Microallelotyping defines the sequence and tempo of allelic losses at tumour suppressor gene loci during colorectal cancer progression.

Microallelotyping of many regions from individual colorectal tumours was used to determine the sequence and tempo of allelic loss on 5q, 17p and 18q during neoplastic progression. No allelic losses were found in normal tissues surrounding colorectal neoplasms, but losses occurred abruptly on 5q at the transition from normal colonic epithelium to the benign adenoma, and on 17p at the transition from adenoma to carcinoma, indicating an essential role for these losses in tumour progression. Allelic losses were uniform throughout extensively microdissected benign adenomas and carcinomas. However, substantial allelic heterogeneity was found in high-grade dysplasia, the transition lesion between adenoma and carcinoma. Thus, allelic losses on 5q and 17p are associated with abrupt waves of clonal neoplastic expansion, and high-grade dysplasia is characterized by a high degree of allelic heterogeneity.

CDX2 has tumorigenic potential in the human colon cancer cell lines LOVO and SW48.

CDX2 is a Drosophila caudal-related homeobox transcription factor that is important for the establishment and maintenance of intestinal epithelial cells. CDX2 is a marker of colon cancer, with strong staining in up to 90% of colonic adenocarcinomas. CDX2 heterozygous-null mice develop colonic neoplasms, which have suggested that CDX2 is a tumor suppressor. However, CDX2 has not been reported to affect xenograft growth. Furthermore, CDX2 is rarely mutated in colon cancer, which has led to suggestions that it may play only a minor role as a tumor suppressor in colon cancer. To understand the functional contributions of CDX2 to colon cancer, we disrupted CDX2 in LOVO and SW48 human colon cancer cell lines by targeted homologous recombination. Consistent with the literature, disruption of CDX2 enhanced anchorage-dependent cell proliferation. However, homozygous loss of CDX2 led to significant inhibition of anchorage-independent growth in LOVO cells, and cell lethality in SW48 cells. Further analyses revealed that disruption of CDX2 led to anchorage-independent G1 to S growth arrest and anoikis. In vivo xenograft studies confirmed that disruption of CDX2 inhibited LOVO tumor growth. These data demonstrate that CDX2 mediates anchorage-independent growth and survival. Thus, CDX2 has tumorigenic potential in the human colon cancer cell lines LOVO and SW48.

CDX2 does not suppress tumorigenicity in the human gastric cancer cell line MKN45.

CDX2 is a Drosophila caudal-related homeobox transcription factor that is expressed specifically in the intestine. In mice, ectopic expression of CDX2 in the gastric mucosa gives rise to intestinal metaplasia and in one model, gastric carcinoma. In humans, increased CDX2 expression is associated with gastric intestinal metaplasia and tubular adenocarcinomas. These patterns of expression have shown that CDX2 is important for the initiation of intestinal metaplasia in the gastric mucosa, but the role of CDX2 in established gastric cancer remains unclear. We sought to determine whether CDX2 contributes to tumorigenic potential in established gastric cancer. The CDX2 gene in MKN45 gastric carcinoma cells was disrupted using targeted homologous recombination. The resulting CDX2-/- cells are essentially identical to their parental cells, with the exception of CDX2 ablation. We found no significant differences in the proliferation of CDX2-/- cells compared to CDX2+/+ cells, in vitro or in vivo. Molecular analyses show that loss of CDX2 predominantly altered the expression of genes involved in intestinal glandular differentiation and adhesion. However, there were no microscopic differences in tumor differentiation. We conclude that disruption of CDX2 in MKN45 cells does not significantly affect their tumorigenic potential.

Flat adenomas in a colon cancer-prone kindred.

We describe new pathologic findings in a hereditary nonpolyposis colorectal cancer family. Affected family members developed multiple small adenomas with right-sided predominance; many adenomas had an unusual appearance featuring slightly elevated lesions with adenomatous changes confined to the upper regions of the colonic crypts. We have adopted the previously established term "flat adenoma" for these lesions. This phenotype may be a morphologic marker for at least one subset of hereditary nonpolyposis colorectal cancer.

Smooth muscle tumors of the gastrointestinal tract. What we know now that Stout didn't know.

