A Nomogram for Predicting Lymph Node Metastasis in Submucosal Colorectal Cancer

Cancer is a leading cause of death worldwide, and colorectal cancer (CRC) is the primary cause of death among women and the third leading cause among men in Japan,¹ as well as in most Western countries.² According to the Japanese Society for Cancer of the Colon and Rectum guidelines for the treatment of colorectal cancer (JSCCR Guidelines), the treatment plan should be decided based on the tumor depth and the presence of lymph node (LN) and/or distant metastases, and a curative surgical resection of primary CRC with regional LNs is generally performed in stage I to III CRC.³ In contrast, endoscopic resection can complete the treatment of intramucosal carcinoma because there is no LN metastasis. However, in submucosal (SM) carcinoma, the probability of LN metastasis is approximately 10%,⁴ and surgical resection with LN dissection is generally recommended. In the guidelines, the additional surgical resection is determined by the pathologic findings of the primary tumor resected by endoscopic treatment.³

The additional surgical resection is required because retrospective analyses have determined that the following factors significantly relate to LN metastasis and tumor recurrence^5,6: (1) depth of submucosal invasion greater than 1000 μm; (2) positive lymphovascular invasion; (3) positive endoscopic vertical margin; (4) poorly differentiated adenocarcinoma, signet-ring cell carcinoma, or mucinous carcinoma; and (5) grade 2/3 budding at the site of deepest invasion.

The rates of recurrence and 5-year disease-free survival were 0.8% and 95%, respectively, in the patients who had none of the risk factors listed above after endoscopic resection.⁷ However, these rates were 6.6% and 89.3%, respectively, in the patients who had one or more of these risk factors and were treated by endoscopic resection alone. When these risk factors were combined, the probability of LN metastasis ranged from 7.4% to 46.9%,⁸ and thus the importance of additional surgery is different in each patient.

Therefore, we examined the risk factors retrospectively and used them to develop a new prediction model. This is the first report of the prediction model for lymph node metastasis in SM CRC.

Materials and Methods

Patients and datasets

We retrospectively analyzed 509 consecutive patients with SM CRC that were surgically treated at the Osaka Medical Center for Cancer and Cardiovascular Diseases, Japan, from 1984 to 2012. We excluded 113 patients who lacked any of the required pathologic findings (Fig. 1). In a group of 293 patients who underwent surgery from 1984 to 2008, we assessed the clinicopathologic characteristics [sex, age, preoperative carcinoembryonic antigen (CEA), primary CRC location, macroscopic tumor type, tumor size, histologic grade, SM invasion depth, lymphatic invasion, vascular invasion, and LN metastasis] and recurrence-free survival. These patients were included in a training set (TS) aimed at developing a prediction model for LN metastasis using a logistic regression model. A total of 103 patients treated from 2009 to 2012 were included in a validation set (VS). The prediction model was validated in this independent group. This retrospective study was approved by the Osaka Medical Center for Cancer and Cardiovascular Diseases Ethics Committee, and written informed consent was obtained from all patients.

View Figure

• Download as Powerpoint
• Download Figure

Pathologic examination

All of the resected specimens were fixed in 10% buffered formalin, processed through graded ethanol solutions, and embedded in paraffin blocks. They were examined histologically using hematoxylin and eosin staining. The degree of histologic differentiation, SM invasion depth, lymphatic invasion, and venous invasion were assessed. The SM invasion depth was measured according to the Japanese Classification of Colorectal Carcinoma (8th edition).⁹ When the muscularis mucosae could be identified, the vertical distance from the muscularis mucosae to the deepest level of invasion represented the SM invasion depth (Fig. 2A). When the muscularis mucosae could not be identified, the vertical distance from the superficial aspect of the tumor to the deepest level of invasion was determined (Fig. 2B). Fifty-five patients had only the relative (but not numeric) SM invasion depth data evaluated as sm1, sm2, and sm3, according to the previous Japanese Classification of Colorectal Carcinoma (6th edition).¹⁰ We translated these classifications from our preliminary examination (data not shown) as follows: 1000 μm for sm1, 2000 μm for sm2, and 4000 μm for sm3.

View Figure

• Download as Powerpoint
• Download Figure

Results

The characteristics of all 396 CRC patients, including those of 293 TS patients and 103 VS patients, are shown in Table 1. LN metastasis was evident in 37 (12.6%) and 7 patients (6.5%), respectively, and the respective median tumor size was 16 (range, 5–90 mm) and 17 mm (range, 2–91 mm). Positive lymphatic invasion was observed in 89 (30.3%) and 21 patients (19.4%) and positive vascular invasion was observed in 64 (21.8%) and 24 patients (22.2%), respectively.

Table 1  Clinicopathologic factors of all patients

View Fullscreen

Univariate analysis of the clinicopathologic factors of LN metastasis in TS is shown in Table 2. It revealed the influence of low histologic grade (muc, por, sig; P < 0.001), positive lymphatic invasion (P < 0.001), positive vascular invasion (P = 0.036), and tumor SM invasion depth (P = 0.098). Tumor size and tumor type were not found to be significant. In the TS, the rate of 5-year disease-free survival was 96%, and the cancer-specific survival was 97% (Fig. 3). A nomogram for predicting LN metastasis in SM CRC was constructed using the 4 factors, histologic grade, SM invasion depth, lymphatic invasion, and vascular invasion, through a logistic regression model (Fig. 4). The nomogram can assign the probability of LN metastasis by adding up the scores identified on the points scale for each factor. The total score projected to the bottom scale indicates the probability of LN metastasis. The area under the curve (AUC) in the TS was 0.717. The prediction model was validated by an external dataset in the VS, and the AUC was 0.920.

