Frequency and Incidence
Between 1973 and 2003, a total of 101,192 pancreatic neoplasms in 101,173 patients were identified in the SEER 17 registries. Among these, 101,046 neoplasms were classified as malignant and occurred in 101,029 patients. When we restricted the search to codes of neuroendocrine histology, a total of 1385 neoplasms in 1385 patients were identified. We removed 75 patients who were classified as having tumors of poorly differentiated or anaplastic grade. Thus, a total of 1310 patients had pancreatic islet cell carcinomas in the SEER registries. This consisted of 1.3% of all patients with pancreatic cancers.
By using linked population files, we calculated the incidence of islet cell carcinoma as a rate per 100,000 per year, age-adjusted to year 2000 U.S. standard population. Because the SEER 9, 13, and 17 registries were linked to different population data sets, we computed the age-adjusted incidence rates in three time periods. The age-adjusted incidence in the SEER 9 registries between 1973 and 1991 was .16 per 100,000. For the SEER 13 registries, from 1992 to 1999, an age-adjusted incidence of .14 per 100,000 was observed. Finally, for the period covered by the SEER 17 registries, from 2000 to 2003, the age-adjusted incidence rate was .12 per 100,000. This suggests that on the basis of the current U.S. population estimate of 302 million, approximately 362 cases of malignant islet cell carcinoma will be diagnosed each year. The number of small benign islet cell tumors may be higher. Detailed incidence data by time period, sex, and race are included in Table 1.
Limited Duration Prevalence
Among the population sampled by the SEER 9 registries, 28-year limited duration prevalence for islet cell carcinoma on January 1, 2003, was estimated by the counting method to be 227 (95% CI, 199–259). These data were then projected to the general U.S. population. Data were matched by sex, race, and age to the U.S. standard population. The estimated 28-year limited duration prevalence of islet cell carcinomas on January 1, 2003, in the United States was 2705 cases. In comparison, the 28-year limited duration prevalence for all pancreatic neoplasms regardless of histology was 27,201.14 Thus, although islet cell carcinoma represented 1.3% of pancreatic cancer by incidence, it represented 9.9% of cases by the 28-year limited duration prevalence analyses.
Of the 1310 patients with islet cell carcinoma identified in the SEER database, there were 619 women and 691 men. The majority (1095 cases) were white. African American and other racial groups accounted for 134 and 78 cases, respectively. Details of patient characteristics are included in Table 2. The other racial groups included American Indian/Alaskan Natives, and Asian/Pacific Islanders. In three cases, the race was unknown. We plotted the number of cases by age group at diagnosis in Fig. 1; the peak age distribution was 65 to 69 years. However, the median age at diagnosis was 59 years.
We next examined the effect of race and sex on age at diagnosis. The median age in years at diagnosis for white, African American, and other racial groups was 60 (mean 58, SD 14.9), 55 (mean 54.5, SD 16.5) and 56 (mean 57, SD 16.5), respectively (P = .02, Fig. 2). However, there was no difference in median age at diagnosis based on sex.
Tumor Location and Hormone Production
The location of the primary tumors within the pancreas was described in 868 cases. In 442 cases, the detailed location data is not known. This is partially because the current coding system allows tumor location to be coded as islet of Langerhans, which gives no information about the location of the tumor within the pancreas. In 379 cases, the primary tumor was located in the head of the pancreas; body, tail, and overlapping groups accounted for 103, 278, and 108 cases, respectively.
By ICD-O-3 histology codes, 1117 cases (Table 2) were coded as islet cell or carcinoid (ICD-O-3 = 8150, 8240, 8241). Because these designations can include either serotonin-producing or nonfunctional tumors, it is not possible to determine the secretory status in these cases. Among the known functional tumors, gastrinoma was the most common with 73 cases (22 cases of duodenal gastrinoma not included); insulinoma, glucagonoma, and VIPoma accounted for 49, 26, and 16 cases, respectively. Finally, in 26 cases, the tumors were considered to have mixed endocrine/exocrine histology (ICD-O-3 = 8154, 8243–8245).
Next, we compared the location of the primary tumors within the pancreas by histological classification. We found significant differences in the pattern of primary tumor localization (P = .029); nonfunctional or serotonin-producing tumors (coded as islet cell and carcinoid tumors) were more likely to be located in the head (44%) than in the body (12%) or tail (31%) or overlapping (14%). Among the functional neoplasms, most insulinomas (57%), glucagonomas (53%), and VIPomas (64%) were located in the tail of the pancreas. Gastrinomas were much more likely to be located in the head of the pancreas (63%).
