Immunohistochemically detected tumor cells in the sentinel lymph nodes of patients with breast carcinoma

Biologic metastasis or procedural artifact?




Sentinel lymph node (SLN) biopsy is a new standard of care for patients with breast carcinoma, and allows enhanced pathologic analysis with serial sections and immunohistochemical (IHC) staining for cytokeratins to be performed on a routine basis. However, the significance of SLN micrometastases detected only by IHC is uncertain. Are these tumor cells truly markers of metastatic potential, or simply evidence of passive displacement by preoperative instrumentation of the tumor site? Here we evaluate whether the pattern of SLN metastasis in breast carcinoma is related to the degree of manipulation at biopsy before surgery, independently of other known predictors.


Among 4016 consecutive eligible patients with breast carcinoma registered in a prospective SLN database at Memorial Sloan Kettering Cancer Center, we noted patient/tumor characteristics, pathologic status of the SLN (negative, positive by hematoxylin and eosin [H&E], or positive only on IHC), and method of previous biopsy (none, fine-needle aspiration biopsy [FNAB], core needle biopsy, or surgical biopsy).


Multivariate analysis showed that the likelihood of an H&E-positive SLN was significantly associated with lymphovascular invasion, tumor size, tumor type, and tumor location, but not with the method of biopsy. In contrast, the likelihood of finding an SLN positive only on IHC was unassociated with any of the four variables above, but was significantly associated with the method of biopsy. After no previous biopsy, FNAB, core needle biopsy, or surgical biopsy, IHC-positive SLN were present in 1.2%, 3.0%, 3.8%, and 4.6% of patients, respectively (P = 0.002).


These data suggest that the frequency of IHC-positive SLN in patients with breast carcinoma 1) is unrelated to conventional predictors of lymph node positivity, 2) is increased after instrumentation of the tumor site, and 3) is increased approximately proportionate to the degree of manipulation. A proportion of IHC-positive SLN were present before biopsy and therefore less likely to be artifactual. Cancer 2004;100:929–34. © 2004 American Cancer Society.

In the decade since the pioneering reports of Krag et al.1 and Giuliano et al.,2 sentinel lymph node (SLN) biopsy has become a new standard of care for axillary lymph node staging in patients with breast carcinoma. A growing number of studies in the literature validate the SLN hypothesis, evaluate criteria for patient selection, establish feasibility in a wide range of practice settings, and address many technical issues. Most importantly, SLN biopsy allows enhanced pathologic analysis of lymph nodes (with serial sectioning and immunohistochemical [IHC] staining) to be performed on a routine basis. Although this methodology has led to validation of the SLN hypothesis,3 improved staging accuracy,4 a reduction in false-negative results,5 and the prediction of non-SLN status in SLN-positive patients,6, 7 it has also fostered controversy regarding the prognostic significance of lymph node micrometastases, especially in patients with small cell clusters or even single tumor cells detected only by IHC. Rosser8 and Carter et al.9 suggest that these “micrometastases” are not biologic metastases at all, but rather the “benign transport” of breast epithelium by preoperative manipulation of the breast and are devoid of clinical significance. This topic assumes particular importance as SLN biopsy becomes more widespread and as a growing proportion of patients with breast carcinoma (87% in a recent audit from the U.K.10) are diagnosed preoperatively by fine-needle aspiration biopsy (FNAB) or core needle biopsy. The goal of the current study is to determine whether the pattern of SLN metastasis is related, independently of other known predictors, to the degree of tumor manipulation before surgery.


Between September 1, 1996 and March 30, 2002, 4232 SLN biopsy procedures were performed for patients with breast carcinoma at Memorial Sloan-Kettering Cancer Center (New York, NY) and entered in a prospective database. After excluding 135 patients in whom the breast primary tumor was unknown or benign (largely patients who underwent prophylactic mastectomy), 76 patients with the tumor quadrant unknown, and 9 patients with an unknown tumor size, 4012 patients with complete data were analyzed.

Of these, 254 had highly suspicious clinical and/or radiographic findings and had SLN biopsy before the concurrent breast procedure. These patients comprised the control group. The remainder (n = 3758) were biopsied before the SLN biopsy: 506 received an FNAB, 1439 received a core needle biopsy, and 1813 received a surgical biopsy. Standard practice required that the SLN biopsy and definitive breast surgery be performed within 6 weeks of the biopsy procedure.

