Article Text

Tumour cell invasion into blood vessels is significantly related to breast cancer subtypes and decreased survival
  1. Tor A Klingen1,2,
  2. Ying Chen1,2,3,
  3. Ingunn M Stefansson1,4,
  4. Gøril Knutsvik1,4,
  5. Karin Collett1,4,
  6. Anne L Abrahamsen5,
  7. Hildegunn Aase6,
  8. Hans Aas7,
  9. Turid Aas8,
  10. Elisabeth Wik1,4,
  11. Lars A Akslen1,4
  1. 1Department of Clinical Medicine, Section for Pathology and Centre for Cancer Biomarkers CCBIO, University of Bergen, Bergen, Norway
  2. 2Department of Pathology, Vestfold Hospital Trust, Tønsberg, Norway
  3. 3Department of Pathology, Akershus University Hospital, Lørenskog, Norway
  4. 4Department of Pathology, Haukeland University Hospital, Bergen, Norway
  5. 5Department of Radiology, Vestfold Hospital Trust, Tønsberg, Norway
  6. 6Department of Radiology, Haukeland University Hospital, Bergen, Norway
  7. 7Department of Surgery, Vestfold Hospital Trust, Tønsberg, Norway
  8. 8Department of Surgery, Haukeland University Hospital, Bergen, Norway
  1. Correspondence to Professor Lars A Akslen, Department of Clinical Medicine, Section for Pathology and Centre for Cancer Biomarkers CCBIO, University of Bergen, Haukeland University Hospital, Bergen N-5021, Norway; lars.akslen{at}


Aims Vascular invasion in breast cancer is associated with increased risk of recurrence, metastases and death from disease. However, there are few studies discriminating between blood vessel invasion (BVI) and lymphatic vessel involvement (LVI).

Methods A population-based series of 282 breast cancers was examined (200 screen-detected and 82 interval patients) with respect to BVI and LVI in addition to basic features and molecular subtypes, using CD31 and D2-40 antibodies. This series is part of the prospective Norwegian Breast Cancer Screening Program.

Results The frequency of LVI and BVI was 25% and 15%, respectively. BVI was associated with HER2-positive and basal-like tumours, and several features of aggressive breast cancer, whereas LVI showed weaker associations. BVI was the strongest factor to predict interval cancer presentation. BVI showed significant associations with recurrence-free survival and disease-specific survival in univariate and multivariate analyses, whereas LVI was not significant.

Conclusions Our findings indicate that BVI by tumour cells is strongly associated with aggressive tumour features including a basal-like phenotype, and BVI was an independent prognostic factor in contrast to what was found for LVI.


Statistics from


Breast cancers are diverse with respect to morphological features and biological characteristics. One important hallmark is the ability of tumour cells to migrate into vascular channels, as an early step of metastatic spread.1 In breast tumours, vascular invasion is considered to be lymphatic vessel involvement (LVI) more often than blood vessel invasion (BVI),2 but there are few studies in this field. LVI is particularly important as a prognostic factor in early stage breast cancer.3 ,4 Vascular invasion, as observed on standard sections, is associated with increased risk of tumour recurrence, metastasis and death from disease.2 ,5 Interestingly, recent findings by Gujam et al6 highlighted that immunohistochemistry is necessary to discriminate between BVI and LVI, and this distinction improves the prognostic value of vascular invasion compared with standard sections.

Here, we studied the frequency and prognostic significance of tumour cell invasion into lymphatic or blood vasculature, to see whether there are differences across breast cancer molecular subtypes. We also wanted to compare the pattern of vascular invasion between screening-detected and interval-presenting breast cancers.

