Actuarial disease-free survival curves. A, Survival by Union Internationale Contre le Cancer (UICC)
stage. B, Survival in patients with positive (pN0[i+]) disease vs those with negative (pN0[i−] disease) lymph node micrometastasis.
C, Survival in patients with positive (R1[cy+]) disease vs those with negative (R0 disease) findings of tumor cells in the peritoneal cavity.
D, Survival in patients with positive (pM0[i+] disease) vs those with negative (pM0[i−] disease) findings of epithelial cells in bone marrow.
Typical examples of positive findings in lymph nodes, the peritoneal cavity, and bone marrow (immunoalkaline phosphatase, original magnification × 400). A, Lymph node,
antibody cytokeratin 18. B, Peritoneal cavity, antibody Ber-EP4, a ubiquitous epithelial-specific, but not tumor-specific, antigen. C,
Bone marrow, antibody A45-B/B3, an anticytokeratin antibody. D, Bone marrow, antibody Ber-EP4.
Steinert R, Hantschick M, Vieth M, Gastinger I, Kühnel F, Lippert H, Reymond MA. Influence of Subclinical Tumor Spreading on Survival After Curative Surgery for Colorectal Cancer. Arch Surg. 2008;143(2):122-128. doi:10.1001/archsurg.2007.49
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
To determine epithelial cell dissemination in patients with localized colorectal cancer.
Prospective observational study.
Two hundred twenty-two patients operated on for colorectal cancer.
Main Outcome Measures
Epithelial cell dissemination was determined using immunohistochemistry or cytology in histologically negative lymph nodes, the peritoneal cavity, and bone marrow. Prognostic significance was determined in relation to 140 clinicopathological variables. Median follow-up was 61 months.
Of 140 patients who underwent curative surgery; 25 (17.9%) died of cancer-related causes; 10 (7.1%), of other causes; and 11 (7.8%)
developed local recurrence. Tumor cells were present in the peritoneal cavity of 22% of patients, but this finding had only borderline influence on disease-free survival (P = .07).
Lymph node micrometastases correlated with T category but not with survival. The presence of epithelial cells in the bone marrow was detected in 64% of patients but was not associated with tumor stage or survival. Multivariate analysis failed to identify occult tumor cell dissemination into any body compartment as an independent prognostic factor of disease-free survival.
Tumor cells disseminate into various body compartments in early stages of disease. In about two-thirds of patients, tumor cells are left in the body after so-called curative surgery. However, the presence of minimal residual disease has no independent prognostic significance in relation to established risk factors for tumor progression. Thus,
other factors, such as the presence of a cellular metastatic phenotype and/or ineffective immunological response, must play an important role.
In the United States, 147 000 new patients were expected to develop colorectal cancer (CRC) in 2004,1 and surgery will be their only hope for healing.
However, about 25% of patients will develop metachronous distant metastases after curative surgery,2 suggesting the presence at the time of tumor resection of minimal residual disease that cannot be removed surgically. This is the rationale for adjuvant therapy, which is recommended in patients with tumoral lymph node involvement.3
New methods and standards are being developed to identify and treat patients at risk for metachronous metastasis. For example, the presence of lymph node micrometastasis can be documented in some patients with histologically node-negative CRC using immunohistochemistry4 or molecular techniques. Tumor cells can spread into the peritoneal cavity5 and tumor cell dissemination has also been verified in the bone marrow6 of patients with CRC.
So far, differences of terms and techniques, as well as the lack of large prospective clinical studies, have prevented wide acceptance of the presence of disseminated tumor cells as a prognostic factor for recurrence in CRC.7 The aim of this prospective observational study, started 9 years ago, was to assess simultaneously the prognostic significance of tumor cells spreading into different body compartments (lymph nodes, peritoneal cavity,
and bone marrow) in CRC.
