| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Abdominal Imaging |
1 Department of Radiology, Seoul National University Hospital, 28, Yongon-Dong,
Chongro-gu, Seoul 110-744, South Korea.
2 Department of Radiology, Asan Medical Center, University of Ulsan College of
Medicine, Seoul, South Korea.
Received December 29, 2003;
accepted after revision March 8, 2004.
Address correspondence to J. M. Lee
(leejm{at}radcom.snu.ac.kr).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. The medical records at our institution of 81 patients with a histologic diagnosis of gastrointestinal stromal tumor of the stomach were reviewed. Two radiologists retrospectively reviewed the CT findings by consensus with respect to lesion size, contour, tumor growth pattern, enhancing pattern, degree of enhancement, mesenteric fat infiltration, ulceration, calcification, lymphadenopathy, direct invasion to adjacent organ, and distant metastasis. Categoric variables were compared using the chi-square or Fisher's exact test. Multiple stepwise logistic regression analysis by means of forward selection was performed to determine significant predictors of high mitotic rate. Univariate analysis and multivariate analysis were also performed in a subgroup of 36 tumors with maximal diameter of 5 cm or smaller.
RESULTS. Size, presence of an ulcer, mesenteric fat infiltration, direct organ invasion, and metastasis were more frequently observed during univariate analysis in tumors with a high mitotic rate (p < 0.05). With stepwise logistic regression analysis, the size (odds ratio, 2.57; 95% CI; 1.42–4.67) was the only significant predictor of a high mitotic rate. In a subgroup of 36 tumors 5 cm or smaller, differentiation of benign from malignant tumors was not possible using CT.
CONCLUSION. Although presence of an ulcer, mesenteric fat infiltration, direct organ invasion, and metastasis were more frequently observed in tumors with a high mitotic rate, no CT feature, other than size, was found to have predictive value with respect to malignant gastrointestinal stromal tumors of the stomach.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The radiologic findings of gastrointestinal stromal tumors have been recently described in the radiology literature [1, 4–7], and these descriptions are similar to those of the leiomyomas and leiomyosarcomas previously described [8–12]. However, few articles have attempted to correlate CT findings with tumor grade. Chun et al. [11] reported that CT features of size, contour, enhancing pattern, mesenteric fat infiltration, ulceration, regional lymphadenopathy, and exophytic growth pattern could be used reliably to differentiate between malignant and benign tumors, and Tateishi et al. [5] reported that an extrinsic epicenter and an unclear boundary were the most significant predictors of high-grade gastrointestinal stromal tumors. Other authors have shown that no correlation between CT findings and malignant potential can be established unless an obvious local invasion or metastatic lesion is seen [1, 13].
Anatomic site, size (maximum diameter in centimeters), and mitotic rate were recently suggested as criteria for the prediction of gastrointestinal stromal tumor malignancy [14]. However, to our knowledge, the correlations between these new criteria and CT features have not been evaluated. The purpose of this study was to describe the CT findings of gastrointestinal stromal tumors of the stomach and to determine whether some CT features are useful for predicting malignancy in these tumors.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
All 81 tumors were positive by KIT immunohistochemical staining. In 40 patients, immunoreactivity to KIT was prospectively verified since its introduction. In five patients, immunoreactivity to KIT was seen in biopsy specimens of a metastatic lesion during the follow-up period. In 36 patients who were diagnosed before the introduction of KIT immunohistochemistry, immunohistochemical staining was retrospectively performed using paraffin blocks. Tumors were considered probably benign if they were 5 cm or smaller and contained no more than five mitoses per 50 high-power fields [14]. Tumors were considered malignant if they were larger than 10 cm or had more than five mitoses per 50 high-power fields. Tumors were considered to have uncertain or low malignant potential if they were larger than 5 but smaller than 10 cm and had no more than five mitoses per 50 high-power fields [14]. In three patients who had metastatic lesions in the liver and who underwent biopsy only, tumors were considered malignant despite an uncountable mitotic rate of small biopsy specimen.