Arthur Purdy Stout and his co-workers, in several publications, raised two important issues concerning gut stromal tumors. First, they felt that all were of smooth muscle origin. Recent ultrastructural and immunohistochemical studies suggest that the component cells are basically undifferentiated, and there is only occasional emergence of smooth muscle features and, in some tumors, possibly features of other cell types as well, such as Schwann cells. Second, Stout felt that the high mitotic rate was the best predictor of malignancy, but he recognized that some tumors, even with low rates, could metastasize. Surprisingly, recent studies, even those covering large series, have done little to dispute these contentions. However, current data suggest that the diagnosis of malignancy can be made using multiple parameters, not all of which must be present in every sarcoma. These parameters include, in addition to mitotic rate, the size, gross invasion of adjacent organs, and cellularity, and all of these must be modified according to the location in the gut and the pattern of growth.

Stromal tumors of the duodenum. A histologic and immunohistochemical study of 20 cases.

Using cell size, cell density, and microscopic growth pattern, 20 duodenal stromal tumors were initially separated into benign and malignant categories. The 10 histologic benign tumors had uniform spindle cells, low cellularity, and an organoid pattern. All had round eosinophilic collagen blobs scattered among the spindle cells, were 4.5 cm or less in maximum diameter, and had two or fewer mitoses per 50 high-power fields (HPF). None metastasized or recurred during a median follow-up of 7 years. In contrast, the 10 histologically malignant tumors were highly cellular, all had two or more mitoses per 50 HPF, and all but one had diameters of 4.5 cm or greater, the exception being 4 cm. Eight cases also had benign-appearing areas, usually submucosal. Eight patients died with disease a median of 31 months after resection, almost all with liver metastases. One patient is alive with metastasis at 13 years. The patient with the 4-cm malignant tumor is disease free at 49 months. All 15 cases were strongly vimentin positive, 11 had S-100 protein, and seven had the CD34 marker. None were desmin or actin positive. No immunophenotype separated benign from malignant. The proliferation marker, proliferating cell nuclear antigen, correlated with histologic diagnosis and clinical outcome, but Ki-67 did not. Based on light microscopic features alone, benign and malignant duodenal stromal tumors can be separated from each other. Tumors with large cells and an organoid pattern are predictably benign; in this study, these tumors measured 4.5 cm or less in diameter and had fewer than 2 mitoses per 50 HPF. Highly cellular tumors with small cells and little or no organoid pattern are malignant. They usually have a diameter greater than 4.5 cm and more than two mitoses per 50 HPF, and they are usually fatal. Immunostaining for cytoplasmic proteins and proliferation markers offers no additional prognostic information to the light microscopic appearances. These conclusions apply only to duodenal tumors; whether they also apply to stromal tumors of the jejunum and ileum is not known.

Inflammatory cloacogenic polyp. A unique inflammatory lesion of the anal transitional zone.

This report describes the clinical and pathologic features of eight cases of an unusual inflammatory polyp arising from the transitional zone of the anus. This lesion has been designated as the inflammatory cloacogenic polyp. Rectal bleeding is the most common presenting clinical symptom. The polyp is usually located on the anterior wall of the anal canal. Morphologically, it is characterized by a tubulovillous pattern of growth, superficial ulceration, displaced groups of crypts into submucosa, and extension of chronically inflamed fibromuscular stroma into the lamina propria. Clinical and morphologic similarities with the solitary rectal ulcer syndrome suggests that prolapse of transitional zone mucosa may be important in its pathogenesis. Simple surgical resection appears to be the treatment of choice.

Ulcerative colitis: patterns of involvement in colorectal biopsies and changes with time.

Chronic inflammation, both endoscopic and histologic, in a contiguous and symmetric distribution is said to be important in distinguishing ulcerative colitis (UC) from Crohn's disease. Little is known whether this rule holds during the course of the disease and whether endoscopic/histologic correlation persists. In this study, we analyzed histologic patterns of UC in sequential sets of biopsy specimens to assess whether endoscopic and histologic findings correlate with time and treatment and to see whether distribution changes. Two hundred seventeen sets of colorectal biopsy specimens from 797 sites from 41 patients with clinical UC were studied and correlated with endoscopic findings. Each biopsy specimen was classified as definite or suspicious for chronic colitis or normal. Two histologic patterns of disease were identified: (1) diffuse, when all areas in all pieces from a biopsy segment had clear-cut colitis and (2) nondiffuse, when not all pieces were involved or single pieces had disease and normal mucosa both. Of 41 patients, the maximal extent of histologic disease was pancolitis in 30; 25 had less extensive disease at some point in the course. The maximal extent was left-sided in eight patients, seven of whom had less extent at some point. Of the three patients in whom the maximal extent was proctosigmoiditis, in one the inflammation disappeared. Seventy percent of the biopsy sites had diffuse patterns and 30% had nondiffuse. Histologic and endoscopic disease reverted to normal in 22 and 24 of 41 patients, respectively. Endoscopic and histologic findings were similar in 65% of the biopsy sites. Our results indicate that in long-standing UC (1) histologic disease may revert to normal mucosa, (2) because endoscopy alone may be insufficient to identify the mucosa as normal, biopsies should also be performed on the endoscopically normal mucosa, (3) the full extent of UC often is not established by a single set of biopsies, and (4) nondiffuse chronic inflammation and rectal sparing occurs in UC and are not necessarily markers of Crohn's disease.

