Lymph node involvement is an important prognostic factor in patients with colorectal cancer, and many studies have indicated that the number and location of metastatic nodes affect prognosis.1–4 In the latest TNM system,5,6 patients with histologically involved nodes have been subdivided into two groups: those with one to three involved nodes (N1), and those with four or more (N2).

Recently, new technologies have been developed, including immunohistochemistry and other special techniques, such as in situ hybridization, which allow the identification of carcinoma cells in lymph nodes, previously undetected by routine histology.7–9 These relatively new and highly sensitive approaches, although proved to be useful and valid in research settings, are not yet available in routine clinical practice. Moreover, the clinical significance of these findings is still controversial.

Routine histologic diagnosis is based only on the identification of cancer cells in lymph nodes. Pathologists diagnose the lymph nodes as positive or negative, but further meticulous morphologic investigation is rarely performed. Komuta et al.10 have reported the prognostic value of extracapsular invasion of positive lymph nodes in patients with resectable colorectal cancer; however, it seems that their method of categorization based on histopathologic findings is complex.

This study was designed to refine the prognostic subsets in Stage III colorectal cancer patients and, for this purpose, we retrospectively reviewed a consecutive series of patients who underwent curative resection and were histologically proved to be node-positive. All the positive lymph nodes were evaluated concerning the size and patterns of microscopic tumor growth, especially focusing on lymph node capsular invasion, and their prognostic value was investigated.

Patients and Methods

During the period from April 1995 to December 2000, 406 consecutive patients with primary colorectal cancer received curative (R0) resection at the Division of Colorectal Surgery, the International Medical Center of Japan. All patients had T1, T2, T3, or T4 adenocarcinoma, and patients with distant metastasis at the time of diagnosis and those with rectal cancer that had circumferential resection margin of <1 mm were excluded from the present study. Regional lymph node metastases were present in 155 cases (38.2 percent), which constituted the study group.

None of these 155 cases received any kind of neoadjuvant therapy, including preoperative radiation therapy. At that time, in our surgical division, there was no established protocol of adjuvant chemotherapy, and therefore, it was not routinely recommended to patients even if they were node-positive. Treatment with oral 5-fluorouracil alone was the most common protocol adopted.

The information extracted from the clinical record of each patient was as follows: date of birth; date of surgery; date of last follow-up; date of first recurrence; location of the primary tumor (cecum, ascending, transverse, descending, sigmoid colon, or rectum); site of first recurrence (liver, lung, local, bone, peritoneal, or other); administration of adjuvant therapy; and preoperative serum carcinoembryonic antigen (CEA) level.

Follow-up was performed in accordance with the division’s protocol and was based on periodic evaluations of the patients, concerning the serum CEA levels (every 3 months), abdominal/pelvic CT scan or ultrasonography (every 6 months), and chest x-ray (every 12 months). The follow-up findings were confirmed for all patients on September 30, 2004.

Pathologic Review

As a part of our routine practice, specimens were delivered to the pathology division in the fresh state and, immediately after the operation, we grossly examined the specimens and dissected the lymph nodes by careful palpation.

The microscopic slides of all specimens and the surgical pathology reports were reviewed for the following information: tumor depth of invasion; tumor size (maximum diameter of the primary lesion); predominant histologic tumor grade (well, moderately, poorly differentiated, and mucinous); grade of lymphatic or venous invasion (focal or extensive); number of lymph nodes recovered; number of lymph nodes that contained metastases. If a tumor nodule of ≥3 mm in diameter was identified in the mesorectum or mesocolon and there was no histologic evidence of residual nodal tissue, it was classified as a regional lymph node metastasis, in accordance with the rules of the fifth edition of TNM staging.11

In addition to the routine pathologic examination, all of the positive lymph nodes were retrospectively reviewed and evaluated for the following findings: size of intranodal tumor (sum of tumor diameter in lymph nodes per case); pattern of tumor growth (infiltrative or expansive); and presence of capsular invasion (defined as growth of adenocarcinoma through or beyond the lymph node capsule and evaluated as present or absent; Fig. 1). We defined the size of intranodal tumor as the sum of maximum dimensions of the tumor nodules measured on the slide. If infiltrative tumor growth or capsular invasion was observed in at least one node, we categorized the case as “infiltrative” and, therefore, “with capsular invasion.” After the appearance of the node was characterized, interobserver agreement between the two independent observers (HY and YS, with 16 and 28 years, respectively, of experience in surgery) was measured.

