—  SHORT COURSE  —

A PRACTICAL APPROACH TO GASTROINTESTINAL PATHOLOGY


GASTROINTESTINAL STROMAL TUMORS
OF THE STOMACH AND SMALL BOWEL

John R. Goldblum, M.D.





Case 11 - Benign Spindle - Cell GIST (Cellular Leiomyoma)
Case 11 - Benign Spindle - Cell GIST (Cellular Leiomyoma)
Case 11 - Benign Spindle - Cell GIST (Cellular Leiomyoma)


Case 12 - Malignant Spindle - Cell GIST (Leiomyosarcoma)
Case 12 - Malignant Spindle - Cell GIST (Leiomyosarcoma)

Perhaps no other area in surgical pathology has caused as much confusion and controversy as has the subject of gastrointestinal stromal tumors (GIST). These tumors are a heterogeneous group of neoplasms, and prediction of clinical behavior requires a multiparametric evaluation of gross and microscopic features. However, the same criteria for malignancy do not apply to stromal tumors from different sites within the gastrointestinal tract, and the relative importance of each of these features is somewhat controversial.

In the past, cellular spindle cell neoplasms of the gastrointestinal tract have often been referred to as "leiomyoma" or "leiomyosarcoma" based upon their purported morphologic resemblance to smooth muscle tumors in other anatomic sites. With the advent of electron microscopy and immunohistochemistry, an enormous number of publications attempted to analyze the characteristics of the constituent cells of these tumors. While a number of studies suggested that these tumors were either derived from or differentiated towards smooth muscle, an equal number of studies suggested neural derivation or differentiation. As discussed below, more recent evidence suggests that the vast majority (if not all) of these tumors are either derived from or differentiate toward interstitial cells of Cajal (ICC). The non-committal term "gastrointestinal stromal tumor" has been adopted to describe this heterogeneous group of neoplasms and is currently the term of preference. Although a wide array of mesenchymal tumors may arise within the gastrointestinal tract, GIST are by far the most common mesenchymal tumors of this site.

Clinical Features of GIST
These tumors predominantly arise in middle-aged or elderly individuals and are quite rare in young adults, although there are reports of these tumors arising in very young patients, including congenital tumors.1  The stomach is the most common site, as approximately two-thirds of these tumors arise in this location. Less than one-third of GIST arise in the small intestine, particularly the duodenum, and less than 10% occur in the esophagus, colon and anorectum.2  It must also be kept in mind that identical tumors may arise in extra-gastrointestinal sites, including the omentum and mesentery.3,4  There does not appear to be a gender predilection, except for those rare gastric epithelioid neoplasms that arise in the setting of Carney's triad, nearly all of which occur in females.5  There are no known risk factors, although there may be some genetic influence given rare reports of familial GIST,6  as well as tumors associated with von Recklinghausen's disease (neurofibromatosis type 1; NF1).

The presenting manifestations depend upon the site of involvement within the gastrointestinal tract, the size of the tumor and the portion of the gut wall from which the tumor arises. A significant percentage of benign tumors are asymptomatic and are found incidentally at surgery performed for other reasons, probably because many of these lesions are quite small. However, it must be kept in mind that even benign tumors can cause symptoms, the most common of which include gastrointestinal bleeding and abdominal pain. Malignant tumors are rarely asymptomatic;7  most present with either gastrointestinal bleeding, abdominal pain or an abdominal mass. These signs and symptoms may lead to endoscopy and biopsy, and in some cases, a histologic diagnosis of GIST can be made if a deep enough biopsy is obtained or if the neoplasm infiltrates the overlying mucosa. Tumors that infiltrate the mucosa (and are thus identifiable in an endoscopic biopsy specimen) are virtually always malignant, although in general it is not necessary to distinguish a benign from malignant GIST on an endoscopic biopsy specimen, or for that matter in a frozen section. Radiographic imaging studies including barium contrast examination, computed tomography and endoscopic ultrasound are commonly used to evaluate and potentially diagnose these neoplasms. In fact, it has recently been suggested that a combination of features identified by endoscopic ultrasound may be highly predictive of malignant tumors.8  Some tumors can be diagnosed by fine needle aspiration cytology, although separation of benign from malignant is usually not possible with this technique.9 

Gross Features of GIST
These tumors may be situated within any portion of the gut wall but most are centered within the submucosa or muscularis propria. Some are predominantly extramural, and extremely large tumors may even extend to or infiltrate adjacent organs.

Most GIST are well circumscribed; some are multinodular. The overlying mucosa can be intact or ulcerated. On cut section, these tumors lack the characteristic gross appearance of uterine smooth muscle tumors in that they lack a bulging, whorled cut surface; instead, the cut surface is typically granular and may show areas of hemorrhage, necrosis or cystic change.

Microscopic Features of GIST
These tumors show a spectrum of histologic features and have distinctive appearances depending upon where they arise within the gastrointestinal tract. Thus, rather than discussing these tumors as a group, they will be separated by site.

