Search tips
Search criteria 


Logo of jclinpathJournal of Clinical PathologyVisit this articleSubmit a manuscriptReceive email alertsContact usBMJ
J Clin Pathol. 2003 April; 56(4): 271–276.
PMCID: PMC1769930

There is more than one kind of myofibroblast: analysis of CD34 expression in benign, in situ, and invasive breast lesions


Aims: Smooth muscle actin (SMA) positive myofibroblasts have been implicated in tumour invasion; however, acquisition of SMA is not limited to peritumorous fibroblasts and other changes in fibroblasts may be more specifically related to the malignant environment. CD34 is a sialomucin expressed by normal breast fibroblasts but lost in invasive carcinomas. The aim of this study was to establish the relation between CD34 and SMA expression in breast fibroblasts and to analyse whether loss of CD34 is specific for invasive disease.

Methods: Immunohistochemistry for CD34 and SMA was performed on 135 cases including 10 normal, 10 fibroadenomas, 40 infiltrating ductal carcinomas, 55 cases of ductal carcinoma in situ (DCIS), and 20 radial scar/complex sclerosing lesions. The relation between staining pattern and histopathological features was recorded as positive, negative, or reduced.

Results: Fibroblasts around all normal duct–lobule units and those showing epithelial hyperplasia were CD34 positive and mainly SMA negative. In fibroadenomas, fibroblasts retained CD34 but acquired SMA expression. In contrast, fibroblasts around invasive carcinoma were CD34 positive and SMA negative. In DCIS, loss of CD34 was significantly more frequent in high grade tumours than in low or intermediate grade ones (p < 0.001). The acquisition of SMA was seen more frequently than the loss of CD34, particularly in non-high grade DCIS. In all radial scars, fibroblasts were SMA positive but CD34 negative, and a similar pattern was seen in stromal cells in areas of fibrosis following core biopsy.

Conclusions: These results show that SMA positive myofibroblasts exhibit variable expression of CD34, indicating that these markers are not coordinately controlled. Loss of CD34 is strongly related to the malignant phenotype, in both invasive and preinvasive disease, but is not entirely specific because radial scar fibroblasts and fibroblasts in reactive fibrosis exhibit a similar phenotype. The functional relevance of altered CD34 expression is unclear but the very focal changes implicate local signalling mechanisms probably of epithelial origin.

Keywords: CD34, breast, myofibroblast

There is increasing evidence to indicate that the tumour microenvironment exerts a major modulatory effect on epithelial tumours and that the stroma makes an important contribution to the process of tumour progression.1–4 In vitro studies have demonstrated for many tumours that tumour cell–fibroblast interactions enhance tumour cell invasion,5–7 and in vitro and tissue studies indicate that this may be mediated at least in part through the stimulation of proteolytic enzymes,8–11 with the fibroblasts being the major source of many of these enzymes, such as the matrix metalloproteinases (MMPs).12,13 The suggestion that stromal changes may be important in tumour progression is also reflected in the identification of loss of heterozygosity in the stromal compartment of tumours, some of which may be expected to result in functional changes.14 In several systems, the stroma has been shown to have a profound effect on normal and tumour cell behaviour,15–17 and it has been suggested that stroma may play a dominant role in regulating breast epithelial cell function.18

The stroma around invasive breast tumours is known to differ from normal breast, with alterations in stromal protein synthesis19 and expression of MMP.20,21 Many of these features are attributed to activated fibroblasts, termed myofibroblasts, reflecting their acquisition of α smooth muscle actin (αSMA) expression. However, myofibroblastic differentiation is not specific for malignancy, being characteristic of stromal cells in healing wounds22,23 and also being seen in benign lesions of the breast.24 Furthermore, the term myofibroblast implies a homogeneous cell population, but there is increasing recognition of the heterogeneity of fibroblast populations,25–27 and it is therefore highly likely that not all myofibroblasts are the same, and there may be other aspects of the fibroblast phenotype that more closely reflect their functional differentiation.