Materials and methods

Patient series

The Norwegian Breast Cancer Screening Program (NBCSP) was initiated in 1996 and covers women aged 50–69 years.7 We studied a consecutive population-based series from Vestfold County in Eastern Norway. Vestfold comprises approximately 5% of the Norwegian population with around 230 000 inhabitants. The NBCSP was implemented in this county in 2004, and a total of 37 977 women participated during the study period 2004–2009. During this period, 204 invasive screen-detected cancers and 85 interval breast cancers were diagnosed after the prevalent and subsequent screening rounds. Of these, three cases were excluded (one screen-detected cancer had no residual tumour tissue; one screen-detected case represented a malignant phyllodes tumour; one interval cancer had multiple metastases, and no biopsy or surgery was performed). Four patients had simultaneous tumours in both breasts (two screening patients and two interval patients). The Nottingham Prognostic Index, NPI (three cases)8 and proliferation by Ki67 (one case with equal NPI) were used to select the tumour with worst prognosis to be included in our study. In 10 patients, only core biopsies were available, and we decided to include these cases. In total, 282 invasive cancers were available for this population-based study, 199 screen-detected and 83 interval cancers (median tumour diameter 14.0 mm). The median follow-up period was 72 months (range 2–117 months).

Regarding primary treatment, 204 resections (72%) and 78 mastectomies (26%) were performed; 218 patients received radiation (77%), 156 received endocrine treatment (55%) and 77 received chemotherapy (27%). In 12 patients, no information on radiation, endocrine treatment or chemotherapy was available. Distant metastases were observed at follow-up in 34 cases (12%), and 24 patients (8%) died of breast cancer.

Basic variables

Traditional prognostic factors such as largest tumour diameter, tumour type, histological grade and lymph node status were recorded. Radiological size estimate was included for cases when tumour diameter was not available from pathology reports (as in patients with locally advanced or multifocal disease). Cases were divided into two groups according to tumour diameter ≥2 cm or <2 cm. Tumour type was included as ductal versus others, and histological grade was recorded according to the modified Bloom-Richardson criteria.9

Tissue microarrays

Tissue microarray (TMA) blocks were made as previously described.10 ,11 In this study, TMA slides were used for assessment of CK5/6, P-cadherin and Ki67 expression. Our research group has previously shown that Ki67 is prognostic over a wide range of cut-off points and for different sample types including TMA sections.12


Tissue specimens were fixed in 10% buffered formalin and embedded in paraffin. All original slides were reviewed (by TAK), and one representative block, including tumours with adjacent peritumorous tissue, was selected for further staining. In 10 cases, preoperative tumour biopsies were used because surgical specimens had very limited tumour tissue.

Vascular invasion is more often found in peritumorous than intratumorous areas.2 ,3 In standard sections, it can be difficult to distinguish lymphatic channels from blood vessels. It might also be difficult to distinguish vascular invasion from retraction artefacts around tumour cells, carcinoma in situ or small groups of invasive tumour cells. Immunohistochemical markers increase the sensitivity of detecting vascular involvement compared with routine sections.2 ,13–15

Immunohistochemical staining was performed on standard 4–5 μm sections (not TMA sections) using D2-40 antibody (M3619; Dako, Glostrup, Denmark) for lymphatic vessels and CD31 antibody (clone JC70A, M0823; Dako) for blood vessel staining.14 Sections were dewaxed with xylene and rehydrated in ethanol before microwave antigen retrieval. For D2-40, sections were boiled in Tris-EDTA buffer pH 9, 10 min at 750 W followed by 15 min at 350 W and then incubated with D2-40 antibody diluted 1:100 for 30 min at room temperature. For CD31, sections were boiled in target retrieval solution pH 6.0 (Dako), 10 min at 750 W followed by 20 min at 350 W and then incubated with the monoclonal antibody CD31 diluted 1:25 for 60 min at room temperature. Staining was carried out using the mouse ENVision Kit (Dako) with diaminobenzidine peroxidase (Dako) as substrate. Finally, the sections were counterstained with Dako REAL haematoxylin for 1 min (Dako).

All stained slides were examined by one experienced breast pathologist (TAK), and the observer was blinded for clinicopathological information. We recorded LVI to be present if tumour tissue was located within more than one D2-40-positive structure14 with weak or negative CD31 staining (figure 1A, B). BVI was reported when tumour cells were detected in one or more CD31-positive and D2-40-negative vessels (figure 2A, B).

