Subepithelial Tumors of the Esophagus and Stomach

Gastrointestinal Stromal Tumor

Most mesenchymal GI tumors are pale, firm, spherical, or ovoid structures embedded in the wall of the affected organ. The microscopic appearance of muscle-like eosinophilic, spindle-shaped cells in uniform sheets and the proximity of the tumors to the muscular wall layers led early observers to believe that these tumors were of myogenic origin, hence the name “leiomyoma” and its variations (e.g., leiomyosarcoma, leiomyoblastoma). However, it later became clear that these neoplasms not only are not of obvious myogenic origin but also often lack any specific markers of differentiation whatsoever. Immunohistochemical analyses showed variable expression of smooth muscle features such as desmin and actin, and neural proteins such as S-100. For the sake of clarity, these lesions came to be referred to as gastrointestinal stromal tumors (GISTs), an acknowledgment that they originate in mesenchymal stroma. Endoscopically, GISTs appear as a dome-shaped, firm subepithelial mass. Central umbilication or frank ulceration is common, and there may be a lobulated or irregular appearance (see Fig. 31.1 ; Video 31.1 ). GISTs are noted to arise from the fourth wall layer with EUS imaging ( Fig. 31.3 ).

A significant breakthrough occurred with the discovery that most GI stromal tumors stain positive for a specific membrane protein, designated CD117. The protein was subsequently identified as one of a class of stem cell factor (SCF) receptors. SCF receptors can exist as either soluble or transmembrane receptors; CD117 was subsequently identified as KIT, a transmembrane tyrosine kinase receptor first found in a feline sarcoma (hence “kit” for kitten).

Abnormal cell growth is a fundamental element of cancer physiology, and hyperfunction of the KIT receptor can lead to neoplasia. This concept is supported by the observation that nearly all GISTs express KIT. It has further been observed that the interstitial cells of Cajal (ICC) share some phenotypic and ultrastructural similarities with GIST and normally express the KIT receptor; this observation has led to the current hypothesis that GISTs arise from ICC or from ICC precursor cells. This hypothesis is bolstered by the observation that ICC cells are seen outside the GI tract, and likewise extra-intestinal GIST have been described Finally, it has been noted that this gain-of-function mutation is not found in true leiomyomas. Some pathologists have suggested that CD117 positivity is required to confirm a diagnosis of GIST. It is now known, however, that a few otherwise obvious GISTs do not express KIT. Many of these tumors have been identified as expressing platelet-derived growth factor receptor α (PDGFRα), another distinct tyrosine kinase receptor that also mediates many of the same growth and antiapoptotic proliferation pathways. Several gain-of-function mutations have been reported in the PDGFRα gene, and the resulting tumors are otherwise indistinguishable from GISTs that are KIT-positive. There also remains a small subset of GISTs that express neither KIT nor PDGFRα; presumably other pathophysiologic pathways exist to account for these, including other tyrosine kinase gain-of-function mutations.

Nearly all upper GI tumors of this type occur in the stomach, but duodenal lesions have been described. Most esophageal stromal tumors lack the CD117 protein and may be true leiomyomas. The tumors are most often solitary except in the case of specific disease entities such as Carney's triad (GIST, pulmonary chondroma, and extra-adrenal paraganglioma). Germline KIT mutation kindreds have been described, however, in which case multiple GISTs are seen. Giant sizes of greater than 10 cm have been noted, but most tumors are less than 3 cm.

Pathologically, the tumor usually consists of uniform pale tissue, although hemorrhagic and necrotic areas may be seen. Microscopically, the cells are spindle shaped with uniform nuclei and general cytologic uniformity. Some cell groups may show epithelioid configurations (closely packed polygonal cells), and there may be nuclear pleomorphism. It has been observed more recently that the histologic pattern is sometimes related to the nature of the underlying genetic abnormality; KIT mutations at exon 13 or 17 more often show spindle cell morphology, whereas PDGFRα GISTs often exhibit epithelioid histology. Ultrastructural cellular abnormalities have also been linked to specific gene mutations. Recent research (2015) has linked some specific mutations with clinical features and outcomes; for instance, KIT exon 9 mutations show higher risk of progression and lower 5-year relapse-free survival than those with exon 11 mutations. It has long been known that malignant behavior in GISTs is difficult to predict given the relatively bland cytology and slow growth of these neoplasms. It has been reported that even small, benign-appearing stromal tumors have been known to metastasize. This discovery led to considerable confusion about the appropriate criteria to categorize these tumors as benign or malignant. Older pathologic scoring systems relied on numerous histologic features and were plagued with problems. More recent attention has focused on the size of the tumor and the number of mitoses observed (mitotic index), at least in part because these are easily quantifiable findings. In one study of 100 cases, tumors with more than 5 mitoses per 10 high-power fields (HPFs) were significantly more likely to metastasize, although 40% of malignant lesions in that study had fewer mitoses. In another study, multivariate analysis of various clinical and pathologic features in 122 specimens showed that more than 10 mitoses/50 HPFs correlated with poor outcome, whereas site, epithelioid histology, and tumor size were not independently predictive.