“It has been suggested that there is an inverse relation between CD34 expression and myofibroblastic differentiation”

CD34 is a transmembrane, highly glycosylated protein expressed by haemopoietic stem/progenitor cells (HPSCs),28 endothelial cells,29 and mesenchymal cells at several different sites, including breast.30 On endothelial cells, CD34 acts as a ligand for L-selectin,31 although this is not the case in HPSCs, where its function remains unclear. The CD34 positive fibroblast in many organs is thought to represent an uncommitted cell capable of multidirectional mesenchymal differentiation,32 and it has been suggested that there is an inverse relation between CD34 expression and myofibroblastic differentiation.24 Whereas the normal mammary stroma comprises CD34 positive spindle cells, CD34 expression is lost on the fibroblasts surrounding invasive carcinomas,33 although it is retained by endothelial cells, making it a useful marker for studies of angiogenesis.34

The aim of our study was to examine the relation between CD34 and αSMA expression in breast fibroblasts, and to investigate whether the loss of CD34 expression is a more sensitive or specific marker of malignancy in the breast than αSMA expression, with particular emphasis on expression patterns in preinvasive and at risk lesions.


Immunohistochemistry for CD34 and αSMA was performed on serial sections taken from normal breast (n = 10), fibroadenoma (n = 10), infiltrating ductal carcinoma (n = 40), ductal carcinoma in situ (n = 55), and radial scar/complex sclerosing lesions (n = 20). A standard avidin biotin complex technique, without prior antigen retrieval, was used incorporating 10 minutes’ incubation in 60% H2O2 to block endogenous peroxidase. The primary antibodies were diluted in Dako antibody diluent (Qbend-10; Dako, Ely, Cambridgeshire, UK) at 1/100 for CD34 and 1/600 for αSMA. Staining of blood vessels was used as an internal positive control for CD34 and of normal myoepithelial cells for αSMA. For each case, serial sections were also stained with haematoxylin and eosin and in a proportion of cases serial sections were stained with anti-CD31 (Dako) as a further marker of endothelial cells.

Fibroblast reactivity for CD34 and αSMA was recorded as positive, negative, and—where there was focal loss—as reduced. Because many of the sections contained ducts and lobules showing a range of histopathological features, the relation between staining pattern and pathology was interpreted for each duct–lobular unit. Thus, in addition to 40 infiltrating ductal carcinomas (IDCs) and 10 fibroadenomas, 301 normal duct–lobular units were scored, 535 ducts with ductal carcinoma in situ (DCIS; 60 low grade, 77 intermediate, and 398 high grade), six cases showing microinvasion, 88 duct–lobular units exhibiting usual type epithelial hyperplasia (HUT), six with atypical ductal hyperplasia (ADH), 12 lobular carcinoma in situ (LCIS), 11 apocrine cysts, and 24 radial scars, four of which contained areas of reactive fibrosis following preoperative core biopsy (table 11).

Table 1
Summary of the pattern of expression of CD34 and α smooth muscle antigen (αSMA) in fibroblasts in relation to histopathological features

The relation between the staining pattern and the different histological features was compared using the χ2 test, and p < 0.05 was considered significant.


In the normal breast, both intralobular and extralobular fibroblasts exhibited strong uniform staining for CD34 and were αSMA negative. This pattern was maintained in the presence of other pathology, including adjacent carcinoma; thus, in all 301 normal duct–lobular units fibroblasts were CD34 positive and, with the exception of 23 cases, αSMA negative (fig 1A,B1A,B).). Similarly, fibroblasts around glands showing HUT, LCIS, or apocrine cyst formation were CD34 positive/αSMA negative (fig 1C,D1C,D),), with the exception of focal αSMA positivity around nine of 79 ducts with epithelial hyperplasia. Fibroadenomas showed a homogeneous pattern of staining, with fibroblasts being strongly CD34 positive and αSMA positive. In contrast, all IDCs and foci of microinvasion exhibited consistent loss of CD34 expression on fibroblasts, but acquisition of αSMA (fig 1E,F1E,F).). Discrete staining for CD34 was evident on small calibre blood vessels within the stroma, confirmed by their expression of the vascular antigen CD31.

Figure 1
Fibroblasts are CD34 positive around (A) normal breast ducts and lobules and (C) hyperplasia of usual type, but are negative for α smooth muscle actin (αSMA) (B, D). Small calibre blood vessels are also CD34 positive (arrowhead in A and ...

