Radiologist Performance in Differentiating Polypoid Early From Advanced Gastric Cancer Using Specific CT Criteria: Emphasis on Dimpling Sign

Transcription

1 Gastrointestinal Imaging Original Research Lee et al. CT of Gastric Cancer Gastrointestinal Imaging Original Research Eun Sun Lee 1 Se Hyung Kim 1,2 Jae Young Lee 1,2 Soo Jin Kim 1 Min A Kim 3 Jeong Min Lee 1,2 Joon Koo Han 1,2 Byung Ihn Choi 1,2 Lee ES, Kim SH, Lee JY, et al. Keywords: CT, neoplasm, performance test, stomach DOI: /AJR Received December 27, 2008; accepted after revision June 4, Supported by a grant from the Seoul National University Hospital Research Fund No Department of Radiology, Seoul National University Hospital, Seoul National University College of Medicine, 101 Daehangno, Jongno-gu, Seoul, , Korea. Address correspondence to S. H. Kim (shkim@radcom. snu.ac.kr). 2 Institute of Radiation Medicine, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Korea. 3 Department of Pathology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Korea. AJR 2009; 193: X/09/ American Roentgen Ray Society Radiologist Performance in Differentiating Polypoid Early From Advanced Gastric Cancer Using Specific CT Criteria: Emphasis on Dimpling Sign OBJECTIVE. The purpose of this study was to retrospectively determine whether there are specific CT features that can be used to differentiate polypoid early from advanced gastric cancer and to assess the performance of radiologists using specific CT findings for differentiation. MATERIALS AND METHODS. A review of medical records yielded the cases of 46 patients, 27 with polypoid early gastric cancer and 19 with polypoid advanced gastric cancer, whose CT scans were available for review. Two radiologists retrospectively reviewed the CT images for the presence and depth of dimpling at the tumor base, the presence of vessel invagination at the dimpling site, thickening of the low-attenuating outer layer, perigastric infiltration, and transmural full-thickness enhancement of the lesion. Individual CT findings relevant as predictors were determined with univariate and multivariate analyses. Individual review of CT scans subsequently was performed by two other radiologists, who were blinded to tumor stage but aware of the results of univariate and multivariate analyses. Individual performance was evaluated with receiver operating characteristic analysis. RESULTS. The presence of severe dimpling greater than 3.5 mm at the base of the tumor (odds ratio, 31.3) had the highest odds ratio for differentiating early from advanced gastric cancer, followed by vessel invagination (odds ratio, 12.3), the presence of dimpling (odds ratio, 9.8), perigastric infiltration (odds ratio, 5.2), and transmural full-thickness enhancement (odds ratio, 4.8). Multivariate analysis showed that the presence of dimpling greater than 3.5 mm was the only independent variable that differentiated polypoid advanced gastric cancer from polypoid early gastric cancer (p = 0.001). Subsequent differentiation of advanced from early gastric cancer with the described CT findings was very good, yielding areas under the receiver operating characteristic analysis curves of and for the two observers. CONCLUSION. Greater than 3.5 mm dimpling and other ancillary CT findings are helpful in differentiating polypoid advanced gastric cancer from polypoid early gastric cancer and contribute to good individual accuracy for differentiation. G astric cancer is one of the most common causes of cancer-related death in East Asia, South America, and Eastern Europe [1 3]. In the United States, the incidence of gastric cancer has decreased since the 1950s, but the incidence of adenocarcinoma of the gastric cardia is increasing [4]. Compared with patients with advanced gastric cancer, who have a dismal prognosis, reported 5-year survival rates being only 3 21%, patients with early gastric cancer have 5-year survival rates of % after curative resection [5 9]. Early gastric cancer with a negligible risk of lymph node metastasis can be managed with minimally invasive methods, such as laparoscopic gastric surgery and therapeutic endoscopic resection [3, 10 13]. These treatments allow the patient to avoid the risks of morbidity and mortality caused by extensive surgery and are associated with long-term benefit in terms of quality of life [14]. Because the management and prognosis of gastric cancer are closely related to tumor depth and nodal involvement, preoperative differentiation of early from advanced gastric cancer has become important. CT is the primary technique for staging of gastric cancer. Even though technical advances such as MDCT scanners, 3D reconstruction software, and negative oral contrast agents and hypotonic agents for CT, have been made, CT frequently does not depict two or three layers of the gastric wall, which is critical for differentiation of early from advanced gastric cancer [15, 16]. However, 3D surface-rendered and virtual endoscopic 1546 AJR:193, December 2009

