Human Pathology (2015) 46, 246–254
www.elsevier.com/locate/humpath
Original contribution
A histomorphologic predictive model for axillary lymph node metastasis in preoperative breast cancer core needle biopsy according to intrinsic subtypes☆,☆☆ Su Hyun Yoo MD a , In Ae Park MD, PhD b , Yul Ri Chung MD b , Hyojin Kim MD b , Keehwan Lee MD b , Dong-Young Noh MD, PhD c , Seock-Ah Im MD, PhD d , Wonshik Han MD, PhD c , Hyeong-Gon Moon MD, PhD c , Kyung-Hun Lee MD d , Han Suk Ryu MD, PhD b,⁎ a
Department of Pathology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, Republic of Korea Department of Pathology, Seoul National University Hospital, Seoul, Republic of Korea c Department of Surgery, Seoul National University Hospital, Seoul, Republic of Korea d Department of Internal Medicine, Seoul National University Hospital, Seoul, Republic of Korea b
Received 13 August 2014; revised 6 October 2014; accepted 31 October 2014
Keywords: Breast cancer; Lymph node metastasis; Predictive model; Histopathology; Nomogram
Summary The aim of this study is construction of a pathologic nomogram that can predict axillary lymph node metastasis (LNM) for each intrinsic subtype of breast cancer with regard to histologic characteristics in breast core needle biopsy (CNB) for use in routine practice. A total of 534 CNBs with invasive ductal carcinoma classified into 5 intrinsic subtypes were enrolled. Eighteen clinicopathological characteristics and 8 molecular markers used in CNB were evaluated for construction of the best predictive model of LNM. In addition to conventional parameters including tumor multiplicity (P b .001), tumor size (P b .001), high histologic grade (P = .035), and lymphatic invasion (P = .017), micropapillary structure (P b .001), the presence of small cell–like crush artifact (P = .001), and overexpression of HER2 (P = .090) and p53 (P = .087) were proven to be independent predictive factors for LNM. A combination of 8 statistically independent parameters yielded the strongest predictive performance with an area under the curve of 0.760 for LNM. A combination of 6 independent variables, including tumor number, tumor size, histologic grade, lymphatic invasion, micropapillary structure, and small cell–like crush artifact produced the best predictive performance for LNM in luminal A intrinsic subtype (area under the curve, 0.791). Thus, adding these combinations of clinical and morphologic parameters in preoperative CNB is expected to enhance the accuracy of prediction of LNM in breast cancer, which might serve as another valuable tool in determining optimal surgical strategies for breast cancer patients. © 2015 Elsevier Inc. All rights reserved.
☆
Competing interests: The authors declare that they have no conflict of interests. Funding/Support: This study was supported by grant number 04-2013-0840 from Seoul National University Hospital Research Fund. ⁎ Corresponding author at: Department of Pathology, Seoul National University Hospital, Seoul National University College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul, 110-744, Republic of Korea. E-mail address: [email protected] (H. S. Ryu).
☆☆
http://dx.doi.org/10.1016/j.humpath.2014.10.017 0046-8177/© 2015 Elsevier Inc. All rights reserved.
Breast biopsy and lymph node metastasis
1. Introduction The presence of axillary lymph node metastasis (LNM) is crucial in predicting clinical outcomes in breast cancer [1,2]. Although axillary lymph node dissection in patients with nodal metastasis is a standard procedure, it carries the risk of complications such as pain, loss of sensation, swelling, infection, impaired shoulder mobility, and lymphedema [3]. Although sentinel lymph node biopsy serves an optimal choice to selectively determine further axillary lymph node dissection, it has been reported to have a 5% false-negative rate in node-positive patients in breast cancer [4]. This falsenegative rate increased when only 1 sentinel lymph node was removed, thereby arousing debates as to how many lymph nodes should be removed [5]. Therefore, if there were certain ancillary methods to predict LNM, it might be helpful in reducing the false-negative rate as well as aid in making more accurate surgical decisions. Several predictive models for LNM have been introduced, many of which are being widely accepted for planning optimal surgical intervention in breast cancer patients [6,7]. Nevertheless, a predictive model or a pathologic nomogram for preoperative biopsy specimens incorporating characteristic histomorphologic parameters has not yet been evaluated thus far. Core needle biopsy (CNB) is an economic, valuable, and easily accessible preoperative diagnostic procedure used in daily practice. Although recent advances in diagnostic tools include molecular/genetic methods such as the nextgeneration sequencing, their application in daily diagnostic processes is extremely limited due to high costs. In this study, we examined the utility of preoperative CNB not only as a diagnostic tool but also as a predictor of LNM in breast cancer patients. In addition to the distinct morphologic characteristics of micropapillary structures and retraction tissue artifact, which are already known morphologic parameters related with LNM [8], we also adopted the small cell–like crush artifact observed in biopsy samples and constructed a model with the best predictive performance for LNM with a combination of these morphologic and other clinicopathological parameters found significant in breast cancer.
2. Materials and methods 2.1. Case selection Cases to be included in the study were selected from the pathologic archives of Seoul National University Hospital: we identified 5810 cases with CNB specimens over a 3-year period (January 2010 to December 2012). Among those patients, 534 cases fulfilling the following criteria were enrolled: (1) those who received axillary lymph node dissection (510 cases, 95.5% underwent sentinel lymph node biopsy) in subsequent surgical excision allowing evaluation of cancer metastasis; (2) those with the final
247 diagnosis of invasive ductal carcinoma; (3) CNB samples with at least 4 well-preserved cores; and (4) cases that had been grouped into intrinsic subtypes with immunostains. Patients who received neoadjuvant chemotherapy after undergoing CNB were excluded for precise evaluation of LNM. All the cases in the study underwent surgical excision. Thus, although we were able to guess each cancer to be either a solitary nodule or having multiple nodules through imaging studies before surgery, their solitary/multiplicity status was confirmed and changed according to the postoperative pathologic evaluations. Biopsy was performed using either an 11- or 14-gauge automated needle–assisted or with an 8- or 10-gauge vacuum–assisted device. All specimens fixed in 10% buffered formalin were paraffin embedded and hematoxylin and eosin (H&E) stained. To improve the accuracy of the results, 2 pathologists (S. H. Y. and H. S. R.) who were blinded to the clinical details of the patients reviewed histomorphologic findings. The patients' clinicopathological information was obtained from the patient database. This study was approved by the Institutional Review Board of Seoul National University Hospital.
2.2. Histomorphologic evaluations The following histologic factors were evaluated: micropapillary structure, retraction tissue artifact, small cell–like crush artifact of tumor cells, cribriform architecture, solid sheet growth showing syncytial arrangements of tumor cells with no tubule formation, spindling or multinucleated giant cells, presence of extracellular mucin, intratumoral calcification, tumor necrosis, loose myxoid fibrous stroma found in early phases of fibroadenoma, lymphatic invasion, neural invasion, and ductal carcinoma in situ (DCIS) component. Histologic grading was categorized into 1 to 3 based on Nottingham combined histologic grading system described by Elston and Ellis [9]. Micropapillary structure was defined as a nest of tumor cells forming discrete tubular or solid structures floating in the middle of an empty space (detached from the stroma) with or without immunohistochemical positivity for epithelial membrane antigen [10,11] (Fig. 1A and D). The presence of micropapillary structure was defined when more than 50% of the structure was identified in the entire core. Retraction tissue artifact was defined as the narrow cleft between the tumor glands and the surrounding stroma [12,13] (Fig. 1B and E). A tumor was considered to have a retraction tissue artifact when such histomorphologic finding occupied greater than or equal to 30% of the entire core biopsied. Small cell–like crush artifact was defined as infiltrative tumor cells with small-sized hyperchromatic nuclei and scanty cytoplasm resulting in a high nuclear/cytoplasmic ratio and crush artifact [14]. The tumor cells were distributed as either infiltrative cells showing streaming pattern or conglomerated clusters without tubule formation accompanied by severe crush artifact (Fig. 1C and F). Such small
248
S. H. Yoo et al.
Fig. 1 A and D, Micropapillary structures showing tumor nests devoid of fibrovascular cores, floating in empty stromal spaces. B and E, Retraction tissue artifact with nest of tumor cells being separated from the surrounding stroma by a clear space. C and F, Small and dark-stained cells with crushing artifact, designated “small cell–like crush artifact,” are infiltrating into the stroma. H&E stain, original magnification ×100 (A-C) and ×200 (D-F).
cell–like crush artifact was considered to be present when the feature was identified in greater than or equal to 10% of the entire core. Mucinous component and spindle or multinucleated giant cell differentiation were considered to be present if found in greater than or equal to 5% of the given cores. To assess lymphatic invasion of the tumor cells, immunohistochemical staining for D2-40 (1:100 dilution; DakoCytomation, Carpinteria, CA) was performed. The D240 immunostaining was only done for cases that were suspicious for lymphatic invasion or required discrimination from retraction artifact.
