Outcome of nonfunctioning pancreatic neuroendocrine tumors after initial surveillance or surgical resection: a single-center observational study

Introduction

In recent years, small pancreatic neuroendocrine tumors (pNETs) have been increasingly detected by chance, partially because of the widespread use of high-resolution imaging techniques for screening programs or nonspecific abdominal symptoms. Thus, patients are often diagnosed in an early and asymptomatic disease stage, with a trend toward smaller tumor size and a lower grade of malignancy [1-3].

Between 60% and 90% of all pNETs are nonfunctioning (nf) pNETs, referring to tumors without clinical symptoms of hormonal hypersecretion. More than half of them present metastatic disease at diagnosis, and approximately 20% are locally advanced [4,5]. The 2017 WHO classification distinguishes well-differentiated NETs from poorly differentiated neuroendocrine carcinomas (NECs). According to the Ki67 proliferation index, pNETs are further subclassified into G1 (<3%), G2 (3-20%), and G3 lesions (>20%) [6].

As surgical resection carries a high risk of morbidity and mortality, current guidelines consider observation a reasonable alternative strategy for G1 or low-grade G2 pNETs ≤2 cm [5,7]. However, this wait-and-watch strategy is controversial. Strong evidence for long-term safety is lacking, and available data seem in part inconsistent. Thus, active surveillance based on high-resolution imaging is mandatory and may be associated with considerable costs. In some retrospective studies, the risk of malignancy in small nf pNETs is estimated at 5-8% [8-10]. In contrast, a recently published review of 6 retrospective studies, including a total of 344 patients with small sporadic nf pNETs, showed that approximately one-fifth of them (pooled estimate 22%) had an increase in lesion size during follow up (range 32-45 months), while none of the patients developed metastatic disease. The rate of secondary surgery in these studies ranged between 3% and 25% [11].

The principle aim of the present work was to characterize the natural behavior of small nf pNETs under careful observation and confront the data with the results and outcomes of patients undergoing upfront surgery.

Patients and methods

Results

During the observation period of 18 years, 263 patients with suspected pNETs were identified. At the moment of diagnosis, metastatic or locally advanced disease was observed in 79 and 7 patients, respectively. Eighty-one patients had lesions >2 cm, 66 of whom were treated by surgical resection, whereas 15 did not undergo surgery because of age, comorbidities or patient refusal. Ten patients were treated by EUS-guided radiofrequency ablation. Twenty-three patients had insufficient or inadequate follow up (<12 months, severe comorbidities or short life expectancy). In 10 patients, histological diagnosis was missing. Thirty-nine patients treated with the surveillance strategy were finally included, and 14 patients underwent surgery immediately after diagnosis (flowchart patient inclusion, Fig. 1).

Figure 1 Flowchart patient inclusion

pNET pancreatic neuroendocrine tumor; RFA, radiofrequency ablation; FU, follow up

Outcomes in patients with careful surveillance strategy

The demographic characteristics of the 39 patients in whom a careful watch-and-wait strategy was employed are outlined in Table 1. Their mean age was 62.1 years (SD 10.1; range 36-83 years); 22 (56%) of them were male, and 17 (44%) were female. Nineteen of 43 lesions (44%) were localized in the head/neck, and 24 (56%) were localized in the body/tail region of the pancreas. Twenty-six percent of all lesions were cystic (11/43). Most patients had single lesions, except for 3/39 (7.7%). Two patients were diagnosed with a hereditary syndrome (MEN1, Von Hippel Lindau), but both of them had single pancreatic lesions.

Table 1 Clinical characteristics at diagnosis of patients with pNET ≤2 cm and expectant management (43 lesions in 39 patients)

The mean lesion size at the moment of diagnosis was 12.6 mm (SD 3.9; range 5-20 mm). Ten of 43 (23%) lesions were <10 mm, and 33 (77%) were 10-20 mm. Histological analysis revealed 32 G1 tumors (82%) and 7 G2 tumors (7/39 (18%), Ki67 range: 3-7%).

