Upper gastrointestinal subepithelial tumors (SETs) consist of a group of mesenchymal lesions and include a variety of tumor types, such as leiomyomas, gastrointestinal stromal tumors (GISTs), granular cell tumors, schwannomas, leiomyosarcomas, and aberrant pancreas. SETs located in the esophagogastric junction (EGJ) are extremely rare, leiomyomas and GISTs occupy predominately among these tumors. Resection of tumors in such location remains challenging because of the specific anatomy location. For GISTs with high potential malignancy, curative surgical resection with an adequate margin is required. Partial or wedge resection, even proximal gastrectomy or total gastrectomy, are often applied for GIST located in the EGJ due to the potential risk of stricture and leakage. However, leiomyomas, as benign tumors, can be treated with conservative therapy or enucleation or tumor-everting resection instead of extended resection.[3,4] Hence, to distinguish these 2 tumors in the preoperative stage is important for surgeons to select the optimal therapeutic strategies.
The accurate differentiation between leiomyoma and GIST is quite difficult before surgery, and the final diagnosis is depended on postoperative pathological examination. Endoscopic biopsy or endoscopic ultrasound (EUS)-guided fine needle aspiration biopsy may provide useful information preoperatively, but they often provide inadequate tissue yield and increase the risk of tumor rupture and seeding. Recently, several studies were reported by using CT characteristics to differentiate gastric GISTs from other benign SETs.[6–8] Several characteristic CT features that could differentiate leiomyomas from GISTs in the gastric cardia were identified. However, the reliability of these results was limited by the extremely small-sized study population and lack of important clinical parameters in previous studies, further investigation is warranted.
Therefore, in this retrospective analysis, we analyzed the clinical and CT features of leiomyomas and GISTs in the EGJ in relatively larger population, with the aim of better distinguishing these 2 tumors preoperatively.
2.1 Patient selection
This was a single-center, retrospective study. This study was approved by the Institutional Review Board of the West China Hospital of Sichuan University, and informed consents were obtained from each patient in this cohort. All procedures for this study were carried out in accordance with the approved guidelines and regulations. Subjects for this study were selected from a total of 1943 patients with esophago-gastric GISTs or leiomyomas presenting to West China Hospital of Sichuan University between April 2010 and July 2018 (Fig. 1). The inclusion criteria were as follows:
- 1. pathologic diagnosis of leiomyomas or GISTs;
- 2. located in the esophagogastric junction;
- 3. underwent surgical resection (endoscopic, laparoscopic, or open surgery) at our hospital.
Exclusion criteria were:
- 1. tiny (<1 cm) incidental GISTs or leiomyomas found in other operations;
- 2. recurrent tumors receiving reoperation.
First, 49 patients with leiomyomas and 87 patients with GISTs were identified in our study for clinical analysis. Second, analysis of CT images between leiomyoma and GIST was conducted in patients with available preoperative contrast-enhanced CT images. Fifteen patients with leiomyomas and 14 patients with GISTs were excluded because there were no available CT images. Finally, 34 patients with leiomyomas and 73 patients with GISTs were included in our study for CT image analysis.
2.2 Acquisition of CT images
Contrast-enhanced CT scans were performed using a SOMATOM Definition Flash unit (Siemens Healthineers, Erlangen, Germany). To achieve gastric distension, 1000 ml tap water was ingested for each patient before the CT examination. Intravenous nonionic contrast agent (2 ml/kg; iopromide, Ultravist, Schering) were injected in an antecubital vein at a flow rate of 3 ml/second. Scan parameters were as follows: 128 × 0.6 mm collimation, gantry rotation time of 0.28 second, 120 kV and 210 mAs. Contrast-enhanced images were available at 25 to 35 seconds (hepatic arterial phase), 60 to 70 seconds (portal venous phase), and 3 to 5 minutes (delayed phase) after initiation of IV injection of contrast material. All images were reconstructed with slice thickness thinner than 5 mm for statistical analysis.
2.3 Image analysis
Two gastrointestinal radiologists (with 5 and 8 years of experience, respectively) retrospectively and independently reviewed the CT findings, and discrepancies were resolved according to their consensus. The 2 radiologists knew that patients had either leiomyomas or GISTs in the EGJ, but who were blinded to the clinical data and the histopathologic results. The images were reviewed on a vendor-specific postprocessing software (Syngo.Via, Siemens, Germany).
