Clinicopathological features and prognosis of omental gastrointestinal stromal tumor: evaluation of a pooled case series

Clinicopathological features and prognosis of omental GISTs are limited due to the extremely rare incidence. Therefore, the aim of the present study was to investigate the clinicopathological features and prognosis of omental GISTs. Omental GISTs cases were obtained from our center and from case reports and clinical studies extracted from MEDLINE. Clinicopathological features and survivals were analyzed. A total of 99 cases of omental GISTs were enrolled in the present study. Omental GISTs occurred predominantly in greater omentum (78/99, 78.8%). The majority of tumors exceeded 10 cm in diameter (67/98, 68.3%) and were high risk (88/96, 91.7%). Histological type was correlated with tumor location and mutational status. The five year DFS and DSS was 86.3% and 80.6%, respectively. Mitotic index was risk factor for prognosis of omental GISTs. Prognosis of omental GISTs was worse than that of gastric GISTs by Kaplan-Meier analysis. However, multivariate analysis showed that the prognosis was comparable between the two groups. The majority of omental GISTs were large and high risk. Mitotic index was risk factor for prognosis of omental GISTs. The prognosis was comparable between omental and gastric GISTs.

The tumors ranged from 0.7 to 40 cm in maximum diameter (median, 13.0 cm; mean, 14.1 cm). Forty-two patients displayed spindle cell morphology ( Table 1. Survival data of 63 patients were eventually selected for analysis using exclusion criteria described in the materials and methods. The follow up time ranged from 2 to 134 months (mean, 36.6 months; median, 21.1 months). Seven patients showed recurrence or metastasis, 6 patients suffered from GIST related deaths. The 1-, 3-and 5-year DFS was 90.8%, 86.3% and 86.3%, respectively. The 1-, 3-and 5-year DSS was 100.0%, 87.9% and 80.6%, respectively. The DFS and DSS of omental GISTs were analyzed using Kaplan-Meier survival analyses and shown in Fig. 1.
The relationship between clinicopathological features were analyzed (data not shown). The histological type was correlated with tumor location and mutational status. The ratio of epithelioid and mixed morphology of greater omental GISTs were significantly higher than that of lesser omental GISTs (P = 0.036). The epithelioid and mixed morphology were significantly correlated with PDGFRA mutation (P < 0.001).
Prognostic factors for DFS and DSS of omental GISTs according to univariate analysis were shown in Table 2. The results showed that the tumor size and mitotic index were risk factors for DFS of omental GISTs, and mitotic index was the only risk factor for DSS of omental GISTs. The DFS and DSS of omental GISTs according to tumor size and mitotic index were shown in Figs 2 and 3.
The clinicopathological features of 99 omental GISTs including age, gender, tumor size, histological type, mitotic index and NIH risk category were compared with 297 gastric GISTs in our institution ( Table 3). The results showed that the distribution of tumor size, histological type and NIH risk category were significantly different between omental and gastric GISTs (all with P < 0.001).
In order to compare the prognosis of omental GISTs with gastric GISTs, survivals of 63 cases of omental GISTs and 217 cases of gastric GISTs with follow up data were analyzed. The results showed that the DFS and DSS of omental GISTs were significantly worse than that of gastric GISTs (Fig. 4). Further, multivariate analysis was performed to evaluate the prognostic value of locations (Table 4). The results showed that location was not an independent risk factor for prognosis of omental and gastric GISTs.

Discussion
Clinicopathological features and prognosis of omental GISTs are limited due to the extremely rare incidence. Therefore, the aim of the present study was to investigate the clinicopathological features and prognosis of omental GISTs from our center and from literatures in MEDLINE. The present study represents the largest analysis of omental GISTs and indicates some features significantly associated with omental GISTs.