At the end of radiological follow up (mean FU 47±26 months; range 12-144 months, see Table 2), the mean tumor size was 15.1 mm (±7.6; range 5-38). A slight change in size (<20%) was observed in 3 lesions (2 G1 lesions and 1 G2 lesion with a Ki67 index of 5%). Twelve patients showed a significant change in the size of the lesions. Based on EUS-FNB, 6 of them were considered G1 lesions and 6 G2 lesions (Ki67 range: 3-7%). Half of the growing lesions reached a size >2 cm by the end of follow up. Fig. 2 shows the Kaplan-Meier estimation of tumor progression in patients under surveillance. There was a significant difference according to the tumor grade: all G2 lesions progressed during follow up (P=0.0018).

Table 2 Outcome of 39 patients with expectative management for 43 pNETs ≤2cm

Figure 2 Progression of small pNETs under surveillance, classified as grade 1 or grade 2. Patients without progression were right-censored on the day of last medical follow up

pNET, pancreatic neuroendocrine tumor

As a consequence of progression in tumor size or the presence of worrisome features (main pancreatic duct dilation) during follow up, 6 patients underwent secondary pancreatic resection, as outlined in Table 3. The mean time from diagnosis until surgery was 39.5 (±27.5) months. Three patients had lymph node-positive disease on resected specimens, but lymph node metastasis was not observed in any of them on preoperative imaging studies. Bearing in mind the time gap between initial evaluation by EUS-FNB and the histopathology of the surgery specimens, we found confirmation of tumor grade in 4 patients and initial overestimation and underestimation in 1 patient. In 1 patient, an initial G2 lesion (Ki67 index 4%) was classified as a G3 lesion after surgical resection (Ki67 index 22%; 66 months of surveillance).

Table 3 Secondary resection after initial surveillance in 6 patients with pNETs ≤2 cm

Pancreatic surgery was followed by severe complications in 2/6 patients (33.3%), who developed pancreatic fistula with erosion of splenic vessels. In one case, hemorrhage was treated successfully, whereas bleeding could not be stopped by means of embolization or surgery in a 58-year-old male patient who died shortly thereafter.

No disease-related death or distant metastasis was observed by the end of follow up in non-resected patients undergoing ongoing surveillance.

Outcome in patients treated by upfront surgery

Fourteen patients underwent pancreatic surgery for neuroendocrine tumors ≤2 cm between November 2002 and March 2017 (Table 4; mean age 57±12.6 years, 64% (9/14) female). All patients had sporadic pNETs. The mean size of the pancreatic lesions was 10.8 mm (±3.7 mm). Five of them (36%) had worrisome features on preoperative imagery studies (dilation of the main pancreatic duct), and 1 patient underwent pancreatic resection because of diagnostic uncertainty (Table 4). The others were treated by surgery, according to the institutional policy or patient preference. In 5 patients, caudal pancreatectomy was performed, while 3 patients underwent pancreaticoduodenectomy, median pancreatectomy or enucleation.

Table 4 Outcomes of patients treated by surgery for nf pNET ≤2 cm (n=14)

Preoperative EUS-FNB was performed in 5/14 patients. The Ki67 index was confirmed in 2 patients after surgery. Overestimation of the Ki67 index was observed in 1 patient, and underestimation was observed in 2 patients. In one of them, histopathological analysis of the resected specimen caused a real change in tumor grading (G1 to G2). Details are shown in Table 4. One patient had lymph node-positive disease confirmed in the resected specimen.

In patients with “worrisome features” (5/14), the post-surgical histopathological diagnosis was G1 or low grade G2 pNET without lymph-node metastasis during surgery in 80% (4/5 patients).

Eight of 14 patients (57%) suffered at least 1 postoperative complication (postoperative pancreatic fistula, n=3; collection, n=2; stenosis of the hepaticojejunal anastomosis, n=2; pleural effusions and tachyarrhythmia, n=1).

