How to Choose the Best Procedure Type for Duodenal Neuroendocrine Tumors With a Maximum Diameter of 1 to 2 Cm a Propensity Score Matching Analysis Based on the SEER Database

Background: The treatment plan for duodenal neuroendocrine tumors (d-NETs) with a diameter between 1 and 2 cm is still controversy. Aim: To compare the effects of local endoscopic resection and radical resection on the prognosis of d-NETs with a maximum diameter of 1-2 cm. Methods: 286 eligible patients were identied from the SEER database. Propensity score matching (PSM) was done to match patients 1:1 on clinicopathological characteristics. Kaplan-Meier analysis was used to analyze the factors affecting the prognosis. Results: Before PSM, there was no signicant difference in the cancer-specic survival (CSS) between the two groups (P = 0.595), but the tumor size, T stage, N stage, and M stage were signicantly different between the two groups (all P < 0.05). After 1:1 PSM, the differences in clinicopathological characteristics between the two groups were signicantly reduced (all P > 0.05). Survival analysis showed that only the tumor grade was correlated with the prognosis (P = 0); surgical method and other clinicopathological characteristics were not correlated with the prognosis (all P > 0.05). Conclusion: The surgical approach had no signicant effect on the prognosis of d-NET patients with a maximum diameter of 1-2 cm and without lymph node metastasis.


Background
Duodenal neuroendocrine tumors (d-NETs), a type of tumor originating from neuroendocrine cells, account for approximately 3% of the total gastrointestinal neuroendocrine tumors. The incidence of d-NETs has increased signi cantly in recent years with the popularization of endoscopy [1,2]. Surgical resection is the only treatment that can cure d-NETs. Currently, the choice of surgical regimen is mainly based on tumor size, tissue grading, and the presence of lymph node metastasis on imaging [3,4]. For tumors < 1 cm in diameter that are not in the periampullary region and that have no suspicious lymph node metastasis, endoscopic resection is recommended. For tumors larger than 2 cm in diameter, radical resection is recommended. For tumors in the periampullary region having different biological behaviors from neuroendocrine tumors (NETs) in other regions, local resection and lymph node biopsy or radical resection are required. However, for tumors with a diameter between 1 and 2 cm, due to the lack of prospective studies and large clinical studies, the speci c treatment plan is still not standardized [3,4].
Therefore, further studies with large sample sizes are needed to con rm whether endoscopic treatment or surgical resection is best.
The Surveillance, Epidemiology, and End Results (SEER) database is a tumor registry database established in the 1970s in the United States, covering approximately 28% of cancer patients in the United States. The database provides important data support for clinical research and clinical decision-making.
However, since the SEER database includes data from multiple cancer registration centers, the real-world 2. Results 2.1 General condition and survival analysis before matching Data of 3709 pathologically diagnosed d-NET patients from 2004 to 2016 were obtained from the SEER database. After they were screened by the above inclusion and exclusion criteria, 286 d-NET patients met the inclusion criteria, including 130 patients with local resection and 156 patients with radical resection noncontrolled data are imbalanced to a certain extent and have many missing values. To balance the baseline characteristics between groups, this study collected the clinicopathological and prognostic data of d-NET patients with a tumor of 1-2 cm in diameter in the SEER database. The propensity score matching (PSM) method was used to match the characteristics between the groups to investigate whether radical resection improved the long-term prognosis of patients. were enrolled. Exclusion criteria: (1) the tumor grade was unknown or undifferentiated and anaplastic for Grade IV; (2) lymph node metastasis or distant metastasis was unknown; (3) the cause of death was unclear or death was not tumor-related; (4) the d-NET was combined with other tumors; (5) since our aim was looking at long-term outcomes, we excluded patients who died from surgery or whose survival time was less than 1 month. Due to the strict register-based nature of the study, informed consent was waived. Moreover, the study was exempted from Institutional Review Board approval, in view of the SEER's use of unidenti able patient information.
1. 1.2 Demographic characteristics, such as sex, age, race, and marital status, and pathological characteristics, such as tissue grade, tumor stage (T stage), and node stage (N stage), were collected. According to detailed information such as tumor size and local extension provided by the SEER database, the TNM status of patients was rejudged according to the 7th edition of the American Joint Committee on Cancer (AJCC) staging system. All data collection and statistical calculations were independently completed by two authors (Jiebin Xie and Huanyu Tang).

