Colorectal cancer is one of the most common malignancies worldwide with ∼490,000 of affected patients dying every year.1 Patients with colorectal cancer UICC stage II are generally considered to be at low risk for developing recurrent or metastatic disease.2 Therefore, patients with colon cancer in this stage, in contrast to patients with positive lymph nodes (UICC stage III), are not recommended to undergo routine adjuvant chemotherapy.2, 3, 4, 5, 6 For patients with rectal cancer, there is now a trend towards neoadjuvant radiation therapy. In rectal cancer patients with negative lymph node status, who have undergone neoadjuvant radiation, the value of postoperative chemotherapy again is not yet adequately defined.3 Despite the low tumor stage some of the stage II patients subsequently develop recurrent disease. It remains under debate, whether all stage II patients or at least stage II patients with additional risk factors should receive adjuvant chemotherapy.7, 8, 9 It is therefore of high importance to define prognostic criteria for this patient group to help identify high-risk patients for tumor relapse, who might benefit from adjuvant therapeutic regimes.10, 11
Many molecular markers have been evaluated as putative prognostic markers in patients with colorectal cancer; however, none of these are currently in clinical use regarding the decision whether a patient with colorectal cancer stage II should receive adjuvant chemotherapy.1 Tumor cell dissemination and formation of metastases is known to be a sequence of complex events, therefore, single molecular markers might not be able to serve as adequate prognostic markers. In addition, other factors, such as the immune competence of the patients, will influence prognosis.12 We, have therefore used a different approach hypothesizing that detection of circulating tumor cells might identify patients at high risk for relapse. The potential advantage of this approach is that tumor and patient related factors that may be of prognostic relevance would both influence the presence of circulating tumor cells, therefore, detection of these cells is a logical approach to identify patients at higher oncological risk. PCR based protocols have been proven to be sensitive and specific assays for detecting disseminated colorectal cancer cells. The prognostic significance of these cells, however, remains under critical debate in the literature.13 Recently, we demonstrated the sensitivity and specificity of a CK 20 reverse transcriptase-polymerase chain reaction (RT-PCR) assay in detecting colorectal cancer cells in blood and bone marrow.14, 15, 16, 17, 18, 19
The aim of our prospective study was to evaluate the prognostic significance of disseminated tumor cells detected by CK 20 RT-PCR in blood and bone marrow in patients with stage II colorectal cancer.
Material and methods
Included in this study were 90 consecutive patients undergoing elective curative (R0) resection of colorectal adenocarcinoma stage II at the Department of Surgery, University of Heidelberg, Germany, between September 1995 and February 2000. Patients with other malignant disease in their medical history were excluded. Informed consent was obtained from all patients; the study protocol was approved by the ethics committee of the University of Heidelberg. The PCR-results of a subgroup of these patients (n = 25) were already reported in earlier works from our group.14, 15
Tumors located less than 16 cm above the anal verge were classified as rectal cancers. Tumor height of rectal cancers was determined using rigid endoscopy performed by the surgeon preoperatively. All patients were staged preoperatively with colonoscopy, abdominal ultrasound and/or CT scan and chest X-ray.
No patient received treatment prior to surgery. According to recommendations for stage II patients, valid at the time patients were included in the study, adjuvant therapy was not recommended for patients with colon cancer and adjuvant chemoradiation was recommended for patients with rectal cancer. Adjuvant therapy was for the most part performed at the discretion of the local oncologist.
The length of follow-up was calculated from the date of operation at our institution. Follow-up was done according to the guidelines of the German Cancer Society, which consists of periodic clinical evaluation, serum CEA levels, imaging of the abdomen, chest X-ray and colonoscopy.
Occurrence of distant metastases and local recurrences, follow-up time and reason for death was obtained for each patient to assess relapse-free and disease-specific survival.
Tumor stage and grade were classified according to the 5th edition of the TNM classification of the UICC (International Union Against Cancer).20
Blood and bone marrow samples
Through a central venous catheter in the superior vena cava, which is routinely placed during colorectal cancer operations in our hospital, 3 blood samples (10 ml) were obtained from each patient: the first after induction of anesthesia, the second after resection of the tumor and the third 24 hr after the operation. Bone marrow samples (10 ml) were obtained after induction of general anesthesia by aspiration from both iliac crests. The blood and bone marrow samples were diluted with 10 ml phosphate buffered saline (PBS). After density centrifugation through Ficoll-Paque (Pharmacia) (30 min, 400 g), mononuclear cells were harvested from the interphase and washed twice in PBS. The cell pellet was then shock frozen in liquid nitrogen and stored at −70°C.
