- Email: [email protected]
Foregut carcinoids: A clinical and biochemical analysis
Foregut carcinoids: A clinical and biochemical analysis Paul M. Kirshbom, MD, Aftab R. Kherani, MD, Mark W. Onaitis, MD, Ankie Hata, BS, Theresa E. Kehoe, MD, Carol Feldman, RN, Jerome M. Feldman, MD, and Douglas S. Tyler, MD, Durham, NC
Background. Gastrointestinal foregut carcinoids make up a small percentage (3% to 6%) of all reported carcinoids. Because these tumors are so uncommon, comparisons between the subtypes have been difficult. The goal of this study was to compare the hormonal and clinical characteristics of gastric, duodenal, and pancreatic carcinoids. Methods. A prospective database of approximately 750 carcinoid patients seen by one author over 25 years was reviewed, and the 104 patients with gastric (33), duodenal (17), or pancreatic (54) carcinoids were selected as the subgroup for analysis. These patients were compared with regard to hormone levels, clinical course, treatment, and survival. Results. Duodenal carcinoids exhibited significantly lower serotoninergic hormone levels than did the gastric and pancreatic carcinoids (urine 5-hydroxyindoleacetic acid [mg/24 h], 5 ± 1 vs 16 ± 5 and 47 ± 12, respectively, P = .03). Pancreatic carcinoids presented with more advanced stage (distant metastases 87% vs 42% and 20% for gastric and duodenal, respectively) and had worse outcomes than patients with gastric and duodenal tumors with 10-year survivals of 10%, 59%, and 58%, respectively ( P = .003). Conclusions. Pancreatic carcinoids produce higher levels of serotoninergic hormones and have a significantly higher stage and worse outcome than other foregut carcinoids. This study demonstrates that the organ of origin is an important determinant of hormonal activity and clinical course for patients with foregut carcinoids. (Surgery 1999;126:1105--10.) From the Departments of Surgery and Endocrinology, Duke University Medical Center, Durham, NC
THE TERM KARZINOIDE was first used by Oberndorfer1 in 1907 to describe a group of gastrointestinal tumors that were notably less aggressive than the more common adenocarcinomas. Further characterization of these tumors revealed that they were endocrine in origin with characteristic argentaffin silver staining.2 Carcinoids arise most commonly in the gastrointestinal tract and bronchopulmonary tree, but they have been reported in a variety of locations including the gallbladder, retroperitoneum, and ovaries.3 Although the cell of origin from which carcinoids arise has been a subject of debate, there is little doubt that these tumors are biochemically members of the amine precursor decarboxylase system. They are capable of producing a multitude of pepPresented at the 20th Annual Meeting of the American Association of Endocrine Surgeons, New Haven, Conn, May 24, 1999. Reprint requests: Douglas S. Tyler, MD, Department of Surgery, PO Box 3118, Duke University Medical Center, Durham, NC 27710. Copyright © 1999 by Mosby, Inc. 0039-6060/99/$8.00 + 0
11/6/101430
tide hormones, the most common being serotonin and the serotoninergic peptides.4 It is thought that release of these hormones into the systemic circulation leads to the characteristic carcinoid syndrome consisting of flushing, diarrhea, wheezing, and right-sided heart disease.4,5 Carcinoid tumors are typically classified by the location of origin within the embryonic subdivisions of the gut: foregut, midgut, and hindgut. Prior reports have clearly demonstrated differences in the biochemical and clinical activity of carcinoids arising in different locations.3,6-8 The most common subgroup, midgut, has been well described in the literature, but the less common foregut carcinoids have not received as much attention. The goals of this study were to define the biochemical and clinical characteristics of the gastrointestinal (ie, nonpulmonary) foregut carcinoids, including gastric, duodenal, and pancreatic. PATIENTS AND METHODS Patients. Between 1970 and 1997 more than 750 patients with carcinoid tumors were evaluated by one author (J.M.F.) at the Duke University Medical SURGERY 1105
1106 Kirshbom et al
Surgery December 1999
Fig 1. Survival curves for foregut carcinoid patients divided by site of origin.