The following CT finding were analyzed: tumor size, tumor shape (round or ovoid, irregular), surface (regular or irregular), growth pattern, enhancement pattern, enhancement degree of the tumor, presence of surface ulcer, presence of calcification, presence of internal low attenuation areas (necrosis, gas, hemorrhage or cystic degeneration). Growth patterns were classified as endoluminal, exophytic, or mixed. Enhancement patterns included homogeneous or heterogeneous enhancement. The enhancement degrees included low, intermediate, and high, which were defined in comparison to the enhancement of back muscles. A CT attenuation value less than 20 HU within the tumor was considered as the presence of internal low attenuation. When measuring the degree of tumor enhancement, CT attenuation values of the tumors were measured in Hounsfield units using 16 to 18 mm2 circular regions of interests (ROIs). Three ROIs were placed on the most strongly enhanced portion of the tumors in each hepatic arterial, portal venous, and delayed phase. The mean tumor attenuation was measured as the average values of 3 ROI values.
2.4 Pathological evaluation
The final diagnoses of all enrolled patients were determined according to the specific immunohistochemical (IHC) analysis in conjunction with molecular genetic analysis. For IHC analysis, positive expression of c-KIT (CD117) and/or DOG-1 is consistent with GISTs, and positive expression of smooth muscle actin (SMA) and desmin indicates leiomyoma. For c-KIT/DOG-1 double negative tumors that cannot exclude GISTs morphologically, molecular genetic analyses that harbor a KIT or PDGFRA mutation can also be diagnosed as GISTs.
2.5 Statistical analysis
All statistical analyses were performed using IBM SPSS Statistics for windows (V21.0, IBM corp, NY, USA). Continuous variables were summarized with mean ± standard deviation (SD), dichotomous variables were expressed as frequency and percentages. To analyze difference of CT and clinicopathological findings between GISTs and leiomyomas, the student t test or variance (ANOVA) were performed for continuous variables, whereas the χ2 test or Fisher exact test were performed for categorical variables. For significant clinical and CT features, the optimal cutoff values for these date for differentiating leiomyomas from GISTs were determined by the receiver operating characteristic (ROC) curves (highest sum of the sensitivity and the specificity). Each patient was given 1 point for each significant feature, and a GIST-risk score was calculated as the sum of significant features. The optimal GIST-risk scoring model was established when the sum of the specificity and the sensitivity was highest, and the optimal cut-off point of the score was calculated for differentiating leiomyomas from GISTs. A P value of <.05 was considered statistically significant, and a 95% confidence interval (CI) was reported for each variable.
3.1 Clinical characteristics in total participants
A total of 49 and 87 patients with leiomyomas and GISTs were identified in this study. The demographic and clinical features of patients are listed in Table 1. Patients with leiomyomas were younger than those with GISTs (46.82 years vs 55.44 years, P < .001). The optimal cutoff value of 46.5 years for the age yielded a sensitivity of 80.5%, a specificity of 57.1%, and an area under the receiver operating characteristic curve (AUC) of 0.681. The mean size of tumor was larger in GISTs group than leiomyomas group significantly (5.19 ± 2.9 cm vs 4.05 ± 2.45 cm, P = .024). The finding of surface ulcer (P = .001) during operation was more frequently observed in GISTs. In addition, open laparotomy was the predominant surgical strategy for leiomyomas and GISTs, although endoscopic resection was more frequently applied to leiomyomas (P = .011). No statistically significant differences were found in gender, symptom, resection range, growth pattern, and tumor shape between the patients with leiomyomas and those with GISTs (P > .05).
