Quantification of microvessel density of breast carcinoma: an assessment of the inter- and intraobserver variation

2The Breast (1999) 8, 251–256 © 1999 Harcourt Publisher Ltd

ORIGINAL ARTICLE

Quantification of microvessel density of breast carcinoma: an assessment of the inter- and intraobserver variation P. Ingeholm*, L. Pedersen† and S. Holck* *Department of Pathology, Hillerød Hospital, Denmark †Department of Oncology, Rigshospitalet, Denmark S U M M A R Y. The density profile of microvessels (MVD) has recently emerged as a prognostically independent morphological marker of various malignancies, including carcinoma of the breast. MVD-scoring may thus afford the pathologist the opportunity to identify subsets of early stage breast cancer patients, which may benefit from adjuvant therapy. Reproducibility studies are, however, mandatory. Fifty sections of invasive breast carcinoma, immunolabelled for Factor VIII-related antigen, were scored independently by two observers for the number of vessel-profiles (v-p) in the following manner: 1) 250 randomly selected high power fields (HPF) were scored by two analysts to evaluate the agreement of the counting per se; and 2) the hot spot(s), i.e. the zones considered the most vessel rich, in the 50 sections were identified and scored twice to evaluate the agreement on selecting hot spots. When the observers evaluated MVD in the very same 250 HPF a median deviation of 2 v-p (10%) was produced. The interobserver disparity was further accentuated when the participants independently had to identify the hot spots with a median deviation of 6 v-p (13%). Intraobserver variation was approximately equal to the interobserver discordance. Divergences were most conspicuous in fields with a complex vasculature. In conclusion the v-p scoring resulted in substantial inter- and intraobserver variation both in selecting hot spots and in performing the scoring per se. Alternative-scoring techniques, including strict stereological principles, should be considered in the morphological evaluation of angiogenesis. © 1999 Harcourt Publishers Ltd

INTRODUCTION

There is a need to identify a marker delineating a subset of early stage carcinomas which may benefit from adjuvant therapy. Whereas some investigators have failed to demonstrate a correlation between MVD and metastatic potential25–31 others have documented such an association.12,32–34 In view of the promising results and their possible therapeutic impacts, rigorous reproducibility studies concerning MVD are warranted.35,36 Current knowledge of inter- and intraobserver variability of MVD scoring is limited and contradictory, with most papers and abstracts including only brief remarks on the subject.29,31–33,37–40 Communications focusing primarily on this issue are few,25 and none has previously dealt with the two potential sources of errors separately: 1) the identification of the hot spots; and 2) the v-p scoring per se. In the paper these two aspects have been studied in an observer agreement study of a series of invasive breast carcinomas.

Tumour angiogenesis in a solid neoplasm1 is a sine qua non for the unrestricted growth and potential for dissemination.2,3 The potent angiogenic effect exerted by certain growth factors substantiates this notion.4–7 Hence, vascular aspects may prove useful prognosticators of malignancies. To this end, the vessel-related urokinase plasminogen activatorsystem has been shown to predict the malignant potential of different subgroups of breast carcinoma.8–10 Another approach evaluating microvessel density (MVD) has produced comparable encouraging results. Its significance as an important and possibly independent prognostic marker has been the subject of several recent publications on carcinoma of the breast.4,11–13 Similar associations have been documented for carcinoma of the prostate,14,15 head and neck,16 lung (non-small cell type),13,17 and gastrointestinal tract.18–20 Ductal carcinoma in situ of comedo subtype, a more aggressive type of non-invasive ductal carcinoma of the breast,21,22 is also more commonly associated with a high MVD when compared to other types of in situ carcinoma.23,24

MATERIALS AND METHODS Fifty sections of invasive carcinomas of the breast were immunolabelled for factor VIII-related antigen as follows.

Address correspondence to: Dr S. Holck, Department of Pathology, Hillerød Hospital, 3400 Hillerød, Denmark.

251

The Breast the qualitative aspects of vessels that might lead to disagreement.