Table I. Age and gender distribution Origin Stomach Duodenum Pancreas
Number
Gender (M/F)
Age at diagnosis (y) (range)
33 17 54
13/20 13/4 35/21
57.9 (26-79) 56.4 (30-86) 52.2 (28-77)
Center and the Durham Veterans Affairs Hospital. This study focused on 104 patients with carcinoids arising in the stomach (33), duodenum (17), or pancreas (54). Hormone levels and patient demographics were collected and maintained prospectively in a database. The medical records of 76 (73%) patients from this group were retrospectively reviewed to determine the tumor size and stage at diagnosis. Regional disease was defined as disease confined to regional lymph nodes or limited direct local spread. Distant disease was defined as the presence of liver or other distant organ metastases, peritoneal spread, or diffuse retroperitoneal involvement. Follow-up data for 98 patients (94%) were obtained from the medical records, patient interviews, or serial contacts by one author (J.M.F.). Hormone levels. The majority of the patients (98%) had 24-hour urine specimens, blood samples, or both collected at the time of initial evaluation at Duke University Medical Center. The urine samples were stored at 3°C and assayed for 5-hydroxyindoleacetic acid (5-HIAA) and serotonin (5-HT) levels using spectrophotometric and radioenzymatic assays, respectively.9,10 Serotonin levels were also
measured in the serum and platelets.4,9 After 1988 high pressure liquid chromatography was used to measure all serotonin levels.11 These data were prospectively collected and maintained in a database including patient demographics, site of tumor origin, and hormone levels at presentation. Subgroup classifications. Carcinoid tumors are notoriously heterogeneous with regard to hormonal activity and clinical course both between and within individual organs. Therefore the criteria used to define a carcinoid tumor varies from organ to organ. For example, the distinction between pancreatic carcinoid tumors and pancreatic islet cell tumors can be difficult or impossible based on routine histology. In this study, patients with pancreatic neuroendocrine tumors that were histologically consistent with carcinoid tumors were only included in the carcinoid subgroup if they had documented excess production of serotonin by at least 1 of the 4 assays described. Duodenal carcinoids were defined as tumors with histology consistent with carcinoid tumors with at least one of the following: elevated chromogranin A levels, evidence of neurosecretory granules on electron microscopy, or evidence of increased serotonin production by 1 of the 4 assays described. Histology alone without either serotonin or chromogranin A elevation was not sufficient for inclusion. The criteria used for inclusion in the gastric carcinoid group were the same as for the duodenal carcinoids. These patients were then further subclassified into those with a reported history of
Kirshbom et al 1107
Surgery Volume 126, Number 6
Fig 2. Survival curves for foregut carcinoid patients divided by stage at diagnosis. Local, no evidence of metastases; Regional, regional lymph nodes or limited local spread; Distant, distant metastases, peritoneal studding, or diffuse retroperitoneal spread.
Fig 3. Survival curves for gastric carcinoid patients divided by subtype.
either chronic atrophic gastritis (CAG) or pernicious anemia (PA) combined in one group (CAGPA group) and those without a history of CAG or PA, called the Sporadic group. These groups correspond to the groups labeled type 1 and type 3, respectively, in the literature. We could not identify any patients with clear evidence of multiple endocrine neoplasia syndrome, so a group corresponding to type 2 is not included. Statistics. Kruskal-Wallis one-way analysis of variance (ANOVA) was used to compare groups. Kaplan-Meier analysis was used to evaluate survival from time of diagnosis to death. Mantel-Cox logrank test was used to compare survival curves between groups. Cox proportional hazard method was used to perform multivariate analysis. A P value
less than .05 was considered significant for all tests. The computer program Statistica for Windows Version 5 (Statsoft, Inc, Tulsa, Okla) was used for all ANOVA evaluations and survival curve generation, and SAS software was used for Cox proportional hazards. RESULTS Demographics. Patient demographics including site of origin, gender, and age are shown in Table I. The gender distribution is consistent with prior reports in that there is a male preponderance for all gastrointestinal sites except for gastric carcinoids, which occur more frequently in women. The age at diagnosis was not significantly different between groups.
1108 Kirshbom et al
Surgery December 1999
Table II. Presenting symptoms Origin (n)
Pain
Stomach (24) CAG/PA (11) Sporadic (13) Duodenum (9) Pancreas (33)
9 (38%) 3 (27%) 6 (46%) 4 (44%) 21 (64%)
Diarrhea
Flushing
Nausea/ vomiting
Gastrointestinal bleeding/guaiac +
5 (21%) 4 (36%) 1 (8%) 1 (11%) 15 (45%)
6 (25%) 2 (18%) 4 (31%) 0 2 (6%)
4 (17%) 2 (18%) 2 (15%) 5 (56%) 2 (6%)
5 (21%) 3 (27%) 3 (23%) 5 (56%) 1 (3%)
Table III. Stage of carcinoids at diagnosis Origin (n)
Local
Stomach (26) 15 (58%) CAG/PA (11) 9 (82%) Sporadic (15) 6 (40%) Duodenum(10) 7 (70%) Pancreas (38) 3 (8%) Total (74) 25 (34%)
Regional
Distant
0 0 0 1 (10%) 2 (5%) 3 (4%)
11 (42%) 2 (18%) 9 (60%) 2 (20%) 33 (87%) 46 (62%)
Clinical symptoms and staging. The most commonly noted symptoms at presentation included abdominal pain, flushing, diarrhea, nausea and vomiting, and evidence of gastrointestinal bleeding (guaiac positive stools, anemia, or frank gastrointestinal bleeding). The incidence of these symptoms is summarized in Table II. Gastric carcinoids presented with flushing as a significant symptom more frequently than the other subtypes (25% vs 0% and 6% for duodenal and pancreatic, respectively), with the highest incidence occurring in the sporadic gastric carcinoids (31%). Diarrhea, on the other hand, occurred much more frequently in the hypergastrinemic CAG/PA group than in the sporadic carcinoids. Pancreatic carcinoids typically presented with pain and diarrhea. For the patients whose medical records were available, staging data were abstracted and are presented in Table III. Pancreatic carcinoids presented with distant metastases more frequently than the other groups (87% vs 42% and 20% for gastric and duodenal, respectively). Gastric and duodenal carcinoids presented with local disease the majority of the time (58% and 70%, respectively). Pancreatic carcinoids were significantly larger at presentation than the other groups (5.3 ± 1.0 cm vs 2.1 ± 0.3 cm and 3.1 ± 0.5 cm for duodenal and gastric, respectively, P = .009). Table III also demonstrates the stage at presentation for the gastric carcinoid subgroups. Those patients with associated hypergastrinemic states (CAG and PA) were found to have significantly lower rates of distant metastases at presentation (18%)