2 CT of Gastric Cancer images, which mimic upper gastrointestinal radiographic series and conventional endoscopy, respectively, can give a hint about differentiation, particularly in depressed gastric tumors. At endoscopy, a dam formation that represents fusion of more than three folds around the depressed lesion usually indicates the presence of advanced rather than early gastric cancer. In contrast, clubbing, abrupt cutting, a rattail appearance, and fusion of the folds favor the diagnosis of early rather than advanced gastric cancer. To our knowledge, however, there have been no points on which to differentiate polypoid early from advanced gastric cancer at CT or endoscopy. In the field of pathology, it is well known that the presence of dimpling or puckering of the outer wall of the gastrointestinal tract in a gross specimen usually indicates involvement of the muscle layer of the bowel wall. Kim et al. [17] suggested that in the diagnosis of polypoid cancer of the gallbladder, the dimpling sign on MDCT images may be a helpful finding in differentiating T2 from T1 lesions. We therefore hypothesized that several CT features, including the dimpling sign, may be helpful for differentiating polypoid early from advanced gastric cancer. The purpose of our study was twofold. The primary aim was to determine whether there are specific CT features that can be used to differentiate polypoid early gastric cancer from polypoid advanced gastric cancer. The second aim was to assess the performance of radiologists in the differentiation when they used specific CT findings. Materials and Methods Patient Population The institutional review board of our hospital approved this retrospective study, and the patients informed consent was not required. We reviewed the electronic medical records (radiologic, surgical, and pathologic reports) from 2000 to 2007 at our hospital and identified the cases of 101 patients with the confirmed or suspected diagnosis of polypoid gastric lesions. The conditions of 13 of the patients were finally diagnosed as benign mucosal tumor (n = 7), submucosal tumor (n = 4), and benign gastric ulcer (n = 2), and these patients were excluded from the study. Among the other 88 cases of gastric cancer, 40 were finally confirmed at pathologic examination as nonpolypoid gastric cancer, and those patients were excluded. Two other patients were excluded because of the poor quality of CT images. Forty-six patients (32 men, 14 women; mean age, 65.3 years; range, years) with surgically confirmed polypoid gastric cancer and adequate abdominal CT images available for review made up the study sample. Twentyseven patients had early gastric cancer, and 19 had advanced gastric cancer. Among the four morphologic types of advanced gastric cancer according to Borrmann s classification [18], Borrmann type 1 (fungating, polypoid) and type 2 (ulcerofungating) lesions were classified as polypoid advanced gastric cancer. According to the rules for macroscopic classification of early gastric cancer of the Japan Gastroenterological Endoscopy Society [19], type 1 (protruded) early gastric cancer was designated polypoid early gastric cancer. More specifically, lesions with a thickness more than twice that of normal mucosa were defined as type 1 early gastric cancer. This definition is a contradistinction from type 2a early gastric cancer, which has a thickness up to twice that of normal mucosa. Morphologically, polypoid early and advanced gastric cancers consisted of lesions protruding into the gastric lumen with a sharply defined margin with or without surface ulceration. TABLE 1: CT Parameters CT Scanner CT Acquisition Immediately before CT, all patients were asked to drink more than 1,000 ml of water (n = 33) or two packets of an effervescent agent (n = 13) to attain adequate gastric distention. Fifteen minutes before CT, all patients received 10 mg of butylscopolamine bromide (Buscopan, Boehringer Ingelheim) through an antecubital vein to minimize bowel peristalsis and to facilitate hypotonia. Most CT examinations (35/46, 76%) were performed with one of the following four MDCT scanners: LightSpeed Ultra, GE Healthcare (n = 14); Sensation 16, Siemens Healthcare (n = 9); Brilliance 64, Philips Healthcare (n = 6); and Mx8000, Philips Healthcare (n = 6). The other 11 CT examinations were performed with one of two single-detector CT scanners: Somatom Plus 4, Siemens Healthcare (n = 9) and HiSpeed Advantage, GE Healthcare (n = 2). The scanning parameters are shown in Table 1. For clinical interpretation, MDCT images were reconstructed with a slice thickness of 3 mm and reconstruction interval of 3 mm. For 3D reconstruction, MDCT images were reconstructed with a slice thickness of 1 mm and reconstruction interval of 1 mm. In addition to axial MDCT images, coronal and sagittal multiplanar reformation (MPR) images and 3D surface-shaded volume-rendered images were generated by expert 3D technicians. For single-detector CT, a slice thickness of 5 mm and reconstruction interval of 5 mm were used. CT acquisition was performed in two positions: supine (n = 7), prone (n = 10), or left posterior oblique (n = 29) and right decubitus. The details regarding CT are described in the report of a previous study [20]. We chose the left posterior oblique position to obtain better distention of the lower half of the stomach [20]. The position for the initial CT acquisition (supine, prone, left posterior oblique) was chosen on the basis of the type of oral contrast agent used and the location of the tumor at endoscopy. CT images were obtained seconds after injection of 120 ml of iopromide (Ultravist 370, Bayer Schering Pharma) at a rate of 3 ml/s through an automatic power injector. After acquisition in the left posterior oblique position, patients were asked to move to the right decubitus position, and after the default delay, acquisition was repeated. Image Analysis Two radiologists aware of the diagnosis of gastric cancer but blinded to the final histopathologic stage of gastric cancer reviewed the CT images in consensus. All CT images were reviewed on 21- inch (53 cm) monitors at a PACS workstation. The following features were analyzed: presence and depth (millimeters) of dimpling at the base of the