Novel Prognostic Role of Serum CA19-9 Identied in Colorectal Terminal Tumor Site and Tubular Adenocarcinoma

It is still illusive that the origination and contribution of different primary tumor sites in Colorectal Cancer (CRC). A total 1039 consecutive CRC patient’s proles were collected and investigated the prediction role of multiple biomarkers during survival period. The correlation and survival analysis were applied to explore the prognostic value of clinical traits (primary tumor sites, and tumor subtypes, the incidence rate) and multiple biomarkers. The sigmoid and rectum (73%) demonstrated the highest incidence rate among other primary tumor sites (27%). Tubular adenocarcinoma, mainly rise from the terminal bowel tumor sites, has signicantly worse clinical outcome, lower 5-year survival rate and shorter overall survival time. Serum level of CA19-9 was signicantly positive correlated with sigmoid and rectum originated-CRC (Cor =0.88, p=2.2e-16). Higher serum level of CA19-9 (>37 U/mL) was signicantly associated with terminal tumor sites, tubular adenocarcinoma and tumor with non-inltrating. It could be applied for the precise diagnosis of CRC and terminal bowel cancer, especially in the tubular adenocarcinoma. Therefore, terminal tumor sites (sigmoid and rectum) demonstrated potential prognostic value in early screening for CRC patients. CA19-9 could be a promising biomarker for the diagnosis of terminal colorectal cancer, especially for the Tubular adenocarcinomas originated from sigmoid and rectum, and improve the clinical outcome.


Introduction
Colorectal cancer (CRC) is a global health threaten for all populations in the world. The incidence rate of CRC has been fast growing, leading it to be commonly death-caused cancer in the world 1,2 . Especially in the past ten years, the morbidity and mortality of CRC increased to the top ten among many types of cancers in China 3 . CRC is treated as complex diseases: over 90% of colorectal carcinomas are identi ed as adenocarcinomas that originate from epithelial cells of the colorectal mucosa. The typically intestinal subtypes of adenocarcinoma contain tubular adenocarcinoma, mucinous adenocarcinoma, adenocarcinoma with necrosis, and mixed adenocarcinoma 4 . Tubular adenocarcinoma, with high cellular density and quick cell proliferation, was the dominant pathologic type of CRC 5 .
Beyond the cellular characterizations, these clinical traits were established as potential indicators to predict the clinical outcome, including age, gender and tumor stage, etc. The incidence rate and gender preference were varied with huge difference among the colorectal carcinomas, demonstrated by investigations in Western and Eastern populations 3,6,7 . The majority (80%) of patients with colon cancer was diagnosed over 60-years-age old and the incidence rate of male's patients was higher in than the female's 6 . In addition, the primary tumor sites of CRC indicated the different contribution in overall survival of patients. CRC patients who are with tumor metastasized in left-side (distal, splenic exure to rectosigmoid) have better survival rate than those whose tumor metastasized in right-side (proximal, cecum to transverse). Similarly, patients with metastasized rectal cancer showed better clinical outcome than those with metastasized colon cancer 8 . However, the exactly prognostic value of primary tumor sites has not been well established to predict the clinical outcome of CRC in previously studies.
Except clinical traits, serum biomarkers have been broadly used in prediction of clinical outcomes.
Multiple biomarkers, including carcinoembryonic antigen (CEA), C-reactive protein (CRP) and carbohydrate antigen 19 − 9 (CA19-9), have been successfully translated into evaluation and management of the patients with CRC in clinical practice 9 . Among them, CEA has been regarded as a promising prognostic biomarker for tumor progress and metastasis in CRC, and integrated into patients monitoring in clinical application 10 . Serum level of CRP was adopted as potential biomarker to evaluate the potential risk of colorectal cancer. However, it is still not clear that the positive correlation between elevated serum CRP level and CRC 11 . Serum concentration of CA19-9 is a well-known tumor biomarker to screen and detect the carcinomas in the digestive tract, and about 18% of CRC cases were associated with higher serum CA19-9 levels. Although CA19-9 has been applied as a tumor biomarker in colon carcinoma over 40 years 12 , especially worked as prognostic indicator of advance stage 13 and metastatic colorectal cancer 14 , it was still controversial that the speci city and sensitivity of CA19-9 applied in the CRC clinical screening 15,16 .
In this study, we retrieved to investigate over 1000 clinical pro les of the hospitalized CRC patients in Zhenjiang, a local area in Jiangsu province of China. Multiple clinical traits and biomarkers levels were associated with these CRC patient's pro les. The serum level of CA19-9 was signi cantly positive associated with a poor prognosis and high recurrence rate, especially the CRC patients that primary tumor located at terminal sites (Rectum and sigmoid). These results suggest that CA19-9 work as potential precise prognostic indicator for tubular adenocarcinoma and tumor non-in ltration in terminal colorectal cancer.