2.3. Immunohistochemistry Immunohistochemical staining was performed for subtyping of the breast cancers. Formalin-fixed, paraffinembedded tissue sections of all included patients were immunohistochemically stained using a Benchmark automatic immunostaining device (Ventana Medical System, Tucson, AZ). The samples were incubated with antibodies to estrogen receptors (ERs, 1:100, 1D5; Novocastra Laboratories, Newcastle, UK), progesterone receptors (PRs, 1:200, PgR636; Dako, Glostrup, Denmark), HER2 (Ventana), bcl-2 (1:200; Ventana), p53 (1:500, DO-7; Dako), epidermal growth factor receptor (EGFR) (Ventana), cytokeratin 5/6 (CK5/6, 1:50; Ventana), and Ki-67 (1:100; Dako). The sections were subsequently incubated with biotinylated antimouse immunoglobulins, peroxidase-labeled streptavidin (LSAB kit; Dako), and 3,3′-diaminobenzidine. Slides
were counterstained using Harris hematoxylin. Nuclear expression of tumor cells was interpreted as positive for ER, PR, bcl-2, and p53, whereas membrane staining of tumor cells was considered positive for HER2, EGFR, and CK5/6. Immunohistochemical staining was evaluated based on the location and percentage or intensity of positively stained cells. Immunohistochemical staining for ER and PR expression was counted and categorized as positive when greater than or equal to 1% of the tumor cells were stained according to the 2010 American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines [15]. Immunohistochemical expression of HER2 was assessed based on 2013 ASCO/CAP guidelines [16]. p53 and bcl-2 expressions were regarded as positive if greater than or equal to 25% of the tumor cells showed nuclear staining [17]. CK5/6 and EGFR (HER1) were scored positive when more than 10% of cytoplasmic staining was observed in the tumor cells [18,19]. The Ki-67 labeling index was counted as a percentage of immunopositivity with a cutoff value of 14% [20]. All of the immunohistochemical stainings were reviewed by 2 experienced breast pathologists (H. S. R. and I. A. P.) to improve accuracy.
2.4. Fluorescence in situ hybridization assays for HER2 gene amplification Presence of HER2 amplification was determined with an additional fluorescence in situ hybridization assay using PathVysion assay (Abbott Molecular, Downers Grove, IL) in cases with equivocal HER2 protein expression (IHC 2+) on
Breast biopsy and lymph node metastasis Table 1
249
Comparisons of continuous and categorical variables using the Student t test and χ2 test
Variables
Age (y, mean ± SD) Overall necrosis (%, mean ± SD) DCIS component (%, mean ± SD) Loose fibrous stroma (%, mean ± SD) ER expression (%, mean ± SD) PR expression (%, mean ± SD) Tumor number Single Multiple Tumor size b2 cm ≥2 cm Histologic grade of invasive carcinoma Absent Present Nuclear grade of DCIS Absent Present Lymphatic invasion Absent Present Neural invasion Absent Present Micropapillary structure Absent Present Retraction tissue artifact Absent Present Small cell–like crush artifact Absent Present Cribriform architecture Absent Present Solid sheet growth Absent Present Spindle or multinucleated giant cell Absent Present Mucinous component Absent Present Intratumoral calcification Absent Present HER2 expression Absent Present Ki-67 labeling index b14% ≥14%
LNM Breast cancer with multiple nodules (n = 534)
Breast cancer with solitary nodule a (n = 437)
No (n = 317)
No (n = 283)
53.86 ± 9.46 2.17 ± 8.99 6.56 ± 15.53 12.50 ± 18.02 58.04 ± 39.83 29.50 ± 36.63
Yes (n = 217) 52.29 ± 10.95 2.05 ± 7.90 3.99 ± 11.14 11.48 ± 17.00 55.12 ± 42.06 20.59 ± 32.45
P .096 .865 .026 .660 .417 .003
54.17 ± 10.70 1.94 ± 8.36 5.68 ± 14.82 12.29 ± 17.91 59.25 ± 39.63 30.01 ± 36.82
Yes (n = 154) 52.75 ± .9.74 2.18 ± 8.34 3.38 ± 9.05 12.41 ± 18.51 55.73 ± 42.20 18.45 ± 30.53
P .175 .782 .046 .948 .395 .001
284 (64.8%) 33 (34.7%)
155 (35.2%) 62 (65.3%)
b.001
251 (69.5%) 66 (38.2%)
110 (30.5%) 107 (61.8%)
b.001
223 (74.3%) 60 (43.8%)
77 (25.7%) 77 (56.2%)
b.001
186 (72.1%) 131 (47.5%)
72 (27.9%) 145 (52.5%)
b.001
168 (76.4%) 115 (53.0%)
52 (23.6%) 102 (47.0%)
b.001
204 (57.6%) 113 (62.8%)
150 (42.4%) 67 (37.2%)
.252
193 (64.1%) 90 (66.2%)
108 (35.9%) 46 (33.8%)
.677
307 (60.8%) 10 (34.5%)
198 (39.2%) 19 (65.5%)
.005
275 (66.3%) 8 (36.4%)
140 (33.7%) 14 (63.6%)
.004
304 (59.8%) 13 (50.0%)
204 (40.2%) 13 (50.0%)
.319
271 (65.0%) 12 (60.0%)
146 (35.0%) 8 (40.0%)
.648
286 (64.4%) 31 (34.4%)
158 (35.6%) 59 (65.6%)
b.001
255 (68.7%) 28 (42.4%)
116 (31.3%) 38 (57.6%)
b.001
222 (65.3%) 95 (49.0%)
118 (34.7%) 99 (51.0%)
b.001
199 (69.3%) 84 (56.0%)
88 (30.7%) 66 (44.0%)
.006
263 (63.2%) 54 (45.8%)
153 (36.8%) 64 (54.2%)
.001
231 (69.2%) 52 (50.5%)
103 (30.8%) 51 (49.5%)
.001
261 (56.7%) 56 (75.7%)
199 (43.3%) 18 (24.3%)
.002
234 (62.1%) 49 (81.7%)
143 (37.9%) 11 (18.3%)
.003
268 (58.0%) 49 (68.1%)
194 (42.0%) 23 (31.9%)
.106
237 (63.5%) 46 (71.9%)
136 (36.5%) 18 (28.1%)
.197
297 (58.9%) 20 (66.7%)
207 (41.1%) 10 (33.3%)
.402
264 (64.5%) 19 (67.9%)
145 (35.5%) 9 (32.1%)
.723
305 (58.9%) 12 (75.0%)
213 (41.1%) 4 (25.0%)
.301
272 (64.3%) 11 (78.6%)
151 (35.7%) 3 (21.4%)
.396
259 (59.1%) 58 (60.4%)
179 (40.9%) 38 (39.6%)
.876
223 (64.2%) 60 (71.4%)
130 (35.8%) 24 (28.6%)
.579
282 (62.8%) 35 (41.2%)
167 (37.2%) 50 (58.8%)
b.001
256 (68.3%) 27 (43.5%)
119 (31.7%) 35 (56.5%)
b.001
265 (61.3%) 52 (51.0%)
167 (38.7%) 50 (49.0%)
.055
239 (67.3%) 44 (53.7%)
116 (32.7%) 38 (46.3%)
.020
(continued on next page)
250
S. H. Yoo et al.
Table 1 (continued) Variables
CK5/6 expression Absent Present Bcl2 expression Absent Present p53 expression Absent Present EGFR expression Absent Present Total
LNM Breast cancer with multiple nodules (n = 534)
Breast cancer with solitary nodule a (n = 437)
No (n = 317)
Yes (n = 217)
No (n = 283)
Yes (n = 154)
297 (58.8%) 20 (69.0%)
208 (41.2%) 9 (31.2%)
.279
266 (64.3%) 17 (73.9%)
148 (35.7%) 6 (26.1%)
.345
73 (50.0%) 244 (62.9%)
73 (50.0%) 144 (37.1%)
.007
59 (52.2%) 224 (69.1%)
54 (47.8%) 100 (30.9%)
.001
212 (64.8%) 105 (50.7%)
115 (35.2%) 102 (49.3%)
.001
191 (70.2%) 92 (55.8%)
81 (29.8%) 73 (44.2%)
.002
290 (60.5%) 27 (49.1%) 317
189 (39.5%) 28 (50.9%) 217
.101
260 (66.2%) 23 (52.3%) 283
133 (33.8%) 21 (47.7%) 154
.067
P
P
Abbreviations: CK, cytokeratin; DCIS, ductal carcinoma in situ; EGFR, epidermal growth factor receptor; ER, estrogen receptor; LNM, lymph node metastasis; PR, progesterone receptor. a Breast cancer excluding multiple nodules.
immunohistochemistry, as previously described. Interpretation for HER2 gene amplification by fluorescence in situ hybridization assay was confirmed based on the greater than or equal to 2.0 ratio of HER2 gene copy number to chromosome 17 copy number or greater than or equal to 6.0 average HER2 signal per tumor cell as defined in the 2013 ASCO/CAP guidelines [16].
were independent predictors of LNM. P b .05 was considered statistically significant except in backward selection methods for which P b .2 was considered statistically significant. Statistical analyses were performed using the SPSS software (SPSS Standard version, 17.0; SPSS, Chicago, IL).