Figure 1

Histological images of BVI and LVI by immunohistochemistry. Histological images of BVI from the same area (1A and 1B) with tumour tissue in one centrally located D2-40-negative (1A) and CD31-positive blood vessel (1B). The slit-like structure in the periphery is a lymphatic vessel positive for D2-40 and only weakly positive for CD31 (×400). Histological images of LVI from another area (2A and 2B) with tumour tissue in two D2-40-positive (2A) and CD31 weakly positive lymphatic vessels (2B, ×400).

Figure 2

Distribution of vascular invasion in luminal versus non-luminal tumours. Frequency of blood vessel invasion (BVI) (A) and lymphatic vessel invasion (LVI) (B) according to breast cancer subtype (per cent of cases).

Evaluation of CK5/6, P-cadherin and Ki67, as well as oestrogen receptor, progesterone receptor (PR) and HER2 status has been described previously.10 CK5/6 and/or P-cadherin positivity was used to define basal-like differentiation,16 ,17 and for these markers, cases were evaluated for intensity of staining and expression area. A staining index (values 0–9), obtained as a product of staining intensity (0–3) and proportion of immunopositive tumour cells (≤10%=1, 10–50%=2, >50%=3) were calculated. Tumours with a staining index ≥1 for CK5/6 or ≥4 for P-cadherin (or both) were considered as basal-like tumours.

Surrogate markers for molecular subtypes of breast cancer were defined according to the St Gallen Consensus 2013.18 The cut-off point for oestrogen receptors and PRs was 1% in the present study.

Statistical analyses

All statistical analyses were performed by IBM SPSS Statistics, V.21.0 (IBM, Armonk, New York, USA). A probability of <0.05 was considered statistically significant. Associations between different categorical variables were assessed by Pearson's χ2 test. A total of 282 patients were accessible for survival analysis. Univariate analyses of time to recurrence for patients without metastases at the time of diagnosis (recurrence-free survival, RFS) and time to death from breast cancer (disease-specific survival, DSS) were performed using the Kaplan-Meier method (log-rank test), with time of diagnosis as entry date. Patients who died from other causes were censored at time of death. The Cox proportional hazards method was used for multivariate survival analyses. The variables were visually examined by a log-minus-log plot to check for proportionality before incorporation into multivariate models. A multivariate analysis was conducted for BVI and LVI, together with standard prognostic variables. Logistic regression (enter method) was applied to assess the ability of various markers to predict the basal-like phenotype and mode of tumour presentation.


BVI and LVI are associated with clinicopathological characteristics

In our series, 45 tumours (16%) showed LVI only, 18 (6%) had BVI only and 25 (9%) showed both BVI and LVI according to our criteria, giving overall frequencies of LVI and BVI of 25% and 15%, respectively. Table 1 shows that both BVI and LVI were associated with aggressive tumour features, such as larger tumour size (only BVI), higher histological grade, lymph node-positive tumours and higher Ki67. In addition, BVI and LVI were significantly associated (OR=1.9, p<0.001).

Table 1

Associations between different categories of vascular invasion by immunohistochemistry and various clinicopathological variables (n=282, except for nodal status, n=281)

Regarding treatment, both BVI (p=0.001) and LVI (p=0.02) were associated with patient subgroups that were treated by chemotherapy. In contrast, no associations were found for radiation therapy or endocrine therapy.

Vascular invasion is associated with molecular subtype

BVI was significantly associated with molecular subtype according to the St Gallen Consensus 2013, being lower among luminal (HER2-negative) cases compared with the other categories (table 2, figure 2A). For LVI, a similar pattern was found (table 2, figure 2B). HER2-positive and Ki67-high tumours showed stronger associations with BVI than LVI (tables 1 and 2).

Table 2

Associations between different categories of vascular invasion and molecular subgroups (n=282)

Basal-like tumours accounted for 19% in our series. The basal-like phenotype showed a strong association with BVI (OR=4.2; p<0.001); the association with LVI was also significant (OR=3.0; p=0.001) (table 1).