RESULTS Scoring of MVD in 250 randomly selected microscopic fields caused substantial disagreement with a median difference of 2.0 v-p (10%) and a range of 0–14 v-p (0–67%), the 5% and 95% fractils being 0–6 v-p (0–33%). It appears from Figure 1 that in only 72 HPF (28,8%) a score difference of less than 6% was obtained, but in 120 HPF (48%) the score divergence exceeded 10%. The results of the ‘hot spot’ scoring are summarized in Table 1. The extra variable in this part of the study increased interobserver disagreement that reached 81% in one case. Interestingly, no difference was observed between intra- and interobserver disagreement (data not shown). Thus, concordance was not enhanced in any of the 4 intraobserver studies. Another unexpected finding emerged when comparing the first scoring round based on randomly selected fields with hot spot scoring. In 11 of 25 sections, the highest scores were recorded by the random method by at least one of the observers and in 7 of these sections the counts of both observers were highest by random selection (data not

80 70

72 58

60 50

55

40 21 1

2

2 >50

3

46-50

9 41-45

21-25

16-20

11-15

6-10

26-30

8

10 0

36-40

19

31-35

30 20

0-5

The sections were dewaxed and placed in a graded series of alcohols and treated with 0.1% protease for 5 min. Endogenous peroxidase was blocked by 1% hydrogen peroxide for 20 min. Non-immune swine serum at a dilution of 1:20 was applied for 5 min to reduce non-specific background staining. Subsequently sections were incubated for 30 min at room temperature with polyclonal rat anti-human factor VIII related antigen (Dakopatt, Copenhagen, Denmark) followed by incubation for 30 min with the appropriate peroxidase–conjugated secondary antibody. The bound peroxidase activity was localized using 3-amino9-ethylcarbazol for 10 min as a chromogen. Sections were finally counterstained with haematoxylin and mounted. Labelling of native vessels served as a reliable positive control. Negative controls employed tissue sections stained with swine serum in place of the primary antibody. Controls were run for every section. Quantification was performed on a one-headed Leitz microscope, on a 400× field (40× objective, 10× ocular) corresponding to an area of 0.261 mm2. All labelled structures, regardless of a recognizable lumen, were counted, including those touching the edge of the field. During the planning stage, including an initial training session, problem cases were thoroughly discussed. The prime objective of the morphometric evaluation was to determine observer variation of the scoring process. The counting was performed by two of the authors (PI and SH, having 2 and 20 years of experience in pathology), testing two potential variables: the scoring per se and the identification of hot spot(s) (the area(s) considered to have the highest MVD). Accordingly, the vessel scoring fell in two parts and included the following 1100 measurements: 1) In 250 randomly selected fields (chosen with the image being slightly out of focus) the numbers of vessel profiles (v-p) were counted by both observers. For each field the difference expressed in percent (calculated as the absolute difference relative to the average of the analysts two scorings) in the number of v-p counted was calculated. The median difference for all 250 fields examined were determined. 2) Each observer independently attempted to identify the hot spots of 50 sections, by first screening at low power (× 100 magnification) and secondly counting the number of v-p. For each section three hot spots were identified. The hot spot with the highest MVD, as well as the average value of MVD for the three fields, were determined. The difference in the number of v-p counted was calculated as described above. After an interval of 2–4 months, the latter procedure was repeated, to determine intraobserver reproducibility. Upon completion of the scoring rounds the material was reviewed in an attempt to identify features such as

Number of fields scored

252

Difference in percent of v-p scoring Fig. 1 Interobserver variation in scoring number of vessel profiles (v-p) in 250 randomly selected fields of invasive breast carcinoma.

Table 1 Inter- and intraobserver variation of hot spot scoring of 50 sections of invasive breast carcinomas (v-p1 in the 3 fields with the highest density was counted) V-p score

Highest2 Average3

Examiners PI-PI SH-SH PI-SH PI-PI SH–SH PI-SH

Difference in v-p scoing Median (%) Range (%) 4(13) 6 (22) 6 (16) 3 (13) 4 (17) 3 (13)

0–22 (0–63) 0–25 (0–72) 0–30 (0–81) 0–23 (0–62) 0–15 (0–59) 0–15 (0–74)

1 V-p: Vessel profiles. 2The single highest value. 3The average of 3 scorings (the 3 fields with the highest values).

Microvessel density in breast carcinoma 253 shown). There was no tendency to indicate that one of the observers consistently provided a higher score. During quantification it became evident that some qualitative features of the microvasculature did influence results. Scoring of zones comprising a maze-like vessel pattern (Fig. 2) was cumbersome and resulted in the substantial disagreement. Such complex vasculature generally occupied the peripheral portion of tumours and may represent the bona fide hot spots. Difficulties were similarly experienced when scoring indistinctly labelled vessels. The latter was not an effect of suboptimal technique. Sections comprising weakly outlined tumour vessels contrasting crisp, vivid labelling of peritumorous vessels did exist (Fig. 3). Rather, the smudgy, indistinct appearance may be a quality inherent in some tumour vessels. Extensive reticulated vascular patterns coupled with indistinct less intense immunolabelling characterized most, though not all the sections with substantial inter observer divergence. The counting of areas

Fig. 2 A rich network of tortuous curvilinear and branching, vasoformative structures predominates. Scoring of such areas is a difficult task (immunoperoxidase stain for factor VIII-related antigen).

