Clinicopathological Features and Clinical Outcomes of Primary Hepatic Gastrointestinal Stromal Tumors: Evaluation of a Pooled Case Series


 Background: Due to the extremely rare incidence, data of clinicopathological features and prognosis of primary hepatic gastrointestinal stromal tumors (GISTs) is limited. Methods: 36 cases of hepatic GISTs were from the literature, PUBMED, EMBASE, China National Knowledge Infrastructure and WANFANG DATA, and 1 case came from our center. Clinicopathological features and outcomes were analyzed between 37 hepatic GISTs and 254 gastric GISTs from our center. Results: A majority of hepatic GISTs exceeded 5 cm (83.7%), displayed mixed density (69.4%) and spindle morphology (74.2%) and were classified as high risk (91.4%). Larger tumors of hepatic GISTs were likely to display mixed lesion and tumors with mixed lesion were prone to be classified as high risk. In comparisons to gastric GISTs, hepatic GISTs differed from gastric GISTs in tumor size, main symptoms, histologic type, mitotic index, CD34 expression, NIH risk classification. In patients with hepatic GISTs, 5-year DFS and DSS rates were 19.4% and 53.7%, which were worse than that of gastric GISTs (P＜ 0.001), especially for those with tumor size exceeding 5 cm or mitotic indices exceeding 5/50 HPF (P < 0.001). Multivariable analysis showed location and NIH risk classification were independent prognostic factors for DFS in patients with GISTs, and size and location were significantly associated with DSS. Conclusions: Hepatic GISTs distinguished from gastric GISTs in respect to clinicopathological features and outcomes. Mitotic index exceeding 5/50 HPF or tumor size exceeding 5 cm may be important factor to distinguish hepatic GISTs from gastric GISTs in DFS and DSS.


Background
Gastrointestinal stromal tumor (GIST) with some malignant potential is the most common gastrointestinal mesenchymal tumor and considered to originate from naive interstitial cells that differentiate into Cajal cells(ICC). [1][2][3] Frequent mutations in genes for c-KIT and platelet-derived growth factor receptor alpha (PDGFRA) may promote GISTs by activating or auto-phosphorylating their downstream signaling pathways. [4][5] About 70% of gastrointestinal stromal tumors show spindle-cell morphology, followed by epithelioid (20%) and mixed type (10%). 6 Immunohistochemical detection of CD117, CD34, DOG-1 plays an important role in the diagnosis of GIST. [7][8] Although GISTs account for only 1-2% of gastrointestinal malignancies, they can occur anywhere in the gastrointestinal tract, 9 which most were frequently located in the stomach(45%-65%), followed by the small intestine(15-25%), colon and rectum (5-10% in), and the esophagus(5-10%). 5 However, GISTs also occur outside of the gastrointestinal tract (pancreas, liver, gallbladder, omentum, mesentery, retroperitoneum, vagina, prostate, etc.) as primary tumor, which are identi ed as extra-gastrointestinal stromal tumors (EGISTs). [10][11][12][13][14][15] Studies of primary hepatic GISTs are really rare, most of which are case reports with limited data. In this study, we evaluated clinicopathological features and prognosis of primary hepatic GISTs in order to optimize the diagnosis and treatment. types, mitotic index, Ki-67 expression, immunohistochemistry (IHC) detection, lymph node metastasis, gene mutation status, surgical intervention, adjuvant therapy, and modi ed National Institutes of Health (NIH) classi cation (very low risk, low risk, intermediate risk or high risk).
To analyze primary hepatic GISTs prognosis and make comparison of outcome between gastric and hepatic GISTs, we included cases for which (1) patients were diagnosed with primary GIST of the liver or stomach, without GISTs of other organs, (2) didn't suffer from other malignant tumors, (3) had no distant metastasis, (4) underwent R0 resections (6) the tumor was integrally removed (7) with follow-up data (8) didn't received preoperative adjuvant imatinib therapy.

Statistical analysis
All continuous values are presented as mean ± standard deviation. Statistical analysis was conducted using the Chi-square test or Fisher's test for categorical variables. The multivariable Cox proportional hazard model and univariate analysis were used to analyze prognostic risk factors in patients with GISTs. The Kaplan-Meier method was used to calculate survival curves for disease-free survival (DFS) and disease-speci c survival (DSS); differences between the curves were assessed using the log-rank test. Data was proceeded by SPSS 23.0 software. The P value was considered to be statistically signi cant at the 5% level.
Tumors density was correlated with tumor size, and NIH risk classi cation (Table 2). A majority (84.2%) of tumors with the maximum diameter exceeding 10 cm, displayed mixed lesion (P = 0.002). Most highrisk tumors (77.4%) were presented with mixed lesion, whereas all low-risk tumors displayed solid lesion (P < 0.001). Density of tumor is determined by the imaging characteristics or pathological ndings. KIT, c-KIT proto-oncogene protein; PDGFRA, platelet-derived growth; factor receptor alpha; NIH, National Institutes of Health.
Diseases-free survival (DFS) time and disease-speci c survival (DSS) time were compared between primary hepatic and gastric GISTs (Fig. 1). Patients were followed for a median period of 13.0 months (range: 1-189 months; mean: 32.4 months;  Fig. 2, Fig. 3). Univariate and multivariable analysis showed location and NIH risk classi cation to be independent prognostic factors for DFS, and size and location were signi cantly associated with DSS (Table 5).