Results
Clinical characterization of cases These original cases were selected from diagnosed CRC patients, and these cases with missing the key parameters or clinical markers were excluded for next step general and speci c level analysis. The process and characteristic analyses were demonstrated by ow chart ( Figure.1). Characterization of these cases was summarized, including corresponding clinical features, 5-year survival rate and tumor distribution of primary site (Table.1). First of all, the quali ed the patients were divided into four agegroups (Table.1), including below 50-years-old group (124 cases, 12.3%), 50-60-years-old group (273 cases, 27.1%), 61~70-years-old group (326 cases, 32.4%) and over 70-years-old group (282 cases, 28.0%).
The majority of cases (882 cases, 87.7%, p<0.001) were signi cantly distributed in the range of over 50years-old (Figure.S1A). The 5-year survival rate of over 60-year-old patients (median 55.5 months, 44%) was signi cantly lower than below 60-year-old patients [median 62 months, 52%, log rank p=0.0004] (Figure.S1B). It suggests that population of over 60-year older patients have higher risk to develop into CRC and lower survival rate. There is not signi cantly differences in the ve-year survival rates between gender groups.
Combined gender bias with primary tumor sites analysis, male patients (58%) had signi cantly higher incident rate than female patients (42%), especially in the terminal tumor sites. The distribution ratio of male patients (61%, 86 cases; 57%, 341 cases) is higher than female patients (39%, 54 cases; 43%, 256 cases) in sigmoid cancer and rectal cancer, respectively ( Figure 3C&D). Combined age with primary tumor sites analysis, 74.8% of 50-70-years-old cases were distributed in terminal tumor sites, 25.2% of patients allocated in others tumor sites. The diagnosis rate of early stage terminal colorectal cancer (41.1%) was signi cantly higher than that of other primary tumor sites (31.2%) ( Table 2), while the advance stage of terminal colorectal cancer (58.9%) was signi cantly lower than that other primary tumor sites (68.8%).
The OS and ve-year survival rates of terminal tumor sites (sigmoid and rectum, 48.4%) were similar with other primary tumor sites (48.3%).
Through tumor subtype analysis, 988 patients were identi ed with pathohistological feature and classi ed as subtype of tubular adenocarcinoma, mucinous adenocarcinoma, adenocarcinoma, adenocarcinoma with necrosis and mixed type (without tubular adenocarcinoma, contained more than two other subtypes), respectively (Figure.S1C). Among four subtypes of adenocarcinoma, the survival rate of tubular adenocarcinoma patients (median 38 months, 5%) was signi cantly lower than others histological tumors (median 60.5 months, 50%, log rank p =2e-15, Figure. (Table.S1), including leukocyte, blood platelet, neutrophil, lymphocyte, monocyte, did not show signi cantly difference between normal and tumor patients. In addition, the patients (99%) with higher level of CA19-9 (CA19-9-2) were mainly distributed in the terminal tumor sites of colorectal cancer, and only small portion allocated in other primary tumor sites (Table.1, Figure.4E). However, there was no signi cant difference in overall survival time of the two different CA19-9 levels patient groups. Combined with the tumor stage analysis, lower CA19-9 level patients (73%) were more likely to be associated with advanced stage than higher CA19-9 level patients (58%). The serum level of CA19-9 was weak positive correlated with tumor non-in ltrating state ( Figure.S4C, r=0.07, p=1.2e-6). Furthermore, the majority of patients diagnosed as tubular adenocarcinoma (81%) were associated with the higher level of CA19-9 ( Figure.4F). Combined several markers' analysis with primary tumor sites ( Figure.4D), CA19-9 may work as a promising biomarker for terminal bowel cancer and tubular adenocarcinoma of CRC.

