Pancreatic acinar-like adenocarcinoma of the proximal stomach invading the esophagus

Human Pathology (2012) 43, 911–920

www.elsevier.com/locate/humpath

Original contribution

Pancreatic acinar-like adenocarcinoma of the proximal stomach invading the esophagus☆,☆☆ Qin Huang MD, PhD a, b, d,⁎, Jason S. Gold MD c, d , Jiong Shi MD, PhD a , Xiangshan Fan MD, MS a , Hongyan Wu BS a , Anning Feng MD, MS a , Qiang Zhou BS a a

Department of Pathology, Nanjing Drum Tower Hospital, Nanjing 210008, China Department of Pathology and Laboratory Medicine, Veterans Affairs Boston Health System, West Roxbury, MA 02132, USA c Department of Surgery, Veterans Affairs Boston Healthcare System, West Roxbury, MA 02132, USA d Harvard Medical School, West Roxbury, MA 02132, USA b

Received 26 May 2011; revised 25 July 2011; accepted 27 July 2011

Keywords: Stomach; Cancer; Pancreatic acinar metaplasia; Gastroesophageal junction

Summary The aim of this study was to systematically investigate clinicopathologic features of the recently described pancreatic acinar-like adenocarcinoma of the proximal stomach invading the esophagus (n = 43). Patient median age was 66 years (range, 51-90 years). The male-to-female ratio was 7.6. Grossly, pancreatic acinar-like adenocarcinoma tumors were nonencapsulated with the median size of 5.5 cm (range, 2-10.5). Bormann's types 1 to 4 tumors were in 7%, 9%, 67%, and 16% cases, respectively. Frank necrosis, hemorrhage, and cysts were rare or absent. Lymphovascular (81%), perineural (74%), and lymph node (81%) invasions were more common in the pancreatic acinar-like adenocarcinoma than in the non–pancreatic acinar-like adenocarcinoma (n = 94) groups. Microscopically, pancreatic acinar-like adenocarcinoma tumors showed acinar (78%), micropapillary (12%), microcystic, solid, trabecular, and mixed neuroendocrine or signet ring (33%) patterns of growth. No adenosquamous differentiation was noted in the pancreatic acinar-like adenocarcinoma group. Nuclei were round to oval with thickened nuclear membrane, stippled chromatin, and single prominent nucleoli. Mitotic figures were variable. The cytoplasm was moderate, eosinophilic, finely granular, and diffusely immunoreactive to the α1-chymotrypsin antibody in all cases to various degrees. Tumor stroma was nondesmoplastic, delicate, and fibrovascular. Pancreatic acinar-like adenocarcinoma tumors staged pI, pII, pIII, and pIV were in 2%, 21%, 70%, and 7% of cases, respectively. The median number of follow-up months after surgery was 29. The 2-year survival rate was 67%, lower than that (73%) in the non–pancreatic acinar-like adenocarcinoma group. A worse overall survival trend was found for patients in the pancreatic acinar-like adenocarcinoma than in non–pancreatic acinar-like adenocarcinoma groups, but the difference was not statistically significant. Age older than 75 years and overall pathology stage were independent risk factors. Published by Elsevier Inc.

☆ Funding sources: This research project was financed partially by grants from the Science and Technology Development Project of the Nanjing City (ZKX05013, ZKX07011), China, a special grant from the Nanjing Drum Tower Hospital, Nanjing, China, and a Career Development Award-2 from the VA Biomedical Laboratory Research and Development Service (JSG). ☆☆ Disclosure/Conflicts of Interest: No conflicts of interest to disclose for all authors. ⁎ Corresponding author. Department of Pathology and Laboratory Medicine, Veterans Affairs Boston Healthcare System and Harvard Medical School, West Roxbury, MA 02132, USA. E-mail addresses: [email protected], [email protected] (Q. Huang).

