Predicting Progression, Recurrence, and Survival in Pancreatic Neuroendocrine Tumors: A Single Center Analysis of 174 Patients

Introduction The European Neuroendocrine Tumor Society, ENETS, reports variables of prognostic significance in pancreatic neuroendocrine tumors (PNET). However, studies have short follow-ups, and the optimal treatment remains controversial. We aimed to determine overall survival (OS), progression-free survival (PFS) after conservative treatment, and recurrence-free survival (RFS) after surgery and further to find predictors of aggressive PNET behavior to support treatment decisions. Methods 174 patients with PNET treated at Aarhus University Hospital from 2011 to 2021 were included in a retrospective cohort study. Patients were divided into surgically resected (SUR, n=91) and medically or conservatively treated (MED, n=83). Variables were tested in univariate and multivariate survival analysis. Median follow-up time was 3.4 years in the MED group and 4.5 years in the SUR group. Results The 5-year OS was 95% and 65% for the SUR and MED groups, respectively. The 5-year RFS in the SUR group was 80% whereas the 5-year PFS in the MED group was 41%. Larger tumor size, Ki67 index, tumor grade, and stage were predictive of shorter OS, RFS, and PFS. Further, chromogranin A was a predictor of OS. Larger tumor size was associated with higher stage and grade. Only 1 of 28 patients with stage 1 disease and size ≤2 cm developed progression on a watch-and-wait strategy during a median follow-up of 36 months. Conclusion This study supported the ENETS staging and grading system to be useful to predict OS, PFS, and RFS in PNET. Further, our data support that small, localized, low-grade PNETS can be followed with active surveillance.


INTRODUCTION
Pancreatic neuroendocrine tumors (PNETs) are rare tumors, constituting 1%-2% of all pancreatic cancers and up to 10% of all NETs (1)(2)(3). The incidence has increased in the past 30 years and there has been a significant improvement in survival (3)(4)(5)(6). PNETs are classified as functioning (F)-or non-functioning (NF) according to the potential hormone production. NF-PNETs comprise at least 70% and are discovered either incidentally or due to symptoms as a sign of advanced disease (7,8).
Primary investigation of newly discovered tumors involves staging and grading. The European Neuroendocrine Tumor Society (ENETS) has proposed a staging system, which is widely recognized to predict survival (9). Micro-radical surgery is considered the only curative treatment and is associated with increased overall survival (OS). However, the management is complex, and the benefits of surgery must be weighed against the relatively high risk of perioperative morbidity (3,(10)(11)(12). Surgery is considered in early-stage disease but is often contraindicated in widely metastatic disease or in patients with a poor performance status (13,14).
The incidental detection of asymptomatic PNETs is increasing along with the availability and sensitivity of imaging techniques. These pancreatic incidentalomas (PI) are small and resectable but often have indolent biology. Therefore, welldifferentiated PNETs ≤2 cm are often managed conservatively (10,15). However, approximately 10% of tumors ≤2 cm have lymph node involvement (16,17), and a meta-analysis suggested a survival benefit for surgery even in smaller tumors (18).
Thus, our study aimed to investigate tumor characteristics in patients with PNETs related to both prognosis and aggressiveness based on the ENETS guidelines. Further, we wished to perform a subgroup analysis of localized PNETS ≤2 cm to test the current guidelines recommending a watch-andwait strategy in small PNETS (19).

METHODS
In this single-center retrospective study, we identified 174 patients with PNET referred to Aarhus ENETS NET center of excellence from 2011 to September 2021. Ninety-one were surgically resected (SUR group) and 83 were medically treated and/or actively followed with a watchful wait strategy (MED group) according to the ENETS guidelines (19). The study was approved as a quality assurance project by The Central Denmark Region Committees on Health Research Ethics. Diagnosis was based on histology or somatostatin receptor-based imaging. See Figure 1.
Data were collected from the online record system, Electronic Patient Journal, in 2021 through the unique Civil Personal Registration numbers, given to all Danish citizens and residents (20). It was managed using REDCap electronic data capture tools (21) hosted at Aarhus University.

Outcome
Patients were followed until the time of death or until the end of follow-up on September 25, 2021. Median follow-up time was 3.4 years in the MED group and 4.5 years in the SUR group. The primary endpoints were OS, progression-free survival (PFS) in the MED group, and recurrence-free survival (RFS) in the SUR group.

Statistical Analysis
Data were analyzed using Stata 17 (StataCorp, College Station, TX USA). OS was time from diagnosis to death; PFS was time from diagnosis to progression, determined as clinical progression at a multidisciplinary NET tumor board meeting; and RFS was time from surgery to recurrence, or until the end of follow-up. Patients followed for less than 3, 5, and 10 years were censored in the respective survival analysis. OS, PFS, and RFS were calculated with Kaplan-Meier methodology, and log-rank tests compared categorical variables across subgroups. Cox proportional hazard models were used to estimate HR with 95% confidence interval for all significant variables in a univariate analysis and finally in a multivariate analysis to identify independent predictors. Subgroups and association in-between variables were compared using simple ttests. P-values ≤0.05 were considered statistically significant.

RESULTS
Of 174 patients, 48% were females and the average age was 65.8 ± 12 and 54.7 ± 13 in the MED and SUR group, respectively. Clinicopathological characteristics are summarized in Table 1.
In the overall MED group, 53% experienced progression and 29% died of PNET. In the SUR group, 16% had recurrence and 6% died of PNET. The 5-year OS was 65% in the MED group versus 95% in the SUR group (p ≤ 0.05). Due to selection bias, the MED and SUR groups are not directly comparable. The MED group had more advanced disease, more co-morbidities, and higher age ( Table 1).

