Goblet Cell Carcinoid Tumor, Mixed Goblet Cell Carcinoid-Adenocarcinoma, and Adenocarcinoma of the Appendix: Comparison of Clinicopathologic Features and Prognosis

We searched the computerized database of the Department of Pathology at MD Anderson Cancer Center (MDACC), Houston, Texas, to identify cases of GCT, mixed GCT-adenocarcinoma, or poorly differentiated adenocarcinomas of the appendix (with and without a minor GCT component) from January 1999 through October 2007. Both outside consultations and patients referred to MDACC for treatment were included in the study, but we excluded all cases in which the primary appendiceal lesion was not seen by an MDACC pathologist (ie, metastasis from an appendiceal primary but the appendix was not reviewed). The study was approved by the MDACC institutional review board.

We recorded the following clinical features, based on review of the outside paperwork, MDACC medical records, and the US Social Security Death Index: race, gender, age at first surgical procedure, presenting symptoms, type of initial surgery (appendectomy, right ileocolectomy, or other), chemotherapy recommendations, tumor recurrence, and length of overall survival.

For comparison purposes, we divided tumors containing any element of GCT into 3 groups based on the amount of adenocarcinoma component (adapted from Burke et al⁸ ): group 1, GCTs or GCTs with less than 25% adenocarcinoma (GCTs with an adenocarcinoma component comprising <25% of the tumor; Figure 1, A through C); group 2, GCTs with 25% to 50% adenocarcinoma (GCTs with an adenocarcinoma component between 25% and 50% of the tumor; Figure 2, A and B); and group 3, more than 50% adenocarcinoma (GCTs with an adenocarcinoma component >50% of the tumor; Figure 3, A and B). For comparison, we included group 4, poorly differentiated appendiceal adenocarcinoma (signet-ring cell, mucinous, solid, or glandular patterns) without a recognizable GCT component. A component of GCT was defined as the presence of small, discrete nests of cells, some with small lumens, containing an admixture of neuroendocrine cells, goblet cells, and Paneth cells in an organoid pattern. The cells had no to minimal cytologic atypia, and nuclear polarity was arranged toward the periphery of the nest/gland. The presence of a dysplastic precursor lesion (ie, adenoma) was incompatible with a GCT (unless 2 different processes could be clearly delineated). It is difficult to exclude a small component of adenocarcinomatous component in GCT because single cells or small clusters of neoplastic cells may be present from tangential sectioning or when the tumor infiltrates through the muscularis propria in seemingly single cells that resemble signet-ring cells; alternatively, it may represent a small focus of an adenocarcinomatous component. A threshold of less than a 25% adenocarcinomatous component was used in group 1 tumors in a prior study⁸  and our study because a minor component of adenocarcinoma cannot be excluded. All group-1 tumors in our study had no or a very focal (<5%) area that could be classified as an adenocarcinomatous component. When GCT was in the muscularis propria, a cohesive linear arrangement of neoplastic cells that resembled signet-ring cells was allowed. Small mucin pools in the tumors were still compatible with GCT if the tumor cells fulfilled the other criteria.

As described above, adenocarcinoma was considered to be present when there were individual dyshesive cells, solid sheets of cells, infiltrative cords of cells (not within muscularis propria or larger cords incompatible with GCT), or a complex glandular architecture (irregular, angulated glands, cribriform glands, or tufting). In addition, clusters of cells simulating GCT but with increased cytologic or architectural atypia beyond typical goblet cell carcinoid nests (enlarged or irregular nests/glands, increased cytologic atypia, increased mitotic activity) were also considered areas of adenocarcinoma. The presence of destructive invasion or desmoplasia was also considered areas of adenocarcinoma. The carcinomatous component was subtyped into signet-ring cell type and nonsignet-ring cell type based on the predominant component (>50% of carcinomatous component). All of the available sections of the appendiceal tumor were evaluated for the presence of adenocarcinomatous and GCT components, and the percentage of adenocarcinomatous component was estimated from the mean percentage of all sections.

As a control group, poorly differentiated adenocarcinomas without an identifiable GCT component of the appendix during the study period were also evaluated. Low-grade mucinous tumors (adenocarcinomas) and neuroendocrine tumors (carcinoid tumors) were excluded from this study.

