Gastrointestinal Stromal Tumor (GIST) Workshop,
April 2 and 3, 2001, NIH, Bethesda, MD
This is essentially the same text that appears in:
Berman JJ and O'Leary TJ. Gastrointestinal stromal tumor
workshop. Human Pathology 32:578-582, 2001.
Gastrointestinal Stromal Tumor (GIST) has emerged in the past year as a prototypical neoplasm
that responds to therapy directed against a single target molecule - the KIT receptor tyrosine
kinase protein. Although GIST seldom responds to conventional chemotherapeutic agents, early
experience with the tyrosine kinase inhibitor, STI-571 (Gleevec,
Novartis, Basel, Switzerland) has been extremely encouraging. Early results have appeared in a
recent case report in the April 5, 2001 New England Journal of Medicine 1, and in early clinical
trials from the United States and Europe reported at the plenary session of the American Society
of Clinical Oncology in San Francisco, on May 14. STI571 is one of the earliest examples of a
"non-toxic" chemotherapeutic agent (an agent whose anti-cancer activity is NOT predicated on a
cytotoxic mechanism). STI-571 had already demonstrated clinical value in BCR-ABL-positive
leukemias. Early clinical results in GIST are so encouraging that oncologists may soon be
wrestling with the opportunity of referring every patient with malignant GIST into clinical trials
with STI-571. In order to ensure appropriate treatment, pathologists need to understand the
biology and treatment of this tumor and to have standard methods and criteria for providing
diagnosis (GIST or not GIST) and consistent prognostic classification (high risk of metastasis or
low risk of metastasis).
The rationale for including GIST in clinical trials with a tyrosine kinase inhibitor resulted from a
series of investigations demonstrating that gain-of-function KIT gene mutations are found in
many GISTs and that these mutations result in constitutive activation of the KIT protein - a
tyrosine kinase receptor - that stimulates proliferation of GIST tumor cells and may inhibit
apoptotic cell death. STI-571 is highly effective in vitro in reducing KIT tyrosine kinase activity.
GIST cells generally express the KIT protein (CD117); this enables the pathologist to utilize
immunohistochemical reactivity for the KIT protein as a marker for discriminating GIST from
other mesenchymal gastrointestinal neoplasms. Within the normal bowel wall, the interstitial
cells of Cajal (which are of the same cell lineage that undergoes neoplastic transformation to give
rise to GISTs) also express the KIT protein. KIT is thus a rare example of a proto-oncogene that
serves as a relatively specific marker for histogenesis and serves as a substrate for targeted drug
therapy; this is surely a model that will become more common in the future.
On April 2 and 3, a GIST Workshop was held at NIH.2 The workshop had several goals: 1) To
review the state of the art in GIST research, 2) to identify tissue resources (GIST tissue
repositories) and centers of specialized experience and expertise, 3) to provide general
recommendations to the pathology community regarding GIST diagnosis and prognosis, 4) to
identify important questions for future research and 5) to facilitate collaborative research
activities. To report the findings of this workshop in a timely manner, we agreed to immediately
prepare a short workshop summary that reviews some of the workshop highlights having greatest
importance to pathologists. Following this, we hope to prepare a supplemental issue of Human
Pathology that contains the scientific presentations from the workshop as well as a consensus
guideline (from the pathology subgroup) for reporting GISTs.
Day 1 of the workshop reviewed clinical features of GIST (Ronald P. DeMatteo, Memorial
Sloan-Kettering Cancer Center), histopathology of GIST (Henry Appelman, University of
Michigan), GIST cytogenetics (Jonathan Fletcher, Brigham and Women's Hospital), GIST
diagnosis (Christopher Fletcher, Brigham and Women's Hospital), GIST prognostic factors
(Leslie Sobin, AFIP), GIST biology (Markku Miettinin, AFIP) , and GIST treatment with STI-571 (George DeMetri, Dana-Farber Cancer Center).
Day 2 included talks on KIT mutations (Brian Rubin, U of Washington), the biological
mechanisms of KIT gene effects (Jack Longley, Columbia U), STI-571 as a modulator of KIT
tyrosine kinase activity (Mike Heinrich, Portland VA Medical Center), and genomic
characteristics of GIST (Wael-El-Rifai, U. of Virginia).
