Background
In this work, we evaluated the epidemiology and the change of tendency in recent years of non-functional pancreatic neuroendocrine carcinoma at a population level. And use the APC (annual percentage changes) to quantize this change as well as set forth its significance. With advancements diagnosis and treatment, such as improvements in imaging technology, surgical techniques, and adjuvant therapies, overall survival and trends in incidence and mortality for NF-pNETs may change. Our study used annual percent change (APC) to evaluate the incidence and incidence-based (IB) mortality trends as well as the prognostic factors in recent years (2000–2014) for NF-pNETs using a population identified through the SEER Programme database in the United States (US).
Pancreatic neuroendocrine tumour (pNET) is a heterogeneous tumour derived from peptide neurons and neuroendocrine cells [
1] that is increasing in incidence year by year. These tumours are more indolent than pancreatic adenocarcinoma [
1] and are considered to confer better survival outcomes than pancreatic adenocarcinoma. However, patients with metastatic disease are known to have a poor prognosis [
2]. pNETs demonstrate a certain degree of malignancy [
3] and have an important status in pancreatic carcinoma.
pNETs can be divided into functional (F-pNET) or non-functional (NF-pNET) pNETs depending on their ability to secrete biologically active hormones and to cause characteristic symptoms [
4]. NF-pNETs represent the following three types: NF-pNETs that not produce hormones; NF-PNETs that produce hormones at a low enough level to not cause symptoms; and NF-PNETs that produce hormones such as pancreatic polypeptide, chromogranin A, ghrelin, calcitonin or neurotensin that do not cause symptoms [
4].
NF-pNETs are typically indolent and accidentally discovered in most patients [
5]. There are no good diagnostic methods to identify this potentially malignant cancer in early stages [
3], and this disease is usually diagnosed by pathologic examination after operation or metastasis of the disease itself [
6]. Chromogranin A and pancreatic polypeptide (PP) [
3,
7] can be useful for diagnosis; nevertheless, NF-pNETs are typically diagnosed at advanced stages because of their indolent nature and slow growth, which cause a delay in symptoms [
8,
9].
Regardless of metastatic disease, surgical resection of NF-pNETs is considered beneficial for overall survival [
8,
10‐
12]. The predictors of survival include tumour grade, disease stage, tumour size, histopathological type, and lymph node status [
10,
11,
13]. In particular, tumour grade is an important predictor of survival [
14,
15].
NF-pNETs account for approximately 2% of all pancreatic malignancies [
11]. In addition, there are very few studies, especially population-based studies, on the epidemiology and natural history of NF-pNETs. This might be because of the rarity of the disease and the complexity in its classification, as well as the lack of understanding of clinical and prognostic features of NF-pNETs.
Although there are some studies based on the Surveillance, Epidemiology, and End Results (SEER) Programme using the International Classification of Disease for Oncology (ICD-O)-3 codes (8013/3, 8246/3, and 8150/3), they had some deficiencies. Such as the study by AJ Moser [
16], who found that enucleation is equivalent to resection and identified enucleation as a factor contributing to survival benefit. However, with advancements diagnosis and treatment, such as improvements in imaging technology, surgical techniques, and adjuvant therapies, overall survival and trends in incidence and mortality for NF-pNETs may change. Our study used annual percent change (APC) to evaluate the incidence and incidence-based (IB) mortality trends as well as the prognostic factors in recent years (2000–2014) for NF-pNETs using a population identified through the SEER Programme database in the United States (US).
Methods
Data source
SEER is an authoritative source of information on cancer incidence and survival in the United States. SEER currently collects and publishes cancer incidence and survival data from population-based cancer registries covering approximately 28% of the U.S. population and is maintained by the National Cancer Institute. The SEER Program is the only comprehensive source of population-based information in the United States that includes stage of cancer at the time of diagnosis, incidence and survival data. The mortality data reported by SEER are provided by the National Center for Health Statistics. The geographic areas of the data was based on the reporting system and population difference. The population data used in calculating cancer rates has through appropriate criteria before abstrcation.
We used a retrospective cohort study to the data from the SEER database, which is in term of the November 2016 submission. Data is obtained from 2000 to 2014 from the SEER 18 registries (San Francisco-Oakland SMSA, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, Atlanta, San Jose-Monterey, Los Angeles, Alaska Natives, Rural Georgia, California excluding SF/SJM/LA, Kentucky, Louisiana, Greater Georgia and New Jersey with adjustment for the areas impacted by Hurricanes Katrina and Rita).
