Abstract
Background
Gastrointestinal stromal tumors (GISTs) can undergo malignant transformation, and thus, the risk grading assessment of postoperative patients is highly critical. This study sought to investigate the correlation between ultrasonic characteristics, clinicopathological features, and biological risk grading in patients with gastric GISTs and to determine whether the preoperative prediction of biological risk grading is feasible.
Methods
The ultrasonic characteristics and clinicopathological data of gastric filling in 92 patients with GISTs confirmed by surgical pathology were retrospectively analyzed, the influencing factors of the biological risk classification of GISTs were assessed through univariate and multivariate analyses, and a prediction model was constructed. The receiver operating characteristic curve was plotted to analyze the predictive value of the logistic regression model.
Results
Univariate analysis revealed that melena was significantly more common in the high-risk group (P<0.05). Tumor size, morphology, echogenicity, calcification, and cystic changes also differed significantly between the risk groups (P<0.05), while location, growth pattern, blood flow grade, and ulceration did not (P>0.05). Multivariate analysis indicated that the independent risk predictors were tumor size [odds ratio (OR) =0.028; P=0.002] and echogenicity (OR =0.092; P=0.011). The derived logistic model (area under the curve =0.934; 95% confidence interval: 0.887–0.981) showed high sensitivity (76.4%) and specificity (97.3%). In terms of pathological findings, the Ki-67 index and mitotic count correlated strongly with risk level (P<0.05) and may serve as key prognostic markers.
Conclusions
Ultrasound-based tumor size and echogenicity are robust preoperative indicators for gastric GIST risk classification. The proposed model demonstrated excellent predictive performance and may be a practical tool for clinical assessment.
Keywords: Gastrointestinal stromal tumors (GISTs), ultrasonic characteristics, linicopathologic features, biological risk, prediction model
Introduction
Gastrointestinal stromal tumors (GISTs) are the most common gastrointestinal mesenchymal tumors. Originating from Cajal stromal cells, GIST accounts for about 1–3% of all primary gastrointestinal malignancies (1,2) and can occur in any part of the digestive tract, with GISTs of the stomach accounting for 60–70% of cases (3). With the advances in diagnostic technology achieved in recent years, a greater number of GISTs are being incidentally detected during routine examinations. Gastric GISTs are tumors with considerable malignant potential, with their biological behavior varying from benign to malignant (4). The probability of metastasis or tumor-related death in patients with high-risk GISTs is as high as 90%, with 11–47% of these patients reported to have distant organ metastases (5) at the time of initial diagnosis; meanwhile, patients with low-risk GISTs can survive for a long period after resection, with recurrence or metastasis being relatively rare. According to the risk classification criteria for primary GISTs proposed in the “Chinese consensus guidelines for diagnosis and management of gastrointestinal stromal tumor” (2017 edition) (6), the biological risk of GISTs is divided into very-low-risk, low-risk, medium-risk, and high-risk. Very-low-risk and low-risk GISTs are the most common in clinic and can usually be surgically removed or regularly followed up for observation. Meanwhile, medium- and high-risk GISTs often require surgical resection or combined targeted therapy.
Diagnosis of GISTs is based on imaging techniques and is confirmed via histopathological examination. A number of imaging modalities can be used to identify gastric GISTs, including endoscopy, computed tomography (CT), positron emission tomography (PET), and magnetic resonance imaging (MRI) (7,8). Endoscopic ultrasonography (EUS) is an important technique used in the diagnostic examination of GISTs. Research suggests that EUS is effective and highly accurate in the preoperative diagnosis of gastric GISTs (9). However, the patient’s discomfort during the examination limits its application, and determining the origin may be difficult when the mass is exophytic and large. Endoscopic ultrasound-guided fine needle aspiration (EUS-FNA) biopsy is generally an effective method for the preoperative biopsy of GISTs. However, in one prospective study, only 46% of patients were found to be suitable candidates for EUS-FNA biopsy (10). Complications are also an important tradeoff to be considered in the application of EUS-FNA, with approximately 22% of patients experiencing bleeding associated with EUS-FNA procedures, and some developing severe abscesses.
