Micrometastasis Volume in Lymph Nodes Determines Disease Recurrence Rate of Stage II Colorectal Cancer: A Prospective Multicenter Trial

Metastasis to regional lymph nodes is a reliable prognostic factor, adopted globally in the tumor–node–metastasis (TNM) staging system and used for clinical decision-making regarding postoperative adjuvant chemotherapy (1–3). Studies have shown the presence of occult cancer metastasis (also designated as micrometastasis) in the lymph nodes of colorectal cancer patients (4).

With IHC and molecular genetics methods, occult tumor cells in regional lymph nodes are detected in 25% to 50% of patients with node-negative colorectal cancer (5–18). Because up to 25% of patients with node-negative colorectal cancer ultimately die as a result of disease relapse, occult cancer metastasis has been suspected as a potential marker for systemic spread of tumor cells (1). However, the prognostic value of molecular tumor cell detection in patients with node-negative colorectal cancer has remained uncertain because of lack of evidence from prospective studies (5–18).

National Comprehensive Cancer Network (NCCN) clinical practice guidelines in oncology (version 4, 2013; ref. 19) recommends that detection of cancer cells by IHC or by the molecular detection of cancer cells in regional lymph nodes should be considered investigational because most of evidence cited in this guideline is constructed by retrospective studies which include our works on distribution and prognostic impact of micrometastases in node-negative colorectal cancers by IHC for cytokeratins or carcinoembryonic antigen (CEA)-targeted RT-PCR (7, 20). In addition, a recent meta-analysis by Rahbari and colleagues emphasized that prospective studies are required to confirm their results that micrometastasis was associated with poor overall survival (OS) and shorter disease-free survival (DFS) in node-negative colorectal cancers (21).

In this study, we conducted a prospective multicenter clinical trial for a definite prognostic marker in patients with node-negative stage II colorectal cancer. To search for the clinically appropriate threshold of high risk for disease recurrence, we employed molecular detection of CEA mRNA by qRT-PCR in addition to conventional qualitative RT-PCR (the so-called “a band” method). A total of 296 patients with pathologic stage II colorectal cancer were eventually analyzed for their prognosis. After a long follow-up period (median follow-up, 82.3 months), we uncover the prognostic impact of micrometastases in stage II colorectal cancer.

After obtaining written informed consent on lymph node sampling and the subsequent clinical trial, lymph node samples were collected during surgery from 419 potential candidates who were preoperatively evaluated as stage II or stage III [International Union Against Cancer (UICC) TNM Classification - the 7th version; ref. 3] by examination of colonoscopy and CT scan, with performance status 0-2 (age, 20–75) from November 2001 to December 2005. After surgery, 315 patients diagnosed with pathologically node-negative stage II colorectal cancer were enrolled (Fig. 1). Experienced surgical staff from the 21 hospitals participated in this study. The study protocol was approved by each local institutional review board and this study was conducted in accordance with the Good Clinical Practice guidelines and the Declaration of Helsinki.

Lymph nodes were collected, fixed in 10% buffered formalin, and embedded in paraffin for routine histopathologic examination. Mean lymph node sampling number inspected in each colorectal cancer case was 16.0. Of them, 5 pericolic nodes adjacent to the tumor were further examined for micrometastasis to detect initial cancer spread. For this purpose, a half of lymph node was collected using a RNase-free disposable lymph node sampling kit within 3 hours after surgical resection as concluded by our early work (22), being preserved in a tube containing RNA later (Ambion) at −20°C until RNA extraction (Supplementary Fig. S1).

Once pathologic tests indicated node-negative stage II colorectal cancer, lymph nodes were minced in TRIzol Reagent (Invitrogen). RNA extraction was performed as described previously (23, 24).For assessment of RNA quality, RNA was electrophoresed on a 0.8% agarose mini-gel, and ribosomal RNAs at 28S and 18S were visualized with ethidium bromide (Supplementary Fig. S1). Among 315 enrolled stage II colorectal cancer patients, lymph node samples from 11 patients showed degradation (3.5%; Fig. 1).

Reverse transcription (RT) was performed as described previously (22–24).

Conventional PCR and the primer sequences for PCR amplification of CEA and the housekeeping gene porphobilinogen deaminase (PBGD) were described previously (7, 22–23, 25–27). PCR products were electrophoresed on a 2% agarose gel and visualized with ethidium bromide (Supplementary Fig. S1). To ensure the reproducibility of PCR, the CEA control panel was produced using a serial dilution of cDNA from MKN45 and LoVo cells (22) and referenced to estimate the positivity of the CEA band in colorectal cancer cases (Supplementary Fig. S1).