Attempts to correlate tumor marker status such as CD117 positivity with malignant behavior have produced generally confusing or negative results, as have studies of specific KIT mutations and other tumor markers. Older GIST scoring systems have been abandoned following a National Institutes of Health (NIH) consensus conference, which defined malignancy risk based on size and mitotic index alone. The NIH criteria divide tumors into four categories of malignant risk ( Table 31.3 ), an acknowledgment that even the most innocent lesion poses a slight but definite risk of malignant behavior.

If all such neoplasms entail malignant risk, prudence might dictate that they should always be resected. However, more recent data suggest that GISTs are more common than previously thought. Up to 10% of resection and autopsy specimens contain such tumors, and microscopic GISTs (also called seedling GISTs, minimal GISTs, or GIST tumorlets) can be seen in 35% of some patient groups. These incidental and microscopic GISTs display the same KIT and PDGFRα mutations as their larger counterparts. The finding that synchronous GISTs from the same patient regularly show different gene mutations (and are independent sporadic GISTs) confuses the picture further. GISTs that were earlier classified as metastatic or recurrent may have been distinct neoplastic events. If it turns out that most small and asymptomatic GISTs stay that way, a conservative approach may be the most prudent.

A second breakthrough in GISTs has been the development of tyrosine kinase inhibitors such as imatinib mesylate, which are effective in reducing KIT enzyme activity and are useful for tumor treatment. Imatinib targets the specific abnormal enzyme activity in the neoplasm and does not rely on generalized cytotoxicity for its effect. In an open-label study of 147 patients with unresectable malignant GISTs, an overall response rate of 38% was seen. Among responders, results are often dramatic. A number of newer compounds with similar tyrosine kinase inhibitory function have been examined in GISTs and have been found to be effective. Sunitinib and regorafenib have shown effectiveness, particularly in imatinib-resistant tumors, as have a variety of agents that target downstream oncogenetic proteins. The recognition of the malignant potential of GIST, combined with the availability of effective treatment even for unresectable disease, has compelled new thinking in the accurate diagnosis of this neoplasm.

Glomus Tumors

Glomus tumors are mesenchymal tumors that generally occur in the skin; a morphologically similar lesion has long been known in the stomach. They arise from modified cells of the glomus body that regulate arteriolar blood flow. They are typically located in the antrum and are generally small, although they can range up to 5 cm. These usually appear as circumscribed hypoechoic masses in the third or fourth layer and can have a characteristic peripheral halo around them. There may have arterial enhancement on computed tomography (CT) scanning similar to hemangiomas. An immunohistochemical study of 32 cases showed that all examined glomus tumors were negative for desmin, S-100, CD34, and KIT, suggesting a different histogenesis from leiomyomas, neuromas, and GISTs. They will typically stain positive for actin. A larger retrospective study of 1894 resected gastric mesenchymal tumors revealed 11 gastric glomus tumors. This same study showed that all glomus tumors examined stained positive for α-smooth muscle actin, laminin, collagen type IV, and vimentin. No C-kit mutations were present. Previous studies have suggested a malignancy risk that is low but not zero. Another study suggested that metastases were found in 38% of glomus tumors (52 cases) and that the possible risk factors for malignancy included deep location, size greater than 2 cm, atypical mitotic figures, moderate-to-high nuclear grade, and greater than 5 mitotic figures per 50 HPFs. Typically, the main goals in diagnosing glomus tumors of the stomach have been to differentiate them from other tumors such as GISTs.

Neural Tumors

Technically, most tumors that appear to be of neural origin continue to be classified as stromal tumors, although the more recent revolution in GIST understanding brought about by the discovery of CD117 mutations has led to confusing terminology in some cases, including neural tumors. Many experts consider stromal tumors that are positive for S-100 and negative for CD117 to be of neural origin.

Neural tumors may represent glial cell proliferation, sometimes in combination with other neural elements. Neuroma , neurofibroma , and schwannoma are largely interchangeable terms, although some pathologists observe differences among these lesions. When ganglion cells are present, the term ganglioneuroma is often applied. They seem to arise from ganglion cells in either the myenteric plexus or submucosal plexus. Schwannomas are slow-growing tumors derived from Schwann cells, which are typically found in the subcutaneous tissue of the distal extremities or head/neck. They are extremely rare in the GI tract, but the stomach is the most common site. Typical histology reveals atypical spindle cells in a microtrabecular pattern, S-100 positivity, negative glial fibrillary acidic protein, and negative c-KIT, as well as negative staining for CD34, desmin, and smooth muscle actin. Previous studies have found no malignant variants. Incidence of Schwannomas is predominantly based on case reports, but has been estimated to be 0.2% of all gastric tumors.