Key words
pancreas - neuroendocrine tumour - bone metastases
Introduction
Pancreatic neuroendocrine tumors (PanNETs) are rare neoplasms with an increasing
annual incidence of 0.48/100 000 [1].
Surgical removal is the only curative therapy. At presentation, a majority of
patients have unresectable disease due to local extension or metastases. Hence, only
palliative therapy can be offered [2]. The
clinical course of differentiated PanNETs is variable. Grading and staging play a
pivotal role in NET prognosis and management. According to the 2017 classification
of the World Health Organization (WHO), well-differentiated PanNETs are classified
as G1, G2, or G3 tumors, depending on the Ki-67 proliferation marker. The presence
of distant metastases of any site corresponds to stage IV, whereas stages I to III
are characterized by increasing locoregional manifestation, with (IIIb) or without
lymphonodular metastases (I–IIIa) [3].
The presence of other factors influencing survival is likely, but these are poorly
understood.
In general, PanNETs can metastasize to any organ, with the liver being most commonly
affected [4]. In contrast to soft tissue
metastases, bone metastases (BM) are particularly challenging because their
detection depends on the type of imaging used. Combining 68Ga-DOTATOC
positron emission tomography (PET) with concurrent contrast-enhanced X-ray computed
tomography (ceCT) is considered the gold standard for the detection of BM with a
sensitivity of 100% and a specificity of 89% [5]. In contrast, ceCT alone, which is
frequently used in neuroendocrine tumor (NET) imaging, has a poor sensitivity of
47% with a specificity of 49% [5]
[6].
To date, the incidence of BM in PanNETs is not well defined. Previous reports came
from highly diverse NET cohorts and imaging methods. For example, a recent analysis
of 14 685 gastrointestinal neuroendocrine neoplasm (GI-NEN) patients enrolled in the
US Surveillance, Epidemiology, and End Results (SEER) database between 1973 and 2015
found a BM rate of 5.7% in stage IV patients, but detection methods were not
disclosed [7]. Similarly, in the Spanish
national NET database, a BM rate of 5.2% in gastro-entero-pancreatic
(GEP)-NEN in general and of 4.3% in PanNET in particular was reported [8]. In contrast, in institutional series from
academic centers, BM rates in NET patients as high as 26.0% were observed,
with varying degrees of tumor differentiation, different primary tumor localizations
and imaging detection methods reported in each study [9]
[10]
[11]. To date, all studies have
suffered from a lack of standardization in terms of tumor characteristics, NET
primary location, grading, staging, clinical course and imaging modality. As a
result, there is a lack of information about the true prevalence of BMs in PanNETs
and uncertainty regarding their potential relevance for NET prognosis and
treatment.
We therefore aimed (1) to determine the true prevalence of BM in a histologically
defined group of differentiated PanNETs using 68Ga-DOTATOC PET/CT
as the gold standard for BM imaging; (2) to investigate the influence of BM on the
course of the disease; and (3) to identify clinical complications arising from
PanNET BM.
Patients and Methods
Patients were identified from our prospective NET database at the European
Neuroendocrine Tumour Society (ENETS) Center of Excellence, Department of
Endocrinology, Diabetes and Metabolism, University Hospital Essen. Eligible patients
included those with histologically confirmed differentiated PanNETs who were treated
at our department between January 2009 and January 2021. All patients underwent
contrast-enhanced 68Ga-DOTATOC PET/CT at initial presentation and
at subsequent follow-up. Patients with incomplete data were excluded from further
analysis. To ensure consistency, scheduling of visits as well as indication for
therapies was determined according to ENETS guidelines by an experienced,
multidisciplinary tumor board (MTB). All staging was performed in-house at our
center. The presence of BM was divided into BM with morphological evidence,
pathological tracer uptake only or morphological evidence combined with pathological
tracer uptake. Skeletal-related events (SREs) were defined as the presence of
pathological fractures, bone surgery, bone radiation and/or metastatic
spinal cord compression.
Data were reported as the number of patients (percentage of the group) for the
categorical data and the median (95% confidence interval (CI)) for
quantitative variables unless otherwise stated. SREs and antiresorptive therapy were
compared using Fisher’s exact test. Overall survival (OS) was computed as
the time from initial diagnosis to death from any cause. Patients who were still
alive were censored at the last visit. OS was estimated using the
Kaplan–Meier method and compared with the results from the log-rank test. A
Cox proportional hazards model was used to calculate hazard ratios and to assess
independent predictors of OS. The tests were two-tailed, and results at
p<0.05 were interpreted as statistically significant. All
statistical analyses were performed using the Statistical Package for the Social
Sciences version 26.0 software program (IBM Corporation, Armonk, NY, USA).