tumor, presence of vessel invagination at the dimpling site, thickening of the low-attenuating outer layer, perigastric infiltration, and transmural fullthickness enhancement of the lesion. Dimpling was defined as indentation of the outer margin of the gastric wall on CT images [17] (Fig. 1). If dimpling was present at the base of the tumor, the radiologists were asked to measure its depth with an electronic caliper on the PACS monitor. Measurement was made from the expected and imaginary outer Detector Configuration Pitch Rotation Time (s) Kilovoltage (kvp) Tube Current Time Product (mas) Brilliance mm Sensation mm LightSpeed Ultra mm Mx mm Somatom Plus 4 5 mm HiSpeed 5 mm Note Brilliance 64 and Mx800 manufactured by Philips Healthcare, Sensation 16 and Somatom Plus 4 manufactured by Siemens Healthcare, and LightSpeed Ultra and HiSpeed manufactured by GE Healthcare. AJR:193, December

3 Lee et al. A Fig. 1 Dimpling sign in two patients. A, 62-year-old woman with polypoid advanced gastric cancer. CT scan shows dimpling (arrow), which is defined as indentation of outer margin of gastric wall. B, 70-year-old man with polypoid advanced gastric cancer. CT scan shows depth of dimpling measured with electronic caliper on PACS monitor. Measurement is made from expected and imaginary outer margin (dotted line) of gastric wall to deepest real outer margin of tumor. In this case, depth of dimpling is 8.32 mm (solid line). margin of the gastric wall to the deepest real outer margin (Fig. 1). To differentiate dimpling and contractions due to peristaltic movement of the stomach, symmetric wall thickening and mild luminal narrowing with a preserved layered pattern in an area independent of gastric cancer were defined as contractions. If perigastric vessels invaginated into the dimpling base, the lesion was considered to have vessel invagination (Fig. 2). The presence of spiculation in perigastric fat around the tumor was considered to indicate perigastric infiltration. If the tumor exhibited transmural enhancement and did not have the normal two or three layers, the lesion was considered to have transmural full-thickness enhancement. For image analysis, MPR images and axial images were used for 27 of the 35 patients who underwent MDCT. Any discrepancy was solved by consensus. In a subsequent independent reading session, two other abdominal radiologists (11 and 4 years of experience in abdominal radiology) blinded to the diagnosis but informed of the results of univariate and multivariate analyses derived from the consensus reading session reviewed the CT images. They were asked to interpret CT images in terms of the features analyzed during the previous consensus reading sessions, that is, presence and depth of dimpling at the base of the tumor, presence of vessel invagination at the dimpling site, thickening of the lowattenuating outer layer, perigastric infiltration, and transmural full-thickness enhancement of the lesion. They also were asked to determine whether the lesion was early or advanced gastric cancer using a 5-point confidence scoring system: 1, definitely early gastric cancer; 2, probably early gastric cancer; 3, possibly advanced gastric cancer; 4, probably advanced gastric cancer; 5, definitely advanced gastric cancer. A score 3 or greater was chosen as the cutoff point for defining advanced gastric cancer. In analysis of the causes of misinterpretation, discrepancies in differentiation were solved by the two reviewers and a third radiologist in consensus. Pathologic Analysis All resected specimens were retrospectively reviewed by one specialized pathologist with 10 years of experience in gastrointestinal pathology. She was asked to report the T category. T category was based on the pathologic TNM (ptnm) system developed by the American Joint Committee on Cancer and the International Union Against Cancer [21] (Table 2). More specifically, for early gastric cancer (T1), the depth of tumor invasion was subcategorized into mucosal cancer and submucosal cancer. Submucosal tumor invasion was further subdivided subjectively into sm1, sm2, and TABLE 2: Pathologic T Categorization, American Joint Committee on Cancer, 2002 T Category pt1 pt2 pt3 pt4 Fig year-old woman with polypoid advanced gastric cancer. Coronal reformatted CT image shows perigastric vessels (arrow) invaginating into dimpling base. Lesion therefore is considered to exhibit vessel invagination. Description Tumor confined to mucosa or submucosal layer Tumor extending into muscle proper or subserosa Tumor extending through serosa without involving contiguous structures Tumor extending to contiguous structures B 1548 AJR:193, December 2009

4 CT of Gastric Cancer sm3, representing invasion to the upper, middle, and lower thirds of the submucosa. The presence and degree (mild versus severe) of submucosal fibrosis also were determined. Causes of misclassification by the two independent reviewers were analyzed on the basis of the pathologic results. Statistical Analysis Univariate statistical differences between the CT features of polypoid early and those of advanced gastric cancers were analyzed with the chi-square test or Fisher s exact test for ordinal data. Sensitivity and specificity analyses were performed for a range of depth of dimpling (in millimeters) in early and advanced gastric cancer to generate a receiver operating characteristic (ROC) curve and to determine the optimal cutoff of depth of dimpling for differentiating polypoid early from advanced gastric cancer. The optimal cutoff value was defined as the point on the ROC curve closest to the 0% falsepositive and 