2.5. Definition of breast cancer immunohistochemical intrinsic subtypes
3. Results
Intrinsic subtypes of breast cancer were defined using immunohistochemical profiles in accordance with previous studies [21-23]. The 5 immunohistochemical phenotypes were as follows: luminal A (ER+ and/or PR+, HER2−, Ki-67 b14%), luminal B with HER2 negative (ER+ and/or PR+, HER2−, Ki67 ≥14%), luminal B with HER2 positive (ER+ and/or PR+, HER2+, any Ki-67), HER2 enriched (ER−, PR−, HER2+), and basal-like (ER−, PR−, HER2−, CK5/6+ and/or EGFR+).
2.6. Statistical analyses The χ2 and Fisher exact tests were used to analyze the statistical significance of the association between LNM status and categorical parameters. The Student t test was used to compare the continuous variables. Logistic regression analysis was performed to evaluate influential risk factors for LNM. To identify factors that were independently predictive of the risk of LNM, multivariate logistic regression analysis was performed with backward selection methods to avoid overfitting. Receiver operating characteristic (ROC) curves were constructed, and areas under the curves (AUCs) were calculated to determine optimal cutoff values for all possible combinations of the variables that
3.1. Clinicopathological characteristics of breast cancer and their correlations with LNM Clinicopathological variables of the patients are as summarized in Table 1. Of the 534 cases with invasive ductal carcinoma, 217 cases (40.6%) had LNM. LNM was associated with the following variables: less amount of DCIS component (P = .026), negative expression of PR (P = .003), tumor multiplicity (P b .001), larger tumor size of more than 2 cm (P b .001), higher histologic grade of invasive carcinoma cells (P b .001), lymphatic invasion (P = .005), micropapillary structure (P b .001), retraction tissue artifact (P b .001), small cell–like crush artifact (P = .001), less cribriform architecture (P = .002), HER2 overexpression (P b .001), bcl-2 overexpression (P = .007), and p53 overexpression (P = .001). The above 13 parameters that had been found to have significant correlations with LNM were also shown to be associated with positive nodal status in univariate analyses (Table 2). When we also evaluated the association of these variables with LNM in breast cancer with a solitary nodule, a higher Ki-67 labeling index (P = .020) along with the other 12 variables that were revealed to have significant correlations in breast cancer with multiple nodules was associated with LNM in solitary breast cancers
Breast biopsy and lymph node metastasis
251
(Tables 1 and 2). We further examined clinicopathological variables correlating with LNM for each intrinsic subtype of breast cancer. Comprehensive results for the predictive variables of LNM in each intrinsic subtype of breast cancer including those with single nodules only are summarized in Supplementary Table 1.
3.2. Predictive models for LNM in breast cancer CNB Of the 13 clinicopathological variables associated with positive nodal status in 534 breast carcinomas, the 8 factors were found to be independent parameters in multivariate logistic regression analysis (Table 3). Using these 7 independent predictive variables for breast cancer with solitary nodule and 8 variables for breast cancer with multiple nodules, prediction models were constructed and calibrated using Hosmer-Lemeshow goodness-of-fit test (P = .462 and P =
Table 2 of LNM Variables
Univariate logistic regression tests for prediction OR
95% CI
Table 3 of LNM
Multivariate logistic regression tests for prediction
Variables
OR
Predictive analysis for breast cancer with Tumor number 3.666 Tumor size 3.308 Histologic grade of 1.592 invasive carcinoma Lymphatic invasion 2.930 Micropapillary structure 2.950 Small cell–like crush artifact 2.273 HER2 expression 1.615 p53 expression 1.438 Predictive analysis for breast cancer with Tumor size 3.489 Lymphatic invasion 3.206 Micropapillary structure 3.019 Small cell–like crush artifact 2.223 HER2 expression 2.428 Ki-67 labeling index 1.698 p53 expression 1.487
95% CI
P
multiple nodules 2.183-6.155 b.001 2.181-5.016 b.001 1.033-2.454 .035 1.210-7.092 .017 1.741-5.009 b.001 1.421-3.635 .001 0.928-2.811 .090 0.948-2.180 .087 solitary nodule 2.220-5.482 b.001 1.193-8.616 .021 1.683-5.413 b.001 1.357-3.643 .002 1.319-4.468 .004 0.973-2.966 .063 0.928-2.381 .099
Abbreviations: CI, confidence interval; LNM, lymph node metastasis; OR, odds ratio.
P
Predictive analysis for breast cancer with multiple nodules DCIS component (%, mean ± SD) 0.985 0.972-0.999 .041 PR immunohistochemical 0.993 0.987-0.998 .004 expression (%, mean ± SD) Tumor number 3.900 2.468-6.165 b.001 Tumor size 3.900 2.468-6.165 b.001 Histologic grade of 2.799 1.962-3.992 b.001 invasive carcinoma Lymphatic invasion 2.841 1.295-6.233 .009 Micropapillary structure 3.430 1.972-5.968 b.001 Retraction tissue artifact 1.908 1.346-2.704 b.001 Small cell–like crush artifact 2.324 1.573-3.435 b.001 Cribriform architecture 0.431 0.245-0.757 .003 HER2 overexpression 2.412 1.504-3.869 b.001 Bcl2 overexpression 0.590 0.402-0.867 .007 p53 expression 1.791 1.256-2.553 .001 Predictive analysis for breast cancer with solitary nodule DCIS component (%, mean ± SD) 0.985 0.967-1.002 .088 PR immunohistochemical 0.990 0.984-0.996 .001 expression (%, mean ± SD) Tumor size 3.717 2.429-5.688 b.001 Histologic grade of 2.094 1.353-3.242 .001 invasive carcinoma Lymphatic invasion 3.437 1.409-8.389 .007 Micropapillary structure 2.983 1.747-5.095 b.001 Retraction tissue artifact 1.777 1.181-2.673 .006 Small cell–like crush artifact 2.200 1.402-3.452 .001 Cribriform architecture 0.367 0.185-0.730 .004 HER2 expression 2.789 1.614-4.819 b.001 Ki-67 labeling index 1.779 1.093-2.897 .020 Bcl2 expression 0.488 0.315-0.756 .001 p53 expression 1.871 1.251-2.798 .002 Abbreviations: CI, confidence interval; DCIS, ductal carcinoma in situ; LNM, lymph node metastasis; OR, odds ratio.
.762), and AUC was 0.735 and 0.760 in breast cancer with solitary and multiple nodules, respectively (Figs. 2 and 3). Using the independent parameters obtained from multivariate logistic regression analyses in each breast carcinoma subtype, predictive models for LNM were evaluated for each subtype (Supplementary Table 2). As a result, 6 independent variables including tumor multiplicity (odds ratio [OR], 6.271; P b .001), increased tumor size (OR, 5.077; P b .001), increased histologic grade of invasive carcinoma (OR, 1.917; P = .030), presence of lymphatic invasion (OR, 3.881; P = .027), micropapillary structures (OR, 2.719; P = .004), and small cell–like crush artifact (OR, 2.057; P = .033) were used to construct a predictive model for LNM luminal A intrinsic subtype, and ROC was constructed from AUC (Hosmer-Lemeshow goodness-of-fit test, P = .697) with combination of 6 variables, which yielded AUC of 0.791. In luminal A intrinsic subtype with single nodules only, the 5 combined predictive models yielded AUC of 0.742 on ROC. Predictive models for luminal B (both HER2− and HER2+ subtypes) and basal-like subtype were not constructed because we were unable to discover independent parameters for these subtypes on statistical analysis.
4. Discussion In this study, we investigated pathologic and molecular features that help predict LNM in preoperative breast cancer biopsy and observed 8 significant predictors composed of 6 clinicopathological and 2 molecular
252
S. H. Yoo et al.
Fig. 2
ROC curve for breast cancer with solitary nodule.
parameters. The AUC using the combination of these 8 features was 0.760, which showed predictive performance comparable to those of the nomograms previously reported [6,7,24]. These previous predictive models were developed using mainly clinical or radiologic parameters with only some conventional pathologic variables such as histologic grade, ER status, and lymphatic invasion [7,24]. In this study, we focused specifically on pathologic parameters, including histomorphologic features and molecular markers, which are known driver genes in carcinogenesis of mammary carcinoma. Three distinct histomorphologic features were evaluated in our predictive model, which are small cell–like crush artifact, micropapillary structure, and retraction tissue artifact.
Fig. 3
We defined the small cell–like crush artifact as tumor cells with a hyperchromatic, small nucleus and scanty cytoplasm and combined crushing artifact, which is frequently found in small cell carcinoma of the lung. Although a few cases of primary small cell carcinoma of the breast have been reported, all of our cases in this study were proven by a panel of immunohistochemical stains [25]. In this study, all of the 118 invasive breast carcinomas with small cell features showed no expression on immunohistochemistry for the most commonly used neuroendocrine markers including CD56, synaptophysin, and chromogranin that were used to discriminate small cell carcinoma of the breast in this study (data not shown). Crushing artifact has been known to be found in approximately 27% of breast CNB specimens during tissue sampling [26], which is
ROC curve for breast cancer with multiple nodules.
