Real-world treatment outcomes for pancreatic neuroendocrine tumours: a single tertiary referral institute experience

Abstract

Background

Pancreatic neuroendocrine tumours (PNETs) are rare neoplasms of the pancreas. Real-world treatment outcomes in Asia are seldom reported. The aim of this study was to provide the real-world treatment outcomes of PNETs at a single tertiary cancer centre in Hong Kong over an 18-year period.

Materials and methods

This was a retrospective cohort study that recruits patients with local or metastatic PNETs treated at Prince of Wales Hospital, Hong Kong between 2005 and 2023. Five-year overall survival (OS), progression-free survival (PFS), recurrence-free survival (RFS) of the surgical cohort, and survival outcomes in patients with metastatic disease were summarized.

Results

A total of 98 patients were recruited (median age: 59 years, 51% male). Among them, 18 patients (18.4%) had functioning tumours, of which majority were insulinoma (72.2%). Most tumours (64.8%) were classified as grade 1 disease. More than 80% of patients had early-stage disease on presentation, with only 14 patients presented with stage IV disease. Five-year OS for the entire cohort was 88.4% and the 5-year PFS was 73.7%. Older age [hazard ratio (HR) 1.06, 95% confidence interval (CI) 1.00-1.13, P = 0.04] and larger tumour size (HR 1.20, 95% CI 1.00-1.44, P = 0.05) were associated with worse OS, whereas patients with metastatic disease had a shorter PFS (HR 7.95, 95% CI 2.55-24.8, P < 0.001). In the cohort treated with curative surgery, metastatic disease (HR 7.49, 95% CI 1.40-40.0, P = 0.018) and tumour grade (HR 8.03, 95% CI 1.58-40.8, P = 0.012) were associated with worse RFS. Microscopic margin involvement did not influence RFS in patients treated with curative surgery. In the 20 patients who received systemic therapy, everolimus showed a trend towards improved PFS compared with chemotherapy (5-year PFS: 46.9% versus 16.7%).

Conclusion

Patients with PNETs in Hong Kong have early presentation and excellent prognosis with most patients being eligible for curative surgery reaching 90% of 5-year OS.

Highlights

• •Five-year OS for PNET is excellent, at nearly 90%.
• •Following curative surgery, metastatic disease and tumour grade were associated with shorter PFS.
• •Microscopic margin involvement was not associated with RFS.

Introduction

Pancreatic neuroendocrine tumours (PNETs) are a heterogeneous group of epithelial neoplasms that represent ∼5% of all pancreatic tumours.^1,2 Despite being much less common than ductal adenocarcinoma of pancreas, increasing incidence of PNETs have been observed, predominantly due to the increasing use of cross-sectional imaging.³ PNETs can be classified as either functional or non-functional, depending on whether they secrete a hormone that contributes to a clinical syndrome.^4,5 Because the majority of PNETs (60%-90%) are non-functional (i.e. not associated with hormone secretions), some patients may only present to the clinic with very advanced disease and thus not suitable for curative therapy.⁶

Treatment strategies for PNETs have not undergone major changes in the past two decades. Surgery is the mainstay of treatment for patients with localized disease, although observation could be an option for small (<2 cm) and non-functioning tumours.⁵ Patients with functioning disease are recommended to undergo surgery as it can cure both the disease and its related hormone secretion-related clinical symptoms.⁴ For unresectable disease, there are various systemic treatment options, including somatostatin analogue (SSA), chemotherapy, targeted therapy and peptide receptor radionuclide therapy depending on the tumour burden and the aggressiveness of disease.⁷

Although guidelines exist for the management of localized and unresectable PNETs, differences exist in patient selection and treatment sequence.^4,5,7, 8, 9 There is also a paucity of studies reporting the real-world outcome in the management of PNETs, especially in the Asian population. In this study, we present real-world data regarding the clinical characteristics and treatment outcome of Hong Kong Chinese patients with PNETs who were managed in a tertiary cancer centre in Hong Kong.