100% true-positive marks. Multivariate stepwise logistic regression analysis was used to determine the factors most predictive of a differential diagnosis between polypoid early and advanced gastric cancer. Individual performance of the two radiologists in differentiating polypoid early from advanced gastric cancer was evaluated with ROC analysis, and interobserver agreement was determined with kappa or weighted kappa statistics. We considered a value greater than 0.81 to represent almost perfect agreement and values of and to represent substantial or moderate agreement. Values less than 0.40 were considered to represent fair agreement [22]. For early gastric cancer, pathologic N category was recorded. Univariate statistical differences in CT features between early gastric cancer with and early gastric cancer without lymph node metastasis were analyzed with the chi-square test or Fisher s exact test for ordinal data and Mann-Whitney U test for continuous data. Analyses were performed with statistical software (SPSS version 12.0, SPSS). A value of p < 0.05 was considered statistically significant. Results Pathologic Examination Eighteen of the 19 cases of advanced gastric cancer were categorized T2 and the other case, T3. Nine of the 18 T2 lesions invaded the muscle proper, and the other nine, the subserosa. The mean size of advanced gastric tumors was 6.3 cm (range, cm). Among 27 cases of early gastric cancer, 13 were mucosal and 14 were submucosal. Six patients had sm1 tumors; five, sm2 tumors; and three, sm3 tumors. Among 27 patients with early gastric cancer, three had mild and three had severe submucosal fibrosis. Microscopic examination revealed no submucosal fibrosis in the other 21 patients with early gastric cancer. The mean size of early gastric tumors was 3.7 cm (range, cm). The presence of lymph node metastasis was confirmed in six of the 27 patients (22%) with early gastric cancer. The lesions were confirmed category N1 according to the classification of the American Joint Committee on Cancer. Two patients had one involved node; one patient, two involved nodes; one patient, four involved nodes; and two patients, five involved nodes. CT Features Differentiating Polypoid Early and Advanced Gastric Cancer The mean depths of gastric wall dimpling were 1.1 ± 2.4 (SD) mm for early gastric cancer and 5.2 ± 3.4 mm for advanced gastric cancer. The difference was statistically significant (p < 0.001). A scatterplot of the depth of dimpling is shown in Figure 3. To differentiate polypoid early from advanced gastric cancer, ROC analysis was performed at different cutoff levels for depth of dimpling. The area under the ROC curve (A z ) was largest when the cutoff value for depth of dimpling was set at 3.5 mm. With the optimal cutoff level of depth of dimpling determined to be 3.5 mm, sensitivity of 79% (15/19) and specificity of 96% (26/27) were achieved. Baseline differences in CT variables between polypoid early and advanced gastric cancer were significant for severe dimpling greater than 3.5 mm (p < 0.001), vessel invagination at the dimpling site (p = 0.001), dimpling (p = 0.002), perigastric infiltration (p = 0.025), and transmural full-thickness enhancement of the lesion (p = 0.029) (Table 3). The presence of greater than 3.5 mm dimpling at the base of the tumor (odds ratio, 31.3) achieved the highest odds ratio for differentiation, followed by vessel invagination (odds ratio, 12.3), presence of dimpling (odds ratio, 9.8), perigastric infiltration (odds ratio, 5.2), and transmural full-thickness enhancement (odds ratio, 4.8). The presence of thickening of the low-attenuating outer layer was not significantly different between the two groups. Multivariate logistic regression analysis for the five significant variables showed that the presence of greater than 3.5- TABLE 3: Results of Univariate Analysis of CT Findings for Differentiating Polypoid Early and Advanced Gastric Cancers Variable Early Gastric Cancer (n = 27) Advanced Gastric Cancer (n = 19) p Odds Ratio > 3.5 mm severe dimpling 1 (3.7) 15 (78.9) < Vessel invagination 2 (7.4) 10 (52.6) Dimpling 9 (33.3) 15 (78.9) Perigastric infiltration 1 (3.7) 5 (26.3) Transmural full-thickness wall enhancement Thickening of the low-attenuation outer layer 7 (25.9) 11 (57.9) (33.3) 7 (36.8) Note Values are number of patients with percentage in parentheses. Values of p calculated with chi-square or Fisher s exact test. Depth of Dimpling (mm) Early Gastric Cancer Advanced Gastric Cancer Fig. 3 Scatterplot shows depth of dimpling in early and advanced gastric cancers. AJR:193, December

5 Lee et al. mm dimpling at the base of the tumor was the only independent variable that differentiated polypoid early from advanced gastric cancer (p = 0.001). Examples are shown in Figures 4 and 5. The mean size of six tumors of early gastric cancer with lymph node metastasis (5.2 cm) was significantly larger than that of tumors without lymph node metastasis (3.3 cm) (p = 0.038). All six cases of early gastric cancer with lymph node metastasis were submucosal, but the 21 cases of early gastric cancer without lymph node metastasis included 13 cases of mucosal cancer and eight of submucosal cancer (p = 0.016). There was no difference between early gastric cancer with and early gastric cancer without lymph node metastasis according to presence of dimpling, vessel invagination, depth of dimpling, and pathologic subtype (p > 0.05). Observer Performance Tests The findings of interobserver agreement on recognizing the distinguishing features of polypoid early and advanced gastric cancer and differentiating the tumor types are