A total of 222 patients scheduled for surgery at the Carl-Thiem Academic Hospital, Cottbus, Germany, because of suspicion of CRC were consecutively included in the study between September 1997 and June 2000. Informed consent was obtained from all patients. Two patients refused bone marrow biopsy.
Diagnostic screening for distant organ metastases included carcinoembryonic antigen (CEA) serum level, chest radiography, liver ultrasonography,
and coloscopy. All patients underwent laparotomy with subsequent tumor resection according to the guidelines of the German Surgical Society (Table 1).8 Diagnosis of CRC was verified by histological examination, and patients were staged according to the TNM classification.9
After surgery, patients underwent clinical examination, liver ultrasonography, and CEA testing every 6 months. All patients with Union Internationale Contre le Cancer (UICC) stage III disease received systemic adjuvant chemotherapy with 5-fluorouracil and folic acid according to the Mayo schema.10
Of 222 patients in whom a bone marrow biopsy specimen was obtained,
the following patients were excluded from further analysis: 36 patients with metastatic disease (UICC stage IV), 10 patients with benign disease (polyps), 9 patients with a history of a previous carcinoma, 8 patients with a synchronous noncolorectal cancer, 6 patients whose disease was downstaged after neoadjuvant therapy, 5 patients with R1 resection,
5 patients who developed a second malignant tumor during follow-up,
and 3 who died during the postoperative course.
Data on 140 clinical and pathological variables (list of variables on file), including all established risk factors for tumor progression,
were prospectively collected for each patient.
In the 90 lymph node–negative patients, a mean value of 12.2 lymph nodes were prepared for examination with cytokeratin 18
staining (in total 1108 examined lymph nodes). Paraffin blocs were cut in 4-μm sections and paraffin was removed with xylol and washed in ethanol. Slides were incubated with Pronase E (Serva, Heidelberg,
Germany; 0.05% in Tris-buffered saline) at 37°C for 10 minutes and washed. Then, slides were incubated as follows: 30 minutes with the primary antibody cytokeratin 18 (Sigma-Aldrich, Glostrup, Denmark)
in a 1:1000 dilution; 30 minutes with the rabbit–anti-mouse antibody (Dako, Hamburg, Germany) in a 1:30 dilution in dilution medium;
and 30 minutes with immunoalkaline phosphatase complex in a 1:50 dilution (Dako). Subsequently, slides were stained with fuchsin (Dako). Nuclei were counterstained with hematoxylin-eosin (hospital pharmacy). In the course of the study, the lymph node staining procedure was automated using an autostainer (Dako Autostainer; DakoCytomation, Hamburg, Germany).
Lymph nodes were considered micrometastatic when clustered, but isolated tumor cells also were observed.
Immediately before surgery, bone marrow was aspirated from the right and left upper iliac crest. After centrifugation through a Ficoll-Hypaque gradient (density, 1.077 g/mol; Pharmacia, Uppsala, Sweden) at 1000 × g for 20 minutes at 4°C, interface cells were sampled. Erythrocytes were removed using a lysis puffer (hospital pharmacy). Then, 2 × 105 mononuclear cells were centrifuged at 300 × g for 5 minutes on each slide. Per patient, 20 cytospins (10 right,
10 left) were screened: one-half were stained with the A45-B/B3 antibody and the other half, with Ber-EP4. A45-B/B3 is an anticytokeratin antibody.
Ber-EP4 is a ubiquitous epithelial-specific, but not tumor-specific,
antigen. Staining was performed using immunoalkaline phosphatase.
Nuclei were counterstained with Hämalaun (hospital pharmacy).
An automated cellular imaging system (ACIS; Chromavision, Paseo Cerveza,
California) was routinely used for primary analysis. Then, results were controlled with a Jenalumar microscope (Carl-Zeiss-Jena, Jena,
Germany) by 2 independent pathologists. A bone marrow sample was scored “positive,” if 1 or more positive cells were detected.
Positive and negative controls for bone marrow specimens were included.