When including BVI, LVI and triple-negative (oestrogen receptor/PR/HER2-negative) status in a multivariate logistic regression analysis, both BVI and triple-negative status independently predicted the basal-like phenotype (OR 3.6 95% CI 1.6 to 8.4, p=0.002, for BVI and OR 29.4 95% CI 8.9 to 96.6, p<0.0005, for triple-negative phenotype), whereas LVI did not independently predict this phenotype.

Vascular invasion is associated with mode of presentation

Table 1 shows associations between BVI or LVI and mode of presentation (screening vs interval detection). BVI was significantly more frequent in interval cases than what was seen among screen-detected cases (OR=6.4, p<0.001). A statistically significant association between LVI and interval cancers was also seen, although weaker (OR=1.9, p=0.020).

Including BVI, LVI, Ki67, CK5/6 and P-cadherin in a multivariate logistic regression model, BVI was a strong predictor of interval-detected cancer (OR 5.5 CI 2.4 to 12.7, p<0.0005) and high P-cadherin was also significant (OR 10.1 CI 3.5 to 29.0, p<0.0005), while none of the other variables reached statistical significance.

Survival analysis

By univariate analysis, BVI was significantly associated with reduced RFS and DSS (p=0.001 and p=0.002, respectively, figure 3). LVI was not significantly associated with survival (figure 3).

Figure 3

Estimated recurrence-free survival and disease-specific survival according to presence of blood vessel invasion (BVI) and lymphatic vessel invasion (LVI). Survival curves are presented for BVI (A and C) and LVI (B and D). Survival curves are estimated by the Kaplan-Meier method using the log-rank test for differences between subgroups. For each category, number of events/total number of cases are given.

Basic histopathological markers such as tumour diameter, histological grade and lymph node status together with BVI and LVI were included in a multivariate analysis of RFS (table 3). BVI was significantly associated with shorter RFS (HR=2.3) as was also lymph node metastases (HR=4.2), whereas LVI was not statistically significant.

Table 3

Univariate and multivariate recurrence-free survival analyses (Cox proportional hazards method) of pathological variables and vascular invasion (BVI and LVI) by immunohistochemistry (n=282)

Among basal-like tumours (n=53), BVI-negative cases had significantly better RFS and DSS when compared with BVI-positive cases (p=0.012 and p=0.003, respectively), whereas there was no difference for LVI. Also, among the luminal/Her2-negative cases, BVI-negative tumours were related to better RFS (p=0.028) when compared with BVI-positive cases, but this difference was not seen for LVI. Among HER2-positive cases, no significant difference was found for BVI or LVI.


A recent review concluded that vascular invasion based on standard sections is associated with RFS and breast cancer-specific survival.15 However, a potentially different impact of BVI as compared with LVI has not been well established, especially not in relation to the recently defined molecular subtypes. This might be due to the lack of firm criteria to separate BVI and LVI. Here, we examined the significance of these features by endothelial markers (CD31, D2-40) in a population-based series from the prospective NBCSP.

BVI, present in 15% of the cases, showed strong associations with non-luminal tumours such as the basal-like, triple-negative and HER2-positive subgroups. In survival analysis, BVI was significantly associated with RFS and breast cancer-specific survival, whereas LVI was not. When adjusting for basic factors, BVI was an independent prognostic marker, indicating that this feature might be recorded in breast cancer diagnostics, although more studies need to confirm our findings. Also, development of even more specific markers for blood vessels would be desirable in a routine setting to identify patients at a higher risk for early systemic spread. To speculate, it would be possible to imagine certain approaches for early systemic therapy among cases with BVI.

We previously reported that basal-like breast cancers appear to have increased angiogenesis with more microvessel proliferation and higher frequency of the glomeruloid microvascular pattern when compared with other breast cancer subtypes.16 ,19 These findings suggest a possible relationship between increased angiogenesis and BVI among basal-like breast cancers.