Discussion
As primary hepatic GISTs that occur outside the gastrointestinal tract are extremely rare, there was only one case in our database, which includes a total of 632 cases of GISTs from 2006 to 2019 (14 years).
Most literature consists of case reports, or metastatic hepatic GISTs, with limited data. Compared to previous study (22 cases), the current study represents the largest number of hepatic GISTs consisting of 37 cases, and even evaluated the relationship between tumor density and clinicopathological features. 45 It implied that mitotic index exceeding 5/50 HPF or tumor size exceeding 5 cm was the important point to distinguish primary hepatic GISTs from gastric GISTs in DFS and DSS, which previous research had never reported. The result was also in accordance with the conclusion of modi ed NIH risk classi cation that gastric and non-gastric GISTs with tumor size less than 5 cm or mitotic indices less than 5/50 HPF shared the same risk classi cation (very low or low risk), however, for those with greater tumor sizes or higher mitotic index, non-gastric GISTs showed higher risk classi cation than gastric GISTs.  49 Together, these studies imply the possibility of primary hepatic GISTs. Lack of speci c clinical manifestations, the main symptoms were abdominal pain, discomfort, abdominal lesion, and abdominal distension. We report the positive expression of CD117 and CD34 was 97.1% and 66.7% in hepatic GISTs, which were in accordance with the results of previous studies about gastric GISTs (CD117, 95%; CD34, 70%). 50 The lesion distribution of hepatic GISTs included left lobe, right lobe and caudate lobe.
In the study, 25 cases (69.4%) showed mixed lesion and 4 cases (11.2%) showed cystic lesion. Mixed or cystic lesions of the liver are likely to be hepatic carcinoma, metastatic hepatic tumors, hepatic cyst, hepatic hemangioma, liver abscess, hepatic mucous cyst. The CT images of primary hepatic GISTs were similar to those of metastatic hepatic GISTs, presenting with heterogeneous hypodense lesions due to necrosis, hemorrhage and cystic degeneration of the mass. The CT image of our patient with heterogeneous hypodense mass was shown in Fig. 4. Majority studies demonstrated hepatic GISTs showed peripheral enhancement in the arterial phase and progressive concentric enhancement in the portal venous phase. 45 Our report of primary hepatic GIST was consistent with previous study with progressive concentric enhancement in CT images. However, Hyum reported a case of primary hepatic GISTs showed the same degree of enhancement compared to live parenchyma in the arterial phase and persistent enhancement in the portal venous and delayed phase. 32 And he deemed GISTs were consisted of a wide range of pathologic spectrum, contributing to the discrepancy of different enhancement degree and pattern.
The present study even took consideration of the relationship between tumor density (solid, cystic and mixed lesion) and clinicopathological features, and discovered larger tumors were likely to display mixed density and tumors with mixed lesion were associated with high risk classi cation. Tumor density may be an indicator of risk classi cation for primary hepatic GIST. Size, mitotic count, anatomical location and rupture are important factors in evaluating the prognosis. 51 Larger tumors were likely to display mixed density, perhaps due to insu cient blood supply, leading to tumor necrosis or cystic degeneration.
Tumors with mixed lesion were associated with high risk classi cation because of the connection between larger size and risk classi cation.
The radiological features of heterogenous hypodense and hypervascular nature of the tumor indicated hepatic GISTs, but also suggested other diagnoses, such as hepatocellular carcinoma. GISTs was de nitively diagnosed by the histopathological examination of a tissue sample. The different cell morphology (spindle, epithelioid or mixed), distinctive immunohistochemical detection of CD117, CD34 and Dog-1 and molecular alterations features of c-KIT and PDGFEA genes mainly formed the diagnosis criteria of GIST. Li et al. demonstrated that (EUE-FNA) was the quickest and least painful way to obtain an accurate tissue diagnosis of hepatic GISTs. 44 However, because of the small amount of obtained tissue, diagnosing gene mutations status is di cult by this method and bleeding or implantation metastasis is possible. Most cases involving the study were diagnosis of primary hepatic GIST, by postoperative histopathological examination and few cases without operation were diagnosed by ne needle aspiration cytology.
Bleeding, bloating, abdominal discomfort, organ dysfunction and distant metastasis resulting in multiple organ failure were the main reasons for patients to seek for medical care. Compared with gastric GISTs, bleeding was less common in hepatic GISTs (P < 0.001) as well as spindle morphology (P < 0.001). Primary hepatic GISTs were most frequently newly diagnosed as higher risk than gastric GISTs (91.8% vs 34.8%, P < 0.001). Surgery was considered to be rst-line therapy for resectable GISTs. R0 resections signi cantly improved survival of patients with primary hepatic GISTs and reduced postoperative recurrence rate. 52 It was reported extracorporeal hepatic resection and auto-transplantation (ECHRA) is an alternative method for tumors with large size, lesion extension or location adjacent to critical structures. [53][54] In our study, 25 patients had received R0 resection, and 5-years DFS and DSS rate were 19.4% and 53.7% in patients with follow-up data (19 patients), including one who underwent ECHRA and survived for 22 years, the longest disease-speci c survival as far reported. ECHRA is likely to be a promising therapeutic strategy for large or awkwardly located hepatic GISTs with no metastasis.
Radiofrequency ablation was performed in 2 cases with very low risk classi cation, one case with 17 months of disease-free time. Otherwise, imatinib, a tyrosine kinase inhibitor against c-KIT and PDGRA, was widely used to treat unresectable GISTs, post-surgical relapses or metastases after surgery, or postoperative adjuvant therapy, which had been shown to improve survival of GISTs. 55 E cacy of target therapy was associated with mutation status. Though wild type GISTs patients were treated with imatinib with a worse response, imatinib was also considered to improve the overall disease control rate (> 60%). [56][57][58] Sunitinib was recommended as a second-line treatment for recurrent, metastatic or unresectable GISTs and the rst choice for generalized progression after standard-dose imatinib therapy. As non-gastric GISTs have higher risk of recurrence than gastric GISTs, 3 years of adjuvant therapy was found that tumor size, mitotic rate and location independently predicted recurrence after resection of primary stromal tumors. 59 And previous studies ever reported they were important malignant potential predictors of GIST. 60 Our multivariate analysis of prognostic factors of hepatic and gastric GISTs showed location and NIH risk classi cation to be independent prognostic factors for DFS and location and size were signi cantly associated with DSS. Compared with gastric GISTs, hepatic GISTs suffered from worse 5-year DFS rate (19.4% vs 80.9%, P < 0.001) and DSS rate (53.7% vs 92.9%, P < 0.05).
Our study has some limitations, mainly for its retrospective design and incomplete data collection. Firstly, the small hepatic GISTs sample size may have produced selection bias, leading to possible false-positive or -negative results. Secondly, we didn't make comparisons of survival data by tumor size less than 5 cm and mitotic index less than 5/50 HPF for limited data acquisition. Finally, we also did not perform subgroup analyzes of clinicopathological features and prognosis between hepatic and gastric GISTs, or between primary hepatic GISTs and other EGISTs, such as pancreas or omentum. Optimal treatment for hepatic GISTs should therefore be further studied.