1.3 Statistical analysis
Cancer-speci c survival (CSS) was taken as the study endpoint. The chi-squared test or t test was used to quantify the differences between surgical groups. The Kaplan-Meier method was used for survival analysis. The above statistical analysis was performed using SPSS 22.0. P < 0.05 was considered statistically signi cant. The MatchIt package of R software v3.6.3 was used to perform the 1:1 PSM with a caliper value set to 0.1. The nearest-neighbor matching method was used to match the baseline characteristic differences between the two groups.
( Fig. 1). Among all tumors, 61.19% were T2, 85.67% were G1, 75.17% were Carcinoid tumor (ICD-O-3 histology code: 8240). The total N1 proportion was 22.03%, including 35.9% in the radical resection group and 4.62% in the local resection group ( Table 1). The nal follow-up time was November 2018. The overall 5-year CSS was 93.2 ± 2.1%. The 5-year CSS of patients in the local resection group was 92.9 ± 3.7%. The 5-year CSS of patients in the radical resection group was 93.4 ± 2.3%. Kaplan-Meier survival analysis showed no signi cant difference in survival rate between the two groups before PSM (P = 0.595, Fig. 3a). 2.2 Comparison of baseline data and the prognosis of the local resection and radical resection groups before and after matching Before the matching, there was no signi cant difference in mean age between the local resection group (64.38 ± 12.00 years) and the radical resection group (61.68 ± 12.24 years, P = 0.062). The mean tumor size in the local resection group (12.88 ± 3. 6 mm) was signi cantly smaller than that in the radical resection group (14.20 ± 3. 29 mm, P = 0.001). The T stage, the N stage, the M stage, and the TNM stage in the radical resection group were signi cantly higher than in the local resection group (all P < 0.05). The differences in race, marriage, sex, tissue type, and degree of differentiation were not statistically signi cant between two groups. To eliminate the differences in baseline characteristics between the two groups, PSM was used to balance all clinicopathological valuables and demographic characteristics (Fig. 2). A total of 154 patients were selected according to the chosen 1:1 ratio, including 77 in each group. On this basis, the comparison of patients in the matched groups showed that the differences in clinicopathological characteristics were signi cantly reduced, and none of the above characteristics were signi cantly different between two groups after matching (Table 1).

Survival analysis and subgroup analysis after matching
After matching, the 5-year CSS of patients in the local resection and radical resection groups was 96.2 ± 2.8% and 98.4 ± 1.6%, respectively. Survival analysis showed no signi cant difference in CSS between the two groups (P = 0.558, Fig. 3b). Except for the degree of differentiation (P = 0, Fig. 3c), the surgical approach, and other clinicopathological features, such as age, sex, T stage, and tissue type, were not correlated with prognosis (Table 2). Further survival analysis of nonmetastatic d-NET patients with T2 + stage showed that the 5-year CSS was 98.1 ± 8.18% in the local resection group and 97.8 ± 2.2% in the radical treatment group, which was still not signi cant (P = 0.973 Fig. 3d).