presented in Table 4. In all of the CT findings thought to be significantly different between the two tumor types, there was substantial to moderate agreement between radiologists (κ = ). There was moderate agreement in differentiating the two tumor types (weighted κ = 0.528). The individual radiologists performance results are shown in Table 5. The individual accuracy of radiologists in differentiation of polypoid early from advanced gastric cancer was evaluated according to A z values, which Fig year-old man with correctly interpreted findings of early gastric cancer. Axial CT scan shows polypoid and homogeneously enhancing lesion (arrows) with transmural full-thickness enhancement in posterior wall of lower part of gastric body. No dimpling, vessel invagination, or perigastric infiltration was present. Two radiologists interpreted lesion as early gastric cancer. Pathologic examination confirmed type 1 early gastric cancer confined to mucosal layer (T1a) (not shown). A were calculated with nonparametric analysis. Differentiation was accurate for both radiologists (A z = and 0.827). To calculate the sensitivity and specificity of CT in differentiation, we considered a radiologist s degree TABLE 4: Interobserver Agreement on CT Findings and Differentiation CT Finding > 3.5 mm severe dimpling (substantial) Vessel invagination Dimpling Perigastric infiltration Transmural full-thickness wall enhancement Differentiation of tumor types Fig year-old man with correctly interpreted advanced gastric cancer. A, Axial CT scan shows polypoid and homogeneously enhancing mass with transmural full-thickness enhancement of posterior wall of lower part of gastric body. Severe dimpling (5.01 mm) is present in outer wall of stomach. Vessel invagination and perigastric infiltration are not evident. Two radiologists interpreted lesion as advanced gastric cancer. B, Photograph of microscopic section obtained after subtotal gastrectomy shows polypoid mass (arrows) with dimpling (dotted line) at base of tumor. Pathologic examination confirmed Borrmann type 2 advanced gastric cancer with subserosal invasion (T2b). κ (moderate) (substantial) (moderate) (moderate) (moderate) B 1550 AJR:193, December 2009

6 CT of Gastric Cancer TABLE 5: Area Under the Receiver Operating Characteristic Curve, Sensitivity, Specificity, and Positive and Negative Predictive Values of CT in Differentiation of Polypoid Early and Advanced Gastric Cancer TABLE 6: Causes of Overstaging and Understaging (n = 8) Final Diagnosis Depth of Invasion Tumor Size (cm) Overstaged cases Parameter Radiologist 1 Radiologist 2 After Consensus Area under the receiver (95% CI, ) (95% CI, ) operating characteristic curve Sensitivity (%) 74 (14/19) 74 (14/19) 79 (15/19) Specificity (%) 78 (21/27) 82 (22/27) 82 (22/27) Positive predictive value (%) 70 (14/20) 74 (14/19) 75 (15/20) Negative predictive value (%) 81 (21/26) 82 (22/27) 85 (22/26) Note Data in parentheses are raw numbers. of confidence greater than 3 to be positive for advanced gastric cancer. Use of this threshold resulted in sensitivity of 74% (14/19) and 74% (14/19) and specificity of 78% (21/27) and 82% (22/27) for radiologists 1 and 2. After consensus review, sensitivity and specificity were 79% (15/19) and 82% (22/27). The radiologists overstaged five cases of early gastric cancer as advanced gastric cancer and understaged four cases of advanced gastric cancer as early gastric cancer. The details on incorrectly differentiated lesions are presented in Table 6. At retrospective review of pathologic and CT findings, the causes of overstaging of five cases of early gastric cancer as advanced gastric cancer were submucosal fibrosis in three patients (Fig. 6), gravity due to nondependent location in one patient (Fig. 7), and associated intussusception in one patient. The causes of understaging of four cases of advanced gastric cancer as early gastric cancer were the tumor pressing on but not infiltrating the muscle proper layer in one case (Fig. 8), which should be considered the same as invasion and therefore T2 according to the sixth edition of American Joint Committee on Cancer classification; unusual location of the pyloric ring in one case; small tumor size (2 cm) in one case; and unknown cause in one case. Among the 11 patients who underwent CT with a single-detector scanner, seven had early gastric cancer and four had advanced gastric cancer. The distribution regarding T category was not significantly different (p = 1.00). In all but one of the 11 patients, disease was correctly classified. The other patient had overstaged early gastric cancer with 3.4-mm dimpling. Discussion Because the management and prognosis of gastric cancer are closely related to tumor depth and nodal involvement, the importance of preoperative differentiation between early and advanced gastric cancer has increased. Many investigators have found through clinicopathologic correlation that the prognoses of early and advanced gastric cancers after treatment differ significantly [5 9]. Our data suggest that CT can be used for accurate differentiation of polypoid gastric cancer into early and advanced. Numerous reports have been published on early and advanced gastric cancer, but to the best of our knowledge, our study is the only evaluation of the performance of radiologists in differentiating polypoid gastric cancers using specific CT criteria. We found that several CT findings, including the presence of severe dimpling greater than 3.5 mm, vessel invagination, dimpling, perigastric infiltration, and transmural full-thickness enhancement, showed statistically significant