B

A Fig. 3 This composite picture of central zone (A) and peripheral rim (B) of an invasive carcinoma illustrates a weak immunoexpression of the tumour vessels (arrows) (A), in contrast to the vivid labelling of vessels in the peritumoral fatty tissue. Incidentally, note the rather low MVD of the tumour area (A) as compared to the peritumoral stroma (immunoperoxidase stain for factor VIII-related antigen).

254

The Breast

Fig. 4 This patchy pattern of microvessels makes scoring feasible (immunoperoxidase stain for factor VIII-related antigen).

with patchily distributed, intensely stained vessels (Fig. 4) proved more consistent.

DISCUSSION Several factors may theorectically influence the result of vessel scoring. Some of these are linked to handling of the tissue and choice of antibody to highlight the vasculature. Several authors4,12,41 have discussed the latter aspect. Antibodies against factor VIII-related antigen have been used frequently;24,26,42 others have employed antibodies against CD3443 or CD3144. It has been emphasized that the sensitivity of CD31 exceeds that of factor VIII,32 whereas CD31 in some studies has been found to be less specific that factor VIII.45 Other uncertainties concern the scoring process, which is dependable of the observers ability to a) locate the most vessel-rich zones; and b) record the labelled structures. The current study has focused on these observerrelated aspects of vascular scoring. It was our primary intention to design a study in such a manner that observer variation was kept to a minimum. To this end the choice of the appropriate magnification was discussed. With few exceptions,38 most workers have used a magnification of approximately 200×4,32,42,43,46 corresponding to an area of 0.384–0.75 mm2. It was our presumption that reproducible results would not be obtainable if counts significantly exceeded 50. For this reason we refrained from using the recommended larger counting area which frequently embraced 50, occasionally more than 100 v-p. The risk of scoring the same vessel twice, omitting other vessels is undoubtedly significant. Despite this precaution, disagreement seemed substantial even when we assessed exactly the same fields. Thus, diver-

gences occasionally exceeded 50%. If vessel scoring is introduced in daily routine morphological evaluation of breast cancer, the method should be as simple as possible. For this reason, our initial inter- and intraobserver studies included no special device, not even an eyepiece graticule. We considered this acceptable, since several authors reporting favourable results, similarly used field counting.4,47,48 With the experience gained in this study, we realize that in subsequent comparative observer studies Chalkley Counting, as suggested by Fox et al.44 and Vermeulen et al,49 may improve results. As expected, greater variation was evident during the second part of the study, in which another (non-random) variable (the identification of hot spots) was added. Of particular note was the fact that the intraobserver reproducibility of the two participants, an experienced and a less experienced pathologist, was comparable. In the comprehensive study of Axelsson et al.,25 considerable variability in v-p counting, performed by two analysts was also found. In a study of 10 sections, van Hoef et al.31 did not observe a discordance exceeding 10 and 5 v-p at 100× and 200× magnification, respectively. Our dismal results are also in contrast to those of Horak et al.32 who found an interobserver discordance that never exceeded 10%. Further details, such as number of cases included and the mode of calculating the percentage, were, however, not included in their paper. The sources of discordant results are probably complex and multifactorial. As expressed by van Diest et al.50 motivation is essential in reproducibility studies in general. Furthermore, some degree of training prior to initiation of a study is desirable.33,40 Such prerequisites were fulfilled in our study. Some investigators studying MVD, have excluded sections (2 to nearly 10%) that were deemed technically unsatisfactory.1,4,24,48 The substantial divergence observed in 11% of cases scored by Klijanienko et al.39 was attributed to inadequate fixation and/or tissue processing. In the present study no selection was undertaken, conceivably contributing to our level of disagreement. Our lack of concordance seemed, however, linked to the quality of the vessels more so than to the technical quality of slide preparation. During the course of this study it became apparent that not all zones of the carcinomas lend themselves to angiogenic scoring. This applies primarily to elaborate vascular structures forming a reticulated pattern. A comparable observation on the existence of complex patterns of tumour vessels in breast carcinomas was recently reported and illustrated.51 In previous papers on angiogenesis of invasive carcinoma, the issue of vascular patterns has, however, not been directly addressed, although a comment on branching structures counted as single vessels has been made.31 Generally any labelled structure, with the exception of larg-

