CT Features of Colorectal Schwannomas: Differentiation from Gastrointestinal Stromal Tumors

Purpose To find differential CT features of colorectal schwannomas from gastrointestinal stromal tumors (GISTs). Materials and Methods CT features of 13 pathologically proven colorectal schwannomas and 21 GISTs were retrospectively reviewed. The following CT items were analyzed: size, longitudinal and transverse location, shape, margin, homogeneity, necrosis, surface ulceration, calcification, degree of attenuation, the presence of enlarged lymph node (LN), and metastasis. Among the features, significant variables were evaluated using univariate statistical tests. The optimal cut-off point of tumor size was obtained by ROC analysis. Binary logistic regression analysis was used to find the most independent CT variables. Results Small size, non-rectum location, smooth margin, homogeneous high attenuation without necrosis, and the presence of enlarged LNs were found to be significant variables to differentiate schwannomas from GISTs (P<0.05). The optimized cut-off point for tumor size in distinguishing GISTs from schwannomas was 3.9 cm (AUC = 0.808, sensitivity = 66.7%, specificity = 92.3%, P<0.0001). Binary regression analysis revealed that only non-rectum location remained independent predictor for schwannomas differentiated from GISTs (odds ratio = 31.667, P = 0.001). Conclusion Colorectal schwannomas usually located in non-rectum and appear as small subepithelial nodules showing homogeneous high attenuation and smooth margin. Schwannomas exclusively accompany with enlarged LNs.


Introduction
Schwannomas are rare tumors that develop from Schwann cells. The colon is the least common site for intestinal involvement and the stomach is the most [1,2]. The exact prevalence of colonic schwannoma is unclear; however, it has been indirectly estimated that schwannomas are significantly less common than gastrointestinal stromal tumors (GISTs) by a ratio of 8-100:1 [3,4].
Colorectal schwannomas have received limited attention from clinicians as well as radiologists due to their rarity. However, imaging technique advances and increased incidences of routine health check-ups have increased the number of incidentally found colorectal mesenchymal tumors. Most colorectal mesenchymal tumors are GISTs; however, other benign mesenchymal tumors including schwannomas can be found. The exact preoperative differentiation between GIST and schwannoma has clinical significance because colorectal schwannomas are completely benign and totally different from the malignant potential of even small GISTs. However, to the best of our knowledge there have been no reports addressing this issue.
Choi et al. recently reported that gastric schwannomas show distinctive CT features differentiating them from gastric GISTs [5]. Gastric schwannomas appear as a homogeneous low attenuating mass without necrosis and frequently accompany with enlarged lymph nodes (LNs) on CT [5]. We hypothesize that colorectal schwannomas also have similar differential CT features from colorectal GISTs. Therefore, the purpose of this study is to find differential CT findings of colorectal schwannomas from GISTs.
were enrolled: 13 patients with colorectal schwannomas (7 men and 6 women; mean age, 61.8 years; range, 45-76 years) and 21 patients with GISTs (15 men and 6 women; mean age, 60.5 years; range, 37-76 years). All tumors were pathologically confirmed by either surgery (n = 32) or biopsy (n = 2). Type of surgery was recorded for patients who underwent surgery. In cases of GISTs, risk classification according to Miettinen and Lasota was performed [6]. This classification defines three risk groups for colorectal GIST: no risk GIST ( 2 cm and exhibiting 5 mitoses per 50 high-power fields [HPFs]), low risk GIST (2-5 cm and exhibiting 5 mitoses per 50 HPFs), and high risk GIST (either > 5 cm or exhibiting > 5 mitosis per 50 high-power fields). The institutional review board of our center as well as three other institutes approved this retrospective study. The requirement for informed consent was waived.
All patients received an iodinated contrast agent injection (Ultravist 370, Bayer Schering Pharma, Berlin, Germany) using an automatic power injector at a rate of 3-5 mL/sec and a dose of 1.5 mL/kg. All CT scans in 13 patients with schwannomas were obtained in single portal phase 60-70 seconds after contrast administration while CT scans in 4 patients with GISTs were obtained in both arterial and portal phases. Arterial phase images were scanned at 13-17 seconds after attenuation of the descending thoracic aorta reached 100 Hounsfield units (HU) using a bolus tracking technique.