Written informed patient consent and approval for data collection and analysis were
obtained upon admission to our institution. The study was approved by the local
ethics committee (18–8367-BO).
Results
Patient characteristics
Based on our prospective NEN database, 314 consecutive patients with
histologically confirmed differentiated PanNETs who were treated between January
2009 and January 2021 were identified. Of these, 149 patients (47%) were
females, and 165 (53%) were males. The median age of subjects at the
initial diagnosis of PanNET was 54 years (14–85 years, range). The Ki67
index was available for 287 patients. Among these, 98 patients (31%) had
grade 1 tumors, 161 (51%) had grade 2 tumors and 28 (9%) had
differentiated grade 3 tumors according to the World Health Organization (WHO)
2017 criteria [3]. The median follow-up
time of this basic cohort was 44 months (39–49, 95% CI).
All further evaluations refer to the group of stage IV patients. According to the
UICC staging, 171 of 314 subjects (55%) had distant metastasized (stage
IV) PanNETs ([Table 1]). Of all stage IV
patients, 35 patients (21%) had G1-NETs, 104 (61%) had G2 tumors
and 24 (14%) had differentiated G3 tumors. Differentiated PanNETs were
diagnosed in 8 patients (5%) without the availability of a Ki-67 index.
The median age at PanNET diagnosis was 56 years (14–81 years, range).
The median disease duration from initial PanNET diagnosis to first
68Ga-DOTATOC PET/CT was 6 months (3–15 months,
95% CI). Nine patients (5%) had hereditary tumors (multiple
endocrine neoplasia type 1). Twenty patients (12%) had functioning
PanNETs (9 insulinomas, 9 gastrinomas and 2 VIPomas). The median follow-up of
the stage IV cohort was 47 months (40.0–55.0, 95% CI) ([Table 1]).
Table 1 Patient characteristics of the stage IV PanNET
cohort (n=171).
Stage IV PanNET cohort
|
n (%)
|
No. of patients
|
171 (100)
|
Sex
|
Male
|
86 (50.3)
|
Female
|
85 (49.7)
|
Age at PanNET diagnosis (years)
|
56 (14–81, range)
|
Median duration from PanNET diagnosis to initial
68Ga-DOTATOC- PET/CT (months)
|
6.0 (3.0–15.0, 95% CI)
|
Median follow-up (months)
|
47.0 (40.0–55.0, 95% CI)
|
PanNET tumor grade (WHO 2017)
|
G1
|
35 (20.5)
|
G2
|
104 (60.8)
|
G3
|
24 (14.0)
|
unknown
|
8 (4.7)
|
Functioning PanNET
|
20 (11.7)
|
Insulinoma
|
9
|
Gastrinoma
|
9
|
VIPoma
|
2
|
Hereditary tumor syndrome – Multiple endocrine
neoplasia type 1
|
10 (5.9)
|
Age, disease duration and follow-up are given as median, categorical
parameters as absolute and relative frequencies.
Manifestation of bone metastases
BM manifested in 62 of the 171 stage IV patients (36.3%) ([Table 2]). In 49 patients, BMs were
detected on initial 68Ga-DOTATOC PET/CT. An additional 13
patients developed BM during follow-up. The median interval between the initial
diagnosis of PanNET and the first detection of BM was 21.5 months
(9.0–35.0, 95% CI). Ten patients (16.1%) had synchronous
osseous metastasis at the time of PanNET diagnosis, and BM occurred in another
14 patients (22.6%) within the first year after diagnosis. The longest
interval between PanNET diagnosis and manifestation of bone metastases was 20.2
years in a patient with multiple endocrine neoplasia type 1 (MEN1) ([Table 2]).
Table 2 Comparison of metastatic patients with and without
BM.
Stage IV PanNET cohort
|
Patients with BM, n (%)
|
Patients without BM, n (%)
|
No. of patients (%)
|
62 (100)
|
109 (100)
|
Sex
|
Male
|
32 (51.6)
|
54 (49.5)
|
Female
|
30 (48.4)
|
55 (50.5)
|
Age at PanNET diagnosis (years)
|
57 (17–81, range)
|
55 (14–79, range)
|
Time between PanNET diagnosis and initial detection of BM
(months)
|
21.5 (9–35, 95% CI)
|
–
|
Time between PanNET diagnosis and first
68Ga-DOTATOC- PET/CT (months)
|
15.5 (5–27, 95% CI)
|
4 (3–13, 95% CI)
|
Median follow-up (months)
|
44 (31–67, 95% CI)
|
48 (40–56, 95% CI)
|
PanNET tumor grade (WHO 2017)
|
G1
|
10 (16.1)
|
25 (22.9)
|
G2
|
39 (62.9)
|
65 (59.6)
|
G3
|
9 (14.5)
|
15 (13.8)
|
unknown
|
4 ( 6.5)
|
4 ( 3.7)
|
Functioning PanNET
|
7 (11.3)
|
13 (11.9)
|
Insulinoma
|
5
|
4
|
Gastrinoma
|
2
|
7
|
VIPoma
|
0
|
2
|
Hereditary tumor syndrome – Multiple endocrine
neoplasia type 1
|
3 (4.8)
|
7 (6.4)
|
Age, disease duration and follow-up are given as median, categorical
parameters as absolute and relative frequencies.