In each case, we recorded the presence or absence of appendiceal perforation (defined as disruption of the wall, which was identified either grossly or as transmural necrosis by microscopic evaluation), margin status of the initial surgery, residual tumor after completion resection (if appendectomy), pathologic stage (after completion resection or debulking), metastatic sites, and (if performed) neuroendocrine differentiation by immunohistochemistry. Tumors were staged according to recommendations from the American Joint Committee on Cancer for appendiceal carcinomas.³⁷  If available, immunohistochemical stains for chromogranin and synaptophysin were reviewed.

Statistical analyses were performed using SPSS software (version 12.0; SPSS, Chicago, Illinois). Comparisons of categoric variables were evaluated by Fisher exact tests, and continuous variables were analyzed with 2-sided Student t tests. Survival was evaluated by the Kaplan-Meier method and with Cox proportional hazards models. P < .05 was considered statistically significant.

The final study population comprised 142 patients, including 23 (16%) cases in group 1; 27 (19%) in group 2; 24 (17%) in group 3; and 68 (48%) in group 4. Of those, 53 (37%) were slide consultations sent from outside institutions, 82 (58%) were patients referred to MDACC for treatment, and 7 (5%) were patients who had their primary resections performed at MDACC.

Clinical and histopathology features are compared in Table 1. Acute appendicitis (acute, right, lower-quadrant abdominal pain and tenderness) was the most frequent clinical presentation for patients in group 1 (65%; 11 of 17) and group 2 (54%; 7 of 13), compared with patients in group 3 (24%; 4 of 17) or group 4 (15%; 8 of 55) (P = .01). In contrast, chronic abdominal pain was the most common clinical presentation in groups 3 and 4, followed by vaginal bleeding, abdominal fullness, bowel obstruction, polyuria, and testicular mass, among others. Fourteen patients of those for whom presenting symptoms were documented (n = 102; 13% overall [group 1, 12%, 2 of 17; group 2, 31%, 4 of 13; group 3, 18%, 3 of 17; and group 4, 9%, 5 of 55]) were asymptomatic from their appendiceal tumor, and the tumor was discovered incidentally upon investigation of unrelated conditions or during screening procedures. There were no statistically significant differences in race, mean age at initial resection, or sex among the 4 groups.

Appendectomy was the initial surgical procedure in 20 of 23 (87%) of group 1, 19 of 27 (70%) of group 2, and 12 of 24 (50%) of group 3 patients, compared with 24 of 68 (35%) of the group 4 patients (P = .01). The adenocarcinomatous component for each group was subclassified into signet-ring cell and nonsignet-ring cell components. Group 1 tumors were not subclassified because those cases had no or only a small carcinomatous component. Group 2 had 14 signet-ring cell types (52%) and 13 nonsignet-ring cell types (48%; 3 poorly differentiated mucinous [23%], one moderately differentiated glandular [8%], and 9 poorly differentiated adenocarcinomas [69%]), group 3 had 8 signet-ring cell types (33%) and 16 nonsignet-ring cell types (67%; 6 poorly differentiated mucinous [38%], one moderately differentiated glandular [6%], and 9 poorly differentiated adenocarcinomas [56%]), and group 4 had 23 signet-ring cell types (34%) and 45 nonsignet-ring cell types (66%; 19 poorly differentiated mucinous [42%] and 26 poorly differentiated adenocarcinomas [58%]), respectively. An appendiceal tubulovillous adenoma was present in a poorly differentiated adenocarcinoma without a GCT component (group 4); no precursor lesion was present in other tumors with or without a GCT component in groups 1 to 4.

In the initial surgical specimen, serosal involvement by the tumor or extension to an adjacent organ (pT4) was present in 11 of 22 (50%) of patients from group 1; 9 of 24 (38%) from group 2, 16 of 21 (71%) from group 3, and 44 of 63 (70%) from group 4 (P = .01). All other patients presented with involvement of the periappendiceal adipose tissue (pT3), except one patient from group 4, who presented with a tumor confined to muscularis propria (pT2). Appendiceal perforation was more frequent in group 1 (39%; 9 of 23) tumors than in tumors from group 2 (19%; 5 of 27), group 3 (8%; 2 of 24), or group 4 (7%; 5 of 68) (P = .01). Margin status on the initial surgical specimen was available in 71 cases and was positive in 27 (38%). Specimens from groups 1 and 2 had higher rates of positive margins than did those from groups 3 and 4 (56% [5 of 9] and 50% [6 of 12] versus 36% [5 of 14] and 31% [11 of 36], respectively). All of the specimens from groups 1 and 2 with positive margins were appendectomies, except in one group 2 case, where both the ileal and colonic margins showed involvement by tumor. Distant metastases did not occur in group 1 patients at their initial surgical procedure but were present in 4% (1 of 27) of group 2, 67% (16 of 24) of group 3, and 78% (53 of 68) of group 4 (P = .01).