than did patients with sporadic carcinoids (60%, P = .03). The CAG/PA tumors were also significantly smaller than the sporadic gastric carcinoids (2.0 ± 0.7 cm vs 4.3 ± 0.8 cm, respectively, P < .05). Hormone levels. The levels of 5-HIAA and 5-HT in the urine and 5-HT in the serum and platelets are summarized in Table IV. Although the levels of all hormones were higher in the gastric and pancreatic groups than in the duodenal group, this difference was only statistically significant for the urine 5-HIAA and serum 5-HT levels. Comparison of the hormone levels in patients with locoregional disease as opposed to those with distant metastases at presentation reveals that all hormone levels are higher in the patients with distant disease (Table V). Although these differences approach significance for all groups, only the serum 5-HT groups are statistically significantly different (P = .05). Survival. The survival curves for all of the patients for whom follow-up data were available (94%), divided by site of origin, are displayed in Figure 1. Patients with pancreatic carcinoids had significantly shorter survival times, with 10-year survivals of 15% vs 58% and 59% for gastric and duodenal carcinoids, respectively (P < .0001). Survival curves for these patients divided by stage at presentation are displayed in Figure 2. There is a significant difference in survival between patients who present with locoregional disease as opposed to those with distant metastases. The 10-year survival for these groups was 10% for patients with distant disease versus 95% for those with local disease (P < .0001). Eightyear survival was 72% for patients with regional disease with no patients followed out to 10 years thus far. Multivariate analysis including age, gender, origin, and stage revealed that only stage was a significant predictor of outcome (P = .001). Origin was not found to be a factor when stage was included in the analysis. DISCUSSION Carcinoid tumors are relatively rare neuroendocrine malignancies that most commonly arise in the gastrointestinal tract. The overall incidence of
Kirshbom et al 1109
Surgery Volume 126, Number 6 Table IV. Hormone levels of carcinoid patients by site of origin Origin Stomach Duodenum Pancreas P value Normal
Urine 5-HIAA (mg/24 h) 16 ± 5 (29) 5 ± 1 (16) 47 ± 12 (54) .03 2-8
Urine 5-HT (µg/24 h) 2070 ± 1043 (27) 184 ± 65 (14) 1347 ± 314 (52) .24 50-220
Serum 5-HT (pmol/mL)
Platelet 5-HT (pmol/mg protein)
1942 ± 606 (24) 515 ± 141 (10) 2868 ± 369 (41) .02 300-2200
1899 ± 573 (21) 568 ± 133 (8) 2603 ± 428 (40) .10 100-1500
Mean ± SEM (n), P value between groups by ANOVA.
Table V. Hormone levels of carcinoid patients by stage at presentation Stage Locoregional Distant P value Normal
Urine 5-HIAA (mg/24 h) 12 ± 7 (24) 41 ± 11 (45) .07 2-8
Urine 5-HT (µg/24 h) 253 ± 88 (21) 1287 ± 369 (44) .06 50-220
Serum 5-HT (pmol/mL)
Platelet 5-HT (pmol/mg protein)
1589 ± 497 (21) 3074 ± 500 (34) .05 300-2200
1411 ± 477 (20) 2353 ± 479 (34) .20 100-1500
Mean ± SEM (n), P value between groups by ANOVA.
these tumors is approximately 1 to 2 cases per 100,000 population.3,12 In all reported series, the majority of the cases have been of midgut origin, with a relatively small percentage arising in the foregut organs. The largest series of carcinoids in the literature reported by Modlin and Sandor,3 for example, included 204 gastric and 46 pancreatic carcinoids or 5% of the total with no mention of duodenal carcinoids. As a result of the overall rarity of carcinoid tumors and the relatively small percentage of carcinoids that arise in the foregut, there has been a scarcity of information available concerning the biochemical and clinical behavior of these tumors. As with all clinical studies, the patient population must be considered before the data can be extrapolated to other groups. Many of these patients were referred to a tertiary care center for treatment of advanced carcinoid tumors. Despite this selection bias, the distribution of stages is quite similar to the SEER group review,3 with localized tumor present in 58% and 8% of the gastric and pancreatic carcinoids, respectively, in this study as opposed to 53% and 9% in the SEER study. It is also important to recognize that this is not a natural history study because these patients received multiple modalities and protocols of treatment. Recognizing these limitations, this study demonstrates several important points. First, pancreatic carcinoids have a worse outcome than other foregut carcinoids. This finding can be explained
in several ways. Pancreatic carcinoids, because of their location, tend to produce relatively vague, nonspecific symptoms resulting in larger, more advanced tumors at presentation. After correction for stage at presentation, tumor location was not predictive of outcome using multivariate analysis. Tumor stage, which is significantly more advanced in pancreatic carcinoids, strongly influences outcome. Therefore, our study suggests that pancreatic carcinoids are not intrinsically more aggressive than other carcinoids. Another interesting point is that gastric carcinoid patients exhibit the characteristic flushing response more commonly than the other foregut carcinoids (23% vs 0% and 6% for duodenal and pancreatic, respectively). This finding is surprising given the more advanced stage and higher levels of serotoninergic peptides produced by pancreatic carcinoids; however, it does lend credence to the hypothesis that the distinctive flushing response produced by gastric carcinoids is the result of a nonserotoninergic hormone (histamine, bradykinin, etc).6 Gastric carcinoids associated with CAG and PA were smaller than sporadic carcinoids and they tended to behave less aggressively (Fig 3). It has not been established whether these differences are the result of underlying biologic differences between these tumor subtypes or whether they merely represent a lead-time error resulting from increased surveillance of the CAG-PA group.
1110 Kirshbom et al
In conclusion, although foregut carcinoids have a similar prognosis when corrected for stage, site of origin is an important determinant of outcome. Despite the high levels of serotoninergic hormones produced by pancreatic carcinoids, they tend to grow to large sizes and metastasize before presentation. This results in significantly shorter life expectancy for patients with pancreatic carcinoids when compared with gastric and duodenal carcinoid tumors. REFERENCES 1. Oberndorfer S. Karzinoide tumoren des Dunndarms. Frankfurt Zeitschr Pathol 1907;1:426-30. 2. Grimelius L, Wilander E. Silver stains in the study of endocrine cells of the gut and pancreas. Invest Cell Pathol 1980;3:3-12. 3. Modlin IM, Sandor A. An analysis of 8305 cases of carcinoid tumors. Cancer 1997;79:813-29. 4. Feldman JM, O’Dorisio TM. Role of neuropeptides and serotonin in the diagnosis of carcinoid tumors. Am J Med 1986;81:41-8. 5. Moertel CG. An odyssey in the land of small tumors. J Clin Oncol 1987;5:1503-22. 6. Akerstrom G. Management of carcinoid tumors of the stomach, duodenum, and pancreas. World J Surg 1996;20:173-82. 7. Greenberg RS, Baumgarten DA, Clark WS, Isacson P, McKeen K. Prognostic factors for gastrointestinal and bronchopulmonary carcinoid tumors. Cancer 1987;60:2476-83. 8. Kirshbom PM, Kherani AR, Onaitis MW, Feldman JM, Tyler DS. Carcinoids of unknown origin: comparative analysis with foregut, midgut, and hindgut carcinoids. Surgery 1998;124:1063-70. 9. Hussain MN, Sole MJ. A simple specific radioenzymatic assay for picogram quantities of serotonin or acetyl serotonin in biological fluids or tissues. Anal Biochem 1981;111:105-10. 10. Udenfriend S, Titus E, Weissbach H. The identification of 5hydroxyindoleacetic acid in normal urine and a method for its assay. J Biol Chem 1955;216:499-505. 11. Tagari PC, Boullin DJ, Davies CL. Simplified determination of serotonin in plasma by liquid chromatography with electrochemical detection. Clin Chem 1984;30:131-5. 12. Godwin JD. Carcinoid tumors: an analysis of 2837 cases. Cancer 1975;36:560-9.
DISCUSSION Dr Hans-Dietrich Roeher (Düsseldorf, Germany). It is well known that carcinoids are accompanied by other malignancies of nonendocrine origin. You lost a couple of patients during the follow-up; was it due to other malignancies? Dr Kirshbom. There were several patients with prior and subsequent malignancies. None of the patients in the localized groups actually died of malignant disease. Several patients dropped off for nonmalignant causes, coronary disease, et cetera.
Surgery December 1999
Dr Geoffrey B. Thompson (Rochester, Minn). I was somewhat surprised by the number of patients with distant metastases in your gastric carcinoid group. In our experience with patients with CAG/PA, secondary hypergastrinemia and gastric carcinoids (ECL tumors), we have not seen nodal metastases, distant metastases, or death due to tumor. Was there anything atypical about the size or histology of these tumors, as described by Carney? Typically, these tumors are very small, less than 2 cm, and are frequently multiple. We start to worry when they get up over 2 cm or there is atypical histology, but this is very unusual. Have you looked at this in more detail? We generally do not recommend any treatment in this subgroup of gastric carcinoid patients, because spontaneous regression can occur without antrectomy. Dr Kirshbom. That also surprised me. Looking through the data, it seems that most of the patients with distant disease and CAG were very early in the series, and I question whether screening was as aggressive at that time as it was later in the series. As you noted, the majority of the patients with CAG had very small tumors. The average was less than 2 cm, somewhere around 1 cm, as opposed to much larger tumors for the sporadic cases. But again it was a relatively small percentage of patients with distant disease. Dr Norman W. Thompson (Ann Arbor, Mich). I am a little confused about the semantics. For instance, in your overall group with gastric carcinoids, I think you are including what many of us would call ECL tumors. In your group with duodenal neuroendocrine tumors, how do you define carcinoids? Were these immunohistochemically stained to rule out somatostatinomas, which are frequently found in the duodenum, and how did you differentiate them from gastrinomas? They can all be grouped together if a pathologist looks at them without knowledge of their function or specific stains. If they are not stained, then we don’t know exactly what we are dealing with or what to call them. Dr Kirshbom. These were all stained. Dr Feldman is fairly obsessive about his documentation of differentiation between pancreatic carcinoids and pancreatic islet cells, for example. He personally evaluated those patients and their slides. As far as the duodenal carcinoids, patients with Zollinger-Ellison syndrome and evidence of hypergastrinemia were excluded. These are patients who did not have evidence of elevated hormone levels. Dr Janice L. Pasieka (Calgary, Alberta, Canada). Do you have any data on chromogranin A? In the last couple of years it has become the tumor marker that many are promoting for carcinoid, and I am wondering whether you have any data on it compared with urinary 5-hydroxyindoleacetic acid. Dr Kirshbom. Unfortunately, I do not. Dr Feldman was in the opposing camp stressing the use of serotonin levels as opposed to chromogranin levels, and he did not measure them in these studies.
Background. Gastrointestinal foregut carcinoids make up a small percentage (3% to 6%) of all reported carcinoids. Because these tumors are so uncommon, comparisons between the subtypes have been difficult. The goal of this study was to compare the hormonal and clinical characteristics of gastric, duodenal, and pancreatic carcinoids. Methods. A prospective database of approximately 750 carcinoid patients seen by one author over 25 years was reviewed, and the 104 patients with gastric (33), duodenal (17), or pancreatic (54) carcinoids were selected as the subgroup for analysis. These patients were compared with regard to hormone levels, clinical course, treatment, and survival. Results. Duodenal carcinoids exhibited significantly lower serotoninergic hormone levels than did the gastric and pancreatic carcinoids (urine 5-hydroxyindoleacetic acid [mg/24 h], 5 ± 1 vs 16 ± 5 and 47 ± 12, respectively, P = .03). Pancreatic carcinoids presented with more advanced stage (distant metastases 87% vs 42% and 20% for gastric and duodenal, respectively) and had worse outcomes than patients with gastric and duodenal tumors with 10-year survivals of 10%, 59%, and 58%, respectively ( P = .003). Conclusions. Pancreatic carcinoids produce higher levels of serotoninergic hormones and have a significantly higher stage and worse outcome than other foregut carcinoids. This study demonstrates that the organ of origin is an important determinant of hormonal activity and clinical course for patients with foregut carcinoids. (Surgery 1999;126:1105--10.) From the Departments of Surgery and Endocrinology, Duke University Medical Center, Durham, NC
THE TERM KARZINOIDE was first used by Oberndorfer1 in 1907 to describe a group of gastrointestinal tumors that were notably less aggressive than the more common adenocarcinomas. Further characterization of these tumors revealed that they were endocrine in origin with characteristic argentaffin silver staining.2 Carcinoids arise most commonly in the gastrointestinal tract and bronchopulmonary tree, but they have been reported in a variety of locations including the gallbladder, retroperitoneum, and ovaries.3 Although the cell of origin from which carcinoids arise has been a subject of debate, there is little doubt that these tumors are biochemically members of the amine precursor decarboxylase system. They are capable of producing a multitude of pepPresented at the 20th Annual Meeting of the American Association of Endocrine Surgeons, New Haven, Conn, May 24, 1999. Reprint requests: Douglas S. Tyler, MD, Department of Surgery, PO Box 3118, Duke University Medical Center, Durham, NC 27710. Copyright © 1999 by Mosby, Inc. 0039-6060/99/$8.00 + 0
11/6/101430
tide hormones, the most common being serotonin and the serotoninergic peptides.4 It is thought that release of these hormones into the systemic circulation leads to the characteristic carcinoid syndrome consisting of flushing, diarrhea, wheezing, and right-sided heart disease.4,5 Carcinoid tumors are typically classified by the location of origin within the embryonic subdivisions of the gut: foregut, midgut, and hindgut. Prior reports have clearly demonstrated differences in the biochemical and clinical activity of carcinoids arising in different locations.3,6-8 The most common subgroup, midgut, has been well described in the literature, but the less common foregut carcinoids have not received as much attention. The goals of this study were to define the biochemical and clinical characteristics of the gastrointestinal (ie, nonpulmonary) foregut carcinoids, including gastric, duodenal, and pancreatic. PATIENTS AND METHODS Patients. Between 1970 and 1997 more than 750 patients with carcinoid tumors were evaluated by one author (J.M.F.) at the Duke University Medical SURGERY 1105
1106 Kirshbom et al
Surgery December 1999
Fig 1. Survival curves for foregut carcinoid patients divided by site of origin.
Table I. Age and gender distribution Origin Stomach Duodenum Pancreas
Number
Gender (M/F)
Age at diagnosis (y) (range)
33 17 54
13/20 13/4 35/21
57.9 (26-79) 56.4 (30-86) 52.2 (28-77)
Center and the Durham Veterans Affairs Hospital. This study focused on 104 patients with carcinoids arising in the stomach (33), duodenum (17), or pancreas (54). Hormone levels and patient demographics were collected and maintained prospectively in a database. The medical records of 76 (73%) patients from this group were retrospectively reviewed to determine the tumor size and stage at diagnosis. Regional disease was defined as disease confined to regional lymph nodes or limited direct local spread. Distant disease was defined as the presence of liver or other distant organ metastases, peritoneal spread, or diffuse retroperitoneal involvement. Follow-up data for 98 patients (94%) were obtained from the medical records, patient interviews, or serial contacts by one author (J.M.F.). Hormone levels. The majority of the patients (98%) had 24-hour urine specimens, blood samples, or both collected at the time of initial evaluation at Duke University Medical Center. The urine samples were stored at 3°C and assayed for 5-hydroxyindoleacetic acid (5-HIAA) and serotonin (5-HT) levels using spectrophotometric and radioenzymatic assays, respectively.9,10 Serotonin levels were also
measured in the serum and platelets.4,9 After 1988 high pressure liquid chromatography was used to measure all serotonin levels.11 These data were prospectively collected and maintained in a database including patient demographics, site of tumor origin, and hormone levels at presentation. Subgroup classifications. Carcinoid tumors are notoriously heterogeneous with regard to hormonal activity and clinical course both between and within individual organs. Therefore the criteria used to define a carcinoid tumor varies from organ to organ. For example, the distinction between pancreatic carcinoid tumors and pancreatic islet cell tumors can be difficult or impossible based on routine histology. In this study, patients with pancreatic neuroendocrine tumors that were histologically consistent with carcinoid tumors were only included in the carcinoid subgroup if they had documented excess production of serotonin by at least 1 of the 4 assays described. Duodenal carcinoids were defined as tumors with histology consistent with carcinoid tumors with at least one of the following: elevated chromogranin A levels, evidence of neurosecretory granules on electron microscopy, or evidence of increased serotonin production by 1 of the 4 assays described. Histology alone without either serotonin or chromogranin A elevation was not sufficient for inclusion. The criteria used for inclusion in the gastric carcinoid group were the same as for the duodenal carcinoids. These patients were then further subclassified into those with a reported history of
Kirshbom et al 1107
Surgery Volume 126, Number 6
Fig 2. Survival curves for foregut carcinoid patients divided by stage at diagnosis. Local, no evidence of metastases; Regional, regional lymph nodes or limited local spread; Distant, distant metastases, peritoneal studding, or diffuse retroperitoneal spread.
Fig 3. Survival curves for gastric carcinoid patients divided by subtype.
either chronic atrophic gastritis (CAG) or pernicious anemia (PA) combined in one group (CAGPA group) and those without a history of CAG or PA, called the Sporadic group. These groups correspond to the groups labeled type 1 and type 3, respectively, in the literature. We could not identify any patients with clear evidence of multiple endocrine neoplasia syndrome, so a group corresponding to type 2 is not included. Statistics. Kruskal-Wallis one-way analysis of variance (ANOVA) was used to compare groups. Kaplan-Meier analysis was used to evaluate survival from time of diagnosis to death. Mantel-Cox logrank test was used to compare survival curves between groups. Cox proportional hazard method was used to perform multivariate analysis. A P value
less than .05 was considered significant for all tests. The computer program Statistica for Windows Version 5 (Statsoft, Inc, Tulsa, Okla) was used for all ANOVA evaluations and survival curve generation, and SAS software was used for Cox proportional hazards. RESULTS Demographics. Patient demographics including site of origin, gender, and age are shown in Table I. The gender distribution is consistent with prior reports in that there is a male preponderance for all gastrointestinal sites except for gastric carcinoids, which occur more frequently in women. The age at diagnosis was not significantly different between groups.
1108 Kirshbom et al
Surgery December 1999
Table II. Presenting symptoms Origin (n)
Pain
Stomach (24) CAG/PA (11) Sporadic (13) Duodenum (9) Pancreas (33)
9 (38%) 3 (27%) 6 (46%) 4 (44%) 21 (64%)
Diarrhea
Flushing
Nausea/ vomiting
Gastrointestinal bleeding/guaiac +
5 (21%) 4 (36%) 1 (8%) 1 (11%) 15 (45%)
6 (25%) 2 (18%) 4 (31%) 0 2 (6%)
4 (17%) 2 (18%) 2 (15%) 5 (56%) 2 (6%)
5 (21%) 3 (27%) 3 (23%) 5 (56%) 1 (3%)
Table III. Stage of carcinoids at diagnosis Origin (n)
Local
Stomach (26) 15 (58%) CAG/PA (11) 9 (82%) Sporadic (15) 6 (40%) Duodenum(10) 7 (70%) Pancreas (38) 3 (8%) Total (74) 25 (34%)
Regional
Distant
0 0 0 1 (10%) 2 (5%) 3 (4%)
11 (42%) 2 (18%) 9 (60%) 2 (20%) 33 (87%) 46 (62%)
Clinical symptoms and staging. The most commonly noted symptoms at presentation included abdominal pain, flushing, diarrhea, nausea and vomiting, and evidence of gastrointestinal bleeding (guaiac positive stools, anemia, or frank gastrointestinal bleeding). The incidence of these symptoms is summarized in Table II. Gastric carcinoids presented with flushing as a significant symptom more frequently than the other subtypes (25% vs 0% and 6% for duodenal and pancreatic, respectively), with the highest incidence occurring in the sporadic gastric carcinoids (31%). Diarrhea, on the other hand, occurred much more frequently in the hypergastrinemic CAG/PA group than in the sporadic carcinoids. Pancreatic carcinoids typically presented with pain and diarrhea. For the patients whose medical records were available, staging data were abstracted and are presented in Table III. Pancreatic carcinoids presented with distant metastases more frequently than the other groups (87% vs 42% and 20% for gastric and duodenal, respectively). Gastric and duodenal carcinoids presented with local disease the majority of the time (58% and 70%, respectively). Pancreatic carcinoids were significantly larger at presentation than the other groups (5.3 ± 1.0 cm vs 2.1 ± 0.3 cm and 3.1 ± 0.5 cm for duodenal and gastric, respectively, P = .009). Table III also demonstrates the stage at presentation for the gastric carcinoid subgroups. Those patients with associated hypergastrinemic states (CAG and PA) were found to have significantly lower rates of distant metastases at presentation (18%)
than did patients with sporadic carcinoids (60%, P = .03). The CAG/PA tumors were also significantly smaller than the sporadic gastric carcinoids (2.0 ± 0.7 cm vs 4.3 ± 0.8 cm, respectively, P < .05). Hormone levels. The levels of 5-HIAA and 5-HT in the urine and 5-HT in the serum and platelets are summarized in Table IV. Although the levels of all hormones were higher in the gastric and pancreatic groups than in the duodenal group, this difference was only statistically significant for the urine 5-HIAA and serum 5-HT levels. Comparison of the hormone levels in patients with locoregional disease as opposed to those with distant metastases at presentation reveals that all hormone levels are higher in the patients with distant disease (Table V). Although these differences approach significance for all groups, only the serum 5-HT groups are statistically significantly different (P = .05). Survival. The survival curves for all of the patients for whom follow-up data were available (94%), divided by site of origin, are displayed in Figure 1. Patients with pancreatic carcinoids had significantly shorter survival times, with 10-year survivals of 15% vs 58% and 59% for gastric and duodenal carcinoids, respectively (P < .0001). Survival curves for these patients divided by stage at presentation are displayed in Figure 2. There is a significant difference in survival between patients who present with locoregional disease as opposed to those with distant metastases. The 10-year survival for these groups was 10% for patients with distant disease versus 95% for those with local disease (P < .0001). Eightyear survival was 72% for patients with regional disease with no patients followed out to 10 years thus far. Multivariate analysis including age, gender, origin, and stage revealed that only stage was a significant predictor of outcome (P = .001). Origin was not found to be a factor when stage was included in the analysis. DISCUSSION Carcinoid tumors are relatively rare neuroendocrine malignancies that most commonly arise in the gastrointestinal tract. The overall incidence of
Kirshbom et al 1109
Surgery Volume 126, Number 6 Table IV. Hormone levels of carcinoid patients by site of origin Origin Stomach Duodenum Pancreas P value Normal
Urine 5-HIAA (mg/24 h) 16 ± 5 (29) 5 ± 1 (16) 47 ± 12 (54) .03 2-8
Urine 5-HT (µg/24 h) 2070 ± 1043 (27) 184 ± 65 (14) 1347 ± 314 (52) .24 50-220
Serum 5-HT (pmol/mL)
Platelet 5-HT (pmol/mg protein)
1942 ± 606 (24) 515 ± 141 (10) 2868 ± 369 (41) .02 300-2200
1899 ± 573 (21) 568 ± 133 (8) 2603 ± 428 (40) .10 100-1500
Mean ± SEM (n), P value between groups by ANOVA.
Table V. Hormone levels of carcinoid patients by stage at presentation Stage Locoregional Distant P value Normal
Urine 5-HIAA (mg/24 h) 12 ± 7 (24) 41 ± 11 (45) .07 2-8
Urine 5-HT (µg/24 h) 253 ± 88 (21) 1287 ± 369 (44) .06 50-220
Serum 5-HT (pmol/mL)
Platelet 5-HT (pmol/mg protein)
1589 ± 497 (21) 3074 ± 500 (34) .05 300-2200
1411 ± 477 (20) 2353 ± 479 (34) .20 100-1500
Mean ± SEM (n), P value between groups by ANOVA.
these tumors is approximately 1 to 2 cases per 100,000 population.3,12 In all reported series, the majority of the cases have been of midgut origin, with a relatively small percentage arising in the foregut organs. The largest series of carcinoids in the literature reported by Modlin and Sandor,3 for example, included 204 gastric and 46 pancreatic carcinoids or 5% of the total with no mention of duodenal carcinoids. As a result of the overall rarity of carcinoid tumors and the relatively small percentage of carcinoids that arise in the foregut, there has been a scarcity of information available concerning the biochemical and clinical behavior of these tumors. As with all clinical studies, the patient population must be considered before the data can be extrapolated to other groups. Many of these patients were referred to a tertiary care center for treatment of advanced carcinoid tumors. Despite this selection bias, the distribution of stages is quite similar to the SEER group review,3 with localized tumor present in 58% and 8% of the gastric and pancreatic carcinoids, respectively, in this study as opposed to 53% and 9% in the SEER study. It is also important to recognize that this is not a natural history study because these patients received multiple modalities and protocols of treatment. Recognizing these limitations, this study demonstrates several important points. First, pancreatic carcinoids have a worse outcome than other foregut carcinoids. This finding can be explained
in several ways. Pancreatic carcinoids, because of their location, tend to produce relatively vague, nonspecific symptoms resulting in larger, more advanced tumors at presentation. After correction for stage at presentation, tumor location was not predictive of outcome using multivariate analysis. Tumor stage, which is significantly more advanced in pancreatic carcinoids, strongly influences outcome. Therefore, our study suggests that pancreatic carcinoids are not intrinsically more aggressive than other carcinoids. Another interesting point is that gastric carcinoid patients exhibit the characteristic flushing response more commonly than the other foregut carcinoids (23% vs 0% and 6% for duodenal and pancreatic, respectively). This finding is surprising given the more advanced stage and higher levels of serotoninergic peptides produced by pancreatic carcinoids; however, it does lend credence to the hypothesis that the distinctive flushing response produced by gastric carcinoids is the result of a nonserotoninergic hormone (histamine, bradykinin, etc).6 Gastric carcinoids associated with CAG and PA were smaller than sporadic carcinoids and they tended to behave less aggressively (Fig 3). It has not been established whether these differences are the result of underlying biologic differences between these tumor subtypes or whether they merely represent a lead-time error resulting from increased surveillance of the CAG-PA group.