Materials and methods

Design

This is a retrospective, single-centre, observational study examining the treatment outcomes of patients with PNETs at the Prince of Wales Hospital in Hong Kong, a tertiary referral cancer centre serving a catchment population of ∼1 million people. The centre features a multidisciplinary neuroendocrine tumour clinic and tumour board. In addition to its local population, the centre receives referrals for complicated cases of neuroendocrine tumours from other institutions in Hong Kong. The diagnosis and management of PNETs were decided by a team consisting of oncologists, surgeons, pathologists, endocrinologists and radiologists.

Eligibility criteria

In this study, patients with a histologically confirmed diagnosis of PNETs from 1 January 2005 to 31 December 2023 were recruited. These patients were identified by a pathologist with a fellowship qualification (HHWL) in the pathology archives at the Department of Anatomical and Cellular Pathology. Patients with a pathological diagnosis of pancreatic neuroendocrine carcinoma were excluded from the study. The clinical notes of each patient were accessed via the clinical management system of the hospital authority, which is the statutory body managing all public hospitals in Hong Kong. Patients’ demographics, presenting symptoms, biochemical tests, pathology reports, imaging results, treatment progress and survival data were collected. Typically, for patients who received surgery, they will be followed up at an interval of 3-6 months in the first 3 years, and yearly afterwards. Imaging for surveillance is typically arranged in the first 5 years at 3- to 12-month intervals. For patients on systemic treatment, they will be followed up typically 4-6 weeks during treatment, with imaging carried out at an interval between 3 and 6 months. For first-line systemic therapies, patients were generally treated with everolimus before 2017. After 2017, emerging data at that time supported the use of chemotherapy such as temozolomide and/or capecitabine to treat low to intermediate disease. The choice between everolimus and chemotherapy was mainly due to affordability and aggressiveness of the disease as indicated by the grade and Ki-67 index.

The data cut-off was made on 31 December 2024. Ethics approval was obtained from the Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (reference number: CREC 2024.003). Patient consent was waived by the ethics committee due to the retrospective nature of the study.

Outcomes

The primary objective was to access the 5-year overall survival (OS) rates for PNET patients. The current study explored other outcomes including progression-free survival (PFS) based on RECIST 1.1 of the entire cohort, recurrence-free survival (RFS) of the surgically treated cohort and the survival outcomes for patients treated with systemic therapy.

Statistical analyses

Baseline patients’ demographics are presented in tables. Student’s t-test was used to compare continuous variables and the chi-square test (or Fisher’s exact test when appropriate) was used to compare categorical variables. Survival outcomes were estimated by the Kaplan–Meier method. For OS, patients were censored if they did not experience death at 10 years. For PFS or RFS, patients were censored at 5 years if they did not experience progression, recurrence or death. Predictors of survival outcome were assessed by univariate and multivariable Cox proportional hazard models. Variables with P value < 0.1 in the univariate analysis were included in the multivariable model. Statistical significance was defined as a two-sided P value < 0.05. All statistical analyses were carried out using R statistical software (v4.2.0; R Core Team 2022).¹⁰

Results

Patients

Ninety-eight patients were recruited in the study (Table 1). There were 50 males and 48 females. The median age of diagnosis was 59 years [interquartile range (IQR) 51-65 years]. Fifty-five patients (56%) presented with symptoms that led to diagnosis of PNET, with the top three most common symptoms being the presence of epigastric pain (25.5%), hypoglycaemia (13.3%) and weight loss (5.1%). Majority of patients had non-functional PNET (81.6%). There were 18 patients (18.4%) with functioning PNET, in which 13 of them (72.2%) were insulinoma. The median size of tumour was 2.4 cm (IQR 1.3-3.8 cm). Most tumours arose from the body of the pancreas (61.2%), followed by the head of the pancreas (30.6%). Lymph node involvement was present in seven patients (7.1%). More than 80% of patients had stage I or II disease on presentation, and ∼14.2% of patients had stage IV disease on presentation. Seven patients did not have histological grade of PNET assessed due to insufficient quantity of tissue for further analyses beyond the initial histological diagnosis. In the remaining 91 patients, grade 1, 2 and 3 PNETs were observed in 64.8%, 34.1% and 1.1% of cases, respectively. Most patients underwent surgery (77.6%) as their first treatment options; nine patients (9.2%) opted for observation as the primary treatment (Table 1). There were no significant difference in patient characteristics between those with functioning and non-functioning tumours (Supplementary Table S1, available at https://doi.org/10.1016/j.esmogo.2025.100293sw).

