Occult Axillary Node Metastases in Breast Cancer Are Prognostically Significant: Results in 368 Node-Negative Patients With 20-Year Follow-Up

VOLUME 26 NUMBER 11 APRIL 10 2008 JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T Occult Axillary Node Metastases in Breast Cancer Are Prognostically Significant: Results in 368 Node-Negative Patients With 20-Year Follow-Up Lee K. Tan, Dilip Giri, Amanda J. Hummer, Katherine S. Panageas, Edi Brogi, Larry Norton, Clifford Hudis, Patrick I. Borgen, and Hiram S. Cody III From the Department of Pathology; Department of Epidemiology and Biostatistics; Division of Breast Oncology, Department of Medicine; Breast Service, Department of Surgery; and the Memorial Sloan-Kettering Cancer Center, New York, NY. Submitted May 17, 2007; accepted January 8, 2008; published online ahead of print at www.jco.org on March 10, 2008. Supported by a grant from the Director s Fund, Evelyn H. Lauder Breast Center. Presented in part at the 38th Annual Meeting of the American Society of Clinical Oncology, May 18-21, 2002, Orlando, FL; and Cody HS III, Borgen PI, Tan LK: Ann Surg Oncol 11:227S- 230S, 2004. Authors disclosures of potential conflicts of interest and author contributions are found at the end of this article. Corresponding author: Hiram S. Cody III, MD, Breast Service, Department of Surgery, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021; e-mail: codyh@mskcc.org. 2008 by American Society of Clinical Oncology 0732-183X/08/2611-1803/$20.00 DOI: 10.1200/JCO.2007.12.6425 A B S T R A C T Purpose In breast cancer, sentinel lymph node (SLN) biopsy allows the routine performance of serial sections and/or immunohistochemical (IHC) staining to detect occult metastases missed by conventional techniques. However, there is no consensus regarding the optimal method for pathologic examination of SLN, or the prognostic significance of SLN micrometastases. Patients and Methods In 368 patients with axillary node-negative invasive breast cancer, treated between 1976 and 1978 by mastectomy, axillary dissection, and no systemic therapy, we reexamined the axillary tissue blocks following our current pathologic protocol for SLN. Occult lymph node metastases were categorized by pattern of staining (immunohistochemically positive or negative [IHC ], hematoxylin-eosin staining positive or negative [H&E ]), number of positive nodes (0, 1, 1), number of metastatic cells (0, 1 to 20, 21 to 100, 100), and largest cluster size ( 0.2 mm [pn0 i ], 0.3 to 2.0 mm [pn1 mi ], 2.0 mm [pn1a]). We report 20-year results as overall survival (OS), disease-free survival (DFS), and disease-specific death (DSD). Results A total of 23% of patients (83 of 368) were converted to node-positive. Of these, 73% were 0.2 mm in size (pn0 i ), 20% were 0.3 to 2.0 mm (pn1 mi ), and 6% were more than 2 mm (pn1a). On univariate and multivariate analysis, pattern of staining, number of positive nodes, number of metastatic cells, and cluster size were all significantly related to both DFS and DSD. On multivariate analysis, each of these measures had significance comparable to, or greater than, tumor size, grade or lymphovascular invasion. Conclusion In breast cancer patients staged node-negative by conventional single-section pathology, occult axillary node metastases detected by our current pathologic protocol for SLN are prognostically significant. J Clin Oncol 26:1803-1809. 2008 by American Society of Clinical Oncology INTRODUCTION Sentinel lymph node (SLN) biopsy has largely replaced axillary lymph node dissection (ALND) as the preferred method of lymph node staging for breast cancer at many centers in the US and worldwide. Viewed as a surgical technique, an extensive literature (including 69 observational studies 1 and four randomized trials 2-5 of SLN biopsy validated by a backup ALND) has established that SLN biopsy is feasible, accurate, safe, less morbid than ALND, and suitable for virtually all patients with stage I to IIIa invasive disease. Viewed as a pathologic technique, SLN biopsy allows the routine performance of lymph node serial sections and/or immunohistochemical stains, methods that have helped to validate the SLN hypothesis, 6,7 increase staging accuracy, 8 reduce false-negative rates, 9 and enhance the prediction of non-sln status in SLN-positive patients. 10 These same methods have also fostered intense debate over (1) the proper method for pathologic examination of SLN, and (2) the prognostic significance of SLN micrometastases, especially those detected only by IHC staining or as isolated tumor cells. 11 There is to date no consensus on either issue, and here we address both. In a cohort of 368 patients with node-negative invasive breast cancer treated between 1976 and 1978 at our institution by mastectomy, ALND, and no systemic therapy, we have restaged the negative axillary nodes 2008 by American Society of Clinical Oncology 1803