differences between polypoid early and advanced gastric cancer. Polypoid advanced gastric cancer usually appears as a polypoid mass with greater than 3.5 mm dimpling and vessel invagination at the base of the tumor, perigastric infiltration, and transmural full-thickness enhancement. Among these characteristics, greater than 3.5 mm dimpling was identified as the only independent variable differentiating advanced from early gastric cancer, having a high odds ratio (31.329) and specificity (96%) and acceptable sensitivity (79%). More specifically, 15 of 19 polypoid advanced gastric cancers (79%) exhibited severe dimpling greater than 3.5 mm at the base of the tumor, whereas only one of 27 polypoid early gastric cancers (4%) had severe dimpling. In addition, inter- Depth of Dimpling (mm) Submucosal Fibrosis Location Causes of Incorrect Staging Early gastric cancer Submucosa Severe Nondependent Submucosal fibrosis Early gastric cancer Mucosa Nondependent Gravity Early gastric cancer Submucosa Severe Dependent Submucosal fibrosis Early gastric cancer Submucosa Mild Dependent Submucosal fibrosis Early gastric cancer Submucosa Nondependent Intussusception Understaged cases Advanced gastric cancer Muscle proper Dependent Pushing border to muscle proper layer Advanced gastric cancer Subserosa Severe Pyloric ring Pyloric ring Advanced gastric cancer Subserosa Mild Dependent Small size Advanced gastric cancer Subserosa Nondependent Unknown Note Dash [ ] indicates no evidence of submucosal fibrosis. AJR:193, December

7 Lee et al. observer agreement was best (κ= 0.692) for determining the presence of severe dimpling. Therefore, we can say that polypoid gastric cancer should not be categorized as early gastric cancer when the tumor has greater than 3.5 mm dimpling at its base. However, the risk of understaging outweighs that of overstaging in that minimally invasive treatment can be used only on tumors confined to the mucosa and submucosa, that is, early gastric cancer, that have low risk of nodal and distant metastasis. For this reason, moderate sensitivity in the diagnosis of advanced gastric cancer may not be acceptable for patients with gastric cancer except those whose poor condition makes aggressive surgery unacceptable. A C Fig year-old woman with early gastric cancer overstaged as advanced gastric cancer owing to severe submucosal fibrosis. A, Axial CT scan shows homogeneously enhancing polypoid mass (asterisk) with 3.18-mm-deep dimpling (arrow, inset) at base of tumor. B, CT scan 1 cm below A vessel invagination (arrow). Both radiologists classified lesion as advanced gastric cancer. C, Photograph of gross specimen from subtotal gastrectomy shows polypoid mass (asterisk) in posterior wall of gastric antrum. D, Photomicrograph confirms presence of type 1 early gastric cancer with submucosal invasion (T1b, sm3) (arrow). Retrospective and careful review of microscopic slide revealed severe submucosal and intramuscular fibrosis (asterisks). In our study, four of 19 cases of advanced gastric cancer (21.1%) were understaged at consensus reading by two radiologists. Not all of the tumors exhibited dimpling, vessel invagination, and perigastric infiltration. Only one tumor had a thickened low-attenuating outer wall, and two tumors had transmural full-thickness wall enhancement. Even though univariate analysis revealed that it was one of the significant CT features in differentiating advanced from early gastric cancer, transmural full-thickness wall enhancement was neither sensitive nor specific for advanced gastric cancer. Indeed, 25.9% (7/27) of early gastric lesions exhibited the finding whereas 42.1% (8/19) of advanced gastric lesions did not. Even though several causes of understaging were proposed at retrospective careful review of both pathologic and CT examinations, we believe that the risk of understaging was not easily overcome with only careful review of CT images. The use of endoscopic ultrasound may be a potential solution for this purpose. Despite its operator-dependent nature, endoscopic ultrasound has been considered the most accurate technique for determining the T category of gastrointestinal cancer. Therefore, limitation of the use of endoscopic ultrasound to patients with polypoid gastric tumors without severe dimpling can be helpful for minimizing the risk of understaging. Further studies should be conducted to verify B D 1552 AJR:193, December 2009

8 CT of Gastric Cancer the role and performance of endoscopic ultrasound in this regard. In the field of pathology, it is well known that the presence of dimpling or puckering in the outer wall of the gastrointestinal tract of gross specimens usually indicates involvement of the muscle layer of the bowel wall. Kim et al. [17] proved that dimpling and vessel or fat invagination at the base of gallbladder tumors are differential points between A Fig year old woman with 5.5-cm lesion of early gastric cancer misclassified as advanced gastric cancer because of effect of gravity. A, Axial CT scan shows pedunculated polypoid mass (asterisk) in anterior wall of lower part of gastric body. Arrow indicates 3.42-mm-deep dimpling (arrow, inset) at base of tumor. Both radiologists classified lesion as advanced gastric cancer. B, Photomicrograph confirms presence of type 1 early gastric cancer (asterisks) confined to mucosa (T1a). Arrows indicate intact muscularis mucosa. Even on retrospective review, pathologist was not able to find evidence of submucosal fibrosis, which can induce indentation of gastric wall. Cause of misclassification therefore was thought to be effect of gravity due to somewhat large