Breast biopsy and lymph node metastasis concordant with the results of our study (22.1%). Although the small cell feature with crushing artifact is a characteristic finding of small cell lung carcinoma with a reported higher risk of annual mortality compared to non–small cell lung carcinoma [27], this feature cannot be applied to our results due to discordant results on immunohistochemical stains. In this study, invasive carcinoma with solid sheet growth frequently observed in basal-like intrinsic subtype showed weak correlations with nodal involvement, which is similar to the results from a previous study [28]. In our opinion, this finding might be related to the cohesive nature of the tumor nests found in basal-like subtype in contrast to other invasive carcinomas with small cell carcinoma–like growth composed of relatively less cohesive and scattered tumor cell growth with higher frequencies of LNM. To our knowledge, the significance of such a finding has not been unveiled in breast cancer, especially with respect to LNM. Although further evaluation is needed, small-sized cells with scanty cytoplasm and crush artifact showed prominent infiltrating growth into the stroma compared to invasive breast carcinoma containing abundant cytoplasm with larger nuclei, which might also be considered a novel morphologic characteristic for invasiveness in breast carcinomas. Two other histomorphologic features of micropapillary structure and retraction tissue artifact have been suggested in prior studies [8,29,30]. Micropapillary growth pattern has been known to have a strong correlation with positive nodal status [8,11,30,31]. Acs et al [29] suggested that micropapillary structure and retraction tissue artifact in CNB were valuable predictive factors of LNM, which are findings consistent with those of the present study. Although a previous report suggested that retraction artifacts should be considered as a true prelymphatic space involved by tumor cell nests [12], in many cases, their biologic characteristics remain ambiguous. In addition, it is difficult to discriminate micropapillary growth from retraction tissue artifacts on histologic examinations. In our experience, however, micropapillary structure has a nest of tumor cells forming discrete tubular or solid structures floating symmetrically in the middle of an empty space. This feature can be highlighted by EMA IHC stain and similar with serous micropapillary carcinoma of the ovary. On the other hand, retraction artifact has more irregularly demarcated clear space and more solid structures floating asymmetrically in an empty space. Finally, we failed to present predictive models for all 5 of the intrinsic subtypes due to the limited number of cases, except for luminal A subtype with the 6-variable model including small cell–like crush artifact and micropapillary structures, which yielded an AUC of 0.791. Further studies incorporating larger cohorts are necessary for analysis and formulation of the predictive model of lymph node involvement in each intrinsic subtype. In conclusion, we propose a novel predictive model of LNM in invasive ductal carcinoma for preoperative CNB specimens, a routine diagnostic material encountered in daily practice. The combinations of the 8 and 6 variables
253 comprising characteristic pathologic findings on CNB revealed relatively good predictive performance (AUC, 0.760 and 0.791) in cases including all of the intrinsic subtypes of breast cancer as well as those including only luminal A intrinsic subtype of breast cancer. Our predictive model based on pathologic parameters for preoperative CNB may be used as a new nomogram in invasive ductal carcinoma of the breast for selecting optimal therapeutic strategies in patients with breast cancer.
Supplementary data Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.humpath.2014.10.017.
References [1] Fisher ER, Costantino J, Fisher B, Redmond C. Pathologic findings from the National Surgical Adjuvant Breast Project (Protocol 4). Discriminants for 15-year survival. National Surgical Adjuvant Breast and Bowel Project Investigators. Cancer 1993;71:2141-50. [2] Carter CL, Allen C, Henson DE. Relation of tumor size, lymph node status, and survival in 24,740 breast cancer cases. Cancer 1989;63:181-7. [3] Ernst MF, Voogd AC, Balder W, Klinkenbijl JH, Roukema JA. Early and late morbidity associated with axillary levels I-III dissection in breast cancer. J Surg Oncol 2002;79:151-5 [discussion 6]. [4] Chao C, Wong SL, Woo C, et al. Reliable lymphatic drainage to axillary sentinel lymph nodes regardless of tumor location within the breast. Am J Surg 2001;182:307-11. [5] Wong SL, Edwards MJ, Chao C, et al. Sentinel lymph node biopsy for breast cancer: impact of the number of sentinel nodes removed on the false-negative rate. J Am Coll Surg 2001;192:684-9 [discussion 9–91]. [6] Pal A, Provenzano E, Duffy SW, Pinder SE, Purushotham AD. A model for predicting non-sentinel lymph node metastatic disease when the sentinel lymph node is positive. Br J Surg 2008;95:302-9. [7] 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 2003;10:1140-51. [8] Acs G, Paragh G, Chuang ST, Laronga C, Zhang PJ. The presence of micropapillary features and retraction artifact in core needle biopsy material predicts lymph node metastasis in breast carcinoma. Am J Surg Pathol 2009;33:202-10. [9] 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 1991;19:403-10. [10] De la Cruz C, Moriya T, Endoh M, et al. Invasive micropapillary carcinoma of the breast: clinicopathological and immunohistochemical study. Pathol Int 2004;54:90-6. [11] Zekioglu O, Erhan Y, Ciris M, Bayramoglu H, Ozdemir N. Invasive micropapillary carcinoma of the breast: high incidence of lymph node metastasis with extranodal extension and its immunohistochemical profile compared with invasive ductal carcinoma. Histopathology 2004;44:18-23. [12] Zaorsky NG, Patil N, Freedman GM, Tuluc M. Differentiating lymphovascular invasion from retraction artifact on histological specimen of breast carcinoma and their implications on prognosis. J Breast Cancer 2012;15:478-80. [13] Hartveit E. Attenuated cells in breast stroma: the missing lymphatic system of the breast. Histopathology 1990;16:533-43.
254 [14] Travis WD. Advances in neuroendocrine lung tumors. Ann Oncol 2010;21(Suppl 7):vii65-71. [15] Hammond ME, Hayes DF, Dowsett M, et al. American Society of Clinical Oncology/College of American Pathologists guideline recommendations for immunohistochemical testing of estrogen and progesterone receptors in breast cancer. J Clin Oncol 2010;28:2784-95. [16] Wolff AC, Hammond ME, Hicks DG, et al. Recommendations for human epidermal growth factor receptor 2 testing in breast cancer: American Society of Clinical Oncology/College of American Pathologists clinical practice guideline update. J Clin Oncol 2013; 31:3997-4013. [17] Lee HC, Ko H, Seol H, et al. Expression of immunohistochemical markers before and after neoadjuvant chemotherapy in breast carcinoma, and their use as predictors of response. J Breast Cancer 2013;16:395-403. [18] Nielsen TO, Hsu FD, Jensen K, et al. Immunohistochemical and clinical characterization of the basal-like subtype of invasive breast carcinoma. Clin Cancer Res 2004;10:5367-74. [19] Rakha EA, El-Sayed ME, Green AR, Paish EC, Lee AH, Ellis IO. Breast carcinoma with basal differentiation: a proposal for pathology definition based on basal cytokeratin expression. Histopathology 2007;50:434-8. [20] Cheang MC, Chia SK, Voduc D, et al. Ki67 index, HER2 status, and prognosis of patients with luminal B breast cancer. J Natl Cancer Inst 2009;101:736-50. [21] Carey LA, Perou CM, Livasy CA, et al. Race, breast cancer subtypes, and survival in the Carolina Breast Cancer Study. JAMA 2006;295: 2492-502. [22] Bertolo C, Guerrero D, Vicente F, et al. Differences and molecular immunohistochemical parameters in the subtypes of infiltrating ductal breast cancer. Am J Clin Pathol 2008;130:414-24.
S. H. Yoo et al. [23] Goldhirsch A, Wood WC, Coates AS, Gelber RD, Thurlimann B, Senn HJ. Strategies for subtypes—dealing with the diversity of breast cancer: highlights of the St. Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2011. Ann Oncol 2011; 22:1736-47. [24] Cho J, Han W, Lee JW, et al. A scoring system to predict nonsentinel lymph node status in breast cancer patients with metastatic sentinel lymph nodes: a comparison with other scoring systems. Ann Surg Oncol 2008;15:2278-86. [25] Shin SJ, DeLellis RA, Ying L, Rosen PP. Small cell carcinoma of the breast: a clinicopathologic and immunohistochemical study of nine patients. Am J Surg Pathol 2000;24:1231-8. [26] Grimm EE, Schmidt RA, Swanson PE, Dintzis SM, Allison KH. Achieving 95% cross-methodological concordance in HER2 testing: causes and implications of discordant cases. Am J Clin Pathol 2010; 134:284-92. [27] Merrill RM, Henson DE, Barnes M. Conditional survival among patients with carcinoma of the lung. Chest 1999;116:697-703. [28] Howland NK, Driver TD, Sedrak MP, et al. Lymph node involvement in immunohistochemistry-based molecular classifications of breast cancer. J Surg Res 2013;185:697-703. [29] Acs G, Dumoff KL, Solin LJ, Pasha T, Xu X, Zhang PJ. Extensive retraction artifact correlates with lymphatic invasion and nodal metastasis and predicts poor outcome in early stage breast carcinoma. Am J Surg Pathol 2007;31:129-40. [30] Farid RM, Ahmed NS. Histolopathologic approach to direct the management of the axilla in early breast cancer patients. APMIS 2014; 122:1032-41. [31] Guo X, Chen L, Lang R, Fan Y, Zhang X, Fu L. Invasive micropapillary carcinoma of the breast: association of pathologic features with lymph node metastasis. Am J Clin Pathol 2006;126:740-6.