Tan et al using our current pathologic protocol for SLN biopsy, and we report the results with 20-year follow-up. PATIENTS AND METHODS Patient Population Over a 2-year period (1976-1978), 744 consecutive unselected nodenegative breast cancer patients were treated on the Breast Service, Department of Surgery, at Memorial Sloan-Kettering Cancer Center (MSKCC; New York, NY); 519 had invasive mammary adenocarcinoma and no prior history of cancer, were treated by mastectomy (radical or modified radical) with an axillary lymph node dissection (ALND) yielding at least five lymph nodes, and received no systemic or radiation therapy. Paraffin blocks were available for review in 368 of the 519 patients, and these 368 comprise our study cohort. This study was conducted under a Waiver of Authorization from our institutional review board. Pathologic Review All diagnoses were initially rendered by MSKCC pathologists, and the original tumor slides were available for rereview in 80% of cases (294 of 368), confirming the diagnosis of invasive carcinoma, and noting tumor size, histologic type, multifocality, and the presence of lymphovascular invasion (LVI). Invasive ductal carcinomas were graded by the modified Bloom-Richardson system. 12 The patient and tumor characteristics among patients with missing axillary blocks (n 151) or missing primary tumor slides (n 74) were substantially similar to those of the fully reviewed cases. Reanalysis of Axillary Nodes For all 368 patients, a total of 2,470 paraffin blocks from the axillary lymph nodes were re-examined using our current pathologic protocol for SLN, as described in detail previously. 13,14 From each paraffin block, two adjacent 5- m sections were cut at each of two levels 50 m apart. At each level, one slide was stained with hematoxylin-eosin (H&E) and the other with immunohistochemistry (IHC), using the anticytokeratin AE1:AE3 (Ventana Medical Systems Inc, Tucson, AZ), for a total of four slides per block. The H&E sections were reviewed and the findings recorded before review of the IHC sections. In cases with isolated IHC-positive cells or microscopic clusters, the cytologic features of the micrometastases (size and nuclear features) were compared with those of the primary tumor, and were accepted as metastases only if they were concordant. Nodal metastases were evaluated for (1) pattern of spread (single cells versus clusters); (2) size of the largest individual cluster to the nearest 0.1 mm; (3) location (subcapsular versus parenchymal), and (4) total number of tumor cells present. When metastases were found in both levels or in multiple nodes, the number of tumor cells at each level of all involved nodes was added up to estimate total tumor burden. Each patient was categorized by pattern of staining (IHC /H&E, IHC /H&E, and IHC /H&E ), number of positive nodes (0, 1, 1), estimated total number of metastatic cells (0, 1 to 20, 21 to 100, 100), and size of largest metastatic cluster ( 0.2 mm [pn0 i ], 0.3-2.0 mm [pn1 mi ], 2 mm [pn1a]), following the most recent American Joint Committee on Cancer (AJCC) Staging Manual (ed 6). 15 Statistical Methods The study end points were overall survival (OS), disease-free survival (DFS) and disease-specific death (DSD). OS was calculated from date of mastectomy to date of death or last-follow-up, and DFS was calculated from the date of mastectomy to date of breast cancer recurrence or last follow-up. DSD was calculated from date of mastectomy to date of death resulting from breast cancer, and death resulting from other causes (approximately half of all deaths) was treated as a competing risk. For OS and DFS, Kaplan-Meier methods were used to estimate survival rates, and Cox proportional hazards models were used to estimate relative risk and to fit multivariate models. For DSD, the cumulative incidence for the competing risk model was estimated using the methods of Gray, 16 and competing risk regression was used to fit multivariate models. 