The following summary of workshop discussions is written particularly for pathologists who
have an immediate need for guidance on reporting GI soft tissue tumors. This workshop
summary should not be interpreted as a set of recommendations endorsed by the workshop
participants, and any assertions made should be considered tentative. A set of guidelines
(written by the GIST workshop pathology subgroup) for reporting on GIST tumors is planned for
a supplemental issue of Human Pathology and will be written with the understanding that new
information related to response-predictive markers based on data derived from current and future
clinical trials using STI-571 or other agents will impact on standards for GIST diagnosis. The
workshop presentations are also planned for inclusion in the supplemental issue.
Diagnosis of GIST
GISTs occur predominantly in middle-aged or older persons. Most tumors (70%) occur in the
stomach and about 20-30% occur in the small intestine. A small percentage of GISTs arise
elsewhere in the GI tract, omentum, or retroperitoneum. GISTs tend to be primarily intramural
tumors, usually involving submucosa and muscularis propria in continuity and often extending
into the mesentery or subserosa. Pure submucosal GISTs have been reported. GISTs are often
clinically silent until they reach a large size, bleed or rupture. Like other soft tissue sarcomas,
they only rarely metastasize to lymph nodes, and thus surgical lymphadenectomy is seldom
There are differences in the frequency, typical morphology and biology of GISTs depending on
their site of origin. GISTs can be single or multiple (as in Carney's triad or, rarely,
neurofibromatosis), spindled or epithelioid, or mixed cell type. Tumors of the small intestine,
including the duodenum, often contain extracellular collagen bundles, called skeinoid fibers.
Most spindle cell lesions of esophagus are leiomyomas and stain negatively for CD117. GISTs
rarely occur in the esophagus. In the experience of some workshop participants, large GISTs of
stomach that are attached to the greater curvature usually demonstrate benign behavior, despite
their size. Most duodenal GISTs occur in the second part of the duodenum where they push or
infiltrate into the pancreas. These tumors and other small bowel GISTs have a high propensity
for malignant behavior, compared to gastric GISTs, which are usually low-risk. Rectal tumors
are usually deep intramural tumors with homogeneous morphology. Local recurrences, probably
reflecting inadequate initial resections, are common in rectal GISTs.
Resected GIST tumors should undergo a gross examination similar to that of other GI tumors.
The exact anatomic location of the tumor should be noted. The size of the tumor should be
measured along its greatest axis, and the status of resection margins should be recorded. The
general experience of the workshop participants was that true local recurrences (excluding
mesenteric or peritoneal spread) occur only with unresectable tumors and with tumors in which
the resection margins are grossly positive; recurrence follows rapidly (several months) in these
cases. The main surgical objective is to achieve complete gross excision only. There is no
known benefit in obtaining wide margins, and this empiric finding provides a rationale for
routinely performing conservative wedge resections of gastric GISTs. Microscopic examination
of resection specimen margins is recommended.
For the diagnosis of spindle cell tumors arising in the GI tract, the general consensus of the
workshop pathologists was that CD117 immunostaining should be performed. A positive CD117
on a spindle cell lesion of GI tract confirms the diagnosis of GIST when morphologic and clinical
features of the tumor are consistent with GIST. CD117 expression in GIST is generally seen in
the majority of cells. Staining is typically cytoplasmic, strong and diffuse, but often shows dot-like accentuation. Focal positivity (in rare cells) or activity involving groups of cells comprising
under 10% of the total number of tumor cells should be interpreted cautiously. Scattered single
cells positive for CD117 are likely to be mast cells. Mast cells and insterstitial cells of Cajal are
the only gut cells that are typically CD117 positive. It is helpful to perform CD117 staining on
tumor sections that include mucosa, since gut mucosa will contain at least a few mast cells that
can serve as a positive control.
Most pathologists present believed that diagnosis of GIST requires immunohistochemical
confirmation by CD117. Two of the pathologists at the workshop did not believe that CD117
positivity is absolutely required in all cases of GIST, suggesting that an experienced pathologist
may, on rare occasion, rely on clinical and morphologic features for GIST diagnosis, even when
the CD117 immunostaining is deferred or the results of the staining are negative. Nevertheless,
all present believed that those who do not have extensive experience with GIST should perform
CD117 immunostaining before making this diagnosis. Pathologists who do not have access to
CD117 immunostaining, or who do not perform the assay frequently enough to feel comfortable
with the technical performance of the assay, should have the case examined by a pathologist
experienced with GISTs. It should be noted that patients entered into current STI-571 trials for
GIST must have tumors with CD117 positivity and must have central pathology review.