0046-8177/$ – see front matter. Published by Elsevier Inc. doi:10.1016/j.humpath.2011.07.024

912

1. Introduction In the region of the gastroesophageal junction (GEJ), pancreatic acinar metaplasia (PAM) is frequently observed in the form of lobules, aggregates, or single cells bordering mucus glands [1-3]. Its incidence ranges from 9% to 24% in biopsies [2,4-6] and 30% to 61% in surgical resection specimens [3,7]. Similar to pancreatic exocrine glands, PAM glandular cells feature uniformly round to oval nuclei with basophilic cytoplasm in the basal compartment and eosinophilic cytoplasm in the apical compartment. By immunohistochemistry, PAM cells are diffusely immunoreactive to antibodies against specific pancreatic acinar protein antigens such as α1-chymotrypsin [3,8], α1-trypsin [8,9], and, to a lesser degree, lipases and amylases [8,10]. Some cells also express both pancreatic exocrine and endocrine molecules, suggesting an aberrant stem cell differentiation [4,11]. PAM in the GEJ region is considered to result from chronic inflammation [1,3,7,12], autoimmune gastritis [13], and/or gastroesophageal reflux disease [6,14]. Some investigators believe PAM to be congenital in nature [2,14] or related to aging [6]. There is a theoretical possibility that gastroesophageal reflux disease and Barrett esophagus can be exacerbated by the pancreatic enzymes secreted by PAM glands [4,10]. Recently, in the course of investigating carcinomas of the proximal stomach with esophageal involvement in Chinese patients [15], we noted that some tumors exhibited pancreatic acinar-like architecture and that tumor cells showed dense basophilic-eosinophilic cytoplasm rich in fine granules that were diffusely immunoreactive to the α1-chymotrypsin antibody [8,15]. We hypothesized that some adenocarcinomas in the proximal stomach might have pancreatic acinar-like differentiation. Herein, we systematically investigated clinicopathologic characteristics of pancreatic acinar-like adenocarcinoma (PALA) in tumors arising in the proximal stomach with esophageal involvement in Chinese patients treated at a high-volume medical center in Nanjing, China.

2. Materials and methods 2.1. Patient selection Consecutive surgical resections with a final histopathologic diagnosis of gastric cardiac carcinoma were identified by search of electronic pathology records of the Nanjing Drum Tower Hospital for in the period from May 2004 through October 2009. Pathology reports and patient medical records were reviewed. Patients had completed preoperative workups including upper gastrointestinal endoscopic and imaging studies. Inclusion criteria consisted of: (1) the tumor epicenter was within 5 cm below the GEJ line that was defined as the proximal end of gastric longitudinal

Q. Huang et al. mucosal folds [16] and (2) the tumor invaded into the distal esophagus. Thus, selected cases fulfilled the American Joint Committee on Cancer (AJCC) criteria for cancer of the GEJ. Exclusion criteria were as follows: (1) the patient had neoadjuvant therapy before resection; (2) follow-up information was not available; and (3) archival tumor tissues were insufficient for immunostains. Patient survival information was acquired through telephone interviews of patients or immediate family members by one of the authors blinded to tumor pathology. The study protocol was approved by the medical ethics committee of the Nanjing Drum Tower Hospital in Nanjing, China.

2.2. Histopathology Pathology reports, available gross tumor images, and hematoxylin-eosin stained slides (median, 10 per case; range, 4-25) were reviewed. Gross tumor features were classified, according to the Bormann's criteria, into 4 types: 1, polypoid; 2, fungating with central necrosis and a defined border; 3, ulcerated with an infiltrative border; and 4, flat with an indistinct border [15] (Fig. 1A-C). Tumors with a napkin-ring growth pattern around the GEJ were centrally ulcerated with an infiltrative border and thus were classified as type 3 tumors (Fig. 1D). Classification of tumor histology and pathologic staging was performed based on the AJCC 7th Edition Guidelines for Gastric Cancer [17]. Adenocarcinoma, not otherwise specified, was defined as the predominant portion (N75%) of the estimated tumor mass showing glandular, acinar, tubular, papillary, trabecular, solid, or microcystic architectural patterns [15]. Adenocarcinomas with a micropapillary pattern in over 5% of the tumor volume were grouped separately for outcome analysis. Neuroendocrine carcinoma featured organoid, trabecular, nested, and solid patterns and immunoreactivity to the synaptophysin antibody in over 50% of neoplastic cells. Mucinous or signet ring cell carcinomas were defined as tumors with mucinous and/or signet ring cell features in over 50% of the tumor volume. Adenosquamous carcinoma was diagnosed based on a predominant (N80%) glandular or squamous differentiation pattern. Tumors classified as mixed histology patterns showed more than 1 type of histology features [15,18,19]. A subset of adenocarcinomas with histologic features suggestive of pancreatic acinar differentiation with a compact acinar growth pattern, centrally or basally located round to oval nuclei, stippled chromatin, single prominent nucleoli, and dark basophilic-eosinophilic granular cytoplasm were further tested for their immunoreactivity to the α1-chymotrypsin in all cases and α1-trypsin antibodies in some cases. Those tumors immunoreactive to the anti–α1-chymotrypsin antibody were classified as PALA [15,18,19]. The status of lymphovascular and perineural invasion was recorded.