Our technique of SLN biopsy has been described in detail.11 Briefly, we used a combined dye-isotope mapping technique (intradermally injected unfiltered 99mTc sulfur colloid and intraparenchymally injected isosulfan blue dye), removing all blue, hot, or palpably suspicious lymph nodes. For patients with a negative intraoperative frozen section, all SLN were analyzed with serial sections according to a standard methodology. Two adjacent sections (one stained with hemtoxylin and eosin [H&E] and the other with IHC using the anticytokeratin AE1:AE3) were taken from the paraffin block at each of 2 levels 50 μm apart.

The primary characteristics were tumor type (duct carcinoma in situ [DCIS] vs. invasive), tumor size (T1a,b [0.1–1 cm], T1c [1.1–2 cm], and T2–3 [>2 cm]), lymphovascular invasion (LVI) (absent vs. present), tumor location (upper-outer quadrant vs. other), and type of previous biopsy (none, FNAB, core needle, or surgical).

The SLN findings were classified as negative, H&E positive, or IHC positive (positive only on IHC). The category of H&E positive was inclusive of cases in which the tumor cells were first found by IHC, as long as they could be confirmed by H&E. To be IHC positive, the SLN must have been positive by IHC only and the cells in the SLN were required to have malignant morphology consistent with the primary tumor type. All cases of artifactual or normal cytokeratin staining (including epithelial debris, dendritic cells, and epithelial cell rests) were classified as negative. We did not analyze overall SLN positivity under the presumption that the determinants of H&E and IHC positivity might be different.

Statistical Methods

The univariate significance of differences in SLN positivity was assessed by the Fisher exact test or the chi-square test for binary or categoric covariates, or by the Spearman rank correlation for ordered covariates. Multivariate analysis applied logistic regression to covariates that significantly predicted SLN positivity in the univariate model.


Table 1 compares the control, FNAB, core needle, and surgical biopsy subgroups. As expected, the frequency of LVI and the proportion of T2–3 lesions, invasive tumor type, and positive SLN was significantly higher in patients diagnosed by FNAB, all of whom would have had palpable tumors. It is noteworthy that 45% of all patients were diagnosed by a previous surgical biopsy. This finding reflects the referral pattern of our institution; our own practice is to make a preoperative diagnosis by FNAB or image-guided biopsy (thereby performing a single operation) whenever possible. Patients with DCIS comprised 7% of this series versus approximately 25% of all new breast carcinoma cases. Our policy has been to perform SLN biopsy for DCIS selectively in the subset of patients at the highest risk of having occult invasion, primarily patients with solid masses and/or extensive lesions requiring mastectomy.12

Table 1. Tumor Characteristics by Method of Biopsy
VariablesMethod of biopsyP value
None (n = 258) (%)FNAB (n = 506) (%)Core needle biopsy (n = 1439) (%)Excisional biopsy (n = 1813) (%)
  1. FNAB: fine-needle aspiration biopsy; LVI: lymphovascular invasion; SLN: sentinel lymph node; DCIS: ductal carcinoma in situ; none: no previous biopsy; T: primary tumor.

 Present 49 (19)143 (28) 275 (19) 340 (19)< 0.001
 Absent209 (81)363 (72)1194 (81)1473 (81) 
T size     
 T1a-b104 (40)111 (22) 641 (45) 882 (49) 
 T1c110 (43)221 (44) 516 (36) 662 (37)< 0.001
 T2-3 44 (17)174 (34) 283 (19) 269 (14) 
T location     
 Upper outer177 (69)333 (66)1025 (71)1245 (69)0.12
 Other 81 (31)173 (34) 414 (29) 568 (31) 
T type     
 Invasive239 (93)490 (97)1326 (92)1664 (92)0.001
 DCIS 19 (7) 16 (3) 113 (8) 149 (8) 
SLN status     
 Positive 69 (27) 181 (36) 401 (28) 533 (29)0.006
 Negative189 (73)325 (64)1038 (72)1280 (71) 

Tables 2 and 3 compare rates of SLN positivity by H&E (Table 2) and by IHC only (Table 3) using both univariate and multivariate analyses. Table 2 shows that although SLN were significantly more likely to be H&E positive in the FNAB subgroup than in the control, core needle biopsy, or surgical biopsy subgroups (reflecting palpability bias), the method of biopsy was unrelated to H&E positivity after correcting for other known predictors, all of which remained jointly significant, in the multivariate analysis. Although logistic regression allows prediction of the likelihood of H&E positivity by multiplying odds ratios (according to the formulas (absolute odds) = (relative odds) × (baseline odds) and P = (odds)/(1 + odds) × 100%), and the statistical significance of these results is very strong (P < 0.0001), the predictive accuracy as measured by sensitivity and specificity is only moderate.

