging of breast cancer mastectomy for pN0 SN status.
For pN1-macro-metastases it is not possible to propose to avoid cALND. When SN was involved by ITC or micro-metastases, omission of cALND is still controversial and should have a negative prognosis impact in relation with a down staging and under treatment.
Background A decrease of axillary lymph node dissection (ALND) rate since the development and validation of sentinel lymph node biopsy (SLNB) for non-involved sentinel node (SN) 1 has been observed. More recently, since results of ACOSOG Z0011, IBCSG 23-01 and AATRM trials 2,3,4 , complementary ALND (cALND) was questioned in some situations for involved-SN. However, few studies were reported about SLNB accuracy for patients who need total mastectomy.
In Z0011 trial 2 patients undergoing upfront conservative surgery followed by systemic adjuvant therapy for unifocal breast cancer (BC) with 1 or 2 SN involved by micro or macro-metastases, without capsular rupture, were included. Results were discussed concerning the limits and the biases of this study 5, 6. Omission of cALND was held in some teams and recommendations 7,8 , underlining the strict conditions of possible omission of cALND. An evaluation in selected patients considered at high-risk was reported 9 . But some points remains unclear, particularly for patients who required total mastectomy. The possibility of cALND avoidance for patients who required total mastectomy is a timely topic. Very few patients with mastectomy were included in randomized trial 3,4 and only for patients with pN0(i+) or pN1mi SN (86 patients in IBCSG 23-01 trial and 18 in AATRM trial). For this reason we designed the SERC trial to compare outcomes in patients with SN-involvement treated with cALND or no further treatment to the axilla with larger inclusion criteria 10,11 .
The aim of this study was to examine involved non-sentinel node (NSN) rate, axillary recurrence (AR) rate and mainly survival according to SN status among a multiinstitutional cohort of patients who required total mastectomy.

Methods
We conducted a retrospective analysis of 2423 consecutive patients managed between March 1999 and December 2012 in 9 specialized breast centers, referred for mastectomy and SLNB (not included in SERC trial) among a breast cancer data base of 23145 patients.
Total mastectomy was usually proposed to manage multi focal tumors, small invasive BC with a large in situ component, patients with a very small breast volume and patient's choice for total mastectomy.
All patients included in this study have been managed for early BC <= cT2-N0, without pre-operative treatment before SLNB and total mastectomy. We excluded patients with axillary cN1 or T3-4. SLNB was performed using combined isotopic and colorimetric detection or isotopic detection alone with peri-tumoral and/or sub areolar injection 11 .
Although the methods used for SN histological examination were not standardized in the protocol, all sites proceeded similarly: serial sections were performed every 200 microns and stained with standard HE. The number of sections was six to ten, or pursued until node exhaustion in case of large SN. Additional IHC analysis was done in case of negative results at standard examination. For lymph nodes (LN) identified by cALND, routine HE analysis was performed 10,11 .
We have analyzed involved non-sentinel node (NSN) rate at cALND and predictive factors of NSN involvement. Then we have evaluated axillary recurrence (AR) rate, overall survival (OS) and disease-free survival (DFS) according to cALND or not. We used standard descriptive statistics (mean, standard deviation [SD], median and range for quantitative variables, count and frequency for categorical variables) to describe patients and tumors characteristics. In univariate analyses, comparisons were performed using Chi Square.
Multivariable analysis was performed using binary logistic regression. Survival analysis was performed using Log Rank test for univariate analysis and Cox model for multivariate analysis. Overall survival was defined as the time elapsed between surgery and death from any cause. Disease-free survival was calculated from the date of surgery to the first date of loco-regional recurrence, distant recurrence or death from any cause, whichever occurred first.
All statistical analyses were conducted using SPSS 16.0. All statistical tests were twosided. The level of statistical significance was set at a p value of 0.05.
All procedures performed in this study involving human participants were done in accordance with the French ethical standards and with the 2008 Helsinki declaration. This work was approved by our institutional review board (IPC Comité d'Orientation Stratégique).