size of mass and location in anterior and nondependent wall of stomach. T2 and T1 lesions. The staging of gallbladder cancer is somewhat different from that of gastric cancer. In gallbladder cancer, category T1b, which does not include dimpling, indicates tumor invasion into the muscle layer below the mucosa and lamina propria. This finding would correspond to T2a gastric cancer. In other words, severe dimpling with vessel or fat invagination in gallbladder cancer usually occurs when the tumor penetrates A all the way through the muscle layer ( T2 of gallbladder cancer), which corresponds to T2b gastric cancer. Such a difference may be due to the difference in thicknesses of the wall and muscle layers of the two organs. Because the muscle layer of the gallbladder is thinner than that of the stomach, only transmural infiltration ( T2) by the tumor in gallbladder cancer may induce dimpling at the base of the tumor. Fig year-old woman with advanced gastric cancer understaged as early gastric cancer. A, CT scan shows 4-cm well-enhancing polypoid mass (asterisk) with transmural full-thickness enhancement on lesser curvature side of lower part of gastric body. No dimpling, vessel invagination, or perigastric infiltration was found at CT. Both radiologists interpreted lesion as early gastric cancer. B, Photomicrograph confirms presence of Borrmann type 1 advanced gastric cancer with invasion of muscle proper (PM) (T2a). Tumor (arrows) did not infiltrate but pushed against muscle proper layer. Subtle invasion to muscle layer is thought to be responsible for absence of dimpling in tumor. B B AJR:193, December

9 Lee et al. Although it is well known that some gastric tumors, particularly those of early gastric cancer, frequently are not visualized, even at MDCT, the radiologists in our study did not have difficulty detecting all lesions, including 27 cases of early gastric cancer. We believe this success may be due to the better detection rate of CT for the protruded type of early gastric cancer (type 1) than the slightly elevated (type 2a), flat (type 2b), depressed (type 2c), and excavated (type 3) types of tumors [23]. Furthermore, the high spatial resolution and isotropic MPR capability of MDCT may be responsible for the excellent detection performance with CT. In our study, use of highresolution MPR images may have helped the radiologists analyze differential CT features, including dimpling, particularly for tumors on the horizontally oriented portion of the gastric wall, such as the greater curvature and gastric angle, in which lesions cannot be exactly evaluated because of partial volume-averaging effects [24, 25]. In the cases of five patients in our study, early gastric cancer was overstaged as advanced gastric cancer at CT. Three of these cases (60%) were submucosal with varying degrees of submucosal fibrosis (Fig. 6). Because, owing to its strong desmoplastic reaction, submucosal fibrosis is a well-known cause of overstaging of early as advanced gastric cancer regardless of morphologic tumor type, this result is not unexpected. One early gastric tumor confined to the mucosal layer also was misclassified as advanced gastric cancer (Fig. 7). The cause of misclassification in this case was the effect of gravity. The misclassified mass was somewhat large (5.5 cm) and located on the anterior, nondependent wall of the stomach. The other case of early gastric cancer was overstaged owing to the presence of gastrogastric intussusception caused by a polypoid mass. In the other two cases, we found no pathologic evidence of submucosal fibrosis that might have induced indentation of the gastric wall. Also in our study, four cases of advanced gastric cancer were misclassified as early gastric cancer. None of the tumors exhibited dimpling at the base, even though two exhibited submucosal fibrosis. At retrospective pathologic examination, we found that one understaged advanced gastric tumor pushed against the muscle proper layer but did not infiltrate into it (Fig. 8). Such subtle invasion to the muscle proper layer might have been responsible for the absence of dimpling in this case. Another understaged advanced gastric tumor was considered misclassified owing to its location. This tumor was located in the pyloric ring, in which radiologists have difficulty staging tumors owing to frequent collapse and strong peristaltic motion. Another understaged advanced gastric tumor was considered misclassified owing to its small size (2 cm). Even though it invaded the subserosal layer, because of its small size, we believed that tumor did not exhibit CT findings proven to be related to advanced gastric cancer. In the fourth case, we were not able to find the exact cause of understaging, even at careful pathologic and CT review. Our study was limited to a degree by its retrospective nature, which was unavoidable owing to the relative rarity of polypoid gastric cancer. In addition, potential selection bias might have been present toward patients with less-advanced disease because we recruited all patients with surgically confirmed lesions. However, we believe that we tried to minimize most biases by including relatively large numbers of tumors. We believe inclusion of individual performance evaluation by independent reviews made our findings and conclusions more applicable to routine clinical practice. Second, MPR images were not obtained for all patients because CT was performed with a single-detector scanner for some patients. In such cases, a partial volume-averaging effect can occur in determination of CT features, especially for tumors on the horizontally