Association of Age With Tumor Grade in Patients With Digestive Tumors: A Study Based on US SEER Registry

Background: Previous studies have not demonstrated an independent association of age and tumor grade. In this study, we aimed to explore the relationship between age and tumor grade (differentiation) in patients with digestive tumors. Methods: Surveillance, Epidemiology and End Results (SEER) 18 registry database for 1973 through 2015 was retrieved for the present study. Both piecewise and non-piecewise linear regression model were utilized to examine the relationship between age and tumor grade. Results: The present study included a total of 938,145 patients with 13 types of primary malignancies of the digestive system. In non-piecewise regression analyses, older age was associated with higher tumor grades in patients with esophageal squamous cell carcinoma (P < 0.0001) and anal squamous cell carcinoma (P < 0.0001). In contrast, older age was related to lower tumor grades in patients with esophageal adenocarcinoma (P = 0.0177), gastric adenocarcinoma (P < 0.0001), pancreatic adenocarcinoma (P < 0.0001) and rectal adenocarcinoma (P < 0.0001). In piecewise regression analyses, positive associations of age and tumor grade were only observed in specic age groups in some types of tumor, e.g., anal adenocarcinoma (> 66 years), gallbladder adenocarcinoma (> 49 years), pancreatic adenocarcinoma (> 56 years), hepatocellular carcinoma (35-90 years), rectal adenocarcinoma (<59 years in White; < 52 years in Non-White) and anal squamous cell carcinoma (<51 years in White; < 58 years in Non-White). Patients with well-differentiated tumors had better long-term prognoses compared to those with poorly-differentiated tumors (all P < 0.05). Conclusion: The patterns of relationship between age and tumor grade were different in patients with different types of digestive tumor, which may be a reection of the distinct molecular subtypes of these tumors.


Introduction
Pathologic examination determines the grade (degree of differentiation) of the tumor. The grade measures how closely the tumor cells resemble the parent tissue (organ of origin). [1] In some types of solid tumors, traditionally, the pathologists have adopted a four-grade system (well-, moderate-, poorly-and un-differentiated) to classify the tumors. [2] Well-differentiated tumor cells closely resemble the tissue from the organ of origin. Poorly-differentiated and un-differentiated tumor cells are disorganized and abnormal looking. They bear little (poorly-differentiated) or no (un-differentiated) resemblance to the tissue from the organ of origin. These similarities/differences are based on cytology, nuclear (or nucleolar) features, pattern (architecture), or a combination of these elements.
Tumor grade represents a gestalt of all subtle changes in molecular level, re ecting proliferation and aggressiveness. [3][4][5][6][7][8] The distribution of tumor grade were different between younger and older patients in some types of tumor. [9][10][11][12] However, previous studies did not illustrate the independent association between age and tumor grade. In the present study, we aimed to explore the relationship between age and tumor grade in 13 types of solid tumors of the digestive system.

Patient Selection
Surveillance, Epidemiology and End Results (SEER) 18 registry database for 1973 through 2015 was retrieved for the present study. The SEER program collects data regarding cancer incidence, patient demographics, tumor parameters, patient treatment and survival from 18 population-based cancer registries, covering approximately 28% of the US population (seer.cancer.gov/about/overview.html). Primary cancer site and histology were coded based on the third edition of the International Classi cation of Diseases for Oncology (ICD-O-3) in SEER 18. We identi ed 938,145 patients with 13 types of primary malignancies of the digestive system according to ICD-O-3 site code and histologic code from the SEER database (Supplementary Table 1). The 13 types of tumors included esophageal adenocarcinoma, esophageal squamous cell carcinoma, gastric adenocarcinoma, hepatocellular carcinoma, intrahepatic cholangiocarcinoma, extrahepatic cholangiocarcinoma, carcinoma of the gallbladder, pancreatic adenocarcinoma, small bowel adenocarcinoma, colon adenocarcinoma, rectal adenocarcinoma, anal adenocarcinoma and anal squamous cell carcinoma. We excluded patients without histologic diagnosis and patients without complete survival data and follow-up information.
We extracted demographic and clinicopathological data, including gender, age, race and tumor grade (differentiation) from SEER database. Race was divided into white, black and others. Tumor grade was divided into grade I (well differentiated), grade II (moderately differentiated), grade III (poorly differentiated) and grade IV (undifferentiated).