Microvessel density in breast carcinoma 255 er native vessels, are counted4 with the noteworthy modification mentioned in some papers1,19,33 that any vessel which is isolated from other vessels is recorded. Though not directly mentioned, complex vascular structures were apparently not included in these scorings. Confluent vascular structures comparable to the maze-like pattern described in this paper were reported in one study of intraductal breast carcinoma and were omitted in quantification.24 In this way reproducibility studies should yield more acceptable results. We are, however, uncertain whether such scoring technique, limited to easily countable areas will prove a valid prognostic information. When correlating MVD of random zones with MVD in hot spot areas an unexpected observation emerged. In comparing our scoring one would have expected that values in the hot spot study would equal or exceed those of the scoring of random fields. This was, however, not consistently the case. One could claim that a more meticulous hot-spot analysis would produce results closer to reality. The view that we, during hot spot searching, subconsciously omitted reticulated vascular zones, which are difficult to count, but probably represent the true hot spots, should, however, be considered. Incidentally, this complex vascular pattern can easily be conceived as a result of an arterial type of growth, involving dichotomous branching and budding of the endothelium, as has previously been found to characterise neovasculature of invasive breast cancer.1 In conclusion, we found significant observer discrepancy in the morphological assessment of angiogenesis, that was attributable to difficulties both in identifying the hot spots and in performing the counting of v-p per se, which should be taken into consideration when alternative histological techniques in the evaluation of MVD are sought. It is conceivable that the reproducibility could be enhanced by confining the scoring to areas with well defined v-p in optimally immunostained sections and in particular by omitting areas with complex tortuous vascular structures. One may, however, question whether this is an acceptable way to circumvent this problem. The difficulties associated with v-p scoring that have been delineated may in part explain why some observers have failed to confirm the prognostic value of MVD.25–31 References 1. Weidner N, Weinstein R, eds. Advances in Pathology and Laboratory Medicine. The relationship of tumor angiogenesis and metastasis with emphasis on invasive breast carcinoma. Chicago: Mosby, 1992; p. 101–22. 2. Folkman J, Watson K, Ingber D, Hanahan D. Induction of angiogenesis during the transition from hyperplasia to neoplsia. Nature 1989; 339: 58–61. 3. Mahadevan V, Hart I R. Metastasis and angiogenesis. Acta Oncol 1990; 29: 97–103. 4. Bosari S, Lee A K C, De Lellis R A, Wiley B D, Heatley G J,

5. 6. 7. 8.

9.

10. 11.

12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

23. 24. 25. 26.

