Jejunal Gastrointestinal Stromal Tumor: A Cause of Obscure Gastrointestinal Bleeding

Gastrointestinal (GI) hemorrhage is considered obscure1 when conventional investigations (esophagogastroduodenoscopy and colonoscopy) fail to detect bleeding lesions. On average, 27% of patients with obscure GI bleeding have small intestinal lesions.2 Diagnosing these lesions is frequently difficult because they tend to be inaccessible to routine endoscopy. Small bowel barium studies, radioactive isotope bleeding scans, selective visceral angiography, intraoperative enteroscopy, exploratory laparotomy, and more recently wireless capsule endoscopy3,4 all have variable sensitivities and specificities for detecting small intestinal lesions.

A 42-year-old woman presented with recurrent episodes of GI hemorrhage. Other than a remote cholecystectomy, there was no other pertinent past medical history including no history of a bleeding diathesis. She initially presented with 2 days of hematochezia associated with mild lower abdominal cramping and a hemoglobin and hematocrit of 4 mg/dL and 13%, respectively. Physical examination on admission revealed conjunctival pallor and orthostasis. There were no stigmata of liver disease or oral telangiectasias. Workup included an esophagogastroduodenoscopy with negative results, a colonoscopy showing blood throughout the colon, fresh bleeding from the terminal ileum, and biopsies yielding normal ileal mucosa, a bleeding scan suggesting a possible source in the left lower quadrant of the abdomen, a negative Meckel scan, and small bowel follow-through examination. After receiving an 8-unit packed red blood cell transfusion, the patient was discharged but readmitted 10 days later with recurrent GI hemorrhage. A celiac and superior mesenteric artery angiogram failed to reveal a source of bleeding, and after a 5-unit blood transfusion, the patient was transferred to our institution. A few hours after admission, she had recurrent hemorrhage. A repeat bleeding scan, angiogram, colonoscopy, and esophagogastroduodenoscopy with small bowel enteroscopy (to the proximal jejunum) were all unsuccessful in establishing an etiology. A laparoscopic evaluation of the small bowel with a planned laparoscopy-assisted enteroscopy revealed an 8-cm highly vascular midjejunal mass on the antimesenteric border of the bowel. The tumor was excised (Figure 1) and identified histologically as a gastrointestinal stromal tumor (GIST). The patient has done well postoperatively.

This case illustrates the diagnostic challenge presented by hemorrhage in the small intestine, even in a patient with recurrent massive bleeding. In as many as 5% of patients with obscure GI bleeding, a source cannot be identified despite extensive examination.5 When a lesion cannot be identified after standard upper endoscopy and colonoscopy, further evaluation depends on the briskness of bleeding (see algorithm). In those with active (brisk) bleeding, technetium-99 radionuclide scanning or angiography should be performed. In patients with a subacute presentation (or intermittent bleeding), the focus of investigation should be broadened to include the small intestine. The lesions most commonly identified as bleeding sites in the small bowel include tumors and vascular ectasias, which vary in frequency depending on age. In patients between 30 and 50 years of age, tumors are the most common abnormalities, whereas in patients less than 25 years of age, Meckel diverticula are the most common source of small bowel bleeding.6 Vascular ectasias predominate in older patients. Other rare causes in the differential diagnosis of obscure GI hemorrhage include hemosuccus pancreaticus, hemobilia, aortoenteric fistula, Dieulafoy lesion, extraesophageal varices, and diverticula (especially of the small intestine).