Table 1.

Baseline characteristics of the entire cohort

Clinical characteristics	N = 98
Age, years (range)	59 (51-65)
Gender, n (%)
Male	50 (51)
Female	48 (49)
ECOG PS, n (%)
0	84 (86)
≥1	14 (14)
Symptomatic presentation, n (%)	55 (56)
Epigastric pain	25 (26)
Hypoglycaemia	13 (13)
Weight loss	5 (5)
Functionality, n (%)
Non-functional	80 (82)
Gastrinoma	1 (1)
Glucagonoma	2 (2)
Insulinoma	13 (13)
Carcinoid syndrome	2 (2)
Location, n (%)
Head/neck	30 (31)
Body	60 (61)
Tail	8 (8)
Size, cm, median (IQR)	2.4 (1.3-3.8)
<2, n (%)	37 (38)
2-4, n (%)	27 (27)
≥4, n (%)	34 (35)
Lymph node involvement, n (%)	7 (7)
Grade, n (%)
1	59 (60)
2	31 (32)
3	1 (1)
Missing	7 (7)
Stage (AJCC eighth edition), n (%)
I	36 (37)
II	44 (45)
III	4 (4)
IV	14 (14)
Baseline Ca19-9, U/mL, median (IQR)a	13 (7-28)
Baseline CgA, μg/L, median (IQR)a	106 (58.2-375.5)
First-line treatment, n (%)
Observation	9 (9)
Surgery	76 (78)
Systemic treatment	9 (9)
Ablation	4 (4)
Any curative first-line treatment, n (%)	76 (78)

AJCC, American Joint Committee on Cancer; CgA, chromogranin A; ECOG PS, Eastern Cooperative Oncology Group performance status; IQR, interquartile range.

^aForty-one patients had baseline Ca19-9 available and 20 patients had baseline CgA available.

Outcomes

For the entire cohort, the 5-year OS rate was 88.4% with a total of nine deaths observed (Figure 1A). The 5-year OS rates by stage were 93.3%, 88.3%, 100% and 75.7%, respectively, for stage I, II, III and IV disease (Figure 1B). The 5-year OS rate by grade was 92.3% for grade 1 disease and 82.7% for grade 2 disease, respectively (Figure 1C). The 5-year OS rate was 86.1% for non-functioning tumours and 100% for functioning tumours (Figure 1D). In terms of PFS, the 5-year PFS rate was 72.2% for the entire cohort, with 22 patients experiencing progression or death (Figure 2A). The 5-year PFS rates by stage were 89.6%, 81.9%, 25.0% and 23.1%, respectively, for stage I, II, III and IV disease (Figure 2B). The 5-year PFS rates for grade 1 and grade 2 disease were 89.1% and 42.9%, respectively (Figure 2C). The 5-year PFS rate was 74.8% for non-functioning tumours and 58.4% for functioning tumours (Figure 2D).

Figure 1.

Overall survival in PNET patients. (A) Overall survival in PNET patients for the entire cohort. (B) Overall survival in PNET patients by stage for the entire cohort. (C) Overall survival in PNET patients by grade for the entire cohort. (D) Overall survival in PNET patients by functionality for the entire cohort. PNET, pancreatic neuroendocrine tumour.

Figure 2.

Progression-free survival in PNET patients. (A) Progression-free survival in PNET patients for the entire cohort. (B) Progression-free survival in PNET patients by stage for the entire cohort. (C) Progression-free survival for PNET patients by grade for the entire cohort. (D) Progression-free survival for PNET patients by functionality for the entire cohort. PNET, pancreatic neuroendocrine tumour.