Thirty-nine of the 75 patients (52%) whose initial surgery was appendectomy subsequently underwent a major resection, such as right ileocolectomy. Residual tumor (in the intestinal wall, peri-intestinal soft tissue, and/or lymph nodes) was detected in 20 cases, including 5 of 14 (36%) in group 1; 7 of 11 (64%) in group 2; 2 of 5 (40%) in group 3; and 6 of 9 (67%) in group 4 (P = .40).

Immunohistochemistry for chromogranin A, synaptophysin, or both was available in 87 cases, of which 73 (84%) were positive (chromogranin A in 68 of 87 [78%] and synaptophysin in 59 of 76 [78%]). All 16 tumors (100%) in group 1 and 24 of 25 tumors (96%) in group 2 had immunoreactivity for neuroendocrine markers compared with 15 of 20 tumors (75%) in group 3 and 11 of 20 tumors (55%) in group 4 (P = .01). Those differences were present for both chromogranin and synaptophysin (see Table 1). Most of these tumors had patchy staining for neuroendocrine markers, and there was no significant difference in the staining pattern of goblet cell carcinoid and adenocarcinoma components (data not shown).

Tumors in groups 3 and 4 were at a more advanced stage (final stage based on the definitive resection) compared with those in groups 1 and 2. In group 1, 87% (20 of 23) of the patients had stage II tumors, and only 4% (1 of 23) were at stage IV; in group 2, 67% (18 of 27) were stage II, and 22% (6 of 27) were stage IV; in group 3, 29% (7 of 24) were stage II, and 67% (16 of 24) were stage IV; and in group 4, only 19% (13 of 68) were stage II, whereas 75% (51 of 68) were stage IV (P = .01).

Chemotherapy was recommended for 80 of the 97 patients (82%), including 6 of 16 patients (38%) in group 1; 9 of 10 patients (90%) in group 2; 14 of 17 patients (82%) in group 3; and 51 of 54 patients (94%) in group 4 (P = .01). No data were available in the other 45 cases. Intraperitoneal chemotherapy was performed in 7 patients (1 in group 1, 1 in group 2, 2 in group 3 and 3 in group 4).

Tumors recurred (among patients with available follow up [n = 44; 31%]) in 3 of 17 patients (18%) in group 1; 6 of 12 patients (50%) in group 2; 6 of 17 patients (35%) in group 3; and 9 of 53 patients (17%) in group 4 (P = .08). Mean times between initial surgery and disease recurrence were 51.5 months (range, 20–102), 34.0 months (range, 2–107), 25.8 months (range, 9–85), and 16.4 months (range, 1–61) for groups 1, 2, 3, and 4, respectively. The most common sites of recurrence were peritonealized soft tissue (including the parietal peritoneum, omentum, and abdominal wall) or pelvic organs (such as ovaries and rectum).

The overall survival was significant for tumor type and stage. The mean (SD) overall survival for patients in groups 1, 2, 3, and 4 was 83.8 (34.6) months, 60.6 (30.3) months, 45.6 (39.7), and 33.6 (27.6) months, respectively (P = .01 log-rank method; Figure 4, A). The overall survival for patients with stage II, III, and IV disease was 109.5 (8.5) months, 62.2 (15.3) months, and 32.7 (5.9) months, respectively (P = .01 log-rank method; Figure 4, B). Even after excluding group 4 patients, the difference in overall survival was significant for patients in groups 1, 2, and 3 (P = .01 log-rank method) and was borderline significant for stage (P = .048 log-rank method).