Fluorescence PCR was performed using the Light Cycler (Roche Diagnostics), as described previously (22, 24, 26).The standard curves for quantification of CEA or PBGD mRNAs were drawn using 10-fold dilutions of cDNA from MKN45 cells (22, 26).We confirmed that CEA mRNA was reproducibly measurable in the range of a dilution of 1 × 10⁻¹ to 1 × 10⁻⁴ (Supplementary Fig. S2).

Postoperative adjuvant chemotherapy was randomly applied to the colorectal cancer patients with positive band for CEA mRNA using the oral 5-fluorouracil derivative HCFU (1-hexylcarbamoyl-5- fluorouracil; refs. 28, 29). Thus, if the patients agreed, they were randomly allocated to either chemotherapy (HCFU 300 mg/day for 1 year) or no treatment (Fig. 1). If they did not agree to allocation, we followed each patient's desire and the study protocol was fixed that they were not excluded from the study. The protocol prohibited use of chemotherapy in CEA band-negative cases.

Among 304 patients whose CEA levels in lymph nodes were determined, several patients were judged as not meeting criteria for the purpose of this study. These included protocol violation (applied chemotherapy in CEA band-negative case, n = 5), insufficient follow-up period (5 months; CEA band-positive case, n = 1), and early death from other disease (2 months and 10 months, respectively; CEA band-negative cases, n = 2). As a result, 296 colorectal cancer patients were analyzed for their disease recurrence and survival (Fig. 1). The patients were periodically monitored by physiologic examination, CT scan, chest x-ray, colonoscopy, and blood tests. The median follow-up period was 82.3 (5.3–133.2) months.

This study was registered with University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR). Clinical trial information is referenced by UMIN C000000185.

Statistical analysis was performed using the StatView J-5.0 program (Abacus Concepts, Inc). The Kaplan–Meier method was used to estimate tumor recurrence and cancer-specific survival, and the log-rank test was used to examine statistical significance. A Cox proportional hazards model was used to assess the risk ratio under simultaneous contribution from several covariates. The associations between the discrete variables were assessed using the χ² test. Mean values were compared using the Mann–Whitney test. All data were expressed as the mean ± SD. Values of P < 0.05 denoted the presence of a statistically significant difference.

The positivity for the CEA band was assessed in reference to the control panel using CEA-expressing cells (Supplementary Fig. S1). Of 304 stage II colorectal cancer patients, 73 cases (24.0%) displayed a clear CEA band with expression of the PBDG gene, and the remaining 231 cases did not by conventional RT-PCR assays (Fig. 1).

CEA transcripts were measured by qRT-PCR and normalized by PBDG gene expression. The CEA/PBDG value, designated as the micrometastasis volume (MMV) ranged widely from 0 (group a, n = 90) to >1 × 10⁻⁴ (group b, n = 111), >1 × 10⁻³ (group c, n = 62), and >1 × 10⁻² (group d, n = 33). We grouped categories c and d as the high MMV group (n = 95, 32.1%) and categories a and b into the low MMV group (n = 201, 67.9%). A clinicopathologic survey indicated that high MMV was significantly associated with the presence of lymphatic invasion (P = 0.030) and smaller tumor size (P = 0.009). MMV was not associated with the presence of ASCO risk factors (Table 1).

After exclusion of inappropriate cases (n = 8, Fig. 1), 72 CEA band-positive cases and 224 CEA band-negative cases were analyzed for DFS and OS. We found that the CEA band-positive cases had significantly worse 5-year DFS and final OS than the CEA band-negative cases (P = 0.021, 0.030, respectively, Fig. 2A and C), but there was no significant difference in 5-year OS (P = 0.287, Fig. 2B). Of CEA band-positive cases, 27 cases received postoperative chemotherapy by random allocation using HCFU for 1 year, whereas 27 cases were randomly allocated to no treatment. There was no significant difference in 5-year DFS and 5-year OS between the two groups (Supplementary Fig. S4). There was no difference in mean follow-up period in the CEA band-positive and band-negative groups (P = 0.660; mean ± SD, 79.2 ± 24.4 vs. 79.9 ± 22.9 months).

Survival analyses indicated that the high MMV group had significantly worse 5-year DFS, 5-year OS, and final OS as compared with the low MMV group (P = 0.001, 0.016, 0.003; Fig. 3A–C). Irrespective of chemotherapy applied to high MMV cases, high MMV group consistently showed significantly worse 5-year DFS as compared with the low MMV group with no treatment (P = 0.009, 0.007; Supplementary Fig. S5). Similar results were observed in analyses for OS (P = 0.003, 0.026; data not shown). There was no difference in mean follow-up period in high MMV and lowMMV groups (P = 0.8471; mean ± SD, 86.0 ± 25.8 vs. 79.9 ± 22.1 months).