17 All statistical analyses were performed using R (http:// www.r-project.org/) and SAS Software (SAS Institute, Cary, NC). RESULTS Patient Population The clinicopathologic features of the cohort are representative of contemporaneous node-negative breast cancer patients (Table 1). Seventy-five patients died as a result of other causes, 76 died as a result of (or with) disease, and 217 were disease free at last follow-up. Median Table 1. Patient Characteristics Variable No. % Age, years Median 57 Range 24-83 50 123 33 50 245 67 Laterality Right 211 57 Left 157 43 Tumor type Ductal 319 87 Lobular 49 13 Grade I 58 16 II 134 36 III 110 30 Lobular 49 13 Missing 17 5 Multifocality Absent 339 92 Present 29 8 Lymphovascular invasion Absent 318 86 Present 50 14 1.0 64 17 1.1-2.0 180 49 2.0 123 34 Missing 1 1 Converted to node positive No 285 77 Yes 83 23 Pattern of staining IHC /H&E 285 77 IHC /H&E 50 14 IHC /H&E 33 9 No. of nodes positive 0 285 77 1 60 17 1 23 6 No. of metastatic cells 0 285 77 1-20 23 6 21-100 17 5 100 43 12 Largest cluster size Negative 285 77 0.2 mm (pn0 i ) 61 17 0.3-2.0 mm (pn1 mi ) 17 5 2.0 mm (pn1a) 5 1 Abbreviations: IHC, immunohistochemistry; H&E, hematoxylin-eosin. 1804 2008 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Occult Axillary Node Metastases in Breast Cancer Cells and Cluster Size Table 2. Number of Metastatic Cells and Largest Cluster Size, by Method of Staining Pattern of Staining IHC /H&E IHC /H&E IHC /H&E No. % No. % No. % No. of metastatic cells 0 285 100 285 1-20 22 44 1 3 23 21-100 16 32 1 3 17 100 12 24 31 94 43 Largest cluster size Negative (pn0) 285 100 285 0.2 mm (pno i ) 49 98 12 36 61 0.3-2.0 mm (pn0 mi ) 1 2 16 49 17 2.0 mm (pn1a) 0 0 5 15 5 Total 285 50 33 368 Abbreviations: IHC, immunohistochemistry; H&E, hematoxylin-eosin. Total follow-up among survivors was 17.6 years (range, 3 to 25 years), and was less than 5 years in only two patients. No patients were lost to follow-up. Detection of Axillary Node Metastases A mean of 17 nodes (range, 5 to 43 nodes) were examined per patient, and 23% of patients (83 of 368) were converted to node positive. Among these 83, one positive node was found in 71% (59 of 83), two in 16% (13 of 83), three in 6% (five of 83), and more than three in 6% (five of 83). Among patients converted to node-positive, occult metastases were first detected by H&E in 40% of cases (33 of 83), and by IHC in 60% (50 of 83; Table 1). The largest cluster size of the occult metastases was no more than 0.2 mm (pn0 i ) in 73% (61 of 83), was 0.3 to 2.0 mm (pn1 mi ) in 20% (17 of 83), and was more than 2 mm (pn1a) in 6% (five of 83). In 77% (63 of 83), the positive node/nodes were in level I. Metastases to level II with level I negative were found in 11% (nine of 83). Because there were no metastases isolated to level III in the first 79 cases examined, level III nodes were not further examined in the remaining patients. Among patients converted to node-positive, 67% (56 of 83) had invasive duct, 24% (20 of 83) had invasive lobular, 4% (three of 83) had mixed ductal-lobular, and 5% (four of 83) had cancer of other types (mucinous, tubular, papillary, medullary, and metaplastic). Across the entire study, conversion to node positive was noted in 20% of invasive duct (56 of 286), 41% of invasive lobular (20 of 49), 33% of mixed duct-lobular (three of nine), and 17% of other types (four of 24). Patients with invasive lobular carcinoma comprised 13% of the study overall, 15% of those with occult metastases detected by H&E, and 36% of those with occult metastases detected by IHC. Location and Extent of Nodal Metastases Nodal metastases involved subcapsular sinuses in 48% (40 of 83) and lymph node parenchyma in 52% (43 of 83); cases with both subcapsular and parenchymal metastases were designated as parenchymal. Metastases occurred as clusters in 67% (56 of 83) and as single cells in 33% (27 of 83). Number of metastatic tumor clusters per patient ranged from one (14 cases) to more than 100 (four cases). Invasive lobular cancers were over-represented among patients with single-cell (59%) versus clustered metastases (7%). In Table 2 we cross-tabulate number of metastatic cells and largest cluster size by pattern of staining. Among the 50 patients who were IHC /H&E, the number of metastatic cells varied widely, but 98% had a cluster