Predictors of OS and PFS in the MED Group
The 83 patients in the MED group comprised 20 patients who received no initial treatment. The mean tumor size was 1.3 ± 10 cm. Further, the group comprised 63 patients, medically treated from the time of diagnosis. The mean tumor size was 4.6 ± 4 cm.

Correlation Between Tumor Size, Stage, and Grade
Larger tumor size was related to lymph node positivity, metastatic disease, and higher grade (p ≤ 0.05, Table 4).

Insulinomas and Repeated Analysis After Exclusion
The SUR group comprised 15 insulinomas, and among these, there were no deaths. Only one patient experienced progression.
Insulinomas have an excellent prognosis and all analyses were repeated after exclusion of these. Ki67 index and NEC-G3 were still significant predictors of survival in the non-insulinoma SUR group (p ≤ 0.05). Ki67, G1-and NEC-G3, tumor infiltration, lymph node positivity, and tumor stage were still predictors of recurrence (p ≤ 0.05). The 5-year OS and RFS in the noninsulinoma SUR group were reduced to 94% and 76%, respectively.

Tumors ≤2 cm
Fifty-nine patients with tumors ≤2 cm had localized G1 disease; 31 in the SUR group and 28 in the MED group. All 28 patients in the MED group were followed with a watch-and-wait strategy, and the group comprised 16 patients who were followed without treatment and 12 patients who were followed on Somatostatin Analogues. During a median follow-up of 36 months only one in 28 patients in the MED group experienced progression while 1 in 31 patients in the SUR group experienced recurrence during a median follow-up of 56.5 months. The SUR and MED groups differed in age, PI, and F-PNET ( Table 5).

DISCUSSION
This large single-center cross-sectional study from an ENETS center of excellence demonstrated that high TNM-stage is a significant predictor of both RFS in surgically treated patients and PFS and OS in patients treated medically or with no treatment. This is in agreement with previous studies demonstrating that stage IV disease is the strongest predictor of a poor prognosis regardless of any other variable (23)(24)(25). We further demonstrated that high tumor grade is a strong, negative predictor of OS. Both tumor stage and grade are widely used for prognostic assessment, and the ENETS classification system for PNET has been evaluated previously (23,(25)(26)(27)(28)(29). In line with Ekeblad et al. and Scarpa et al. (23,26), we found no significant difference between stage I and II disease ( Table 2). This was also the case for G1 and G2 tumors and this may be caused by type 2 error. Meanwhile, Ki67 was a significant predictor of OS both in the MED and in the SUR groups. This is supported by Panzuto et al. (30) who also found that an increase in the Ki67 index was associated with poorer survival. Brooks et al. showed that surgery in PNET is an independent predictor of OS (31). Our findings support these data as we demonstrated a higher 5-year OS after surgery (95% versus 65%). However, our study is non-randomized and retrospective, and the selection of the patients biases our results. Surgery was performed in younger patients with lower grade and stage  tumors and results should be interpreted with caution and with this selection bias in mind. F-and NF-PNETs are suggested to differ in aggressiveness and hence have a different prognosis (7,23,25,26,28). We were unable to demonstrate this. Insulinomas have an excellent prognosis after surgery, and we therefore, tried to exclude these and repeat all analyses. The 5-year OS remained excellent in the SUR group and only minor changes in the RFS were observed.
A recent meta-analysis demonstrated that post-surgery recurrence was higher in patients with high grade tumors, lymph node involvement, or vascular invasion (32). This is in accordance with our study, demonstrating that patients with higher stage and grade have a decreased RFS after surgery and therefore warrant closer follow-up.
In agreement with previous literature, PIs had a more indolent behavior compared to those that were symptomatic at diagnosis. The PFS in our study was longer, and they were more likely to be lymph node negative and have a low grade. This coheres with the fact that they are discovered early (8,10,16,17,24).

Tumors ≤2 cm
The 2016 ENETS Consensus Guidelines (19) suggest a conservative approach in non-metastatic, NF-tumors ≤2 cm. Our findings support that tumor size is a predictor of aggressive behavior. We demonstrated that larger tumor size was predictive of both higher stage and grade but also the presence of lymph node metastasis. Smaller size was also associated with lower RFS after surgery ( Table 3). This all agrees with previous studies (23,(25)(26)(27)32 (4,16). Overall, they showed that PI can display aggressive behavior despite small size. Further, a meta-analysis from 2017 demonstrated survival benefits in tumors ≤2 cm (18).
In our study, 59 patients had stage I tumors ≤2 cm and 53% underwent surgery. The SUR group comprised more functioning tumors and younger patients than the MED group. Only 3% of the resected tumors ≤2 cm showed recurrence. This supports a recent study from Sallinen et al. who demonstrated an excellent disease-free survival after surgery (33). As long-term results after surgery are excellent, the outcome of non-operative management of tumors ≤2 cm is of paramount interest. The 28 patients in the MED group were followed with a watch-and-wait strategy. Only one patient (3%) experienced progression. Although the total number is limited, our findings support a conservative approach in accordance with the 2016 ENETS Consensus Guidelines (19).
In conclusion, in this large cohort of PNETs we demonstrated that high TNM-stage, tumor grade, Ki67 index, size, CgA, and symptomatic discovery are negative prognostic predictors of survival. Further, the surgically treated group had the highest survival, and we support the guidelines recommending surgery when predictors of aggressive tumor behavior are present. Further, we believe that a watch-and-wait strategy with active surveillance can be followed in patients with low grade, low stage NF-PNET ≤2 cm (19).

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ACKNOWLEDGMENTS
The content of the manuscript has previously appeared in a thesis by SK. 5 | Clinicopathological data in 59 patients with stage I pancreatic neuroendocrine tumors ≤2 cm divided into surgically treated (SUR group, n = 31) and medically treated or non-treated (MED group, n = 28).