Among various clinical and pathologic characteristics, presence of lymphatic invasion was a significant risk factor for shorter 5-year DFS (P = 0.033) in addition to high MMV (P = 0.001), but ASCO risk factor was not (Table 2 and 3). Multivariate analyses indicated that high MMV alone was retained as an independent risk factor (P = 0.004, Table 2). As for 5-year OS and final OS, high MMV alone was significantly associated with poor prognosis (P = 0.020, 0.004; risk ratio 3.097, 3.340, respectively; Table 3).

In addition, the micrometastasis-free colorectal cancer group (group a) had a favorable prognosis. Five-year DFS was 93.3%; 5-year OS and final OS were both 97.8% (Supplementary Fig. S3A–S3C).

Among 296 patients, 51 developed disease recurrence after surgery. The recurrent organ sites are listed in Supplementary Fig. S6. Like more advanced-stage patients, stage II patients often had hematogenous metastasis to liver and lung.

We have demonstrated clinical significance of micrometastasis in stage II colorectal cancer as a result of long-term follow-up. We are not aware of other reports on prospective multicenter studies that followed patient outcomes for a duration as long as our clinical trial (median, 82. 2 months). We found that MMV was a particularly useful biomarker for discriminating both high risk (group c+d) and low risk (group a) for disease recurrence in stage II colorectal cancer. The recurrent mode was mainly hematogenous metastasis to liver and lung in stage II colorectal cancer patients, which is the serious problem as is the case with stage III colorectal cancer.

To detect occult tumor cells in lymph nodes, we employed CEA-targeted molecular detection by RT-PCR because our retrospective study showed that micrometastasis detected by RT-PCR, but not by IHC, was associated with poor prognosis in stage II colorectal cancer (7). In routine practice, RT-PCR is more convenient than IHC because IHC of occult cancer cells requires the extraordinary labor of preparing slides and searching minimal tumor cells under the microscopy. Moreover, the advantage of the molecular genetic technique is that it allows examination of a large quantity of lymph nodes; in this way, unstable features of IHC caused by the number of inspected slices can be avoided (20).

In addition to the routine pathologic lymph node examination (mean inspection number, 16.0 lymph nodes per case), we examined representative 5 lymph nodes near the primary tumor for detection of micrometastasis. We made this choice based on detailed lymph node mapping of micrometastasis by IHC that showed that the initial occult tumor spread was found at pericolic nodes adjacent to the tumor (level 1) in 25 of 26 node-negative colorectal cancer patients (96.2%; Supplementary Fig. S7), which suggests that micrometastasis is initially generated around the tumor. We also considered that molecular inspection of additional 5 lymph nodes on current pathologic examination would be clinically applicable as a routine practice.

Through our previous translational studies (22), we decided that lymph nodes should be collected within 3 hours after tumor resection to maintain the quality of RNA materials. Surgeons were instructed before the start of this clinical trial to retrieve lymph nodes using the RNase-free disposable lymph node sampling kit (Supplementary Fig. S1; ref. 22). As a result, RNA quality was preserved in 304 of 315 colorectal cancer cases (96.5%), which we think is a sufficiently acceptable range in clinics. For clinical application, qRT-PCR is suitable because the clinical useful threshold was better provided by qRT-PCR, than the conventional RT-PCR, suggesting that adjustment of CEA value by the appropriate internal control is essentially important.

Studies have not supported a rationale for application of chemotherapy for all stage II colorectal cancer patients (30, 31).However, 25% of stage II patients ultimately die from the disease recurrence (1).This dilemma has been carried over for decades. ESMO (European Society for Medical Oncology) guideline and a proposal from ASCO in 2004 introduced several prognostic factors such as perforation, T4 lesions, poorly differentiated histology, inadequately sampled nodes to identify high-risk stage II colon cancer (30, 31). Univariate analysis showed that high MMV and lymphatic invasion were indicative of shorter 5 year, but ACSO risk factor was not a significant prognostic parameter. Multivariate analysis indicated that high MMV alone was an independent prognostic factor in stage II colorectal cancer.

A study by O'Connor and colleagues showed that 75% of the 24,847 stage II cancers had one or more poor prognostic features of perforation, T4 stage, poor histology, or others, and that survival benefit from adjuvant chemotherapy was not observed even with any poor prognostic features (32). The findings indicate that prognostic factors do not always predict patients' survival benefit by adjuvant chemotherapy. Although we applied postoperative chemotherapy by randomization to CEA band-positive groups, no significant difference was noted in patients' prognosis (Supplementary Fig. S4). In addition, survival analyses of the entire CEA band-positive groups (31 chemotherapy vs. 41 no treatment) showed similarly insignificant results (Supplementary Fig. S8). This is because of small size of the PCR-positive group and more likely to be due to a sort of adopted therapeutics. The current standard adjuvant therapy including FOLFOX, capecitabine, or UFT tegafur-uracil/UZEL (33–35) did not emerge in clinics those days when this study started. We used HCFU which was later proved to be less effective even than UFT alone (36)and it is not available today. Therefore, to obtain the correct answer, a standard chemotherapy should be applied to the more numbers of PCR-positive cases in the next stage.