Image Analysis
Morphological features and enhancement patterns of the tumors were independently assessed by two radiologists in consensus (K.J.H and K.S.H with 2 and 17 years of experience) on a picture archiving and communication system (PACS) workstation monitor (m-view, INFINITT, Seoul, Korea). The radiologists knew that all patients were diagnosed as having either colorectal schwannomas or GISTs, but were blinded to the exact diagnosis.
The following CT features were analyzed: size (cm), longitudinal and transverse location of the tumor, shape, margin, homogeneity, the presence of necrosis, surface ulceration, calcification, degree of enhancement on each CT phase, the presence of enlarged lymph node (LN), and metastasis. Longitudinal location was divided into cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum while transverse location was divided into endophytic, dumbbell, and exophytic. Tumor shape was evaluated as round or oval and tumor margin was defined as smooth or lobulated. Degrees of enhancement were compared to that of normal back muscle and defined as high-, iso-, or low-attenuation on the arterial and portal phases. LN was considered enlarged when its long diameter exceeded 8 mm.

Statistical Analysis
Statistical analyses were performed with SPSS 22.0 for Windows (SPSS Inc., Chicago, IL, USA). Chi-square or Fisher's exact test for categorical variables and the Mann-Whitney U test for continuous variables compared the prevalence of each CT feature between schwannomas and GISTs. A binary logistic regression analyses with a forward LR method were used to assess the most significant CT feature in differentiating schwannomas from GISTs. Receiver operating characteristic (ROC) analysis was performed to find the optimal cut-off value of tumor size. A value of P < 0.05 was considered statistically significant.

CT Features
Binary logistic regression analysis using a forward LR method showed that only the nonrectum location remained independent predictors for colorectal schwannomas differentiated from GISTs (odds ratio = 31.667; 95% confidence interval, 4.523-221.722; P = 0.001).