CT Technique
In 71 patients, CT examinations were performed using a Somatom Plus-4
scanner (Siemens; n = 46), a HiSpeed Advantage scanner (GE
Healthcare; n = 17), or an MX8000 4-MDCT scanner (Marconi Medical
Systems; n = 8) at our institution. In seven patients, 20 mg of
hyoscine butylbromide (Boehringer Ingelheim Korea) was administered IV to
reduce gastric peristalsis, and 8 g of an oral effervescent agent (Tae Joon
Top Effervescent G Granule, Tae Joon Pharm) with minimal water were then
administered orally to obtain gastric distention immediately before
examination. Sixty-four patients drank 400–800 mL of pure tap water
instead of an antiperistaltic agent and oral effervescent agent immediately
before undergoing scanning. Each patient received 120 mL of a nonionic
contrast material ([iopromide] Ultravist 370, Schering Korea) through an
18-gauge angiographic catheter inserted in a forearm vein. The contrast
material was injected at a rate of 3 mL/sec using an automatic power injector.
Helical CT was performed using a single-detector scanner with the following
parameters: 5- to 7-mm collimation, 1:1 table pitch, and 5- to 7-mm
reconstruction intervals. If an MX8000 scanner was used, the parameters were
2.5-mm detector collimation, 20 mm/sec table speed, 3.2-mm slice thickness,
and 1.6-mm reconstruction interval. In 14 patients, biphasic helical CT scans
were obtained at 30 sec (arterial phase) and at 70 sec (portal venous phase)
after the initiation of the contrast material injection. In 57 patients,
monophasic helical CT scans were obtained with a 60- to 70-sec scan delay
(portal venous phase). An unenhanced image was obtained in only one
patient.
Ten patients were referred from other institutions with initial hard-copy CT scans acquired using a technique that differed from ours; for example, using iodinated orally administered contrast material (n = 3) and no water taken orally (n = 5).
Image Analysis
Two radiologists, blinded to the true histology, reviewed all radiologic
studies retrospectively and reached a consensus regarding the final
interpretation. We evaluated only the CT scans obtained during the portal
venous phase for tumor evaluation and all CT scans for metastasis evaluation.
The reviewers evaluated each CT scan with respect to lesion size, contour,
tumor growth pattern, enhancing pattern, degree of enhancement, mesenteric fat
infiltration, ulceration, calcification, regional lymphadenopathy, direct
invasion to adjacent organ, and distant metastasis.
The diameter of each tumor was measured in round lesions, and the greatest dimension, in ovoid or lobulated lesions. Lesion contours were classified as either round or lobulated. Growth patterns were classified as endoluminal, exophytic, or mixed. Endoluminal growth was defined as present if the tumor mass attached to the bowel wall was completely confined to the bowel lumen without bulging into the extraluminal space. Conversely, exophytic growth was defined as a mass confined to the extraluminal space without bulging into the bowel lumen, although extrinsic indentation could be observed. Mixed growth pattern was defined as a typical dumbbell appearance. Enhancement patterns (i.e., homogeneous or heterogeneous) were assessed subjectively. The degrees of enhancement of the soft-tissue elements of a tumor were judged against that of the muscle and liver: poor enhancement, identical to or less than that of the muscle; moderate enhancement, more than that of the muscle and less than that of the liver; and good enhancement, identical to or more than that of the liver. Mesenteric fat infiltration was considered present when the margin of the mass was indistinct because of streaky increased densities. Ulceration was considered present when a focal tissue defect filled with air or fluid or when contrast material was found on the endoluminal surface of the mass lesion. The presence of calcification was required to document high-attenuation foci within the mass lesion. Regional lymphadenopathy was considered present if soft-tissue nodular lesions larger than 1 cm in the short-axis diameter were observed.
Statistical Analysis
Tumor grade depends on both tumor size and mitotic rate
[14]. Because tumor size can
be accurately measured on a CT scan, we evaluated the correlations between
each CT feature and mitotic rate. A low mitotic rate was defined as no more
than five mitoses per 50 high-power fields, and a high mitotic rate, as more
than five mitoses per 50 high-power fields
[14]. In three patients who
underwent liver biopsy, the mitotic rate was considered high, although the
mitotic rates were uncountable because of a small biopsy specimen.
Categoric variables were compared using the chi-square or Fisher's exact test. The level of significance used for inclusion in the model in multiple logistic regression was less than 0.20, and odds ratios and 95% confidence intervals (CIs) were calculated. Multiple stepwise logistic regression analysis by means of forward selection was used to identify significant predictors of a high mitotic rate. In a subgroup of 36 tumors 5 cm or smaller, univariate analysis and multivariate analysis were performed in the same manner.