www.elsevier.com/locate/humpath
Original contribution
A histomorphologic predictive model for axillary lymph node metastasis in preoperative breast cancer core needle biopsy according to intrinsic subtypes☆,☆☆ Su Hyun Yoo MD a , In Ae Park MD, PhD b , Yul Ri Chung MD b , Hyojin Kim MD b , Keehwan Lee MD b , Dong-Young Noh MD, PhD c , Seock-Ah Im MD, PhD d , Wonshik Han MD, PhD c , Hyeong-Gon Moon MD, PhD c , Kyung-Hun Lee MD d , Han Suk Ryu MD, PhD b,⁎ a
Department of Pathology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, Republic of Korea Department of Pathology, Seoul National University Hospital, Seoul, Republic of Korea c Department of Surgery, Seoul National University Hospital, Seoul, Republic of Korea d Department of Internal Medicine, Seoul National University Hospital, Seoul, Republic of Korea b
Received 13 August 2014; revised 6 October 2014; accepted 31 October 2014
Keywords: Breast cancer; Lymph node metastasis; Predictive model; Histopathology; Nomogram
Summary The aim of this study is construction of a pathologic nomogram that can predict axillary lymph node metastasis (LNM) for each intrinsic subtype of breast cancer with regard to histologic characteristics in breast core needle biopsy (CNB) for use in routine practice. A total of 534 CNBs with invasive ductal carcinoma classified into 5 intrinsic subtypes were enrolled. Eighteen clinicopathological characteristics and 8 molecular markers used in CNB were evaluated for construction of the best predictive model of LNM. In addition to conventional parameters including tumor multiplicity (P b .001), tumor size (P b .001), high histologic grade (P = .035), and lymphatic invasion (P = .017), micropapillary structure (P b .001), the presence of small cell–like crush artifact (P = .001), and overexpression of HER2 (P = .090) and p53 (P = .087) were proven to be independent predictive factors for LNM. A combination of 8 statistically independent parameters yielded the strongest predictive performance with an area under the curve of 0.760 for LNM. A combination of 6 independent variables, including tumor number, tumor size, histologic grade, lymphatic invasion, micropapillary structure, and small cell–like crush artifact produced the best predictive performance for LNM in luminal A intrinsic subtype (area under the curve, 0.791). Thus, adding these combinations of clinical and morphologic parameters in preoperative CNB is expected to enhance the accuracy of prediction of LNM in breast cancer, which might serve as another valuable tool in determining optimal surgical strategies for breast cancer patients. © 2015 Elsevier Inc. All rights reserved.
☆
Competing interests: The authors declare that they have no conflict of interests. Funding/Support: This study was supported by grant number 04-2013-0840 from Seoul National University Hospital Research Fund. ⁎ Corresponding author at: Department of Pathology, Seoul National University Hospital, Seoul National University College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul, 110-744, Republic of Korea. E-mail address: [email protected] (H. S. Ryu).
☆☆
http://dx.doi.org/10.1016/j.humpath.2014.10.017 0046-8177/© 2015 Elsevier Inc. All rights reserved.
Breast biopsy and lymph node metastasis
1. Introduction The presence of axillary lymph node metastasis (LNM) is crucial in predicting clinical outcomes in breast cancer [1,2]. Although axillary lymph node dissection in patients with nodal metastasis is a standard procedure, it carries the risk of complications such as pain, loss of sensation, swelling, infection, impaired shoulder mobility, and lymphedema [3]. Although sentinel lymph node biopsy serves an optimal choice to selectively determine further axillary lymph node dissection, it has been reported to have a 5% false-negative rate in node-positive patients in breast cancer [4]. This falsenegative rate increased when only 1 sentinel lymph node was removed, thereby arousing debates as to how many lymph nodes should be removed [5]. Therefore, if there were certain ancillary methods to predict LNM, it might be helpful in reducing the false-negative rate as well as aid in making more accurate surgical decisions. Several predictive models for LNM have been introduced, many of which are being widely accepted for planning optimal surgical intervention in breast cancer patients [6,7]. Nevertheless, a predictive model or a pathologic nomogram for preoperative biopsy specimens incorporating characteristic histomorphologic parameters has not yet been evaluated thus far. Core needle biopsy (CNB) is an economic, valuable, and easily accessible preoperative diagnostic procedure used in daily practice. Although recent advances in diagnostic tools include molecular/genetic methods such as the nextgeneration sequencing, their application in daily diagnostic processes is extremely limited due to high costs. In this study, we examined the utility of preoperative CNB not only as a diagnostic tool but also as a predictor of LNM in breast cancer patients. In addition to the distinct morphologic characteristics of micropapillary structures and retraction tissue artifact, which are already known morphologic parameters related with LNM [8], we also adopted the small cell–like crush artifact observed in biopsy samples and constructed a model with the best predictive performance for LNM with a combination of these morphologic and other clinicopathological parameters found significant in breast cancer.
2. Materials and methods 2.1. Case selection Cases to be included in the study were selected from the pathologic archives of Seoul National University Hospital: we identified 5810 cases with CNB specimens over a 3-year period (January 2010 to December 2012). Among those patients, 534 cases fulfilling the following criteria were enrolled: (1) those who received axillary lymph node dissection (510 cases, 95.5% underwent sentinel lymph node biopsy) in subsequent surgical excision allowing evaluation of cancer metastasis; (2) those with the final
247 diagnosis of invasive ductal carcinoma; (3) CNB samples with at least 4 well-preserved cores; and (4) cases that had been grouped into intrinsic subtypes with immunostains. Patients who received neoadjuvant chemotherapy after undergoing CNB were excluded for precise evaluation of LNM. All the cases in the study underwent surgical excision. Thus, although we were able to guess each cancer to be either a solitary nodule or having multiple nodules through imaging studies before surgery, their solitary/multiplicity status was confirmed and changed according to the postoperative pathologic evaluations. Biopsy was performed using either an 11- or 14-gauge automated needle–assisted or with an 8- or 10-gauge vacuum–assisted device. All specimens fixed in 10% buffered formalin were paraffin embedded and hematoxylin and eosin (H&E) stained. To improve the accuracy of the results, 2 pathologists (S. H. Y. and H. S. R.) who were blinded to the clinical details of the patients reviewed histomorphologic findings. The patients' clinicopathological information was obtained from the patient database. This study was approved by the Institutional Review Board of Seoul National University Hospital.
2.2. Histomorphologic evaluations The following histologic factors were evaluated: micropapillary structure, retraction tissue artifact, small cell–like crush artifact of tumor cells, cribriform architecture, solid sheet growth showing syncytial arrangements of tumor cells with no tubule formation, spindling or multinucleated giant cells, presence of extracellular mucin, intratumoral calcification, tumor necrosis, loose myxoid fibrous stroma found in early phases of fibroadenoma, lymphatic invasion, neural invasion, and ductal carcinoma in situ (DCIS) component. Histologic grading was categorized into 1 to 3 based on Nottingham combined histologic grading system described by Elston and Ellis [9]. Micropapillary structure was defined as a nest of tumor cells forming discrete tubular or solid structures floating in the middle of an empty space (detached from the stroma) with or without immunohistochemical positivity for epithelial membrane antigen [10,11] (Fig. 1A and D). The presence of micropapillary structure was defined when more than 50% of the structure was identified in the entire core. Retraction tissue artifact was defined as the narrow cleft between the tumor glands and the surrounding stroma [12,13] (Fig. 1B and E). A tumor was considered to have a retraction tissue artifact when such histomorphologic finding occupied greater than or equal to 30% of the entire core biopsied. Small cell–like crush artifact was defined as infiltrative tumor cells with small-sized hyperchromatic nuclei and scanty cytoplasm resulting in a high nuclear/cytoplasmic ratio and crush artifact [14]. The tumor cells were distributed as either infiltrative cells showing streaming pattern or conglomerated clusters without tubule formation accompanied by severe crush artifact (Fig. 1C and F). Such small
248
S. H. Yoo et al.
Fig. 1 A and D, Micropapillary structures showing tumor nests devoid of fibrovascular cores, floating in empty stromal spaces. B and E, Retraction tissue artifact with nest of tumor cells being separated from the surrounding stroma by a clear space. C and F, Small and dark-stained cells with crushing artifact, designated “small cell–like crush artifact,” are infiltrating into the stroma. H&E stain, original magnification ×100 (A-C) and ×200 (D-F).
cell–like crush artifact was considered to be present when the feature was identified in greater than or equal to 10% of the entire core. Mucinous component and spindle or multinucleated giant cell differentiation were considered to be present if found in greater than or equal to 5% of the given cores. To assess lymphatic invasion of the tumor cells, immunohistochemical staining for D2-40 (1:100 dilution; DakoCytomation, Carpinteria, CA) was performed. The D240 immunostaining was only done for cases that were suspicious for lymphatic invasion or required discrimination from retraction artifact.