PALA of the proximal stomach invading the esophagus

913

Fig. 1 Bormann's gross appearance of representative PALAs of the proximal stomach invading the esophagus. A, Type 1, polypoid with central necrosis. B, Type 3, ulcerated with necrotic center and an indistinct border. C, Type 4, flat with an infiltrative border. D, Type 3, the napkin-ring pattern with central necrosis, an indistinct border, and fleshy tan cut surface.

2.3. Immunohistochemistry Immunohistochemical staining was performed with standard methods [15]. Briefly, an archival tumor tissue block with the highest estimated tumor volume was selected and cut at 5 μm in thickness. Tumor tissue sections were treated first in the heat-induced epitope retrieval EDTA buffer (Dako, Glostrup, Denmark) and then with appropriate primary antibodies at 4°C overnight (Table 1). Immunoreactivity of cells was visualized with the Dako Envision + Kit-HRP (Dako) after hematoxylin counterstaining. Normal pancreatic tissue and carcinoid tumors served as positive controls for antibodies against α1-chymotrypsin, α1-trypsin, and synaptophysin antigenes, respectively. Human tonsils and normal horse serum without a primary antibody served as negative controls. Immunoreactivity of neoplastic cells was semiquantitatively scored by 2 experienced pathologists independently, blinded to clinical information or patient

survival data. The results were reported either as positive or negative in the best oriented and stained area of the section. A positive anti–α1-chymotrypsin or anti–α1-trypsin immunostain was defined as intensely diffuse, granular cytoplasmic stain in over 10% of neoplastic cells on the section. The specificity of α1-chymotrypsin and α1-trypsin antibodies was confirmed by negative immunoreactivity in

Table 1

Immunohistochemistry

Antibody

Clone

Manufacturer

Dilution

α1-chymotrypsin

Rabbit polyclonal Rabbit polyclonal 27G12

Zymed Lab, San Francisco, CA DAKO, Glostrup, Denmark Novocastra, Newcastle, UK

1:100

α1-trypsin synaptophysin

1:2000 1:200

914

Q. Huang et al.

9 different tumor cases in which PALA was not suspected. Because of identical immunostain results between α1chymotrypsin and α1-trypsin antibodies, the α1-trypsin immunostain was not carried out in all cases. A positive synaptophysin immunostain was defined as intense cytoplasmic immunoreactivity in over 50% of neoplastic cells on the section.

Table 2

2.4. Periodic acid Schiff stain with diastase digestion In 40 PALA and 9 non-PALA tumors, periodic acid Schiff (PAS) stains with and without diastase digestion were carried out. In 3 remaining PALA tumors, tumor blocks were not available for this stain. Positive stains were

Clinicopathologic features

Variable

All cases

PALA

Total no. Male sex (%) Median age in y (range) H pylori–positive cases (%) Bormann's type (%) 1 2 3 4 Predominant histology (%) Adenocarcinoma, NOS Adenosquamous carcinoma Micropapillary adenocarcinoma Mucinous carcinoma Neuroendocrine carcinoma Signet ring cell carcinoma Mixed carcinoma Median tumor size in cm (range) Lymphovascular invasion (%) Perineural invasion (%) AJCC pathologic stage (%) pT T1 T2 T3 T4 pN N0 N1 N2 N3 pM M0 M1 Overall I IA IB II IIA IIB III IIIA IIIB IIIC IV

137 103 (75) 65 (47-90) 57 (42)

43 38 (88) 66 (51-90) 17 (40)

94 65 (69) 65 (47-84) 40 (43)

7 (5.1) 11 (8.0) 101 (74) 18 (13)

3 (7.0) 4 (9.3) 29 (67) 7 (16)

4 (4.3) 7 (7.4) 72 (77) 11 (12)

Abbreviation: NOS, not otherwise specified.