In addition, the relationships between tumor size and growth pattern or enhancement pattern were assessed by means of linear-by-linear association. Statistical analyses were performed with computer software (SPSS, version 10.0; Statistical Package for the Social Sciences). A p value of less than 0.05 was considered to indicate a statistically significant difference.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The CT features of all 81 patients are summarized in Table 1, and those of a subgroup of 36 tumors 5 cm or smaller are summarized in Table 2. The sizes of 81 tumors ranged from 1 to 23 cm (mean, 8.1 ± 5.3 cm). Twenty-six cases (32%) were classified, according to the gold standard, as benign (Fig. 1), eight (10%) as having low malignant potential, and 47 (58%) as malignant (Figs. 2 and 3). The sizes of benign tumors ranged from 1 to 5 cm (mean, 3.7 ± 0.9 cm); those of tumors with low malignant potential, from 5.5 to 10 cm (mean, 7.4 ± 1.7 cm); and those of malignant tumors, from 1.5 to 23 cm (mean, 10.8 ± 5.6 cm). The locations of the 81 tumors were the fundus (n = 25; 30.9%), the body (n = 44; 54.3%), and the antrum (n = 12; 14.8%). Heterogeneous tumor enhancement was present in 51 tumors (63%), and central fluid attenuation was present in 40 (49.4%). Calcification (n = 4) was seen only in malignant lesions.
|
|
|
|
|
Nine patients (11%) had metastatic lesions on CT scans: Of this group, six patients were diagnosed with liver metastases, one of whom also had peritoneal seeding, and an additional three patients were diagnosed with peritoneal seeding. Two of nine patients with metastasis had gastric tumors smaller than 5 cm. On CT scans, the attenuation of metastatic lesions within the liver was lower than that of the normal surrounding parenchyma in five patients (Fig. 4), and in the sixth patient, the metastatic nodule was isoattenuating on the portal venous phase and highly attenuating on the arterial phase (Fig. 5A, 5B). Liver metastases were visualized as a peripheral soft-tissue component with central fluid attenuation on portal venous phase CT scans in three patients.
|
|
|
Size and presence of an ulcer, mesenteric fat infiltration, direct organ invasion, and metastasis were more frequently observed with univariate analysis in tumors with a high mitotic rate (p < 0.05). Stepwise logistic regression analysis showed that only size (odds ratio, 2.57; 95% CI, 1.42–4.67) was a significant predictor of a high mitotic rate. In a subgroup of 36 tumors 5 cm or smaller, univariate analysis revealed that all variables were above the p = 0.20 level; therefore, a multivariate logistic regression model was not used.
The relationship between tumor size and growth pattern is shown in Table 3. The larger the size of a tumor, the more often it grows exophytically (p < 0.00). The relationship between tumor size and the enhancement pattern is shown in Table 4. Thirty homogeneous tumors were found to range from 1 to 11 cm (mean, 4.4 ± 2.3 cm); and 51 heterogeneous tumors, from 2.5 to 23 cm (mean, 10.3 ± 5.4 cm). The greater the size of a tumor, the more often the tumor showed a heterogeneous enhancement pattern (p < 0.00).
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In our series, univariate analysis revealed that size, presence of an
ulcer, mesenteric fat infiltration, direct organ invasion, and metastasis were
more frequently observed in tumors with a high mitotic rate. In general, these
observations agreed with previous reports
[5,
11]. Multivariate analysis was
not performed in one report
[11], and Tateishi et al.
[5], using a multiple stepwise
logistic regression model, reported that an extrinsic epicenter and an unclear
boundary were the most significant predictors of a high-grade tumor. In our
series, however, multiple logistic regression analysis by means of forward
selection showed that tumor size is the only significant predictor of a high
mitotic rate. Tumors of intestinal origin tend to have more aggressive
behavior and thus a worse prognosis than do tumors originating in the stomach
[15]. The criteria for
malignancy differed according to whether tumors were gastric or intestinal
[14]. Although gastric tumors
were considered malignant if they were greater than 10 cm or had more than
five mitoses per 50 high-power fields, intestinal tumors were considered
malignant if they were greater than 5 cm or had more than five mitoses per 50
high-power fields. Previous reports included both gastric and intestinal
tumors [5,
11], but our series includes
only gastric tumors, which may provide an explanation for discrepancies
concerning predictors. Moreover, in a subgroup of 36 small tumors (
5 cm),
no CT features were found to usefully discriminate malignant tumors from
benign ones, and 10 (28%) of these 36 small tumors were malignant. These
results should make us hesitate about predicting malignancy in clinical
practice.
