Clinicopathological factors associated with synchronous distant metastasis and prognosis of stage T1 colorectal cancer patients

It is rare and understudied for patients with stage T1 colorectal cancer to have synchronous distant metastasis. This study was to determine the clinicopathological factors associated with distant metastasis and prognosis. T1 colorectal cancer patients diagnosed between 2010 and 2015 were obtained from the SEER database. Logistic regression was applied to determine risk factors related to distant metastasis. Cox-proportional hazard models were used to identify the prognostic factors for patients with distant metastasis. Among 21,321 patients identified, 359 (1.8%) had synchronous distant metastasis and 1807 (8.5%) had lymph node metastasis. Multivariate analysis revealed that younger age, positive serum CEA, larger tumor size, positive tumor deposit, perineural invasion, lymph node metastasis, histology of non-adenocarcinoma and poorer differentiation were significantly associated with the increased risk of synchronous distant metastasis. Older age, female, Black, positive CEA, positive lymph node metastasis, positive tumor deposit, larger tumor size, no chemotherapy, inadequate lymph node harvesting and no metastasectomy were correlated with worse survival in these patients with synchronous distant metastasis. Patients with metastasis to the liver displayed the highest rate of positive CEA. We conclude that T1 colorectal cancer patients with multiple risk factors need thorough examinations to exclude synchronous distant metastasis. Chemotherapy, adequate lymph node cleaning and metastasectomy are associated with improved survival for those patients with distant metastases. Positive serum CEA may be useful in predicting distant metastases in patients at stage T1.

Prognostic factors for stage T1 colorectal patients with synchronous distant metastasis. Cox-proportional hazard regression models were applied to analyze risk factors associated with survival in patients with distant metastases. Both univariate and multivariate results showed that age, gender, race, CEA status, tumor size, histology, grade, tumor deposit, number of lymph node harvested, lymph node metastasis, metastatic sites, chemotherapy and metastasectomy were significantly associated with cancer specific survival ( Table 3). The survival curves of the risk factors are presented in Fig. 2A-F and Supplemental Fig. S2A-S2F).
Univariate and multivariate survival analysis revealed that age, race, histology, tumor grade, tumor deposit, harvested lymph node number, lymph node metastases, metastatic sites, chemotherapy and metastasectomy were associated with overall survival in stage T1 colorectal patients with synchronous distant metastasis (Table 3).
Tumor histology and survival. It was surprising that mucinous adenocarcinoma was associated with improved cancer specific survival and overall survival in stage T1 colorectal cancer patients with synchronous distant metastases (Table 3 and Fig. 2D). We therefore analyzed the correlation between tumor histology and survival in all stage T1 patients and patients without synchronous distant metastases. The results indicated that adenocarcinoma was associated with better cancer specific and overall survival in all stage T1 patients or patients without synchronous distant metastases (Supplemental data Fig. S3A-S3D).
Serum CEA status and metastatic sites. Stage T1 colorectal cancer patients without distant metastases had a relatively low rate (5.7%) of positive serum CEA. The rate of positive serum CEA was significantly  (Table 4).

Discussion
T1 colorectal cancer patients with synchronous distant metastases have a poor prognosis even after the most aggressive treatment. Due to the very low incidence, their clinicopathological characteristics have not been well characterized. Using the SEER database, this population-based study identified multiple clinicopathological factors associated with the increased risk of synchronous distant metastases: diagnosis at a younger age, positive serum CEA, larger tumor size, positive tumor deposit, perineural invasion and lymph node metastasis, histology of non-adenocarcinoma and poorer differentiation. The predicting model using these risk factors had an area under the ROC of 0.879 in distinguishing patients with distant metastasis from those without distant metastasis.
Most described risk factors have been associated with increased risk of recurrence due to heightened aggressive metastases in advanced staged colorectal cancer; however, very few have been previously reported to be risk factors related to lymph node metastases and recurrence in T1 colorectal cancer patients 5,[7][8][9] . It is noted that a previous study revealed that 4 (2.1%) of 195 consecutive stage T1 colorectal cancer patients had synchronous distant metastasis 7 , a comparable incidence rate as ours. This study indicated that patients with T1 colorectal cancer who have all or most of the risk factors are very likely to have distant metastases. Through combination with molecular markers and advanced screening technology, these risk factors may be useful to distinguish T1 colorectal cancer patients with high risk of distant metastasis, from those with low-risk of distant metastases.