oriented portion of the gastric wall. Third, we obtained CT images in the portal venous and delayed phases but not in the mucosal (late arterial) phase. According to a previous report [26], contrast-enhanced CT images obtained during the late arterial phase (40 seconds after contrast injection) are useful for detection of gastric cancer. In our study, however, because of their unique morphologic characteristics, all malignant polypoid tumors were well visualized in the portal venous phase despite the absence of mucosal phase CT images. Finally, the presence of dimpling or vessel invagination was not proven at pathologic examination. However, because of the varying degrees of stretching and manipulation that occur during fixation with formalin, radiologic pathologic correlation of these findings was not possible. We conclude that with use of specific CT findings, including greater than 3.5-mm severe dimpling, radiologists can differentiate polypoid advanced gastric cancer from polypoid early gastric cancer with good individual accuracy and interobserver agreement. References 1. Fuchs CS, Mayer RJ. Gastric carcinoma. N Engl J Med 1995; 333: Forman D, Goodman KJ. The epidemiology of stomach cancer: correlating the past with the present socioeconomic influences in early life can influence mortality in adult life. BMJ 2000; 320: Lee IJ, Kim SH, Lee JM, et al. Multidetector row computed tomographic gastrography findings after endoscopic submucosal dissection for early gastric cancer: emphasis on time evolution and factors for predicting residual tumor. J Comput Assist Tomogr 2009; 33: Devesa SS, Blot WJ, Fraumeni JF Jr. Changing patterns in the incidence of esophageal and gastric carcinoma in the United States. Cancer 1998; 83: Nakamura K, Ueyama T, Yao T, et al. Pathology and prognosis of gastric carcinoma: findings in 10,000 patients who underwent primary gastrectomy. Cancer 1992; 70: Siewert JR, Bottcher K, Stein HJ, Roder JD. Relevant prognostic factors in gastric cancer: ten-year results of the German Gastric Cancer Study. Ann Surg 1998; 228: Noh SH, Yoo CH, Kim YI, Kim CB, Min JS, Lee KS. Result after a gastrectomy of 2,603 patients with gastric cancer: analysis of survival rate and prognostic factor. J Korean Surg Soc 1998; 55: Msika S, Benhamiche AM, Jouve JL, Rat P, Faivre J. Prognostic factors after curative resection of gastric cancer: a population-based study. Eur J Cancer 2000; 36: Kim JP, Lee JH, Kim SJ, Yu HJ, Yang HK. Clinicopathologic characteristics and prognostic factors in 10,783 patients with gastric cancer. Gastric Cancer 1998; 1: Ludwig K, Klautke G, Bernhard J, et al. Minimally invasive and local treatment for mucosal early gastric cancer. Surg Endosc 2005; 19: Oda I, Saito D, Tada M, et al. A multicenter retrospective study of endoscopic resection for early gastric cancer. Gastric Cancer 2006; 9: Uedo N, Iishi H, Tatsuta M, et al. Longterm outcomes after endoscopic mucosal resection for early gastric cancer. Gastric Cancer 2006; 9: Lee MW, Kim SH, Kim YJ, et al. Gastrointestinal stromal tumor of the stomach: preliminary results of preoperative evaluation with CT gastrography. Abdom Imaging 2008; 33: Sasako M. Risk factors for surgical treatment in the Dutch Gastric Cancer Trial. Br J Surg 1997; 84: Bhandari S, Shim CS, Kim JH, et al. Usefulness of three-dimensional, multidetector row CT (virtual 1554 AJR:193, December 2009

10 CT of Gastric Cancer gastroscopy and multiplanar reconstruction) in the evaluation of gastric cancer: a comparison with conventional endoscopy, EUS, and histopathology. Gastrointest Endosc 2004; 59: Clark CJ, Thirlby RC, Picozzi V Jr, Schembre DB, Cummings FP, Lin E. Current problems in surgery: gastric cancer. Curr Probl Surg 2006; 43: Kim SJ, Lee JM, Lee JY, et al. Accuracy of preoperative T-staging of gallbladder carcinoma using MDCT. AJR 2008; 190: Borrmann R. Geschwulste des Magens und Duodenums. In: Henke F, Lubarsch O, eds. Handbuch der Spezieller Pathologischen Anatomie und Histologie. Berlin: Springer-Verlag, 1926:825 FOR YOUR INFORMATION 19. Japanese Research Society for Gastric Cancer. The general rules for the gastric cancer study and surgery and pathology. Jpn J Surg 1981; 11: Kim SH, Lee JM, Han JK, et al. Effect of adjusted positioning on gastric distention and fluid distribution during CT gastrography. AJR 2005; 185: Greene FL, Page DL, Fleming ID, et al. AJCC manual of staging of cancer, 6th ed. New York: Springer-Verlag, 2002: Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: Kim AY, Kim HJ, Ha HK. Gastric cancer by multidetector row CT: preoperative staging. Abdom Imaging 2005; 30: Hori S, Tsuda K, Murayama S, Matsushita M, Yukawa K, Kozuka T. CT of gastric carcinoma: preliminary results with a new scanning technique. RadioGraphics 1992; 12: Cho JS, Kim JK, Rho SM, Lee HY, Jeong HY, Lee CS. Preoperative assessment of gastric carcinoma: value of two-phase dynamic CT with mechanical IV injection of contrast material. AJR 1994; 163: Chen CY, Hsu JS, Wu DC, et al. Gastric cancer: preoperative local staging with 3D multi-detector row CT correlation with surgical and histopathologic results. Radiology 2007; 242: Realize Your Potential with AJR Integrative Imaging Each quarterly issue of AJR Integrative Imaging offers you a series of peer-reviewed learning opportunities, many with CME credit, which will challenge and enhance your diagnostic and treatment skills. Whether it is the printed version or the online version, AJR Integrative Imaging is an important tool for your professional development. For more information or the latest issue of AJR Integrative Imaging, visit AJR:193, December