After a mean follow-up time of 69 months (range 31-153), 13/14 patients remained recurrence free. Three years after the Whipple procedure (complete resection, G2, 15 mm, 0 N+/8 N), a 64-year-old female patient developed lymph node metastasis (abdominal and mediastinal), which later progressed to liver and bone metastasis. She was treated by palliative radiochemotherapy and died 7 years after surgery.

Discussion

The decision-making process in the management of small nf pNETs remains challenging, and unequivocal criteria to predict the behavior of these tumors are still scarce. Whereas pancreatic surgery bears the risk of overtreatment in indolent neoplasms, nonoperative management by careful observation policy can miss the window of opportunity to cure potentially aggressive tumor disease.

Several studies argue for an indolent disease course of small pNETs, suggesting that a careful “wait-and-watch strategy” seems reasonable in selected patients [12-16]. Interim analyses of the prospective PANDORA and ASPEN trials suggest that a non-operative strategy seems safe, as only a negligible fraction of patients have an increase in tumor size [17,18]. In a recent meta-analysis, 84 of 267 patients with sporadic nf pNET showed increasing tumor size (range across studies 0-51%, pooled estimate 22%), but no patients developed lymph-node or distant metastasis during follow up. The percentage of patients undergoing surgical resection after expectant management varied highly, between 3% and 25% (pooled estimate 12%) [11]. Another meta-analysis revealed 14% of patients undergoing pancreatic surgery after the initial wait-and-watch strategy, with a median time from diagnosis to surgery of 30-41 months [19].

The results of the present work suggest even higher progression rates. Accordingly, approximately one-third of all lesions were found to be growing during follow up (35%). At the end of follow up, half of the growing lesions had reached a size >2 cm. Three of these patients were secondarily treated by surgery and found to have lymph node-positive disease on resected specimens. This does not appear surprising, as tumor size is known to correlate with the probability of lymph node involvement [20].

In a retrospective study on predictors of lymph node involvement in nf pNETs, Partelli et al found intraoperative nodal metastases in 30% of patients, approximately half of which were not visualized on preoperative cross-sectional imaging. However, 39% of all 181 included patients turned out to have G2 lesions, and the radiological tumor size before surgery was >4 cm in 26% of patients [21]. In the present study, the rate of intraoperative discovery of lymph node disease was similar in both groups: in patients who were primarily (1/14 [7.1%] G2 lesion) resected and those who were secondarily (3/39 [7.7%], all G2 lesions) resected.

All pNETs classified as G2 in the present work (n=7; initial tumor size 12-18 mm) progressed over time. Previous studies have shown that the Ki67 index is a key prognostic factor for the disease-specific survival of patients with pNETs [20,22,23]. In contrast, in a multicenter study with 88 patients, the Ki67 index was not found to be an independent predictor of malignancy. Unfortunately, only a minor portion of the currently available studies reporting on the wait-and-watch strategy in small pNETs provide histological grading and the Ki67 index. Furthermore, precise information about the correlation between grading and progression is scarce [13-15,24]. Based on the results of the present study, a careful watch-and-wait strategy for G2 pNETS should be applied with great caution. There should be a low threshold for more invasive treatment in cases of progression during close follow up.

In addition, this preliminary study found that a considerable proportion of G1 pNETs progressed during follow up (8/43, 18.6%). Even though the difference between the mean size of all included pNETs at the beginning and at the end of follow up was only 3 mm, a total of 8 G1 lesions progressed, with a mean increase of 4.6 mm. In a matched case-control study by Sadot et al, 50% of observed lesions progressed, but the median tumor size had not changed at the end of follow up (1.2 cm, P=0.7). Notably, 31% of lesions showed decreasing tumor size. Similarly, in our study, 4/43 lesions regressed (n=4, mean difference in size 5.5±3.1 mm). Until the end of follow up, none of the patients with G1-pNET developed lymph node or distant metastasis, and there was no disease-related death. This is concordant with previously published data [11,14,19].