Characteristics Parameters
Mean (m, ± SD) 36. To date, the largest cases of omental GISTs was reported by Miettinen et al. 5 . The study contained 95 cases. However, it mainly focused on clinicopathological features of omental GISTs. The distribution of age, gender, tumor size, and mitotic index were similar to our present study. However, a few highlights with respect to clinicopathological features were revealed and the prognosis of omental GISTs were analyzed in depth in our present study.
The precise etiology of omental GISTs remains to be clarified. In the study reported by Miettinen et al. 5 , over half of the solitary omental GISTs were attached to or involved the gastrointestinal tract, and the histologic features were similar to gastric or small intestinal GISTs. Thus, they believed that solitary omental GISTs are actually      Table 4. Comparative survival analysis of omental and gastric GISTs using univariate and multivariate analysis.
externally extending gastric or small intestinal GISTs, and many others may have lost their original connection to the stomach or small intestine and become parasitically attached into the omentum. For multiple omental GISTs, they were believed to be metastatic tumors from an overlooked primary tumor. However, it has been reported that GISTs in the omentum are derived from mesenchymal cells that are less differentiated than ICCs 6 . These may be ICC precursors straying into the abdominal cavity 7 . Moreover, Sakurai et al. found that KIT positive bipolar cells were present just beneath the mesothelial cells of the omentum. Thus, the identification of an ICC-counterpart in the omentum is the evidence that omental GISTs may also originate from ICC. They also demonstrated the existence of ICC-like cells focally in the omentum at 21 weeks of human gestation 8 . However, it is unknown whether they develop in situ or migrate from the ICC of the tubular GI tract at particular point in fetal development. Dedemadi et al. reported that there is no difference in the incidence between lesser and greater omentum 9 . However, the incidence of GISTs in the greater omentum was approximately four times as much as that of lesser omentum in our present study. The difference in the incidence between lesser and greater omentum needs further investigation.
In our present study, the majority of omental GISTs exceeded 10 cm in diameter and approximately ninety percent of the tumors were classified as high risk category. The spectrum of tumor size and NIH risk category of omental GISTs were significantly different from that of gastric GISTs in our institution. This may attribute to the absence of specific symptoms when the tumor was not large enough in the omentum or tumors did not invade adjacent gastrointestinal tract. Once the GISTs of the omentum reached a significant size, symptoms will appears including abdominal pain, mass, distension, fatigue and discomfort. Thus, early diagnosis of omental GISTs is very difficult.
Histologically, most GISTs display spindle cell morphology (70%), whereas a minority is of epithelioid (20%) or mixed phenotypes (10%) 10 . However, in the study reported by Miettinen et al. 5 , 53 out of 89 tumors showed spindle cell morphology (59.6%), 28 tumors were epithelioid (31.5%) and 8 tumors were mixed (8.9%). In our present study, 29 tumors displayed epithelioid morphology (33.3%) and 16 tumors displayed mixed morphology (18.4%). The proportion of epithelioid and mixed morphology of omental GISTs were significantly higher than that of gastric GISTs in our center and previous report. This indicated that the constituent ratio of epithelioid and mixed morphology could be various from each other depending on the location of GISTs. Further, we found that the incidence of epithelioid and mixed cell morphology was higher in greater omental GISTs than in lesser omental tumors. This may indicate that the origins of tumors in lesser omentum and greater omentum were different from each other, which needed further investigation.
In 1998, Hirota et al. reported their groundbreaking discovery of KIT mutations in GISTs. It is now established that 70% to 80% of GISTs harbor a KIT gene mutation 11 , and PDGFRA mutations occur in approximately 8% to 10% of gastric GISTs 12 . In the study reported by Miettinen et al. 5 , KIT mutation was detected in 15/36 tumors (41.7%), and PDGFRA mutation was detected in 11/36 tumors (30.6%). However, in our present study, only 9/27 tumors (33.3%) harbored KIT mutation but 14/27 tumors (51.9%) harbored PDGFRA mutation. Although the incidence of PDGFRA mutation in Miettinen's and our report were higher than previous report, the results in our present study was even higher than that in Miettinen's report. It was reported that spindle cell morphology correlates with KIT mutations 13 and epithelioid and mixed cell morphology correlates with PDGFRA mutations 14 . This has also been demonstrated in our present study, the KIT mutations almost exclusively occurred in spindle cell morphology, and PDGFRA mutations almost exclusively occurred in epithelioid cell morphology. This indicated that KIT and PDGFRA mutant GISTs probably represent two distinct clinicopathological and molecular genetic disease entities. However, this needs further investigations in depth. It must be pointed out that the data of mutational analysis is only available in too few cases (27/99, 27.3%) in our present study, which are extremely too low to characterize the mutation spectrum of omental GISTs. The limited data could also result in bias during analyzing the association between cell morphology and mutational status. This was one limitation in our present study.
Besides tumor size and mitotic index, tumor location is also one important risk factor for the prognosis of GISTs 15 , and it was considered that extra-gastrointestinal stromal tumors were more aggressive than gastric GISTs in clinical course. However, the modified NIH risk classification system distinguishes only gastric from non-gastric GISTs, and the prognosis of omental GISTs are not discussed. Thus, we compared the prognosis of omental GISTs with gastric GISTs in our center. We found that the prognosis of omental GISTs was significantly worse than that of gastric GISTs. However, multivariate analysis showed that the prognosis was comparable between omental and gastric GISTs. This indicated that the prognosis of omental GISTs was as considerable as gastric GISTs. The significantly lower survival of omental GISTs than gastric GISTs in Kaplan-Meier survival analysis may attribute to the larger tumor size and higher NIH risk category of omental GISTs compared with gastric GISTs. However, it was inevitable that the extremely low incidence of imatinib therapy in our present study would result in bias during analysis of prognosis of omental GISTs. Thus, the actual disease free survival and disease specific survival of omental GISTs may be more favorable than that in our present study.
There are a few limitations in our present study. Firstly, the present study is a retrospective analysis and the completeness of data is limited. This will influence the analysis of clinicopathological features and prognosis. Secondly, the sample size of omental GISTs was not large enough, which will result in statistical bias. Thirdly, due to the limited sample size of duodenal and small intestinal GISTs in our center, the prognosis of omental GISTs were only compared to that of gastric GISTs.