CEA is a valuable serum biomarker in differential diagnosis, disease monitoring and evaluation of therapeutic efficacy in colorectal cancer. This study showed that patients with positive CEA had a significantly worse survival and a higher risk of distant metastasis. Moreover, T1 colorectal cancer patients without metastasis had a low percentage of had positive serum CEA, which was significantly increased in patients with non-liver distant  www.nature.com/scientificreports/ metastases, and further significantly increased in patients with liver metastases. These results suggest that serum CEA is particularly useful to screen synchronous liver metastasis in patients with T1 colorectal cancer. A previous study reported that CEA was as effective as computed tomography (CT) imaging in assessing the response of chemotherapy in colorectal cancer patients with liver metastases 10 . CEA has been shown to play a direct role in assisting metastasis of colorectal cancer cells to the liver and therefore it is a potential target in the treatment of liver metastasis 11 . The AJCC edition 7-TNM staging defines tumor deposits as a macroscopic or microscopic tumor nest or nodule in adjacent adipose tissue, without lymph node structure 12 . However, tumor deposits are often overlooked in TNM staging, particularly when they coexist with lymph node metastasis. Many clinicians are confused about the actual prognostic impact of tumor deposits. This study indicated that less than 1% of T1 colorectal cancer patients had tumor deposits. Multivariate Cox proportional hazard analysis showed that tumor deposits were significantly associated with an increased risk of distant metastasis, and adverse survival outcomes. Multiple studies and reviews have reported that tumor deposits are associated with increased local recurrence and distant metastasis rates, and overall survival in colorectal cancer patients [13][14][15] . Due to the limit of the database, it is not possible to determine other characteristics of tumor deposits (such as their location, diameter, and shape) and their correlation with prognosis.
Primary colorectal mucinous adenocarcinoma and signet-ring cell carcinoma are two rare subtypes of colorectal cancer with a poorer prognosis than typical adenocarcinoma 16,17 . Mucinous carcinoma may respond poorly to chemotherapy than adenocarcinoma, as mucins may be an obstacle to drug delivery. This study revealed that mucinous adenocarcinoma was associated with worse survival in all T1 patients or patients without synchronous distant metastases. However, it was associated with improved cancer survival in T1 colorectal cancer patients www.nature.com/scientificreports/ with synchronous distant metastases. The underlying mechanism is still unknown. This finding needs to be corroborated in future studies. The association between mucinous adenocarcinoma and improved prognosis is inconsistent with results in colorectal cancer patients at other stages, as reported previously 16,[18][19][20][21][22] . Perineural invasion is a pathological process in which the tumor infiltrates the nerve structure and spreads along the nerve sheath 23 . The perineural invasion has been served as an important risk factor related to recurrence and a low survival rate in colorectal cancers and other cancers [24][25][26] . Perineural invasion is also being used as a novel target to block tumor progression and improve survival 23 . This study found a very low rate (0.9%) of perineural invasion in T1 colorectal cancer patients. It was significantly correlated with the increased risk of synchronous distant metastases, but not associated with the survival in these T1 colorectal cancer patients with synchronous distant metastases.
National Comprehensive Cancer Network (NCCN) guidelines recommend sampling at least 12 lymph nodes for adequate staging of colorectal cancer 27 . Adequate lymph node cleaning has been associated with better prognosis in colorectal cancer patients at other stages 28,29 , but not in others 30,31 . This study revealed that harvesting over 12 lymph nodes was associated with increased survival in stage T1 colorectal cancer patients with distant metastasis, which supports harvesting an adequate number of lymph nodes to improve the survival in these high-risk stage T1 patients.