Figure 1
figure 1

Microscopic findings of metastasized lymph nodes. A. Tumor shows expansive growth and does not extend beyond the capsule (×40). B. Tumor shows infiltrative growth and does not extend beyond the capsule (×40). C. Tumor shows infiltrative growth and extends beyond the capsule (×40→×100).

Full size image

Statistical Analysis

Recurrence-free period was defined as the interval from surgery to the time recurrent disease was diagnosed. For the evaluation of cancer-specific survival, only cancer-related deaths were considered events and data on the patients who died from other causes or who were still alive at the end of our study were censored. Recurrence-free and cancer-specific survival rates were calculated by the Kaplan-Meier method. The log-rank test was used to evaluate differences between survival curves. Univariate analysis was conducted with regard to the following variables: location, size, depth of invasion, tumor grade, and lymphovascular invasion of main tumors; number of positive nodes; and patterns of lymph node involvement. Multivariate analysis was performed for variables that were statistically significant in univariate analysis using the Cox proportional hazards model and stepwise logistic regression. Associations were analyzed using the chi-squared test or Fisher’s exact probability test (two-sided) for categoric variables and Mann-Whitney U test was used to compare means. Interobserver agreement for the microscopic assessment of intranodal tumor growth pattern or capsular invasion was assessed by use of the κ statistic. P<0.05 was considered a statistically significant association. Statistical analysis was performed with SPSS® software, version 12.0 (SPSS Japan Inc., Tokyo, Japan).

Results

Microscopic Findings

Of the 155 Stage III patients, the mean and median number of lymph nodes recovered per case was 29 and 27, respectively (range, 1–107). The mean and median number of positive lymph nodes per case was 3.1 and 2, respectively (range, 1–23). Fifty-six patients had one positive node, 45 patients had two, 12 patients had three, and 42 patients had four or more. A total of 480 positive nodes were examined.

The size of intranodal tumor per case ranged from 0.3 to 100 (median, 8) mm. Infiltrative growth pattern of intranodal tumor was observed in 86 cases (55.5 percent), whereas expansive growth pattern was seen in 69 cases (44.5 percent). Lymph node capsular invasion was present in 75 cases (48.4 percent) and absent in 80 cases (51.6 percent). Tumor nodules without histologic evidence of residual lymph node were observed in 17 cases (11 percent). Because all of them had an irregular contour or accompanied lymph nodes with capsular invasion, they were categorized into the capsular invasion positive group in this study. Interobserver agreement for assessing the microscopic findings of lymph nodes was 85 percent, with a κ score of 0.693.

Patient Characteristics

Table 1 shows the clinicopathologic features of 155 patients whose positive lymph nodes were analyzed for the presence of lymph node capsular invasion. The mean size of the primary tumor was 4.9 (range, 1.3–17) cm in maximum diameter. Six tumors were T1, 10 were T2, 133 were T3, and 6 were T4.

Table 1 Clinicopathologic Characteristics of 155 Stage III Patients Analyzed for Presence of Lymph Node Capsular Invasion
Full size table

Clinical Outcomes

The overall median follow-up period was 4.9 years (range, 4 months to 8.9 years). All the patients with a follow-up period of less than one year were free of disease and died of causes other than cancer, and none of the patients died of surgical complications.

During this period, 41 cases (26.5 percent) among the 155 Stage III patients developed recurrent disease. Initial recurrence occurred in the liver in 20 cases, in the lungs in 9 cases, in the local regions (including anastomotic or pelvic recurrence) in 6 cases, in the peritoneum in 5 cases, and in other organs in 6 cases. The median time period until the recurrent disease became clinically apparent was 12 months (range, 2 months to 4.1 years). Twenty-eight cancer-related deaths were observed, and 16 patients died of other causes during the period.