Gastric filling ultrasound (GFUS) is a noninvasive imaging examination that allows the patient to ingest contrast media and thus fill the gastric cavity. This eliminates gastric interference and makes the gastric cavity produce a uniform to medium-to-high echo, thus providing a suitable acoustic window for the clear visualization of the gastric wall and surrounding structures. In this manner, GFUS is able to accurately visualize the five-layer structure of the gastric wall and assess the basic characteristics of the lesion (e.g., size, morphology, and anatomical origin) as well as tumor blood flow signals (11). This method has been widely recommended and applied in the diagnosis of gastric lesions, significantly improved the clinical diagnostic value of gastrointestinal ultrasound (12), and demonstrated high accuracy in the diagnosis of gastric GISTs (13).
The risk assessment system for primary resectable gastric GISTs is primarily based on postoperative pathological analysis, but this cannot determine the malignant potential of the tumor in advance and involves a delay in the guidance of treatment. Therefore, the ability to accurately assess the preoperative risk and screen high-risk cases for the selection of appropriate treatment strategy in patients with gastric GISTs remains a critical need. Few studies have been conducted on the evaluation of gastric GIST risk based on postoperative pathological immunohistochemistry results, clinical data, and GFUS image characteristics. In order to further clarify the risk classification of patients with gastric GISTs and improve the basis for the formulation of individualized treatment plans, the ultrasound characteristics and immunohistochemistry results of 92 patients with gastric GISTs confirmed by pathology and immunohistochemistry were retrospectively analyzed. The principal objectives of this study were to determine the correlation between ultrasound image features and immunohistochemistry results of gastric GISTs of varying risk and to evaluate the clinical application value of ultrasound in predicting the risk grading of gastric GISTs before surgery. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1306/rc).
Methods
General information
A total of 166 patients diagnosed with gastric GISTs by GFUS examination from January 2017 to October 2024 were initially enrolled, among whom 92 diagnosed with gastric GISTs via endoscopic resection or surgical treatment and pathological and immunohistochemical examination were ultimately included, with their ultrasound characteristics and clinicopathological data being retrospectively analyzed. The inclusion criteria were as follows: (I) a lesion originating from the gastric wall; (II) GFUS examination performed before surgery; (III) endoscopic or surgical resection treatment; (IV) pathological confirmation of primary gastric GIST after surgery; (V) complete clinical and pathological data; and (VI) no targeted therapy. Meanwhile, the exclusion criteria were as follows: (I) targeted therapy drugs administered prior to lesion examination; (II) non-gastric GISTs; and (III) incomplete data (Figure 1). This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Ethics Committee of Sichuan Cancer Hospital (approval No. SCCHEC-03-2014-013). The requirement for informed consent was waived due to the retrospective nature of the analysis.
Figure 1.
Flowchart of patient screening for study inclusion. GFUS, gastric filling ultrasound; GIST, gastrointestinal stromal tumor.
Clinical data and study design
The ultrasound images of all patients were acquired from the hospital’s imaging workstation. The clinical data and pathological results of all patients were collected and recorded from the electronic medical record system of the hospital, including mitotic figures and immunohistochemistry-related indicators [CD34, CD117, discovered on GIST-1 (DOG-1), vimentin, S100, smooth muscle actin (SMA), desmin, and Ki-67]. This study adopted the risk classification criteria from “Chinese consensus guidelines for diagnosis and management of gastrointestinal stromal tumor” (2017 edition) (6) as the sole and decisive reference standard. Accordingly, the patients were divided into four risk groups, including very-low-, low-, medium-, and high-risk, with 11, 26, 34, and 21 cases, respectively. To improve statistical efficiency and ensure group balance, the very-low-risk and low-risk groups were combined into a low-risk group, while the medium- and high-risk groups were combined into a high-risk group. This combined grouping served as the primary endpoint of the study, as these merged categories share common characteristics in clinical management and biological behavior. The correlation between the ultrasound characteristics and immunohistochemistry results of gastric GISTs of varying risk status was analyzed, univariate and multivariate analyses were conducted to identify independent predictors, and the feasibility of the preoperative prediction of GIST biological risk classification based on the relevant parameters was also evaluated.
Ultrasound examination
All suspected GIST lesions were examined with a standardized ultrasound protocol that remained consistent throughout the study. The protocol clearly specified pre-examination preparation, patient positioning, scanning planes, equipment settings, and the dose and administration method of contrast agents. The ultrasound diagnostic instruments used in this study were the MyLab Twice (Esaote, Genoa, Italy) and EPIQ 5/7 (Philips, Amsterdam, the Netherlands), both uniformly equipped with abdominal probes with a frequency range of 2.0–6.0 MHz. The standardized scanning parameters included a mechanical index (MI) of 0.6–0.9 and a gain of 60–80 dB. The scanning depth was individually adjusted according to the patient’s body type and lesion location and ranged from 8 to 16 cm. Blinded assessment of all ultrasound images (from different devices) by two physicians (>5 years) was applied to control for potential bias and to confirm that device differences did not systematically affect feature analysis.
Before the examination, the patient fasted for at least 8 hours and was instructed to take 500 mL of oral gastrointestinal ultrasound contrast agent. The stomach was systematically and comprehensively examined in different positions, including the sitting, supine, and left and right positions.
The following lesion information was observed and recorded: (I) location (fundus, body, incisura angularis, or antrum); (II) maximum diameter (≤2, >2–≤5, >5–≤10, or >10 cm); (III) growth mode, (intraluminal type, extraluminal type, or intraluminal–extraluminal type); (IV) morphology (regular or irregular); (V) internal echo (homogeneous or heterogeneous); (VI) cystic change (yes or no), (VII) calcification (yes or no); (VIII) ulcer on the surface (yes or no); and (IX) Adler blood flow grade (grade 0= no blood flow signal in the lesion; grade 1= small amount of blood flow in the lesion, with 1 or 2 punctate or thin, short rod-shaped blood vessels being visible; grade 2= moderate blood flow in the lesion, with 3 or 4 punctate or 1 elongated blood vessel visible and the length of the blood vessel being close to or exceeding the radius of the lesion; grade 3= abundant blood flow in the lesion, with more than 5 punctate or 2 elongated blood vessels visible).
Statistical methods
Data are expressed as the mean ± standard deviation (SD). The independent samples t-test or analysis of variance was used to compare groups in terms of normally distributed data, while nonparametric tests were used for comparisons with data with a nonnormal distribution. Count data are expressed as the number of cases with percentages, and the Chi-squared test was used for comparing these data between groups; data with an expected value of <5 in cells were analyzed with the Fisher’s exact test. The proportion of missing data was minimal (<2%), and missing data were processed via column-wise deletion. Multicollinearity was assessed through use of the variance inflation factor. Univariate logistic regression was used to evaluate the relationship between each feature and the risk grade of gastric GISTs. Multivariate analysis was performed via binary logistic regression analysis to identify significant independent predictors. Interobserver consistency was assessed with the Cohen’s kappa (κ) value. Interobserver agreement was determined on the basis of the Fleiss κ value as follows: marginal, 0–0.40; medium, 0.41–0.60; good, 0.61–0.80; and excellent, 0.81–1.00.
Areas under the receiver operating characteristic (ROC) curve (AUCs) were calculated to evaluate the predictive performance of the model. The calibration of the final model was evaluated with the Hosmer-Lemeshow (HL) test. Internally and externally validated, bias-corrected AUCs and calibration curves were constructed via bootstrap resampling. Decision-curve analysis (DCA) was employed to evaluate clinical performance. The associations between ultrasound features and the risk stratification of gastric GISTs were quantified as odds ratios (ORs) with 95% confidence intervals (CIs). P<0.05 indicated a statistically significant difference. All statistical analyses were performed with SPSS v.19.0 (IBM Corp., Armonk, NY, USA; RRID: SCR_002865) and R software v.4.5.1 (The R Foundation for Statistical Computing, Vienna, Austria).
Results
Baseline data
A total of 92 patients were included in this study. According to the postoperative pathological examination, there were 11 very-low-risk cases, 26 low-risk cases, 34 medium-risk cases, and 21 high-risk cases; thus, there were 37 cases in the low-risk group and 55 cases in the high-risk group. The main clinical manifestations of the patients included abdominal pain, abdominal distention, melena, hematemesis, and fatigue, occurring in 31.5% (29/92), 14.1% (13/92), 19.6% (18/92), 6.5% (6/92), and 4.3% (4/92) of the patients, respectively. Univariate analysis showed that age and gender had no significant impact on the risk grading of postoperative pathological outcomes (P>0.05). However, the proportion of patients with melena symptoms in the high-risk group was significantly higher than that in the low-risk group, suggesting that melena symptoms have an important reference value for the risk classification of gastric GISTs (P<0.05). The results of the analysis are shown in Table 1.
Table 1. Comparison of clinical data between the low-risk and high-risk gastric GIST groups.
| Variable | Biological risk classification | χ2 | P | |
|---|---|---|---|---|
| Low-risk group (n=37) | High-risk group (n=55) | |||
| Age, years | 0.955 | 0.328 | ||
| <60 | 15 (40.5) | 28 (50.9) | ||
| ≥60 | 22 (59.5) | 27 (49.1) | ||
| Gender | 1.978 | 0.16 | ||
| Male | 16 (43.2) | 32 (58.2) | ||
| Female | 21 (56.8) | 23 (31.8) | ||
| Symptom | ||||
| Abdominal pain | 12 (32.4) | 17 (30.9) | 0.024 | 0.877 |
| Abdominal distention | 6 (16.2) | 7 (12.7) | 0.222 | 0.638 |
| Melena | 2 (5.4) | 16 (29.1) | 7.885 | 0.005 |
| Hematemesis† | 0 (0.0) | 6 (10.9) | 0.078 | |
| Fatigue† | 1 (2.7) | 3 (5.5) | 0.646 | |
| Asymptomatic | 17 (45.9) | 15 (27.3) | 3.4 | 0.065 |
Data are presented as number (%). †, Fisher’s exact test. GIST, gastrointestinal stromal tumor.
Analysis of ultrasound features
A comparison of the ultrasound features between the low-risk and high-risk groups of gastric GISTs is presented in Table 2. In the low-risk group, the tumors were relatively small, with a maximum diameter ranging from 2 to 5 cm. Internally, 62.2% exhibited homogeneous echogenicity, while 16.2% showed calcification, and cystic change was rare (8.1%). Additionally, 91.9% of the tumors had a regular morphology, and 97.3% displayed grade 0–1 blood flow (Figure 2A,2B). In contrast, the high-risk group featured larger tumors, with 74.5% having a maximum diameter greater than 5 cm. The internal echogenicity was predominantly heterogeneous (96.4%), and calcification (36.4%) and cystic changes (43.6%) were more frequently observed (Figure 2C,2D).
Table 2. Univariate analysis of ultrasound characteristics in patients with gastric GISTs stratified by risk.
| Variable | Biological risk classification | χ2 | P | |
|---|---|---|---|---|
| Low-risk group (n=37) | High-risk group (n=55) | |||
| Maximum diameter (cm) | 46.015 | <0.001 | ||
| ≤2 | 11 (29.7) | 1 (1.8) | ||
| >2–≤5 | 25 (67.6) | 13 (23.6) | ||
| >5–≤10 | 1 (2.7) | 35 (63.6) | ||
| >10 | 0 (0.0) | 6 (10.9) | ||
| Tumor site† | 2.279 | 0.545 | ||
| Fundus | 15 (40.5) | 25 (45.5) | ||
| Body | 14 (37.8) | 24 (43.6) | ||
| Incisura angularis | 1 (2.7) | 1 (1.8) | ||
| Antrum | 7 (18.9) | 5 (9.1) | ||
| Growth pattern | 5.973 | 0.05 | ||
| Intraluminal | 26 (70.3) | 25 (45.5) | ||
| Intraluminal-extraluminal | 8 (21.6) | 18 (32.7) | ||
| Extraluminal | 3 (8.1) | 12 (21.8) | ||
| Morphology | 18.137 | <0.001 | ||
| Regular | 34 (91.9) | 27 (49.1) | ||
| Irregular | 3 (8.1) | 28 (50.9) | ||
| Echo | 38.285 | <0.001 | ||
| Homogeneous | 23 (62.2) | 2 (3.6) | ||
| Heterogeneous | 14 (37.8) | 53 (96.4) | ||
| Ulcer | 3 (8.1) | 12 (21.8) | 3.047 | 0.081 |
| Calcification | 6 (16.2) | 20 (36.4) | 4.429 | 0.035 |
| Cystic change | 3 (8.1) | 24 (43.6) | 13.466 | <0.001 |
| Adler blood flow grade† | 0.137 | |||
| Grade 0–1 | 36 (97.3) | 48 (87.3) | ||
| Grade 2–3 | 1 (2.7) | 7 (12.7) | ||
Data are presented as number (%). Percentages may not sum to 100 due to rounding. †, Fisher’s exact test. GIST, gastrointestinal stromal tumor.
Figure 2.
Ultrasound features of gastric GISTs with different risk levels. (A) A tumor classified as very-low-risk measuring 1.5 cm × 1.2 cm × 1.0 cm. It was protruding into the gastric cavity and was small, oval, homogeneous, and hypoechoic, with regular morphology. (B) A tumor classified as low-risk measuring 2.7 cm × 3.8 cm × 2.2 cm. It was protruding into the gastric cavity and was small, round, homogeneous, and hypoechoic, with clear borders. (C) A tumor classified as medium-risk measuring 4.4 cm × 3.7 cm × 3.9 cm. It was protruding into the gastric cavity and was large, round, heterogeneous, and hypoechoic, with clear boundaries and internal cystic changes. (D) A tumor classified as high-risk measuring 11.5 cm × 8.6 cm × 7.7 cm. It was large and irregular in shape and exhibited heterogeneous echogenicity and internal cystic changes. GIST, gastrointestinal stromal tumor.
Interobserver agreement for ultrasound features of gastric GISTs
The κ values for morphology (κ =0.794; 95% CI: 0.667–0.921), echogenicity (κ =0.786; 95% CI: 0.645–0.927), blood flow grade (κ =0.708; 95% CI: 0.573–0.843), calcification (κ =0.762; 95% CI: 0.615–0.909), and cystic change (κ =0.743; 95% CI: 0.594–0.892) indicated excellent interobserver agreement.
Univariate analysis
Univariate analysis indicates no significant difference in the risk of tumor location, growth mode, blood flow grade, or surface ulcer presence between the low-risk and high-risk groups (P>0.05). However, there were significant differences in the maximum diameter, morphology, echogenicity, calcification, and cystic changes between the two groups (P<0.05), as shown in Table 2.
Multivariate logistic regression analysis of gastric GISTs of different risk
Multivariate binary logistic regression analysis was conducted on variables demonstrating statistical significance in the univariate analysis. The results identified lesion maximum diameter >5 cm (P=0.002) and echo uniformity (P=0.011) as independent predictive factors for differentiating gastric GISTs risk levels based on ultrasonic characteristics. Specifically, the larger the tumor volume and greater the heterogeneity in the internal echo, the greater the likelihood of gastric GISTs being classified as high-risk. In contrast, tumor morphology (P=0.095), cystic changes (P=0.773), and calcification (P=0.881) were not significantly correlated with risk grading, as detailed in Table 3.
Table 3. Multivariate binary logistic regression analysis of patients with gastric GISTs with ultrasound characteristics and different risk levels.
| Variable | β | SE | Wald χ2 | P | OR (95% CI) |
|---|---|---|---|---|---|
| Maximum diameter | −3.573 | 1.158 | 9.516 | 0.002* | 0.028 (0.003–0.272) |
| Morphology | −1.531 | 0.916 | 2.794 | 0.095 | 0.216 (0.036–1.302) |
| Echo | −2.386 | 0.944 | 6.393 | 0.011* | 0.092 (0.014–0.585) |
| Calcification | 0.126 | 0.84 | 0.023 | 0.881 | 1.134 (0.219–5.885) |
| Cystic change | 0.296 | 1.026 | 0.083 | 0.773 | 1.344 (0.180–10.035) |
| Constant | 4.416 | 1.397 | 9.997 | 0.002 | 82.784 (–) |
*, P<0.05 indicates that the predictive variable is independently associated with the risk of gastric GISTs. CI, confidence interval; GIST, gastrointestinal stromal tumor; OR, odds ratio; SE, standard error; β, regression coefficient.
Predictive performance of the logistic regression model
Based on the above analysis, the following logistic regression equation was established: logistic (P) =4.416–3.573 × maximum diameter − 2.386 × echo. The goodness-of-fit of the model was evaluated via the HL test, and the results (χ2=2.483 and P=0.963) indicated good consistency between the model’s predicted values and the observed values. With a predicted probability of P=0.50 serving as the threshold, the model’s sensitivity (SE) for distinguishing high-risk from low-risk gastric GISTs was 76.4%, and its specificity (SP) was 97.3%. Furthermore, as shown in Figure 3, AUC of the model reached 0.934 (95% CI: 0.887–0.981; P<0.001), further confirming the excellent discriminative ability of the model.
Figure 3.
ROC curve of the logistic regression model for predicting the risk grade of gastric GISTs. AUC, area under the curve; CI, confidence interval; GIST, gastrointestinal stromal tumor; ROC, receiver operating characteristic.
The model’s calibration was evaluated via the HL goodness-of-fit test and calibration curve (Bootstrap method, n=1,000) (Figure 4). The results (χ2 =4.65; P=0.75; mean absolute error =0.019; concordance index =0.92; 95% CI: 0.891–0.949) indicated that the logistic regression model approached ideal predictive performance.
Figure 4.
Calibration curve of the risk prediction model for gastric GISTs. The 45° diagonal line represents perfect agreement between the nomogram-predicted probability (X-axis) and the observed value (Y-axis). GIST, gastrointestinal stromal tumor.
Nomograms for predicting the risk of gastric GISTs
Figure 5 displays the nomogram for predicting gastric GIST risk, constructed using R software. This tool assigns a specific scoring scale to each predictor. In practice, points are allocated based on an individual patient’s characteristics; these points are then summed to obtain a total score, which corresponds to the patient’s probability of risk stratification on the total points axis. Analysis revealed that maximum tumor diameter was the most influential predictor in the model.
Figure 5.
Nomogram for predicting the risk of gastric GISTs. GIST, gastrointestinal stromal tumor.
Analysis of immunohistochemistry results
All patients in this study underwent postoperative pathological and immunohistochemical testing. The expression of DOG-1 was positive in all patients, and so it was not included in the statistical analysis. In terms of immunohistochemistry, 86 cases (93.5%) were positive for CD34, 91 (98.9%) positive for CD117, 64 (69.6%) positive for succinate dehydrogenase B (SDHB), 28 (30.4%) positive for SMA, 12 (13.0%) positive for vimentin, and 8 (9.0%) positive for desmin. Univariate analysis showed that the positive expressions of CD34, CD117, SMA, SDHB, vimentin, desmin, and S100 were not significantly correlated with the postoperative risk grade of gastric GISTs (P>0.05). However, the Ki-67 index and the number of pathological mitotic figures were significantly correlated with the risk level (P<0.05). Further multivariate binary logistic regression analysis showed that the mitotic count was an independent influencing factor for distinguishing gastric GISTs of different risk (P<0.05). The detailed results from the analysis are provided in Table 4.
Table 4. Univariate analysis of the immunohistochemical indicators of gastric GISTs at different risk levels.
| Variable | Biological risk classification | χ2 | P | |
|---|---|---|---|---|
| Low-risk group (n=37) | High-risk group (n=55) | |||
| CD34† | 0.078 | |||
| + | 37 (100.0) | 49 (89.1) | ||
| − | 0 (0.0) | 6 (10.9) | ||
| CD117† | >0.99 | |||
| + | 37 (100.0) | 54 (98.2) | ||
| − | 0 (0.0) | 1 (1.8) | ||
| SDHB | 0.117 | 0.733 | ||
| + | 25 (67.6) | 39 (70.9) | ||
| − | 12 (32.4) | 16 (29.1) | ||
| SMA | 0.803 | 0.37 | ||
| + | 12 (32.4) | 16 (29.1) | ||
| − | 25 (67.6) | 39 (70.9) | ||
| Desmin† | >0.99 | |||
| + | 3 (8.1) | 5 (9.1) | ||
| − | 34 (91.9) | 50 (90.9) | ||
| Vimentin† | 0.535 | |||
| + | 6 (10.9) | 6 (10.9) | ||
| − | 31 (89.1) | 49 (89.1) | ||
| S100† | 0.646 | |||
| + | 1 (2.7) | 3 (5.5) | ||
| − | 36 (97.3) | 52 (94.5) | ||
| Ki-67 | 10.416 | 0.005 | ||
| ≤5% | 30 (81.1) | 29 (52.7) | ||
| 6–10% | 7 (18.9) | 16 (29.1) | ||
| >10% | 0 (0.0) | 10 (18.2) | ||
| Mitotic figures | 23.958 | <0.001 | ||
| ≤5/50 HPF | 34 (91.9) | 23 (41.8) | ||
| 6–10/50 HPF | 3 (8.1) | 22 (40.0) | ||
| >10/50 HPF | 0 (0.0) | 10 (18.2) | ||
Data are presented as number (%). †, Fisher’s exact test. DOG-1 was positive in all patients and was therefore excluded from statistical analysis. DOG-1, discovered on GIST-1; GIST, gastrointestinal stromal tumor; HPF, high-power field; SDHB, succinate dehydrogenase B; SMA, smooth muscle actin.