It has been suggested that the basal-like phenotype may be related to non-lymphatic spread,20 and recent findings indicate a reduced risk of axillary lymphatic spread in triple-negative breast cancer.21 Here, 11% of the cases with negative lymph nodes showed presence of BVI, indicating the possibility of direct haematogenic spread in some patients. It is well established that although the presence of metastases in axillary lymph nodes predicts development of distant metastases, 20–30% of patients with node-negative breast cancer develop metastatic spread at distant sites.22 Notably, studies of disseminated tumour cells from the bone marrow, as well as expression profiles of primary tumour cells, suggest that haematogenous spread is often an early event in tumour progression.23 Early systemic dissemination of breast cancer cells is associated with a specific expression signature and the molecular pathways associated with primary haematogenous spread and lymphatic dissemination appears to be different.24

As expected, LVI was more frequently observed (25%) when compared with BVI (15%), although the difference was smaller than that reported by others.2 ,15 Previous studies have shown marked variation in the frequency of LVI, with rates of 9–50% using H&E sections, and 32–41% by endothelial markers. For BVI, corresponding rates of 4–46% and 4–29%, respectively, have been reported.2 ,15 ,25 ,26 Here, endothelial markers (CD31, D2-40) were used. Most cases of LVI visualised by D2-40 were apparently easy to interpret, although some D2-40 staining in myoepithelial cells might represent a source of misclassification. Since LVI might be confused with small lesions of carcinoma in situ in rare cases, we decided to report the presence of LVI when tumour tissue was located within more than one D2-40-positive structure.

Contrary to our findings, Gujam et al6 demonstrated independent association between LVI and survival, also in the multivariate setting In that study, only invasive ductal carcinomas were examined, and patients below 50 years were included. Also, the cohort had larger and more high-grade tumours, with frequent triple-negative and Her2-positive phenotypes, and with different distribution of the therapies given, compared with our study. These differences might have contributed to the different survival pattern for patients with LVI-positive tumours.

As a novel finding, we found a significantly stronger association between BVI and tumour detection during mammography screening intervals than was the case for LVI. Our findings might in part explain why interval cancers behave more aggressively when compared with screen-detected cases. It has been shown that differences in standard prognostic factors such as tumour diameter, histological grade and lymph node status, as well as molecular subtypes, do not fully predict the difference between interval and screening cases.27–30 Possibly, the presence of BVI can explain some of this contrast.

In conclusion, our findings indicate that BVI by tumour cells is strongly associated with aggressive tumour subtypes (HER2-positive, basal-like, triple-negative) and BVI was independently associated with reduced survival, in contrast to LVI. Also, BVI was related to interval breast cancer presentation compared with screen-detected tumours. Based on these findings, it might be of importance to examine the presence of BVI in breast cancers.

Take home messages

  • Blood vessel invasion (BVI) by tumour cells is strongly related to aggressive tumour subtypes.

  • BVI is more frequent in interval breast cancer compared with screen-detected tumours.

  • In this series, BVI is an independent prognostic factor in contrast to lymphatic involvement.


The authors thank Gerd Lillian Hallseth, Randi Hope Lavik and Bendik Nordanger for excellent technical assistance. The authors also thank the bioengineers at Vestfold Hospital Trust for technical assistance.


View Abstract


  • Handling editor Cheok Soon Lee

  • Contributors LAA, TAK, YC and EW contributed in study design, data interpretation and writing the manuscript. IMS, GK, KC, ALA, HiA, HaA and TA participated in data collection and interpretation.

  • Funding TAK and YC were supported by the Vestfold Hospital Research Fund for this study. This work was partly supported by the Research Council of Norway through its Centres of Excellence funding scheme, project number 223250, and by grants from Helse Vest Research Fund and the Norwegian Cancer Society (LAA).

  • Competing interests None declared.

  • Ethics approval The study was approved by the Regional Ethics Committee of Eastern Norway (REK registration number S-08685d).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.