The median follow-up period for those patients who did not develop recurrence was 5.4 years (range, 4 months to 8.9 years). The median follow-up period for those who were still alive was 5.5 years.

Analysis of Recurrence

The features that were associated with recurrent disease in univariate analysis were as follows: higher grade of venous invasion; greater number (≥4) of positive lymph nodes; presence of lymph node capsular invasion; and infiltrative intranodal growth pattern. Figure 2 shows the Kaplan-Meier curves of recurrence-free rates for those features. Multivariate analysis using Cox proportional hazards model revealed lymph node capsular invasion to be the only significant prognostic factor associated with recurrence (hazard ratio, 3.812; 95 percent confidence interval, 1.638–8.871; Table 2). The five-year recurrence-free rates of patients with and without lymph node capsular invasion were 56.1 and 88 percent, respectively.

Figure 2
figure 2

Recurrence-free rates according to the factors that show a statistically significant difference. A. Grade of venous invasion of primary tumor. B. Number of positive nodes. C. Lymph node capsular invasion. D. Intranodal tumor growth pattern.

Full size image
Table 2 Multivariate Analysis of Features Associated with Recurrent Disease
Full size table

Among the 75 patients with lymph node capsular invasion, 32 patients developed recurrence, 26 of whom (81.3 percent) recurred in distant organs. Of the 80 patients without lymph node capsular invasion, on the other hand, only 9 recurred and 8 of them (88.9 percent) developed distant metastasis. No significant association was observed between lymph node capsular invasion and site of initial recurrence (distant or local).

Analysis of Survival

Univariate analysis for cancer-specific survival showed statistical association with the following features: higher grade of lymphatic and venous invasion; greater number (≥4) of positive lymph nodes; presence of lymph node capsular invasion; and infiltrative intranodal growth pattern. Among those, the following features remained as the significant prognostic factors associated with survival in multivariate analysis: lymph node capsular invasion; number of positive lymph nodes; and venous invasion (hazard ratio and 95 percent confidence interval are shown in Table 3). Kaplan-Meier curves of cancer-specific survival for those features are presented in Figure 3. The five-year survival rates of patients with and without lymph node capsular invasion were 70.4 and 90.5 percent, respectively.

Table 3 Multivariate Analysis of Features Associated with Cancer-specific Survival
Full size table
Figure 3
figure 3

Cancer-specific survival rates according to the factors that show a statistically significant difference. A. Grade of venous invasion of primary tumor. B. Number of positive nodes. C. Lymph node capsular invasion.

Full size image

Factors Associated with Lymph Node Capsular Invasion

The features that were associated with the presence of lymph node capsular invasion were as follows: depth of primary tumor; lymphatic and venous invasion; number of lymph node metastases; intranodal tumor growth pattern; and treatment with adjuvant chemotherapy (Table 1).

Discussion

In the current study, we identified a novel histopathologic finding that can serve as a significant prognostic determinant based on the biologic behavior of Stage III tumors. Our results show that lymph node capsular invasion, determined by routine histology, can be a more critical variable in predicting prognosis than other variables, such as T status, lymphovascular invasion, and number of positive nodes. They also indicate that patients with capsular invasion should be regarded as a high-risk group for recurrence and, therefore, should be the candidates for adjuvant chemotherapy along with intensive follow-up.

There are many investigators who have reported a close relationship between survival and number of involved lymph nodes.1–4,12,13 In the latest International Union Against Cancer/American Joint Committee on Cancer (UICC/AJCC) TNM staging,5,6 patients with histologically involved nodes have been subdivided into two groups: those with one to three involved nodes (N1), and those with four or more involved nodes (N2). Wong et al.14 reported that the number of involved nodes, rather than the volume of metastatic involvement of the regional lymph nodes, determines outcome and the presence or absence of nodal metastases represents the most important prognostic factor in colorectal cancer; our results are congruent with their report.