Seventy-one patients consented to bone marrow puncture. Of the blood samples, 89 preoperative samples, 88 intraoperative samples and 82 postoperative samples could be evaluated by CK 20 RT-PCR. The missing data in regard to the blood samples are either due to unsuccessful RNA extraction or reverse transcription or due to organizational or patient related reasons, resulting in non-acquisition of the blood samples.
RNA extraction from peripheral mononuclear blood cells, from bone marrow samples and frozen tissue sections of tumors was done as previously described.14 CK 20-RT-PCR was performed as previously described.14 PCR-products were analyzed by electrophoresis on 2% agarose gels. RNA quality and performance of reverse transcription of all analyzed samples was confirmed by RT-PCR amplification of glyceraldehyde phosphate dehydrogenase (GAPDH) transcripts as previously described.14
Sensitivity and specificity of CK 20 RT-PCR
The sensitivity of the CK 20 RT-PCR assay was determined in previous cell spiking experiments, allowing the detection of 10 colon cancer cells (HT 29) in 10 ml blood.14
The specificity of the CK 20-RT-PCR assay was determined in previous studies, in which 174 blood samples from 98 healthy individuals and bone marrow samples from 30 patients without malignant disease consistently tested negative for CK 20 expression.14, 15, 21
Statistical computations were done using the software package JMP (JMP, Cary, NC). Survival was estimated according to the Kaplan-Meier method and compared using the log-rank test.22 A multivariate proportional hazards model was built using the variables that were significant prognostic parameters in the univariate analysis (p ≤ 0.05 at univariate analysis).23 A result was considered statistically significant when the p-value was less than or equal to 5% (p ≤ 0.05).
The patient characteristics of the 90 patients included in this study are displayed in Table I. The median age of the patients was 66 (range: 29–85 years). The median number of lymph nodes removed was 19 (range: 6–65). Of 52 patients with colon cancer, 5 received adjuvant 5-FU-based chemotherapy; of 38 patients with rectal cancer, 21 received adjuvant chemoradiation and 2 adjuvant radiation therapy.
Table I. Distribution of Clinical, Pathological and Treatment Factors in 90 Patients With Stage II Colorectal Cancer Undergoing R0-Resection
No. of patients
Lymphatic vessel invasion
Mucin producing carcinoma
Number of removed lymph nodes
≥12 lymph nodes
<12 lymph nodes
Tumor cell detection in blood and bone marrow samples
Tumor cells were detected in preoperative blood samples of 21 of 89 (24%) patients. Twenty-two of 88 (25%) patients revealed disseminated tumor cells in blood samples taken intraoperatively. Twenty-eight of 82 (34%) patients showed tumor cells in postoperative blood samples (taken 24 hr after the operation). Of the 11 patients with a tumor recurrence, 9 patients (82%) had at least 1 positive blood sample; 5 of 11 patients (45%) revealed a positive preoperative blood sample, 5 of 11 patients (45%) a positive intraoperative blood sample and 8 of 11 patients (73%) a positive postoperative blood sample.
Twenty of 71 (28%) patients tested positive for disseminated tumor cells in their bone marrow samples. Patients with positive blood samples had positive bone marrow samples in 35% of cases compared to 20% in patients with negative blood samples (p = ns).