The most variable pattern of expression was seen on fibroblasts surrounding ducts with DCIS; however, the loss of CD34 expression was significantly more frequent in high nuclear grade compared with low or intermediate grade samples (p < 0.001). In high grade DCIS, fibroblasts were CD34 negative in 78% cases compared with 29 of 77 (33%) intermediate and 21 of 60 (35%) cases of low grade DCIS (table 11;; fig 22).). In some cases, partial loss of CD34 was evident around a duct. The incidence of reduced staining was again variable between the grades; however, only 12% of cases of high grade DCIS retained normal CD34 expression compared with 22 of 77 (29%) intermediate grade and 28 of 60 (47%) low grade cases. Staining of serial sections of 20 of these cases with CD31 demonstrated clearly the pattern of small blood vessel formation around the ducts, and these vessels were easily distinguishable from the fibroblast population. In most cases, loss of CD34 expression was accompanied by the acquisition of αSMA; however, this relation was not absolute, with gain of αSMA being more frequent than loss of CD34 (table 11),), and only 27% of periductal fibroblasts in low grade DCIS were αSMA negative, compared with 47% retaining expression of CD34. Again, a mixed pattern of staining was evident in relation to ADH, with a proportion of CD34 negative fibroblasts in half the cases; however, in none of the six cases was there evidence of αSMA staining.

Figure 2
In ducts displaying features of high grade DCIS, 78% were associated with (A, C) CD34 negative periductal fibroblasts and (B, D) the acquisition of α smooth muscle actin (αSMA). (C) Black arrows indicate retained CD34 on small blood vessels ...

In the radial scars, most fibroblasts were CD34 negative/αSMA positive (fig 33),), although some cases displayed CD34 positive/αSMA positive fibroblasts at the periphery of the lesion. The pattern was similar in those radial scars admixed with DCIS (n = 4) and those not associated with malignant change (n = 20). Interestingly, in four cases, areas of reactive fibrosis were evident from preoperative core biopsy, and in these sites the fibroblasts were uniformly CD34 negative/αSMA positive (fig 33).

Figure 3
The predominant pattern in radial scars was (A, C) loss of CD34 expression associated with (B, C) acquisition of α smooth muscle actin (αSMA) expression. Strong αSMA expression was also evident around the glandular structures in ...


CD34 is a transmembrane glycoprotein that is thought to be involved in the modulation of cell adhesion and signal transduction,28 and is expressed by mesenchymal cells at several sites, including the normal mammary stroma.30,33 Loss of CD34 by mesenchymal cells has been described in several situations where there is malignant transformation of the mesenchymal population. Malignant phyllodes tumours of the breast exhibit lower levels of CD34 expression than benign phyllodes or fibroadenomas,24,33,35 and CD34 is lost in sarcomas arising within CD34 positive dermatofibrosarcoma protruberans.36 Loss of fibroblast CD34 has also been described in non-neoplastic fibroblasts around epithelial tumours, such as basal cell carcinoma37 and colorectal carcinoma.38

“Loss of CD34 may be related to invasive potential”

Our study has shown that fibroblast CD34 expression is consistently lost in invasive breast carcinomas that include microinvasion, and in a high proportion of cases of DCIS, particularly high grade lesions, which are thought to be more likely to progress to invasion.39 Loss of expression is also seen in a proportion of ADH, but not around glandular structures showing LCIS. This is of interest because ADH and DCIS are regarded as premalignant lesions, whereas LCIS is considered to confer an increased risk of developing carcinoma, but the risk relates to the development of carcinoma in either breast, not to the site of the LCIS.40 This raises the possibility that loss of CD34 may be related to invasive potential. The change in CD34 expression is very localised, with loss around a duct containing DCIS, but retained expression around adjacent normal breast glands, and this strongly implicates epithelial–mesenchymal interactions in the control of expression, as has been suggested previously in the case of phyllodes tumours.33 What determines the loss of CD34 is of interest because not all DCIS cases show loss, and this may point to different functional states of the neoplastic epithelium, which are important in determining their invasive potential. However, whereas it has previously been suggested that the loss of CD34 is specific to malignancy,41 we show that this is not the case. Although most benign lesions, including HUT, retain CD34 expression, loss of CD34 was a consistent finding in the fibroblasts associated with areas of fibrosis following core biopsy and in radial scars; in this last case, whether they were associated with malignancy or not. Although on the surface this makes CD34 less valuable as a diagnostic marker of malignancy—particularly in sclerotic lesions—it raises interesting questions as to the state of differentiation of fibroblasts in different breast lesions, and the potential relevance of this in terms of fibroblast function. Some studies suggest that radial scars are associated with an increased risk of development of malignancy,42 and several other studies have shown similar alteration in the stroma of radial scars as seen in invasive carcinomas, such as increased hyaluronic acid43 and increased expression of ED-A fibronectin and vascular endothelial growth factor.44 The importance of loss of expression of CD34 on stromal cells in reactive fibrosis is unclear, although it could possibly be related to the terminal differentiation of these cells.