Discussion
Study results showed that several CT features, i.e., smaller size, non-rectum location, smooth margin, homogeneously high attenuation without necrosis, and the presence of LN enlargement were statistically significant in the differentiation of colorectal schwannomas from GISTs. In addition, binary logistic regression analysis revealed that non-rectum location was the sole CT discriminator of schwannomas from GISTs (odds ratio = 31.667). The results were similar to previous reports. According to several publications by Miettinen et al., only 5% (1/ 20) colorectal schwannomas were located in the rectum, while 78.2% (133/170) of colorectal  GISTs were located in the rectum, the third common site for GISTs following the stomach and small bowel [4,8,9]. The median size of the colorectal schwannomas was also reported to be 3 cm (range, 0.5-5 cm), while it was 6 cm (range, 0.5-15 cm) in colorectal GISTs [4,8,9]. Mean size (6.3 cm) of our GISTs was significantly larger than that (2.4 cm) of schwannomas (P = 0.001). When the cut-off value of tumor size was set at 3.9 cm in our study, AUC, sensitivity, and specificity were 0.808, 66.7% (14/21), and 92.3% (12/13), respectively. Therefore, we should pay attention to the possibility of GISTs because of their high malignant potential when we encounter a larger tumor (> 3.9 cm) located in the rectum.
Only one colorectal schwannoma showed heterogeneous enhancement with intra-tumoral necrosis in our study. The absence of necrosis in colorectal schwannomas is a well-known feature that has never been reported on imaging studies in the pathologic field [4]. This should On portal phase axial CT images, a 3.9 cm round mass (arrow, A) is noted at the descending colon. The mass shows homogeneously higher attenuation than back muscle with smooth margin. Note the two enlarged enhancing lymph nodes (arrowheads) at pericolic area. (C). A photograph of the cut surface of gross specimen obtained after left hemicolectomy shows a well-defined endophytic subepithelial mass in the colon. Tumor cells are diffuse and strongly S-100 protein positive on immunohistochemistry and confirmed as colonic schwannoma involving the submucosa and proper muscle (not shown).
doi:10.1371/journal.pone.0166377.g002 not be surprising considering the relatively slow speed of tumor growth in benign tumors which is typically on par with that of neovascularization. Conversely, in malignant tumors such as GISTs, the speed of tumor growth often outstrips neovascularization and leads to central necrosis. Several previous reports indicate 37.5%-81.8% of rectal GISTs show heterogeneous enhancement with internal hemorrhage or necrosis [10,11]. Indeed, half of GISTs (10/ 21, 47.6%) in our study had intra-tumoral necrosis. We believe that the results of this study suggest and confirm that the presence of necrosis is a differentiating CT feature of colorectal GISTs from schwannomas that have already been well-established in the field of pathology.
The frequency of LN enlargement is low; however, it was exclusively found in colorectal schwannomas (4/13, 30.8%). This result partly coincides with previous studies in which gastric schwannomas (75%-81.3%) more frequently accompany with enlarged LNs than gastric GISTs (5%-28.6%) [5,12]. The incidence of enlarged LNs in previous studies was higher than in our study as they used a less strict size criteria for lymph node enlargement, i.e., 5 mm in short axis diameter in the previous study [12] versus 8 mm in long axis in our study. The larger tumor size (mean size ± standard deviation, 6.3 ± 1.8 cm) in the previous study might be also responsible for a higher incidence of enlarged LNs in the schwannoma group [5]. Pathophysiology of accompanying LN enlargement has not been clearly investigated; however, we hypothesize that lymphoid cuffing around the tumor, which is a characteristic histologic feature of gastrointestinal schwannomas, might be related to this phenomena. Several pathologists insist that lymphoid cuffing might be the result of cytokines systemically secreted by tumor cells that induce the chemokinesis of lymphocytes [1,13]. Contrary to schwannoma, it is well-known that GISTs seldom accompany with LN metastasis [10,11]. Therefore, schwannoma should be first considered when we encounter a subepithelial mass with enlarged LNs in the colorectum. This observation and hypothesis should be further investigated.
Contrary to our expectations and the previous report, the degree of enhancement is higher in colorectal schwannomas than in GISTs [5]. Most schwannomas (11/13, 84.6%) showed high attenuation than adjacent back muscle, while two thirds of GISTs (15/21, 71.4%) showed isoattenuation. The degree of attenuation on portal phase was mostly iso or low attenuation in gastric lesions and was not significantly different between the two tumor groups [5]. We do  not exactly know why colorectal schwannomas show high attenuation; however, a smaller size and non-rectum location of schwannoma might be responsible for the high enhancement.
There are several limitations in our study. First, statistical power may have been comparatively weak due to the relatively small sample size and inherent limitations of the retrospective nature of our study. However, our study should be considered useful as it is the first report systematically analyzing the differential CT features of colorectal schwannomas from GISTs. Second, CT protocols were not standardized because the patients with colorectal schwannomas were collected from four different hospitals. However, we believe the limitations on the use of different CT protocols may be insignificant because most CT scans were performed using MDCT scanners and with reconstruction interval and slice thicknesses that were less than or equal to 5 mm. Third, we did not analyze the diagnostic performance of the radiologists to differentiate between two tumor groups on CT. Finally, the results might be overestimated because we did not include other subepithelial tumors such as leiomyoma or neuroendocrine tumor. Further studies recruiting various kinds of colorectal subepithelial neoplasms are strongly warranted.
In conclusion, colorectal schwannomas are usually located in non-rectum and appear as small subepithelial nodules showing homogeneous high attenuation without necrosis and a smooth margin. Schwannomas exclusively accompany with enlarged lymph nodes. Using these CT findings, colorectal schwannomas can be differentiated from GISTs with high diagnostic performance.