The adenocarcinomatous component for each group was subclassified into signet-ring cell and nonsignet-ring cell types. The overall survivals in group 2 to 4 were analyzed by subtype of adenocarcinomatous component. The mean (SD) overall survival of patients with signet-ring cell adenocarcinomas was 74.9 (10.6) months compared with 70.7 (10.5) months in nonsignet-ring cell types for patients in group 2 (P = .79; Figure 5, A); 25.1 (3.8) months compared with 72.3 (16.5) months for patients in group 3 (P = .01; Figure 5, B); and 34.3 (7.9) months compared with 40.6 (5.5) months for patients in group 4 (P = .34; Figure 5, C), respectively.

By univariate analyses (Table 2), clinicopathologic features that predicted overall survival included older age (P = .01), tumor category groups 1 and 2 (P = .01 and P = .01, respectively, compared with group 3), stage II and stage III (P = .01 and P = .03, respectively, compared with stage IV), and appendiceal perforation (P = .01). The difference in tumor category was present even after exclusion of group 4 tumors (P = .01) and was also present between group 1 and group 2 tumors (P = .02). By multivariate analyses, only stage and tumor category were independent predictors of overall survival using group 3 and 4 tumors as the reference group (model 1, Table 3), but the difference for patients in group 2 was of borderline significance (P = .049). However, using only group 3 as the reference group, stage and only group 1 tumor category were significant (model 2). By log-rank method and univariate and multivariate analyses, there were no significant differences in overall survival between group 3 and group 4 tumors.

The GCT has been considered a low-grade malignancy with a prognosis better than appendiceal adenocarcinoma but worse than classic carcinoid tumor.^2,5,18  In the most-recent, large study of these tumors, Tang et al³⁵  found a 77% overall disease-specific survival among 63 patients with GCT, which was more favorable than the survival of a control group of 28 conventional appendiceal adenocarcinomas, even when comparing only patients with stage IV disease. However, it is well-known that GCTs do not behave uniformly, and cases with a more-aggressive course have been described.^† Poor prognostic features have been reported to include lymph node metastases, incomplete excision at the base of the appendix, extra-appendiceal spread, increased mitoses (>2 per 10 high-power fields), and a large (>50%) component of adenocarcinoma,¹⁹  but only 2 prior studies,^8,35  to our knowledge, have provided specific criteria for risk stratification in GCTs based on histologic findings and clinical follow-up.

In 1990, Burke et al⁸  reported that all 22 GCTs with 25% or less adenocarcinoma (ADCA) were without metastasis at a mean follow-up of 19 months, whereas 8 of 10 patients with >50% ADCA had died of disease at a mean follow-up of 16 months (no tumors had 26–49% ADCA in that study), and proposed that GCTs with large carcinomatous components should be termed mixed carcinoid-adenocarcinomas. More recently, Tang et al³⁵  proposed that GCTs be divided into 3 groups: group A, typical GCTs without ADCA; group B, GCTs with dyshesive signet-ring cells or large clusters of signet-ring cell ADCA; and group C, GCTs with more than 1 mm² of gland-forming, undifferentiated, or sheetlike growth of poorly differentiated ADCA. In their study, overall disease-specific survival was 96% in group A, 73% in group B, and only 14% in group C.

Our current study of 142 appendiceal tumors further supports that the amount of carcinomatous component correlates with the clinical features and disease stage and is a major predictor of survival. We subcategorized poorly differentiated adenocarcinomas with or without GCT into separate categories but did not find any significant survival differences between the 2 groups. We classified tumors based on adenocarcinoma component using criteria similar to those of Burke et al.⁸  Initially, we hypothesized that intermediate amounts (25%–50%) of ADCA in a GCT (group 2 patients) would confer a better prognosis—similar to the better prognosis seen in GCTs (group 1). Our outcome data did not support that hypothesis and instead revealed that patients in group 2 have an intermediate overall survival that is worse than those in group 1 but marginally better than those in group 3. The patients who had signet-ring cell adenocarcinomas with a GCT component in group 3 had a worse prognosis compared with patients with other poorly differentiated adenocarcinomas with a GCT component. In contrast, Tang et al³⁵  reported better survival in patients with signet-ring cell adenocarcinomas compared with poorly differentiated nonsignet-ring cell adenocarcinomas. Differences in the patient population, histologic criteria for subclassification, or treatment modalities may account for these differences.