Take home messages

  • Smooth muscle actin positive myofibroblasts express CD34 to varying degrees, indicating that these markers are not coordinately controlled
  • The loss of CD34 is strongly related to the malignant phenotype, in both invasive and preinvasive disease, but is not entirely specific because radial scar fibroblasts and fibroblasts in reactive fibrosis exhibit a similar phenotype
  • The functional relevance of altered CD34 expression is unclear but the very focal changes implicate local signalling mechanisms, probably epithelial in origin
  • In vitro studies should be undertaken to establish the role of CD34 in the breast fibroblast

“It is essential that the changes in fibroblast phenotype associated with malignancy are carefully dissected, not only to validate their use as possible diagnostic markers, but also to establish their potential as therapeutic targets”

The importance of changes in fibroblast function in promoting tumour progression is increasingly recognised and such a tumour promoting effect is frequently attributed to the activation of fibroblasts to αSMA positive myofibroblasts. However, our study clearly demonstrates that acquisition of the myofibroblast phenotype is not indicative of malignancy in the breast, and it also demonstrates that not all myofibroblasts have the same phenotype, some being CD34 positive, as in fibroadenomas, and others being CD34 negative, as in invasive and many in situ lesions. The potential for the stromal response to tumours becoming a target for treatment is increasing, with clinical trials in place for fibroblast derived factors such as MMP45 and Tenascin.46 However, it is essential that the changes in fibroblast phenotype associated with malignancy are carefully dissected, not only to validate their use as possible diagnostic markers, but also to establish their potential as therapeutic targets.

The functional relevance of fibroblast CD34 expression and its loss is not clear, although it may represent a change from a multipotent mesenchymal cell to a committed cell type. However, the variable association between CD34 and αSMA expression demonstrates that not all myofibroblasts are the same, and that the loss of CD34 appears to be more closely related, although clearly not specific, to malignancy than the acquisition of αSMA. This suggests that αSMA may not be the most important marker of fibroblast function, and in vitro studies are now indicated to establish the role of CD34 in the breast fibroblast.


  • ADH, atypical ductal hyperplasia
  • DCIS, ductal carcinoma in situ
  • HPSC, haemopoietic stem/progenitor cell
  • HUT, usual type epithelial hyperplasia
  • IDC, infiltrating ductal carcinoma
  • LCIS, lobular carcinoma in situ
  • MMP, matrix metalloproteinase
  • SMA, smooth muscle actin