size of no more than 0.2 mm (pn0 i ). For the 33 patients who were IHC /H&E, 94% had more than 100 cells and only 15% (five of 33) had a cluster size of more than 2 mm in size (pn1a). In Table A1 we cross-tabulate number of metastatic cells by cluster size, confirming that the greatest variation in number of metastatic cells was in patients with metastases of no more than 0.2 mm in size, and that all patients with metastases more than 0.2 mm in size had more than 100 cells. Survival Analyses Table 3 and Tables A2 and A3 (online only) report the results of univariate analyses for OS, DFS, and DSD. Pattern of staining is significant for all three outcome measures. Conversion to node positive, method of staining, number of positive nodes, number of metastatic cells, cluster size, tumor size, grade, and LVI are all highly significant for DFS and DSD. Table 4 and Tables A4 and A5 (online only) report the results of multivariate analyses for OS, DFS, and DSD. Because pattern of staining, number of metastatic cells, number of positive nodes, and cluster size all estimate lymph node tumor burden and are therefore correlated with each other, each of these variables is modeled separately ( models 1-4 ). All are highly significant for all three outcome measures, as are LVI, grade and tumor size for DFS and DSD. OS, DFS, and DSD by method of staining are reported in Table A6 (online only), in Figure 1, and in Figures A1 and A2 (online only). OS, DFS, and DSD by cluster size are reported in Table A7 (online only), in Figure 2, and in Figures A3 and A4 (online only). On the basis of the relatively small number of total cases (368) and of patients converted to node positive, (83) 95% CI are wide, but there is a clear trend toward worse outcomes and especially disease-specific outcomes (DFS and DSD), with increasing lymph node tumor burden. www.jco.org 2008 by American Society of Clinical Oncology 1805

Tan et al Variable Table 3. Univariate Analysis of DFS 15-Year DFS (%) 95% CI RR 95% CI P Age, years 50 73 64 to 80 1.0.32 50 78 72 to 83 0.8 0.5 to 1.2 Tumor type Ductal 78 73 to 82 1.0.08 Lobular 66 49 to 78 1.6 0.9 to 2.8 Grade I 95 85 to 98 1.0.002 II 69 60 to 77 6.5 2.0 to 21.2 III 76 67 to 83 5.1 1.5 to 16.7 Lobular 66 49 to 78 7.6 2.2 to 25.9 LVI Absent 80 74 to 84 1.0.001 Present 56 41 to 69 2.2 1.4 to 3.6 1.0 87 76 to 93 1.0.01 1.1-2.0 78 71 to 84 1.7 0.8 to 3.7 2.0 68 59 to 76 2.7 1.3 to 5.9 Converted to node positive No 81 76 to 85 1.0.001 Yes 59 48 to 70 2.3 1.5 to 3.6 Pattern of staining IHC /H&E 81 76 to 85 1.0.001 IHC /H&E 66 50 to 78 1.9 1.1 to 3.3 IHC /H&E 49 31 to 65 3.1 1.8 to 5.4 No. of nodes positive 0 81 75to85 1.0.001 1 58 44 to 70 2.4 1.5 to 3.8 1 64 41 to 80 2.2 1.0 to 4.6 No. of metastatic cells 0 81 76to86 1.0.001 1-20 60 34 to 78 2.0 0.9 to 4.2 21-100 82 55 to 94 1.0 0.3 to 3.2 100 50 35 to 65 3.1 1.9 to 5.2 Largest cluster size Negative 81 76 to 85 1.0.001 0.2 mm 64 50 to 75 2.0 1.2 to 3.3 0.3-2.0 mm 41 19 to 63 3.9 2.0 to 7.8 2.0 mm 75 13 to 96 1.2 0.2 to 8.8 Abbreviations: DFS, disease-free survival; RR, relative risk; LVI, lymphovascular invasion; IHC, immunohistochemistry; H&E, hematoxylin-eosin. n 5. DISCUSSION In breast cancer, conventional histopathologic assessment of lymph nodes consists of a single H&E-stained section per node. This technique, still endorsed by the College of American Pathologists as standard care, 18 has long been recognized as flawed. Dowlatshahi et al 19 have reviewed 25 separate articles spanning a 50-year period in which negative axillary lymph nodes were re-sectioned to determine the prognostic significance of missed nodal metastases. Although the investigators used various methods of serial sectioning and/or IHC staining, 24 of 25 studies identified lymph node metastases in 7% to 42% of patients initially staged as node negative. Although many of the studies were small and found no prognostic significance for lymph node micrometastases, six of the seven studies with more than 100 patients found DFS and/or OS to be significantly worse among patients with occult nodal disease. In the largest of these (n 921), the Ludwig investigators used H&E-stained serial sections to identify occult metastases in 9% of cases and demonstrated significantly worse DFS and OS at 5 years, 20 and in a detailed reanalysis, at 10 years as well. 21 Unfortunately, the first Ludwig study required the examination of about 