Gastric GIST
These tumors can essentially be divided into four main categories including benign (cellular spindle cell tumor) and malignant (spindle cell sarcoma) spindled tumors and benign and malignant epithelioid tumors. Each of these tumor types has a characteristic morphologic appearance and usually can be separated from one another by assessing a combination of histologic features. As such, the characteristic features of each of these tumor types will be discussed below, followed by a review of some of the morphologic parameters that are useful in separating benign from malignant gastric GIST.

Cellular Spindle Cell Tumor
These tumors are composed of highly cellular spindled cells with abundant pale to eosinophilic fibrillar cytoplasm. The cells are uniform in size and shape and are typically arranged into short fascicles or whorls with frequent nuclear palisading reminiscent of that seen in schwannoma. Perinuclear vacuoles are often present and indent the nucleus at one pole; these vacuoles are an artifact of fixation since they are not present in frozen sections. The nuclei are uniform with evenly distributed chromatin and inconspicuous nucleoli, and mitotic activity is typically low with two or fewer mitotic figures/50 HPF in virtually all cases. The cells are often separated by hyalinized or calcified stroma as well as large areas of liquefactive necrosis with pools of acellular material separating perivascular tumor islands.

Tumors with this characteristic pattern virtually always act in a clinically benign fashion. In the series of 48 "cellular leiomyomas" with follow-up reported by Appelman and Helwig,10  only one tumor metastasized. Although this metastatic lesion histologically resembled all of the other tumors, it was 17 cm and had 5 mitotic figures/50 HPF.

Spindle Cell Sarcoma
When compared to the cellular spindle cell tumor, spindle cell sarcomas are generally larger and more densely cellular as the constituent cells tend to be more crowded because they have less cytoplasm. The nuclei vary more in size and shape and are often vesicular in appearance. Perinuclear vacuoles are less conspicuous or even absent, and broad areas of tumor cell necrosis are typical. The cells may be arranged into fascicular or storiform growth patterns. Mitotic figures are easily found, usually more than 10 mitoses/50 HPF. Using a combination of these features, spindle cell sarcomas can usually be readily distinguished from cellular spindle cell tumors.11 

Benign Epithelioid Gastric Stromal Tumor
These tumors are the most common stromal tumor encountered in the stomach and are composed predominantly of epithelioid cells with arrangement into nests or sheets. The cells often have a condensed rim of eosinophilic cytoplasm adjacent to the nucleus with peripheral cytoplasmic clearing that is not present in frozen sections or on ultrastructural examination.12  Some cases are composed of cells with abundant eosinophilic cytoplasm, and the nucleus is often pushed to an eccentric location. The nuclei are round with small nucleoli, although scattered multinucleated giant cells or cells with bizarre nuclei may be present. Mitotic figures are exceedingly rare (usually two or fewer mitotic figures/50 HPF). Stromal alterations including liquefaction, hyalinization and calcification may also be seen.

Malignant Epithelioid Gastric Stromal Tumor
These tumors are composed of cells with less cytoplasm, and thus the cells are more tightly packed resulting in a more cellular appearance at low magnification. Cellular arrangement varies between different areas of the same tumor, as the cells may be arranged into large sheets or small acinus-like clusters deposited in an acid mucopolysaccharide-rich myxoid stroma. The nuclei are more hyperchromatic than those seen in the benign counterpart, and in general, the nuclei are actually monotonous in appearance. Rare examples of this tumor are diffusely pleomorphic, but scattered bizarre cells are more commonly seen in benign epithelioid stromal tumors. Mitotic activity may overlap that seen in benign epithelioid tumors, and thus, unless mitotic figures are numerous, they cannot be used alone to separate benign from malignant epithelioid gastric stromal tumors. Importantly, benign-appearing areas are often encountered in these tumors, and thus extensive sampling of these lesions is required to identify the malignant component. A subgroup of gastric epithelioid stromal tumors occurs in patients with Carney's triad (malignant gastric epithelioid stromal tumor, pulmonary chondroma and functioning extra-adrenal paraganglioma).5  These patients tend to be young women, and the tumors are often multifocal, purely epithelioid and have a low risk of metastasis. Even in those patients who develop metastatic disease, long-term survival is the rule.

Assessment of Individual Parameters

Tumor Size
Although sarcomas tend to be larger than benign stromal tumors, there is significant overlap in the individual case, and size alone cannot be used to distinguish benign from malignant. However, when a gastric stromal tumor is felt to be histologically malignant based upon other parameters, tumor size is the single best predictor of metastatic risk. For example, in a study of 44 malignant gastric stromal tumors by Appelman and Helwig, metastases occurred in 20% of those tumors that were less than 6 cm and in 85% of those that were 6 cm or larger.11  Roy and Summers studied the relationship between tumor size and metastatic potential in 211 reported malignant gastric stromal tumors as well as 30 of their own.13  Fifteen percent of sarcomas less than 2.5 cm, 29% of those between 2.5 and 5 cm, 65% of those between 5 and 10 cm, and virtually all sarcomas larger than 10 cm metastasized. No other histologic parameter correlated as strongly with metastatic risk or survival. Thus, it is clear that larger malignant gastric stromal tumors metastasize more often than smaller ones, comparable to the relationship between tumor size and metastases for almost all other malignancies.