Non-PALA

P .02 .34 .85 .72

.03 71 (52) 8 (5.8) 18 (13) 9 (6.6) 1 (0.7) 4 (2.9) 26 (19) 5 (0.5-12) 82 (60) 81 (59)

23 (54) 0 5 (12) 1 (2.3) 0 0 14 (33) 5.5 (2-10.5) 35 (81) 32 (74)

48 (51) 8 (8.5) 13 (14) 8 (8.5) 1 (1.1) 4 (4.3) 12 (13) 4.75 (0.5-12) 47 (50) 49 (52)

.04 b.001 .02 .07

4 (2.9) 12 (8.8) 121 (88)

0 1 (2.3) 42 (98)

4 (4.3) 11 (12) 79 (84)

36 (26) 20 (15) 36 (26) 45 (33)

8 (19) 2 (4.7) 11 (26) 22 (51)

28 (30) 18 (19) 25 (27) 23 (24)

127 (93) 10 (7.3)

40 (93) 3 (7.0)

87 (93) 7 (7.4)

11 (8.0) 4 (2.9) 7 (5.1) 48 (35) 25 (18) 23 (17) 66 (48) 30 (22) 28 (20) 8 (5.8) 12 (8.8)

1 (2.3) 0 1 (2.3) 9 (21) 7 (16) 2 (4.7) 30 (70) 10 (23) 14 (33) 6 (14) 3 (7.0)

10 (11) 4 (4.3) 6 (6.4) 39 (42) 18 (19) 21 (22) 36 (38) 20 (21) 14 (15) 2 (2.1) 9 (9.6)

.007

.61

.005

PALA of the proximal stomach invading the esophagus defined as the presence of PAS-positive, diastase-resistant cytoplasmic granules.

2.5. Statistical analysis Based on histopathologic and immunohistochemical features, patients were divided into 2 groups: PALA and non-PALA. Differences in clinicopathologic variables and patient survival after surgical resection were analyzed and compared between the 2 groups independently by 2 investigators. The χ 2 and Fisher exact tests were used to compare differences between categorical variables. KaplanMeier survival curves were calculated and compared with the log-rank test. Cox regression analysis was used for uniand multivariate analyses. P b .05 was considered statistically significant. Statistical Package for Social Sciences (version 13.0; SPSS Inc, Chicago, IL) statistical software was used for analysis.

915

3. Results 3.1. Demographic and clinical features During the study period, 172 consecutive surgical resections were performed for proximal gastric carcinoma invading the esophagus. Among them, 32 patients were lost to follow-up, 3 did not have sufficient tumor tissues for immunostains, and 137 were eligible for the study. Based on histopathologic features, PALA was suspected in 48 and confirmed by immunohistochemistry in 43 (31%). The remaining 94 cases (69%) were classified as the nonPALA group. Patient median age was 65 years for the entire cohort and not significantly different between the PALA and nonPALA groups. The male-to-female ratio was significantly greater in the PALA (7.6) than in the non-PALA (2.4) groups (Table 2).

Fig. 2 Representative histology patterns of PALAs of the proximal stomach invading the esophagus show acinar (A and B), micropapillary (C), and microcystic-acinar (D) patterns. Note that neoplastic cells show prominent single nucleoli in B and intense cytoplasmic immunoreactivity to the α1-chymotrypsin antibody in inset of A, C, and D (hematoxylin-eosin stains in A-D; original magnification ×100 in A, ×200 in C and D, and insets, and ×400 in B).

916

3.2. Pathologic characteristics Tumor size was slightly larger in the PALA than in the non-PALA groups (Table 2). There were no significant differences in the Bormann's gross type and Helicobacter pylori infection rates between the 2 groups. Grossly, PALA tumors were nonencapsulated, centrally ulcerated, and bulky. The bulk of the tumor mass was centered in the proximal stomach, with a minor component of the tumor invading into the distal esophagus in all cases (Fig. 1). On the cut surface, PALA tumors were fleshy and tan with infiltrative borders (Fig. 1B-D). Unlike pancreatic acinar carcinoma [8], gross hemorrhage, frank necrosis, or discernable cysts were inconspicuous or absent. Microscopically, PALA tumors were hypercellular, but tumor lobules or nests were not prominent at low-power examination (Fig. 2A). The tumor stroma was delicate with a rich network of thin-walled capillaries. Desmoplastic stromal reaction was inconspicuous. Neoplastic cells were most frequently arranged in compact acini with minute lumens