The SEER Programme registries routinely collect data on patient demographics, cancer features and cancer-associated treatment. The SEER database has coded surgery intervention as a variable, which mentions whether no surgery or resection is performed. However, details of chemotherapy are not included in this database.
Study population
We used the SEER database for descriptive analysis and determination of incidence trends. For descriptive analysis, in the SEER database, we used ICD-O-3 codes (8013/3, 8150/3, and 8240/3–8249/3) and site codes (C25.0-C25.4 and C25.7-C25.9) from 2000 to 2014. Patients who were diagnosed within 1 month before death (diagnosis reported on the death certificate or diagnosed at autopsy) were included in the analysis of incidence trends. However, these patients were excluded from survival analyses because survival time in the SEER database is calculated in months and not days, and if we used these data, these cases would be treated to have a survival duration of zero. In addition, we excluded patients who were diagnosed with other primary malignant tumours because we wanted to minimize the chances of cases misdiagnosed as NF-pNET because of metastatic disease to the pancreas. Furthermore, we excluded patients who died due to causes other than NF-pNETs because unrelated causes of death were precluded. The SEER historic stages were used for disease stage classification in our study instead of the American Joint Committee on Cancer staging system, as we did not have access to the latter during our study period. The SEER stage classification provides consistent definitions over time and are as follows: localized (the tumour is confined to the primary site); regional (the tumour has spread to regional lymph nodes); and distant (the cancer has metastasized).
Statistical analysis
We used SEER*Stat software (version 8.3.4) to analyse our data, including incidence, IB mortality, and survival. All rates were age adjusted according to the 2000 US standard population. Standard mortality statistics were not available because the death certificate does not include the histology of cancer. We obtained IB mortality data by associating the features of the incident cancer to the information on the death certificate. IB mortality enables the stratification of mortality by variables associated with cancer onset.
IB mortality were used. Because in a selected year, NF-pNETs IB mortality is a proportion of the total number of deaths caused by NF-pNETs. These people, defined as those who died of NF-pNETs, must be those who had earlier been diagnosed with NF-pNETs, rather than being diagnosed with NF-pNETs simply because of an autopsy.
We used the Joinpoint Regression Analysis program (version 4.5.0.1) from NCI (
https://surveillance.cancer.gov/joinpoint/) to examine trends in incidence and IB mortality for NF-pNET. This program can calculate incidence and IB mortality using a model that is a segmented in a log-linear form. The program can calculate APC in age-adjusted incidence and IB mortality in each segment and present 95% confidence interval (CI) values. The software can represent the slope of the curve in the calculated model, and a linear regression model can be fitted to the last line segment when the incidence is predicted to be stabilised.
For survival analysis, we divided the population into three subgroups by the year of diagnosis as follows: 2000–2004, 2005–2009, and 2010–2014. Median survival and survival rates were calculated for the whole group and the four subgroups. The P value reported for the analysis of survival trends is for all the three subgroups. Additionally, we performed survival analysis based on treatment and stage at diagnosis. We used the Kaplan-Meier method to calculate the cumulative survival rates and the log-rank test to compare the survival curves.
Independent predictors of mortality were determined by Cox proportional hazard regression. The covariates analysed included patient age, sex, and ethnicity, tumour grade, stage, size and site and treatment. All P values were two tailed, and P < 0.05 was considered statistically significant for all tests. Stata (version 11.0; StataCorp LP) was used for all survival and other analyses.
Discussion
NF-pNETs account for a certain proportion of pancreatic tumours [
8]. Some pathological types of NF-pNETs are also fatal [
14,
17,
18]. Therefore, the epidemiology, treatment and prognosis of the disease are worth studying. RR Salem [
19] have previously studied the incidence and predictors of survival for small NF-pNETs (< 2 cm). Zhou H [
20] explored the impact of ethnicity on this disease; AJ Moser [
16] reported the epidemiological and prognostic factors of NF-pNET from 1973 to 2004; in addition, many studies have focused on the incidence of NF-pNETs but lack data on annual changes. It is important to improve our understanding of the epidemiology, natural history and prognostic factors associated with this disease.
To the best of our knowledge, this is the largest population-based study using the SEER database from 2000 to 2014 focused on the incidence and prognosis of NF-pNET in the US. An updated evaluation of the incidence and prognosis, as well as trends in increasing incidence and improved survival, was presented over time.