Mitotic Counts
Virtually all studies have found that increased mitotic counts are associated with decreased patient survival.14,15  In a large study of 127 epithelioid gastric stromal tumors by Appelman and Helwig,16  all tumors with more than 10 mitoses/50 HPF ultimately metastasized. However, 13% of tumors with between 1 and 5 mitoses/50 HPF and 2% of tumors with no identifiable mitotic figures in 50 HPF metastasized. However, it is difficult to compare results among different studies since different methods of counting mitotic figures were used, including the number of fields counted (10 vs 50 HPF) and the actual size of the HPF. Factors such as fixation time and the type of fixative may also account for variability in mitotic counts.17  Reproducibility in counting mitoses from one observer to the next is not optimal. Finally, there may be significant intratumoral variability; one area of the tumor may have few mitotic figures and another area may be mitotically active. In general, mitotic counts seem to be highest in the most cellular areas of the neoplasm, and most studies perform mitotic counts in the most cellular portions of the tumors. In the study by Trupiano et al, patients whose tumors had 5 or more mitoses/50 HPF were significantly more likely to develop metastatic disease or die as a direct result of their tumor.18  However, it must be kept in mind that some malignant tumors have very few mitotic figures (fewer than 5/50 HPF), and thus this parameter cannot be used alone to separate benign from malignant, except when mitotic counts are high (10 or more mitoses/50 HPF).

Other methods of measuring cellular proliferation, including antibodies to proliferating cell nuclear antigen (PCNA) and the Ki-67 antigen (MIB 1) and nucleolar organizing region content (AgNOR), have been evaluated in this group of tumors. Virtually all studies utilizing these techniques have found a correlation between cellular proliferation and prognosis,19-21  but it is unclear whether these techniques provide additional prognostic information over mitotic counts alone. In addition, the lack of uniform technical and quantitative methods limits the utility of these markers.

Cellularity
Cellularity has been found to be an important histologic parameter in assessing clinical behavior of gastric stromal tumors. However, the same cellularity cannot be used across the board for all morphologic types in distinguishing benign from malignant. For instance, cellular spindle cell tumors of the stomach, by definition, are highly cellular, and this feature alone does not appear to be useful in separating benign from malignant gastric spindle cell tumors. On the other hand, cellularity is among the most important features in distinguishing benign from malignant gastric epithelioid neoplasms. Determination of cellularity is admittedly difficult and more subjective than other parameters. The separation of levels of cellularity (low vs moderate and moderate vs high) requires experience with these tumors. Furthermore, cellularity may vary significantly within different portions of the same tumor, and it is unclear how much of the tumor has to have "high cellularity" to have an adverse effect on prognosis.

Other Morphologic Parameters
Several other features should also be evaluated in every GIST and are part of the multiparametric analysis required to accurately predict clinical behavior. Mucosal invasion with infiltration of tumor cells between and around mucosal glands is a highly specific marker of a malignant tumor.18,22-24  However, this is not a particularly sensitive marker of malignancy, since many malignant GIST do not infiltrate the overlying mucosa in this manner. Furthermore, the presence of mucosal ulceration may make this evaluation more difficult since an ulcer can obscure the site of mucosal infiltration. Although tumor cell necrosis is more commonly seen in malignant tumors, large benign tumors may also have areas of necrosis. Marked nuclear atypia in a spindle cell lesion suggests a sarcoma. On the other hand, when evaluating gastric epithelioid tumors, scattered bizarre or multinucleated cells are more characteristic of benign lesions. Invasion of adjacent organs, although an uncommon finding, also appears to correlate with an aggressive clinical course. Finally, knowledge of whether or not the tumor has metastasized at the time of surgical excision is the single most important prognostic variable. Although intuitively obvious, this is information that is not often sought or known by the surgical pathologist at the time of evaluation.

Sarcoma Grading
There is no uniform grading scheme that is used to separate low-grade from high-grade GIST. It is unclear whether features that are commonly utilized in the grading of soft tissue sarcomas, such as cellularity, nuclear pleomorphism, mitotic counts and necrosis, are also useful in grading GIST. Thus, we do not routinely grade these sarcomas but, rather, use tumor size as the most reliable indicator of metastatic risk in tumors that are felt to be histologically malignant, based upon the aforementioned parameters.

Summary of Gastric GIST
The vast majority of gastric GIST can be comfortably categorized into one of the four major subtypes mentioned above. It has been our experience, based on our consultation practice, that the greatest difficulty pathologists have is not telling which tumors are malignant, but telling which are benign. Because of this uncertainty, tumors that are clearly benign are often diagnosed as being either of uncertain malignant potential or as overtly malignant. Tumors that are likely to act in a clinically benign fashion can be diagnosed using a constellation of histologic features, as outlined above, thus minimizing the number of tumors that are of "borderline malignancy" or "uncertain malignant potential."

Small Bowel GIST
The small bowel is the second most common site of GIST, as approximately one-third of such tumors arise in this location. Although these tumors occur throughout the small bowel, a disproportionately large number occur in the duodenum given its short length in comparison to the jejunum and ileum. Unlike gastric GIST, most small bowel GIST are composed of spindled cells, and predominantly epithelioid variants are rare. As a group, a greater percentage of small bowel GIST are malignant when compared to those found in the stomach. These tumors have different histologic features than gastric GIST and thus should be evaluated independently.