Q. Huang et al. (Fig. 2A and B) or in micropapillary (Fig. 2C) and solid patterns but infrequently in microcystic (Fig. 3A), signet ring (Fig. 3B), trabecular (Fig. 3C), single-cell (Fig. 3D), and nested organoid (Fig. 4) patterns. The acinar pattern was present at least focally in most cases. The organoid pattern was noted in 6 cases, one of which showed focal comedo necrosis. In general, the nuclei of neoplastic cells in PALA tumors were almost uniformly round to oval and showed mild pleomorphism, thickened nuclear membrane, stippled chromatin, and single prominent nucleoli in most cases (Fig. 2B). Mitotic figures were infrequent. As a rule, the cytoplasm was moderately abundant, conspicuously dark basophilic-eosinophilic, and finely granular. The degree of cytoplasmic immunoreactivity to the α1-chymotrypsin antibody varied among PALA tumors (Figs. 2 and 3, insets). Usually, tumors with solid and dense acinar patterns showed sheet-like intense immunoreactivity in almost all neoplastic cells. In contrast, in poorly differentiated tumors, focal yet intense immunoreactivity was conspicuous in single cells or abortive glands (Fig. 3D). None of 9 non-PALA tumors were

Fig. 3 Representative histology patterns of PALAs of the proximal stomach invading the esophagus exhibit microcystic (A), mixed signet ring (B), trabecular (C), and single-cell (D) features (hematoxylin-eosin stains in A-D. Insets in A-D; immunostains for the α1-chymotrypsin antigen; original magnification ×200 in A-D and insets and ×400 in B).

PALA of the proximal stomach invading the esophagus

917

Fig. 4 Mixed differentiation of a PALA of the proximal stomach invading the esophagus features an organoid growth pattern with neoplastic cells intensely immunoreactive to both α1-chymotrypsin (A) and synaptophysin (B) antibodies (hematoxylin-eosin stains in A and B; original magnification ×100 in A and insets of A and B and ×600 in B).

immunoreactive to either the α1-chymotrypsin or α1-trypsin antibody. PAS stain with diastase digestion showed positive cytoplasmic granules in neoplastic cells in 33 (83%) of 40 cases (Fig. 5) but negative results in all 9 non-PALA tumors. In the benign gastric tissue adjacent to PALA tumors, there was no morphological evidence of heterotopic pancreas or PAM. According to the patient medical records, there were no similar tumors in the pancreas and no direct connections between the GEJ tumors and the pancreas in any patient of this cohort. Unlike non-PALA tumors (Table 2), PALA tumors did not exhibit adenosquamous differentiation. Two PALA tumors (4.7%) demonstrated features of mixed neuroendo-

crine differentiation with an organoid pattern and strong immunoreactivity to both α1-chymotrypsin and synaptophysin antibodies (Fig. 4, insets). Nine PALA tumors (21%) demonstrated mixed signet ring differentiation. In all PALA tumors with mixed neuroendocrine or signet ring differentiation, features of adenocarcinoma were predominant. Overall, the distribution of predominant histology types was significantly different between PALA and nonPALA groups. Compared with non-PALA tumors (Table 2), PALA tumors showed more frequent lymphovascular invasion, more advanced pN and overall stages with pIII and pIV in 77% of cases, and a trend toward more advanced pT stage.

Fig. 5 One PALA of the proximal stomach invading the esophagus with acinar features (A) demonstrates strong immunoreactivity to the α1-chymotrypsin antibody (B) and positive PAS stain with diastase digestion (C) (hematoxylin-eosin stain in A; original magnification ×200 in A-C).

918

Q. Huang et al.

Table 3 Univariate and multivariate analyses of variables associated with survival in patients with PALA of the proximal stomach Variable Male sex Age N75 y

Univariate Multivariate Relative risk P P (95% CI) .30 .04

.006

H pylori positivity .13 Bormann's gross type .41 Histology type .23 Tumor size N5 cm .07 pT .63 pN .005 pM .01 Overall AJCC stage b.001

NS NS NS NS b.001

Lymphovascular invasion Perineural invasion PALA

NS

2.53 (1.31-4.89)

2.05 (1.38-3.04)

.47 .38 .71

Abbreviations: CI, confidence interval; NS, not significant.