Immediately after laparotomy, 200 mL of saline was injected into the peritoneal cavity, and the samples obtained were cytocentrifuged at 1000g for 5 minutes. The supernatant was discarded, the cell pellet was rediluted in cell wash with bovine serum albumin, erythrocytes were removed using a lysis puffer (hospital pharmacy), and samples were centrifugated at 300 × g during 5 minutes and resuspended in cell wash.
After overnight air-drying, slides were stained with Ber-EP4 antibody.
Staining was performed using immunoalkaline phosphatase. Per patient,
2 to 4 cytospins (depending on the cell quantity) were analyzed by 2 independent pathologists. The smears were classified according to their cytologic features and the presence of typical staining. The pathologists were not aware of the histopathological findings.
Follow-up was achieved in all patients (100%) using a standardized questionnaire. Follow-up information was obtained from Carl-Thiem Academic Hospital; Tumor Center Brandenburg, Cottbus; Institute for Quality Control in Operative Medicine, University of Magdeburg, Magdeburg,
Germany; general practitioners; and personal contact with patients.
Last follow-up was January 31, 2004. All patients with UICC stage III disease received adjuvant therapy according to the National Institutes of Health recommendations.3
The primary end point was cancer-related survival as measured from the date of surgery to the time of last follow-up or death. Kaplan-Meier life-table curves were constructed to estimate cancer-related survival.
The distributions of the patients with and without Ber-EP4–
and/or cytokeratin 18–positive tumor cells were compared using log-rank statistics. Locoregional relapse was considered a secondary end point and was measured accordingly. We used Cox proportional hazards analysis to estimate the simultaneous prognostic effect of variables.
The variables were entered stepwise forward to assess the independent prognostic value of bone marrow, lymph node, and peritoneal cavity micrometastasis compared with other prognostically relevant variables.
To compare categorical variables, we used the χ2 test.
Differences between groups were considered significant if the P values were less than .05 in a 2-sided test.
The characteristics of the 140 patients are detailed in Table 2. Median follow-up was 61.0 months (range, 43-95 months). Twenty-nine patients (20.7%) developed metachronous metastases. Eleven patients developed local recurrence, of whom 8
had simultaneous distant metastases. During the follow-up period,
25 patients (17.9%) died of cancer-related causes and 10 (7.1%), of other causes. Actuarial, stage-specific disease-free survival curves are shown in Figure 1. Five-year disease-free survival was 74.5% for patients with UICC stage I disease,
72.3% for patients with stage II disease, and 44.1% for patients with stage III disease. Median survival could not be calculated for all stages together since more than half of patients were alive by the end of January 2004.
Established factors showed prognostic significance in univariate analysis: elevated preoperative CEA level correlated with poor prognosis (P = .01), as did increasing tumor size (P = .008), lymph node involvement (P < .001), presence of tumor stenosis (P = .003), and lymphatic invasion (P < .001). Using multivariate analysis, only lymph node involvement (P < .001)
and tumor stenosis (P = .003) remained significant, and a trend for poor prognosis was observed for lymphatic invasion (P = .11).
In addition, tumor cell dissemination was investigated in different body compartments. Figure 2 shows typical examples of positive findings in lymph nodes, the peritoneal cavity, and bone marrow.
Micrometastasis could be found in the lymph nodes of 41 patients of 90 patients (45.6%) classified as having N0 disease with conventional histological methods. In these patients, we found a significant correlation between the presence of lymph node micrometastasis and tumor infiltration (P = .02), tumor cell budding (P = .05), tumor shape (insular vs circular, P = .003), and macroscopic tumor type (exophytic vs flat, P = .01).
Cytokeratin 18– and Ber-EP4–positive cells could be detected in the bone marrow specimens of the majority of patients (89 of 140 patients [63.6%]), but the presence of such cells was not related to UICC stage (67% of those with stage I disease had positive results; 63% of those with stage II disease, and 61% of those with stage III disease). No correlation was observed with established histopathological parameters.