Benign Small Bowel GIST
The "classic" benign small bowel GIST is small (<5 cm) and is composed of a uniform population of cytologically bland spindled cells with abundant eosinophilic cytoplasm and an overall low tumor cellularity. The cells are usually divided into nests by fine fibrovascular septa resulting in an organoid growth pattern reminiscent of paraganglioma. Eosinophilic collagen globules (so-called skeinoid fibers) are characteristic and are often numerous. Mitotic figures are low (fewer than 5/50 HPF), and there is an absence of tumor cell necrosis and mucosal invasion. Tumors that fulfill all of the above criteria are predictably benign.22-24 

Malignant Small Bowel GIST
Most malignant small bowel GIST are composed of a highly cellular proliferation of spindled cells with more overt cytologic atypia than that seen in their benign counterparts. The nuclei are larger and have more coarsely clumped chromatin, and mitotic figures are usually easily identified (more than 5/50 HPF). The cells are typically arranged into longer fascicles, as opposed to the uniform organoid growth pattern seen in benign tumors. Skeinoid fibers are few in number or completely absent, and many (but not all) malignant small bowel GIST have tumor cell necrosis and/or mucosal infiltration. In addition, tumors with a conspicuous epithelioid component comprising more than 25% of the tumor are virtually always malignant.22,23  Importantly, many malignant GIST have areas that resemble those seen in benign small bowel GIST, raising the possibility of a sequence of progression analogous to that seen in the dysplasia-carcinoma sequence found in some epithelial tumors. Most small bowel GIST can be categorized as either benign or malignant, and like the stomach, only rare small bowel GIST fit into the "borderline" or "uncertain malignant potential" category.

Immunohistochemical Features of GIST
The literature on the immunohistochemical features of GIST has evolved over the past twenty years. Earlier studies focused on the expression of myogenic and neural antigens, and the results of these studies were often discrepant.25-27  Frequently, CD34, a hematopoietic stem cell marker, was found to be expressed in the majority (approximately 70%) of GIST from all sites.28-31 

More recently, the product of the c-kit gene (CD117) has been detected in the majority of GIST from all sites regardless of morphologic features and degree of malignancy (Fig 15).32-35  In fact, some authors have equated immunoreactivity for CD117 as definitional of GIST.2  The c-kit gene encodes for a protein that belongs to the family of receptor tyrosine kinases and is a transmembrane growth factor receptor for stem cell factor.36  This antigen marks interstitial cells of Cajal (ICC), gastrointestinal pacemaker cells that regulate intestinal motility and are normally found in an around the myenteric plexus as well as the outer smooth muscle layer.37,38  Because of the similar immunohistochemical and ultrastructural features of ICC and GIST, the latter have been proposed to be derived from or differentiate toward ICC.32,33  Given that ICC have myoid and neural features by electron microscopy, this hypothesis is attractive in explaining years of conflicting immunohistochemical and ultrastructural data regarding the histogenesis of GIST. An alternative hypothesis is that GIST originate from primitive stem cells that can differentiate into ICC and smooth muscle cells.39  Regardless, the absence of CD117 staining in a tumor suspected of being a GIST should engender consideration of an alternative diagnosis. Furthermore, as discussed below, given the therapeutic trials aimed at countering the activation of KIT tyrosine kinase activity, it has become important if not essential to document the expression of this antigen in these tumors.40 

Myoid markers may be found in a subset of GIST. Up to 20 to 30% of these tumors stain for smooth muscle actin, either focally or diffusely; expression of this antigen seems to be reciprocally related to CD34 expression.2  Desmin is found in less than 10% of these tumors, and staining is typically limited to scattered tumor cells with more prominent staining in epithelioid neoplasms. Heavy caldesmon, an actin-binding cytoskeleton-associated protein, is also detected in a subset of these tumors, supporting at least partial smooth muscle differentiation.41 

Neural antigens have also been reported in some of these tumors; S-100 protein may be found in up to 10% of GIST, typically with focal immunoreactivity. These tumors are usually negative for neurofilament protein and glial fibrillary acidic protein. Finally, although scattered tumor cells may stain for cytokeratins, particularly in malignant epithelioid GIST, the co-expression of CD117 should prevent confusion with a carcinoma.

Genetic Features of GIST
Activating mutations of the c-kit gene, most commonly in exon 11 (juxtamembrane domain), result in ligand-independent activation of the tyrosine kinase activity, and this event appears to be central in the pathogenesis of these tumors.43  Much less commonly, mutations in exon 9 (extracellular domain) and 13 (kinase domain) can be detected.43,44  There are reports of patients with familial GIST with activating germline c-kit mutations.6  Mutations in this gene are not found in other tumors that may be potentially confused with GIST, including leiomyosarcomas, fibromatoses and nerve sheath tumors. Some have found c-kit mutations to preferentially occur in malignant GIST;45,46  Taniguchi et al found these mutations to be of independent prognostic significance.47  These activating mutations have become targets for therapeutic intervention using tyrosine kinase inhibitors (STI571) with initial promising results.48  Immunohistochemical and/or molecular genetic evaluation of the KIT protein or c-kit gene may soon become a routine part of the evaluation of these tumors in order to determine who is eligible or likely to benefit from this therapy.