Table 2. SLN Positive on H&E
CharacteristicsH&E+/total 1045/4016 (26%)Univariate P valueMultivariate logistic regression Odds ratio (95% CI)Multivariate P value
  1. SLN: sentinel lymph node; H&E: hematoxylin and eosin; CI: confidence interval; FNAB: fine-needle aspiration biopsy; LVI: lymphovascular invasion; DCIS: ductal carcinoma in situ; UOQ: upper outer quadrant.

 Absent 598/3212 (19%)< 0.00011.0< 0.0001
 Present 427/804 (53%) 3.52 (2.96–4.19) 
 T1a-b 224/1740 (13%)< 0.00011.0< 0.0001
 T1c 446/1509 (30%) 2.08 (1.72–2.52) 
 T2-3 355/767 (46%) 3.68 (2.96–4.58) 
 Invasive1007/3719 (27%)< 0.00011.00.002
 DCIS  18/297 (6%) 0.46 (0.28–0.75) 
 Other quadrant 273/1236 (22%)0.0011.00.003
 UOQ 752/2780 (27%) 1.29 (1.09–1.53) 
Biopsy type    
 None  66/258 (26%)0.0031.000.55
 FNAB 164/506 (32%) 1.00 (0.69–1.44) 
 Core needle 345/1439 (24%) 0.92 (0.66–1.27) 
 Excision biopsy 450/1813 (25%) 1.04 (0.75–1.44) 
Table 3. SLN Positive only on IHC
CharacteristicsIHC+/total 156/4016 (3.9%)Univariate P valueLogistic regression Odds ratio (95% CI)
  • SLN: sentinel lymph node; IHC: immunohistochemistry; CI: confidence interval; FNAB: fine-needle aspiration biopsy; LVI: lymphovascular invasion; DCIS: ductal carcinoma in situ; UOQ: upper outer quadrant.

  • a

    As an ordered variable.

  • b

    As a categoric variable.

 Absent128/3212 (4%)> 0.4
 Present 28/804 (3.5%)  
 T1a+b 59/1740 (3.4%)> 0.5
 T1c 63/1509 (4.0%)  
 T2 30/631 (4.8%)  
 T3  6/136 (4.4%)  
 Invasive146/3719 (3.9%)> 0.9
 DCIS 10/297 (3.4%)  
 Other quadrant 45/1236 (3.6%)> 0.5
 UOQ111/2780 (4.0%)  
Biopsy type   
 None  3/258 (1.2%)0.002a1.0
 FNAB 15/506 (3.0%)0.014b2.62 (0.75–9.1)
 Core needle 54/1439 (3.8%) 3.32 (1.03–10.7)
 Excision biopsy 84/1813 (4.6%) 4.36 (1.37–13.9)

The results in Table 3 contrast sharply with those in Table 2. IHC positivity was unrelated to the presence of LVI or to tumor size, type, and location. The proportion of IHC positives was strikingly similar in all of these subgroups and even for the invasive and high-risk patients with DCIS (i.e., 3.9% and 3.4%, respectively). However, IHC positivity was significantly related to biopsy type: 1.2% of patients with no previous biopsy, 3.0% after FNAB, 3.6% after core needle biopsy, and 4.6% after surgical biopsy had positive SLN only on IHC (P = 0.002 as an ordered variable and P = 0.014 as a categoric variable).


Three major randomized trials13–15 have reported 20–25-year results, supporting the hypothesis of Fisher16 that long-term survival in patients with localized breast carcinoma is largely governed by the presence of occult distant metastases and not by variations in local treatment. As a result, surgical treatment has become more conservative and systemic adjuvant therapy more widespread. In this setting, will increased scrutiny for occult metastases better identify high-risk patients or will it simply lead to overtreatment?