Gut spindle cell tumors that are negative for CD117 should generally be stained for desmin and
S-100, which are helpful in supporting the diagnosis of smooth muscle and neural tumors,
respectively. Both desmin and S-100 are generally negative in GISTs. Occasional desmin
positive cells within GIST tumors usually are non-neoplastic smooth muscle cells infiltrated by
GIST. Obviously-malignant atypical spindle cell tumors that demonstrate S-100 protein may
sometimes be metastatic melanomas. Confirmation with additional melanoma immunostains and
clinical correlation can be used to confirm this diagnosis. Among the lesions to be considered in
the differential diagnosis when all these immunostains are negative are desmoid fibromatoses,
solitary fibrous tumors, and inflammatory myofibroblastic tumors. As with other tumors,
immunohistochemical staining itself is adjunctive to, not a replacement for, careful morphologic
examination and clinical correlation.
Our understanding of GIST prognostic/predictive markers is certain to change in the next few
years. However, the group made a number of recommendations that can be used today by
general surgical pathologists. Once a diagnosis of GIST is made in a tumor with no known
metastases, the group recommended that in routine use two prognostic factors - tumor size and
mitotic index - provide the most useful histologic indication of the future biologic behavior of the
tumor. For this group of tumors, the mitotic index is measured by counting the number of
mitotic figures in 50 high powered fields and expressing the result as the number of mitoses per
50 high power fields). Workshop participants did not recommend grading GISTs, but suggested
that tumors be categorized by their relative risk of malignant behavior.
Although high cellularity is more prevalent in high-risk tumors and paucicellularity more
common in low-risk tumors, the consensus was that cellularity was not a valuable independent
prognostic indicator. Tumor necrosis, cystic change, nuclear atypia, infiltrative growth pattern,
tumor vascularity and degree of intensity of CD117 positivity were not considered to be proven
independent determinants of biologic behavior. Although an infiltrative growth pattern may be
found in low-risk and high-risk GISTs, it was noted that a pattern of infiltration into the lamina
propria, though rarely identified, was thought to be a useful criteria of malignancy. The criteria
for malignancy based on tumor size and mitotic count vary depending on tumor location. For
example, GIST arising in the small intestine demonstrate malignant behavior at lower levels of
mitotic activity than do GISTS arising in the stomach.
As yet, the group has not come to agreement on specific cut-points to distinguish GIST tumors as
low risk , intermediate risk (i.e. indeterminate) or high risk (i.e. malignant) based on size and
mitotic count. There was serious concern by some of the pathologists that no GIST can be
definitively labelled benign, particularly if it presents clinically as a mass. However, other
pathologists felt comfortable equating "low-risk" and "benign" tumors. The argument was raised
that small tumors (less than 2 cm) are occasionally found in stomachs and small intestine
resected for reasons unrelated to the GIST tumor, and that experience would suggest that such
patients never have GIST metastases (i.e. their GISTs are benign). Other pathologists were
skeptical, suggesting that the risk of malignant behavior in GISTs has no zero-threshold,
deprecating the term "benign", and favoring the term "low risk". There was group agreement for
categorizing GIST tumors at the extremes of biologic behavior. Small tumors (less than 2 cm)
with no mitotic figures are at very low risk for metastasis. Large tumors (> 5 cm) with a high
mitotic count (> 50 mitotic figures per 50 high power fields) commonly metastasize. However,
small tumors (i.e. tumors of any size) occurring in the intestine with >2 mitotic figures per 50
high power field were considered high risk (e.g. malignant) by some participating pathologists.
GIST Cytogenetics and Molecular Biology
The hallmark cytogenetic finding in both benign and malignant GISTs is monosomy of
chromosome 14; loss of chromosome 22 is also frequent. Nevertheless, karyotyping is not a
reliable method by which to establish the diagnosis of GIST. Not all GISTs can be successfully
karyotyped, and a "GIST-like" cytogenetic profile is found occasionally in malignant peripheral
nerve sheath tumors and in smooth muscle tumors. Various chromosomal deletions, including
deletions of 1p and 9p, coincide with malignant progression in GISTs. Interestingly,
chromosome 9p is the site of several tumor suppressor genes, including the cell cycle regulator
Using comparative genomic hybridization, DNA copy number changes can be observed in almost
every GIST with benign GISTs having fewer changes than malignant GISTs. Losses of DNA
copy numbers at 1p, 14q, and 22q are the most frequent changes seen in benign and malignant
GISTs. These changes are rarely seen in other mesenchymal tumors, such as schwannoma,
leiomyoma, and leiomyosarcoma. Gains and high-level amplifications at 5p, 8q, 17q, and 20q
were significantly more frequent in malignant and metastatic GISTs than in benign GISTs.
