Langenbeck's Archives of Surgery

, Volume 393, Issue 6, pp 911–917

Pancreatic cancer: a generalized disease—prognostic impact of cancer cell dissemination

Authors

    • Department of General, Visceral- and Thoracic-SurgeryUniversity Medical Centre of Hamburg-Eppendorf
  • T. Strate
    • Department of General, Visceral- and Thoracic-SurgeryUniversity Medical Centre of Hamburg-Eppendorf
  • E. F. Yekebas
    • Department of General, Visceral- and Thoracic-SurgeryUniversity Medical Centre of Hamburg-Eppendorf
  • J. R. Izbicki
    • Department of General, Visceral- and Thoracic-SurgeryUniversity Medical Centre of Hamburg-Eppendorf
Original Article

DOI: 10.1007/s00423-007-0278-y

Cite this article as:
Bogoevski, D., Strate, T., Yekebas, E.F. et al. Langenbecks Arch Surg (2008) 393: 911. doi:10.1007/s00423-007-0278-y

Abstract

Background

Pancreatic ductal adenocarcinoma is the fifth leading cause of death among all malignancies, leading to approximately 40,000 deaths each year in Europe. The annual incidence rate for all types of pancreatic cancer is approximately nine new cases per 100,000 people, ranking it as the 11th among all cancers. Stage, grade and resection margin status are currently accepted as the most accurate pathologic variables predicting survival. All classification systems fail prognostically to distinguish between different stages. Even in patients with seemingly early tumours (T1, N0), the likelihood of relapse is high. This reflects the shortcomings of the pathologic staging to sufficiently discriminate patients with a high risk to develop tumour recurrence from those that carry a lower risk.

Results

On the other hand, none of the currently used systems includes or takes into consideration the role of disseminated tumour cells neither in the lymph nodes nor in the bone marrow. Occult residual tumour disease is suggested when either bone marrow or lymph nodes, from which tumour relapse may originate, are affected by micrometastatic lesions undetectable by conventional histopathology. For detection, antibodies against tumour-associated targets can be used to detect individual epithelial tumour cells both in lymph nodes and in bone marrow. The clinical significance of these immunohistochemical analyses is still controversial. Various monoclonal antibodies are still in use for micrometastatic detection, thus contributing to the incongruity of data and validity of results. These assays have been rarely used in patients with pancreatic carcinoma.

Conclusion

The presence or absence of lymph-node metastases can predict the likelihood of survival for most, if not all, patients with pancreatic ductal adenocancer and the likelihood that metastases will develop at distant sites.

Keywords

Pancreatic adenocarcinomaMicrometastasesImmunohistochemistry

Pancreatic ductal adenocarcinoma is the fifth leading cause of death among all malignancies [1], leading to approximately 40,000 deaths each year in Europe [2]. The annual incidence rate for all types of pancreatic cancer is approximately nine new cases per 100,000 people, ranking it as the 11th among all cancers [1]. The peak incidence occurs in the seventh and eighth decades of life, with the average age at diagnosis being 60 to 65 years [3].

Stage, grade and resection margin status are currently accepted as the most accurate pathologic variables predicting survival. A number of schemes for the staging of pancreatic carcinoma have been proposed in the past. The newest version of the American Joint Committee on Cancer cancer staging manual was published in 2002 [4] (Tables 1 and 2). Because only a minority of patients with pancreatic cancer undergoes surgical resection, a single TNM classification must apply to both clinical and pathologic staging. The definitions of TNM have changed from past versions, with specific changes made to the T classification and to the definition of stage III disease.
Table 1

The latest version of the American Joint Committee on Cancer staging manual for pancreatic carcinoma

AJCC

Code

Description

Primary tumour (T)

Tx

Primary tumour cannot be assessed

T0

No evidence of primary tumour

Tis

Carcinoma in situ

T1

Tumour limited to pancreas and 2 cm or less in greatest dimension

T2

Tumour limited to pancreas and more than 2 cm in greatest dimension

T3

Tumour extends beyond pancreas but does not involve celiac axis or superior mesenteric artery

T4

Tumour invades the celiac or the superior mesenteric artery (unresectable primary tumour)

Regional lymph nodes (N)

Nx

Regional lymph nodes cannot be assessed

N0

No regional lymph-node metastases

N1

Regional lymph-node metastases

Distant metastases (M)

Mx

Distant metastases cannot be assessed

M0

No distant metastases

M1

Distant metastases

From [4].