BM occurred at each grading according to WHO criteria: G1 (10/35;
28.6%), G2 (39/104; 37.5%), and G3 PanNET (9/24;
37.5%). Neither the proliferation marker Ki-67 nor the intrapancreatic
location of the primary tumor (pancreatic head, body, or tail) was associated
with the occurrence of BM ([Fig. 1]).
Fig. 1 Grading of PanNET in patients with and without BM.
Seventeen patients (17/49, 35%) presented initially solely with tracer
uptake of BM on 68Ga-DOTATOC PET, without morphological correlate on
the corresponding ceCT ([Fig. 2]). In 9
of those cases (9/17, 53%), BM became morphologically evident on
ceCT during follow-up. The median time from pathologic tracer uptake to
visualization of a morphologic correlate on ceCT was 8 months (4–13
months, range).
Fig. 2 Presentation of BM on initial 68Ga-DOTATOC
PET/CT (n=49).
At the last follow-up, BM was visible by both tracer uptake and morphology in 50
patients (50/62, 81%). Ten patients (10/62, 16%) merely showed
pathological tracer uptake of BM. In two cases (2/62, 3%), BM were
morphologically visible without tracer accumulation. Forty-four patients
(71%) showed multifocal manifestations, and 18 (29%) showed
unifocal manifestations of BM. The most common sites were the spine (37/62,
60%) and pelvis (21/62, 34%).
Morphology of BM, skeletal-related events and therapy
The morphology of BM was available for 32 patients. Twenty-three patients showed
osteoblastic metastases (23/32, 72%), while 5 had osteolytic metastases
(5/32, 16%). In 4 cases, the morphology was mixed (4/32,
13%).
Skeletal-related events (SREs) were detected in 11 of the 62 patients
(18%), including pathological fractures, bone surgery, spinal cord
compression and bone radiation. There was a trend toward shorter survival in
patients with an SRE (41 months, 0.0–85.1, 95% CI) versus
patients with BM but without an SRE (67 months, 52.7–81.3, 95%
CI). However, this difference did not reach statistical significance
(p=0.185, log-rank test).
Thirty-six patients received antiresorptive therapy, that is, bisphosphonates or
denosumab. Among these, 2 patients (2/36, 6%) experienced an SRE. In
contrast, 26 patients did not receive bone-specific therapy. In this group, 9
SREs (9/26, 35%) occurred (odds ratio 9.0, p=0.0054,
Fisher’s exact test) ([Fig.
3]).
Fig. 3 Proportion of skeletal-related events (SRE) in PanNET
patients with and without antiresorptive therapy.
Overall survival (OS)
The median OS of the stage IV PanNET cohort was 82.0 months (53.6–110.4
months, 95% CI). The median OS was significantly longer at 116.0 months
(87.7–144.4 months, 95% CI) in patients with distant metastases
other than BM compared with 63.0 months (50.0–76.1 months, 95%
CI) in patients with BM (p=0.016, log-rank test) ([Fig. 4]). The median OS was 121.0 months
(79.7–162.3; 95% CI) in G1 PanNET patients, 103.0 months
(68.8–137.2; 95% CI) in G2 PanNET patients and 59.0 months
(34.6–83.4; 95% CI) in G3 PanNET patients (p=0.003,
log-rank test). Mortality was increased 1.8-fold when comparing the risk in G1
vs. G2 NET and G2 vs. G3 NET. At the same time, mortality risk was increased by
1.65 in patients with BM compared to those with distant metastases other than
BM. Interestingly, the time of BM manifestation had no impact on mortality risk.
Multivariate analysis confirmed these results.
Fig. 4 Median overall survival in PanNET patients with BM and
distant metastases other than bone (n=171).
Discussion
Historically, BMs have been considered to be rare in patients with NETs, with
conflicting or lacking data on individual primary locations [9]
[10]
[11]. In this study, we
investigated 171 stage IV PanNET patients from a cohort of 314 consecutive PanNET
patients treated with regular hybrid imaging at a single- center institution.