Conclusions
The majority of omental GISTs occurred in greater omentum, exceeded 10 cm in diameter and were high risk. The incidence of epithelioid cell morphology and PDGFRA mutation were relatively high in omental GISTs. The histological type was correlated with location and mutational status. Mitotic index was risk factor for prognosis of omental GISTs. Omental GISTs differ significantly from gastric GISTs in respect to clinicopathologic features. The prognosis was comparable between omental and gastric GISTs.  67 , and written informed consent was obtained from the two patients in our center.
Data including age, gender, accompanied tumor, symptoms, location, tumor size, surgical intervention, histological type, immunohistochemical features, mutational status, mitotic index, NIH risk category, adjuvant therapy, tumor progression and survival data were recorded. The tumors were categorized into very low, low, intermediate and high risk groups according to the modified NIH risk classification criteria reported by Joensuu et al. 68 . For survival analysis, the inclusion criteria were listed as follows: 1. R0 resection, 2. without distant metastasis, 3. without GIST in other locations, 4. without other malignant tumors, 5. without neoadjuvant imatinib therapy, 6. with follow up data. Due to data acquisition, completeness of data is limited.
Data were processed using SPSS 22.0 for Windows (SPSS Inc., Chicago, IL, USA). Discrete variables were analyzed using the Chi-square test or Fisher's exact test. Numerical variables were expressed as the mean ± SD unless. Significant predictors for survival identified by univariate analysis were further assessed by multivariate analysis. Evaluation of disease-free-survival (DFS) and disease-specific-survival (DSS) were obtained by the Kaplan-Meier method. The P values were considered to be statistically significant at the 5% level.

Ethical approval and informed consent. This study was approved by the Ethics Committee of Xijing
Hospital, and written informed consent was obtained from the two patients in our center.