Stage T1 colorectal cancer patients with synchronous distant metastasis are treated the same as patients with distant metastasis at other stages 32 . In the past decade, the efficacy of systematic chemotherapy, particularly with the recently developed personalized target therapy, has dramatically improved in the treatment of patients with metastatic colorectal cancer. However, the benefit of chemotherapy has not been previously studied in T1 patients with synchronous distant metastasis. In this study, 70.3% of patients with synchronous distant metastasis received chemotherapy, whereas only 7.3% of patients without distant metastasis received chemotherapy. Chemotherapy was associated with a significantly improved survival in stage T1 colorectal cancer patients with distant metastasis. Similar to this finding, a previous study using the National Cancer Database reported that adjuvant chemotherapy significantly improved survival in stage T1 colorectal patients with lymph node metastasis 33 . The survival benefit of chemotherapy supports its use for these high-risk T1 colorectal cancer patients.
Distant metastases are the main cause of death in colorectal cancer patients. Both metastatic burden and its involvement of vital organs are important prognostic factors for metastatic colorectal cancer 34 . This study indicated that patients with metastases to multiple-sites, or bone, or brain had significantly worst survival compared to those with liver or lung metastasis alone. Among all patients with synchronous distant metastasis, 56 (14.8%) patients received metastasectomy. Multivariate survival analyses using Cox proportional hazards models revealed that metastasectomy was significantly associated with increased survival in stage T1 colorectal cancer patients with synchronous distant metastasis. The metastasectomy in metastatic colorectal cancer demonstrated the survival beneficial in previous studies [34][35][36][37] . Simultaneous colorectal and minor hepatic resections are considered to be safe for most patients with synchronous liver metastases 38 .

Patients and methods
The study cohort was obtained from the SEER database using SEER*Stat 8.3.6. According to the Site Recode Classifications, patients with colon (C18.0-18.9, and C26.0) and rectal (C19.9 and C20.9) cancers were identified from the SEER database. The inclusion criteria used for this study were: (a) the first primary tumor; (b) age of 18 or older; (c) diagnosed between 2010 and 2015; (d) surgery was performed with pathological confirmation; (e) pathological stage T1 according to AJCC 7th edition. The exclusion criteria were: (a) patients received preoperative radiation (downstage T stage); (b) histology other than adenocarcinoma, mucinous adenocarcinoma and signet ring cell carcinoma; (c) patients with unknown survival status or time (Fig. 3).
The following variables were gathered: age at diagnosis, gender, marital status, race, the status of serum CEA, tumor size, histologic type, differentiation, the status of lymph node metastasis, number of lymph nodes harvested, metastatic sites, cancer specific survival and overall survival months. The histological types were categorized using the ICD-O-3 (International Classification of Disease for Oncology, 3rd edition) coding schema: conventional adenocarcinoma (8010, 8020-8022, 8140-8141, 8144-8145, 8210-8211, 8220-8221, 8230-8231, 8260-8263), mucinous adenocarcinoma (8470, 8472-8473, 8480-8481); and signet-ring cell carcinoma (8490) 39 . The tumor grades were grouped as: Grade I (well differentiated); Grade II (moderately differentiated); Grade III (poorly differentiated); Grade IV (undifferentiated or anaplastic lesions). The metastasectomy was defined by "non-primary surgical procedure to a distant site" in the record of "surgery other regional/distant (2003 +)" 40 . The survival time was calculated from the month of diagnosis to death or censored. The last follow-up month in the database was December of 2016.
This study was based on public data from the SEER database without interacting with human subjects or using personal identifying information. This research was therefore exempted from review by the Human Subjects Committee of Institutional Review Board of our hospital. Statistical analysis. The frequency and percentage were calculated for clinicopathological variables and a Chi-square test was used to compare the patients with and without distant metastasis. Kaplan-Meier curves were generated for cancer specific and overall survival and the significance was assessed using a log-rank test. Cox proportional hazards models were used in the univariate and multivariate analyses. Two-sided p values < 0.05 were considered statistically significant. The SAS software (Version 9.3, SAS Institute Inc., Cary, NC, USA) was used to analyze data.
Ethics approval and consent to participate. As the SEER dataset is publicly available and de-identified, therefore, the ethical approval was waived by the ethics committee of our hospital.