Follow-up and univariate characterization of prognostic factors
The median follow-up for all patients was 58 months with a range from 3 to 81 months. The actuarial 5-year relapse-free survival was 85% (95% CI, 77–93%), and the actuarial 5-year disease-specific survival was 93% (95% CI, 87–99%). At the time of the last follow-up, 70 patients (78%) had no evidence of disease (5 of these patients had recurred, however, were rendered free of disease by surgical resection), 6 patients (7%) had died of disease and 12 patients (13%) had died of unrelated causes. Two patients (2%) were lost to follow-up; these patients were censored at the time of last follow up (35 and 50 months postoperatively). Table II depicts the univariate analysis of the different factors for relapse-free and disease-specific survival. Regarding relapse-free survival T-category, number of removed lymph nodes and tumor cell detection in blood samples showed prognostic relevance. The actuarial 5-year relapse-free survival was 93% compared to 70% in patients with negative versus positive postoperative blood samples (p = 0.003) (Fig. 1). Number of removed lymph nodes and tumor cell detection in postoperative blood samples were also prognostic criteria for disease-specific survival. In patients with negative postoperative blood samples, the actuarial 5-year disease-specific survival was 100% compared to 77% in patients with a positive postoperative blood sample (p = 0.0006) (Fig. 1). Positive postoperative blood samples were associated with a worse disease-specific survival irregardless of whether the patients had received adjuvant therapy or not. This was also the case when looking at colon and rectal cancer patients separately (Figs. 2 and 3). Positive postoperative blood samples were also associated with a worse relapse-free survival in patients with rectal cancer and those receiving adjuvant therapy.
Table II. Univariate Analysis of Prognostic Factors in 90 Patients With Stage II Colorectal Cancer Undergoing R0-Resection
Multivariate characterization of prognostic factors
Separate multivariate analyses were performed for tumor cell detection in any blood sample (pre-, intra- and postoperative sample) and for tumor cell detection in postoperative blood samples.
The model for relapse-free survival is shown in Table III. This model identified tumor cell detection in blood samples as an independent risk factor. The T-category reached statistical significance only in the model analyzing tumor cell detection in any blood sample. The number of removed lymph nodes failed to reach statistical significance, but showed a strong trend.
Table III. Multivariate Analysis of Factors Associated With Relapse-Free Survival Including Tumor Cell Detection in Any Blood Sample (Model I) or In Postoperative Blood Samples (Model II)
Hazard ratio (95% CI)
Hazard ratio (95% CI)
ns, not significant.
2.2 (1.1, 4.1)
1.8 (0.8, 3.2)
Number of removed lymph nodes
1.7 (0.8, 3.2)
2.0 (0.9, 4.0)
Tumor cell detection
any blood sample
2.1 (1.1, 5.4)
Tumor cell detection
Postop. blood sample
2.4 (1.3, 5.3)
Table IV depicts the results of the multivariate model for disease-specific survival. Number of removed lymph nodes, and tumor cell detection in postoperative blood samples were independently associated with disease-specific survival.
Table IV. Multivariate Analysis of Factors Associated With Disease-Specific Survival
Hazard ratio (95% CI)
Number of removed lymph nodes
7.7 (2.5, 38)
Tumor cell detection
Postop. blood sample
6.4 (2.2, 34)
In contrast to colorectal cancer patients with positive lymph nodes (stage III), where the benefit of adjuvant chemotherapy has been proven, the use of adjuvant chemotherapy for patients with stage II remains controversial. This is due to the favorable prognosis of stage II patients; optimized surgery without adjuvant chemotherapy can result in disease-specific survival rates of over 90% in this patient group.24 Therefore, many patients who will never develop recurrent disease will have to undergo adjuvant chemotherapy to treat one patient who might benefit from adjuvant therapy. This dilemma is reflected by the available data from randomized trials studying the effect of adjuvant chemotherapy in colorectal cancer stage II patients; no consistent significant benefit of adjuvant chemotherapy has yet been demonstrated.3, 4, 5, 6, 7, 8, 10 Some authors, however, suggest that chemotherapy can lead to a similar improvement of survival for stage II patients as demonstrated for stage III patients.25 For patients with rectal cancer, the situation is even more complicated as there are several different radiation regimes available. For patients receiving neoadjuvant radiation, the role of adjuvant chemotherapy is currently unclear, again underlining the importance of defining valid prognostic criteria for these patients. Prognostic markers may help to identify a subgroup of stage II patients at high risk for disease relapse who may also benefit from adjuvant therapy. Several clinical, pathological and molecular prognostic factors have been discussed in the literature; however, no single factor has yet been identified which is generally accepted as an indicator for initiating adjuvant chemotherapy.1, 4, 8, 11, 25, 26 Using single molecular markers to describe the prognosis of patients with colorectal cancer might not be successful, as tumor cell dissemination and formation of metastases are the result of multiple steps involving tumor and patient related factors. In addition, cancers may be very heterogeneous; therefore, assessment of the primary tumor may not accurately predict the presence or behavior of disseminated cancer cells, which are the primary target of adjuvant chemotherapy. The occurrence of circulating tumor cells is the result of 2 basic competing factors: first, the ability of the cancer cells to disseminate, and secondly, the effectiveness of tumor cell elimination mechanisms, especially of the immune system. Therefore, detection of circulating cancer cells seems to be a logical approach to assess prognosis of the patients to define surrogate markers for indicating adjuvant therapy.