2.3. Immunohistochemistry Immunohistochemical staining was performed for subtyping of the breast cancers. Formalin-fixed, paraffinembedded tissue sections of all included patients were immunohistochemically stained using a Benchmark automatic immunostaining device (Ventana Medical System, Tucson, AZ). The samples were incubated with antibodies to estrogen receptors (ERs, 1:100, 1D5; Novocastra Laboratories, Newcastle, UK), progesterone receptors (PRs, 1:200, PgR636; Dako, Glostrup, Denmark), HER2 (Ventana), bcl-2 (1:200; Ventana), p53 (1:500, DO-7; Dako), epidermal growth factor receptor (EGFR) (Ventana), cytokeratin 5/6 (CK5/6, 1:50; Ventana), and Ki-67 (1:100; Dako). The sections were subsequently incubated with biotinylated antimouse immunoglobulins, peroxidase-labeled streptavidin (LSAB kit; Dako), and 3,3′-diaminobenzidine. Slides
were counterstained using Harris hematoxylin. Nuclear expression of tumor cells was interpreted as positive for ER, PR, bcl-2, and p53, whereas membrane staining of tumor cells was considered positive for HER2, EGFR, and CK5/6. Immunohistochemical staining was evaluated based on the location and percentage or intensity of positively stained cells. Immunohistochemical staining for ER and PR expression was counted and categorized as positive when greater than or equal to 1% of the tumor cells were stained according to the 2010 American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines [15]. Immunohistochemical expression of HER2 was assessed based on 2013 ASCO/CAP guidelines [16]. p53 and bcl-2 expressions were regarded as positive if greater than or equal to 25% of the tumor cells showed nuclear staining [17]. CK5/6 and EGFR (HER1) were scored positive when more than 10% of cytoplasmic staining was observed in the tumor cells [18,19]. The Ki-67 labeling index was counted as a percentage of immunopositivity with a cutoff value of 14% [20]. All of the immunohistochemical stainings were reviewed by 2 experienced breast pathologists (H. S. R. and I. A. P.) to improve accuracy.
2.4. Fluorescence in situ hybridization assays for HER2 gene amplification Presence of HER2 amplification was determined with an additional fluorescence in situ hybridization assay using PathVysion assay (Abbott Molecular, Downers Grove, IL) in cases with equivocal HER2 protein expression (IHC 2+) on
Breast biopsy and lymph node metastasis Table 1
249
Comparisons of continuous and categorical variables using the Student t test and χ2 test
Variables
Age (y, mean ± SD) Overall necrosis (%, mean ± SD) DCIS component (%, mean ± SD) Loose fibrous stroma (%, mean ± SD) ER expression (%, mean ± SD) PR expression (%, mean ± SD) Tumor number Single Multiple Tumor size b2 cm ≥2 cm Histologic grade of invasive carcinoma Absent Present Nuclear grade of DCIS Absent Present Lymphatic invasion Absent Present Neural invasion Absent Present Micropapillary structure Absent Present Retraction tissue artifact Absent Present Small cell–like crush artifact Absent Present Cribriform architecture Absent Present Solid sheet growth Absent Present Spindle or multinucleated giant cell Absent Present Mucinous component Absent Present Intratumoral calcification Absent Present HER2 expression Absent Present Ki-67 labeling index b14% ≥14%
LNM Breast cancer with multiple nodules (n = 534)
Breast cancer with solitary nodule a (n = 437)
No (n = 317)
No (n = 283)
53.86 ± 9.46 2.17 ± 8.99 6.56 ± 15.53 12.50 ± 18.02 58.04 ± 39.83 29.50 ± 36.63
Yes (n = 217) 52.29 ± 10.95 2.05 ± 7.90 3.99 ± 11.14 11.48 ± 17.00 55.12 ± 42.06 20.59 ± 32.45
P .096 .865 .026 .660 .417 .003
54.17 ± 10.70 1.94 ± 8.36 5.68 ± 14.82 12.29 ± 17.91 59.25 ± 39.63 30.01 ± 36.82
Yes (n = 154) 52.75 ± .9.74 2.18 ± 8.34 3.38 ± 9.05 12.41 ± 18.51 55.73 ± 42.20 18.45 ± 30.53
P .175 .782 .046 .948 .395 .001
284 (64.8%) 33 (34.7%)
155 (35.2%) 62 (65.3%)
b.001
251 (69.5%) 66 (38.2%)
110 (30.5%) 107 (61.8%)
b.001
223 (74.3%) 60 (43.8%)
77 (25.7%) 77 (56.2%)
b.001
186 (72.1%) 131 (47.5%)
72 (27.9%) 145 (52.5%)
b.001
168 (76.4%) 115 (53.0%)
52 (23.6%) 102 (47.0%)
b.001
204 (57.6%) 113 (62.8%)
150 (42.4%) 67 (37.2%)
.252
193 (64.1%) 90 (66.2%)
108 (35.9%) 46 (33.8%)
.677
307 (60.8%) 10 (34.5%)
198 (39.2%) 19 (65.5%)
.005
275 (66.3%) 8 (36.4%)
140 (33.7%) 14 (63.6%)
.004
304 (59.8%) 13 (50.0%)
204 (40.2%) 13 (50.0%)
.319
271 (65.0%) 12 (60.0%)
146 (35.0%) 8 (40.0%)
.648
286 (64.4%) 31 (34.4%)
158 (35.6%) 59 (65.6%)
b.001
255 (68.7%) 28 (42.4%)
116 (31.3%) 38 (57.6%)
b.001
222 (65.3%) 95 (49.0%)
118 (34.7%) 99 (51.0%)
b.001
199 (69.3%) 84 (56.0%)
88 (30.7%) 66 (44.0%)
.006
263 (63.2%) 54 (45.8%)
153 (36.8%) 64 (54.2%)
.001
231 (69.2%) 52 (50.5%)
103 (30.8%) 51 (49.5%)
.001
261 (56.7%) 56 (75.7%)
199 (43.3%) 18 (24.3%)
.002
234 (62.1%) 49 (81.7%)
143 (37.9%) 11 (18.3%)
.003
268 (58.0%) 49 (68.1%)
194 (42.0%) 23 (31.9%)
.106
237 (63.5%) 46 (71.9%)
136 (36.5%) 18 (28.1%)
.197
297 (58.9%) 20 (66.7%)
207 (41.1%) 10 (33.3%)
.402
264 (64.5%) 19 (67.9%)
145 (35.5%) 9 (32.1%)
.723
305 (58.9%) 12 (75.0%)
213 (41.1%) 4 (25.0%)
.301
272 (64.3%) 11 (78.6%)
151 (35.7%) 3 (21.4%)
.396
259 (59.1%) 58 (60.4%)
179 (40.9%) 38 (39.6%)
.876
223 (64.2%) 60 (71.4%)
130 (35.8%) 24 (28.6%)
.579
282 (62.8%) 35 (41.2%)
167 (37.2%) 50 (58.8%)
b.001
256 (68.3%) 27 (43.5%)
119 (31.7%) 35 (56.5%)
b.001
265 (61.3%) 52 (51.0%)
167 (38.7%) 50 (49.0%)
.055
239 (67.3%) 44 (53.7%)
116 (32.7%) 38 (46.3%)
.020
(continued on next page)
250
S. H. Yoo et al.
Table 1 (continued) Variables
CK5/6 expression Absent Present Bcl2 expression Absent Present p53 expression Absent Present EGFR expression Absent Present Total
LNM Breast cancer with multiple nodules (n = 534)
Breast cancer with solitary nodule a (n = 437)
No (n = 317)
Yes (n = 217)
No (n = 283)
Yes (n = 154)
297 (58.8%) 20 (69.0%)
208 (41.2%) 9 (31.2%)
.279
266 (64.3%) 17 (73.9%)
148 (35.7%) 6 (26.1%)
.345
73 (50.0%) 244 (62.9%)
73 (50.0%) 144 (37.1%)
.007
59 (52.2%) 224 (69.1%)
54 (47.8%) 100 (30.9%)
.001
212 (64.8%) 105 (50.7%)
115 (35.2%) 102 (49.3%)
.001
191 (70.2%) 92 (55.8%)
81 (29.8%) 73 (44.2%)
.002
290 (60.5%) 27 (49.1%) 317
189 (39.5%) 28 (50.9%) 217
.101
260 (66.2%) 23 (52.3%) 283
133 (33.8%) 21 (47.7%) 154
.067
P
P
Abbreviations: CK, cytokeratin; DCIS, ductal carcinoma in situ; EGFR, epidermal growth factor receptor; ER, estrogen receptor; LNM, lymph node metastasis; PR, progesterone receptor. a Breast cancer excluding multiple nodules.
immunohistochemistry, as previously described. Interpretation for HER2 gene amplification by fluorescence in situ hybridization assay was confirmed based on the greater than or equal to 2.0 ratio of HER2 gene copy number to chromosome 17 copy number or greater than or equal to 6.0 average HER2 signal per tumor cell as defined in the 2013 ASCO/CAP guidelines [16].
were independent predictors of LNM. P b .05 was considered statistically significant except in backward selection methods for which P b .2 was considered statistically significant. Statistical analyses were performed using the SPSS software (SPSS Standard version, 17.0; SPSS, Chicago, IL).