Of the 1310 cases, 125 (10%) were not staged (Table 2). For the remaining 1185 cases, 179 (14%) were localized; 295 (23%) were classified as regional; and 711 (54%) were classified as distant. When we compared stage by ICD-O-3 histology, we found that most carcinomas were metastatic at the time of diagnosis, including islet cell carcinomas (61%), insulinomas (61%), glucagonomas (56%), and gastrinomas (60%). A smaller percentage of VIPomas (47%) were metastatic (P < .001). This may be the result of the presence of massive diarrhea in VIPoma patients, which may bring them to medical attention earlier. The higher rate of metastases among insulinoma patients observed in this study is likely because most small insulinoma are considered benign and not reported to SEER. Next, we compared stage by tumor location and found that tumors located at the head of the pancreas trended toward a lower rate of distant disease (48% head vs. 57% body, 58% tail, and 60% overlapping) and a higher rate of regional disease (34% head vs. 27% body, 23% tail, 27% overlapping). However, the difference was not statistically significant (P = .063).
For survival analyses, we excluded 36 cases that were identified at autopsy or on the basis of death certificates only. The median overall survival for all 1274 remaining cases was 38 months (95% CI, 34–43). When compared by the log rank test, SEER stage predicted patient outcome (P < .001). The median survival for patients with localized, regional, and distant islet cell carcinoma was 124 months, 70 months, and 23 months, respectively (Fig. 3). One-year, 3-year, 5-year, and 10-year survival rates are listed in Table 3. The median survival for those cases that were not staged was 50 (95% CI, 34–66) months. Relative risk was calculated by Cox proportional modeling. Compared to the group with localized disease, patients with regional and distant disease had 1.56- and 3.50-fold of increased risk of death during the period (1973–2005) included in this study.
We then examined potential prognostic factors for survival duration that were based on data available from the SEER database. For these analyses, we stratified patients by stage. Patients with missing stage data were excluded from the analyses. When compared against the group designated as islet cell or carcinoid by ICD-O-3 histology codes, patients with gastrinoma (P = .001) and VIPoma (P = .044) had longer survival durations after adjusting for the effect of stage (Table 4), while the group with mixed histology experienced worse survival. The difference was not statistically significant, which is likely because of the small number of cases in this category.
The location of the primary tumor may have a marked effect on the time of presentation (head tumors may obstruct the bile duct and cause visible jaundice) and resectability, as well as on surgical morbidity and mortality. Therefore, we next examined the prognostic role of the location of the primary tumor within the pancreas (Table 4). In our analyses, tumors located at the head of the pancreas were less likely to be associated with distant metastasis. The rates of distant metastasis for primary tumors located in the head, body, tail, and overlapping locations were 48%, 57%, 58%, and 60%, respectively. These differences were not statistically significant (P = .063). However, once adjusted for stage, primary tumor location in the pancreatic head was associated with a worse prognosis than the pancreatic body (P = .037). Similarly, tumors in the pancreatic head tended to be associated with worse survival compared with those in the pancreatic tail (P = .056); however, the difference was not statistically significant. Patients with primary tumors that were classified as overlapping had the worst prognosis (P = .041). In some cases, an overlapping lesion may indicate a larger primary tumor that covered a larger portion of the pancreas.
Age is found to be a predictor of outcome in a variety of malignancies. We examined the effect of age at diagnosis on overall survival stratified by stage (Table 4, Fig. 4). In the Kaplan-Meier analysis, we separated the patients into two groups on the basis of median age. We observed decreasing survival with increasing age (P < .001). Similarly, we also examined the effect of age as a continuous variable in Cox proportional hazard modeling and found increasing age to be a predictor of poor outcome (P < .001).
Next, we examined the survival duration of patients with islet cell carcinoma by year of diagnosis. There has been an improvement in survival over time. Whether analyzed as a continuous variable by Cox proportional hazard modeling or divided into discrete time durations, the observed difference was statistically significant (P = .001). Sex and race did not significantly affect survival. The details of these analyses are included in Table 4.
Finally, we performed multivariate survival analyses by Cox proportional hazard modeling. SEER stage, primary tumor localization, ICD-O-3 histology groups, and age at diagnosis were entered into the model. In multivariate analysis, the ICD-O-3 histology group was not a statistically significant predictor of outcome. All other variables retained statistical significance; the most important predictor of outcome was stage. Compared to patients with localized disease, patients with regional (HR = 1.44) and distant (HR = 3.40) disease had decreased survival duration. Location of the primary tumor within the pancreas (P = .032) and age at diagnosis (P < .001) also remained significant predictors of outcome (Table 5).