1,600 slides to identify each additional node-positive patient, an impossible workload for the routine examination of an ALND specimen containing 15 to 20 nodes. With the advent of SLN biopsy, and a pathologic specimen containing a median of two to three nodes, enhanced pathologic examination became feasible for the first time and seemed advantageous for a number of reasons. First, enhanced pathology has allowed validation of the SLN hypothesis; in elegant complementary studies, Turner 6 demonstrated that a negative SLN is highly predictive that the remaining nodes are negative, whereas Weaver 7 showed that the SLN is the node most likely to be positive. Second, enhanced pathology seemed to increase staging accuracy by identifying additional node-positive patients within each category of tumor size. 8 Third, Liberman 9 showed that the false-negative rate of SLN biopsy was significantly lower for those validation studies that used IHC staining, compared with those which did not (8% v 3%, respectively). Finally, enhanced pathology allowed more accurate prediction of non-sln status 10 ; for patients whose SLN were positive only on IHC (compared with patients whose SLN was positive on routine H&E), the chance of finding additional positive nodes on completion ALND was substantially smaller. This information has been incorporated into a multivariate nomogram by Van Zee et al, 10 and has allowed a growing number of SLN-positive patients in our own practice to avoid ALND altogether. 22 Enhanced pathologic examination of SLN has also become controversial. In the absence of confirmatory level 1 evidence, are the so-called advantages of pathologic upstaging and reduced falsenegative rate clinically relevant? Most importantly, are SLN micrometastases, especially those detected only by IHC or as isolated tumor cells, prognostically significant? The present study addresses this issue by drawing on a historic cohort of node-negative breast cancer patients: all had radical surgery, none had systemic adjuvant therapy, and median follow-up was 17.6 years. First, we demonstrate the feasibility of our pathologic technique. 13,14 For patients with negative SLN frozen section, five additional slides are generated per SLN (one frozen section control, plus one H&E and one IHC-stained section from each of two levels 50 m apart). There is at present no consensus method of SLN examination, and the spectrum ranges from routine single-section H&E (as recommended by the College of American Pathologists) 18 to exhaustive frozen section in which the SLN is examined in its entirety (entailing 30 or more sections). 23 Our method represents a deliberate compromise between these two extremes, and has proven suitable for multiple surgeons and pathologists in our high-volume practice. Second, on serial sectioning/ihc, we found metastases in 23% of patients (83 of 368) originally staged as node negative. Although 40% of these were first seen on H&E, some were first identified by IHC and confirmed by H&E after the fact. Even for those who would dispute the clinical relevance of SLN metastases detected by IHC, the benefit of IHC in helping to detect H&E-positive disease will be apparent to any experienced pathologist and is indisputable. 1806 2008 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Occult Axillary Node Metastases in Breast Cancer Table 4. Multivariate Analysis of Disease-Free Survival Model 1 Model 2 Model 3 Model 4 Variable RR 95% CI P Variable RR 95% CI P Variable RR 95% CI P Variable RR 95% CI P Staining No. of cells No. of nodes Cluster size IHC /H&E 1.0.001 0 1.0.001 0 1.0.006 Negative 1.0.001 IHC /H&E 1.5 0.9 to 2.7 1-21 1.5 0.7 to 3.3 1 2.2 1.3 to 3.6 0.2 mm 1.7 1.0 to 2.9 IHC /H&E 3.3 1.8 to 6.0 21-100 1.1 0.3 to 3.6 1 1.9 0.9 to 4.0 0.3-2.0 mm 4.2 2.1 to 8.5 100 2.8 1.7 to 4.8 2.0 mm 1.5 0.2 to 11.1 LVI LVI LVI LVI Absent 1.0.002 Absent 1.0.002 Absent 1.0.001 Absent 1.0.001 Present 2.3 1.4 to 3.8 Present 2.3 1.4 to 3.8 Present 2.4 1.4 to 4.0 Present 2.4 1.4 to 4.0 Grade Grade Grade Grade I 1.0.008 I 1.0.009 I 1.0.01 I 1.0.01 II 6.7 2.0 to 22.0 II 6.2 1.9 to 20.3 II 5.9 1.8 to 19.2 II 6.1 1.9 to 20.1 III 5.2 1.5 to 17.9 III 4.7 1.4 to 16.1 III 4.2 1.3 to 14.3 III 4.8 1.4 to 16.1 Lobular 8.4 2.4 to 29.9 Lobular 8.0 2.3 to 28.2 Lobular 7.0 2.0 to 24.3 Lobular 7.7 2.2 to 27.3 1.0 1.0.01 1.0 1.0.02 1.0 1.0.03 1.0 1.0.01 1.1-2.0 0.9 0.4 to 2.0 1.1-2.0 0.9 0.7 to 2.0 