3.2 CT features between GISTs and leiomyomas
The CT findings between leiomyomas and GISTs are summarized in Table 2. The tumor long diameter (LD) was significantly larger in GISTs compared to leiomyomas (4.90 ± 2.40 cm vs 3.69 ± 1.48 cm, P = .002) with an optimal cutoff value of 4.5 cm (sensitivity: 53.4%, specificity: 76.5%) and an AUC of 0.681. The tumor short diameter (SD) was also significantly larger in GISTs compared to leiomyomas (3.77 ± 1.82 cm versus 2.78 ± 1.24 cm, P = .005). The LD/SD ratio, tumor shape, growth pattern, and surface were not significantly different between patients with leiomyomas and those with GISTs (P > .05). There was a statistically significant higher rate of surface ulcer (53.4% vs 26.5%, P = .009), heterogenous enhancement (56.2% vs 26.5%, P = .004), and high degree enhancement (45.2% vs 2.9%, P < .001) among GISTs compared to leiomyomas. The rate of intralesional low attenuation was also higher among GIST (47.9% vs 20.16%, P = .007). However, the presence of calcification (20.6% vs 5.5%, P = .035) was more common in leiomyomas. Mean CT attention value was significantly lower in leiomyomas than that in GISTs (67.57 ± 11.05 HU vs 75.60 ± 20.05 HU, P = .031). The cutoff value of for 69.2 HU for the CT attention value yielded a sensitivity of 61.6%, a specificity of 64.7% and an AUC of 0.632.
3.3 Exploration of a novel scoring system for identifying GISTs
Considering the clinical and CT features involved in our study, there were 8 significant characteristics that could differentiate GISTs from leiomyomas: age >46.5 years, tumor long diameter >4.5 cm, heterogeneous enhancement, high degree enhancement, mean CT attenuation > 69.2 HU, presences of intralesional low attenuation and surface ulcer, absences of calcification. Each patient in this study was given 1 point for each significant feature, the total score of each patient was determined based on how many significant features he had. The optimal number of significant features differentiating leiomyomas and GISTs was explored using ROC analysis. We found that when all 8 significant features were used in combination, a highest AUC of 0.844 was achieved with an optimal cutoff value of 3.5 (sensitivity: 76.7%, specificity: 76.5%, Fig. 2). Representative CT images of leiomyomas and GIST are illustrated in Figs. 3 and 4.
With the widespread use of endoscopic examination, the detection rate of upper gastrointestinal SETs has obviously increased. According to the literature, a majority of SETs are leiomyomas and GISTs confirmed by pathological examination. Leiomyomas are benign fibromuscular tumors originating from the inner circular layer of the muscularis propria and are mostly located in the distal third of the esophagus. GISTs, the most common mesenchymal tumors of gastrointestinal tract, are potential malignancy arising from the intestinal cells of Cajal. For patients with GISTs larger than 2 cm, resection is recommended according to the guidelines of the National Comprehensive Cancer Network. For other benign SETs, tumor size >3 cm or symptomatic require surgical resection, as recommended by current guidelines. The present methods available for these tumors include endoscopic resection, open or laparoscopic wedge resection.[13,14] However, tumors located in the EGJ still remain a stimulating challenge for the surgeon. Proximal or total gastrectomy instead of wedge resection is typically chosen because of the potential stricture of gastroesophageal junction after surgery. In our study, 49.0% (24/49) of leiomyomas and 51.7% (45/87) of GISTs were treated with proximal gastrectomy. No statistically significant differences were found in resection range between GISTs and leiomyomas, even though leiomyomas are benign tumors. Partial or wedge resection, even proximal gastrectomy, is often chosen for GISTs located in such location, whereas benign leiomyomas can be treated with conservative therapy, enucleation, or tumor-everting resection, which could decrease the possibility of postoperative complications. Hence, exact preoperative differentiation between GISTs and leiomyomas may obviate the overtreatment of leiomyomas.
Our study indicates that 8 clinical and CT findings: older age (>46.5 years), tumor long diameter >4.5 cm, heterogeneous enhancement, high degree enhancement, mean CT attenuation > 69.2 HU, presences of intralesional low attenuation and surface ulcer, absence of calcification are significant features for GISTs in differentiation from leiomyomas. When all 8 significant features were used in combination using a scoring system, GISTs could be differentiated from leiomyomas with a relative high sensitivity (76.7%) and specificity (76.5%). However, there are still a small fraction of tumors could not be distinguished successfully in the preoperative setting. For symptomatic tumors or tumors with large size (>2 cm) in the EGJ, active surgical intervention is recommended in our study.
Our study results demonstrated that older age (>46.5 years) and larger long diameter (>4.5 cm) were more often observed in patients with GISTs than those with leiomyomas, which are concordant with those of previous studies.[7,15] In a single-center review of subepithelial esophagus tumors, the author found that patients with leiomyomas were younger than those with GISTs or leiomyosarcomas, and leiomyomas were smaller at presentation compared with GISTs and leiomyosarcomas. Contrary to the study reported by Yang et al, the LD/SD ratio was not a significant differential CT feature between GISTs and leiomyomas in our study. Most of leiomyomas and GISTs in our study had a round or ovoid shape, which is in line with a previous study.