Occurrence of BM
BM has been reported in 3.6–26.0% of NEN patients using
heterogeneous examination methods and in different primary tumor localizations
([Table 3]). The highest prevalence
to date was reported by Scharf et al. [9],
who also included 92 PanNET patients. However, a proportion of poorly
differentiated NENs and the multitude of screening techniques with different
sensitivities limit this study ([Table
3]). Therefore, the authors concluded that their study, like others, may
underestimate the true prevalence of BM in NET [9]. In fact, we demonstrate a significantly higher frequency of BM.
In total, 36.3% of stage IV PanNET patients showed BM at the time of the
last follow-up. We consider this to be due to several reasons. The focus on
advanced NETs results in an increased prevalence of metastases. However, the
26% prevalence reported by Scharf et al. [9] also referred to a stage IV cohort;
therefore, additional causes are likely. The exclusive testing of pancreatic
NETs, which may have a higher incidence of BM than other NETs, is an option.
Previous studies that employed scintigraphy or conventional imaging demonstrated
a lower BM prevalence. In contrast, the exclusive use of superior SSTR-based
multi-phase contrast-enhanced hybrid imaging at initial diagnosis and follow-up
may reflect the real prevalence of BM more accurately. Thus, known restrictions
can be eliminated; selection bias in terms of availability of methods or
correlation of health status with a specific imaging technique can be excluded.
At the same time, our approach is independent of the distribution of different
disease stages in the cohort.
Table 3 Frequency of bone metastases (BM) in
neuroendocrine neoplasia (NEN) patients, reported in the
literature.
Entity
|
Cohort Size, n
|
Stage
|
Sample
|
BM Frequency
|
Imaging
|
Origin
|
Year
|
Reference
|
GEP NEN, CUP NEN, Other
|
668
|
I–IV
|
random, 87 centers
|
6.4%
|
CT, SRS, MRI
|
France
|
2009
|
Lombard-Bohas et al. [13]
|
GEP NEN, CUP NEN
|
837
|
I–IV
|
random, 46 centers
|
5.4%
|
CT, MRI, SRS
|
Spain
|
2010
|
Garcia-Carbonero et al. [14]
|
GEP NEN, Other NEN, Paraganglioma
|
691
|
I–IV
|
random, 2 centers
|
11.9%
|
X-ray, CT, MRI, bone scintigraphy, MIBG scan, PET scan
|
US
|
2015
|
Van Loon et al. [11]
|
GEP NEN, Other NEN, Pulmonary NEN
|
341
|
I–IV
|
retrospective institutional database
|
11.7%
|
CT, SRS, X-ray, MRI, bone scintigraphy
|
US
|
2015
|
Kavecansky et al. [10]
|
GEP NEN, Other NEN, Pulmonary NEN
|
7.334
|
I–IV
|
Swedish Cancer Registry
|
3.6%
|
not defined
|
Sweden
|
2016
|
Riihimäki et al. [4]
|
GEP NEN, CUP NEN, Pulmonary NEN
|
327
|
IV
|
institutional database
|
26.0%
|
MRI, DOTATOC PET/CT, SRS, CT, bone scintigraphy,
other PET/CT, X-ray
|
Germany
|
2017
|
Scharf et al. [9]
|
(677 all stages)
|
Differentiated PanNET
|
171
|
IV
|
prospective institutional registry, consecutive patients
|
36.3%
|
68Ga-DOTATOC PET/CT
|
Germany
|
2022
|
Current study
|
(314 all stages)
|
GEP: GastroEnteroPancreatic; CUP: Cancer of Unknown Primary; CT: Computed
tomography; SRS: Somatostatin receptor scintigraphy; MRI: Magnetic
resonance imaging; MIBG: 131I/123I-Metaiodobenzylguanidine; PET:
Positron emission tomography.
Interestingly, the majority of patients with bone metastases developed them
within the first two years after PanNET diagnosis. Together with the shorter
survival of these patients, this indicates that PanNET with BM is a more
aggressive subentity of PanNET rather than a stochastic coincidence in the
course of the disease. In 35% of cases, BM was initially confirmed
solely by tracer uptake. After a median of 8 months, morphological lesions were
visible in the majority of these patients. This underlines the importance of
hybrid imaging not only at initial diagnosis but also during follow-up. A more
aggressive course is thus detected earlier.