1.1-2.0 1.0 0.4 to 2.1 1.1-2.0 0.9 0.4 to 2.0 2.0 1.8 0.8 to 3.9 2.0 1.8 0.8 to 3.7 2.0 1.9 0.9 to 3.9 2.0 1.8 0.8 to 3.9 Abbreviations: RR, relative risk; LVI, lymphovascular invasion; IHC, immunohistochemistry; H&E, hematoxylin-eosin. Third, our data suggest that routine single-section H&E is adequate for the detection of lymph node macrometastases larger than 2 mm (pn1a disease). Among our 368 patients initially classified as node negative, only 1% (five of 368) were found on re-examination to have lymph node macrometastases (Table A1). Serial sectioning/ihc is therefore of value primarily for the detection of micrometastases (0.3 to 2 mm, pn mi ) and submicrometastases ( 0.2 mm, pn0 i ). Fourth, among patients converted to node positive, we identified a single positive node in 71%. These are of course SLN, and this figure closely matches our entire experience of SLN biopsy, in which approximately 70% of patients have had disease limited to the SLN. 24 We also identified skip metastases isolated to level II in 11% of node-positive cases. The term skip metastasis is a misnomer; since these represent cases of direct lymphatic drainage from the breast to axillary level II, skip metastases are, in reality, SLN. Fifth, we identify a higher rate of conversion to nodepositive for invasive lobular than for invasive duct carcinomas (40% v 20%), over-representation of invasive lobular among IHC-detected versus H&E-detected disease (36% v 15%), and over-representation of invasive lobular among patients with single-cell versus clustered metastases (59% v 7%). All are consistent with the loss of e-cadherin expression characteristic of invasive lobular carcinomas, 25,26 and suggest the particular importance of IHC staining for this distinctive entity. Proportion Surviving 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 -IHC/-H&E +IHC/-H&E +IHC/+H&E Proportion Surviving 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Negative 0.2 mm 0.3 2.0 mm 0 5 10 15 20 25 Time (years) No. at risk (# events) -IHC/-H&E 241 (35) 199 (48) 140 (51) 32 (53) +IHC/-H&E 38 (11) 29 (14) 24 (16) 8 (16) +IHC/+H&E 21 (11) 14 (16) 9 (16) 0 (16) Fig 1. Disease-free survival by pattern of lymph node staining, IHC /H&E (n 285) versus IHC /H&E (n 50) versus IHC /H&E (n 33); P.001. IHC, immunohistochemistry; H&E, hematoxylin-eosin. 0 5 10 15 20 25 Time (years) No. at risk (# events) Negative 241 (35) 199 (48) 140 (51) 32 (53) 0.2 mm 46 (14) 34 (19) 28 (21) 8 (21) 0.3 2.0 mm 10 (7) 7 (10) 4 (10) 0 (10) Fig 2. Disease-free survival by largest cluster size, negative (pn0, n 285) versus 0.2 mm (pn0 i,n 61) versus 0.3 to 2.0 mm (pn1 mi,n 17); P.001. Results for cluster size more than 2.0 mm (pn1a, n 5) are not shown. www.jco.org 2008 by American Society of Clinical Oncology 1807

Tan et al Sixth, we examine the site and distribution of nodal metastases. While the site of metastasis was equally likely to be subcapsular or parenchymal, the extent of metastasis varied widely, ranging from one to more than 100 clusters. Among IHC /H&E patients, the number of metastatic cells varied widely, but 98% had pn0 i disease, with a cluster size of no more than 0.2 mm (Table 2). The greatest variation in number of metastatic cells was in patients with no more than 0.2 mm of disease (Table 3), suggesting biologic heterogeneity even within the smallest subcategory of cluster size. Although these observations in general support the most recent AJCC Staging Manual (ed 6), 15 in which nodal metastases are categorized by cluster size, they also illustrate the difficulty of fitting a heterogeneous collection of pathologic findings into a single system of classification, particularly for patients with invasive lobular cancers. Finally, we demonstrate that lymph node micrometastases missed by conventional single-section H&E and detected by our current pathologic protocol for SLN are prognostically significant. These findings are particularly striking for the disease-specific outcome measures DFS and DSD, where every measure of lymph node tumor burden (pattern of staining, number of positive nodes, number of cells, and cluster size) was highly significant on univariate analysis (Tables 3 and A3), and on multivariate analysis retained a significance comparable to, or greater than, tumor size, grade, and LVI (Tables 4 and A5). The prognosis of patients with micrometastases detected by IHC was worse, and the