Silverman M L. Microvessel quantitation and prognosis in invasive breast carcinoma. Human Pathol 1992; 23: 755–61. Brown L F, Berse B, Jackman R W et al. Expression of vascular permeability factor (vascular endot-helial growth factor) and its receptors in breast cancer. Human Pathol 1995; 26: 86–91. Bussolini F, di Renzo M F, Ziche M et al. Hepatocyte growth factor is a potent angiogenic factor which stimulates endothelial cell motility and growth. I Cell Biol 1992; 119: 629–41. Schreiber A B, Winkler M E, Derynck R. Transforming growth factor- : a more potent angiogenic mediator than epidermal growth factor. Science 1986; 232: 1250–3. Duffy M J, Reilly D, McDermott E, O’Higgins N, Fennelly J J, Andreasen P A. Urokinase plasminogen activator as a prognostic marker in different subgroups of patients with breast cancer. Cancer 1994; 74: 2276–80. Grøndahl-Hansen J, Christensen I J, Rosenquist C et al. High levels of urokinase-type plasminogen activator and its inhibitor PAI-1 in cytosolic extracts of breast carcinomas are associated with poor prognosis. Cancer Research 1993; 53: 2513–21. Jänicke F, Schmitt M, Hofter R et al. Urokinase-type plasminogen activator (u-PA) antigen is a predictor of early relapse in breast cancer. Fibrinolysis 1990; 4: 69–78. Gasparini G, Barbareschi M, Boracchi et al. 67-kDa laminin-receptor expression adds prognostic information to intra-tumoral microvessel density in node-negative breast cancer. Int J Cancer 1995; 60: 604–10. Toi M, Kashitani J, Tominaga J. Tumor angiogenesis is an independent prognostic indicator in primary breast carcinoma. Int J Cancer 1993; 55: 371–4. Weidner N. Tumor angiogenesis: Review of current applications in tumor prognostication. Sem Diagn Pathol 1994; 10: 302–13. Bigler S A, Brawer M K, Deering R E. Neovascularization in carcinoma of the prostate: a quantitative morphometric study. Mod Pathol 1992; 5: 50A. Wakui S, Furusato M, Itoh T. Tumor angiogenesis in prostatic carcinoma with and without bone marrow metastasis: a morphometric study. J Pathol 1992; 168: 257–62. Gasparini G, Weidner N, Bevilacqua P. Tumor angiogenesis and p53 protein: correlation with metastasis in head and neck squamous cell carcinoma. Int J Cancer 1993; 55: 739–44. Macchiarini P, Fontoini G, Hardin M J. Relation of neovasculature to metastasis of non-small-cell lung cancer. Lancet 1992; 340: 145–6. Bennett S, Engel C J, O’Malley F, Doig G, Chambers A. Tumor angiogenesis predicts metastasis in invasive colorectal cancer. [Abstract] Lab Invest 1995; 7257A. Saclarides T J, Speziale N J, Drab E, Szeluga D J, Rubin D B. Tumor angiogenesis and rectal carcinoma. Dis Colon Rectum 1994; 37: 921–6. Sanz-Ortega J, Moro E, Stenberg S, Sanz-Esponera J, Merino M J. Angiogenesis as an independent prognostic factor in gastric carcinoma. [Abstract] Lab Invest 1995; 7268A. Bellamy C O C, McDonald C, Salter D M, Chetty U, Andersen T J. Noninvasive ductal carcinoma of the breast: the relevance of histologic categorization. Human Pathol 1993; 24: 16–23. Ottesen G L, Graversen H P, Blichert-Toft M, Zedeler K, Andersen J A. Ductal carcinoma in situ of the female breast – short-term results of a prospective nationwide study. Am J Surg Pathol 1992; 16: 1183–96. Bose S, Lesser M L, Norton L, Rosen P. Immunophenotype of intraductal carcinoma. Arch Pathol Lab Med 1996; 120: 81–5. Guidi A J, Fischer L, Harris J R, Schnitt S J. Microvessel density and distribution in ductal carcinoma in situ of the breast. J Natl Cancer Inst 1994; 86: 614–9. Axelsson K, Ljung B-M, Moore D. Tumor angiogenesis as a prognostic assay for invasive ductal breasr carcinoma. J Natl Cancer Inst 1995; 87: 997–1008. Costello P, McCann A, Carney D N, Dervan P. Lack of prognostic significance of microvessel density in lymph node negative breast carcinoma. [Abstract] Lab Invest 1995; 7216A.