One of the patients treated with upfront surgery in this study suffered disease recurrence 3 years after R0/N0 resection of a G2 lesion. Sadot et al described 77 initially resected patients with pNETs <3 cm, of whom 5 (6%) had recurrent disease after a median follow up of 5.1 years. Since they found a rate of late metastasis and recurrence in 3/39 (7.7%) patients with lesions <2 cm, Haynes et al suggested tumor resection and careful postoperative surveillance, even in small pNETs with favorable pathological findings. However, only 1 of these 3 patients was classified as “benign”, defined as well-differentiated without characteristics of malignant disease [8,14]. In the present study, 5/14 patients underwent upfront surgery because of worrisome features: almost all of them had localized G1 or low grade G2 disease in the resected specimen. In 1 patient, a known G2 lesion (EUS-FNB before surgery) was shown to have lymph-node metastasis on the resected specimen.

However, morbidity and mortality inherent to pancreatic surgery are still considerable [7,25]. In the present study, the pooled surgery-associated complication rate in patients treated by primary or secondary resection was 50%, and 1 patient died following caudal pancreatectomy. Previous studies reported comparable results. In a retrospective review of 55 patients in Tel Aviv treated by upfront surgery for mainly G1 small pNETs (48/55 lesions with Ki-67 index <3%), adverse events were reported in 51%, and 1 patient died because of a complicated pancreatic leak. Parenchymal-sparing resections are considered a reasonable treatment strategy for small low-grade malignant tumors. Nevertheless, enucleation, for instance, is associated with a high complication rate, mainly due to pancreatic fistula. It is important to remember that the importance of regional lymphadenectomy in small pNETs remains controversial and lymph node dissection is not systematically performed [26,27]. Thus, less invasive ablation strategies, such as EUS-guided radiofrequency ablation, should present a promising alternative treatment strategy in well-defined patients. Associated adverse events are often transient in nature and can be managed noninvasively or endoscopically. Of course, long-term outcomes have to be awaited, but they might be comparable to those of surgery, especially if patients are treated at an early stage.

The patient review reported here has certain limitations that deserve discussion. The results are limited by the observational and monocentric study design, with a restricted number of patients. Thus, it was decided to present results mainly in a descriptive manner and provide detailed insight into patient outcomes. Evaluation of tumor progression is principally based on imaging reports from radiologists with experience in pancreatic disease. However, a systematic retrospective review of images was not performed. Follow ups were based on MRI or CT imaging studies, as they seemed most objective and reproducible. However, some patients with follow up by EUS or transabdominal ultrasound have been excluded, even if some recent data suggest that EUS might be the best correlating imaging tool in pNETs <20 mm [28]. Some patients who had irregular or missing external follow up by general practitioners were excluded. Patients with advanced tumor disease other than pNET were not taken into account, especially when life expectancy was significantly limited, and surveillance was focused on another predominant disease. Chromogranin A (CgA) levels were not reported in this study, as their routine use is of limited importance in cases of small NET lesions, and information about CgA was only available for a small number of the included patients. Somatostatin receptor-based imaging studies have not been performed routinely.

A strong point of this study was the histological confirmation of all lesions. Recent studies report adequate tissue sampling during EUS-FNB in 86%. Even if the high variability of Ki-67 in different areas of pNETs impedes an accurate assessment of tumor grading by EUS-fine-needle aspiration/EUS-FNB, there is some evidence that concordance between preoperative and postoperative histological analysis is acceptable, estimated at 80%. Especially for lesions <2 cm and biopsies performed in the recent era the concordance reaches >90% [29]. However, the data reported herein did not allow us to investigate the reliability of EUS-FNB, as there was a long delay between the first EUS-FNB and secondary resection in the observation group, and only a few patients in the upfront surgery group had preoperative EUS-FNB.

In conclusion, the behavior of small nf pNETs is difficult to predict, and there is growing evidence for malignant behavior in a subgroup of patients, even after surgical treatment. Optimal management remains challenging, as pancreatic surgery is associated with significant morbidity.