Candida hepatitis. Histopathologic diagnosis.

Candida hepatitis, usually a manifestation of disseminated candidiasis in immunocompromised patients, is difficult to diagnose antemortem. We studied six patients with proven hepatic candidiasis to assess features helpful in deriving a correct diagnosis. Five patients were immunosuppressed as a result for treatment for leukemia; one was immunosuppressed due to renal transplantation. All had sustained fevers greater than 101 degrees F, elevated alkaline phosphatase levels, and multiple hepatic and splenic defects--presumably abscesses--on abdominal CT scan. Twelve liver biopsies (nine needle, three wedge) were examined. Biopsies from four patients contained identifiable Candida organisms within suppurative granulomas; a biopsy from a fifth patient grew Candida albicans in cultures. In the sixth patient, the first biopsy was culture positive for Candida albicans, and the second biopsy, a fine-needle aspirate, contained Candida organisms and purulent material. In all of the nondiagnostic biopsies, as well as in regions of the diagnostic biopsies around the suppurative granulomas, mass-associated obstructive changes were noted. These included pericentral sinusoidal dilatation and cholestatic inflammation characterized by periportal ductular proliferation with surrounding neutrophils and edema. We conclude that in the appropriate clinical setting, these mass-associated histologic findings are suggestive of adjacent Candida abscesses. Definite diagnosis requires either the identification of Candida organisms within inflammatory hepatic lesions or positive culture of Candida from the liver biopsy.

The histopathologic spectrum of acute self-limited colitis (acute infectious-type colitis).

Acute self-limited colitis (ASLC) is a self-limiting diarrheal illness which is often caused by known infectious agents (Campylobacter, Salmonella, and Shigella), but many cases are of unknown etiology. This report describes the histopathologic features of acute self-limited colitis as related to its natural history. The extent of inflammation and regeneration varies with the duration of the disease. In the peak activity stage (within 0-4 days of onset of bloody diarrhea) there is mucosal edema, cryptitis, crypt ulcers, and abscesses. At the time of resolution (within 6-9 days of onset of bloody diarrhea), regenerative features become apparent along with residual focal neutrophilic cryptitis. In the latter stages of resolution, along with some regenerative features, occasional crypts with transmigrating lymphocytes may be present. A rectal biopsy is diagnostic only in the early stages of the disease. Later in the course, the rectal biopsy from patients with ASLC may be nondiagnostic or may be confused with Crohn's disease due to the persistence of focal cryptitis. In our experience, the presence of crypt distortion and basal plasmacytosis are the two most useful criteria to differentiate chronic ulcerative colitis from ASLC.

Diffuse duodenitis associated with ulcerative colitis.

Backwash ileitis and postcolectomy pouchitis are well-recognized complications of ulcerative colitis (UC), whereas inflammation of the proximal small intestine is not. In contrast, small intestinal disease at any level is common in Crohn's disease (CD). Despite this well-established and accepted dogma, rare cases of histologically proven diffuse duodenitis (DD) associated with UC appear in the literature. In this study, we report our experience with similar cases exhibiting this unusual inflammatory phenomenon. Routine histologic sections from four cases of DD associated with well-documented UC were reviewed and the findings correlated with all available medical records. Multiple endoscopic biopsies showing histologic features of UC and colectomy specimens confirming severe ulcerative pancolitis were available for all cases. Varying degrees of active chronic inflammation and architectural mucosal distortion identical to UC were observed in pre- and postcolectomy duodenal biopsies of one of four and four of four cases, respectively. Similar inflammatory patterns were present postoperatively in the ileum in three of four cases and in the jejunum in one case. Endorectal pull-through (ERPT) procedures were performed in three of four patients and an end-to-end ileorectal anastomosis was done in one patient. Despite extensive upper gastrointestinal tract involvement, none of the patients developed postsurgical Crohn's-like complications during a follow-up period of 12 to 54 months. This suggests that patients with pancolitis and DD do not necessarily have CD, but rather may have UC and, most importantly, that successful ERPT procedures may be performed in these patients.