Table 2  Univariate analyses of lymph node metastasis in TS

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

We performed a multivariate analysis of LN metastasis in the TS, as shown in Table 3. Low histologic grade (muc, por, sig; P = 0.022), SM invasion depth (P = 0.033), and positive lymphatic invasion (P = 0.002) were significant risk factors. However, vascular invasion was not a significant factor (P = 0.525). Another nomogram for predicting LN metastasis in SM CRC was constructed using the 3 significant factors, namely, histologic grade, SM invasion depth, and lymphatic invasion, through a logistic regression model (Fig. 5). The AUC in the TS was 0.720 and 0.920 in the VS.

Table 3  Multivariate analyses of lymph node metastasis in TS

View Fullscreen

Table 4  Clinicopathologic factors of LN metastasis in polypoid type and nonpolypoid type (others)

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

Discussion

Surgical resection is a standard treatment for CRC. Five-year disease-free survival and 5-year overall survival are 95% to 97% and 97% to 99%, respectively, in SM CRC after surgical resection.⁷ This prognosis is good; however, intraoperative and perioperative complications may occur with surgical treatment,^11,12 and severe complications such as anastomotic leakage may be a major cause of death.¹³ The mortality of the colorectal cancer operation has been reported to be approximately 5%.¹¹ Surgical resection including regional LNs is the standard treatment for colorectal cancer; however, the procedure can cause these complications or other functional problems. For example, local excision followed by chemoradiotherapy is preferred for low rectal cancer instead of surgery with total mesorectal excision, and this treatment results in good anal function.¹⁴ Recent reports have suggested that older patients with certain risks could be overtreated, with the possibility of subsequent excess morbidity and mortality; the survival gain is smaller in older compared with younger patients.¹⁵ Therefore, the need for surgical treatment must be evaluated in each individual case, and the tool for predicting LN metastasis in SM CRC is useful.

Recently, several predictive scores have been reported for predicting cancer prognosis, surgical complications, or genetic mutational status.^16–20 These scores can predict each individual patient's prognosis or complications.

In previous reports, the probability of LN metastasis in SM CRC varied across a combination of risk factors. For example, the probability was 9.4% in patients with an SM depth >1000 μm and negative vascular invasion and was 46.9% in patients with an SM depth >1000 μm and low histologic grade.⁸ However, it is recommended in the current Japanese guidelines that both patient groups undergo additional surgical resection.³ In our study, we evaluated the risk factors for LN metastasis in SM CRC and developed a new nomogram for predicting LN metastasis in individual SM CRC cases. Recently, there has been an increasing number of surgeries for colorectal cancer with complications in elderly patients. Therefore, our nomogram may help to decide whether an additional surgical treatment is necessary after endoscopic resection.

The invasion depth is an important risk factor in this study, but the polypoid type (0-I) is difficult to evaluate. If the muscularis mucosae cannot be identified, the measure of the invasion depth will be wide; the tumor volume has greater effect in the polypoid type than the nonpolypoid type, as mentioned above. We also reevaluated the invasion depth according to the type of tumor, including polypoid type (0-I) and nonpolypoid type (0-II and 0-III).⁹ Univariate analysis of the clinicopathologic factors of LN metastasis in the polypoid type is shown in Supplementary Table 4A. There were influences of positive lymphatic invasion (P = 0.004) and low histologic grade (muc, por, sig; P = 0.091). Univariate analysis of the clinicopathologic factors involved in LN metastasis in the nonpolypoid type is shown in Supplementary Table 4B. There were influences of low histologic grade (muc, por, sig; P < 0.001), SM invasion depth (P = 0.044), positive lymphatic invasion (P < 0.001), positive vascular invasion (P < 0.001), and tumor size (P = 0.070). The SM invasion depth and tumor size were considered important factors to predict LN metastases in the nonpolypoid type; however, they were not significant factors in the polypoid type. In addition to the measure of the invasion depth in the polypoid type, the classification of the tumor types is various, because the type often contains some elements, such as “0-I + IIc.” The type of tumor, such as polypoid or nonpolypoid type, should be evaluated to predict LN metastases relating to SM invasion depth with a large number of subjects.

This study has several limitations. The budding grade⁹ was not evaluated in this study because there were so few data in the TS (collected from 1984 to 2008); the budding grade was added to the guidelines in 2009. SM invasion depth was evaluated as a relative invasion depth in 55 patients of the TS, and they were translated into numerical values. In addition, there were a small number of subjects in this study compared with previous studies,^16,18 and this may be related to the insignificant factor of SM invasion depth in univariate analysis and venous invasion in multivariate analysis. A multiple institutional study that enrolls many patients should be conducted. However, our new prediction model of LN metastasis in SM CRC is useful as a clinical tool because this model predicts the probability for the individual patient, resulting in more personalized medical care.

Conclusions

A nomogram predicting LN metastasis was successfully developed through the integration of 4 pathologic factors: invasion depth, lymphatic invasion, vascular invasion, and histologic grade. This tool could help physicians and patients decide on the additional surgical treatments required after endoscopic resection.