pN1 macro metastases SN status
Among 723 patients with SN macro-metastases, cALND was omitted in only 50 patients (6.9%). Among patients with involved-SN number known, only one SN macro-metastases was observed in 372 patients and more than one in 257 patients: 124 patients with only one SN macro-metastases had one or more NSN-involved at cALND (124/332: 37.3%). AC and PMRT were delivered more frequently for patients with cALND ( Table 2). In Cox regression analysis, OS and DFS were not significantly different according to cALND or not ( Table 3, 4). AR rates were 1.0 and 0% respectively for patients with and without cALND (7/673 vs 0/50), 0.9% and 2.5% respectively for patients with and without PMRT (6/683 vs 1/40: p=0.330).
In univariate analysis, age, pT size and SN status were significantly predictive of NSN involvement among patients with pN0(i+) or pN1mi SN and cALND ( Table 5). These factors remained significant in binary logistic regression ( Table 5).
Four groups were determined according to pT tumor size < or >= 20mm and age > or <= 40 years-old with NSN involvement rate for pN0(i+) and pN1mi from 0 to 65%. In binary logistic regression cALND was not significantly associated to these 4 sub-groups and SN status ( Table 6).
For pN0(i+) and pN1mi according to cALND or not, PMRT rate was significantly different only for patients with pT<20mm/age>40 years-old, RNI rate was significantly different for patients with pT<20mm/age>40 years-old and pT<20mm/age<=40 years-old, AC rate was significantly different for patients with pT<20mm/age>40 years-old and pT>=20mm/age>40 years-old (Supplementary Table 1).
In binary logistic regression, PMRT, RNI and AC were significantly associated with cALND but also to SN status and pT/age sub-groups ( Table 6).

Conclusions
We reported from a large retrospective cohort of mastectomy, no OS and DFS significant difference between cALND or not for patients with pN0(i-) SN status and for patients with pN1 macro metastases SN status. However, few patients with pN1 macro metastases SN had no cALND. For patients with pN0(i+) or pN1mi SN, lesser DFS was reported for patients without cALND in comparison with patients with cALND in multivariable analysis (HR: 3.861, p: 0.002) bur without significant difference for OS.   14 . It had been reported that SLNB was accurate for large tumors (1101 with tumors > 20 and <30mm and 748 tumors >=30mm) 15 and for multi-focal multicentric tumors 16 even if LN involvement rate was higher for these patients.
For patients who had a macro-metastases SN, the only reported trial with cALND randomization versus only SLNB 2 had included macro and micro-metastases only for conservative treatment. In AMAROS trial 17  trial with randomization between ALND or no axillary surgery for patients with total mastectomy, no survival impact was observed with a long follow-up 18 . However, patients included in this trial had large tumors, which were very different with patients for whom SLNB is indicated, and any systemic treatment was administered with low survival rates in two arms. Omission of regional treatment with ALND had in consequence no survival impact.
In OTOASOR trial 19 , with randomization between cALND versus axillary radiotherapy, patients with tumors up to 3cm diameter, cN0, unifocal or multifocal and breastconserving treatment or mastectomy were eligible: 48.7% of patients had tumors more  PMRT rate was not lower in our study for pN0(i+) and pN1mi for patients with cALND versus cALND (respectively, 34.1% and 21.3% versus 59.4% and 39.6%). In IBCSG 23-01 trial, no PMRT was realized in all patients with mastectomy. The role of RT in the absence of cALND in patients with invaded SN has been extensively discussed 28 . In the ACOSOG Z0011 trial, adjuvant treatments associated with whole breast irradiation (WBI) using axillary tangential fields, likely contributed to the low rate of node recurrence (1% in the group without ALND). However, WBI with tangential fields and regional RT were specified in only one third of cases 29 . Most series evaluating the SN technique showed that tangential fields include the majority of levels I and II but others reported that standard tangential fields of breast RT include the axillary only to a limited extent 30, 31 .
AR is a rare event corresponding to a strong survival pejorative factor 32 .
In Gentilini et al. study, AR rates were significantly different between patients who received WBI or partial breast irradiation after conservative treatment 33 . At 10-years, AR rate were in IBCSG 23-01 trial 1.2% and 1.2% for 86 mastectomies respectively in arm with ALND (1/44) and without cALND (1/42) but AR rates for patients with conservative treatment with IORT without WBI were 0% (0/79) in arm with cALND and 6.25% (5/80) in arm without cALND. In our study, AR rates were no significantly different according to cALND or not and PMRT or notPMRT is usually indicated for patients with lymph node macro-metastases 34  All procedures performed in this study involving human participants were done in accordance with the French ethical standards and with the 2008 Helsinki declaration.
All included patients provided written informed consent before surgery, including the use of their data for research.

Availability of data and material :
Administrative data and clinical data are compiled in a common database and are available to editors and peer reviewers.