Although conventional prognostic factors, such as vascular invasion and number of positive nodes, proved to be associated with survival in both univariate and multivariate analyses in the present study, they did not have prognostic significance in multivariate analysis with regard to recurrence and lymph node capsular invasion remained as the only significant prognostic factor. These findings indicate that lymph node capsular invasion is a potent prognostic factor associated with recurrence.

The reason why capsular invasion is a significant prognostic factor for distant metastasis and local recurrence is unknown. Given that capsular invasion is associated with depth and lymphovascular invasion of primary tumor and number of positive nodes, it may represent the ability of colorectal tumor cells to disseminate to distant sites, which might be a process distinct from the ability to spread to local lymph nodes. Otherwise, tumor cells penetrate into vascular structures in the lymph node, leading to hematogenous dissemination, or drain into the efferent lymphatics of lymph nodes and eventually into the systemic circulation. In fact, lymph node capsular invasion was identified in all Stage IV patients who had their main tumor resected during the same period (data not shown).

Extracapsular extension of lymph node metastasis has been suggested to be of prognostic value in a variety of solid tumors, such as breast,15,16 head-and-neck,17,18 thyroid,19 prostate,20 skin,21 gynecologic,22 and gastrointestinal cancers.10,23,24 Local control, disease-free survival, and overall survival also may be affected in colorectal cancer.10,23,25 Komuta et al.10 have identified extracapsular invasion of the metastatic nodes as a useful and unfavorable prognostic factor in 84 colorectal cancer patients in terms of recurrence and survival, and our findings support their results. They have claimed, however, that no significant association was found between extracapsular invasion and other clinicopathologic features; moreover, no statistically significant difference was observed in recurrence or survival between N1 and N2 groups according to TNM classification. These observations may be attributable to a smaller number of patients in their study. Heide et al.23 have correlated extracapsular extension of nodal metastasis with poor local control in 96 patients with node-positive rectal cancer. In the present work, a larger number of patients with Dukes C colorectal cancer was examined and, to the best of our knowledge, no studies have previously evaluated the prognostic significance of lymph node capsular invasion using multivariate analysis.

Adequate retrieval and assessment of lymph nodes in colorectal cancer resection specimens is an important component of staging.26–29 The number of nodes examined in a specimen varies substantially from case to case, which is dependent on the anatomic variability or the different operative procedures, as well as on the extent or diligence of pathologic examination.13,26–31 Recent National Cancer Institute guidelines suggest that a minimum of 12 lymph nodes is required to accurately determine whether a patient has positive lymph nodes,32 and it is adopted by the TNM Classification of AJCC and UICC. The fat-clearance technique has been shown to increase the accuracy of lymph node harvest in surgical specimens compared with manual dissection.33,34 More recently, immunohistochemical or genetic techniques have been proposed to identify small clusters of cancer cells (“micrometastases”) within lymph nodes, and ultrastaging, by serial sectioning combined with immunohistochemical techniques, has improved detection of nodal micrometastasis.7–9,35–38 These techniques, however, are time-consuming and labor-intensive for routine clinical use. Moreover, consensus has not yet been achieved regarding the prognostic significance of micrometastasis or isolated tumor cells in lymph nodes identified by those special techniques.39–42 Originally, the Dukes classification and records developed during many decades at St. Mark’s Hospital were based on a single section through each node and so is the latest TNM classification.

In this study, lymph nodes were retrieved manually without fat clearance, a single section was obtained through each node, monoclonal antibodies were not used routinely, and tumor deposits without residual nodal tissue were counted as positive nodes in accordance with the fifth edition of UICC/AJCC staging guidelines. The mean time required to complete the retrieval was approximately 60 minutes. As a result, the median number of lymph nodes harvested in this study was 27, which is far more than those found using manual dissection in previous studies.26,27,29,43 This might be attributable to the extent of lymph node dissection and to our diligence in identifying nodes. Overall node-positive rate in our study population was 38.2 percent, which is consistent with that observed in other studies.27,29

Lymph node capsular invasion may represent various patterns of cancer invasion: afferent or efferent vessel invasion, hilar tissue invasion, or intranodal tumor extending beyond the capsule. It often is difficult to distinguish between them, unless tumor cells show focal invasion. Lymph nodes have unidirectional lymphatic flow, with most of the supply coming from the hilar vessels.15,44,45 Lymph flow enters the lymph node from the multiple afferent lymphatic vessels, into the subcapsular sinus, moves down the lymphatic sinuses, and leaves through the efferent lymphatic vessel in the hilum. Colorectal cancer cells may involve the afferent vessels of the lymph node, leading to direct extension from the afferent vessels to the subcapsular sinus. In addition, the tumor cells may involve the lymph node in a variety of patterns, intrasinusoidal or parenchymal, and also may involve the hilar tissue and efferent vessels. They might sometimes completely replace the lymph node.