Univariate and multivariate analysis

In the univariate analysis, age, tumour size and Eastern Cooperative Oncology Group performance status (ECOG PS) were associated with OS. Functionality of tumours was not included in the analysis because none of the patients with functional tumours experienced death. After multivariate adjustments, only age [hazard ratio (HR) 1.06, 95% confidence interval (CI) 1.00-1.13, P = 0.04] and tumour size (HR 1.20, 95% CI 1.00-1.44, P = 0.05) were independently associated with worse outcome of OS (Table 2). For PFS, tumour size, ECOG PS, metastatic disease and tumour grade were associated with PFS in the univariate analysis. After multivariate adjustment, metastatic disease (HR 7.95, 95% CI 2.55-24.8, P < 0.001) remained as the only independent parameter associated with disease progression (Table 3).

Table 2.

Univariate and multivariable analysis for predictors of overall survival for the entire cohort (n = 98)

	Univariate		Multivariate
	HR (95% CI)	P value	HR (95% CI)	P value
Age	1.07 (1.01-1.14)	0.032	1.06 (1.00-1.13)	0.038
Male gender	0.79 (0.23-2.74)	0.709	—	—
Size	1.14 (1.01-1.30)	0.037	1.20 (1.00-1.44)	0.049
ECOG PS ≥1	7.25 (2.08-25.2)	0.002	3.53 (0.87-14.4)	0.078
Metastatic disease	3.00 (0.85-10.7)	0.089	1.41 (0.35-5.74)	0.630
Grade ≥2	1.56 (0.42-5.83)	0.505	—	—

Bold indicates P-value < 0.05.

CI, confidence interval; ECOG PS, Eastern Cooperative Oncology Group performance status; HR, hazard ratio.

Table 3.

Univariate and multivariable analysis for predictors of progression-free survival for the entire cohort (n = 98)

	Univariate		Multivariate
	HR (95% CI)	P value	HR (95% CI)	P value
Age	1.02 (0.98-1.06)	0.414	—	—
Male gender	0.83 (0.36-1.92)	0.670	—	—
Size	1.13 (1.04-1.24)	0.006	0.98 (0.84-1.13)	0.746
ECOG PS ≥1	3.35 (1.36-8.24)	0.009	1.18 (0.42-3.29)	0.749
Metastatic disease	10.03 (4.25-23.6)	<0.001	7.95 (2.55-24.8)	<0.001
Grade ≥2	5.48 (2.10-14.3)	<0.001	2.22 (0.64-7.65)	0.206
Functional	1.52 (0.56-4.11)	0.414	—	—

Bold indicates P-value < 0.05.

CI, confidence interval; ECOG PS, Eastern Cooperative Oncology Group performance status; HR, hazard ratio.

Subgroup analyses

Seventy-four patients received surgery of curative intent as the primary treatment. Their baseline characteristics are listed in Supplementary Table S2. The 5-year RFS is 82.1%, with 11 recurrences or deaths observed (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmogo.2025.100293sw). The 5-year RFS rates by grade were 96.0% and 51.4%, respectively (Supplementary Figure S2, available at https://doi.org/10.1016/j.esmogo.2025.100293sw). Sixty patients (81.1%) had a clear margin, 13 patients (17.6%) had a positive margin and 1 patient (1.4%) had an unknown margin status. Of the 13 patients with positive margins, 2 of them received adjuvant therapy with 1 receiving radiotherapy and the other 1 receiving chemotherapy, respectively. Four out of the 13 patients with a positive margin developed recurrences, including the two patients who received adjuvant therapy. All of these recurrences occurred in patients with grade 2 or 3 disease. There was one local recurrence and two distant recurrences, and one patient had both local and distant recurrence. The 5-year RFS rates by margin status were 84.0% for patients with a clear margin and 74.0% for patients with a positive margin (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmogo.2025.100293sw). In the surgical cohort, metastatic disease and tumour grade are predictors of recurrence in the univariate model. These two variables remain statistically significant in the multivariable model (metastatic disease: HR 7.49, 95% CI 1.40-40.0, P = 0.018; tumour grade: HR 8.03, 95% CI 1.58-40.8, P = 0.012) (Supplementary Table S3, available at https://doi.org/10.1016/j.esmogo.2025.100293sw).