Conclusion
Majority of primary hepatic GISTs share an immune pro le of positive expression of CD117 and Dog-1, similar to that of gastric GISTs. The 37 hepatic GISTs and 254 gastric GISTs signi cantly differed in tumor size, main symptoms, histologic type, mitotic index, CD34 expression, NIH risk classi cation.
Compared with gastric GISTs, hepatic GISTs were obviously larger in size and showed higher mitotic index. Mitotic index exceeding 5/50 HPF or tumor size exceeding 5 cm may be the critical factor to distinguish hepatic GISTs from gastric GISTs in DFS and DSS. The study did not receive any speci c grant from funding agencies in the public, commercial or not-forpro t sectors.

Abbreviations
Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Authors' contributions XX searched and reviewed the literature, designed the research and wrote the paper. ZGL designed the research and provided the clinical data. ZZC participated in reviewing the literature, designing the research, revising and writing the paper.
Ethics approval and consent to participate This study was approved by the Ethics Committee of Liao Ning Cancer Hospital.

Consent for publication
Written informed consents for publication were obtained from the patients in our center.

Competing of interest
There were no relevant a liations or nancial involvement with any organization or entity with a nancial interest in or nancial con ict with the subject matter or materials discussed in the manuscript.

Figure 1
Comparisons of survival data between gastric and hepatic GISTs. Kaplan-Meier survival analysis was employed to compare difference between groups. a. The comparison of DFS in patients with hepatic and gastric GISTs, P < 0.001. b. The comparison of DSS in patients with hepatic and gastric GISTs, P < 0.001.

Figure 2
Comparisons of disease-free survival and disease-speci c survival associated with mitotic index between gastric and hepatic GISTs. Kaplan-Meier survival analysis was employed to compare difference between groups. a. Comparison of disease-free survival associated with mitotic index exceeding 5/50 HPF Page 23/24 between gastric and hepatic GISTs, P = 0.027. b. Comparison of disease speci c survival associated with mitotic index exceeding 5/50 HPF between gastric and hepatic GISTs, P < 0.001.

Figure 3
Comparisons of disease-free survival and disease-speci c survival associated with tumor size between gastric and hepatic GISTs. Kaplan-Meier survival analysis was employed to compare difference between groups. a. Comparison of disease-free survival associated with tumor size exceeding 5 cm between gastric and hepatic GISTs, P < 0.001. b. Comparison of disease speci c survival associated with mitotic index exceeding 5 cm between gastric and hepatic GISTs, P < 0.001.