In our study, the presence of surface ulcerations was more frequently appeared in GISTs compared to leiomyomas. Surface ulcerations were considered as the result of enlarging tumor restricting mucosal circulation, resulting in its ischemic and damage by gastric acidity.[17,18] In previous studies, the presence of surface ulcer was also regarded as a significant feature of GISTs more than of leiomyomas. Contrary to surface ulcerations, the presence of calcification was commonly found in leiomyomas but was hardly seen in GISTs in our study, which has not been reported. This result differs from previous studies, in which the presence of calcification was not a significant differential CT feature between leiomyomas and GISTs.[7,8] However, we have a larger sample sizes compared to previous studies, and further studies are needed for confirmation.
Intralesional low attention and heterogeneous enhancement were more prevalent in GISTs than in leiomyomas in our study. Intralesional low attention was defined as the presence of intratumoral necrosis or degeneration, hemorrhage, or gas. In GISTs with malignant potential, intralesional low attention, and heterogenous enhancement can be explained by the mistuning of neovascularization with quick growth of tumor.[20,21] Conversely, these features rarely appeared in benign tumors, in which tumor growth rate is line with the neovascularization. Our study also demonstrated that high degree enhancement, and mean CT attention >69.2 HU are significant features to differentiate GISTs from leiomyomas. These results are in accordance with those of previous studies.[8,22,23] The high-grade enhancement in GISTs reveals richer blood supplies and vascularity of tumor, which reflects an aggressive tumor biology.
Based on the significant clinical and CT features, a simple scoring method was established in our study. However, the sensitivity (76.7%) and specificity ((76.5%) of the present scoring method is not higher than that of a previous study, in which Liu et al  used 7 features to differentiate gastric GISTs from non-GISTs and got very high sensitivity (100%) and specificity (72%). Tumor location limited to the esophagogastric junction in our study may contribute to the condition. In addition, our study drew a relatively conservative conclusion although we incorporated a larger population than previous studies. This result suggests the limited differential ability of conventional CT characteristics, and a deep learning-based radiomics may be a better alterative technique and will be the future direction.
There are several limitations to our study. First, the number of GISTs and leiomyomas included in our study is different and patients who did not underwent tumor excision were excluded, which may lead to a certain selection bias. Second, our study included a small number of patients, although which incorporates a larger number than previous studies.
In conclusion, older age (>46.5 years), tumor long diameter >4.5 cm, heterogeneous enhancement, high degree enhancement, mean CT attenuation >69.2 HU, presences of intralesional low attenuation and surface ulcer, and absence of calcification are clinical and CT features highly suggestive of GISTs in differentiation from leiomyomas in the EGJ. When all these clinical and CT features are incorporated in combination using a scoring system, a higher sensitivity and positivity could be achieved for differentiating GISTs from leiomyomas in the EGJ.
Bo Zhang and Dan Cao proposed the idea and conceptualization. Xiaonan Yin and Yuan Yin performed data analysis, experimentation and scientific discussions, and prepared the original draft. Xijiao Liu and Caiwei Yang performed data analysis and scientific discussions. Xin Chen, Chaoyong Shen and Zhixin Chen validated the findings and helped in revision and organization of the paper.
Bo Zhang orcid: 0000-0002-0254-5843.
. Schulz RT, Fabio LC, Franco MC, et al. Predictive features for histology of gastric subepithelial lesions. Arq Gastroenterol 2017;54:11–5.
. Hwang SH, Park DJ, Kim YH, et al. Laparoscopic surgery for submucosal tumors located at the esophagogastric junction and the prepylorus. Surg Endosc 2009;23:1980–7. Epub 02008 May 00410.
. Sun LJ, Chen X, Dai YN, et al. Endoscopic ultrasonography in the diagnosis and treatment strategy choice of esophageal leiomyoma
. Clinics (Sao Paulo) 2017;72:197–201.
. Jiang G, Zhao H, Yang F, et al. Thoracoscopic enucleation of esophageal leiomyoma
: a retrospective study on 40 cases. Dis Esophagus 2009;22:279–83. Epub 02008 Nov 00819.