Achalasia. A morphologic study of 42 resected specimens.

Achalasia is characterized by failure of relaxation of the lower esophageal sphincter and absence of progressive peristalsis in the esophageal body. Few data are available regarding the morphologic features of achalasia, in particular its histologic progression. The esophagi of 42 patients with achalasia treated with total thoracic esophagectomy were examined histologically in order to systematically identify morphologic features of clinically unresponsive achalasia and to determine what could be learned about the disease's evolution. In all cases, myenteric ganglion cells within the esophageal body were markedly diminished, with 20 specimens having none. Twenty specimens had residual ganglion cells in the proximal esophagus, and 15 specimens had a few randomly distributed ganglion cells in the mid- and distal portions of the esophagus. Inflammation within myenteric nerves, present in all cases, generally consisted of a mixture of lymphocytes and eosinophils, occasionally with plasma and mast cells. Focal replacement of myenteric nerves by collagen occurred in all cases, and there was almost complete replacement in several cases. Actual destruction of the residual ganglion cells was not seen. The resected esophagi also shared extramyenteric morphologic features. Some features probably stemmed from physiologic obstruction, such as muscular hypertrophy, mainly of the muscularis propria (all cases), with secondary degeneration and fibrosis (29 cases), and eosinophilia of the muscularis propria (22 cases). Other changes, probably resulting from chronic stasis of ingested materials in the lumen, included diffuse squamous hyperplasia (all cases), lymphocytic mucosal esophagitis (28 cases), lymphocytic inflammation of the lamina propria and submucosa with prominent germinal centers (all cases), and submucosal periductal or glandular inflammation with complete loss of submucosal glands in half of the cases. One patient had high-grade squamous dysplasia, and another had superficially invasive squamous cell carcinoma. A third group of changes was probably due to previous esophagomyotomy, including abnormal gastroesophageal reflux, as shown by pH reflux testing (13 cases) and Barrett's mucosa (four cases). In one case of Barrett's there was low-grade dysplasia. Clinically unresponsive, surgically resected achalasia has almost total loss of ganglion cells, and widespread destruction of myenteric nerves has already occurred. The only active component is myenteric inflammation. However, it cannot be determined whether this inflammation is a manifestation of ongoing nerve destruction or whether it is a secondary phenomenon.

Stromal tumors of the abdominal colon: a clinicopathologic study of 20 cases.

Stromal tumors of the abdominal colon, the least common of all gastrointestinal stromal tumors, have not been well characterized. They have often been lumped with stromal tumors of the anorectum in order to achieve significant numbers for analysis, yet there are no data to prove that stromal tumors from these two sites are the same. In this study, we evaluated 20 colonic stromal tumors to identify clinical, morphologic, and immunophenotypic features that were useful in discriminating between those that had metastasized or caused death from those that had not metastasized or caused death. We found that colonic stromal tumors are morphologically heterogeneous, and the malignant ones are clinically aggressive. They often have metastases at presentation, and cause death in a short time. An infiltrative growth pattern in the muscularis propria, invasion of the mucosa, and high mitotic counts correlated significantly both with metastases and with death from tumor. We also found that dense cellularity correlated significantly with metastases, but not with death, and that coagulative necrosis correlated with death, but not with metastases.

Stromal tumors of the anorectum: a clinicopathologic study of 22 cases.