Comparative genomic hybridization studies have found other genetic alterations in GIST, including losses in chromosomes 14q and 22q.49-51  El-Rifai et al found DNA copy number gains in various loci to occur preferentially in malignant GIST.52  Losses at several other loci including 1p53  and 17p54  (the site of the NF1 gene), have also been detected in some GIST.

Gastrointestinal Autonomic Nerve Tumors (GANT)
GANT were originally described under the name "plexosarcoma" based upon ultrastructural resemblance to cells of the autonomic nervous system.55,56  These tumors have a heterogeneous histologic appearance similar to GIST and may arise anywhere in the gastrointestinal tract. The characteristic ultrastructural features include complex interdigitating cell processes with bulbous synaptic terminals, neurosecretory granules, rudimentary cell junctions and intermediate filaments.57  Although originally believed to be distinct from GIST, recent evidence supports the concept that GANT merely represents a phenotypic variant of GIST. Lee et al found these tumors to be consistently CD117 immunoreactive, and a substantial percentage were found to harbor GIST-specific gain-of-function mutations in the juxtamembrane domain of the c-kit gene.58  Given the rarity of ultrastructural evaluation of GIST, this tumor is likely much more common than previously believed, since they can only be recognized (and in fact are defined) by their characteristic ultrastructural features. Similar to GIST, rare examples of GANT have been associated with Carney's triad (intestinal stromal tumor associated with extra-adrenal paraganglioma and pulmonary chondroma)59  and NF1.60  There is also a report of multiple familial GANT associated with intestinal neuronal dysplasia.61 

Differential Diagnosis of GIST
GIST with epithelioid features may be confused for a lymphoma, melanoma or carcinoma, particularly on a small endoscopic biopsy specimen. Immunohistochemical analysis is invariably required for this distinction. Expression of CD117 would confirm a diagnosis of epithelioid GIST. On very rare occasions, an epithelioid GIST may express CD34 in the absence of CD117. In addition, rare examples of epithelioid GIST may show focal cytokeratin immunoreactivity, and thus the coexpression of either CD34 or CD117 becomes critical in order to avoid misdiagnosing such cases as carcinoma. The absence of strong and diffuse S-100 protein immunoreactivity (as well as staining for melanocytic markers such as HMB45 and Melan-A) would militate against a diagnosis of malignant melanoma.

Spindled GIST must be distinguished from other spindle cell proliferations of the gastrointestinal tract. Intra-abdominal fibromatoses are spindle cell proliferations that usually occur in the mesentery or retroperitoneum, but may on occasion extensively involve the gut wall, mimicking GIST. Grossly, intra-abdominal fibromatoses are firm, tan and homogeneous, thus lacking the typical gross appearance of GIST. The constituent cells are monotonous spindled cells arranged into sweeping fascicles with bland nuclear features and scant mitotic figures. Although some authors have reported CD117 immunoreactivity in a significant percentage of intra-abdominal fibromatoses,62  most studies (and our own experience) have not found this antigen to be expressed in these lesions. These discrepant data have been ascribed to differences in antibody clone and antigen retrieval techniques. The distinction between an intra-abdominal fibromatosis and GIST is of clinical relevance, given that the former lacks metastatic capability, although it does have a propensity for local recurrence.

True smooth muscle tumors of the gastrointestinal tract are much less common than GIST. Benign smooth muscle tumors (leiomyomas) are most common in the esophagus and anorectum, although identical-appearing tumors may occur anywhere throughout the gastrointestinal tract. Overtly malignant smooth muscle tumors (leiomyosarcomas) are exceedingly rare, and there are very few reports describing these tumors as an entity distinct from GIST. Like soft tissue leiomyosarcomas, these tumors are composed of cells with eosinophilic fibrillar cytoplasm and arrangement into fascicles. The cells are negative for CD117 and CD34 but stain strongly and diffusely for smooth muscle actin and often for desmin. The absence of c-kit mutations in leiomyosarcomas supports the contention that they are distinct from GIST.

Benign and malignant nerve sheath tumors are also rare in the gastrointestinal tract. Schwannomas occur most commonly in the stomach, but occasionally arise in the esophagus and intestines. They usually arise within the submucosa and/or muscularis propria and are often covered by an intact mucosa. These tumors are composed of a proliferation of S-100 protein immunoreactive wavy spindled cells, often associated with a sprinkling of lymphocytes and a peripheral lymphoid cuff with germinal centers. Unlike GIST, these cells do not express CD117. Very rarely, a malignant peripheral nerve sheath tumor may arise in the gastrointestinal tract. These tumors resemble their counterparts in the peripheral soft tissues, do not express CD117 and are variably immunoreactive for S-100 protein.