Losses in 9p were not seen in benign tumors, and they were more frequent in metastatic GISTs
than in malignant primary GISTS.
The best-characterized gene alterations associated with GIST are found in the KIT gene. Most
mutations consist of in-frame deletions and single nucleotide substitutions within exon 11, which
codes for the juxtamembrane domain, but alterations have also been described in exons 9 and 13,
which encode portions of the extracellular and kinase domains, respectively. Several studies
suggest that tumors demonstrating exon 11 deletion behave more aggressively than those
without. The clinical significance of the different types of mutation is not yet certain, however, as
no investigation yet published has had sufficient statistical power to detect differences among
mutation types. Studies at different centers show variation in both the frequency with which the
KIT gene is mutated, and in the significance of these mutations. The differences in these studies
may reflect differences in the types of tumors and patients which have been referred to these
institutions, but technical factors cannot be excluded. The clinical role of KIT mutation analysis
in characterization of GISTs thus remains incompletely defined.
It was suggested in the workshop that, in the future, GIST treatment decisions might require
input from both surgical pathology and molecular pathology. It was noted that in vitro studies
show a correlation between STI-571-responsiveness and the exon in which KIT mutation is
found. Hence, GIST response to STI-571 likely depends on which KIT protein domain is altered
by the underlying genomic mutation.
GIST Treatment and STI-571
Until recently, GIST treatment consisted of resection followed by surveillance for metastatic
disease. Chemotherapy and radiation have been ineffective. GIST metastases are most often
found in the liver or the peritoneum. Metastases can occur many years after initial resection, and
metastases do occur in well over half of the patients diagnosed with malignant or high risk
tumors at time of resection. Some patients have benefitted greatly from surgical resection of
localized GIST metastases. Most importantly, development of metastatic GIST no longer carries
the same implication of hopelessness as it did prior to the availability of STI-571.
Early clinical trial data with STI-571 are very encouraging, even for patients with metastatic
disease. Enrollment in an ongoing STI-571 clinical trial should probably be considered for all
GIST patients with metastases. Novartis has indicated that they have sufficient supply of STI-571 to include all GIST patients meeting this criterion into clinical trials.
The appropriate treatment for patients with the diagnosis of high risk (i.e. malignant) GIST but
without metastases is currently undergoing debate in the oncology community. The American
College of Surgeons Oncology Group (ACOSOG) has received approval from the NCI's Cancer
Therapy Evaluation Program (CTEP) for a clinical trial testing the benefit of adjuvant STI-571
therapy in patients after complete resection of high risk (>10 cm or ruptured) primary GIST in
the absence of metastases.
The side-effects of STI-571 treatment have been mostly mild dyspepsia and diarrhea, conditions
that pale in comparison with those of tumor metastases. Oncologists presented with the
availability of a "non-toxic" anti-GIST agent will surely be tempted to use the drug in patients
with a high risk of metastasis who have not yet developed widespread disease. Since
mechanisms by which GIST may escape responsiveness to STI-571 have yet to be completely
elucidated, it is prudent to refer such patients for treatment in controlled clinical trials that can
more clearly delineate the advantages and disadvantages of adjuvant treatment.
Long-term follow-up of STI-571-treated GIST patients does not yet exist, and no information
regarding the likelihood of an initially responsive tumor later developing STI-571 resistance is
available. The pharmacologic rationale for the use of STI-571 is clear since this compound
potently inhibits KIT activity in vitro. However, KIT signaling pathways are known to regulate
many aspects of cellular behavior, including proliferation, apoptosis, adhesion and
differentiation. At present, it is uncertain which pathways determine clinical responses to STI-571. Nonetheless, STI-571 can reduce GIST tumor size, suggesting that the response of GIST to
STI-571 is not merely cytostatic. Within 8 hours of STI-571 therapy, changes in tumors can be
seen on PET scan, suggesting rapid metabolic changes in tumor cells. Histologic examination of
STI-571-treated cases demonstrated that some of the tumors undergo myxomatous change,
leaving small pyknotic nuclei in an eosinophilic myxoid background.