Because histologic types of metastatic deposits in lymph nodes often varied in a patient with Stage III tumor, we found it difficult to determine the predominant type in each case. When examined in each positive node, histologic types of metastatic deposits were not associated with capsular invasion (data not shown).

Microsatellite instability (MSI) status has been known as a prognostic factor in colorectal cancer46,47 and MSI-H status has been reported to correlate with histologic parameters, such as mucinous or medullary histology and Crohn’s-like lymphoid infiltrates, and with right sidedness of the tumor.48,49 Although we did not assess the MSI status of the tumors by use of immunohistochemical or molecular method in the present study, predominant mucinous histology of the main tumor was identified in 11 patients and medullary histology in 2 patients; 41 tumors were located in the right side. These features did not correlate with favorable outcome in our study; moreover, no significant association was observed between these clinicopathologic features and lymph node capsular invasion.

Even the definition of an involved lymph node remains controversial. According to the fifth edition of the UICC/AJCC staging guidelines, a tumor nodule ≥3 mm in diameter in the pericolic or perirectal fat without histologic evidence of residual lymph node is classified as regional lymph node metastasis.11 However, it has been suggested that those nodules often are not derived from destroyed metastatic nodes but are intravascular, perivascular, or perineural extensions of the primary tumor, and represent a feature of poor prognosis, independently of their number and size.25,50,51 Thus, the sixth edition of the UICC/AJCC staging guidelines classified those nodules into two entities: a nodule with a smooth contour is classified in the pN category as a regional lymph node metastasis, whereas a nodule with an irregular contour is classified in the T category coded as V1 or V2 for venous invasion.5,6 In the present study, the authors identified tumor nodules without histologic evidence of residual node in 17 cases (11 percent) from Stage III patients. We, therefore, applied the two different approaches in the analyses according to both guidelines. Fortunately, the results of the two analyses were similar.

A possible limitation of the study is that our study population was relatively small, although it was larger than those of the previous studies by Komuta et al.10 and Heide et al.,23 which might have diluted the significance of conventional prognostic factors leading to accentuate that of lymph node capsular invasion. Furthermore, the favorable outcome of the entire population of Stage III patients in our study (five-year, recurrence-free and cancer-specific survival rates were 72.4 and 73.8 percent, respectively) may have further augmented the effect. It would be of value to further investigate the prognostic significance of lymph node capsular invasion.

In the current study, one experienced surgeon (YS) participated in both the preoperative diagnosis and the operative procedure of all patients. This probably eliminates the variable of surgeon-related prognostic factors noted in previously published multicenter studies. The surgical procedure consisted of standard radical resection in most cases and of limited resection in a few cases.

Patients with TNM Stage III colorectal cancer are a heterogeneous group. Approximately half, without lymph node capsular invasion, have good prognosis similar to that of patients with Stage II disease, whereas the other half, with capsular invasion, have a poor prognosis, indicating the need for adjuvant therapy and intensive follow-up. Surgery and adjuvant chemotherapy are considered the standard treatments for Stage III disease (irrelevant to N-number), which include 5-fluorouracil and leucovorin.52 New drugs, such as irinotecan or oxaliplatin, are coming to be incorporated not only for the advanced stages but also in an adjuvant setting.53 Considering the adverse effects and cost-effectiveness of these drugs as well as the wide spectrum of Stage III disease, better assessment of prognosis in patients with Stage III colorectal cancer could allow individually adapted treatment decisions with the potential of toxicity sparing and therapy intensification.