This study demonstrated that the incidence of NF-pNET is increasing in the US population. In recent years (from 2009 to 2014), the rate of increase in incidence has been significantly high. The acceleration of increase in the incidence of localized disease and well-differentiated tumours without changes in biological behaviours may be indicative of recent improvements in detecting pNETs [
21‐
23] such as the more frequent use of axial imaging and endoscopic ultrasound and good understanding and classification [
11] of NF-pNETs. However, although we demonstrated an acceleration of increase in the incidence of NF-pNETs from 2009 to 2014, the IB mortality presented an deceleration of increase trends in recent years. The deceleration of increase trends in IB mortality was also observed for distant disease. In addition, the median survival over time and all stages is increasing, indicated that a progressive increase in the detection of NF-pNET and appropriate therapy in recent years have led to a better response, resulting in good prognoses. Our finding of deceleration of increase trends IB mortality for NF-pNET is novel, contrary to the improved survival demonstrated in many studies [
16,
19].
The higher rate of increase in the incidence of localized disease than distant disease is indicative of progressive improvements in the detection of NF-pNETs at an early stage and recent advancements in the treatment of distant disease. Additionally, the IB mortality for distant disease decelerate in increase trend. However, the rate of surgery in patients with distant disease significantly decreased during the study period, indicating that recent emergence of curative intervention with non-surgical treatments [
3,
24], such as targeted therapy, chemotherapy, and radiotherapy, has improved the survival of patients with distant disease.
In this study, the increase in incidence was the highest in recent years, and African-American patients displayed the highest incidence. In contrast, both African-American and Caucasian patients displayed a deceleration in increase trend in IB mortality. The possible reasons may be the development of technologies that increased the detection of the disease so that it could be treated at an earlier stage; additionally, early treatment of the disease is more effective than later treatment.
The site of pNETs and NF-pNETs is a known factor that is essential for correct patient management [
25], and it is also an important predictor of outcomes. Despite a steady increase in incidence, trend of the IB mortality for patients with tumours in the tail of the pancreas decelerated in recent years, and this tumour location presented the best outcome, indicating that treatment of tumours located in the tail of the pancreas is worth affirmation. On the other hand, tumours located in the head of the pancreas had the worst outcome, incidence and IB mortality, and tumours in other sites except tail did not present a favourable outcome either, indicating that the treatment of NF-pNETs in sites other than the tail have not displayed ideal therapeutic effects in recent years, and better treatment adjustment is needed to improve prognosis.
In our study, tumour size> 2 cm was independently associated with an increased risk of death. Yoji Kishi [
26] found that for tumour size> 1.5 cm, the risk of metastasis and recurrence significantly increases, but for tumour size< 1.5 cm, there is little risk of metastasis and recurrence. Eric J [
19] found that nodal metastasis occurs in a small proportion of small NF-pNETs (< 2 cm). Thus, consistent with other studies, small tumour size is associated with better survival [
3,
17,
19,
26,
27].
Based on our results, patients undergoing surgery for primary and metastatic tumours had a prolonged survival. Although many factors have been included in our study, younger patients, patients with well- or moderately differentiated tumours and patients with favourable tumour location were more inclined to undergo surgery. It has also been confirmed that surgery can benefit to patients with distant disease [
25].
Nevertheless, our study has several limitations. The SEER database does not provide information on patient comorbidities and clinical details such as biliary obstruction, which may affect prognosis. In any case, we focused on all-cause mortality, and our results may not have been influenced by this limitation. There are many differences between mortality and survival; however, bias in survival analysis does not affect mortality analysis [
28]. In addition, patient selection bias, miscoding, and incorrect data classification should be considered. However, it has been verified that misclassification is very minimal in the SEER database [
29‐
31]. Furthermore, we did not include detailed information on adjuvant treatment such as chemotherapy. Finally, data on tumour size was only obtained after 2004.
Conclusions
The diagnosis and treatment of NF-pNET are changing in the US, with favourable results based on stage, treatment, survival and mortality. Although the incidence of this disease has increased in recent years, IB mortality has a deceleration in increase trend, indicating recent improvements in detecting NF-pNETs on one hand and better treatment of this disease on the other. Even for cases with distant disease, non-surgical treatments have improved prognosis. Moreover, in contrast to that for tumours in the pancreatic head, the treatment for tumours the pancreatic tail has improved, suggesting that there is room for further improvement of survival for patients with tumours in the head of the pancreas. Therefore, early diagnosis and appropriate treatment adjustments continue to be a priority in improving patient outcomes.