Small bowel examination can be accomplished with standard small bowel follow-through, enteroclysis, push enteroscopy, Sonde enteroscopy, wireless capsule endoscopy, and intraoperative enteroscopy. In the case described here, small bowel follow-through examination was nondiagnostic most likely because the jejunal lesion was predominantly extramucosal. This case also illustrates that in young patients with obscure small intestinal bleeding the source is frequently tumors rather than arteriovenous malformations. These tumors are problematic because they often grow in an extraluminal direction and only cause intermittent bleeding if they erode or ulcerate through the small bowel mucosa, as occurred in this patient. In a situation such as this, a noninvasive test, such as a computed tomography scan, may obviate the need for more invasive testing. We did not obtain this imaging study, although it should be considered in the algorithm for obscure GI bleeding, especially in young patients1 (see algorithm). Today, wireless capsule endoscopy3,4 (not commercially available at the time of our patient's workup) can be used to survey the entire small intestinal mucosa and has added another dimension to the noninvasive armamentarium for diagnosis of obscure GI bleeding. This modality is a new, conceptually simple approach in which the capsules pass safely without causing mucosal injury. In the present case, this approach probably would have allowed us to make a diagnosis without using laparoscopy, although that technique would still have been necessary to effect treatment.

GISTs are the most common mesenchymal tumors of the stomach and small intestine.7 They possess unique histologic, immunophenotypic, and molecular genetic features that set them apart from typical leiomyomas and schwannomas.7 GISTs are typically discovered incidentally during endoscopic, radiologic, or surgical procedures or are diagnosed in the evaluation of patients presenting with an abdominal mass, abdominal pain, or GI bleeding. Surgical resection with curative intent is the approach for patients presenting with localized GIST.8 

Grossly, benign GISTs are usually sharply demarcated spherical or ovoid-fusiform nodules 2 to 5 cm in diameter. Malignant GISTs tend to be larger than 10 cm in maximal diameter and often show mucosal ulceration and areas of necrosis or hemorrhage7 (Figure 2). Approximately 10% of all GISTs display malignant behavior.7 In addition to tumor size and presence of necrosis, other features suggestive of malignancy include a high level of mitotic activity, omental or mesenteric seeding, and presence of distant metastases.7 Nevertheless, proof of malignancy can be difficult because most malignant GISTs do not show pronounced cytologic atypia.

Microscopically, 2 principal histologic patterns occur in benign and malignant GISTs, spindle cell (70%) and epithelioid (30%), although some overlap occurs.9 The tumor in this patient exhibited malignant characteristics of large size (8 cm), focal cytologic atypia, and multifocal areas of necrosis. Features of high grade malignancy (ie, high mitotic count, infiltrative tumor edge, and angiolymphatic invasion) were not seen. The tumor was fully excised with negative resection margins.

A number of factors may have contributed to the hemorrhagic potential of this tumor. First, the small bowel location is associated with the highest incidence of bleeding. A review of published cases revealed that 87% of duodenal GISTs and 64% of other small bowel GISTs present with bleeding.8 In comparison, gastric, colonic, and rectal tumor locations have been associated with less than 50% incidence of bleeding.8 Second, tumor size and extension to the mucosal surface are factors that may help to promote GI hemorrhage. Third, most GISTs contain minimal stromal collagen, but delicate thin-walled vessels may be prominent, making stromal hemorrhage a common feature.9 

To better characterize these tumors and to positively identify them as GISTs, immunohistochemical staining is essential. Most GISTs are positive for CD34 (hematopoietic cell progenitor cell antigen) and CD117 (c-Kit protein).10 CD117 is quite specific for GISTs among mesenchymal tumors and is therefore the most useful marker for their definition.11 The expression of the c-Kit receptor tyrosine kinase has allowed for the discrimination of true nonmyogenic GISTs. Because the interstitial cells of Cajal (the GI pacemaker cells that regulate spontaneous GI motility) also express CD117, the relationship of GISTs to these cells has been proposed.12 GISTs are usually negative for smooth muscle actin (although 20% to 30% of GISTs may be positive) and, in contrast to schwannomas, are negative for S100 protein. Immunostained sections from the tumor in this patient (Figure 3) are consistent with the diagnosis of GIST.

This patient is clinically stable at 2 years postresection, and no evidence of recurrent disease has been found on follow-up computed tomography scans. We recommended that she be examined at the oncology clinic every 6 months.