Discussion
In this study, we have collected the rst hands clinical data of 1006 patients from local hospital to discover these risk factors that exactly contributed to the colorectal carcinogenic, pathologic process and explore potential novel treatment strategy. The contribution of patient's age and gender, tumor stage and differentiated grade, lymph node metastasis was evaluated and deeply analyzed. The majority (88%) of colorectal cancer patients were mainly distributed in the over 50-years-old group, which is with twice higher incidence rates than the below 50-years-old patients group. In addition, the 5-year survival rate of patients was signi cant decreased in the group over 60-years-old. Previously cohort studies indicated that millions of patients were diagnosed as colorectal cancer and median age dropped from 70 years to 50 years 3,18 . It suggests that over 50-year-old patient's population have higher risk to develop as CRC and would be listed as screening target during annually physical examination. Through survival analysis, the patients diagnosed with advance stage tumor, higher differentiated grade, accompanied with lymph node metastasis and higher number of lymph node metastasis were signi cantly decreased the overall survival time and 5-year survival rate ( Fig. 2A-C). It was identi ed that the correlation between overall survival time and tumor stage, lymph node metastasis and the number (> 5) of lymph node metastasis, respectively. Furthermore, the number of lymph node metastasis was negative correlated with the patients 5-year survival rate. Compared with the standard of 8th edition AJCC staging system, these results match the previously reported trend, which identi ed the prognostic value of lymph node metastasis and tumor stage in colorectal cancer 19 . In addition, the number of metastasis lymph node could be translated as a clinical measurable predictor for tumor progress and patients' 5-years survival outcome. In the same time, the prognostic value of tumor-in ltration was also evaluated for these cases. The non-tumor-in ltration cases showed signi cantly lower survival rate and shorter overall survival time than tumor-in ltration cases. Furthermore, the majority cases of tubular adenocarcinoma (53/54) were identi ed as non-tumor in ltration. It hints that tumor-in ltration lymphocytes could be as overall survival prognostic biomarker for CRC. It matches the results of previously large-scale population meta-analysis study, which identi ed the high level of tumor-in ltrating lymphocytes linked with better patient's survival rate and worked as prognosis indicator for colorectal cancer 20,21 .
The colon adenocarcinoma mainly rose from adenomatous polyps, including three histologic types of tubular, tubulovillous and villous adenomas. Tubular adenomas contribute about 85% adenomatous polyps and only 5% portion will transform as malignancy 22 . CRC was originated from different anatomic position with varied molecular genetic alternation and different pathogenic mechanism 1,23,24 . The traditional dichotomy of colon and colorectum were facing challenge and potential delayed in early diagnosis and impacted on patient's survival 25 . In this study, 90% of patient's cases were identi ed as adenocarcinoma or mixed subtype, and only small portion cases (6%) were determined as tubular adenocarcinoma with signi cantly lower 5-years survival rate and shorter overall survival time ( Figure.4A). To further explore the heterogeneity of colorectal cancer, this study investigated the distribution of multiple primary tumor sites in CRC cases population. The 73% cases were concentrated on the terminal tumor sites. The rectum and sigmoid (terminal tumor sites) were listed as the top two tumor sites with the highest incidence rate. In addition, the incidence rate of male's patients was signi cantly higher than female's patients (Figure.S2C&D). Although the early diagnosis rate of terminal colorectal cancer (41%) was signi cantly higher than other primary sites (31%), the overall survival time was no signi cant difference among primary tumor sites. Furthermore, the tubular adenocarcinoma cases were mainly distributed in terminal tumor sites ( Figure.4B). It strongly suggests that rectum and sigmoid would be listed as the rst priority screening target for colorectal health during annually physical examination and bene t for early diagnose tubular adenocarcinoma.
Multiple serum biomarkers, like CEA, CA19-9 and CRP, were identi ed as the standard for screening CRC patient in clinical practice 9,10,13,26 . The serum levels of CA19-9 and CEA played differentiated prognostic value in the CRC. The higher serum level of CA19-9 (> 200 U/ml) was reported as signi cant predictor for poor survival of colorectal cancer patients with liver metastasis 27 . The serum level of CEA worked as the best tumor biomarker for chemotherapy drug response prediction, while CA19-9 worked as one of the best prognostic indicators for advanced colorectal carcinoma 13,28 . In this study, we investigated the internal connection and potential prediction of these biomarkers for patient's clinical outcome. Through the correlation analysis, we discovered the signi cantly correlation between serum level of CA19-9 with patient's CRC tumor sites and stage, respectively, while the serum level of CEA was signi cantly positive correlated with CRC tumor stage ( Figure.4C). Furthermore, terminal tumor cases (99%) were mainly associated with abnormal higher level of these markers (CA19-9, CEA, CRP and CEA) ( Figure.4E). More interesting, abnormal higher serum level of CA19-9 was signi cantly positive correlated with terminal tumors, especially associated with tubular adenocarcinoma cases (81%, Figure.4F), and nonlymphocytes-in ltrating tumors (90%, Figure.S4C). It suggests that the serum level of CA19-9 work as more precise prognostic predictor for terminal colorectal cancer.