256

The Breast

27. Goulding H, Nik Abdul Rashid N F, Robertson J F et al. Assessment of angiogenesis in breast carcinoma. Hum Pathol 1995; 26: 1196–200. 28. Hall N R, Fish D E, Hunt N, Goldin R D, Guillou P J, Monson J R T. Is the relationship between angiogenesis and metastasis in breast cancer real? Surg Oncol 1992; 1: 223–9. 29. Marley E, Frost A, Tabbara S. The significance and reproducibility of microvessel quantitation in infiltrating lobular carcinoma (ILC) of the breast. [Abstract] Lab Invest 1995; 7221A. 30. Tan L K, Lesser M L, Arroyo C D, Rosen P P. Intratumoral microcvessel density (MVD) and prognosis in breast carcinoma. [Abstract] Lab Invest 1995; 7226A. 31. van Hoef M E H M, Knox W F, Dhesi S S, Howell A, Schnor A M. Assessment of tumor vascularity as a prognostic factor in lymph node negative invasive breast cancer. Eur J Cancer 1993; 29A: 1141–5. 32. Horak E R, Leek R, Klenk N et al. Angiogenesis, assessed by platelet/endothelial cell adhesion molecule antibodies, as an indicator of node metastases and survival in breast cancer. Lancet 1992; 340: 1120–4. 33. Weidner N, Semple J P, Welch W R, Folkman J. Tumor angiogenesis and metastasis – correlation in invasive breast carcinoma. NEJM 1991; 324: 1–8. 34. Yadrandji S, Dolan P, Visscher D W. Relationship between microvessel density and area in breast carcinomas – an image morphometric study. [Abstract] Lab Invest 1995; 7228A. 35. Gilchrist K W, Kalish L, Gould V E et al. Interobserver reproducibility of histopathological feature in stage II breast cancer – an ECOG study. Breast Cancer Research Treatment 1985; 5: 3–10. 36. Pedersen L, Holck S, Schi¢dt T, Zedeler K, Mouridsen H T. Interand intraobserver variability in the histopathological diagnosis of medullary carcinoma of the breast, and its pro- gnostic implications. Breast Cancer Research Treatment 1989; 14: 91–9. 37. Barbareschi M, Weidner N, Gasparini G. Microvessel density quantification in breast carcinoma. Assessment by light microscopy vs. a computer-aided image analysis system. Appl Immunohistochem 1995; 3: 75–84. 38. De Jong J S, van Diest P J, Baak J P A. Heterogeneity of microvessel counts in breast cancer: a methodological study. [Abstract] 8th Int Symposium on diagnostic quantitative pathology 1994; 28. 39. Klijanienko J, El-Naggar A K, de Braud F et al. Tumor vascularization, mitotic index, histopathologic grade and DNA

40.

41. 42. 43.

44.

45. 46. 47. 48. 49.

50.

51.

ploidy in the assessment of 114 head and neck squamous carcinomas. Cancer 1995; 75: 1649–56. Morelli L, Barbareschi M, Gasparini G et al. Angiogenesis in human breast carcinoma: a comparison between human and computer assisted microvessel quantification. [Abstract] 8th Int Symposium on diagnostic quantitative pathology 1994; 29. Visscher D W, Smilanetz S, Drozdowicz S. Prognostic significance of image morphometric microvessel enumeration in breast carcinoma. Analytical Quant Cytol Histol 1993; 15: 88–92. Miliaras D, Kamas A, Kalekou H. Angiogenesis in invasive breast carcinoma: is it associated with parameters of prognostic significance? Histopathol 1995; 26: 165–9. Vartanian R K, Weidner N. Correlation of intratumoral endothelial cell proliferation with microvessel density (tumour angiogenesis) and tumor cell proliferation in breast carcinoma. Am J Pathol 1994; 114: 1188–94. Fox S B, Leek R, Weekes M P, Whitehouse R, Gatter K C, Harris A L. Quantitation and prognostic value of breast cancer angiogenesis: Comparison of microvessel density, Chalkley count, and computer image analysis. J Pathol 1995; 117: 275–83. Weidner N. Tumor angiogenesis: review of current applications in tumor prognostication. Sem Diagn Pathol 1993; 10: 302–13. Weidner N, Carroll P R, Flax J, Blumenfeld W, Folkman J. Tumor angiogenesis correlates with metastasis in invasive prostate carcinoma, Am J Pathol 1993; 143: 401–9. Toi M, Kashitani J, Tominaga T. Tumor angiogenesis is an independent prognostic indicator in primary breast carcinoma. Int J Cancer 1993; 55: 371–4. Weidner N, Folkman J, Pozza F et al. Tumor angiogenesis: A new significant and independent prognostic indicator in early-stage breast carcinoma. J Natl Cancer Inst 1992; 84: 1875–87. Vermeulen P B, Gasparini G, Fox S B et al. Quantification of angiogenesis in solid tumours: an international consensus on the methodology and criteria of evaluation. Eur J Cancer 1996; 32A: 2474–84. van Diest P J, Baak J P A, Matze-Cok P et al. Reproducibility of mitosis counting in 2,469 breast cancer specimens: Results from a multicenter morphometric mammary carcinoma project. Human Pathol 1992; 23: 603–7. De la Torre M, Nygren P, Larsson R, Lindgren A, Tsuruo T, Bergh J. Expression of the 85-kd membrane protein in primary human breast cancer: relationship to hormone receptor levels, DNA ploidy, and tumor grade. Human Pathol 1995; 26: 180–5.