The frequency of carcinoid syndrome at neuroendocrine tumor
diagnosis: A large population-based study using SEER-Medicare
data
Daniel M. Halperin, MD
*,1
, Chan Shen, PhD
*,2,3
, Arvind Dasari, MD
1
, Ying Xu, MD, MS
2
, Yiyi
Chu, MS
2
, Shouhao Zhou, PhD
3
, Ya-Chen Tina Shih, PhD
2,§
, and James C. Yao, MD
1,§
1
Department of Gastrointestinal Medical Oncology, The University of Texas MD Anderson Cancer
Center, 1515 Holcombe Blvd, Unit 426, Houston, TX 77030
2
Department of Health Services Research, The University of Texas MD Anderson Cancer Center,
1400 Pressler St, Unit 1444, Houston, TX 77030
3
Department of Biostatistics, The University of Texas MD Anderson Cancer Center, 1400 Pressler
St, Unit 1411, Houston, TX 77030
Abstract
BACKGROUND—Neuroendocrine tumors (NETs) can secrete bioactive amines into the
bloodstream causing carcinoid syndrome (CS), with symptoms including flushing and diarrhea.
However, CS frequency in the NET population has never been rigorously evaluated, nor has its
relationship to presenting characteristics. This analysis assessed the proportion of NET patients
with CS and associated clinical factors.
METHODS—We identified patients diagnosed 4/2000–12/2011 from the SEER-Medicare
database, excluding those with pancreatic tumors or small cell/large cell lung cancer, as well as
those without complete data. The incidence of patients with at least two claims of flushing
(782·62), diarrhea (564·5, 787·91), or carcinoid syndrome (259·2) during the three months before
and after NET diagnosis was assessed. We compared demographic and clinical characteristics
Corresponding Author: James C. Yao, MD, Department of Gastrointestinal Medical Oncology, The University of Texas MD Anderson
Cancer Center, 1515 Holcombe Blvd, Unit 426, Houston, TX 77030, Phone: 713.792.2828, Fax: 713.563.0539,
jyao@mdanderson.org.
*
These two authors contributed equally.
§
Full professor
Authors’ Contributions
DMH contributed conception and design, data analysis and interpretation, manuscript writing, and final manuscript approval. CS
contributed conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, and final
manuscript approval. AD contributed conception and design, manuscript writing, and final manuscript approval. YX contributed
conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, and final manuscript
approval. YC contributed data analysis and interpretation, manuscript writing, and final manuscript approval. SZ contributed data
analysis and interpretation, manuscript writing, and final manuscript approval. YTS contributed conception and design, data analysis
and interpretation, manuscript writing, and final manuscript approval. JCY contributed conception and design, data analysis and
interpretation, manuscript writing, and final manuscript approval.
Declaration of interests
Dr. Halperin reports grants from NCI, during the conduct of the study. Dr. Shen has nothing to disclose. Dr. Dasari has nothing to
disclose. Dr. Xu has nothing to disclose. Ms. Chu has nothing to disclose. Dr. Zhou has nothing to disclose. Dr. Shih has nothing to
disclose. Dr. Yao reports grants from NCI, during the conduct of the study.
Previously presented at ASCO Annual Meeting in Chicago, IL, June 2016.
HHS Public Access
Author manuscript
Lancet Oncol
. Author manuscript; available in PMC 2018 July 30.
Published in final edited form as:
Lancet Oncol
. 2017 April ; 18(4): 525–534. doi:10.1016/S1470-2045(17)30110-9.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
between patients with and without CS using chi-squared tests. We used the Cochran-Armitage
trend test to identify trends in CS incidence and cox regression to assess the relationship between
CS and survival.
FINDINGS—Among 9,512 NET patients, 1,922 (19%) had CS. The proportion of NET patients
with CS reported increased from 10·8% (50 of 465 patients) in 2000 to 18·6% (159 of 854
patients) in 2011 (p<0·0001). Patients with CS more likely had regional/distant than localized
disease, grade 1/2 than grade 3/4 histology, and small bowel or cecal primary tumors rather than
lung or colon. Female (p=0·00030) and non-Hispanic white patients (and p < 0·0001) more likely
reported CS. CS was associated with shorter survival (HR 1·1, p = 0·017). Age and patient location
were not correlated with the incidence of CS. Use of octreotide was more common in patients with
CS, while use of chemotherapy and radiation were used more frequently in patients without CS
and use of surgery was not significantly associated with the presence of CS at diagnosis.
INTERPRETATION—This population-based analysis reveals that CS is significantly associated
with grade, stage, primary site, and shorter survival. An improved understanding of the
heterogeneity of presenting syndromes among patients with NETs may permit more tailored
evaluation and management, and enables future research regarding the impact of CS control on
patient survival.
FUNDING—Supported by Ipsen, which sponsored purchasing SEER-Medicare data and funded
analytical support. The company was not involved in data collection, analysis, or interpretation.
This study was not directly funded by NIH.