3.3. Survival The median follow-up after surgical resection was 27 months overall and 36 months in survivors. At last followup, 56 patients (39%) had died. Two-year survival of the PALA group was 67% (95% confidence interval, 51%-83%) compared with 73% (63%-82%) for the non-PALA group, whereas 5-year survival was 45% for the PALA group (25%66%) compared with 30% (12%-48%) for the non-PALA group. On univariate analysis, patient age older than 75 years, higher pN, pM1, and overall stages predicted worse survival (Table 3). On multivariate analysis, only age older than 75 years and overall stage were independent predictors of worse outcomes. Patients with PALA tumors did not have a statistically significant survival difference, compared with those with non-PALA tumors on either univariate or multivariate analysis (Fig. 6).

patterns of growth. Adenosquamous, neuroendocrine, and signet ring cell histologies were absent. This distinct pattern of predominant histology types was statistically different between the PALA and non-PALA groups. PALA tumors were more prone to lymphovascular and perineural invasion and associated with a higher pN stage than the non-PALA counterpart. These clinical pathologic features of PALA tumors mimic those of pancreatic acinar carcinoma [8,20] and are distinctly different from non-PALA tumors and esophageal adenocarcinoma in American white patients [18,19]. In addition, a significant male predominance was detected in the PALA group. Because of these unique clinicopathologic characteristics and strong histologic, histochemical, and immunohistochemical evidence, we propose classifying this subset of adenocarcinomas in the proximal stomach as PALA of the proximal stomach. In this study, pancreatic acinar differentiation in PALA tumors was supported by immunoreactivity to the specific pancreatic acinar α1-chymotrypsin antibody [9,20]. Intuitively, the expression of this powerful protein lytic enzyme in PALA tumors may help facilitate tumor invasion and metastasis and may explain the highly invasive and poorly differentiated features of this cancer. Compared with nonPALA tumors, PALA tumors were more advanced, most staged pT3, and over half staged pN3. PALA of the stomach has been reported previously [21-34], mainly in the distal stomach such as the antrumpylorus [23-27,33] and also rarely in the fundus [22] and body [29-32]. In only 1 case report was the tumor discovered in the GEJ region [34]. All, except for 2, cases were associated with heterotopic pancreas. This may suggest a

4. Discussion The current study documents the clinicopathologic features of PALA of the proximal stomach invading the esophagus and compares these features to those of nonPALA tumors. Different from non-PALA tumors, PALA tumors showed less heterogeneous histopathology but more advanced pathologic stage. Grossly, PALA tumors were bulky with a predominantly ulcerated, infiltrative appearance. Microscopically, neoplastic cells were characteristically arranged in packed acini and admixed, to some extent, with solid, micropapillary, trabecular, microcystic, signet ring, single-cell, and organoid

Fig. 6 The Kaplan-Meier survival analysis shows a trend of worse outcomes in the PALA than in the non-PALA group, but the difference was not statistically significant. Crossed marks represent censored cases.

PALA of the proximal stomach invading the esophagus different mechanism of pathogenesis for PALA of the proximal stomach as compared with PALA of the distal stomach. There are at least 4 possibilities for the tumorigenesis, origin, and pathobiology of this cancer: (1) the tumor arises in heterotopic pancreas [21,22,25-27,31,33]. This possibility is unlikely. Heterotopic pancreas in the proximal stomach and the distal esophagus is vanishingly rare. There was no residual heterotopic pancreatic tissue detected in any case of this cohort. (2) The tumor is metastatic from a primary pancreatic site. This explanation is also inconsistent with the fact that a pancreatic tumor was not identified by preoperative imaging studies or at the time of surgery in any patient. In addition, histopathologic features of PALA tumors showed a variety of growth patterns that were similar to but slightly different from those of typical pancreatic acinar carcinoma [8,20]. In the latter, hypercellular acinar and solid patterns are predominant, and other morphological patterns such as micropapillary, signet ring, and microcysts are extremely rare or have not been described [8,20]. (3) PALA tumors arise from PAM glands [28]. Ample evidence supports this explanation because PAM is common in the GEJ region [1-3]. Our recent study revealed PAM in about 30% of resection cases, 1 of which also exhibited dysplastic features [3]. Although PAM was not identified in the residual nonneoplastic tissue in this study, it is most likely that PALA tumors might have overgrown the residual PAM because of their advanced stages. This hypothesis remains to be examined in future studies. (4) The tumor derives from pluripotential stem cells in the GEJ region with multiple differentiation capabilities. This hypothesis seems likely and has been suggested by other investigators who observed a mixed pattern of pancreatic exocrine and endocrine differentiations in the same adenocarcinoma [23,24,27-30]. This hypothesis requires the support of solid experimental studies. There are several limitations of this study. First, the small number of PALA tumors and relatively short followup interval may have limited our ability to detect overall survival differences between PALA and non-PALA groups. Second, we did not have the information on the cause of death to calculate disease-free survival (recurrence-free survival in gastric cancer is notoriously unreliable because of variable follow-up and delayed clinical presentation of recurrence). Third, we did not find many PALA tumors at early stages and thus could not evaluate the hypothesis that PAM represents the precursor lesion to PALA tumors in this cohort. Fourth, because of the retrospective nature of this study, we were not able to standardize tissue sampling and might have missed precursor lesions of PALA tumors. Although the study was conducted entirely in Chinese patients, PALA tumors are not unique to the Chinese and may also be found in other ethnic populations. In fact, carcinomas in some published histopathologic studies do show pancreatic acinar-like morphology. For example, a recent article on high-grade neuroendocrine carcinomas in the gastrointestinal tract [35] showed a tumor with typical