In addition to stage, grading based on the Ki-67 proliferation marker is a second
independent prognostic parameter for survival [3]. In line with previous data [9], the occurrence of BM was not dependent on grading in our study
([Fig. 1]). Interestingly, the hazard
ratio for death from any cause increases almost as significantly in the presence
of BM as with higher grading. The presence of BM is almost as strong an
influence on OS as grading and should be considered an independent risk factor
for OS in PanNETs. In addition, we investigated the influence of intrapancreatic
localization of the primary tumor on the presence of BM, following evidence that
localization in the pancreatic tail is associated with a more favorable clinical
course [12]. For the occurrence of BM,
there was no correlation with the intrapancreatic location of the primary
tumor.
One-third of patients with BM presented only by tracer uptake in the first scan,
not morphologically. This rate halved to 16% at the end of the study.
False-positive results may have occurred since tracer accumulation represents
the expression of somatostatin receptors rather than malignancy per se. However,
in the majority of cases, a morphological correlate appeared in ceCT after a
median of 8 months. Thus, a high specificity can be assumed. Published data
indicate a specificity of 89–92% with a sensitivity of
97–100% for 68Ga-DOTATOC PET/CT in NET bone
metastases [5].
SRE, therapy, and morphology of BM
Until now, the frequency of SRE in differentiated PanNETs has not been
elucidated. In a sample of NENs of different primaries, a high rate of
59% was reported for bone-specific symptoms in BM patients [11]. However, this result cannot simply be
transferred. Only twelve PanNETs were involved and different entities, such as
high-grade NEC, pheochromocytoma and NEN of unknown primary, were also included.
Other limitations apply to the results of Scharf et al. [9]. The rate of SRE was reported as
“nearly half of the patients” [9]. However, the number of PanNET patients examined in this work was
low as well; in addition, symptoms could have triggered the need for imaging;
therefore, overreporting seems likely. In contrast, the SRE ratio in our study
was significantly lower in 18% of BM patients. In addition to the
first-time analysis of SRE in differentiated PanNETs, we attribute the lower
rate to the structured use of hybrid imaging, which maps even asymptomatic BMs
sensitively. At the same time, the likelihood of SRE – one in five
patients with BM – is clinically highly relevant. We defined SRE as a
composite endpoint consisting of pathologic fracture, spinal cord compression,
or radiation or surgery to bone. Each of these events can have a lasting impact
on a patient’s quality of life.
It must be emphasized that the rate of SREs was not evenly distributed among BM
patients. In subjects treated with antiresorptive therapy, the probability of
SRE was significantly lower at 6% (2/36) than in those not
treated with bone-specific therapy (9/26, 35%; p=0.0054,
Fisher’s exact test). The use of antiresorptive therapy in PanNETs is
thus associated with a lower rate of SRE. The type of therapy –
bisphosphonates or denosumab – did not influence the outcome. Of course,
it must be considered that the patients were not randomized prospectively.
Nevertheless, we consider the result significant. Although we cannot exclude
that more advanced patients were more likely to be treated with antiresorptive
therapy, they still suffered less SRE.
The majority of our patients had osteoblastic BM. Given the limited number of
patients, morphology had no influence on the occurrence of SRE. Antiresorptive
therapy was similarly distributed in osteoblastic and osteolytic metastases.
Patients who initially showed tracer uptake only virtually always had
osteoblastic BM, suggesting that osteoblastic metastases are more likely not to
become morphologically evident.
Overall survival (OS)
OS between stage IV PanNET patients with and without BM differed significantly
between 63 and 116 months. A significant shortening of OS in patients with BM
compared with stage IV patients without BM was also reported in two of three
published studies. Congruent with our findings, both reported a near halving of
OS, Kavecansky et al. [10] from 98 to 52
months and Scharf et al. [9] from 100.8 to
49.0 months. The approximately one-year longer OS of our cohort in both groups
can be attributed to the biologic behavior of differentiated PanNETs, as the two
aforementioned studies included numerous other NETs ([Table 3]). In addition, improved therapies
are also possible, as our cohort is approximately 10 years more recent than the
two mentioned. Van Loon et al. [11] showed
a shortened OS in patients with BM, but the difference was not significant (62.1
vs. 75.4 months), most likely due to the heterogeneous group of different
primary sites.
It must be noted that the patients who initially showed only tracer uptake in
bone had an OS as limited as the patients with morphologically visible BM. Thus,
the structured use of SSTR-based hybrid imaging allows early identification of a
clinically vulnerable subgroup of PanNETs and should be used routinely.
Conclusion
Bone metastases in advanced PanNETs are not only common but also significant. The
presence of BM diminishes OS similar to higher tumor grade. One in five patients
experiences an SRE that persistently lowers the quality of life. Regarding this,
antiresorptive therapy is associated with a more favorable risk and should be
offered to all patients with BM in PanNETs.