Discussion
Although d-NET is a rare tumor, the incidence rate in recent years has been rising. Seventy-ve percent of d-NET patients have a tumor diameter below 2 cm at the time of initial diagnosis, most of which are con ned to the mucosa and submucosa. For nonampullary tumors smaller than 1 cm with better differentiation, lymph node metastasis rates are relatively low, and the current guidelines recommend endoscopic treatment as the main treatment. Patients with a tumor diameter of 1-2 cm have a higher likelihood of distant lymphatic metastasis and a thinner wall in the duodenum than in other gastrointestinal tracts. Therefore, the rate of surgical resection is currently higher for these patients.
Radical resection can more completely remove the primary lesion and the primary foci. However, due to the special location of the duodenum, the risk of radical resection is greater, the treatment cost is high, the hospital stay is longer, and postoperative complications are more common [5][6][7]. In addition, the differences in the effects of endoscopic vs. surgical treatment on prognosis are still rarely reported. Therefore, there is still much debate about the treatment of d-NET patients with a tumor of 1-2 cm in diameter [3,4].
Margonis et al. [7] studied 146 d-NET patients undergoing different surgical methods and found that the prognosis of d-NET was correlated with tumor grade and metastasis at the time of diagnosis but was not correlated with the surgical method. Our study showed that among the 286 patients who met the inclusion criteria, there was no signi cant difference in the CSS between the two groups before matching.
However, fewer of our patients underwent endoscopic resection than radical resection, and the endoscopic resection was mainly done in patients with older age, smaller tumors, and N0 tumors. Considering the differences in the clinicopathological characteristics, sample size, and risk factors before treatment between the patients in the local resection group and the radical resection group, which led to the di culty of balancing the covariates between the groups, PSM was used to control the bias that resulted. PSM can simultaneously match the characteristics of multiple factors, minimize confounding bias, and better simulate clinical studies, especially when we are unable to do a prospective clinical study or the clinical study is low-quality. Analysis based on a large sample size after PSM has more reference value [8]. In the present study, after the PSM, there was no signi cant difference in the clinicopathological distribution characteristics between the two groups, which improved the reliability of the conclusions of the subsequent analysis.
Endoscopic treatment, as the main treatment method for tumors less than 1 cm, has the signi cant advantages of short operation time, low cost, short hospital stay, and less impact on quality of life [9].
With the progression of ultrasound endoscopy, endoscopic mucosal resection, and endoscopic submucosal dissection, the advantages of endoscopic resection treatment of gastrointestinal NET tumors have become obvious. The analysis of the endoscopic treatment effect and prognosis of gastrointestinal NET patients by Chen [10] showed that endoscopic submucosal dissection can be used for both en bloc and complete resection of lesions ≤ 3.5 cm, and no new metastasis occurred during the follow-up. Our results showed that after matching, surgical method, age, sex, T stage, and tissue type were not correlated with prognosis, but the degree of differentiation was (P = 0), in line with the ndings reported by Margonis et al [7]. To further analyze the survival advantage of radical resection in patients with ≥ T2 d-NET the patients with lymph node metastasis or distant metastasis were excluded for strati ed analysis, the results still showed that there was no difference in survival between patients with radical resection and local treatment. Local endoscopic treatment is still recommended for G1 patients regardless of sex, age, marital status, and T stage.
However, due to the thin wall of the duodenum and its rich blood vessels, most tumors invade the submucosa. Therefore, duodenal endoscopic submucosal dissection is truly challenging, and even Japanese experts think twice before indicating [11], because there are risks of bleeding, perforation, positive margins, and missed metastatic lymph nodes in endoscopic resection treatment [8,9,[12][13][14][15][16]. Some studies reported that the lymphatic metastasis rate of d-NET with a tumor of 1-2 cm in diameter was approximately 60% [17,18]. Our data showed that among the 286 patients included, the T2 rate was 61.19%, which was the highest, the T3 + rate was 11.54%, and the total lymph node positive rate was 22.03%, which included 35.9% of the radical resection group and 4.62% of the local resection group. Our ndings are consistent with the reported studies. The earlier studies may have had higher lymphatic metastasis rates because all of their patients underwent radical resection. Combining those data with ours, approximately 71.33% of d-NET (T1N0, T2N0) patients with a tumor of 1-2 cm in diameter can be treated with local endoscopic resection. However, for patients with T3 + N0, choosing the surgical method is a dilemma. If radical resection is chosen, the risk of surgery is increased, the treatment risk and the length of hospital stays are signi cantly prolonged 7 , and there is no survival bene t over local resection.
However, endoscopic treatment also has the risks of perforation, bleeding, and positive resection margin. It is recently reported that endoscopic resection of full-thickness diseased tissue combined with continuous laparoscopic suture treatment has achieved good clinical outcomes in the treatment of d-NETs [19]. This method can completely remove the lesion tissue, while also detecting the presence of swollen lymph nodes and metastatic lesions in the abdominal cavity and strengthen the suture wound.
This should be the optimal surgical treatment plan for these patients, but few studies have con rmed its short-term and long-term e cacy; thus, more prospective studies are needed. However, for N + patients, radical resection of metastatic lymph nodes should be the standard treatment.
This study has some limitations. First, it was a retrospective study based on the SEER database, and there are confounding factors of mismatching, such as the difference in surgeon skills and the general condition of the patients. Second, to ensure the integrity of the data, we deleted many cases, so the total sample size was small. In particular, the numbers of N1, T3, and T4 patients after matching were small. Third, the advantages and disadvantages of the two surgical methods are important factors in their longterm and short-term e cacy, but the SEER database does not provide information on postoperative complications, which limits the comparison of the short-term e cacy. Although our data are not ideal, PSM performed a good balancing of the clinical and pathological characteristics of the two groups, reducing the selection bias. We also compared the differences in overall survival (OS) between the two groups and achieved consistent results with previous studies. Our results still need to be validated by a prospective, multicenter, randomized controlled study.

Conclusion >
The surgical approach had no signi cant effect on prognosis, and the degree of tumor differentiation was an independent prognostic factor in d-NET patients with a maximum tumor diameter of 1-2 cm.
Local endoscopic resection is still recommended for G1 patients regardless of factors such as sex, age, marital status, and T stage, but for patients with T3 + N0M0, combined laparoendoscopic full-thickness resection may be the best procedure type in the future.

Declarations
Ethics approval and consent to participate This study was exempted from Institutional Review Board approval, in view of the SEER's use of unidenti able patient information. Due to the strict register-based nature of the study, informed consent was waived.

Consent for publication
Not applicable Availability of data and materials The data were obtained from the SEER database.

Competing interests
The authors declare that they have no competing interests    Distribution pro les of the clinicopathologic factors of the patients in the local resection group and radical resection group before and after PSM matching, and distribution pro les of the lter-in and lterout patients.