2.5. Definition of breast cancer immunohistochemical intrinsic subtypes
3. Results
Intrinsic subtypes of breast cancer were defined using immunohistochemical profiles in accordance with previous studies [21-23]. The 5 immunohistochemical phenotypes were as follows: luminal A (ER+ and/or PR+, HER2−, Ki-67 b14%), luminal B with HER2 negative (ER+ and/or PR+, HER2−, Ki67 ≥14%), luminal B with HER2 positive (ER+ and/or PR+, HER2+, any Ki-67), HER2 enriched (ER−, PR−, HER2+), and basal-like (ER−, PR−, HER2−, CK5/6+ and/or EGFR+).
2.6. Statistical analyses The χ2 and Fisher exact tests were used to analyze the statistical significance of the association between LNM status and categorical parameters. The Student t test was used to compare the continuous variables. Logistic regression analysis was performed to evaluate influential risk factors for LNM. To identify factors that were independently predictive of the risk of LNM, multivariate logistic regression analysis was performed with backward selection methods to avoid overfitting. Receiver operating characteristic (ROC) curves were constructed, and areas under the curves (AUCs) were calculated to determine optimal cutoff values for all possible combinations of the variables that
3.1. Clinicopathological characteristics of breast cancer and their correlations with LNM Clinicopathological variables of the patients are as summarized in Table 1. Of the 534 cases with invasive ductal carcinoma, 217 cases (40.6%) had LNM. LNM was associated with the following variables: less amount of DCIS component (P = .026), negative expression of PR (P = .003), tumor multiplicity (P b .001), larger tumor size of more than 2 cm (P b .001), higher histologic grade of invasive carcinoma cells (P b .001), lymphatic invasion (P = .005), micropapillary structure (P b .001), retraction tissue artifact (P b .001), small cell–like crush artifact (P = .001), less cribriform architecture (P = .002), HER2 overexpression (P b .001), bcl-2 overexpression (P = .007), and p53 overexpression (P = .001). The above 13 parameters that had been found to have significant correlations with LNM were also shown to be associated with positive nodal status in univariate analyses (Table 2). When we also evaluated the association of these variables with LNM in breast cancer with a solitary nodule, a higher Ki-67 labeling index (P = .020) along with the other 12 variables that were revealed to have significant correlations in breast cancer with multiple nodules was associated with LNM in solitary breast cancers
Breast biopsy and lymph node metastasis
251
(Tables 1 and 2). We further examined clinicopathological variables correlating with LNM for each intrinsic subtype of breast cancer. Comprehensive results for the predictive variables of LNM in each intrinsic subtype of breast cancer including those with single nodules only are summarized in Supplementary Table 1.
3.2. Predictive models for LNM in breast cancer CNB Of the 13 clinicopathological variables associated with positive nodal status in 534 breast carcinomas, the 8 factors were found to be independent parameters in multivariate logistic regression analysis (Table 3). Using these 7 independent predictive variables for breast cancer with solitary nodule and 8 variables for breast cancer with multiple nodules, prediction models were constructed and calibrated using Hosmer-Lemeshow goodness-of-fit test (P = .462 and P =
Table 2 of LNM Variables
Univariate logistic regression tests for prediction OR
95% CI
Table 3 of LNM
Multivariate logistic regression tests for prediction
Variables
OR
Predictive analysis for breast cancer with Tumor number 3.666 Tumor size 3.308 Histologic grade of 1.592 invasive carcinoma Lymphatic invasion 2.930 Micropapillary structure 2.950 Small cell–like crush artifact 2.273 HER2 expression 1.615 p53 expression 1.438 Predictive analysis for breast cancer with Tumor size 3.489 Lymphatic invasion 3.206 Micropapillary structure 3.019 Small cell–like crush artifact 2.223 HER2 expression 2.428 Ki-67 labeling index 1.698 p53 expression 1.487
95% CI
P
multiple nodules 2.183-6.155 b.001 2.181-5.016 b.001 1.033-2.454 .035 1.210-7.092 .017 1.741-5.009 b.001 1.421-3.635 .001 0.928-2.811 .090 0.948-2.180 .087 solitary nodule 2.220-5.482 b.001 1.193-8.616 .021 1.683-5.413 b.001 1.357-3.643 .002 1.319-4.468 .004 0.973-2.966 .063 0.928-2.381 .099
Abbreviations: CI, confidence interval; LNM, lymph node metastasis; OR, odds ratio.
P
Predictive analysis for breast cancer with multiple nodules DCIS component (%, mean ± SD) 0.985 0.972-0.999 .041 PR immunohistochemical 0.993 0.987-0.998 .004 expression (%, mean ± SD) Tumor number 3.900 2.468-6.165 b.001 Tumor size 3.900 2.468-6.165 b.001 Histologic grade of 2.799 1.962-3.992 b.001 invasive carcinoma Lymphatic invasion 2.841 1.295-6.233 .009 Micropapillary structure 3.430 1.972-5.968 b.001 Retraction tissue artifact 1.908 1.346-2.704 b.001 Small cell–like crush artifact 2.324 1.573-3.435 b.001 Cribriform architecture 0.431 0.245-0.757 .003 HER2 overexpression 2.412 1.504-3.869 b.001 Bcl2 overexpression 0.590 0.402-0.867 .007 p53 expression 1.791 1.256-2.553 .001 Predictive analysis for breast cancer with solitary nodule DCIS component (%, mean ± SD) 0.985 0.967-1.002 .088 PR immunohistochemical 0.990 0.984-0.996 .001 expression (%, mean ± SD) Tumor size 3.717 2.429-5.688 b.001 Histologic grade of 2.094 1.353-3.242 .001 invasive carcinoma Lymphatic invasion 3.437 1.409-8.389 .007 Micropapillary structure 2.983 1.747-5.095 b.001 Retraction tissue artifact 1.777 1.181-2.673 .006 Small cell–like crush artifact 2.200 1.402-3.452 .001 Cribriform architecture 0.367 0.185-0.730 .004 HER2 expression 2.789 1.614-4.819 b.001 Ki-67 labeling index 1.779 1.093-2.897 .020 Bcl2 expression 0.488 0.315-0.756 .001 p53 expression 1.871 1.251-2.798 .002 Abbreviations: CI, confidence interval; DCIS, ductal carcinoma in situ; LNM, lymph node metastasis; OR, odds ratio.
.762), and AUC was 0.735 and 0.760 in breast cancer with solitary and multiple nodules, respectively (Figs. 2 and 3). Using the independent parameters obtained from multivariate logistic regression analyses in each breast carcinoma subtype, predictive models for LNM were evaluated for each subtype (Supplementary Table 2). As a result, 6 independent variables including tumor multiplicity (odds ratio [OR], 6.271; P b .001), increased tumor size (OR, 5.077; P b .001), increased histologic grade of invasive carcinoma (OR, 1.917; P = .030), presence of lymphatic invasion (OR, 3.881; P = .027), micropapillary structures (OR, 2.719; P = .004), and small cell–like crush artifact (OR, 2.057; P = .033) were used to construct a predictive model for LNM luminal A intrinsic subtype, and ROC was constructed from AUC (Hosmer-Lemeshow goodness-of-fit test, P = .697) with combination of 6 variables, which yielded AUC of 0.791. In luminal A intrinsic subtype with single nodules only, the 5 combined predictive models yielded AUC of 0.742 on ROC. Predictive models for luminal B (both HER2− and HER2+ subtypes) and basal-like subtype were not constructed because we were unable to discover independent parameters for these subtypes on statistical analysis.
4. Discussion In this study, we investigated pathologic and molecular features that help predict LNM in preoperative breast cancer biopsy and observed 8 significant predictors composed of 6 clinicopathological and 2 molecular
252
S. H. Yoo et al.
Fig. 2
ROC curve for breast cancer with solitary nodule.
parameters. The AUC using the combination of these 8 features was 0.760, which showed predictive performance comparable to those of the nomograms previously reported [6,7,24]. These previous predictive models were developed using mainly clinical or radiologic parameters with only some conventional pathologic variables such as histologic grade, ER status, and lymphatic invasion [7,24]. In this study, we focused specifically on pathologic parameters, including histomorphologic features and molecular markers, which are known driver genes in carcinogenesis of mammary carcinoma. Three distinct histomorphologic features were evaluated in our predictive model, which are small cell–like crush artifact, micropapillary structure, and retraction tissue artifact.
Fig. 3
We defined the small cell–like crush artifact as tumor cells with a hyperchromatic, small nucleus and scanty cytoplasm and combined crushing artifact, which is frequently found in small cell carcinoma of the lung. Although a few cases of primary small cell carcinoma of the breast have been reported, all of our cases in this study were proven by a panel of immunohistochemical stains [25]. In this study, all of the 118 invasive breast carcinomas with small cell features showed no expression on immunohistochemistry for the most commonly used neuroendocrine markers including CD56, synaptophysin, and chromogranin that were used to discriminate small cell carcinoma of the breast in this study (data not shown). Crushing artifact has been known to be found in approximately 27% of breast CNB specimens during tissue sampling [26], which is
ROC curve for breast cancer with multiple nodules.