1. Ronnov-Jessen L, Petersen OW, Bissell MJ. Cellular changes involved in conversion of normal to malignant breast: importance of the stromal reaction. Physiol Rev 1996;76:69–125. [PubMed]
2. Noel A, Foidart JM. The role of stroma in breast carcinoma growth in vivo. J Mammary Gland Biol Neoplasia 1998;3:215–25. [PubMed]
3. Martin M, Pujuguet P, Martin F. Role of stromal myofibroblasts infiltrating colon cancer in tumour invasion. Pathol Res Pract 1996;192:712–17. [PubMed]
4. Tuxhorn JA, Ayala GE, Rowley DR. Reactive stroma in prostate cancer progression. J Urol 2001;166:2472–83. [PubMed]
5. Popowicz P, Engel G, Wilking N, et al. Differential effects of fibroblast coculture on MCF-7 and MDA-MB-231 breast carcinoma cell invasion through Matrigel. Oncol Rep 1995;2:303–6.
6. Saiki I, Murata J, Yoneda J, et al. Influence of fibroblasts on the invasion and migration of highly or weakly metastatic B16 melanoma cells. Int J Cancer 1994;56:867–73. [PubMed]
7. Dimanche-Boitrel MT, Vakaet L, Pujuguet P, et al. In-vivo and in-vitro invasiveness of a rat colon-cancer cell-line maintaining E-cadherin expression—an enhancing role of tumour-associated myofibroblasts. Int J Cancer 1994;56:512–21. [PubMed]
8. Lengyel E, Gum R, Juarez J, et al. Induction of MR-92,000 type-IV collagenase expression in a squamous cell carcinoma cell line by fibroblasts. Cancer Res 1995;55:963–7. [PubMed]
9. Ornstein DL, MacNab J, Cohn KH. Evidence for tumour–host cooperation in regulating MMP-2 expression in human colon cancer. Clin Exp Met 1999;17:205–12.
10. Polette M, Gilles C, Marchand V, et al. Induction of membrane-type matrix metalloproteinase 1 (MT1-MMP) expression in human fibroblasts by breast adenocarcinoma cells. Clin Exp Metastasis 1997;15:157–63. [PubMed]
11. Boyd RS, Balkwill FR. MMP-2 release and activation in ovarian carcinoma: the role of fibroblasts. Br J Cancer 1999;80:315–21. [PMC free article] [PubMed]
12. Poulsom R, Hanby AM, Pignatelli M, et al. Expression of gelatinase-A and TIMP-2 messenger-RNAs in desmoplastic fibroblasts in both mammary carcinomas and basal cell carcinomas of the skin. J Clin Pathol 1993;46:429–36. [PMC free article] [PubMed]
13. Heppner KJ, Matrisian LM, Jensen RA, et al. Expression of most matrix metalloproteinase family members in breast cancer represents a tumour-induced host response. Am J Pathol 1996;149:273–82. [PubMed]
14. Kurose K, Hoshaw-Woodard S, Adeyinka A, et al. Genetic model of multi-step breast carcinogenesis involving the epithelium and stroma: clues to tumour–microenvironment interactions. Hum Mol Genet 2001;10:1907–13. [PubMed]
15. Cooke PS, Buchanan DL, Young P, et al. Stromal estrogen receptors mediate mitogenic effects of estradiol on uterine epithelium. Proc Natl Acad Sci U S A 1997;94:6536–40.
16. Silberstein GB. Tumour–stromal interactions: role of the stroma in mammary development. Breast Cancer Res 2001;3:218–23.
17. Tuxhorn JA, McAlhany SJ, Dang TD, et al. Stromal cells promote angiogenesis and growth of human prostate tumours in a differential reactive stroma (DRS) xenograft model. Cancer Res 2002;62:3298–307. [PubMed]
18. Shekhar MPV, Werdell J, Santner SJ, et al. Breast stroma plays a dominant regulatory role in breast epithelial growth and differentiation: implications for tumour development and progression. Cancer Res 2001;61:1320–6. [PubMed]
19. Adams M, Jones JL, Walker RA, et al. Changes in Tenascin-C isoform expression in invasive and pre-invasive breast disease. Cancer Res 2002;62:3289–97. [PubMed]
20. Basset P, Bellocq JP, Wolf C, et al. A novel metalloproteinase gene specifically expressed in stromal cells of breast carcinomas. Nature 1990;348:699–704 [PubMed]
21. Jones JL, Glynn P, Walker RA. Expression of MMP-2 and MMP-9, their inhibitors and the activator MT1-MMP in primary breast carcinomas. J Pathol 1999;189:161–8. [PubMed]
22. Gabbiani G, Hirschel BJ, Ryan GB, et al. Granulation tissue as a contractile organ: a study of structure and function. J Exp Med 1972;135:719–34. [PMC free article] [PubMed]
23. Grinnell F. Fibroblasts, myofibroblasts and wound contraction. J Cell Biol 1994;124:401–4. [PMC free article] [PubMed]