One possible explanation for the differences in prognosis related to GCTs may be found in the distinctive morphology of these tumors. In some GCTs with an adenocarcinoma component, the carcinoma is intimately admixed with small, tight clusters of goblet cells, rather than forming a discrete or separate focus. In other areas, the 2 components can be distinct. In addition, the adenocarcinomatous component can have multiple histologic components, including signet-ring cell, mucinous, glandular, and solid components. Even for tumors that many pathologists would consider “pure” GCT, small foci may be found that show features that fulfill the criteria for “adenocarcinoma.”^8,34  Thus, small components of adenocarcinoma are difficult to exclude in these tumors.

In addition to predicting differences in outcome, the proportion of adenocarcinoma was also associated with differences in clinical presentation and disease stage. Most (87%; 20 of 23) patients in group 1 presented with stage II disease, whereas this was true for fewer patients (69%; 18 of 27) in group 2 and only 29% (7 of 24) of patients in group 3 and 19% (13 of 68) of patients in group 4. Conversely, 75% (51 of 68) of group 4, 67% (16 of 24) of group 3, 22% (6 of 27) of group 2, and only 4% (1 of 23) of group 1 presented with stage IV disease. Abdominal pain was the most common symptom in all of 3 subsets of patients in our study, with symptoms compatible with acute appendicitis (right, lower-quadrant pain with acute onset) more commonly occurring in patients in groups 1 and 2, and chronic abdominal pain more common for patients in groups 3 and 4. Our data are consistent with previously published studies^{6,9,22,24,26–28}  that have reported abdominal pain or other symptoms of acute appendicitis as the most common presentation for patients with GCT. Others^8,14,27  have noted that when the patient presents with metastases in the lower abdomen and pelvis, the presentation may be atypical.

The surgical management of patients who have GCT diagnosed on an appendectomy specimen is controversial. Some studies^14,20,24,38  advocate further surgery in all patients, whereas others suggest additional resection only for patients with positive margins, significant atypia, increased mitoses (≥2 per 10 high-power fields), increased Ki-67 staining (>3%), clinically involved lymph nodes, or metastatic disease.^{5,8,11,18,26–28,39}  In our study, 14 of 30 patients (47%) with tumors containing a GCT component had residual disease in right colectomy specimens, either in the intestinal wall, the peri-intestinal soft tissue, and/or lymph nodes, including 36% (5 of 14) of the patients in group 1. This greater frequency of residual disease may represent a referral bias at our institution. In our opinion, right colectomy is warranted for all patients with GCT-related tumors, to ensure completeness of excision and adequate disease staging.

In our study, the most common sites of recurrence were peritonealized soft tissue and pelvic organs, including ovaries. The long length of time to recurrence in GCT has been noted by other authors, some citing recurrence as long as 16 years after initial diagnosis.^9,24  Ovarian metastases are well documented in GCT.^{6,8,12,14,18,22,25}  In one series,²⁶  50% of women had metastases at initial presentation, and in a meta-analysis³⁹  of 13 case series, 50% of metastases in women were to the ovaries. These data have prompted suggestions of prophylactic oophorectomy in postmenopausal women.^8,25,26  A few studies share their experience with chemotherapy in GCT and mixed GCT/adenocarcinoma, including intraperitoneal chemotherapy, and report that temporary control was achieved with some reports suggesting survival benefit,^{14,26,40,41,42}  and others showing a lack of response to systemic chemotherapy.³¹ 

We agree with others that these tumors represent a continuum of GCT and GCT with varying amounts of adenocarcinoma (or adenocarcinoma ex GCT).^8,35  Inclusion of the estimate of the adenocarcinomatous component with a diagnosis of adenocarcinoma ex GCT gives additional prognostic data, which may assist in further management of the patients. The prognosis of patients who have GCT with a minor component of adenocarcinoma (<50%) is slightly better compared with patients who have GCT with a major component of adenocarcinoma (>50%). The latter have prognosis similar to patients who have poorly differentiated appendiceal adenocarcinoma without a GCT component. Long-term prospective studies—preferably a collaboration of multiple centers because of the rarity of these tumors, with documentation of the ratio of goblet cell component to adenocarcinoma component—are needed so that we can further refine our understanding of the GCT/adenocarcinoma continuum and provide more-uniform estimates of prognosis and best treatment options for these patients.