Introduction
Neuroendocrine tumors (NETs) are relatively rare malignancies that can develop from a
diffuse network of neuroendocrine cells throughout the body. Their incidence is increasing,
with the most recent Surveillance, Epidemiology, and End Results (SEER) data showing a 5-
fold increase from 1973 to 2004, with an annual incidence of 5·25/100,000 in the U.S. in
2004. This increase is likely multifactorial and related to improvements in detection and
diagnosis, as well as a true rise in incidence.
1
Given the indolent nature of well-
differentiated NETs, many patients live over 5 years with metastatic disease, and must
endure symptoms for significant periods of time.
Extrapancreatic NETs are historically known as carcinoids due to their “carcinoma-like”
histology. These NETs can secrete bioactive amines causing carcinoid syndrome (CS).
2
These symptoms include wheezing, skin flushing, diarrhea, and fibrotic valvular heart
disease.
3
Consistent with observations that serotonin is a product of the enterochromaffin
cells
4
considered the nonmalignant counterparts of gastrointestinal NETs, serotonin
secretion is associated with CS in NET patients.
5
Similarly, urinary excretion of the
downstream serotonin metabolite, 5-hyrdoxyindolacetic acid, is elevated in patients with CS,
more so than serum serotonin, which fluctuates significantly.
6
The hormone secretion and
associated symptoms can be reduced, but not eliminated, in most patients with somatostatin
analogues such as octreotide.
7–9
However, the frequency of CS among patients with NETs has not been systematically
evaluated. While prior studies have attempted to estimate this frequency, the wide range
Halperin et al.
Page 2
Lancet Oncol
. Author manuscript; available in PMC 2018 July 30.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
identified (3–74%) suggests methodological limitations and the need to establish a more
accurate incidence rate in the broader NET patient population.
10–13
Indeed, prior analyses
were small retrospective studies from single centers or relied on surveys, with potential
ascertainment bias.
While the reported frequency of CS among NET patients has been inconsistent, the impact
of CS on patient quality of life is clear.
2,14–16
NET patients with CS demonstrate marked
impairments in multiple areas including fatigue, general health, and physical function,
compared both to the general population
15,16
and to NET patients without overt CS,
2
using
metrics such as the PROMIS-29 and SF-36. Additionally, smaller studies have suggested
that potentially due to subclinical niacin deficiency from tumor consumption of tryptophan
to produce serotonin,
6,17,18
even NET patients on therapeutic somatostatin analogues have
modest subjective and objective cognitive impairments.
19,20
Given the importance of
understanding the symptom burden of this growing patient population, this study used
SEER-Medicare, a large population-based database, to examine trends in the incidence of
CS and associated symptoms. Demographic and clinical characteristics were compared
between patients with and without CS to determine factors associated with having NETs and
CS. Treatment selection and patient outcomes were also evaluated.
Materials and Methods
Data Source
The data source used for this study is the SEER registry linked with Medicare claims data.
The SEER program of the National Cancer Institute (NCI) provides information on cancer
statistics including tumor characteristics, patient demographics, and cause of death
information for persons diagnosed with cancer in the United States. The SEER registries
cover approximately 28% of the U.S. population and, when linked to Medicare claims data,
the SEER-Medicare data provides detailed patient information such as neighborhood
socioeconomic status (SES), comorbidities, and the usage of health care resources. Of note,
the SEER registry uses 4 grades based on extraction of pathology reports without central
review. Therefore, in cases where pathology reports did not explicitly describe tumors as
WHO grade 1, 2, or 3, extraction based on morphologic description was performed, with
SEER grade I including tumors classified as “well differentiated,” grade II including tumors
classified as “moderately differentiated,” grade III including tumors classified as “poorly
differentiated,” and grade IV including tumors classified as “undifferentiated,” or
“anaplastic.” SEER grade III or IV cancers would be “grade 3 neuroendocrine carcinomas”
in the 2010 WHO nomenclature, which is primarily based on division rate rather than
histologic appearance.
21
The SEER-Medicare database has been widely used is considered
to be representative of the older U.S. population.
22
Study Cohort
The study cohort consisted of 9,512 cancer patients over age 65 who were diagnosed with
NETs between April 1, 2000 and December 31, 2011. NETs were identified by International
Classification of Diseases for Oncology, 3rd Edition (ICD-O-3) codes: 8240, 8241, 8242,
8243, 8244, 8245, 8246, and 8249 (gastrointestinal and pulmonary carcinoids, and
Halperin et al.
Page 3
Lancet Oncol
. Author manuscript; available in PMC 2018 July 30.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
neuroendocrine carcinomas). Pancreatic NET patients were excluded due to overlapping
symptoms with CS, as well as the lack of serotonin-producing cells in the normal pancreas.
Patients less than 65 years old at the time of diagnosis were excluded, as they make up a
minority of patients enrolled in Medicare are defined by specific qualifying conditions,
therefore making these patients less representative of the overall population. We also
required continuous Medicare Parts A and B enrollment (providing coverage for hospital
and physician services) and no Health Maintenance Organization (HMO) enrollment during
the time period between three months before and three months after the NET diagnosis to
ensure consistent data availability and the completeness of medical claims used to identify
CS.