Although we could not determine an efficacy of chemotherapy in high MMV cases, we still consider that such group may be an appropriate target for a postoperative adjuvant chemotherapy for the following reasons. First of all, lymph node metastasis in stage III colorectal cancer is a well-established predictive marker to ensure survival benefit by chemotherapy (33, 37–38). We previously found that one-step nucleic acid amplification (OSNA) of cytokeratin 19 (CK19) mRNA provided a judgment performance equivalent to a 2-mm-interval histopathologic examination of lymph nodes of colorectal cancer (39). In our separate multicenter prospective study, we found by OSNA that tumor volume in lymph nodes of colorectal cancers upstaged from stage II to stage III (17.6%, 13/76) ranged widely. Of interest was that the values were largely overlapped with those in lymph nodes of stage III colorectal cancers (40). The prognosis of these upstaged colorectal cancer patients is uncertain until 3 to 5 years ahead. Importantly, however, the current study uncovered that MMV determined by qRT-PCR for CK19 mRNA was also useful to discern high-risk stage II colorectal cancer patients in the same clinical settings (Supplementary Fig. S9). In this context, OSNA could substitute CEA targeted RT-PCR owing to its simple and rapid system and would be helpful to confirm efficacy of chemotherapy in colorectal cancers upstaged from stage II to stage III in a large-scale trial in the future.

In conclusion, we prospectively demonstrated that molecular detection of MMV adds prognostic value to the conventional histologic survey in stage II colorectal cancer. Our data also provide a concept that stage II is a transitional place between localized stage I and expanding stage III, where colorectal cancer tumors continuously increase MMV in lymph nodes and the risk for tumor recurrence.

No potential conflicts of interest were disclosed.

Conception and design: H. Yamamoto, K. Murata, T. Ohnishi, T. Kato, M. Ikeda, M. Sekimoto, N. Matsuura, M. Monden, M. Mori

Development of methodology: H. Yamamoto, T. Ohnishi, T. Kato, Y. Nakamura, M. Ikeda

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Yamamoto, K. Murata, M. Fukunaga, T. Ohnishi, Y. Miyake, T. Kato, Y. Nakamura, I. Takemasa, M. Ikeda, M. Ohue, M. Monden, M. Mori

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Yamamoto, T. Kato, M. Ohtsuka, Y. Nakamura, M. Sekimoto, N. Matsuura, M. Monden, M. Mori

Writing, review, and/or revision of the manuscript: H. Yamamoto, T. Kato, Y. Nakamura, T. Mizushima, R. Nezu, M. Monden

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Kato

Study supervision: T. Kato, T. Mizushima, R. Nezu, N. Matsuura, M. Monden, Y. Doki, M. Mori

Other (technical advice and patients follow-up): S. Noura

Other participants in the clinical study are:

Takayuki Fukuzaki: Department of Surgery, Osaka Saiseikai Senri Hospital, Osaka, Japan.

Taishi Hata: Department of Surgery, Gastroenterological Surgery, Graduate School of Medicine, Osaka University, Osaka, Japan.

Kimimasa Ikeda: Department of Surgery, Minoh City Hospital, Osaka Japan.

Masakazu Ikenaga: Department of Surgery, Higashiosaka City General Hospital, Osaka, Japan.

Masanori Kishibuchi: Department of Surgery, Shitennoji Hospital, Osaka, Japan.

Toshiharu Morita: Department of Surgery, Toyonaka Municipal Hospital, Osaka, Japan.

Hirofumi Ohta: Department of Surgery, Ikeda Municipal Hospital, Osaka, Japan.

Shu Okamura: Department of Surgery, Suita Municipal Hospital, Osaka, Japan.

Masaki Okuyama: Department of Surgery, Kaizuka City Hospital, Osaka, Japan.

Nobuo Tanaka: Department of Surgery, Itami City Hospital, Hyogo, Japan.

Yoshio Uemura: Department of Surgery, Kinki Central Hospital of the Mutual Aid Association of Public School Teachers, Hyogo, Japan.

Makoto Watase, Ogawa A: Department of Surgery, Tane General Hospital, Osaka, Japan.

This work was supported by MEXT KAKENHI under grant number 10307028.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.