Summary
GIST are a heterogeneous group of tumors composed of either spindled or epithelioid cells that have distinctive histologic appearances and clinical behaviors in different portions of the gastrointestinal tract. These tumors should be evaluated in a site-specific fashion, and prediction of clinical behavior is best achieved by evaluation of multiple parameters. The vast majority of these tumors can be reliably placed into benign or malignant categories using a constellation of morphologic features. These tumors almost uniformly express CD117, and mutations of the c-kit gene are central to their pathogenesis and likely the key to effective therapeutic intervention.

REFERENCES

  1. Bates AW, Feakins RM, Scheimberg I. Congenital gastrointestinal stromal tumour is morphologically indistinguishable from the adult form, but does not express CD117 and carries a favourable prognosis. Histopathology 2000;37:316-322.
  2. Miettinen M, Lasota J. Gastrointestinal stromal tumors - definition, clinical, histological, immunohistochemical and molecular genetic features and differential diagnosis. Virch Arch 2001;438:1-12.
  3. Miettinen M, Moniham JM, Sarlomo-Rikala M, et al. Gastrointestinal stromal tumors/smooth muscle tumors/GISTs in the omentum and mesentery - clinicopathologic and immunohistochemical study of 26 cases. Am J Surg Pathol 1999;22:1109-1118.
  4. Reith JD, Goldblum JR, Lyles RH, et al. Extragastrointestinal (soft tissue) stromal tumors: an analysis of 48 cases with an emphasis on histologic predictors of outcome. Mod Pathol 2000;13:577-585.
  5. Carney JA. Gastric stromal sarcoma, pulmonary chondroma and extra-adrenal paraganglioma (Carney triad): Natural history, adrenal cortical component and possible familial occurrence. Mayo Clin Proc 1999;74:543-552.
  6. Nashida T, Hirota S, Taniguchi M, et al. Familial gastrointestinal stromal tumours with germline mutation of the KIT gene. Nature Genet 1998;19:323-324.
  7. Appelman HD, Lewin K. Mesenchymal tumors and tumor-like proliferations of the stomach. In: AFIP Fascicle, Third Series: Tumors of the Esophagus and Stomach, 1996.
  8. Palazzo L, Landi B, Cellier C, et al. Endosonographic features predictive of benign and malignant gastrointestinal stromal tumours. Gut 2000;46:88-92.
  9. Dodd LG, Nelson RC, Mooney EE, et al. Fine-needle aspiration of gastrointestinal stromal tumors. Am J Clin Pathol 1998;109:439-443.
  10. Appelman HD, Helwig EB. Cellular leiomyomas of the stomach in 49 patients. Arch Pathol Lab Med 1977;101:373-377.
  11. Appelman HD, Helwig EB. Sarcomas of the stomach. Am J Clin Pathol 1977;67:2-10.
  12. Cornog JL. Gastric leiomyoblastoma: a clinical and ultrastructural study. Cancer 1974;34:711-719.
  13. Roy M, Sommers SC. Metastatic potential of gastric leiomyosarcoma. Pathol Res Pract 1989;185:874-877.
  14. Shiu MH, Farr GH, Papachristou DN, et al.. Myosarcomas of the stomach: natural history, prognostic factors and management. Cancer 1982;49:177-187.
  15. Ranchod M, Kempson RL. Smooth muscle tumors of the gastrointestinal tract and retroperitoneum. Cancer 1977;39:255-262.
  16. Appelman HD, Helwig EB. Gastric epithelioid leiomyomas and leiomyosarcomas (leiomyoblastomas). Cancer 1976;38:708-728.
  17. Baak JPA. Editorial: Mitosis counting in tumors. Hum Pathol 1990;21:683-685.
  18. Trupiano JK, Goldblum JR, Stewart RE, et al. Gastric stromal tumors: correlation of features with aggressive clinical behavior. Mod Pathol 2000;13:516A.
  19. Yu CCW, Fletcher CDM, Newman PL, et al. A comparison of proliferating cell nuclear antigen (PCNA) immunostaining, nucleolar organizing region (AgNOR) staining, and histologic grading in gastrointestinal stromal tumors. J Pathol 1992;166:147-152.
  20. Amin MB, Ma CK, Linden MD, et al. Prognostic value of proliferating cell nuclear antigen index in gastric stromal tumors: correlation with mitotic count and clinical outcome. Am J Clin Pathol 1993;100:428-432.
  21. Franquemont DW, Frierson HF. Proliferating cell nuclear antigen immunoreactivity and prognosis of gastrointestinal stromal tumors. Mod Pathol 1995;8:473-477.
  22. Goldblum JR, Appelman HD. Stromal tumors of the duodenum: a histologic and immunohistochemical study of 20 cases. Am J Surg Pathol 1995;19:71-80.
  23. Brainard JA, Goldblum JR. Stromal tumors of the jejunum and ileum: a clinicopathologic study of 39 cases. Am J Surg Pathol 1997;21:407-416.
  24. Tworek JA, Appelman HD, Singleton TP, et al. Stromal tumors of the jejunum and ileum. Mod Pathol 1997;10:200-209.
  25. Pike AM, Lloyd RV, Appelman HD. Cell markers in gastrointestinal stromal tumors. Hum Pathol 1988;19:830-834.
  26. Saul SH, Rast ML, Brooks JJ. The immunohistochemistry of gastrointestinal stromal tumors: evidence supporting an origin from smooth muscle. Am J Surg Pathol 1987;11:464-473.
  27. Franquemont DW, Frierson HF. Muscle differentiation and clinicopathologic features of gastrointestinal stromal tumors. Am J Surg Pathol 1992;16:947-954.
  28. Mikhael AI, Bacchi CE, Zarbo RJ, et al CD34 expression in stromal tumors of the gastrointestinal tract. Appl Immunohistochem 1994;2:89-93.
  29. Monihan JM, Carr NJ, Sobin LH. CD34 immunoexpression in stromal tumors of the gastrointestinal tract and in mesenteric fibromatoses. Histopathology 1994;l25:464-473.
  30. Van de Rijn M, Hendrickson MR, Rouse RV. CD34 expression by gastrointestinal tract stromal tumors. Hum Pathol 1994;25:766-771.
  31. Miettinen M, Virolainen M, Sarlomo-Rikala M. Gastrointestinal stromal tumors: value of CD34 antigen in their identification and separation from true leiomyomas and schwannomas. Am J Surg Pathol 1995;19:207-216.