Lindemann et al. were able to detect disseminated colorectal cancer cells in bone marrow aspirates of patients with colorectal cancer using immunocytological methods.27 The authors also demonstrated the prognostic relevance of these cells; however, other authors could not confirm this finding making the detection of these cells not a generally accepted prognostic marker.13, 26, 28 The prognostic impact of tumor cells circulating in the blood stream is even more controversial. Opponents of the prognostic value of circulating tumor cells argue that implantation of circulating tumor cells is a very inefficient process with less than 0.01% of circulating tumor cells actually establishing metastases.29, 30 Similar to tumor cell detection in lymph nodes and bone marrow, detection of tumor cells in the blood stream is still not an accepted prognostic marker.13, 26, 28 A major problem of most of the published studies is that only small, inhomogeneous patient groups with short follow-up periods were evaluated. Recently published studies report conflicting results regarding the prognostic value of circulating tumor cells.31, 32, 33, 34 Moreover, the methods used for tumor cell detection also need to be taken into account, as sensitivity and specificity are of major importance and may differ significantly.
Our study prospectively examined a homogenous group of patients with stage II colorectal cancer undergoing R0-resection with a median follow up of almost 5 years. We used a CK 20 RT-PCR assay for tumor cell detection in blood and bone marrow samples. In our previous studies, we have demonstrated the sensitivity and specificity of this assay. In these studies, 174 blood samples of 98 individuals and bone marrow samples of 30 patients without malignant disease consistently tested negative for CK 20 expression.14, 15, 16 Blood samples taken intraoperatively from patients having surgery for benign colorectal diseases also revealed a negative finding in these studies.
Some authors argue that colon and rectal cancer should not be studied together as they might show a different biologic behavior. We, therefore, analyzed site of tumor (colon versus rectum) as a potential prognostic marker. No prognostic relevance could be demonstrated in our patient cohort on univariate analysis; therefore, site of tumor did not influence the outcome in this study and was not included in the multivariate model.
When evaluating a new prognostic marker, it is essential to demonstrate independence from other prognostic parameters by multivariate analysis. Multivariate analysis revealed 3 independent prognostic markers in our patient cohort: number of removed lymph nodes, T-category and tumor cell detection in the blood stream. The first 2 parameters are well-defined markers, with proven prognostic relevance in the literature.26, 35 The present study is the first study demonstrating tumor cell detection in blood samples to be an independent prognostic marker in stage II colorectal cancer patients for both relapse-free and disease-specific survival. Of special interest is that tumor cell detection in blood obtained 24 hr postoperatively was the most relevant time point of sampling in regard to prognosis. The 5-year disease-specific survival for patients with a negative postoperative blood sample was 100% compared to only 77% in patients with detectable tumor cells at that sampling time. It is well known from the literature that the majority of circulating tumor cells are cleared from the blood stream within 24 hr.36, 37 Possible reasons for the persistence of disseminated tumor cells in the circulation may be an immune escape mechanism or a higher malignant potential of this specific clone. One could therefore contemplate that circulating tumor cells which survive for a longer period of time in the blood stream have higher prognostic relevance. In contrast to the prognostic significance of tumor cells detected in the blood stream, we were not able to find the same effect for tumor cells detected in bone marrow samples. We can only speculate on the reason for this discrepancy. It is well known that the vast majority of tumor cells in the bone marrow are in a dormant state and do not proliferate.38 One could hypothesize that tumor cells that are able to survive in the blood stream as described above might have a greater prognostic impact compared to dormant tumor cells in the bone marrow. Further studies have to clarify this issue.
In conclusion, the results of this study support a prognostic significance of tumor cells detected in blood samples of patients with stage II colorectal cancer. These findings, once confirmed by further studies, could be used as a basis for conducting a randomized trial evaluating the effect of adjuvant chemotherapy in stage II patients with proven circulating tumor cells.
The authors thank Professor M.W. Büchler, M.D. for his support of our work.