prognosis of those detected by H&E was substantially worse, than that of patients who remained node negative. These differences were most apparent at 10 or more years of follow-up. Our study has some limitations. The design is retrospective. The primary tumor slides were missing in 20% (and the axillary tissue blocks in 30%) of eligible patients. The breast cancers of the 1970s were largely symptomatic, while present-day lesions are increasingly screen-detected at an earlier stage. Despite excellent long-term followup, our results are limited by relatively small numbers of node-positive patients (50 detected by H&E and 33 by IHC). The data reflect our own pathology protocol, and may not apply for institutions which use different methods. Finally, none of our patients received systemic therapy. Our data nevertheless clearly establish the value of our current SLN pathology protocol (serial sections and IHC staining) for reexamining archival material for lymph node metastases missed by conventional single-section H&E, and demonstrate the prognostic significance of those metastases, whether detected by H&E or IHC. We wish to emphasize that our data cannot address the prognostic significance of SLN micrometastases (and especially of pn0 i disease 0.2 mm) in present-day patients, a more favorable group who virtually all receive systemic adjuvant therapy. For current patients who are staged and treated according to routine H&E examination of their SLN, does the information gained from serial sections/ihc either alter prognosis or change treatment? Two prospective trials, American College of Surgeons Oncology Group (ACOSOG) Z0010 27 and National Surgical Adjuvant Breast and Bowel Project B-32, 28 aim to answer this question through a design in which all therapy is given based on routine H&E examination of the SLN, blinding both clinician and patient to the results of serial sections/ihc; if the presence of occult SLN metastases does not affect outcome, then serial sections/ihc are not worthwhile. Each trial is fully accrued at more than 5,000 patients, but will require years to determine whether the information added by enhanced pathologic examination of SLN will confer any advantage. In the meantime, for all of the reasons discussed herein, we would argue that the benefits of SLN examination by serial sections and IHC far outweigh the risks. AUTHORS DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The author(s) indicated no potential conflicts of interest. AUTHOR CONTRIBUTIONS Conception and design: Lee K. Tan, Dilip Giri, Amanda J. Hummer, Katherine S. Panageas, Patrick I. Borgen, Hiram S. Cody III Financial support: Larry Norton, Clifford Hudis, Patrick I. Borgen, Hiram S. Cody III Administrative support: Lee K. Tan, Edi Brogi, Larry Norton, Clifford Hudis, Patrick I. Borgen, Hiram S. Cody III Provision of study materials or patients: Lee K. Tan, Edi Brogi Collection and assembly of data: Lee K. Tan, Dilip Giri, Amanda J. Hummer, Katherine S. Panageas, Edi Brogi, Patrick I. Borgen, Hiram S. Cody III Data analysis and interpretation: Lee K. Tan, Dilip Giri, Amanda J. Hummer, Katherine S. Panageas, Edi Brogi, Patrick I. Borgen, Hiram S. Cody III Manuscript writing: Lee K. Tan, Dilip Giri, Amanda J. Hummer, Katherine S. Panageas, Hiram S. Cody III Final approval of manuscript: Lee K. Tan, Dilip Giri, Amanda J. Hummer, Katherine S. Panageas, Edi Brogi, Larry Norton, Clifford Hudis, Patrick I. Borgen, Hiram S. Cody III REFERENCES 1. Kim T, Giuliano AE, Lyman GH: Lymphatic mapping and sentinel lymph node biopsy in earlystage breast carcinoma. Cancer 106:4-16, 2006 2. Veronesi U, Paganelli G, Viale G, et al: A Randomized comparison of sentinel-node biopsy with routine axillary dissection in breast cancer. N Engl J Med 349:546-553, 2003 3. Krag DN, Julian TB, Harlow SP, et al: NSABP- 32: Phase III, randomized trial comparing axillary resection with sentinel lymph node dissection: A description of the trial. Ann Surg Oncol 11:208S- 210S, 2004 4. Mansel RE, Fallowfield L, Kissin M, et al: Randomized multicenter trial of sentinel node biopsy versus standard axillary treatment in operable breast cancer: The ALMANAC Trial. J Natl Cancer Inst 98:599-609, 2006 5. Purushotham AD, Upponi S, Klevesath MB, et al: Morbidity after sentinel lymph node biopsy in primary breast cancer: Results from a randomized controlled trial. J Clin Oncol 23:4312-4321, 2005 6. Turner