. Nemeth K, Williams C, Rashid M, et al. Oesophageal GIST-A rare breed case report and review of the literature. Int J Surg Case Rep 2015;10:256–9. Epub 2015 Feb 1017.
. Choi YR, Kim SH, Kim SA, et al. Differentiation of large (>/= 5 cm) gastrointestinal stromal tumors from benign subepithelial tumors in the stomach: radiologists’ performance using CT
. Eur J Radiol 2014;83:250–60. Epub 2013 Nov 1017.
. Liu M, Liu L, Jin E. Gastric sub-epithelial tumors: identification of gastrointestinal stromal tumors using CT
with a practical scoring method. Gastric Cancer 2018;9: 018-00908.
. Yang HK, Kim YH, Lee YJ, et al. Leiomyomas in the gastric cardia: CT
findings and differentiation from gastrointestinal stromal tumors. Eur J Radiol 2015;84:1694–700. Epub 2015 May 1621.
. Lee LS, Singhal S, Brinster CJ, et al. Current management of esophageal leiomyoma
. J Am Coll Surg 2004;198:136–46.
. Corless CL, Barnett CM, Heinrich MC. Gastrointestinal stromal tumours: origin and molecular oncology. Nat Rev Cancer 2011;11:865–78.
. Demetri GD, von Mehren M, Antonescu CR, et al. NCCN Task Force report: update on the management of patients with gastrointestinal stromal tumors. J Natl Compr Canc Netw 2010;8: (Suppl 2): S1–41. quiz S42-44.
. Li L, Wang F, Wu B, et al. Endoscopic submucosal dissection of gastric fundus subepithelial tumors originating from the muscularis propria. Exp Ther Med 2013;6:391–5. Epub 2013 Jun 3825.
. Cho JW. Current guidelines in the management of upper gastrointestinal subepithelial tumors. Clin Endosc 2016;49:235–40. Epub 2016 Feb 5922.
. Kim SY, Kim KO. Management of gastric subepithelial tumors: the role of endoscopy. World J Gastrointest Endosc 2016;8:418–24.
. Codipilly DC, Fang H, Alexander JA, et al. Subepithelial esophageal tumors: a single-center review of resected and surveilled lesions. Gastrointest Endosc 2018;87:370–7. Epub 2017 Aug 1014.
. Hur BY, Kim SH, Choi JY, et al. Gastroduodenal glomus tumors: differentiation from other subepithelial lesions based on dynamic contrast-enhanced CT
findings. AJR Am J Roentgenol 2011;197:1351–9.
. Chen TH, Hsu CM, Chu YY, et al. Association of endoscopic ultrasonographic parameters and gastrointestinal stromal tumors (GISTs): can endoscopic ultrasonography be used to screen gastric GISTs for potential malignancy? Scand J Gastroenterol 2016;51:374–7. Epub 00362015 Oct 00365522.
. Iannicelli E, Carbonetti F, Federici GF, et al. Evaluation of the relationships between computed tomography features, pathological findings, and prognostic risk assessment in gastrointestinal stromal tumors. J Comput Assist Tomogr 2017;41:271–8.
. Suster S. Gastrointestinal stromal tumors. Semin Diagn Pathol 1996;13:297–313.
. Miettinen M, Sobin LH, Lasota J. Gastrointestinal stromal tumors of the stomach: a clinicopathologic, immunohistochemical, and molecular genetic study of 1765 cases with long-term follow-up. Am J Surg Pathol V 29 2005;52–68.
. Levy AD, Remotti HE, Thompson WM, et al. Gastrointestinal stromal tumors: radiologic features with pathologic correlation. Radiographics2003 2003;23:283–304.
. Pinaikul S, Woodtichartpreecha P, Kanngurn S, et al. 1189 Gastrointestinal stromal tumor
(GIST): computed tomographic features and correlation of CT
findings with histologic grade. J Med Assoc Thai 2014;97:1189–98.
. Yang TH, Hwang JI, Yang MS, et al. Gastrointestinal stromal tumors: computed tomographic features and prediction of malignant risk from computed tomographic imaging. J Chin Med Assoc 2007;70:367–73.
. Miettinen M, Lasota J. Gastrointestinal stromal tumors--definition, clinical, histological, immunohistochemical, and molecular genetic features and differential diagnosis. Virchows Arch 2001;438:1–2.