Stromal tumors of the anorectum are a rare group of mesenchymal tumors that often have a protracted clinical course. We sought to determine which clinical, morphologic, and immunophenotypic features correlated with an adverse outcome in 22 patients with anorectal stromal tumors. An adverse outcome, defined as either tumor recurrence or metastasis, occurred in nine patients. Seven patients had metastases, two of whom also had local recurrences. Four of these patients also died from their disease. One patient had one local recurrence, and one patient had two local recurrences; neither of these patients had metastases. Recurrences were found as long as 103 months and metastases as late as 117 months after initial presentation. However, for patients without an adverse outcome, maximum follow-up was only 84 months. Thus both recurrence and metastasis may not appear until several years after treatment, indicating that a long-term follow-up period, probably longer than available for many tumors without an adverse outcome in this study, is needed before a patient can be considered to be cured. Tumor size greater than five centimeters correlated with an adverse outcome. However, given the protracted course of these tumors and the relatively limited follow-up available, other features such as location within the muscularis propria, mitotic activity, necrosis, and pleomorphism that did not significantly correlate with an adverse outcome may become significant with longer follow-up periods. We also found that on the basis of morphologic appearance and whether tumors were confined to the submucosa or located within the muscularis propria, anorectal stromal tumors could be divided into three groups, and that the behavior of anorectal stromal tumors may also depend upon their phenotype. The largest group of 17 tumors was located within the muscularis propria, mitotically active, and composed of densely cellular spindle-shaped cells. A second group of two tumors was also located within the muscularis propria and was composed of spindle-shaped cells, but lacked dense cellularity and mitotic activity. The third group was composed of three submucosal, polypoid tumors.

Gastritis: terminology, etiology, and clinicopathological correlations: another biased view.

The histological approach to gastritis, especially the chronic forms, has undergone a series of re-evaluations by different experts over the past decade, mainly because of the recognition of individual disease patterns that have specific clinical and epidemiological implications. The most spectacular of these was the discovery of Helicobacter pylori and its common gastritis, its relation to almost all duodenal peptic ulcers and to most gastric peptic ulcers, its potential as a precursor of first multifocal atrophic gastritis and later tubule-forming gastric carcinomas, and its status as a cause of gastric mucosal lymphomas. During this same decade other classes of gastric reaction and inflammations have been recognized, including chemical injury and lymphocytic gastritis. Also in the same decade the importance of non-steroidal anti-inflammatory drugs (NSAIDs) has emerged as a cause of gastric mucosal injuries. To add emphasis to all these discoveries, biopsies are being performed on stomachs in almost epidemic numbers and each biopsy specimen has the potential of having the features of one or more of these injuries as well as injuries that have yet to be described. To cope with this rapidly expanding gastric inflammatory informational extravaganza, pathologists need some way of dealing with the various entities comfortably and some method of cataloging them in ways that are understandable both to them and to the endoscopists with whom they work. However, if emerging data about the chronic gastritides are correct, it is conceivable that the need to diagnose them, from a strictly clinical standpoint, is limited. Either we may know what is in the biopsy specimen before we see it or what we see may not be important, although it may be intellectually challenging.

Solitary solid stromal gastrointestinal tumors in von Recklinghausen's disease with minimal smooth muscle differentiation.

Neurofibromatosis (von Recklinghausen's disease) is occasionally associated with large, solid stromal tumors of the gastrointestinal tract. The authors examined by electron microscopy two such cases of cellular spindle cell neoplasms of the small bowel histologically that resembled leiomyomas, in an attempt to clarify the cell of origin of these lesions. Ultrastructurally, the tumor cells predominantly contained moderate to large numbers of intracellular filaments, small cell processes, discontinuous adherent dense basement-membrane-like material, and abundant intercellular collagen. Definite fusiform dense bodies or structures highly suggestive of them and pinocytotic vesicles were seen in rare cells of each lesion after viewing multiple blocks. While patients with neurofibromatosis are certainly at risk of developing gastrointestinal Schwann cell neoplasms, these two cases suggest that they are also at risk for developing poorly differentiated stromal tumors, resembling leiomyomas by light microscopy, which may show only characteristic cytoplasmic differentiation of smooth muscle cells after ultrastructural examination of many sections.

Cell markers in gastrointestinal stromal tumors.

Stromal tumors of the gastrointestinal (GI) tract have generated considerable controversy about their direction and level of differentiation, particularly about whether the tumor cells are smooth muscle or Schwann cells. In an attempt to characterize these tumors, the immunohistochemical staining patterns of desmin, vimentin, actin, and S-100 protein were studied in 41 GI stromal tumors, using the avidin-biotin method, and compared with normal host smooth muscle and nerve and with esophageal and uterine leiomyomas. Twenty gastric and one rectal tumor stained diffusely with vimentin and actin, but not with desmin, and had scattered strongly S-100-positive cells that might either be trapped Schwann cells or tumor cells. Twenty small bowel tumors stained similarly to the gastric tumors with regard to vimentin, actin, and desmin, but most (17/20) had a unique, strongly positive geographic staining pattern with S-100. No differences in staining were noted between benign and malignant tumors in either gastric or small bowel sites, and most histologic patterns in tumors from similar locations stained similarly. These results suggest that GI stromal tumors are not truly "leiomyomas and leiomyosarcomas," but relatively undifferentiated tumors, with the expression of various antigens depending on their location in the gut.