  32. Sircar K, Hewlett BR, Huizinga JD, et al. Interstitial cells of Cajal as precursors of gastrointestinal stromal tumors. Am J Surg Pathol 1999;23:377-389.
  33. Kindblom L-G, Remotti HE, Aldenborg F, et al. Gastrointestinal pacemaker cell tumor (GIPACT): gastrointestinal stromal tumors show phenotypic characteristic of the interstitial cells of Cajal. Am J Pathol 1998;152:1259-1269.
  34. Miettinen M, Sobin LH, Sarlomo-Rikala M. Immunohistochemical spectrum of GISTs of different locations and their differential diagnosis. Mod Pathol 2000;13:536-541.
  35. Sarlomo-Rikala M, Kovatich A, Barusevicius A,et al. CD117: a sensitive marker for gastrointestinal stromal tumors that is more specific than CD34. Mod Pathol 1998;11:728-734.
  36. Kitamura Y, Hirota S, Nashida T. Molecular pathology of c-kit proto-oncogene in development of gastrointestinal stromal tumors. Ann Chir Gyne 1998;87:282-286.
  37. Mieda H, Yamagata A, Nishikawa F, et al. Requirement of c-kit for development of intestinal pacemaker system. Development 1992;115:369-375.
  38. Sanders KM. A case for interstitial cells as pacemakers and mediators of neurotransmission in the gastrointestinal tract. Gastroenterology 1996;111:492-515.
  39. Torihashi S, Nishi K, Tokutomi Y, et al. Blockade of KIT signaling induces transdifferentiation of interstitial cells of Cajal to a smooth muscle phenotype. Gastroenterology 1999;117:140-148.
  40. Seymour L. Novel anti-cancer agents in development: exciting prospects and new challenges. Cancer Treat Rev 1999;25:301-312.
  41. Miettinen M, Sarlomo-Rikala M, Kovatich A, et al. Heavy caldesmon and calponin in smooth muscle tumors and gastrointestinal stromal tumors. Mod Pathol 1999;12:756-762.
  42. Hirota S, Isozaki K, Moriyama Y, et al. Gain-of-function mutations of c-kit in human gastrointestinal stromal tumors. Science 1998;279:577-580.
  43. Lasoda J, Wozniak A, Sarlomo-Rikala M, et al. Mutations in exon 9 and 13 of KIT gene are rare events in gastrointestinal stromal tumors. A study of 200 cases. Am J Pathol 2000;157:1091-1095.
  44. Lux M, Rubin BP, Biase TL, et al. KIT extracellular and kinase domain mutations in gastrointestinal stromal tumors. Am J Pathol 2000;156:791-795.
  45. Ernst SI, Hobbs AE, Przygodzki RM, Emory TS, Sobin LH, O'Leary TJ. KIT mutation portends poor prognosis in gastrointestinal stromal/smooth muscle tumors. Lab Invest 1998;78:1633-1636.
  46. Lasoda J, Jasinski M, Sarlomo-Rikala M, et al. Mutations in exon 11 of c-kit occur preferentially in malignant versus benign gastrointestinal stromal tumors and do not occur in leiomyomas and leiomyosarcomas. Am J Pathol 1999;154:53-60.
  47. Taniguchi M, Nashida T, Hirota S, et al. Effect of c-kit mutation on prognosis of gastrointestinal stromal tumors. Cancer Res 1999;59:4297-4300.
  48. Joensuu H, Roberts PF, Sarlomo-Rikala M, et al. Effect of the tyrosine kinase inhibitor STI571 in a patient with metastatic gastrointestinal stromal tumor. N Engl J Med 2001;344:1052-1056.
  49. El-Rifai W, Sarlomo-Rikala M, Miettinen M, et al. DNA copy number losses in chromosome 14: an early change in gastrointestinal stromal tumors. Cancer Res 1996;56:3230-3233.
  50. El-Rifai W, Sarlomo-Rikala M, Andersson L, et al. High resolution deletion mapping of chromosome 14 in stromal tumors of the gastrointestinal tract. Genes Chromosomes Cancer 2000;27:397-401.
  51. Sarlomo-Rikala M, El-Rifai W, Andersson L, et al S. Different patterns of DNA copy number changes in gastrointestinal stromal tumors, leiomyomas and schwannomas. Hum Pathol 1998;29:476-481.
  52. El-Rifai W, Sarlomo-Rikala M, Andersson L, et al. Prognostic significance of DNA copy number changes in benign and malignant GISTs. Cancer Res 2000;60:3899-3903.
  53. O'Leary T, Ernst S, Przygodzki R, et al. Loss of heterozygocity at 1p36 predicts poor prognosis in gastrointestinal stromal/smooth muscle tumors. Lab Invest 1999;79:1461-1467.
  54. Kindblom LG, Remotti HE, Angervall L, et al. Gastrointestinal pacemaker cell tumor - a manifestation of neurofibromatosis 1. Mod Pathol 1999;12:77A.
  55. Walker P, Dvorak AM. Gastrointestinal autonomic nerve (GAN) tumor: ultrastructural evidence for a newly recognized entity. Arch Pathol Lab Med 1986;110:309-316.
  56. Herrera GA, Cerezo L, Jones JE, et al. Gastrointestinal autonomic nerve tumors: "plexosarcomas." Arch Pathol Lab Med 1989;113:846-853.
  57. Lauwers GY, Erlandson RA, Casper ES, et al. Gastrointestinal autonomic nerve tumors: a clinicopathologic, immunohistochemical and ultrastructural study of 12 cases. Am J Surg Pathol 1993;17:887-897.
  58. Lee JR, Joshi V, Griffin JW, et al. Gastrointestinal autonomic nerve tumor: immunohistochemical and molecular identity with gastrointestinal stromal tumor. Am J Surg Pathol 2001;25:979-987.
  59. Sigal A, Carello S, Caterina P, et al. Gastrointestinal autonomic nerve tumors: a clinicopathologic, immunohistochemical and ultrastructural study of 10 cases. Pathology 1994;26:239-247.
  60. Sakaguchi N, Sano K, Ito M, et al. A case of von Recklinghausen's disease with bilateral pheochromocytoma-malignant peripheral nerve sheath tumors of the adrenal and gastrointestinal autonomic nerve tumors. Am J Surg Pathol 1996;20:889-897.
  61. O'Brien P, Kapusta L, Dardick I, et al. Multiple familial gastrointestinal autonomic nerve tumors and small intestinal neuronal dysplasia. Am J Surg Pathol 1999;23:198-204.
  62. Yantiss RK, Spiro IJ, Compton CC, et al. Gastrointestinal stromal tumor versus intra-abdominal fibromatosis of the bowel wall: a clinically important differential diagnosis. Am J Surg Pathol 2000;24:947-957.