Identification of CS
As previously described,
23
we identified the presence of CS by International Classification
of Disease 9
th
Revision (ICD-9) codes: flushing (782·62), diarrhea (564·5, 787·91), and
carcinoid syndrome (259·2). We defined patients to have CS if they had at least two claims
with any of the above mentioned ICD-9 codes during a six month time window between
three months before and three months after the NET diagnosis. We therefore tightened the
case ascertainment criterion for CS, increasing the specificity for an accurate diagnosis, but
sacrificing sensitivity, potentially permitting the relatively rare patients who develop CS
subsequently to go undetected. This concession was necessary to avoid confounding from
coding for diarrhea or CS that arises as a complication of therapies, such as bowel resections
or steatorrhea from somatostatin analogue therapy.
Demographic and clinical factors
The demographic information included age [65–69, 70–74, 75–79, ≥80], gender [male vs.
female], race/ethnicity [non-Hispanic white, non-Hispanic black, Hispanics or all others],
region [Northeast, West, Midwest, South], and urban/rural status [metropolitan vs. non-
metropolitan]. Clinical factors included tumor grade, stage, and primary site, as well as
treatment variables such as octreotide, chemotherapy, radiation, and surgery.
Statistical Analysis
We used the Cochran-Armitage trend test to evaluate the trend in CS. This method was
chosen over parametric regression to avoid the assumption that the change over time could
be described with a simple function. We used the chi-square test to compare demographic
and clinical characteristics between patients with and without CS, as well as the frequency
with which interventions were performed in patients with and without CS and the proportion
of patients experiencing CS by grade, stage, and primary site. Multivariate Cox regression
analysis was used to identify prognostic variables, with plots of log of the cumulative
hazards function versus log of survival time employed to ensure that the proportional
hazards assumption was satisfied (Supplementary Figure 1, pages 2–14). Multivariate
logistic regression was performed to determine the use of specific interventions in patients
with versus without CS while controlling for other explanatory variables. Median survival
was estimated using the Kaplan-Meier method.
Halperin et al.
Page 4
Lancet Oncol
. Author manuscript; available in PMC 2018 July 30.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
All statistical analyses were conducted in SAS 9·3 (SAS Institute, Cary NC). The
Institutional Review Board at The University of Texas MD Anderson Cancer Center
exempted this study for approval because all patients were de-identified.
Role of the funding source
The sponsor had no involvement in the study design, data collection, analysis, or
interpretation, and was furthermore not involved in the writing of the report. DMH, CS, AD,
YX, YC, SZ, YTS, and JCY all had access to all raw data. The corresponding author had
full access to all of the data and the final responsibility to submit for publication.
Results
Overall frequency of reported CS
Figure 1 depicts the selection criteria by which the NET cohort was generated, resulting in
9,512 patients for analysis. The proportion of newly diagnosed NET patients also diagnosed
with CS or its component symptoms was assessed for each year between 2000 and 2011
(Figure 2). Among the 9,512 patients diagnosed with NETs over this period, 1,784 (18·8%)
manifested CS in the six months surrounding the diagnosis of NET. The proportion of
patients diagnosed with CS increased over this time period from 10·8% (50 of 465 patients)
in 2000 to 18·6% (159 of 854) in 2011 (Figure 1, p<0·0001 by Cochran-Armitage test).
CS and demographics
Table 1 contains patient characteristics that were evaluated for association with the diagnosis
of CS. The percentage of patients with CS as compared to those patients without did not
differ significantly with respect to age at diagnosis, geographic region, or metropolitan vs.
rural status. Interestingly, female patients were more likely to have CS diagnosed (1091 of
5449 [20%] compared to 695 of 4063 [17·1%] for males; p=0·00030). Lastly, race/ethnicity
was associated with a significant difference in the reported incidence of CS. CS occurred in
1483 (19·8%) of 7501 non-Hispanic white, 166 (16·3%) of 1016 non-Hispanic black, and
137 (13·8%) of 995 Hispanic or other race cases. These differences were highly statistically
significant (p<0·0001).
CS and NET grade
The proportion of patients exhibiting symptoms of CS was assessed in relation to tumor
grade. Well-differentiated tumors of grade 1 or 2 histology were associated with CS in
22·6% (1302 of 5774) and 17·5% (114 of 651) of cases, respectively (Table 1). Poorly
differentiated tumors of grade 3 or 4 histology were associated with CS in 8·2% (77 of 939)
and 5·1% (18 of 350) of cases, respectively. This difference was highly statistically
significant (p<0·0001).
CS and NET stage
Using SEER staging, the proportion of NETs manifesting CS was evaluated in relation to
extent of disease. Localized NETs were associated with CS in 11·9% (473 of 3965) of cases,
whereas regionally advanced and distant metastatic NETs were associated with carcinoid
Halperin et al.
Page 5
Lancet Oncol
. Author manuscript; available in PMC 2018 July 30.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript