- © 1999 by American Society of Clinical Oncology
Prospective Study of the Clinical Course, Prognostic Factors, Causes of Death, and Survival in Patients With Long-Standing Zollinger-Ellison Syndrome
- Fang Yu,
- David J. Venzon,
- Jose Serrano,
- Stephan U. Goebel,
- John L. Doppman,
- Fathia Gibril and
- Robert T. Jensen
- From the Digestive Diseases Branch, National Institute of Diabetes and Digestive and Kidney Diseases; Biostatistics and Data Management Section, National Cancer Institute; and Diagnostic Radiology, Warren Grant Magnuson Clinical Center, National Institutes of Health, Bethesda, MD.
- Address reprint requests to Dr Robert T. Jensen, National Institutes of Health/National Institute of Diabetes and Digestive and Kidney Diseases/Digestive Diseases Branch, Bldg 10, Rm 9C-103, 10 Center Dr MSC 1804, Bethesda, MD 20892-1804
Abstract
PURPOSE: The long-term clinical course of unselected patients with gastrinomas as well as other functional pancreatic endocrine tumors (PETs) in whom the excess-hormone state is controlled is largely unknown. To address this issue, patients with gastrinomas were assessed.
PATIENTS AND METHODS: Two hundred twelve patients with Zollinger-Ellison syndrome (ZES) were prospectively studied. All had controlled acid hypersecretion and were assessed yearly, with a mean follow-up period of 13.8 ± 0.6 years (range, 0.1 to 31 years). Annual assessments of possible factors that might affect prognosis or treatment approaches were performed, such as those for tumor size and location; the presence, location, and extent of metastases; and the occurrence of ectopic Cushing's syndrome or another PET syndrome. Deaths were categorized as ZES-related or non–ZES-related and classified into different causes.
RESULTS: Thirty-one percent of patients died, all of non–acid-related causes. One half died of a ZES-related cause; they differed from those who died of non–ZES deaths by having a large primary tumor, more frequently a pancreatic tumor; lymph node, liver, or bone metastases; ectopic Cushing's syndrome; or higher gastrin levels. The extent of liver metastases correlated with survival rate. The presence of liver metastases alone only moderately decreased survival time; however, the additional development of bone metastases or ectopic Cushing's syndrome markedly decreased survival rate.
CONCLUSIONS: In ZES, gastrinoma growth is now the main single determinant of long-term survival, with one half of patients dying a gastrinoma-related death and none an acid-related death. Large primary tumors that are pancreatic in location, the development of liver metastases, (especially if associated with bone metastases or Cushing's syndrome), and the extent of liver metastases are all important prognostic factors. The identification of these factors allows the recognition of subgroups that can be used to tailor antitumor treatment approaches.
LITTLE IS KNOWN ABOUT the clinical course of patients with long-standing pancreatic endocrine tumors (PETs), such as gastrinomas. These tumors are not only uncommon, but until recently many patients died of the effects of the excess-hormone state because satisfactory treatments were not available.1-4 Even for patients with Zollinger-Ellison syndrome (ZES), for which total gastrectomy could control the gastric acid hypersecretion,1,5-7 the data from patients with long-standing disease were limited because the mortality of the procedure was high and many patients died before diagnosis.1,5-8 Furthermore, in these older studies, sensitive tumor-imaging methodologies were not available to allow careful disease staging.1,9
At present, a large number of treatment options are available for these patients, including medical control of the excess-hormone state,3,4,10,11 curative resection,3,12 cytoreductive surgery for patients with advanced disease,13,14 treatment of malignant disease with embolization and chemoembolization,15,16 chemotherapy, 3,17,18 interferon,3,19 long-acting somatostatin analogs,3,20 or liver transplantation.21-23 When to apply a given option, how long treatment should be continued, and in what subgroup of patients a particular treatment option should be used is, in general, unclear.3 This confusion has occurred in large part because of a lack of understanding of the clinical course of the tumor per se, the risk factors that correlate with survival time, and the definitions of the causes of death in patients not dying of complications of the inadequately treated excess-hormone state.
To address these questions in the case of gastrinomas, which are one of the most common functional PETs,2,3,24 we report here the results of the analysis of the clinical courses of 212 patients with ZES who had been prospectively observed since 1978. No patients died of acid-related complications; therefore, the course of the disease, independent of the excess-hormone state and the complications of gastric-acid hypersecretion, could be analyzed for identification of the prognostic factors contributing to survival time, the causes of death, and the patient subgroups with altered survival times whose recognition could affect treatment approaches.
PATIENTS AND METHODS
Between December 1978 and December 1997, 212 consecutive patients with ZES were included in this prospective study at the National Institutes of Health (NIH) (Fig 1). This study group represents all patients with ZES who took part in the ongoing Digestive Diseases Branch evaluation of patients with ZES that was approved by the Clinical Research Committee of the National Institute of Diabetes and Digestive and Kidney Diseases, NIH. Seven patients with ZES who were evaluated at the NIH were not included in this protocol, either because of a failure to follow the protocol or because the responsibility for their care was returned to their referring physician and no follow-up studies were available.
The diagnosis of ZES was established as previously defined.25,26 Serum gastrin samples were drawn during fasting states, and levels were determined by Bioscience Laboratories (New York, NY) and Mayo Laboratories (Rochester, MN) as described previously.26 Measurement of basal acid output (BAO) and maximal acid output (MAO) after pentagastrin injection were determined as previously described.27 Secretin and calcium provocative tests were performed as previously described.25
Diagnostic criteria for the presence of multiple endocrine neoplasia, type 1 (MEN-1) in a patient with ZES included establishment of the diagnosis of ZES, as described above, in combination with clinical and/or biochemical evidence for hyperparathyroidism, anterior pituitary disease, or a family history compatible with the MEN-1 syndrome as previously described.2,28 Secondary PET syndromes were defined as previously described and involved the development of a second symptomatic pancreatic endocrine syndrome after the initial presentation of the ZES, including adrenocorticotropin-releasing tumors, vasoactive intestinal peptide–secreting tumors, growth hormone-releasing factor–secreting tumors, insulinomas, glucagonomas, somatostatinomas, or carcinoid syndrome.
Specific Protocol
During an initial evaluation at the NIH, all patients underwent a comprehensive interview and physical examination that paid particular attention to the history and presence of symptoms compatible with gastric acid hypersecretion.2,29 Patients were also questioned with respect to their personal or family history of nephrolithiasis or other symptoms compatible with the presence of the MEN-1 syndrome.28 The onset of ZES was defined as the time of onset of continuous symptoms that were compatible with gastric-acid hypersecretion, as previously described.29 The time of diagnosis of ZES was defined as the time at which the patient first had laboratory values which determined that the symptoms were compatible with this diagnosis. The time from determination of the extent of the disease was calculated from the first time the patient underwent conventional imaging studies (computed tomography [CT] scan, magnetic resonance imaging [MRI], or angiography) to determine whether liver metastases were present. In most patients, the time at which the diagnosis was established and the time at which the first determination of disease extent was performed were the same as the time at which the initial NIH evaluation was performed. For analysis, the duration of ZES was calculated from the time of disease onset, the time of diagnosis until the death of the patient, or to December 31, 1997.
During the initial evaluation at the NIH, the patient underwent an evaluation of acid secretion (BAO, MAO, adequate control of acid secretion by oral and parenteral antisecretory agents), serum gastrin studies to confirm the diagnosis (fasting gastrin levels, secretin and calcium provocative tests), and studies to determine the presence or absence of MEN-1 (urinary and plasma cortisol levels; prolactin level; MRI or CT of the sella turcica; family history; serum and ionized calcium and parathyroid hormone levels). Initially, and at each subsequent NIH admission, patients were screened for Cushing's syndrome, or were screened if symptoms developed that suggested the presence of such a syndrome, by measurment of urinary cortisol, plasma am and pm cortisol, and adrenocorticotropic hormone (ACTH) levels. If these values suggested the presence of Cushing's syndrome, dexamethasone suppression studies, corticotropin-releasing factor testing, and/or petrosal sinus sampling for ACTH and cortisol levels were performed as previously described.30-32 In patients with metastatic disease, the possible presence of a secondary pancreatic endocrine tumor syndrome was assessed during each admission by determination of serum insulin, proinsulin, growth hormone, and glucagon levels, as well as urinary 5-hydroxyindoleacetic acid levels. If secretory diarrhea developed, plasma vasoactive intestinal polypeptide levels were obtained. If any symptoms compatible with the somatostatinoma syndrome developed,33 plasma somatostatin levels were obtained.
Investigations to define the extent and localization of the gastrinoma included esophagogastroduodenoscopy, using an Olympus videoendoscopy system (Olympus Inc., Lake Success, NY). Tumor-imaging studies were performed as previously described9,26,29,34,35 and included a bone scan, abdominal ultrasound, CT with and without intravenous and oral contrast, MRI of the abdomen, and selective abdominal angiography. Since June 1994, somatostatin receptor scintigraphy (SRS) was performed using 6 mCi of indium111[pentetic acid dPhe1] octreotide with single-photon emission CT imaging at 4 and 24 hours after injection as described previously.35,36
If, on the basis of imaging studies, patients were suspected to have metastases to the liver, the diagnosis of metastatic gastrinoma to the liver was confirmed by either CT- or ultrasound-guided percutaneous liver biopsy, laparoscopic-guided biopsy, or during a minilaparotomy.29 In contradistinction, if there was neither evidence of metastatic disease present during the initial evaluation nor the presence of MEN-1, patients underwent an extensive exploratory laparotomy.34,37 One hundred forty-three of the 212 patients underwent explaratory laparotomy for a possible cure. Since 1987, patients have had duodenal transillumination38 and duodenotomy performed routinely at the time of laparotomy.34,37 To determine the disease status after resection but before discharge from the hospital, fasting serum gastrin levels were determined on at least 3 separate days, and a secretin provocation test was performed.26 Postoperatively, patients were assessed at 3 to 6 months and then yearly with studies of drug-controlled acid secretion, studies to determine disease status (fasting gastrin levels, secretin test), and imaging studies (CT, MRI, ultrasound).34,37 Patients with liver metastases were assessed every 3 to 6 months.29,39 Evaluations included assessments of acid-secretion control and studies to assess the growth of the primary tumor and/or metastatic disease (ultrasound, CT scan, MRI, and, if the results were unclear, selective abdominal angiography and bone scanning). Since 1994, patients have also undergone SRS and MRI of the spine to determine the presence of bone metastases.39 Any positive imaging results for possible bone metastases were re-examined every 3 months and—on the basis of clinical history, histologic examination, growth before or after detection, and development of other bone lesions—were classified as to whether they were or were not bone metastases, previously as described.39 If the diagnosis still remained unclear, bone biopsies were performed.39 Patients who did not have tumor resections or liver metastases were re-evaluated yearly.
In patients with liver metastases, after histologic confirmation, no anticancer treatment was given initially, and the growth of the liver metastases was evaluated by repeated imaging studies in 3 to 6 months.39-41 If no growth was seen in recent imaging studies, growth was reassessed at 3- to 6-month intervals. If growth was seen, patients were treated with either interferon (5 × 106 units/day; n = 17)41 or chemotherapy (streptozotocin, fluorouracil, and doxorubicin; n = 29), as previously described.40 Patients who initially had metastases limited to one lobe of the liver or that were considered to be potentially resectable were considered for exploratory laparotomy and partial hepatic resection.14,42 Twenty patients in the study presented here underwent such surgery. For the analysis of the effect of the extent of hepatic metastases on survival time, patients were stratified into four groups: patients with no hepatic metastases, patients with hepatic metastases limited to one hepatic lobe (single-lobe metastases), patients with hepatic metastases in both lobes (< 5 per lobe) (both-lobe metastases), and patients with metastases distributed throughout the liver that were too numerous to be counted. To assess the effect of time of development of hepatic metastases on survival time, patients were stratified into three groups: patients who did not initially have metastases and who had not developed hepatic metastases by the time of the follow-up evaluation; patients who had hepatic metastases present on the initial evaluation (liver metastases when first seen); and patients who did not initially have hepatic metastases but who had developed liver metastases that were histologically proven during a follow-up evaluation (developed liver metastases).
No patients were lost to follow-up. The deceased patients were categorized as to whether the death was ZES-related or not ZES-related, based on whether the ZES was a direct contributing factor in the patient's death. Death due to ZES was defined as a death due to metastatic spread of tumor (n = 29), tumor-related complications (n = 13), or a failure to control gastric-acid hypersecretion (n = 0), leading to death. The causes of ZES-related deaths were further subcategorized into seven groups, including the following: death due to liver metastases with progressive inanition, death due to the development of a secondary pancreatic endocrine tumor syndrome, death due to liver metastases causing hepatic failure, death due to liver metastases with secondary infection, death due to tumor-related embolism, death due to postoperative complications of surgery to resect the tumor, and death due to the presence of a cardiac gastrinoma with arrhythmia. The causes of non–ZES-related deaths were further subcategorized into 11 categories, including the following: death due to alcohol-related illness, the development of an additional malignancy whose occurrence could not be related to the presence of Zollinger-Ellison syndrome but which lead to death, cardiac diseases, lung disease/pneumonia, drug-related causes (including cocaine or other narcotics overdose) that were not related to treatment of advanced metastatic disease, sepsis not related to the presence of ZES, liver disease caused by hepatitis C complicated with hepatic and renal failure, cerebrovascular accidents, pulmonary embolism not related to the presence of ZES, AIDS, and an undetermined exact cause known not to be due to the ZES.
Statistics
Estimated probabilities of survival (Figs 2 through 5) were calculated and plotted according to the method of Kaplan and Meier.43 For the analysis of the data listed in Tables 1, 2, and 3, the Wilcoxon rank-sum test and Fisher's exact test were used. Associations with survival in Table 1 were confirmed by testing with the Cox proportional hazards model.44 The competing risk analysis of ZES-related and non–ZES-related deaths (Fig 2) was made using method A of Lunn and McNeil,45 with cause of death as the only covariate. This analysis tests whether the two cause-specific hazard rates are equal.45 Other survival differences (Tables 4 and 5; Figs 3 and 5) were assessed using the likelihood ratio test of the Cox proportional hazards model.44 Confidence intervals for survival (Tables 4 and 5) were calculated by Rothman's method.46 Consistency with the proportional hazards model (Figs 4 and 5) was confirmed by Chappell's method.47 For the comparison of the survival times of the patients with no liver metastases and those who later developed liver metastases (Fig 3), a conservative estimate of the effect of liver metastases was obtained, using the likelihood ratio test of the Cox proportional hazards model. These data were also analyzed using the likelihood ratio test of the time-dependent Cox proportional hazards model, with the development of liver metastasis as the only covariate. This method was also applied to the survival analyses involving the development of bone metastases and ectopic Cushing's syndrome (Fig 5). The P values in Tables 1 through 5 were also analyzed after correction by the Hochberg's method48 to correct for multiple comparisons.
RESULTS
Clinical, Laboratory, and Tumor Characteristics
Two hundred twelve consecutive patients with ZES were enrolled onto this prospective study, which has lasted 18 years (Fig 1). Sixty-seven patients died during this time; 33 deaths were related to ZES, whereas 34 patients died of unrelated causes (Fig 1). One hundred forty-five patients were alive on the day of completion of the study (Fig 1). The study population resembled other large series of patients with gastrinoma in that it had a slight male predominance (61%), in the percentage of patients with MEN-1 (19%), a mean age in the sixth decade (54 ± 1 years), and in the laboratory studies, with respect to mean fasting gastrin, BAO, and MAO levels.1,2,49,50 The mean duration of follow-up was 13.8 ± 0.6 years since the time of onset of ZES and 8.7 ± 0.4 years from diagnosis, demonstrating a mean delay in diagnosis of 5.1 years from the onset of symptoms (Fig 1; Table 1).
Summary of patients' outcome and disease duration. Of the 212 patients with gastrinoma and ZES included in this study from December 1978 to December 1997, 67 patients died. Deaths were classified as ZES-related or not ZES-related.
Comparison of Clinical and Laboratory Characteristics of Patients Dying of ZES- and Non–ZES-Related Causes*
When the clinical and laboratory characteristics of the 145 living and 67 deceased patients were compared (Table 1), there were no differences in mean age, sex, frequency of MEN-1 incidence, and MAO (Table 1). They did differ in disease duration, BAO, fasting gastrin level, location of the primary tumor, the type of antisecretory drug the patient was last treated with, whether prior gastric acid–reducing surgery or gastrinoma resection was performed, and the presence of liver metastases. Deceased patients had a shorter disease duration from diagnosis or onset, had higher BAO and fasting gastrin levels, were more likely to have had a primary tumor in the pancreas and less likely to have had one in the duodenum or other site, were less likely to have been last treated with omeprazole, were less likely to have had prior gastrinoma resection, and were more likely to have had liver metastases or to have undergone a prior gastric acid–reducing surgical procedure (Table 1). Fifteen patients died before the use of H+-K+ ATPase inhibitors at the NIH in April 1984, and if these patients are removed from the analysis, the proportion of patients taking omeprazole was not significantly different between alive or deceased patients (125 of 145 [86%] v 40 of 52 [76%]; P = .12).
Among the deceased patients, 33 patients died of ZES-related causes and 34 patients died of non–ZES-related causes (Tables 2 and 3). Regarding clinical or laboratory characteristics, there were no differences between the two groups in the percentage with MEN-1, acid-secretory rates (basal or maximal), antisecretory drugs used, the duration of omeprazole treatment, or the percentage who had prior gastric acid–reducing surgery (Table 2). In comparison to patients without ZES-related deaths, patients with ZES-related deaths were younger, more likely to be female, tended to have a shorter disease duration from diagnosis (P = .023) or onset (P = .0007), and tended to have a higher fasting gastrin level (P = .024) (Table 2). In comparing tumor characteristics, patients with a ZES-related death were more likely to have a gastrinoma in the pancreas (P < .0001) and less likely to have a duodenal gastrinoma, and the tumor was likely to be larger (81% > 3 cm) (P < .0001), compared with patients dying of non–ZES-related causes (Table 3). Patients dying of ZES-related causes also had more extensive disease, with a higher percentage of the patients having lymph-node metastases (P = .028), liver metastases (P < .0001), or bone metastases (P < .0001); and were more likely to develop ectopic Cushing's syndrome (P = .0049) but less likely to have another nonendocrine malignancy (Table 3). The rate of occurrence of a secondary pancreatic endocrine tumor syndrome (excluding ectopic Cushing's syndrome) did not differ between patients who died of ZES- or non–ZES-related deaths (P = .17).
Comparison of Clinical and Laboratory Characteristics of Patients Dying of ZES- and Non–ZES-related Causes*
Comparison of Tumor Characteristics of Patients Dying of ZES- and Non–ZES-Related Causes
Survival Rates
The overall survival times of all patients from the time of disease onset are shown in Fig 2, calculated as death due to a ZES-related cause, death due to a non–ZES-related cause, and death due to any cause (Table 4). The survival probability was excellent, with a 15-year survival rate of 83% for ZES-related deaths, 89% for non–ZES-related deaths, and 74% for any cause, which were not significantly different (Table 4).
Survival rates, considering death due to ZES-related causes or death due to non–ZES-related causes as outcomes. Survival times were calculated from onset of disease (top) or from time of diagnosis (bottom). There were 33 ZES-related and 34 non–ZES-related deaths. There was no significant difference between ZES- and non–ZES-related death rates, either with survival from disease onset (P = .9) (top) or from diagnosis (P = .96) (bottom).
Effect of the Presence of Liver Metastases or the Extent of Liver Metastases on Survival Rates in Patients With Gastrinomas
Previous studies have demonstrated that the presence of liver metastases is the most important determinant of long-term survival in patients with ZES.2,8,29,51,52 However, some patients had liver metastases at the time of their initial evaluation (n = 37), whereas other patients developed liver metastases (n = 17) during the follow-up period (6.6 ± 0.6 years [range, 0.1 to 19.4 years] ). We therefore compared the survival rates in patients who never had liver metastases, who had developed liver metastases during follow-up period, or who had liver metastases when initially evaluated (Fig 3; Table 4). Patients who did not develop liver metastases or who did not have liver metastases at the time of their initial evaluation had an excellent long-term survival rate of 93% at 15 years after diagnosis (95% confidence interval, 84% to 97%) (Table 4). Patients who developed liver metastases had a worse survival rate than did patients who never had liver metastases, with a 15-year survival rate of 68% (P < .024) (Table 4; Fig 3). In a time-dependent analysis using the date of diagnosis of liver metastases, the decrease in the survival rate was highly significant (P = .0010). The patients with liver metastases when first seen had the worst survival times, with a 10-year survival rate of only 26%, which was significantly lower than those for patients who developed liver metastases (P = .0002; P = .00002 by time-dependent analysis) and for patients who never developed liver metastases (P < .00001) (Table 4; Fig 3).
Effect of the presence or absence of liver metastases or their development on survival. Survival rates were evaluated using death due to ZES-related causes as the outcome. There were 158 patients with no liver metastases (|a9), and six of whom died; 17 patients (•) who did not initially have liver metastases developed liver metastases during follow-up period, and four died; and of 37 patients with liver metastases present (█) during an initial evaluation, 23 patients have died.
Patients with ZES as well as those with other malignant PETs3,42,53 may have different degrees of metastatic disease in the liver, and it is unknown what effect this may have on survival time. We therefore stratified our patients into those who had metastatic disease limited to a single lobe, those who had limited metastases in both lobes of the liver, those who had diffuse metastases throughout the liver, and those who had no metastatic liver disease (Fig 4; Table 4). The patients without liver metastases had a significantly better survival time than did the patients with liver metastases limited to a single lobe (15-year survival rate, 93% v 58%; P = .027) or limited to both lobes (15-year survival rate, 93% v 67%; P = .036), as calculated using the likelihood ratio test of the Cox proportional hazards model (Fig 4; Table 4). The patients with diffuse liver metastases had the worst survival times, with a 10-year survival rate of only 16%, which was significantly worse than those for patients with limited one- or two-lobe disease (P = .0004) or for patients without liver metastases (P < .00001) (Fig 4; Table 4).
Effect of the extent of the liver metastases on survival. Survival rates were evaluated as described in Fig 3. Of 158 patients with no liver metastases, six died (•); of 14 patients with metastases in a single lobe of the liver, three died (◯); of 13 patients with limited discrete lesions in both lobes, three died (|a9); and of 27 patients with diffuse metastases, 21 died (█).
Bone metastases or the development of ectopic Cushing's syndrome has been reported in small numbers of patients with ZES or other malignant PETs; however, their effects on survival time are unclear.11,30,31,39,54-59 To investigate the effects of these two variables on survival time, the survival curves for patients with or without bone metastases or ectopic Cushing's disease or both together were determined (Fig 5). Twenty-four patients had bone metastases and/or ectopic Cushing's syndrome, and all of these patients also had liver metastases. Twenty patients (9.4% of total) had bone metastases and nine patients (4.2% of total) had ectopic Cushing's syndrome, with 15 patients having bone metastases alone, four patients having ectopic Cushing's syndrome alone, and five patients having both bone metastases and ectopic Cushing's syndrome (Fig 5; Table 3). Bone metastases occurred in 58% and ectopic Cushing's syndrome in 21% of patients dying of a ZES-related cause, which was significantly higher than the occurrence in patients dying of non–ZES-related causes (bone metastases [0%; P = .0001; Table 3] or ectopic Cushing's syndrome [0%; P = .005; Table 3]). Bone metastases or ectopic Cushing's syndrome usually developed late in the disease course, as evident because the mean duration of survival was only 1.9 ± 0.4 years after the diagnosis of bone metastases and 1.7 ± 0.4 years after the onset of ectopic Cushing's syndrome (Table 3). Survival times were compared in the 27 patients with liver metastases but without Cushing's syndrome or bone metastases with those seen in the patients with liver metastases, bone metastases, and/or ectopic Cushing's syndrome (Fig 5; Table 5). Survival rates in the 27 patients with liver metastases without either bone metastases or Cushing's syndrome or both were excellent, with the 10-year survival rate being 83% (95% confidence interval, 62% to 93%) (Fig 5; Table 5), and were significantly better than those for the 20 patients with bone metastases (P < .0001; relative risk, 6.0) and also significantly better than those for the nine patients with ectopic Cushing's syndrome (P = .041; relative risk = 2.7) (Fig 5; Table 5). Therefore the presence of bone metastases or ectopic Cushing's syndrome significantly decreased survival times. The development of bone metastases only significantly (P < .0001) decreased survival times in patients with liver metastases (10-year survival rate, 22% v 83%, respectively; Table 5) as did the development of ectopic Cushing's syndrome alone (P = .027) (10-year survival rate, 0% v 83%, respectively; Table 5). The development of both bone metastases and ectopic Cushing's syndrome significantly decreased survival times in patients with liver metastases (P = .0078); however, survival times for these patients were not significantly worse than those for patients with either bone metastases or ectopic Cushing's syndrome alone (Table 5). In the time-dependent analysis using the dates of diagnosis of bone metastases and ectopic Cushing's syndrome, the decreases in survival time were more significant than those in the analyses based on groupings determined by the patient's eventual diagnosis (P < .00001 for bone metastases only, P = .012 for Cushing's syndrome only, P = .0002 for both, P < .00001 for any bone metastases [with or without Cushing's syndrome], and P = .00005 for any Cushing's syndrome [with or without bone metastases]).
Effect of the presence of bone metastases, liver metastases, or ectopic Cushing's syndrome alone or in combination on survival. Survival were calculated as described in Fig 3. Of 27 patients with liver metastases only, three died; of 15 patients who developed bone metastases, 14 died (•); of four patients who developed Cushing's syndrome alone, three died (|a9); and of five patients who developed both bone metastases and Cushing's syndrome, four died (█).
Effect of the Presence of Liver Metastases, Bone Metastases, or Ectopic Cushing's Syndrome Alone or in Combination on Survival Rates in Patients with Gastrinoma
All of the P values listed in Tables 1 through 5 are two-tailed and are not corrected for the number of tests performed. By use of Hochberg's48 correction method, the survival comparisons in Table 1 with P values entered are all significant at least at the P < .05 level after correction. In Tables 2 through 5, P values between .01 and .05 should be considered suggestive until they are confirmed in independent data.
Causes of Death
Now that gastric-acid hypersecretion can be controlled medically on an acute and long-term basis in almost all patients with ZES, it has been suggested that the natural history of the gastrinoma will become an increasingly important determinant of survival.2,60,61 However, there is little data available for assessment of this proposal. To provide data on the natural history of such patients late in their disease courses, we analyzed the causes of death in detail (Tables 6 and 7). No patient died of an acid-related cause, as in the past,1,2,8 despite the fact that only five (2.3%) of 212 patients had a total gastrectomy, demonstrating the long-term effectiveness of current medical management. In total, 67 (32%) of the 212 patients died, 33 (16%) of whom died of causes directly attributable to ZES and 34 patients of non–ZES-related causes (Fig 1; Tables 2 and 3).
The causes of death in the 33 patients dying of a ZES-related cause are listed in Table 6. The most common cause of death was tumor-induced cachexia leading to wasting and progressive weakness (70%) (Table 6). This only occurred in patients who had extensive liver metastases. The development of a second PET syndrome contributed to the death in ten patients (30%), with seven patients developing ectopic Cushing's syndrome because of the release of ACTH by the gastrinoma, two patients developing insulinomas that directly contributed to their deaths, one patient developing a glucagonoma, and one patient developing both ectopic Cushing's syndrome and the carcinoid syndrome. Four patients (12%) had progressive hepatic failure secondary to extensive liver metastases that contributed to their death. In 6% of patients (two patients) with liver metastases, a tumor-related infection or development of a pulmonary embolism contributed to their deaths. One patient died postoperatively of a cerebrovascular accident secondary to paradoxical emboli from the lower extremities through a previously unrecognized patent foramen ovale, and one patient with a primary cardiac gastrinoma62 died of a cardiac arrhythmia after refusing to consider heart transplantation.
ZES-Related Causes of Death
The causes of death in the 34 patients who died of non–ZES-related causes are listed in Table 7. Two of these patients had liver metastases that did not contribute to their deaths, and one patient had an insulinoma that was successfully resected and did not contribute to his death (Table 3). None had ectopic Cushing's syndrome or bone metastases. The most common contributing factors were alcohol-related diseases (29%) and the development of an additional malignancy (29%). In the ten patients who developed an additional malignancy, four patients developed oronasopharyngeal cancer; two patients, lung cancer; one patient, colon cancer; one patient, prostate cancer; one patient, breast cancer, and one patient, leukemia. Cardiac diseases (myocardial infarction, arrhythmias, and cardiac arrest) contributed to death in four patients (12%). Six percent of patients (two of 34) died of advanced chronic obstructive lung disease/pneumonia. Another 6% of patients (two of 34) died of either a cocaine overdose or another drug. Other causes of non–ZES-related death include one death due to sepsis (3%), one death due to nonprogressive liver disease caused by hepatitis C complicated with hepatic and renal failure (3%), one death due to a cerebrovascular accident (3%), one death due to pulmonary embolism (3%), and one death due to AIDS (3%). In one patient, the principal cause that resulted in death was undetermined, but it was not thought related to be ZES-related.
Non–ZES-Related Causes of Death
DISCUSSION
The purpose of this study was to examine the clinical course of patients with gastrinomas who were observed on a long-term basis or until death to attempt to define the causes of death and to identify the risk factors that alter prognosis or survival. The areas dealt with in this study have important implications for the selection of the appropriate management of patients with gastrinomas, and they may also be applicable to other malignant PETs, which are reported to resemble gastrinomas in their malignant potential, tumor biology, tumor pathology, metastatic patterns, and antitumor treatment.2,3,63,64 There are numerous treatment options reported for advanced gastrinomas and other PETs, including chemotherapy,2,3,17,18,65 embolization and chemoembolization,2,3,15,16 cytoreductive surgery,2,3,14,16 treatment with long-acting somatostatin analogus,2,3,20,65 treatment with interferon,2,3,19 and liver transplantation.2,22,23,66,67 Which treatment option to recommend and, more importantly, when to recommend that it be used is unclear at present.2,3 This lack of agreement exists primarily because of a lack of detailed information on the clinical course of patients with advanced disease. For example, it is not known at present whether the majority of patients have a tumor-related death or die of unrelated causes, or what the exact risk factors for tumor-related deaths are, and therefore it is not possible to determine how aggressive any treatment should be at a given disease stage. There are a number of reasons for this lack of information. First, even though gastrinomas are the most common symptomatic, malignant, PETs,2,3 gastrinoma is still an uncommon disease, with an incidence of one to three new cases per year per million people.2,11 The other malignant PETs are even less common.2,3 Because of the rarity of these tumors, it has been difficult for one center to have sufficient numbers of patients to allow long-term studies of the clinical course of the advanced disease to be conducted. Several studies in patients with gastrinomas have focused on overall disease survival2,3,29; however, they provide little detailed data regarding the disease course and modifying factors in patients with advanced disease. Second, adequate treatment to control the symptoms caused by the excess-hormone state has only existed in recent years with the availability of potent gastric antisecretory drugs for ZES and octreotide (Sandostatin; Novartis, East Hanover, NJ) for other functional PETs.2,3,10 Therefore, until recently, the excess-hormone state was the primary determinant of survival, which prevented long-term survival studies.1-3 Recently, especially in patients with gastrinomas, it has become possible to control gastric-acid hypersecretion in all patients, therefore preventing the development of lethal peptic ulcer disease complications; hence, the clinical course of the gastrinoma can be determined.2,3,68 Third, in older studies, imaging studies that were sufficiently sensitive to detect the extent of tumors either were not used or did not exist at the time of the study. Therefore, in these studies, it was not possible to carefully correlate tumor burden or location with survival time or to assess tumor growth and metastatic pattern in patients with advanced disease. Recently, with the introduction of MRI with short TI inversion recovery images9,69 and SRS, the localization of the primary tumor and assessment of tumor extent has been greatly improved and, therefore, tumor extent and its relation with disease course can be carefully assessed.9,35,39 Fourth, in the past, most patients with gastrinomas were diagnosed late in their disease course, after they had developed full-blown peptic-acid complications or metastatic disease.1,2,70 With the increased awareness of the disease and the widespread availability of gastrin radioimmunoassays, diagnosis of gastrinoma is frequently made earlier,70 which enables physicians to detect individual gastrinomas biochemically before extensive metastatic spread takes place. Early detection allows better assessment of the risk factors that may lead to the development of progressive disease. Fifth, previous studies involving small numbers of patients with gastrinomas have suggested that the presence of bone metastases, liver metastases, ectopic Cushing's syndrome, unresectable disease, and the development of a secondary PET syndrome or the presence of MEN-1 all may be important risk factors affecting survival.2,8,30,39,59,71,72 However, other studies have not found these to be important risk factors,2,73 and their possible significance therefore remains unclear. Last, because of the prolonged survival times of patients with malignant PETs, including gastrinomas,2,51,52,74 compared with those who have nonendocrine tumors, long-term follow-up is required to assess the clinical course of the PETs. Very few such studies have been done, and those that are available involve only small numbers of cases, limiting detailed analysis. The study presented here has none of the above limitations. First, it involves a large number of cases (212 patients) observed longitudinally in one institution, with reassessments at least annually and sometimes more frequently. Second, the acid secretion was well controlled by antisecretory drugs in all phases of the disease, and none of the patients died of acid-related complications. Therefore, the clinical course of the gastrinomas in patients with advanced disease could be assessed for the first time. Third, detailed imaging studies were performed at least yearly, so accurate data exists to compare tumor location and extent to outcome. Fourth, less than one fourth of patients had liver metastases at presentation; therefore, patients with a complete spectrum of the disease, both with advanced and nonadvanced tumor burdens, were observed long-term. Fifth, the presence of bone metastases, ectopic Cushing's syndrome, liver metastases, unresectable disease, and development of a secondary PET syndrome or MEN-1 were assessed regularly using the most sensitive methods so that such conditions could be detected and assessed as possible prognostic factors. Last, all patients observed at the NIH were included in an unselected manner, and the follow-up period was long. The mean follow-up period was 14 years from the onset of the disease, or 9 years from diagnosis, with a range from 1 month to 32 years. During this long-term follow-up period, almost one third of the patients died, resulting in numbers of decreased patients sufficient for the causes of death and possible risk factors to be accurately assessed.
In our study, to identify possible disease-related prognostic factors, different parameters were compared between living and deceased patients and between patients dying of ZES-related causes and those who died of non–ZES-related courses. A number of factors, including age, sex, presence or absence of MEN-1, or MAO, did not differ between patients that were alive or dead during the follow-up. The increased proportion of pancreatic gastrinoma and increased presence of liver metastases in deceased patients are consistent with a number of previous studies that suggest that duodenal gastrinomas are less frequently associated with liver metastases, the presence of which is a very important determinant of long-term survival in almost all studies.2,29,41,75 Even though our results suggest that treatment with histamine H2-receptor antagonists or that having undergone a previous gastric acid–reducing operation is associated with an increased death rate, and a gastrinoma resection with a decreased death rate, these associations are unlikely to be informative and true because they are most likely related to different treatment approaches at different times. More than 80% of patients with liver metastases have nonresectable disease42 and therefore did not undergo surgical exploration, which would strongly contribute to the higher percentage of deceased patients not having undergone surgical exploration. Omeprazole has only been available since 1983 at the NIH and therefore all 15 patients who died before 1984 would have only been treated with histamine H2-receptor antagonists. This thus falsely increased the proportion of deceased patients who had been treated with these drugs. The lack of effect of H+-K+ ATPase inhibitor use on survival time is supported by the results of the comparison when the 15 patients who died before the availability of these drugs in April 1984 are excluded. This analysis demonstrated that the proportion being treated with these drugs was not significantly different between the living or deceased patients. Similarly, gastric acid–reducing surgery was widely used until the late 1970s; therefore, a higher proportion of older patients with ZES diagnosed before this time underwent this procedure and are now deceased.
To search further for possible predictive factors for death from the gastrinoma itself, numerous possible risk factors were compared between patients who died of ZES-related causes and those who died of non–ZES-related causes. In addition to acid-secretory rate, two variables that did not differ between the two groups were the type of antisecretory drug used or whether the patient had undergone a prior gastric acid–reducing operation. These observations support the conclusions discussed above: that differences in these variables between living and deceased patients are likely not important but rather are due to the time of availability of these drugs and the use of this operation. In previous studies, it was unclear whether the presence of MEN-1 altered the disease prognosis of a PET with some studies reporting that it did,8,24,72,76 whereas others reported that it did not.51,52,70,75,77,78 In our study, there was no significant difference between the proportion of patients with MEN-1 who died of a ZES-related death or those who died of a non–ZES-related death, or in living or deceased patients. In various studies,2,29,75,77 the increasing size and location of gastrinomas and other PETs, similar to carcinoid tumors,79 are reported to correlate with malignant potential. Our studies further support this conclusion, demonstrating that pancreatic location and a tumor diameter of greater than 3 cm but not 1 to 3 cm in size are associated with an increased risk of death from a gastrinoma. A number of previous studies2 demonstrate that both of these factors are associated with an increased occurrence of hepatic metastases, which in our study were significantly more frequent (P = .0001) in patients dying of a ZES-related cause. It has been recently proposed that a benign and aggressive form of ZES exists29,51 and that the aggressive form is associated with an increased incidence of hepatic metastases and pancreatic gastrinomas, occurs more frequently in females, and occurs in patients with shorter disease duration. In our study, each of these variables was associated with an increased risk of death due to ZES and therefore support these proposals. In some29,80-82 but not other77 previous studies, the extent of elevation of fasting serum gastrin levels has been a prognostic factor. Our results support the conclusion that higher serum gastrin levels are associated with more malignant disease and decreased survival time, because it was significantly higher in both patients dying of ZES-related causes, compared with those dying of non–ZES-related causes, and overall in patients who died, compared with patients who were alive at the end of this study.
A number of studies report that survival time in patients with gastrinomas is related to the presence or absence of liver metastases.2,8,29,49,51,52 The results in our study confirmed that liver metastases are associated with a much worse prognosis. No previous studies have considered whether the extent of liver metastases or the time of development of liver metastases are important variables affecting prognosis. Previous surgical studies have reported that in patients with gastrinomas and other malignant gastrointestinal neuroendocrine tumors with hepatic metastases, 5% to 15% of the patients had metastases limited primarily to one lobe of the liver,14,42,83 which could affect survival time differently than diffuse metastases would. In the study presented here, sufficient numbers of patients with hepatic metastases were present to allow stratification into three groups to assess the extent of hepatic metastases on survival time. The results demonstrated that different degrees of hepatic metastases had different effects on survival. The presence of hepatic metastases limited to a single hepatic lobe significantly (P = .027) decreased survival time, and the addition of limited metastases to a second hepatic lobe (< five per lobe) did not further decrease the survival time. However, the presence of diffuse metastases significantly (P = .0004) further decreased survival. These results, demonstrating decreased survival with even limited hepatic metastases, support proposals for the aggressive surgical treatment of limited hepatic metastases if complete resection is thought to be possible.3,13,42,84 Furthermore, because diffuse metastases are associated with such a decrease in survival, the question is raised of whether patients with limited metastases that are not resectable but which are growing53 should receive chemotherapy or other antitumor treatments earlier than is frequently recommended at present. Currently, because the patients are frequently asymptomatic, no antitumor treatment is usually given until symptoms develop or unless the tumor grows rapidly.2,3,85,86 Our results demonstrate that in addition to the extent of hepatic metastases, the time of development of hepatic metastases also has an important effect on survival. Patients who do not initially have hepatic metastases but who develop liver metastases during the follow-up had a significantly (P = .02) decreased survival time, compared with patients without liver metastases. However, their survival time was not as poor as that for patients who had liver metastases when first seen. These results support the importance of early surgery in patients with gastrinomas because a recent study87 demonstrated that such surgery significantly (P < .003) decreased the rate of development of hepatic metastases; however, in that study, the survival advantage by such resections was borderline significant (P = .085).87
The development of bone metastases11,30,39,54 is increasingly reported in patients with ZES and other malignant PETs. One study59 has suggested that the development of bone metastases in patients with gastrinoma is predictive of a poor prognosis. In three previous studies, 7% to 12% of patients with gastrinomas, other PETs, or carcinoid tumors that have a similar biologic behavior to that of PETs had bone metastases,24,39,88 and in the present study 9% of patients developed bone metastases. However, bone metastases occurred in 27% (18 of 67 patients) of all patients dying of any cause, in 58% (18 of 33 patients) of patients dying of ZES-related causes, and in 37% (20 of 54 patients) of patients with liver metastases. Therefore, in patients with advanced disease, bone metastases are common and should be routinely sought. Similar to that reported in small series,39,59 in the present study liver metastases were only seen in patients with hepatic metastases. In the different studies, because bone metastases in patients with ZES almost invariably occur in patients with liver metastases, which themselves are associated with a poor prognosis, it has been difficult to define whether the addition of bone metastases is associated with a worse prognosis. Our results demonstrate that the development of bone metastases significantly decreases survival time (P < .005) and that the mean survival duration was only 1.9 years after their development. These results support the conclusion that patients developing bone metastases should undergo aggressive treatment directed against the tumor if survival is to be impacted.
The development of ectopic Cushing's syndrome is reported in up to 5% of patients with ZES,11,30,31 and it has been suggested by some30 but not others73 that it is associated with a poor prognosis. In our study, ectopic Cushing's syndrome due to ACTH production by the gastrinoma developed in 4% (nine of 212) of patients. However, it was present in 17% (nine of 54) of patients with liver metastases, 21% (seven of 33) of patients dying of ZES-related causes, and 25% (five of 20) of patients with bone metastases. It was an independent predictor of poor survival (P < .005), and patients lived a mean of only 1.7 years after its onset. These results support the recommendations that all patients with metastatic gastrinomas to the liver should be monitored at least yearly not only for the development of bone metastases but also for the development of ectopic Cushing's syndrome, and, if diagnosed, these conditions, similar to the patient's tumor, should be aggressively treated.
The cause of death for unselected patients with ZES since the advent of effective medical therapy for gastric-acid hypersecretion in the 1980s is largely unknown. In most early studies, a significant percentage of patients died of acid-related causes.1,6,89,90 In patients surviving total gastrectomy, ZES-related deaths due to tumor progression were reported to be the main cause of death, occurring in 57% to 81% of patients.52,72,80,89,91 In the study presented here, no patients died of acid-related problems. Therefore, the clinical course of long-standing ZES could be evaluated in all patients for the first time, without a loss of patients due to the complications of gastric-acid hypersecretion. Our results demonstrate that during a mean follow-up period of 14 years from onset and 9 years from diagnosis, 32% of all patients will die and that in 49% of those dying, the cause of death is ZES-related. The most common cause of ZES-related death is progression of the hepatic metastases leading to progressive inanition (67% of patients with ZES-related deaths). This was followed by death due to the development of a secondary pancreatic endocrine tumor in 30% (primarily ectopic Cushing's syndrome, in 21%), hepatic failure due to extensive liver metastases (12%), secondary infections (6%), or tumor-related emboli (6%). Postoperative complications leading to death after gastrinoma surgery occurred in only one patient (3%) and are now a rare cause of death, compared with early series before effective treatment of gastric-acid hypersecretion, in which it could exceed 15% to 30%.7,90 In contrast to the early studies discussed above, non–ZES-related causes of death are at present at least as common a cause of death as ZES-related causes. Of the non–ZES-related causes, alcoholism and the development of another malignancy are equally common causes (each causing 29% of the non–ZES-related deaths). Other non–ZES-related causes of death included cardiac disease (12%), advanced lung disease or pneumonia not related to tumor growth (6%), death due to drug overdose (6%), and a single case each (3%) due to sepsis, nonalcoholic liver disease, a cerebrovascular accident, embolism, and AIDS. This increased occurrence of non–ZES-related causes of death is likely to be related to the slowly progressive nature of the gastrinoma in many patients. Even during a mean follow-up period of 15 years since the disease onset, only 26% of all patients in the present study developed liver metastases. This result further supports the suggestion that ZES can be divided into two clinical forms, depending on the clinical course of the disease.29,51,60 One form of ZES pursues a malignant course characterized by the development of liver metastases and a poor 10-year survival rate (30% in one study29) and involves approximately 25% of the patients. In contrast, the other form of ZES pursues a benign course in which liver metastases are uncommon, it is more frequently associated with duodenal gastrinomas, it has an excellent 10-year survival rate (96% in one study29), and it involves 75% of the patients with ZES. In patients without liver metastases who would have the benign form of the disease as classified above, the 15-year survival rate in this study was 93%. Therefore, with this excellent long-term survival rate in the majority of patients with ZES, and because of the usual onset of ZES in the fifth to sixth decades, increasing numbers of patients with ZES will be dying of non–ZES-related causes in the future.
In conclusion, the prospective study presented here reports for the first time the long-term late clinical course of a large cohort of unselected patients with ZES in which no patients died of acid-related problems. During the mean follow-up period of 15 years, 32% of patients died, one half of whom died of ZES-related causes. The ZES-related causes of death were more frequent in patients with higher gastrin levels; shorter disease duration from onset to diagnosis; pancreatic gastrinomas and large tumors; metastases to the lymph nodes, liver, or bone; or ectopic Cushing's syndrome. For patients with liver metastases, the development of bone metastases or tumor-related ectopic Cushing's syndrome was an independent predictor of decreased survival time. Furthermore, the extent of liver metastases and the time of their development were independent predictors of decreased survival time. Last, the causes of ZES-related and non–ZES-related deaths were determined. The identification of the clinical course of the gastrinoma and of the prognostic factors for survival provides important data to allow planning of the timing and courses of treatment.
- Received June 22, 1998.
- Accepted October 8, 1998.