Carbohydrate antigen 19 − 9 (CA19-9), a modi ed Lewis blood group antigen associated with speci c malignancies, increases in patients with gastrointestinal cancers 29 . CA19-9 was utilized as sensitive biomarker to evaluate the adenocarcinomas of the colon and rectum, and advance stage colorectal cancer with metastasis 24,30−32 . However, the correlation between the local expression level of CA19-9 and colorectal terminal tumor sites (sigmoid and rectum) is still unclear. Limitations for evaluation effective of CA19-9 in clinical tumor assessment include several factors. First of all, about 5 ~ 7% individuals in population are Lewis negative, who are with fucosyltransferase defection and do not produce CA19-9 into blood 33 . Thus, their serum levels of CA19-9 keep pretty lower or undetectable even cancer recurrence 16,34 . Secondary, concentration of serum CA19-9 is affected by liver metabolism, or disturbed by environment epidemiology 24 . Thirdly, serum level of CA19-9 also frequently increased in the patients with other cancers such aspancreatic adenocarcinomas 35 , which interferes the clinical diagnostics accuracy for colorectal cancer. In this study, the original patient's data missing the information about the Lewis blood identi cation or more detail of clinical chemotherapy drug treatment. It may cause less patients samples with higher serum level of CA19-9 correlated with terminal tumor sites even associated with advance CRC. Taken these factors into future study, it will be helpful to illuminate the precise application of CA19-9 in clinical assessment for colorectal cancer.
Some limitations were existing in this study. First of all, although it is many years retrieved study, the absence of patient's pathologic characteristic data caused the exclusion of some cases and destroyed the statistics power of large samples. Secondly, lacked the somatic mutations and microsatellite instability status of primary tumor, which indicated the patterns of CRC metastasis, it blocked the deeper exploration of internal connection of CRC. Thirdly, without the clinical chemotherapy detail and multiple prognostic scores, and the serum level CA19-9 of patients before and after chemotherapy, it deterred the precise application as clinical prognostic biomarker. Lastly, missing the genetic mutation status of key oncogenes in patients, such as mutations in KRAS (exons 2-4) and BRAF (V600E), which are mainly associated with metastasis CRC and broadly used in the metastasis CRC patient's management [36][37][38][39] , will impair the prognostic effect of CA19-9. Despite these limitations, our results demonstrated novel sight of CA19-9 as precise biomarker for terminal colorectal cancer, especially for tubular adenocarcinoma. In the future studies, larger patient's size and multiple centers independently studies will conduct to verify the prognostic value of CA19-9 in colorectal tubular adenocarcinoma.

Summary
In this study, our results provided the clinical evidences that the incidence rate and overall survival outcomes were signi cantly difference in the terminal tumor sites and other tumor sites of colorectal cancer. Colorectal tubular adenocarcinoma demonstrates the signi cantly lower survival rate in terminal tumor sites (sigmoid and rectum). The serum level of CA19-9 is a promising precise diagnostic biomarker for the terminal colorectal cancer, especially in the tubular adenocarcinoma. Detection of Serum Biomarkers.
The serum biomarkers detection (CA19-9, CEA, ALB, and CRP) were performed by standard preoperative protocol (within 2 weeks) before surgery. These biomarkers were determined by the local pathological unit for all patients. The serum CA19-9 levels were measured with electrochemiluminescence immunoassay using the Roche Cobas E601 (Roche, Switzerland) immunoassay system. The serum level of CA19-9 below or equal to 37 U/mL was identi ed as normal reference value (de ned as level 1), and greater than 37 U/mL was setup as abnormal value (de ned as level 2). According to the hospital de ned normal range, the threshold values for CEA was 0 ~ 5 µg/mL, CRP was 0 ~ 8 mg/L and ALB was 40 ~ 55 g/L (all patients were associated with abnormal level with < 40 g/L).
Lymph Node Metastases Detection.
Lymph node metastases involvement was diagnosed by PET-CT in all patients. The regional number of lymph node were identi ed and calculated based on PET-CT images.
Statistical Analysis.
Statistical analysis was conducted with R software (version 3.6.1; http://www.Rproject.org) and GraphPad Prism 6 (GraphPad Software, Inc.). The Pearson correlation (r) was employed to measure a linear dependence correlation analysis between two variables of patient's physiological indicators, and demonstrated as correlation Matrix. In the analysis, CA19-9 level below 37 U/mL, other primary tumor sites, early stage, in ltration tumor and normal indicators were de ned as "1"; CA19-9 level over 37 U/mL, terminal tumor sites, advanced stage, non-in ltration tumor and abnormal were de ned as "2". The category study variable was used independent t test or one-way analysis of variance test. The overall survival probabilities were estimated by Kaplan-Meier method and compared using log-rank test. The statistical signi cance level was set as p value < 0.05. Abbreviations CA19-9: Carbohydrate antigen 19-9