Twenty patients received systemic treatment in the entire cohort. Their baseline characteristics are listed in Supplementary Table S4, available at https://doi.org/10.1016/j.esmogo.2025.100293sw. Eleven patients received systemic treatment due to de novo metastatic disease and nine patients received systemic treatment due to unresectable recurrences. Among them, 11 patients had liver metastases. Three had grade 1 disease, and the respective figures for grade 2 and 3 diseases were 16 and 1, respectively. Seven patients received everolimus with or without concurrent SSA as first-line treatment, and eight patients received chemotherapy as first-line treatment (four capecitabine plus temozolomide, two capecitabine alone, one temozolomide alone and one etoposide plus cisplatin). The 5-year OS rates for patients receiving first-line everolimus and chemotherapy were 75.0% and 68.6%, respectively (Supplementary Figure S4, available at https://doi.org/10.1016/j.esmogo.2025.100293sw). The 5-year PFS rates for patients receiving first-line everolimus and chemotherapy treatment were 46.9% and 16.7%, respectively (Supplementary Figure S5, available at https://doi.org/10.1016/j.esmogo.2025.100293sw).

Discussion

To our knowledge, this is the first comprehensive report on the treatment outcomes of patients specifically with PNETs in Hong Kong. One of the major strengths of our study is the high accuracy and completeness of our database. Every patient included in the study had a histologically confirmed diagnosis. Consequently, the distribution of PNET subcategories mirrors the true epidemiological landscape of the region. These important statistics are often absent from existing studies. Our study has comprehensively addressed them in our report, thereby closing an important gap in the literature.

Firstly, we observed majority of patients with PNETs had early presentation, which is similar to population cohorts in both Western^11,12 and Asian regions.¹³ The indolent nature and early presentation of PNETs allow many patients to receive curative surgery, achieving a remarkable 5-year OS of >80% in our cohort and consistent with other modern cohorts.^11,13 This is in contrast with patients suffering from pancreatic ductal adenocarcinoma (PDAC), in which majority of patients present with locally advanced or metastatic disease.¹⁴ For example, according to a SEER database study of patients with PDAC diagnosed between 1973 and 2014, only 7.3% of patients presented with localized disease.¹⁴ According to a nationwide population-based study in Korea including 78 920 PDAC patients between 2006 and 2019, only ∼20% of patients received surgery.¹⁵ Notably, even in patients who received curative surgery, the 5-year OS was still poor for PDAC at ∼15%-30% in both Western and Asian populations.15, 16, 17 This highlights the importance of clarifying the histology for pancreatic lesions as it provides important prognostic information that could be drastically different.

Secondly, we observed that >80% of patients had non-functioning PNETs, while 18% of patients had functioning tumours. This is in line with our previous publication on gastroenteropancreatic neuroendocrine tumour in that ∼91.3% were non-functional.¹⁸ In other population-based studies such as those in Germany or Japan, non-functional PNETs are also the predominant subtype, representing >90% of the total PNETs.^11,19 Insulinoma being the predominant type is also a finding that aligns with the literature.^7,11 We further showed that functionality of tumours does not impact the survival outcomes of patients. In our study, patients with functioning tumours had a numerically worse PFS at 5 years but this was primarily due to higher proportion of patients with functioning tumours presented with metastatic disease, as shown in the multivariate analysis (Table 3). None the less, OS was not affected by whether patients have a functioning or non-functioning tumour. This is in contrast with the classical view that non-functional tumours tend to carry worse survival due to late presentation.²⁰ One possible explanation is the increasing availability of imaging which resulted in stage migration and more non-functioning tumours being picked up at an earlier stage.³ Our study showed that the baseline characteristics of patients with functioning and non-functioning tumours were not statistically significantly different. Moreover, 37% of patients in our study had small tumours with a size <2 cm, and the median size of tumours was 2.4 cm, which is smaller than some of the older surgical series published previously.21, 22, 23

Thirdly, we showed that having a positive microscopic margin in a surgical specimen is not predictive of recurrence for patients who received curative surgery. This is an important message as patients with a positive margin might be subjected to additional adjuvant therapy as they would be classified as high-risk patients. These patients are recommended to consider adjuvant treatment in international guidelines, but no standardized treatment strategy is recommended.^5,9 In fact, our study showed that none of the patients with grade 1 disease and a positive margin developed recurrence. Instead, we showed that patients with grade 2 disease or de novo metastatic disease are at high risk of recurrence, even after curative intent surgery (Supplementary Table S3, available at https://doi.org/10.1016/j.esmogo.2025.100293sw). This highlights an unmet clinical need, and clinical trials on adjuvant therapy should be conducted targeting this group of patients with elevated risk of recurrence.