1110 Kirshbom et al
In conclusion, although foregut carcinoids have a similar prognosis when corrected for stage, site of origin is an important determinant of outcome. Despite the high levels of serotoninergic hormones produced by pancreatic carcinoids, they tend to grow to large sizes and metastasize before presentation. This results in significantly shorter life expectancy for patients with pancreatic carcinoids when compared with gastric and duodenal carcinoid tumors. REFERENCES 1. Oberndorfer S. Karzinoide tumoren des Dunndarms. Frankfurt Zeitschr Pathol 1907;1:426-30. 2. Grimelius L, Wilander E. Silver stains in the study of endocrine cells of the gut and pancreas. Invest Cell Pathol 1980;3:3-12. 3. Modlin IM, Sandor A. An analysis of 8305 cases of carcinoid tumors. Cancer 1997;79:813-29. 4. Feldman JM, O’Dorisio TM. Role of neuropeptides and serotonin in the diagnosis of carcinoid tumors. Am J Med 1986;81:41-8. 5. Moertel CG. An odyssey in the land of small tumors. J Clin Oncol 1987;5:1503-22. 6. Akerstrom G. Management of carcinoid tumors of the stomach, duodenum, and pancreas. World J Surg 1996;20:173-82. 7. Greenberg RS, Baumgarten DA, Clark WS, Isacson P, McKeen K. Prognostic factors for gastrointestinal and bronchopulmonary carcinoid tumors. Cancer 1987;60:2476-83. 8. Kirshbom PM, Kherani AR, Onaitis MW, Feldman JM, Tyler DS. Carcinoids of unknown origin: comparative analysis with foregut, midgut, and hindgut carcinoids. Surgery 1998;124:1063-70. 9. Hussain MN, Sole MJ. A simple specific radioenzymatic assay for picogram quantities of serotonin or acetyl serotonin in biological fluids or tissues. Anal Biochem 1981;111:105-10. 10. Udenfriend S, Titus E, Weissbach H. The identification of 5hydroxyindoleacetic acid in normal urine and a method for its assay. J Biol Chem 1955;216:499-505. 11. Tagari PC, Boullin DJ, Davies CL. Simplified determination of serotonin in plasma by liquid chromatography with electrochemical detection. Clin Chem 1984;30:131-5. 12. Godwin JD. Carcinoid tumors: an analysis of 2837 cases. Cancer 1975;36:560-9.
DISCUSSION Dr Hans-Dietrich Roeher (Düsseldorf, Germany). It is well known that carcinoids are accompanied by other malignancies of nonendocrine origin. You lost a couple of patients during the follow-up; was it due to other malignancies? Dr Kirshbom. There were several patients with prior and subsequent malignancies. None of the patients in the localized groups actually died of malignant disease. Several patients dropped off for nonmalignant causes, coronary disease, et cetera.
Surgery December 1999
Dr Geoffrey B. Thompson (Rochester, Minn). I was somewhat surprised by the number of patients with distant metastases in your gastric carcinoid group. In our experience with patients with CAG/PA, secondary hypergastrinemia and gastric carcinoids (ECL tumors), we have not seen nodal metastases, distant metastases, or death due to tumor. Was there anything atypical about the size or histology of these tumors, as described by Carney? Typically, these tumors are very small, less than 2 cm, and are frequently multiple. We start to worry when they get up over 2 cm or there is atypical histology, but this is very unusual. Have you looked at this in more detail? We generally do not recommend any treatment in this subgroup of gastric carcinoid patients, because spontaneous regression can occur without antrectomy. Dr Kirshbom. That also surprised me. Looking through the data, it seems that most of the patients with distant disease and CAG were very early in the series, and I question whether screening was as aggressive at that time as it was later in the series. As you noted, the majority of the patients with CAG had very small tumors. The average was less than 2 cm, somewhere around 1 cm, as opposed to much larger tumors for the sporadic cases. But again it was a relatively small percentage of patients with distant disease. Dr Norman W. Thompson (Ann Arbor, Mich). I am a little confused about the semantics. For instance, in your overall group with gastric carcinoids, I think you are including what many of us would call ECL tumors. In your group with duodenal neuroendocrine tumors, how do you define carcinoids? Were these immunohistochemically stained to rule out somatostatinomas, which are frequently found in the duodenum, and how did you differentiate them from gastrinomas? They can all be grouped together if a pathologist looks at them without knowledge of their function or specific stains. If they are not stained, then we don’t know exactly what we are dealing with or what to call them. Dr Kirshbom. These were all stained. Dr Feldman is fairly obsessive about his documentation of differentiation between pancreatic carcinoids and pancreatic islet cells, for example. He personally evaluated those patients and their slides. As far as the duodenal carcinoids, patients with Zollinger-Ellison syndrome and evidence of hypergastrinemia were excluded. These are patients who did not have evidence of elevated hormone levels. Dr Janice L. Pasieka (Calgary, Alberta, Canada). Do you have any data on chromogranin A? In the last couple of years it has become the tumor marker that many are promoting for carcinoid, and I am wondering whether you have any data on it compared with urinary 5-hydroxyindoleacetic acid. Dr Kirshbom. Unfortunately, I do not. Dr Feldman was in the opposing camp stressing the use of serotonin levels as opposed to chromogranin levels, and he did not measure them in these studies.