Table 2

Stage grouping

Stage grouping

T

N

M

O

Tis

N0

M0

IA

T1

N0

M0

IB

T2

N0

M0

IIA

T3

N0

M0

IIB

T1–3

N1

M0

III

T4

Any N

M0

IV

Any T

Any N

M1

From [4].

It is also important to note that the extent of resection (R0—complete resection; R1—grossly negative but positive microscopic margins of resection; R2—grossly and microscopically positive margins of resection) is not a part of the TNM staging system, but is of great, if not crucial, prognostic significance.

Another commonly used staging system involves the Union Internationale Contre Le Cancer (UICC) system, which is also based on TMN factors (Table 3) [5]. A more complex stage classification system has been proposed by the Japan Pancreas Society (Table 4), adding other factors to the classification such as serosal invasion (S factor), retroperitoneal invasion (RP factor) and invasion of the portal venous systems (PV factor) [6]. This system has been overly cumbersome and difficult to apply, and it has gained only limited use.
Table 3

UICC staging of pancreatic cancer

Stage grouping

T

N

M

5-Year survival rate (%)

Stage I

T1 or T2

N0

M0

20–40

Stage II

T3

N0

M0

10–25

Stage III

Any T

N1

M0

10–15

Stage IV

Any T

Any N

M1a

0–8

From [5].

T Tumour, N lymph nodes, M distant metastasis, T1 limited to pancreas, T2 extension directly to duodenum, bile duct or peripancreatic tissues, T3 extension directly to stomach, spleen, colon or adjacent large vessels, N0 no lymph-node metastases, N1 lymph-node metastases, M0 no distant metastasis, M1 distant metastasis

Table 4

Japan Pancreas Society stage classification

Stage grouping

T

S

RP

PV

N

M

5-Year survival rate (%)

Stage I

T1

S0

RP0

PV0

N0

M0

35–45

Stage II

T2

S1

Rp1

PV1

N1

M0

15–25

Stage III

T3

S2

RP2

PV2

N2

M0

5–15

Stage IV

T4

S3

RP3

PV3

N3

M1

0–10

From [6].

T Tumour, T1 0 to 2 cm, T2 2.1 to 4 cm, T3 4.1 to 6 cm, T4 6.1 cm, S serosal invasion, RP retroperitoneal invasion, PV portal venous invasion, 0 absence of invasion, 1 suspected invasion, 2 definite invasion, 3 severe invasion, N lymph nodes, N0 no metastasis, N1 primary lymph-node group metastasis, N2 secondary lymph-node group metastasis, N3 tertiary lymph-node group metastasis, M distant metastasis, M0 no distant metastasis, M1 distant metastasis

However, all these classification systems fail prognostically to distinguish between different stages. Even in patients with seemingly early tumours (T1, N0), the likelihood of relapse is high. This reflects the shortcomings of the pathologic staging to sufficiently discriminate patients with a high risk to develop tumour recurrence from those that carry a lower risk. Thus, effort continues to identify new prognosticators of tumour relapse that indicate the need for adjuvant therapy.

On the other hand, none of the currently used systems includes or takes into consideration the role of disseminated tumour cells neither in the lymph nodes nor in the bone marrow. Occult residual tumour disease is suggested when either bone marrow or lymph nodes, from which tumour relapse may originate, are affected by micrometastatic lesions undetectable by conventional histopathology [7]. For detection, antibodies against tumour-associated targets can be used to detect individual epithelial tumour cells both in lymph nodes [811] and in bone marrow [11]. The clinical significance of these immunohistochemical analyses is still controversial [1221]. Various monoclonal antibodies are still in use for micrometastatic detection, thus contributing to the incongruity of data and validity of results. These assays have been rarely used in patients with pancreatic carcinoma [2224].

The presence or absence of lymph-node metastases can predict the likelihood of survival for most, if not all, patients with cancer and the likelihood that metastases will develop at distant sites. For some tumour types, the correlation between lymph-node metastases and distant metastases is strong; e.g. head and neck cancer is a member of this group [25]. The presence of lymph-node metastases in the neck is the strongest prognostic factor, halving the survival rate of these patients. The number of lymph-node metastases and the extra nodal spread are also prognostic factors, particularly for the development of distant metastases. When more than three lymph-node metastases are identified in the neck by histological analysis, the risk for the development of distant metastases is almost 50%, whereas it is only 7% when the patient’s neck is free of lymph-node metastases [25].

The metastatic pattern of other solid tumours, such as breast carcinomas, is different from that of head and neck tumours. Although the presence of metastases in the axillary lymph nodes predicts the development of distant metastases, 20–30% of the patients with breast cancer that are free of axillary lymph-node metastases – even when analysed by very sensitive assays – also develop disease at distant sites [26]. This observation indicates that in patients with breast cancer, the presence of haematogenous tumour cells is not as strongly associated with lymph-node metastasis as it is for patients with head and neck cancer. So, breast tumour cells can bypass the lymph nodes and disseminate directly through the blood to distant organs. This view is supported by recent gene-expression profiling studies showing that the molecular pathways involved in lymphatic and haematogenous dissemination of breast tumours differ [27].

The data concerning metastatic process in patients with pancreatic adenocarcinoma in the literature are confusing. On one hand, there are studies claiming that metastases in regional lymph nodes are of predictive value for developing distant metastasis. On the other hand, there are several studies claiming exactly the opposite that even those patients that were free of local lymph-node involvement at the time of surgery had developed a distant metastasis later.

Metastasis usually starts out as single cells that can escape a morphological identification during the routine histopathological staging of patients with cancer. Many sensitive immunological and molecular procedures have therefore been developed to detect single tumour cells in lymph nodes that drain the primary tumour, in peripheral blood and in distant organs, such as bone marrow. These cells have been controversially referred to as ‘micrometastases’, but to diminish the possibility for confusion, here we will use the more neutral term ‘disseminated tumour cells’. Although many different assays have been developed over the past 10 years to detect disseminated tumour cells, the two main approaches that are used involve either immunocytochemical staining or polymerase chain reaction (PCR) analysis. Immunocytochemical detection assays involve monoclonal antibodies that bind to tumour-associated or histogenic markers—proteins that are expressed by disseminated tumour cells but are absent on the surrounding normal cells. If these assays are sensitive and specific enough, they can also detect a single metastatic cell in the background of millions of normal cells. For epithelial tumours, cytokeratins have become the best marker for disseminated tumour cells. The specificity of this diagnostic approach has been demonstrated by Braun et al. who analysed bone-marrow cells from almost 200 individuals without cancer and more than 550 patients with breast cancer. They detected single cytokeratin-positive cells in only 1% of control individuals, whereas 30–40% of bone-marrow cells from patients with breast cancer were cytokeratin positive [26]. It remains unclear whether cytokeratin-positive cells found in the bone marrow of control patients are normal epithelial cells or tumour cells derived from an unknown primary carcinoma.

In our recently published study [28], we have applied the sensitive immunohistochemical method for detection of cytokeratin-positive cells in pancreatic ductal adenocarcinoma. The key finding of this study was that isolated tumour cells, detectable in lymph nodes by immunohistochemical analysis, are strong independent prognostic factors in pancreatic ductal adenocarcinoma, irrespective of the histopathological N-status. We have analysed a homogeneous entity of patients with pancreatic ductal adenocarcinoma, and none of them received any adjuvant chemoradiation or chemotherapy. Two subsets of patients could be identified in the group of patients who were classified as pN0 through conventional histopathology: the one with a poor 5-year survival probability of 0%, which was close to that of patients with overt nodal involvement (pN1). In the other subset, which had a much better prognosis with a 5-year survival probability of over 55%, nodal microinvolvement was absent. Thus, immunohistochemistry confirmed the cardinal importance of occult tumour cells for the separation of the respective survival curves in pN0-patients (Table 5).
Table 5

Review of the literature for the frequency and impact on overall survival of ‘disseminated tumour cells’ in lymph nodes in patients with pancreatic ductal adenocarcinoma

Author

Year

No. of patients

Marker

% of positive patients

Significance

Hosch et al. [23]

1997

12

IHC Ber-EP4

75

Positive

Ando et al. [32]

1997

13

PCR K-ras

62

Negative

Tamagawa et al. [33]

1997

6

PCR K

83

Positive

Demeure et al. [34]

1998

25

PCR K

68

Negative

Yamaguchi et al. [35]

2000

31

PCR K

71

Negative

Yamada et al. [36]

2000

12

PCR K

58

Positive

Ridwelski et al. [37]

2001

25

ICH CK

60

Negative

Niedergethmann et al. [38]

2002

69

PCR K-ras

17

Positive

Kanemitsu et al. [22]

2003

7

ICH CK

71

Positive

Milsmann et al. [31]

2005

41

IHC Ber-EP4

39

Positive

Bogoevski et al. [28]

2006

106

IHC Ber-EP4

69

Positive

Kurahara et al. [30]

2007

58

AE1/AE3

76

Positive

In another published study from our group [29], even in patients staged as pN1, the detection of occult tumour cells in ‘tumour-free’ lymph nodes had also prognostic influence. This finding was consistent with previous observations of our group showing that in both esophageal [20] and non-small cell lung carcinoma [9], survival is significantly worsened when a histopathological pN1-status is accompanied by nodal microinvolvement. Basically, a pN1-status in solid tumours is considered as a locoregional disease, which can be potentially cured by surgery, although it generally carries a higher risk of systemic dissemination than does a pN0-status. Therefore, the finding that pN1 patients with additional nodal microinvolvement in tumour-free lymph nodes apart from overt lymph-node metastases had significantly shorter recurrence-free and overall survival as compared with pN1-patients without occult tumour cells suggests that immunohistochemistry may be helpful to identify different risk profiles in these patients.

These results are in correlation with the results from other studies. Kanemitsu et al. [22] showed that the overall survival was strongly influenced by the disseminated tumour cells, as was also shown by the most recently published study from Kurahara et al. [30]. The patients without any nodal involvement (excluded by HE and immunohistochemistry) had a 5-year overall survival of 50%, compared to 26% in patients with only disseminated tumour cells detected through immunohistochemistry and 0% in patients with HE nodal involvement. Milsmann et al. [31] had also similar results in their recently published study, where the patients free of nodal involvement had 5-year overall survival probabilities of 53%. In contrast, patients with disseminated tumour cells detected by immunohistochemistry had similar survival as patients with nodal metastases detected through conventional histology (10% vs 9%). The overview of literature is given on Table 5 [22, 28, 3038].

The PCR-based approaches are also used to identify disseminated tumour cells through detection of genetic and epigenetic alterations that are specifically associated with cancer cells [39]. These sequences include tumour-associated mutations in oncogenes such as RAS or in tumour suppressors such as TP53. The PCR approach, however, is labourious and is complicated by the substantial degree of genetic variability between solid tumours. The PCR can also be used to detect tumour-specific mRNA sequences, but this is a challenge as many tumour-associated transcripts can sometimes be expressed by normal cells. The removal of contaminating normal cells, such as granulocytes that express cytokeratin-20, from the tumour cell sample before the assay is performed, or using quantitative RT-PCR assays [40] might be a solution to this problem.

Bone marrow, which can be easily collected from the iliac crest, is the most important site for detecting disseminated epithelial tumour cells. Disseminated tumour cells are present in bone-marrow samples of 20–40% of patients with carcinomas at various primary sites—even in the absence of lymph-node metastases (stage N0) or clinical signs of overt distant metastases (stage M0) [26, 4143]. Bone-marrow samples can also be monitored for the presence of disseminated tumour cells after primary surgical treatment to detect tumour recurrence [44, 45]. Interestingly, the presence of disseminated tumour cells in the bone marrow is useful not only in predicting the development of skeletal metastases but also in predicting the development of metastases in other distant organs, such as lung or liver. This is even true for tumours that rarely show skeletal metastases, such as colon cancer [41]. In fact, disseminated cancer cells have also been found in the bone marrow of patients with head and neck cancer that did not have lymph-node metastases, although the clinical significance of these cells is not clear [46]. So, bone marrow might be an important reservoir that allows for disseminated epithelial tumour cells to adapt and disseminate into other organs. An alternative explanation is that the presence of occult cancer cells in the bone marrow might reflect the general propensity of these cells to disseminate and survive in organs, rather than just in the bone marrow. Until methods are developed to detect the presence of disseminated cancer cells in organs such as the lung or liver, it will not be possible to distinguish between these two possibilities.

In our most recently published study for impact of disseminated tumour cells in pancreatic carcinoma [29], we have also examined the bone marrow of 59 patients for disseminated tumour cells. The lack of a significant difference in recurrence-free and overall survival (median 17 months vs 15 months, p = 0.18) between patients with nodal microinvolvement alone and those with additional involvement of the bone marrow suggests that the key event in pancreatic cancer progression is the spread of tumour cells to the regional lymph nodes. Nodal microinvolvement seems to indicate a systemic disease in pancreatic carcinoma much more accurately than do occult tumour cells in bone marrow. However, our results have to be interpreted cautiously, because bone-marrow findings may have been biased by the fact that only a subset of patients was analysed. On the other hand, the results from Vogel et al. [24] had shown borderline significance between the patients with and without bone-marrow affection (2-year overall survival 0% vs 30%). Therefore, the influence of bone-marrow microinvolvement on the outcome of patients with pancreatic carcinoma needs to be clarified in future studies.

Although chemoradiation and/or chemotherapy for adjuvant treatment of pancreatic carcinoma may have severe side effects [47], in common clinical practice, it is in most instances applied irrespective of tumour stage. This reflects the distrust in the value of conventional tumour-staging nomenclature in terms of reliably predicting the risk of tumour relapse even in patients with early pancreatic cancer (T1, N0). The data exposed here indicate that immunohistochemical assessment of lymph nodes can be used to refine the staging system for pancreatic ductal adenocarcinoma and might help us to identify patients who will not be cured by surgery alone and who need adjuvant therapy. In turn, patients who are true node-negative both in histopathology and in immunohistochemistry have an excellent 5-year survival probability of nearly 60%, even without chemotherapy. Future studies will show whether this prognosis can be further improved by adjuvant therapy.

Peripheral blood, however, is an ideal source for the monitoring of metastatic tumour cells, and many groups have demonstrated the presence of these circulating cells in patients with early-stage cancer without overt metastases. The lifespan of circulating cells in the peripheral blood is short [48]. Little is known about what is required for circulating tumour cells to survive the vigorous passage in blood and the subsequent invasion of organs in patients. Pierga et al. showed a correlation between the presence of cytokeratin-positive cells in peripheral blood and bone marrow [49], although only the presence of cytokeratin-positive cells in the bone marrow could be correlated with metastatic relapse. So, the circulating tumour cells that are able to find their way to the bone marrow and survive there seem to have an increased ability to develop into overt metastases.

Recent technical developments have made it possible to examine the genome of disseminated tumour cells. A combination of immunocytochemistry and fluorescence in situ hybridization has shown that the bone marrow contains disseminated epithelial cells of malignant origin [50]. By developing a new procedure for whole-genome amplification and subsequent comparative genomic hybridization of single immunostained cells, Klein and co-workers demonstrated that cytokeratin-positive cells in the bone marrow of patients with epithelial breast cancer without clinical signs of overt metastases (stage M0) are genetically heterogeneous [51, 52]. This heterogeneity was strikingly reduced with the emergence of clinically evident metastasis (stage M1).

