Clinicopathological features and prognosis of colonic gastrointestinal stromal tumors: evaluation of a pooled case series

Background Due to the extremely rare incidence, data about colonic GISTs are limited. Therefore, aim of the present study was to explore clinicopathological characteristics and prognosis of colonic GISTs. Patients and Methods Colonic GISTs cases were obtained from our center and from case report and clinical studies extracted from MEDLINE. Clinicopathological features and survivals were analyzed. Results There were 79 colonic GISTs patients with a female predominance. The median age was 66 years (range 0.17-84). The median tumor size was 5.8 cm (range 0.5-29). The most common location was sigmoid colon (45.8%), followed by transverse colon (19.5%). The majority of colonic GISTs were high risk (70.8%). Mitotic index was correlated with gender (P = 0.002) and tumor size (P = 0.005), and tumor location was correlated with age (P = 0.017). The five year DFS and DSS were 57.4% and 61.6%, respectively. Mitotic index and NIH risk classification were associated with prognosis of colonic GISTs. However, mitotic index was the only independent risk factor. The distribution of tumor size and NIH risk classification were significantly different between colonic and gastric GISTs (both P = 0.000). The DFS and DSS of colonic GISTs were significantly lower than that of gastric GISTs (P = 0.012 and P = 0.002, respectively). Conclusions The most common location for colonic GISTs was sigmoid colon. Most tumors were high risk. Mitotic index was the only independent risk factor for prognosis of colonic GISTs. Colonic GISTs differ significantly from gastric GISTs in respect to clinicopathological features. The prognosis of colonic GISTs was worse than that of gastric GISTs.

GISTs can occur anywhere in the alimentary tract but most commonly in the stomach (40% to 70%) [5]. GISTs located in the colorectum are relatively rare, representing approximately 5% of all GISTs [6]. GISTs located in colon is much rarer, and it represents only 1-2% of all cases [7]. Thus, studies involving large numbers of colonic GISTs are lacking, and the clinicopathological

Clinical Research Paper
Oncotarget 40736 www.impactjournals.com/oncotarget profiles and prognosis are limited. Therefore, the aim of our present study was to explore clinicopathological characteristics and prognosis of colonic GISTs.
The relationship between clinicopathological features were summarized in Table 2. The mitotic index exceeded 5/50HPF for the majority of male patients but only for the minority of female patients (P = 0.002), and it was positively related with tumor size (P = 0.005). Age was associated with tumor location (P = 0.017). For patients less than 60 years old, the commonest location were transverse and sigmoid colon. For patients more than 60 years old, the commonest location was sigmoid colon only.
Survival data of colonic GISTs were summarized in Table 3. Survival data of 48 patients were eventually  selected for analysis. The median follow-up time was 23.5 months (range from 3 to 149 months). Fifteen cases showed recurrence or metastasis, 18 patients suffered from GIST related deaths. The 1-, 3-and 5-year DSS was 85.0%, 66.0% and 61.6%, respectively. The 1-, 3-and 5-year DFS was 90.8%, 78.0% and 57.4%, respectively. The DFS and DSS of colonic GISTs were shown in Figure  2.
Prognostic factors for DFS and DSS were shown in Table 4. Mitotic index and NIH risk classification were associated with prognosis of colonic GISTs. However, only mitotic index was independent risk factor. The DFS and DSS according to mitotic index and NIH risk classification were shown in Figure 3 and 4. NIH risk classification was not enrolled in logistic regression analysis, although it showed significant correlation with prognosis. Because no patients suffered from recurrence, metastasis or death in NIH risk category 1 and 2. When calculating log of odds ratio, the null frequency resulted in a computational error due to presence of logarithm of zero.
The clinicophathological features of 79 colonic GISTs were compared with 297 gastric GISTs in our center ( Table 5). The results showed that the distribution of tumor size and NIH risk classification were significantly  Oncotarget 40740 www.impactjournals.com/oncotarget different between colonic and gastric GISTs (both P = 0.000). In order to compare the prognosis between colonic and gastric GISTs, the two groups were matched according to tumor size, mitotic index and imatinib treatment. The entire process was shown in Figure 5. Finally, 39 cases of colonic GISTs and 39 cases of gastric GISTs were selected. No intergroup difference was found in age, gender, tumor size, mitotic index and imatinib treatment ( Table 6). The DFS (P = 0.012) and DSS (P = 0.002) of colonic GISTs were significantly lower than that of gastric GISTs (56.0% vs 85.7%, 62.7% vs 96.3%).

dIscussIon
GISTs located in the colon constitute a very rare subset with limited data on the clinicopathological features and prognosis. Therefore, we evaluated 79 cases of colonic GISTs from our center and from literatures in MEDLINE. The present study represents the largest analysis of colonic GISTs.
The study containing 37 cases of colonic GISTs reported by Miettinen et al. [8] was the only one with a much larger number of patients. In the series, the