An extensive literature search17 spanning 50 years documents that more comprehensive pathologic methods (using serial sectioning and/or IHC staining for cytokeratins) will find axillary lymph node metastases in 7–42% of patients with breast carcinoma initially diagnosed as lymph node negative and that these missed metastases may be prognostically significant. While such comprehensive analysis was not logistically feasible in the era of lymph node staging by axillary lymph node dissection (the 1990 Ludwig study,18 using H&E-stained serial sections, required the evaluation of approximately 1600 slides to identify each additional lymph node-positive patient), SLN biopsy, which removes a median of 2–3 lymph nodes, allows comprehensive pathologic analysis on a routine basis and has reactivated debate regarding the significance of lymph node micrometastases in patients with breast carcinoma.

Lymph node metastases in patients with breast carcinoma represent a broad spectrum of pathologic findings (ranging from macroscopic disease, to H&E-detected macrometastases > 2 mm, to H&E and/or IHC-detected micrometastases ≤ 2 mm, to micrometastases ≤ 0.2 mm, to single tumor cells found only by IHC). These gradations almost certainly reflect a spectrum of prognostic significance as well and have been incorporated into the new American Joint Committee on Cancer staging system.19 At one extreme, the 518 and 10-year20 results of the Ludwig study clearly showed that missed lymph node metastases found by H&E staining are indeed prognostically significant, as does a large study of our own21 in which the 10-year disease-free survival was approximately 30% worse in patients with missed metastases found by H&E compared with patients with negative lymph nodes. At the other extreme, the significance of single cells detected by IHC is unknown.22 Conventional chemotherapy in this setting may represent overtreatment. This heterogeneity among micrometastases will confound the interpretation of clinical trials, two of which currently aim to resolve the significance of IHC-detected micrometastases by blinding both patient and clinician to the results of IHC staining.23, 24

Some argue that IHC-detected micrometastases may not be metastases at all. Rosser8 has suggested that manipulation of the breast either during mammography or at surgery could mechanically disperse tumor cells and that such manipulation explains the IHC-positive SLN. Youngson et al.25, 26 have shown that core biopsy can physically displace DCIS into the breast stroma and into lymphatic vessels, artifactually simulating invasion. Finally, Carter et al.9 have reported a series of 15 pathologic consultation cases of “benign transport” in which axillary lymph nodes were sampled after an initial biopsy procedure and found to contain not only tumor cells but also evidence of a traumatic origin (e.g., hemosiderin-laden macrophages, multinucleated giant cells, red blood cells, and/or lymphocytes). They caution that without histologic evidence of trauma, these cells must be assumed to be metastatic.

Our results indicate that the frequency of an IHC-positive SLN is proportional to the degree of preoperative tumor manipulation. This observation is particularly significant because method of biopsy is unrelated to H&E positivity (after correcting for conventional predictors) (Table 2) and conventional predictors of H&E positivity are unassociated with IHC positivity (Table 3). The 1.2% background rate of IHC positivity in controls not previously biopsied (Table 3) suggests that epithelial displacement may account for many IHC-positive SLN and true metastasis for the remainder.

Substantial caveats apply to this interpretation. First, we rarely observe evidence of a traumatic origin as described by Carter et al.9 in our own IHC-positive patients (unpublished data). Second, among patients having SLN biopsy at the time of prophylactic mastectomy, we have not observed any IHC-positive SLN among 41 patients with benign breast pathology and a history of recent breast instrumentation.27 Third, the differences in IHC positivity between biopsy methods are quite small, and, although statistically significant, may lack clinical significance. Fourth, the IHC-positive group comprises only 3.9% of all patients and a minority (14%) of lymph node-positive patients. The majority of lymph node-positive patients are found by H&E and are conventional candidates for systemic therapy. Fifth, IHC-positive cells in the bone marrow of patients with breast carcinoma have an adverse prognostic significance equal to or exceeding that of axillary lymph node metastasis.28, 29 Finally, when IHC-positive tumor cells are found in an SLN (with or without evidence of a traumatic origin) there is no definitive way to distinguish between epithelial displacement and true metastasis.

Although our study suggests that a proportion, and perhaps a minority, of IHC-positive SLN are true metastases and that the remainder represent epithelial displacement, this observation can only be regarded as preliminary. Pending the results of clinical trials currently in progress,23, 24 the treatment of patients with an IHC-positive SLN will remain a matter of clinical judgment, balancing proven evidence, uncertain evidence, and the preferences of the patient. As we stand on the threshold of molecular profiling of breast carcinoma,30, 31 our understanding of this heterogeneous disease remains a tantalizing mixture of both mechanistic and biologic considerations.