24. Silverman JS, Tamsen A. Mammary fibroadenomas and some phyllodes tumour stroma are composed of CD34+ fibroblasts and factor XIIIa+ dendrophages. Histopathology 1996;29:411–19. [PubMed]
25. Schor SL, Schor AM. Clonal heterogeneity in fibroblasts: implications for the control of epithelial–mesenchymal interactions. Bioessays 1987;7:200–4. [PubMed]
26. Gleave M, Hsieh J-T, Gao C, et al. Acceleration of human prostate cancer growth in vivo by factors produced by prostate and bone fibroblasts. Cancer Res 1991;51:3753–61. [PubMed]
27. Lindahl P, Betscholtz C. Not all myofibroblasts are alike: revisiting the role of PDGF-A and PDGF-B using PDGF targeted mice. Curr Opin Nephrol Hypertens 1998;7:21–6. [PubMed]
28. van de Rijn M, Rouse RV. CD34. A review. Appl Immunohistochem 1994;2:71–80.
29. Fina L, Molgaard HV, Robertson D, et al. Expression of the CD34 gene in vascular endothelial cells. Blood 1990;12:2417–26.
30. Yamazaki K, Eyden BP. Ultrastructural and immunohistochemical observations on intralobular fibroblasts of human breast with observations on the CD34 antigen. J Submicrosc Cytol Pathol 1995;27:309–23. [PubMed]
31. Lanza F, Healy L, Sutherland DR. Structural and functional features of the CD34 antigen: an update. J Biol Regul Homeost Agents 2001;15:1–13. [PubMed]
32. Magro G, Michal M, Bisceglia M. Benign spindle cell tumours of the mammary stroma: diagnostic criteria, classification and histogenesis. Pathol Res Pract 2001;197:453–66. [PubMed]
33. Moore T, Lee AHS. Expression of CD34 and bcl-2 in phyllodes tumours, fibroadenomas and spindle cell lesions of the breast. Histopathology 2001;38:62–76. [PubMed]
34. Lee AH, Dublin EA, Bobrow LG. Angiogenesis and expression of thymidine phosphorylase by inflammatory and carcinoma cells in ductal carcinoma in situ of the breast. J Pathol 1999;187:285–90. [PubMed]
35. Chen CM, Chen CJ, Chang CL, et al. CD34, CD117 and actin expression in phyllodes tumour of the breast. J Surg Res 2000;94:84–91. [PubMed]
36. Goldblum JR, Reith JD, Weiss SW. Sarcomas arising in dermatofibrosarcoma protruberans: a reappraisal of biologic behaviour in eighteen cases treated by wide local excision with extended clinical follow up. Am J Surg Pathol 2000;24:1125–30. [PubMed]
37. Kirchmann TT, Prieto VG, Smoller BR. Use of CD34 is assessing the relationship between stroma and tumour in desmoplastic keratinocyte neoplasms. J Cutan Pathol 1995;22:422–6. [PubMed]
38. Nakayama H, Enzan H, Miyazaki E, et al. Differential expression of CD34 in normal colorectal tissue, peritumoural inflammatory tissue and tumour stroma. J Clin Pathol 2000;53:626–9. [PMC free article] [PubMed]
39. Lagios MD. Heterogeneity of duct carcinoma in situ (DCIS): relationship of grade and subtype analysis to local recurrence and risk of invasive transformation. Cancer Lett 1995;90:97–102. [PubMed]
40. Page DL, Kidd TE, Jr, Dupont WD, et al. Lobular neoplasia of the breast: higher risk for subsequent invasive cancer predicted by more extensive disease. Hum Pathol 1991;22:1232–9. [PubMed]
41. Barth PJ, Ebrahimsade S, Ramaswamy A, et al. CD34+ fibrocytes in invasive ductal carcinoma, ductal carcinoma in situ and benign breast lesions. Virchows Arch 2002;440:298–303. [PubMed]
42. Jacobs TW, Byrne C, Colditz G, et al. Radial scars in benign breast biopsy specimens and risk of breast cancer. N Engl J Med 1999;340:430–6. [PubMed]
43. de la Torre M, Wells AF, Bergh J, et al. Localisation of hyaluronan in normal breast tissue, radial scar, and tubular breast carcinoma. Hum Pathol 1993;24:1294–7. [PubMed]
44. Jacobs TW, Schnitt SJ, Tan X, et al. Radial scars of the breast and breast carcinomas have similar alterations in expression of factors involved in vascular stroma formation. Hum Pathol 2002;33:29–38. [PubMed]
45. Parsons SL, Watson SA, Griffin NR, et al. An open phase I/II study of the oral matrix metalloproteinase inhibitor marimastat in patients with inoperable gastric cancer. Ann Oncol 1996;7:47.
46. Cokgor I, Akabani G, Kuan CT, et al. Phase I trial results of iodine 131-labelled anti-tenascin monoclonal antibody 81C6 treatment of patients with newly diagnosed malignant gliomas. J Clin Oncol 2000;18:3862–72. [PubMed]

Articles from Journal of Clinical Pathology are provided here courtesy of BMJ Publishing Group