The stage at which individual cells leave the primary tumour is unclear. In patients with early-stage invasive breast cancer, the cytokeratin-positive cells isolated from the bone marrow had few features in common with those found in their respective primary tumours [53]. A provocative interpretation of this surprising finding is that the disseminated tumour cells separated from their primary tumour at a very early stage. This hypothesis is also supported by the finding that only few disseminated tumour cells in these patients had TP53 mutations, which are associated with the later stages of tumourigenesis [52]. Disseminated tumour cells might therefore evolve independently into overt metastases, driven by the specific selective pressures of the bone-marrow environment [54]. Even more unclear is the potential of these ‘disseminated cells’ to develop into overt metastases. Genetic analyses that compared paired primary and metastatic breast tumour samples confirm the hypothesis that disseminated tumour cells evolve independently from the primary tumour. The patterns of genetic alterations that are observed in overt metastases are often discordant with those of the primary tumour, and differ almost completely in approximately one-third of the cases [55]. During genetic progression of the primary breast tumour, cancer cells might disseminate continuously, acquiring additional genetic alterations after migration into secondary organs such as the bone marrow.

Metastatic spread might follow two models, which are complementary but emphasize specific routes. In the first model, cancer cells disseminate from the primary tumour to the lymph nodes or blood during the early stages of tumour growth. The disseminated tumour cells proliferate and form solid metastases in lymph nodes, whereas tumour cells that spread to distant sites, through the blood, die or remain dormant. At later stages, tumour cells disseminate from the established lymph-node metastases to distant sites, where they are able to form solid metastases. It is possible that this ability is gained during the selection of these cells in the lymph-node environment. As a result, metastasis to other organs is dependent on the presence of lymph-node metastases.

In the second model, cells frequently disseminate, through the blood, from the primary tumour to distant sites, where they progress to overt metastases without previous passage through the lymph nodes. In patients with breast cancer, this haematogenous dissemination seems to be a very early event in tumour progression, so the disseminated tumour cells found in the bone marrow might be considered ‘immature’ tumour cells, as they have a limited lifespan and are usually not proliferating. Unlike in patients with pancreatic ductal carcinoma, it seems that the cascade starts first in the regional lymph nodes and then spreads haematogenous from these sites. When haematogenous dissemination occurs, the cells are far more potent and they grow fast into an overt metastasis.

Early detection of disseminated tumour cells might help to predict which patients are in need of additional systemic therapies after successful surgical resection of the primary tumour. Although all of these therapies are aimed at preventing metastatic relapse, the selection of patients is at present based on their statistical risk of developing tumour recurrence, without knowing whether they actually harbour any disseminated tumour cells. This uncertainty leads to over treatment of patients with cancer with toxic agents that exert severe side effects. For example, only 30–35% of patients with pancreatic carcinoma without any detected disseminated cells in lymph nodes undergo metastatic relapse within 5 years after surgery, but more than 90% of these patients are not supposed to receive chemotherapy according to the new consensus recommendations. So, the need for better selection criteria and the development of more specific and less toxic forms of therapy is obvious. Based on these considerations, it will not be sufficient to simply characterize the primary tumour as a therapeutic target, but it is essential to include disseminated tumour cells into this analysis too. This information is important for the design of clinical trials using biological therapies directed against specific targets. In these trials, sequential screening and monitoring of bone marrow and blood cells could provide early information about the therapeutic efficacy of the tested drug against the disseminated tumour cells. Clearance of disseminated tumour cells from bone marrow could serve as an intermediate end point in clinical trials with anticancer agents. Further molecular and functional description of these cells will be essential to develop and select more efficient forms of systemic therapy. For example, the low proliferative activity of these individual disseminated tumour cells in bone marrow of patients with non-metastatic cancer at the time of primary surgery [56] might explain the relative resistance of these cells to conventional chemotherapy [57]. New therapies that work equally well on proliferating and quiescent tumour cells (for example, immunotherapy) are on the horizon. The identification of the specific genetic and phenotypic changes that occur in disseminated tumour cells could also lead to new targets for antimetastatic therapies [58]. These therapies have the chance to cure patients only if applied in the early stage of micrometastases—before the occurrence of overt metastasis and secondary dissemination.

Copyright information

© Springer-Verlag 2007