Correlation of histopathologic evidence of disease activity with the presence of immunoglobulin-containing cells in the colons of patients with inflammatory bowel disease.

Immunofluorescence of formalin-fixed, paraffin-embedded tissues was performed to study the plasma cell population in 114 colonic specimens from 58 patients. Correlation of the histopathologic stage of disease activity with the isotypes and numbers of immunoglobulin-containing cells in the lamina propria demonstrated highly significant (P less than 0.001) increases in the mean numbers of IgG- (18-fold), IgA- (twofold) and IgM- (sixfold) containing cells in specimens from patients with active inflammatory bowel disease as compared with control specimens. Increased numbers of immunoglobulin-containing cells were uncommon in inactive inflammatory bowel disease and in reactive mucosa. No deposition of immunoglobulin-containing immune complexes was found at any stage of disease activity. These findings suggest that immune complex-mediated damage does not play a major role in the epithelial damage in inflammatory bowel disease. In future studies, it will be of importance to determine whether the antibody from immunoglobulin-containing cells seen in patients with inflammatory bowel disease can effect damage via an antibody-dependent cell-mediated cytotoxicity mechanism.

Dysplasia in inflammatory bowel disease: standardized classification with provisional clinical applications.

Assessment of epithelial dysplasia in ulcerative colitis has been hindered by inconsistencies in and disagreements about nomenclature and interpretation. To resolve these issues, pathologists from ten institutions participated in three exchanges of multiple slides and, following each exchange, in discussions of the results. A classification system for the epithelial changes that occur in ulcerative colitis was developed, which should be applicable to other forms of inflammatory bowel disease as well. The classification makes use of standardized terminology, addresses specific problem areas, and offers practical solutions. The reproducibility of the system was studied by means of examinations of both inter- and intra-observer variations. The clinical implications of the findings were incorporated into suggestions for patient management. The basis of the classification is that the term "dysplasia" is reserved for epithelial changes that are unequivocally neoplastic and may therefore give rise directly to invasive carcinoma. Specimens are categorized as negative, indefinite, or positive for dysplasia. The negative category includes all inflammatory and regenerative lesions and indicates that only continued regular surveillance is required. The indefinite category is applied to epithelial changes that appear to exceed the limits of ordinary regeneration but are insufficient for an unequivocal diagnosis of dysplasia or are associated with other features that prevent such unequivocal diagnosis. Clinically, it indicates that early repeat biopsy is often required to assess the changes more accurately. The positive category is divided into two subcategories: 1) high-grade dysplasia, for which colectomy should be strongly considered after confirmation of the diagnosis, and 2) low-grade dysplasia, which also requires confirmation and early repeat biopsy or colectomy, depending on other findings.

Adenocarcinoma of the distal esophagus and gastric cardia. Comparison of results of transhiatal esophagectomy and thoracoabdominal esophagogastrectomy.

Between 1965 and 1984, 72 patients underwent operation for adenocarcinoma of the distal esophagus or gastric cardia. A standard transthoracic esophagogastrectomy and esophagogastrostomy was performed in 43 and a transhiatal esophagectomy without thoracotomy and partial proximal gastrectomy was performed in 29. There was no significant difference between the two groups in age, sex, or TNM tumor staging. The perioperative complication rate was 86% in the esophagogastrectomy patients and 48% in the transhiatal esophagectomy patients (p less than 0.05). Mortality was higher in the esophagogastrectomy group (14%) than in the transhiatal esophagectomy group (7%). Average operative blood loss was greater in the esophagogastrectomy patients (2,510 versus 1,187 ml). Average postoperative hospitalization was longer for the esophagogastrectomy patients (22.2 days versus 12.3 days). Both differences are statistically significant (p less than 0.05). Late results, as evaluated by life-table analysis, showed no significant difference in survival between the two groups of patients. Because the morbidity and mortality rates of transhiatal esophagectomy are as low as or lower than those for esophagogastrectomy, late survival is as good, and palliation is superior (less suture-line tumor recurrence and reflux esophagitis), we believe that transhiatal esophagectomy is the preferred operative approach in patients with adenocarcinoma of the distal esophagus or gastric cardia.