Dissemination of tumor cells into the peritoneal cavity was investigated in 132 patients by means of Ber-EP4 cytology and 29 patients (22%) had positive findings. Positivity of peritoneal lavage results correlated with UICC stage (P = .05),
lymph node positivity (N+ vs N−, P = .02),
and presence of a tumor stenosis (P = .03).
Peritoneal dissemination increased with tumor size (T1 and T2, 12%
and 13% vs T3 and T4, 26% and 42%). Furthermore, an elevated preoperative CEA level was observed in conjunction with peritoneal tumor dissemination (33% vs 18% in patients with negative results; P = .08).
Presence of peritoneal cell dissemination was associated with a poorer prognosis; however, differences did not reach statistical significance (P = .07) in the univariate analysis. In contrast, the presence of peritoneal dissemination was not associated with an increased local recurrence rate, and lymph node and bone marrow micrometastases were not associated with a poor prognosis. Of 89 patients with occult metastatic cells in bone marrow,
15 (16.8%) died of cancer, whereas of 51 patients without such cells,
10 (19.6%) died of cancer-related causes (risk ratio = 0.86; P = .44). After 61 months' median follow-up,
the presence of micrometastases in bone marrow was neither associated with death from cancer-related causes nor with locoregional relapse.
An old question in oncology is to know whether surgery will be capable of curing the patient. Obviously, the same procedure performed by the same surgeon in 2 different patients with comparable tumors can result in 2 opposite outcomes, namely cure or death following tumor recurrence. Currently, relying on the UICC classification alone,
a patient-specific prognostic statement is not possible. For example,
current guidelines recommend adjuvant therapy for all patients with UICC stage III disease, although only 40% of them will actually develop metachronous metastasis.2
Against this framework, this study is providing, for the first time to our knowledge, a precise documentation of local, locoregional,
and systemic disease in CRC by examining simultaneously the presence of tumor cells in lymph nodes, the peritoneal cavity, and bone marrow.
Not only does this study document the presence of minimal residual disease in different body compartments after curative resection for CRC, but it investigates the prognostic significance of such presence.
We have drawn 3 conclusions from our study. First, about two-thirds of patients with CRC have systemic disease at the time of surgery,
so CRC should be declared a systemic disease, even in the early stages.
Second, about one-quarter of patients developed metachronous metastasis after so-called curative surgery so they cannot be cured by surgery alone, an observation we made previously.2 Third, it was not possible to detect the patients at risk for metachronous metastasis on the basis of the presence of isolated tumor cells at the time of surgery.
This seems puzzling since local, locoregional, or systemic tumor cell dissemination is thought to be a prerequisite for cancer relapse.
However, several hypotheses may explain the present findings.
First, removing the lymphatic drainage area in patients with N0 disease is curative when micrometastasis is detected in lymph nodes.
In the present study, micrometastatic lymph node involvement highly correlated with pT category but did not correlate with poorer survival.
According to current oncological standards, colorectal surgery includes the resection of lymphatic drainage areas, and therefore, it can be assumed that involved lymph nodes are removed during surgery and,
thus, cannot be a relapse site. However, in about a quarter of patients who have undergone curative resection, removing the lymphatic drainage area is not a sufficient cure and these patients will develop distant metastasis in spite of lymphadenectomy.2 Thus, limited surgery without extended lymphadenectomy might be sufficient in this subgroup of patients. This hypothesis should be tested in appropriate studies. Obviously, in these patients, systemic disease is already present at the time of surgery and this problem can only be addressed with systemic adjuvant chemotherapy.
Second, epithelial cells disseminate in the bone marrow of about two-thirds of patients with CRC, but this finding does not imply a dismal prognosis. Not only can this hypothesis be drawn from the results obtained in this cohort of patients, but it is confirmed in the only meta-analysis published on this topic that we found.11 As a consequence, the malignant nature of disseminated epithelial cells in bone marrow should be questioned.