Figure 3: dFs and dss of colonic GIsts by mitotic index.
Oncotarget 40741 www.impactjournals.com/oncotarget commonest site was sigmoid colon, followed by transverse colon, cecum, descending and ascending colon. In our present study, the commonest location was also sigmoid colon, and followed by transverse colon. However, the ratio of cecum, ascending and descending colon were almost equivalent. The distribution of GISTs may  Oncotarget 40742 www.impactjournals.com/oncotarget be attributed to the distribution of ICCs in the colon. However, Hagger et al. reported that the highest density of ICC was at the myenteric plexus of the transverse colon, not the sigmoid colon [9]. Further, we found that the distribution of colonic GISTs was correlated with age of patients. For patients less than 60 years old, the commonest location were transverse and sigmoid colon. For patients more than 60 years old, the commonest location was sigmoid colon only. This may be attributed to the decline of ICC number in the colon with age at a rate of 13% per decade reported by Gomez-Pinilla et al. [10]. They also found that volume of ICC networks decreased more quickly with age in the ascending colon than that in the sigmoid colon. Although the distribution of colonic GISTs in our present study could not fully elucidated by the above studies, they provide clues for the investigation of distribution of colonic GISTs.
The surgical treatment of GISTs is radical resection of the primary tumor with negative microscopic margins. It is well known that lymph node involvement are rare [11], and lymphadenectomy or mesorectal excision is unnecessary. An appropriate segmental en bloc resection is enough for colonic GISTs only if adjacent organs are involved [12]. However, 3 of 17 patients (17.6%) had lymph node metastasis in our present study, which was apparently higher than the previous report. Although the incidence of lymph node metastasis was relatively low, necessity of lymphadenectomy may need careful consideration in the treatment of colonic GISTs.
Even with surgical resection, there is a high risk of recurrence and metastasis. Distant metastases are the most frequent treatment failure for colorectal GISTs and are associated with poor prognosis. The most common site of metastasis in colorectal GISTs is liver, followed by peritoneum. Other locations include pleura, lung, bone and retroperitoneum [13]. In our present study, 9 patients suffered liver metastasis and 7 patients suffered peritoneal metastasis. Whereas GISTs predominantly metastasis to liver and leiomyosarcomas mainly spread to lung [14]. This may assistant in the differential diagnosis of the two tumors.
It was reported that approximately 10-30% of GISTs are regarded as clinically malignant [15], and tumor size and mitotic index are valuable predictors for evaluating malignant potential of GISTs [16]. In the present study, mitotic index more than 5/50HPF and high risk category were associated with poorer prognosis. Tumor location is also a critical risk factor for recurrence after radical surgical resection [17]. However, the modified NIH risk classification system only distinguishes gastric from nongastric tumors. The prognostic features of colonic GISTs are unclear. Considering different distribution of tumor size and NIH risk category between colonic and gastric GISTs, patients were matched in order to compare the prognosis. The survival analysis showed that the DFS and DSS of colonic GISTs were significantly lower than gastric GISTs. It was reported that non-gastric GISTs have similar risk for tumor recurrence [18]. However, the survival of colonic GISTs were not compared with that of duodenum, small intestine or rectum, due to the limited sample size of GISTs in these locations in our center.
There are some limitations of the present study. First, the present study is a retrospective analysis and lacks systematic prospective data. Therefore, completeness of the data is limited. Second, the sample size of colonic GISTs was not large enough, which will result in sampling error. Third, due to the limited sample size of duodenum, small intestine or rectum GISTs in our center, we could not

conclusIons
The commonest location for colonic GISTs was sigmoid colon, followed by transverse colon. Most colonic GISTs are high risk category. Mitotic index was the unique independent predictor for the prognosis of colonic GISTs. Colonic GISTs differ significantly from gastric GISTs in respect to clinicopathological features. The prognosis of colonic GISTs was worse than that of gastric GISTs.  [8,[40][41][42] including 54 cases. To this end, a total of 79 colonic GISTs patients were identified ( Figure  1). In addition, the clinicopathological characteristics and prognosis of 297 cases of gastric GISTs were analyzed and compared with colonic GISTs. This study was approved by the Ethics Committee of Xijing Hospital, and written informed consent was obtained from the three patients in our center.

PAtIents And MetHods
Clinicopathological data including age, gender, accompanied tumor, symptoms, location, tumor size, surgical intervention, histological type, lymph node involvement, mitotic index, immunohistochemical features, mutational status, NIH risk category, adjuvant therapy, tumor recurrence or metastasis and survival data were recorded from hospital medical records or extracted from published reports and studies. The tumors were categorized into very low, low, intermediate and high risk groups according to the modified NIH risk classification criteria [43]. For survival analysis, the exclusion criteria were listed as follows: 1. not receive R0 resection, 2. with tumor rupture during operation, 3. accompanied with distant metastasis, 4. accompanied with GIST in other locations, 5. accompanied with other malignant tumors, 6. with neoadjuvant imatinib therapy, 7. without follow up data. Due to data acquisition, completeness of data is limited.
Data were processed using SPSS 16.0 for Windows (SPSS Inc., Chicago, IL, USA). Numerical variables were expressed as the mean ± SD unless otherwise stated. Discrete variables were analyzed using the Chi-square test or Fisher's exact test. Significant predictors for survival identified by univariate analysis were assessed by multivariate analysis using the logistic regression analysis. Evaluation for disease-free-survival (DFS) and disease-specific-survival (DSS) were obtained by the Kaplan-Meier method. The P value was considered to be statistically significant at the 5% level.

AcknowledGMents
This study was supported in part by grants from the National Natural Scientific Foundation of China [NO. 31100643, 31570907, 81572306, 81502403, XJZT12Z03].

conFlIcts oF Interest
There are no financial or other relations that could lead to a conflict of interest.