Breast biopsy and lymph node metastasis concordant with the results of our study (22.1%). Although the small cell feature with crushing artifact is a characteristic finding of small cell lung carcinoma with a reported higher risk of annual mortality compared to non–small cell lung carcinoma [27], this feature cannot be applied to our results due to discordant results on immunohistochemical stains. In this study, invasive carcinoma with solid sheet growth frequently observed in basal-like intrinsic subtype showed weak correlations with nodal involvement, which is similar to the results from a previous study [28]. In our opinion, this finding might be related to the cohesive nature of the tumor nests found in basal-like subtype in contrast to other invasive carcinomas with small cell carcinoma–like growth composed of relatively less cohesive and scattered tumor cell growth with higher frequencies of LNM. To our knowledge, the significance of such a finding has not been unveiled in breast cancer, especially with respect to LNM. Although further evaluation is needed, small-sized cells with scanty cytoplasm and crush artifact showed prominent infiltrating growth into the stroma compared to invasive breast carcinoma containing abundant cytoplasm with larger nuclei, which might also be considered a novel morphologic characteristic for invasiveness in breast carcinomas. Two other histomorphologic features of micropapillary structure and retraction tissue artifact have been suggested in prior studies [8,29,30]. Micropapillary growth pattern has been known to have a strong correlation with positive nodal status [8,11,30,31]. Acs et al [29] suggested that micropapillary structure and retraction tissue artifact in CNB were valuable predictive factors of LNM, which are findings consistent with those of the present study. Although a previous report suggested that retraction artifacts should be considered as a true prelymphatic space involved by tumor cell nests [12], in many cases, their biologic characteristics remain ambiguous. In addition, it is difficult to discriminate micropapillary growth from retraction tissue artifacts on histologic examinations. In our experience, however, micropapillary structure has a nest of tumor cells forming discrete tubular or solid structures floating symmetrically in the middle of an empty space. This feature can be highlighted by EMA IHC stain and similar with serous micropapillary carcinoma of the ovary. On the other hand, retraction artifact has more irregularly demarcated clear space and more solid structures floating asymmetrically in an empty space. Finally, we failed to present predictive models for all 5 of the intrinsic subtypes due to the limited number of cases, except for luminal A subtype with the 6-variable model including small cell–like crush artifact and micropapillary structures, which yielded an AUC of 0.791. Further studies incorporating larger cohorts are necessary for analysis and formulation of the predictive model of lymph node involvement in each intrinsic subtype. In conclusion, we propose a novel predictive model of LNM in invasive ductal carcinoma for preoperative CNB specimens, a routine diagnostic material encountered in daily practice. The combinations of the 8 and 6 variables
253 comprising characteristic pathologic findings on CNB revealed relatively good predictive performance (AUC, 0.760 and 0.791) in cases including all of the intrinsic subtypes of breast cancer as well as those including only luminal A intrinsic subtype of breast cancer. Our predictive model based on pathologic parameters for preoperative CNB may be used as a new nomogram in invasive ductal carcinoma of the breast for selecting optimal therapeutic strategies in patients with breast cancer.
Supplementary data Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.humpath.2014.10.017.
References [1] Fisher ER, Costantino J, Fisher B, Redmond C. Pathologic findings from the National Surgical Adjuvant Breast Project (Protocol 4). Discriminants for 15-year survival. National Surgical Adjuvant Breast and Bowel Project Investigators. Cancer 1993;71:2141-50. [2] Carter CL, Allen C, Henson DE. Relation of tumor size, lymph node status, and survival in 24,740 breast cancer cases. Cancer 1989;63:181-7. [3] Ernst MF, Voogd AC, Balder W, Klinkenbijl JH, Roukema JA. Early and late morbidity associated with axillary levels I-III dissection in breast cancer. J Surg Oncol 2002;79:151-5 [discussion 6]. [4] Chao C, Wong SL, Woo C, et al. Reliable lymphatic drainage to axillary sentinel lymph nodes regardless of tumor location within the breast. Am J Surg 2001;182:307-11. [5] Wong SL, Edwards MJ, Chao C, et al. Sentinel lymph node biopsy for breast cancer: impact of the number of sentinel nodes removed on the false-negative rate. J Am Coll Surg 2001;192:684-9 [discussion 9–91]. [6] Pal A, Provenzano E, Duffy SW, Pinder SE, Purushotham AD. A model for predicting non-sentinel lymph node metastatic disease when the sentinel lymph node is positive. Br J Surg 2008;95:302-9. [7] 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 2003;10:1140-51. [8] Acs G, Paragh G, Chuang ST, Laronga C, Zhang PJ. The presence of micropapillary features and retraction artifact in core needle biopsy material predicts lymph node metastasis in breast carcinoma. Am J Surg Pathol 2009;33:202-10. [9] 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 1991;19:403-10. [10] De la Cruz C, Moriya T, Endoh M, et al. Invasive micropapillary carcinoma of the breast: clinicopathological and immunohistochemical study. Pathol Int 2004;54:90-6. [11] Zekioglu O, Erhan Y, Ciris M, Bayramoglu H, Ozdemir N. Invasive micropapillary carcinoma of the breast: high incidence of lymph node metastasis with extranodal extension and its immunohistochemical profile compared with invasive ductal carcinoma. Histopathology 2004;44:18-23. [12] Zaorsky NG, Patil N, Freedman GM, Tuluc M. Differentiating lymphovascular invasion from retraction artifact on histological specimen of breast carcinoma and their implications on prognosis. J Breast Cancer 2012;15:478-80. [13] Hartveit E. Attenuated cells in breast stroma: the missing lymphatic system of the breast. Histopathology 1990;16:533-43.
254 [14] Travis WD. Advances in neuroendocrine lung tumors. Ann Oncol 2010;21(Suppl 7):vii65-71. [15] Hammond ME, Hayes DF, Dowsett M, et al. American Society of Clinical Oncology/College of American Pathologists guideline recommendations for immunohistochemical testing of estrogen and progesterone receptors in breast cancer. J Clin Oncol 2010;28:2784-95. [16] Wolff AC, Hammond ME, Hicks DG, et al. Recommendations for human epidermal growth factor receptor 2 testing in breast cancer: American Society of Clinical Oncology/College of American Pathologists clinical practice guideline update. J Clin Oncol 2013; 31:3997-4013. [17] Lee HC, Ko H, Seol H, et al. Expression of immunohistochemical markers before and after neoadjuvant chemotherapy in breast carcinoma, and their use as predictors of response. J Breast Cancer 2013;16:395-403. [18] Nielsen TO, Hsu FD, Jensen K, et al. Immunohistochemical and clinical characterization of the basal-like subtype of invasive breast carcinoma. Clin Cancer Res 2004;10:5367-74. [19] Rakha EA, El-Sayed ME, Green AR, Paish EC, Lee AH, Ellis IO. Breast carcinoma with basal differentiation: a proposal for pathology definition based on basal cytokeratin expression. Histopathology 2007;50:434-8. [20] Cheang MC, Chia SK, Voduc D, et al. Ki67 index, HER2 status, and prognosis of patients with luminal B breast cancer. J Natl Cancer Inst 2009;101:736-50. [21] Carey LA, Perou CM, Livasy CA, et al. Race, breast cancer subtypes, and survival in the Carolina Breast Cancer Study. JAMA 2006;295: 2492-502. [22] Bertolo C, Guerrero D, Vicente F, et al. Differences and molecular immunohistochemical parameters in the subtypes of infiltrating ductal breast cancer. Am J Clin Pathol 2008;130:414-24.
S. H. Yoo et al. [23] Goldhirsch A, Wood WC, Coates AS, Gelber RD, Thurlimann B, Senn HJ. Strategies for subtypes—dealing with the diversity of breast cancer: highlights of the St. Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2011. Ann Oncol 2011; 22:1736-47. [24] Cho J, Han W, Lee JW, et al. A scoring system to predict nonsentinel lymph node status in breast cancer patients with metastatic sentinel lymph nodes: a comparison with other scoring systems. Ann Surg Oncol 2008;15:2278-86. [25] Shin SJ, DeLellis RA, Ying L, Rosen PP. Small cell carcinoma of the breast: a clinicopathologic and immunohistochemical study of nine patients. Am J Surg Pathol 2000;24:1231-8. [26] Grimm EE, Schmidt RA, Swanson PE, Dintzis SM, Allison KH. Achieving 95% cross-methodological concordance in HER2 testing: causes and implications of discordant cases. Am J Clin Pathol 2010; 134:284-92. [27] Merrill RM, Henson DE, Barnes M. Conditional survival among patients with carcinoma of the lung. Chest 1999;116:697-703. [28] Howland NK, Driver TD, Sedrak MP, et al. Lymph node involvement in immunohistochemistry-based molecular classifications of breast cancer. J Surg Res 2013;185:697-703. [29] Acs G, Dumoff KL, Solin LJ, Pasha T, Xu X, Zhang PJ. Extensive retraction artifact correlates with lymphatic invasion and nodal metastasis and predicts poor outcome in early stage breast carcinoma. Am J Surg Pathol 2007;31:129-40. [30] Farid RM, Ahmed NS. Histolopathologic approach to direct the management of the axilla in early breast cancer patients. APMIS 2014; 122:1032-41. [31] Guo X, Chen L, Lang R, Fan Y, Zhang X, Fu L. Invasive micropapillary carcinoma of the breast: association of pathologic features with lymph node metastasis. Am J Clin Pathol 2006;126:740-6.