In our series, 36 (44%) of 81 tumors were small ( 5 cm) gastric tumors.
Small gastric tumors constitute about 10% of all tumors in the West
[1,
6], although they account for
more than half in the East [5,
11]. Gastric cancer is the
second most common cancer in the world but is the most common cancer in Korea
and Japan [16]. Barium studies
and gastric endoscopy are commonly used to detect early gastric cancer during
routine check-ups and in patients complaining of abdominal discomfort. This
situation explains why small tumors are more common in the East than in the
West.
Although some studies in the literature show a male predominance [4, 6], others, like the present series, show no gender predilection [1, 5, 11, 17]. The most common clinical manifestation of a symptomatic gastrointestinal stromal tumor is gastrointestinal bleeding from mucosal ulceration [1]. Patients may present with hematemesis, melena, or signs and symptoms of anemia caused by occult bleeding. In another report from the East, the most common symptoms were pain, followed by masses and gastrointestinal bleeding [5]. In our series, 41 patients went to the hospital because of abdominal pain or discomfort, although 15 patients presented with gastrointestinal bleeding or anemia. Gastrointestinal stromal tumors may be incidental findings in the examination of individuals with abdominal pain, but when they do cause symptoms, pain and discomfort are the most common complaint in the East.
The liver is the most common metastatic site at both presentation and disease relapse [4]. On CT at the portal venous phase, metastases within the liver were usually of lower attenuation than the normal surrounding liver [4]. A metastatic mass in the liver is likely to be heterogeneous and peripherally enhancing like a primary tumor [4]. Low attenuation in the center of metastatic lesions often indicates central necrosis of a solid mass, and a peripherally enhancing portion represents a viable solid tumor. In our series, five patients showed this appearance and one additional patient had a single metastatic nodule that was isoattenuating on portal venous phase CT and hyperattenuating on arterial phase CT. We probably missed some hypervascular liver metastasis, because biphasic CT scans were obtained in only 14 patients. The availability of the tyrosine kinase inhibitor has markedly altered the clinical approach to gastrointestinal stromal tumors because it has been proven effective in metastatic gastrointestinal stromal tumors [18]. Therefore, the detection of liver metastases has become more important than ever. Hepatic arterial phase CT images can provide added value in the evaluation of hypervascular liver metastasis in a carcinoid tumor [19]. Further studies are needed to determine whether biphasic helical CT can detect more liver metastases than monophasic helical CT in patients with gastrointestinal stromal tumors.
Pathologists usually experience difficulty in differentiating benign from malignant tumors. Moreover, mitotically inactive tumors can metastasize, which indicates that a low mitotic count does not necessarily rule out malignant behavior [2]. In our series, one patient with liver metastasis had a 4-cm gastric tumor with a low mitotic rate that was classified as a benign tumor according to the criteria of size and mitotic rate.
Our study has a number of limitations. This was a retrospective review of cases collected over a number of years at a tertiary hospital. CT scans included those obtained at other institutions using different imaging techniques that could influence some of the CT findings used to evaluate these tumors. An unenhanced image was obtained in only one patient. Therefore, we may have missed some calcification or liver lesions that may have been isodense on portal venous phase images. In three patients who had a liver biopsy, the mitotic rate was indeterminable. In statistical analysis, we used mitotic rate instead of malignancy as a dependent variable, because tumor size can be accurately measured on a CT scan, despite the fact that tumor grade depends on a combination of tumor size and mitotic rate.
In conclusion, we found that CT features, other than tumor size, cannot
helpfully predict gastrointestinal stromal tumor malignancy and that the
malignant potential of small gastrointestinal stromal tumors ( 5 cm)
cannot be determined on CT. It would be interesting to find out whether
biphasic helical CT can detect more liver metastasis than monophasic helical
CT.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S ULUSAN, Z KOC, and F KAYASELCUK Gastrointestinal stromal tumours: CT findings Br. J. Radiol., August 1, 2008; 81(968): 618 - 623. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. H. Kim, H. W. Eun, D. E. Goo, C. S. Shim, and Y. H. Auh Imaging of Various Gastric Lesions with 2D MPR and CT Gastrography Performed with Multidetector CT. RadioGraphics, July 1, 2006; 26(4): 1101 - 1116. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