919 histologic features highly suggestive of pancreatic acinarlike differentiation (Fig. 6 of Shia et al [35]). Regrettably, the authors neither specified the organ of origin of that tumor nor performed further studies to detect pancreatic acinar molecules [35]. Finally, although pancreatic acini secrete several proteins, such as α1-chymotrypsin, trypsin, lipase, and amylase, we tested only α1-chymotrypsin for 2 major reasons. First, previous studies have shown that among the widely used pancreatic exocrine antibodies, the α1-chymotrypsin antibody was the most sensitive and specific for pancreatic acinar carcinoma [20], whereas detection of other pancreatic acinar molecules was highly variable [8,20]. Second, we suspect that immunohistochemical detection of other pancreatic acinar molecules may be inconsistent in PALA tumors because of their poor differentiation. In summary, PALA tumors of the proximal stomach show gross, histologic, histochemical, and immunohistochemical features mimicking, but not identical to, pancreatic acinar carcinoma. This tumor subtype is distinctly different from non-PALA tumors. Recognition of this uncommon subtype of adenocarcinomas may help classify tumors in the GEJ region more accurately. Further molecular pathology studies involving large numbers of cases are required to validate our preliminary results, elucidate the underlying mechanism of tumorigenesis, and to better understand the pathobiology of this subtype of GEJ cancers.

References [1] Doglioni C, Laurino L, Dei Tos AP, et al. Pancreatic (acinar) metaplasia of the gastric mucosa. Histology, ultrastructure, immunocytochemistry, and clinicopathologic correlations of 101 cases. Am J Surg Pathol 1993;17:1134-43. [2] Wang HH, Zeroogian JM, Spechler SJ, et al. Prevalence and significance of PAM at the GEJ. Am J Surg Pathol 1996;20:1507-10. [3] Huang Q, Zhang LH. Histopathological features of diseases in esophageal glands in the region of the GEJ in Chinese patients with gastric cardiac cancer involving the esophagus. Pathol Lab Med Int 2010;2:33-40. [4] Krishnamurthy S, Dayal Y. Pancreatic metaplasia in Barrett's esophagus. An immunohistochemical study. Am J Surg Pathol 1995; 19:1172-80. [5] el-Zimaity HM, Verghese VJ, Ramchatesingh J, et al. The gastric cardia in gastro-oesophageal disease. J Clin Pathol 2000;53:619-25. [6] Johansson J, Håkansson HO, Mellblom L, et al. PAM in the distal oesophagus and the gastric cardia: prevalence, predictors and relation to GORD. J Gastroenterol 2010;45:291-9. [7] Sarbia M, Donner A, Gabbert HE. Histopathology of the GEJ: a study on 36 operation specimens. Am J Surg Pathol 2002;26:1207-12. [8] Klimstra DS, Heffess CS, Oertel JE, et al. Acinar cell carcinoma of the pancreas. A clinicopathologic study of 28 cases. Am J Surg Pathol 1992;16:815-37. [9] Ordonez NG. Pancreatic acinar cell carcinoma. Adv Anat Pathol 2001;8:144-59. [10] Takubo K, Arai T, Sawabe M, et al. Structures of the normal esophagus and Barrett's esophagus. Esophagus 2003;1:37-47. [11] Krishnamurthy S, Integlia MJ, Grand RJ, et al. Pancreatic acinar cell clusters in pediatric gastric mucosa. Am J Surg Pathol 1998;22:100-5.