STI-571 will likely be approved by the FDA within the next few months for the treatment of
chronic myelogenous leukemia. Early indicators of a role for STI-571 in the treatment of GIST
may lead oncologists and patients to ask for CD117 immunostaining in other neoplasms. At this
time, there are no data to support a role for STI-571 therapy in the treatment of non-GIST solid
tumors expressing strong CD117 positivity (e.g. adenoid cystic carcinoma, seminoma, small cell
lung carcinoma). Most participants emphasized that not all tumors that express KIT are likely to
respond to STI-571. Tumor response to STI-571 will probably depend on the mechanism of KIT
activation. It was noted that the detection of KIT expression (by immunohistochemical staining
with the CD117) does not indicate KIT activation.
Pathologists are called upon to provide oncologists with information that is critical to the correct
treatment of GIST patients. Currently, that information is limited to an accurate diagnosis and
some estimate of prognosis, based on clinical, anatomic and morphologic (size and mitotic
activity) features. It would seem prudent for pathologists to freeze fresh samples of any GIST
tumors they might accession. Uses for frozen GIST tissue may include ancillary research in
support of Clinical Trial protocols.
In the next few years, we will know much more about which patients with GIST benefit from
treatment and which patients do not. There will be methods to correlate molecular features with
predictors of tumor response. GIST is a prototype for solid tumors treated with a "non-toxic"
agent directed against a specific target molecule. Pathologists can expected a growing role in
characterizing tumors, such as GIST, whose therapy is directed at specific molecular targets
selected from tumor-classified molecular pathways.
Jules Berman, Ph.D., M.D.
Cancer Diagnosis Program
National Cancer Institute, NIH
Timothy J. O'Leary, Ph.D., M.D.
Department of Cellular Pathology and Genetics
Armed Forces Institute of Pathology
Washington, DC 20306
Pathology subgroup: Henry D. Appelman, M.D., Jules J. Berman, Ph.D., M.D., Christopher
Corless, M.D., Ph.D., Christopher D.M. Fletcher, M.D., Jonathan A. Fletcher, M.D., Fred
Gorstein, M.D., Jerzy Lasota, M.D., Ph.D., B. Jack Longley, M.D., Leslie Sobin, M.D., Sharon
W. Weiss, M.D., Markku Miettinen, M.D., Timothy J. O'Leary, M.D., Ph.D., Helen ReMotti,
M.D., Brian Rubin, M.D., Ph.D., Barry Shmookler, M.D.,
Clinical/Sciences Group: Roger L. Aamodt, Ph.D., Cem Akin, M.D., Ph.D., Charles Blanke,
M.D., Ronald P. DeMatteo, M.D. George Demetri, M.D., Wa'el El-Rifai, M.D., Ph.D., Jorge
Gomez, M.D., Ph.D., Michael Heinrich, M.D., John Hoffman, M.D. Burton Eisenberg, M.D.,
Tracy G. Lugo, Ph.D., Paul Meltzer, M.D., Ph.D., Mourad Majidi, Ph.D., Anthony J. Murgo,
M.D., Shreyaskumar Patel, M.D., Sheila E. Taube, Ph.D., Magdalena Thurin, Ph.D., Lisa
McShane, Ph.D., Sandra Silberman, M.D., Ph.D., Michael Radmacher, Ph.D.
We thank Henry Appelman, Wa'el El-Rifai, Ron DeMatteo, Christopher Corless, Christopher
Fletcher, Jonathan Fletcher, Brian Rubin, Helen Remotti and Leslie Sobin for their thoughtful
revisions of the Workshop Summary.
1. Joensuu H, Roberts PJ, Sarlomo-Rikala M, Andersson LC, Tervahartiala P, Tuveson D,
Silberman SL, Capdeville R, Dimitrijevic S, Druker B, Demetri GD. Brief Report: Effect of the
Tyrosine Kinase Inhibitor STI571 in a patient with a metastatic gastrointestinal stromal tumor.
N Engl J Med 344:1052-1056, 2001
2. GIST Workshop. Sponsored by the NIH Office of Rare Diseases, the Armed Forces Institute
of Pathology and the National Cancer Institute. Held in Bethesda, Maryland, April 2 and 3,