TABLE 1
Gastric Stromal Tumors
Cellular Spindle Cell Tumor
  • Highly cellular uniform, long spindled cells
  • Bland nuclei and pale fibrillar cytoplasm
  • Arranged in whorls, fascicles, palisades
  • Perinuclear vacuole which indents nucleus
  • Low mitotic activity


TABLE 2
Gastric Stromal Tumors
Benign Epithelioid Type
  • Sheets of epithelioid cells of low cellularity
  • Abundant cytoplasm
    • Perinuclear eosinophilic condensation
    • Peripheral clearing (artifact)
  • Well-defined cell membranes
  • Scattered bizarre cells
  • Rare mitotic figures


TABLE 3
Gastric Stromal Tumors
Spindle Cell Sarcomas
  • Cellularity overlaps with cellular spindle cell tumor
  • Vesicular, primitive nuclei
  • Increased mitotic counts (>5/50HPF)
  • Tumor cell necrosis
  • Lack perinuclear vacuoles and palisading


TABLE 4
Gastric Stromal Tumors
Malignant Epithelioid Type
  • Sheets of epithelioid cells of high cellularity
    • Less cytoplasm
    • Cells more closely packed
  • Uniform cell population
    • Often lacks bizarre cells
    • Rare cases with diffuse pleomorphism
  • Infrequent mitotic figures
  • Often areas with cellularity of benign tumor
TABLE 5
Gastric Stromal Tumors
Spindle Cell Type

  Benign Malignant
Cellularity high high
Pleomorphism 1+ 1+ = 3+
Mitoses <2/50 usually >5/50
Vacuoles present usually absent
Mucosal invasion absent + / -
Necrosis usually absent+ / -


TABLE 6
Gastric Stromal Tumors
Epithelioid Cell Type

  Benign Malignant
Cellularity low high
Pleomorphism 1+ 1+ - 3+
Mitoses low low or high
Mucosal invasion absent + / -


TABLE 7
Small Bowel Stromal Tumors

  Benign Malignant
Cellularity low low or high
Organoid pattern present focal / absent
Epithelioid cellsabsent + / -
Pleomorphism absent 1+ - 3+
Mitoses <5/50 usually >5/50
Mucosal invasion absent often present
Necrosis absent often present