In this respect, all available detection techniques (immunohistochemical and molecular techniques) use epithelial-specific and not tumor-specific markers. Case reports of bone metastases in CRC are exceptional12 and such metastases are rarely found in routine autopsies.13 Moreover, epithelial cells have been found in the bone marrow of patients with colorectal adenoma.14 However, other observations speak in favor of the malignant nature of disseminated epithelial cells in bone marrow. A cell line was established from such disseminated epithelial cells in CRC and its characterization suggested a malignant phenotype.15 Although the mutational pattern of disseminated epithelial cells differed sometimes from that of paired primary tumors, the presence of K-ras mutations in disseminated tumor cells in bone marrow also documented a malignant genotype.16 Thus, it must be assumed that disseminated epithelial cells in bone marrow are tumor cells, at least in a subgroup of patients, but that these cells do not all have the potential to develop overt metastasis. Finally, a recent meta-analysis of bone marrow in breast cancer17 arrived at exactly the opposite conclusion of the current study on CRC. Taken together, the present findings suggest that the metastasis process is highly inefficient in CRC.
Thus, since it has no independent prognostic significance, the detection of epithelial cells in bone marrow of patients with CRC is not indicated in clinical routine, at least at the present state of knowledge and techniques.
Third, tumor cell dissemination into the peritoneal cavity correlated with disease-free survival. Although it did not reach independent statistical significance, this is an interesting finding since such dissemination is not part of routine staging systems in CRC (although proposed as an additional description in the commentaries of the UICC18). Of course, the curative nature of surgical resections in CRC (R0 resection) can be challenged if remaining tumor cells can be documented in the peritoneal cavity of 14% to 34% of patients at the time of surgery, depending on tumor stage. Since cell dissemination as a consequence of intraoperative tumor perforation has been reported to increase the incidence of local recurrence both in colon19 and in rectal20 cancer surgery, locoregional peritoneal spreading would be expected to be clinically significant.
In summary, the present study conducted in a well-defined, large prospective cohort of patients with a mature and complete follow-up shows that CRC cells disseminate at early tumor stages into various body compartments, a majority of patients with CRC are presenting micrometastatic disease at the time of surgery, and residual tumor cells are left in the body of a significant proportion of patients after curative surgery. However, presence of dissemination is not sufficient by itself to define a patient-specific risk for metachronous metastasis after curative surgery.
Further studies are necessary to clarify the metastatic potential of disseminated tumor cells. In particular, longer follow-up of the various groups studied will be of interest. There is a hope that future studies will be facilitated by the availability of emerging tumor-specific (and not epithelial-specific) protein or genetic markers. This will help better define local and systemic CRC, set the limits of surgery between local and systemic disease, and be helpful in predicting the need for adjuvant systemic therapy.
Correspondence: Marc A. Reymond,
MD, MBA, Department of Surgery, Evangelic Hospital Bielefeld, Burgsteig 13, 33617 Bielefeld, Germany (email@example.com).
Accepted for Publication: May 30, 2006.
Author Contributions:Study concept and design: Steinert, Hantschick, Gastinger,
and Reymond. Acquisition of data: Steinert,
Hantschick, Gastinger, and Kühnel. Analysis and interpretation of data: Steinert, Vieth, Gastinger, and Lippert. Drafting of the manuscript: Steinert,
Hantschick, and Kühnel. Critical revision of the manuscript for important intellectual content: Steinert,
Hantschick, Vieth, Gastinger, Lippert, and Reymond. Statistical analysis: Steinert and Kühnel. Obtained funding: Lippert. Administrative,
technical, and material support: Steinert, Hantschick, Vieth,
Gastinger, and Kühnel. Study supervision: Steinert, Hantschick, Vieth, Gastinger, and Reymond.
Financial Disclosure: None reported.