We used the combination of the lymph endothelium-specific marker D2-40 and the panendothelial marker CD34 to detect and distinguish between LVI and BVI. LVI and BVI were found in 69.5 and 37.9% of patients, respectively. Other authors report LVI and especially BVI in BC to be less frequent: prevalences of LVI and BVI range from 8.8 (Nime et al, 1977
) to 86% (Kahn and Marks, 2002
) and from 4.2 (Lauria et al, 1995
) to 33% (Kato et al, 2002
), respectively. The presence and extent of LVI and BVI correlated with the size of the tumour and only the extent of LVI correlated with the number of blocks investigated. Therefore, the high number of FFPE tissue blocks that was investigated per resection specimen cannot fully explain the increased frequency of LVI and BVI in this study. On the contrary, the use of IHC with endothelial markers to detect LVI and BVI accounts for their high prevalence. Immunohistochemistry is a very sensitive and specific technique. In our study, the presence of LVI or BVI was missed on HE in 71.0 and 67.9% of specimens intratumourally and in 20–25% of specimens peritumourally. Furthermore, also the extent of LVI and BVI was underestimated on HE. Differences in endothelial markers and in study populations further lead to differences in LVI and BVI prevalence. For LVI, the highest prevalences are found in the studies with D2-40 and podoplanin. Kahn et al
demonstrated lymphatic invasion in 44% of LN negative and 86% of LN positive (overall 66%) BC patients (Kahn and Marks, 2002
). Recently, it has been shown that the D2-40 antibody specifically recognises podoplanin (Schacht et al, 2005
) and that both antibodies against podoplanin and D2-40 can be regarded as reference standards for the identification of lymphatic vessels in most settings. Both anti-podoplanin and D2-40 antibodies have a sensitivity and specificity for lymphatic endothelium of over 95% (Evangelou et al, 2005
). We previously showed that in BC, D2-40 is the best marker for lymphatic endothelium (Van der Auwera et al, 2005
). Most studies investigating BVI in BC used FVIII or a combination of FVIII and van Gieson elastica staining. Nevertheless, CD34 or CD31 is prefered to visualise blood vessels in solid tumours (Vermeulen et al, 1996
), as FVIII expression is absent in some tumour capillaries that can lead to underestimation of the presence and extent of BVI (Vermeulen et al, 1995
). As CD34 is not very specific for blood vessels, especially in tumour tissue, the combination with a specific lymphatic endothelium marker such as D2-40 is necessary to differentiate between LVI and BVI. In the present study, more than 50% of lymph vessels did show CD34 expression (data not shown). CD31 might be a more specific endothelial marker than CD34, nevertheless the latter was prefered in this study, as CD31 positivity of inflammatory cells might hamper interpretation.
Recently, peritumoural lymphovascular invasion has been included in the St Gallen guidelines for adjuvant therapy of operable breast carcinoma patients (Goldhirsch et al, 2005
). In these criteria, peritumoural lymphovascular invasion is assessed on HE-stained sections during routine pathological examination, making it impossible to differentiate between LVI and BVI. In our data set, peritumoural ‘vascular' invasion on HE and peritumoural LVI, not BVI were correlated with the presence of LN metastasis, still the only most important prognostic factor in BC. Multivariate analysis showed peritumoural LVI to be the most important determinant for the presence of LN metastases. Together with the increase in negative predictive value to 75% for peritumoural LVI, our results suggest that immunohistochemical detection of lymphovascular invasion and differentiation between LVI and BVI might be of value in clinical practice. Further studies are needed to address this issue.
If LVI and BVI are the morphological correlates of BC cells metastasising, respectively, via the lymphatic and the haematogenous route, our data support BC metastasis models stating that haematogenous and lymphatic metastasis are two complementary but specific metastasis pathways in BC. These newer models are in contrast with older models explaining BC metastasis as a stepwise cascade going from primary tumour via regional LNs to distant sites (Pantel and Brakenhoff, 2004
). Only peritumoural LVI, and not BVI, was associated with the presence of LN metastases. Other authors found a correlation between BVI and LN metastases (Lauria et al, 1995
; Kato et al, 2003
). This might be due to the fact that the methods they used were too insensitive to discriminate between LVI and BVI, leading to misinterpretation of BVI as LVI. Another factor contributing to the correlation between BVI and LN status in some studies might be the association between LVI and BVI. In our study, the presence of LVI was associated with the presence of BVI intratumourally, not peritumourally. As BVI is correlated with angiogenesis (Kato et al, 1999
) and LVI is correlated with lymphangiogenesis (Schoppmann et al, 2004
), the association between LVI and BVI might be due to the cross-interaction between lymphangiogenesis and hemangiogenesis. The hemangiogenic factors vascular endothelial growth factor (VEGF)-A, basic fibroblast growth factor, angiopoietin-1 and -2 and platelet-derived growth factor have been shown to induce lymphangiogenesis and the lymphangiogenic factor VEGF-C can also induce hemangiogenesis (reviewed in Cao, 2005
; Tammela et al, 2005
). In a study of 29 invasive breast carcinomas, Choi et al (2005)
reported a correlation between blood and lymph vessel microvessel density. The presence of a fibrotic focus is a surrogate marker for hypoxia-driven angiogenesis (Colpaert et al, 2003a
) and for lymphangiogenesis in BC (Van der Auwera et al, 2005
). In the present study, the presence of a fibrotic focus was indeed correlated with the presence of both LVI and BVI. The hypothesis that blood and lymph vessels are not just different routes that cancer cells can use to metastasise, but are characterised by a different biology is furthermore sustained by the fact that some patients exclusively show BVI or LVI and by the differences in size and number between LVI and BVI. In BVI, less vessels are involved and the size of the tumour emboli is smaller than in LVI. Very extensive vascular invasion is not found in BVI. To what extent these differences influence the metastatic capacity of both pathways remains to be elucidated.
In conclusion, we demonstrated that the described immunohistochemical technique made it possible to discriminate between BVI and LVI in BC and enabled a more sensitive detection of LVI and BVI and a better assessment of the extent of both than on conventional HE stains. Furthermore, our data demonstrate that most (lympho)vascular invasion in BC is LVI and that lymph vessel tumour emboli are larger than blood vessel tumour emboli. This suggests that LVI and BVI are not just different routes of BC metastasis, but that both pathways are characterised by a different biology.