RR, Ollila DW, Krasne DL, et al: Histologic validation of the sentinel lymph node hypothesis for breast carcinoma. Ann Surg 226:271-278, 1997 7. Weaver DL, Krag DN, Ashikaga T, et al: Pathologic analysis of sentinel and nonsentinel lymph nodes in breast carcinoma: A multicenter study. Cancer 88:1099-1107, 2000 8. Giuliano AE, Dale PS, Turner RR, et al: Improved staging of breast cancer with sentinel lymphadenectomy. Ann Surg 222:394-401, 1995 9. Liberman L: Pathologic analysis of sentinel lymph nodes in breast carcinoma. Cancer 88:971-977, 2000 10. Van Zee KJ, Manasseh DM, Bevilacqua JL, et al: A nomogram for predicting the likelihood of additional nodal metastases in breast cancer patients with a positive sentinel node biopsy. Ann Surg Oncol 10:1140-1151, 2003 11. Hermanek P, Hutter RVP, Sobin LH, et al: Classification of isolated tumor cells and micrometastasis. Cancer 86:2668-2673, 1999 12. Elston CW, Ellis IO: Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: Experience from a large study with long-term follow-up. Histopathology 19: 403-410, 1991 13. Cody HS, Borgen PI: State-of-the-art approaches to sentinel node biopsy for breast cancer: 1808 2008 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Occult Axillary Node Metastases in Breast Cancer Study design, patient selection, technique, and quality control at Memorial Sloan-Kettering Cancer Center. Surg Oncol 8:85-91, 1999 14. Brogi E, Torres-Matundan E, Tan LK, et al: The results of frozen section, touch preparation, and cytological smear are comparable for intraoperative examination of sentinel lymph nodes: A study in 133 breast cancer patients. Ann Surg Oncol 12:173-180, 2005 15. Greene FL, Page DL, Fleming ID, et al: Breast, in Greene FL, Page DL, Fleming ID, et al (eds): AJCC Cancer Staging Manual. New York, NY, Springer, 2002, pp 221-240 16. Gray RJ: A class of k-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 16:1141-1154, 1988 17. Fine JP, Gray RJ: A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 94:496-509, 1999 18. Fitzgibbons PL, Connolly JL, Page DL: Breast. College of American Pathologists Cancer Protocols, 2005:1-26. http://www.cap.org/apps/docs/cancer_ protocols/2005/breast05_pw.pdf 19. Dowlatshahi K, Fan M, Snider HC, et al: Lymph node micrometastases from breast carcinoma: Reviewing the dilemma. Cancer 80:1188-1197, 1997 20. International (Ludwig) Breast Cancer Study Group: Prognostic importance of occult axillary lymph node micrometastases from breast cancers. Lancet 335:1565-1568, 1990 21. Cote RJ, Peterson HF, Chaiwun B, et al: Role of immunohistochemical detection of lymph-node metastases in management of breast cancer. International Breast Cancer Study Group. Lancet 354: 896-900, 1999 22. Park J, Fey JV, Naik AM, et al: A declining rate of completion axillary dissection in sentinel lymph node-positive breast cancer patients is associated with the use of a multivariate nomogram. Ann Surg 245:462-468, 2007 23. Viale G, Bosari S, Mazzarol G, et al: Intraoperative examination of axillary sentinel lymph nodes in breast carcinoma patients. Cancer 85:2433-2438, 1999 24. Naik AM, Fey J, Gemignani M, et al: The risk of axillary relapse after sentinel lymph node biopsy for breast cancer is comparable with that of axillary lymph node dissection: A follow-up study of 4008 procedures. Ann Surg 240:462-468, 2004 25. Moll R, Mitze M, Frixen UH, et al: Differential loss of E-cadherin expression in infiltrating ductal and lobular breast carcinomas. Am J Pathol 143: 1731-1742, 1993 26. Gamallo C, Palacios J, Suarez A, et al: Correlation of E-cadherin expression with differentiation grade and histological type in breast carcinoma. Am J Pathol 142:987-993, 1993 27. Giuliano AE: Z0010: A prognostic study of sentinel node and bone marrow micrometastases in women with clinical T1 or T2 N0 M0 breast cancer. http://www.acosog.org/studies/synopses/z0010_ Synopsis.pdf 28. Krag DN, Anderson SJ, Julian TB, et al: Technical outcomes of sentinel-lymph-node resection and conventional axillary-lymph-node dissection in patients with clinically node-negative breast cancer. Results from NSABP B-32 randomized phase III trial. Lancet Oncol 8:881-888, 2007 Appendix The Appendix is included in the full-text version of this article, available online at www.jco.org. It is not included in the PDF version (via Adobe Reader ). www.jco.org 2008 by American Society of Clinical Oncology 1809