Statistical analysis
Multivariable linear regression analyses were used to examine the association between age and tumor grade. In all regression models, multivariable adjustments were made by including age, gender and race. Then we explored the relationship between age and tumor grade by the smoothing plot, with an adjustment for potential confounders.
We further applied a piecewise linear regression model to examine the threshold effect of age on tumor grade according to the smoothing plot. Model 1 and model 2 were the non-piecewise and piecewise linear regression models, respectively. To evaluate whether there was interaction with gender, race and tumor grade on the one hand and age on the other hand, we evaluated the interaction terms ( e.g. age x gender) in the multivariable models. In case of p < 0.05, strati ed analyses were performed. In this study, patients with colonic adenocarcinoma, rectal adenocarcinoma, hepatocellular carcinoma and anal squamous cell carcinoma were performed subgroup analyses based on sex and race (interaction P values < 0.05). The survival curves were determined by the Kaplan-Meier method and compared by the log-rank test. In survival analyses, grade 1 and 2 were grouped into welldifferentiated and grade 3 and 4 were grouped into poorly-differentiated. All analyses were performed by R (http://www.R-project.org) and EmpowerStats software (www.empowerstats.com, X&Y solutions, Inc. Boston MA).

Patient characteristics
The present study included a total of 938,145 patients with 13 types of primary malignancies of the digestive system. The clinicopathological features of the 13 types of tumor were shown in Table 1. Age were presented as mean ± standard deviance. Sex, race and tumor grade in different types of tumor were shown in numbers. In Fig. 1, age was divided into 10 groups and the number of patients were shown by sex. Association of age and tumor grade in the non-piecewise models As shown in Tables 2 and 3, the effect values (β) of non-piecewise multivariable linear regression analyses were shown in model 1. Older patients with esophageal squamous cell carcinoma (P < 0.0001) and anal squamous cell carcinoma (P < 0.0001) had higher tumor grades compared to younger patients. In addition, older age was also positively associated with higher tumor grades in the female (P < 0.0001) and White (P < 0.0001) patients with colonic adenocarcinoma. In contrast, older patients with esophageal adenocarcinoma (P = 0.0177), gastric adenocarcinoma (P < 0.0001), pancreatic adenocarcinoma (P < 0.0001) and rectal adenocarcinoma (P < 0.0001) had lower tumor grades. In strati ed analyses, older age was negatively associated with tumor grade in male (P < 0.0001) and non-White (P < 0.0001) patients with colonic adenocarcinoma, and in male patients with hepatocellular carcinoma (P < 0.0001).    Notably, for some types of tumor (e.g., colonic adenocarcinoma and esophageal adenocarcinoma), the signi cance was not changed in different age groups, while the effect size of age on tumor grade was different.

Survival analyses based on tumor grade
To explore the prognostic signi cance of tumor grade in these types of tumor, we performed survival analyses based on tumor grade. As shown in Fig. 4 (tumors in the hepatobiliary system) and Fig. 5, the results showed that patients with well-differentiated tumors had better long-term prognoses compared to those with poorlydifferentiated tumors (all P < 0.05). The results were consistent in all types of tumor.

Discussion
Previous publications have not demonstrated the association of age and tumor grade. In this large cross-sectional study, we explored the relationship of age and tumor grade in 13 types of digestive tumors. In the multivariable regression models, both positive and negative associations were observed in different tumor types. Interestingly, in patients with squamous cell carcinoma, older patients tended to have higher tumor grades, while an inverse relationship was observed in most of the adenocarcinoma. Age was negatively associated with tumor grade in patients with esophageal adenocarcinoma, gastric adenocarcinoma, pancreatic adenocarcinoma and colorectal adenocarcinoma. Notably, patients with intra-and extra-hepatic cholangiocarcinoma did not show a positive association between age and tumor grade. In addition, possible interactions with sex and with race on the primary outcome was observed in four types of tumor (Table 3). We also observed better long-term prognoses for patients with well-differentiated tumors, which is consistent with results in previous studies. [3,5] Previous studies have demonstrated the different clinicopathologic characteristics including tumor grade between younger and older patients with tumor. [9,11,12] Several studies showed that younger patients tend to present with more histologically aggressive tumors. [12,13] Figure 1 The number of patients with 13 types of tumor were shown by sex and age (age was divided into ten groups). A,