Finally, our study demonstrated that patients with unresectable PNETs derived a trend towards a better PFS benefit with everolimus over chemotherapy (mainly capecitabine ± temozolomide). Even after excluding the two patients with grade 1 tumour treated with everolimus, patients treated with everolimus still showed an improved PFS compared with chemotherapy (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmogo.2025.100293sw). Everolimus is an inhibitor of mammalian target of rapamycin (mTOR), a serine-threonine kinase that stimulates cell growth, proliferation and angiogenesis.²⁴ Studies on molecular pathogenesis have shown that mutations along the PTEN–AKT–mTOR pathway can occur in up to 60% of PNETs, making everolimus a suitable drug to suppress oncogenic signalling due to the activation of this pathway.²⁵ In the international, phase III, randomized RADIANT-3 study, patients with advanced, progressive, low- or intermediate-grade PNETs were randomly allocated to receive everolimus or best supportive care. Everolimus demonstrated improved PFS at 11 months compared with 4.6 months in the placebo arm.²⁶ In the real-world setting, everolimus was also supported as the most effective treatment compared with other therapies in PNETs in a recent network meta-analysis carried out by the Cochrane Database of Systemic Review.²⁷ On the contrary, high-level evidence on systemic chemotherapy in advanced PNET is limited and mostly came from small prospective studies and real-world experiences.^7,28 The ECOG study was the only prospective, randomized study that examined the efficacy of temozolomide with or without capecitabine in patients with advanced, low- to intermediate-grade PNETs. In the 144 patients who were randomly allocated to the study, patients who received temozolomide and capecitabine (CAPTEM) achieved an improved median PFS of 22.7 months, compared with temozolomide alone at 14.4 months (HR 0.58, P = 0.22). The OS also showed a trend towards an improvement with CAPTEM at 58.7 months compared with 53.8 months with temozolomide alone (HR 0.82, P = 0.42). These data provided support for the use of CAPTEM in advanced, low- to intermediate-grade PNETs.²⁹ The recently published SEQTOR trial, which is a prospective, randomized, phase III study that aimed to determine the optimal sequence in patients with advanced PNETs, showed that the sequence of everolimus followed by chemotherapy streptozocin plus 5-fluorouracil (5-FU) performed better in patients with grade 1 tumours in terms of PFS and OS compared with those treated with streptozocin plus 5-FU followed by everolimus.³⁰ First treatment with streptozocin plus 5-FU only performed better in response rate in patients with grade 2 tumour.³⁰

Our study has several limitations. Firstly, there is no study protocol to guide the frequency of imaging hence the PFS or RFS may have been confounded. However, PNETs are generally slow-growing tumours, and our clinic has the routine of scheduling imaging approximately every half-year, so this bias is expected to be small. Secondly, the sample size of our cohort is small and some of the patients may have insufficient follow-up time to experience events such as death for an indolent disease like PNET, so some of the results should be interpreted with caution. For instance, metastatic disease was a clinically significant adverse prognostic factor for OS in the univariate analysis only but not in the multivariate analysis. This could be attributable to the small sample size (14 patients had metastatic disease) with a small number of events (5 out of 14 patients experienced death). This highlights the challenges in studying PNET and the future need for multicentre collaboration in collecting and analysing real-world treatment outcomes.

In conclusion, in our single-institution experience of nearly 20 years, we showed that patients with PNETs have excellent prognosis. Majority of the tumours presented at the early stage and were amenable to curative treatments. Patients with advanced disease were treated with various systemic treatments, but first-line everolimus may confer a better PFS compared with chemotherapy. These findings should be further confirmed in multicentre studies in the future.