920 [12] Polkowski W, van Lanschot JJ, ten Kate FJ, et al. Intestinal and pancreatic metaplasia at the esophagogastric junction in patients without Barrett's esophagus. Am J Gastroenterol 2000;95:617-25. [13] Jhala NC, Montemor M, Jhala D, et al. Pancreatic acinar cell metaplasia in autoimmune gastritis. Arch Pathol Lab Med 2003;127: 854-7. [14] Integlia MJ, Krishnamurthy S, Berhane R, et al. Pancreatic metaplasia of the gastric mucosa in pediatric patients. Am J Gastroenterol 1997;92:1553-6. [15] Huang Q, Zhang LH. The histopathologic spectrum of carcinomas involving the GEJ in the Chinese. Int J Surg Pathol 2007;15:38-52. [16] Huang Q. Definition of the esophagogastric junction: a critical mini review. Arch Pathol Lab Med 2011;135:384-9. [17] Huang Q, Shi J, Feng AN, et al. Gastric cardiac carcinomas involving the esophagus are more adequately staged as gastric cancers by the 7th edition of the American Joint Commission on Cancer staging system. Mod Pathol 2010;24:138-46. [18] Huang Q, Fan XS, Agoston AT, et al. Comparison of GEJ carcinomas in Chinese versus American patients. Histopathology 2011;59:188-97. [19] Huang Q, Fan XS, Agoston AT, et al. GEJ carcinomas in Chinese patients show distinct clinicopathologic features. Mod Pathol 2009;22(Suppl 1):133A. [20] Hruban RH, Pitman MB, Klimstra DS. Tumors of pancreas. AFIP Atlas of Tumor Pathology, Series 4. Washington, DC: America Registry of Pathology; 2007. p. 194-9. [21] Goldfarb WB, Bennett D, Monafo W. Carcinoma in heterotopic gastric pancreas. Ann Surg 1963;158:56-8. [22] Hickman DM, Frey CF, Carson JW. Adenocarcinoma arising in gastric heterotopic pancreas. West J Med 1981;135:57-62. [23] Fukunaga M. Gastric carcinoma resembling pancreatic mixed acinarendocrine carcinoma. HUM PATHOL 2002;33:569-73. [24] Sun Y, Wasserman PG. Acinar cell carcinoma arising in the stomach: a case report with literature review. HUM PATHOL 2004;35:263-5.

Q. Huang et al. [25] Emerson L, Layfield LJ, Rohr LR, et al. Adenocarcinoma arising in association with gastric heterotopic pancreas: a case report and review of the literature. J Surg Oncol 2004;87:53-7. [26] Papaziogas B, Koutelidakis I, Tsiaousis P, et al. Carcinoma developing in ectopic pancreatic tissue in the stomach: a case report. Cases J 2008;1:249. [27] Mizuno Y, Sumi Y, Nachi S, et al. Acinar cell carcinoma arising from an ectopic pancreas. Surg Today 2007;37:704-7. [28] Ambrosini-Spaltro A, Potì O, De Palma M, et al. Pancreatic-type acinar cell carcinoma of the stomach beneath a focus of pancreatic metaplasia of the gastric mucosa. HUM PATHOL 2009;40:746-9. [29] Kusafuka K, Bando E, Muramatsu K, et al. Pancreatic-type mixed acinar-endocrine carcinoma with alpha-fetoprotein production arising from the stomach: a report of an extremely rare case. Med Mol Morphol 2009;42:167-74. [30] Jain D, Eslami-Varzaneh F, Takano AM, et al. Composite glandular and endocrine tumors of the stomach with pancreatic acinar differentiation. Am J Surg Pathol 2005;29:1524-9. [31] Ura H, Denno R, Hirata K, et al. Carcinoma arising from ectopic pancreas in the stomach: endosonographic detection of malignant change. J Clin Ultrasound 1998;26:265-8. [32] Osanai M, Miyokawa N, Tamaki T, et al. Adenocarcinoma arising in gastric heterotopic pancreas: clinicopathological and immunohistochemical study with genetic analysis of a case. Pathol Int 2001;51: 549-54. [33] Jeong HY, Yang HW, Seo SW, et al. Adenocarcinoma arising from an ectopic pancreas in the stomach. Endoscopy 2002;34:1014-7. [34] Guillou L, Nordback P, Gerber C, et al. Ductal adenocarcinoma arising in a heterotopic pancreas situated in a hiatal hernia. Arch Pathol Lab Med 1994;118:568-71. [35] Shia J, Tang LH, Weiser MR, et al. Is nonsmall cell type high-grade neuroendocrine carcinoma of the tubular gastrointestinal tract a distinct disease entity? Am J Surg Pathol 2008;32:719-31.