Clinical efficacy of the logistic regression model
DCA was used to evaluate the clinical predictive efficacy of the logistic regression model (Figure 6). The model demonstrated the greatest net benefit for predicting the risk of gastric GISTs at a risk threshold probability of 0.01–0.97. Its predictive ability was superior to the extreme reference curves, indicating significant potential for clinical application.
Figure 6.
DCA was used to evaluate the clinical predictive efficacy of the logistic regression model. DCA, decision-curve analysis.
Discussion
Gastric GISTs, the most common pathological type of gastric submucosal tumor, occur most often in older adults (14) and are more common in the gastric corpus and fundus regions (15). Most gastric GISTs lack characteristic clinical manifestations and are typically characterized by nonspecific symptoms such as abdominal pain and bloating, although some cases may be accompanied by symptoms such as bleeding (16). In our study, melena, a first symptom potentially caused by tumor surface ulceration or bleeding due to tumor invasion of blood vessels, demonstrated an important reference value for the risk classification of gastric GISTs. Meanwhile, no significant correlation was found between patient sex or age and pathological risk grade (P>0.05), a finding consistent with previous work (17).
GFUS offers the unique advantages of being radiation-free, convenient, and cost-effective in the diagnosis of gastric GISTs (18) and is suitable for frequent local efficacy monitoring (such as in measuring tumor size and observing internal blood flow changes). GFUS can clearly show the hierarchical structure of the gastric wall and accurately assess the location, size, morphology, and relationship between the tumor and the surrounding tissues, providing an important basis for clinical diagnosis. GFUS further allows for the real-time, dynamic assessment of tumor mobility; meanwhile, color Doppler flow imaging can conveniently visualize the internal blood flow distribution of the tumor with high SE. Gastric GISTs on GFUS primarily originate from the muscularis propria, with small tumors typically appearing as uniformly hypoechoic, while larger tumors appear as moderately hypoechoic and heterogeneous, which may include patchy hyperechoic areas, cystic formations, or irregular anechoic regions (19). In this study, we investigated the correlation between ultrasound image characteristics and the risk of gastric GISTs, and the observed indicators included tumor location, growth mode, maximum diameter, morphology, echo uniformity, cystic changes, calcification, surface ulceration, and blood flow grade. Univariate analysis indicated that the maximum diameter, morphology, homogeneity of echo, cystic changes, and calcification were statistically significant in determining the postoperative risk of patients with gastric GISTs (P<0.05). Gastric GISTs exhibit variable internal echoes depending on their size. Generally, smaller GISTs tend to be more homogeneously hypoechoic. As the tumor size increases, the internal echogenicity often becomes heterogeneous, and cystic changes may even occur (20). Most low-risk gastric GISTs have a more regular shape, appearing round, oval, or oval-like, whereas tumors designated as medium or high-risk are often irregular in shape.
The results of multivariate logistic regression analysis identified the maximum diameter of the tumor and echo uniformity as independent influencing factors for risk classification, which is essentially consistent with the results of a few previous studies (17,21). Moreover, the larger the maximum diameter of the tumor and the less uniform the internal echo, the greater is the risk of high-risk gastric GISTs and postoperative recurrence (22), and patients with these should be further evaluated after the corresponding surgical treatment to determine the need for adjuvant therapy.
In a study by Chen et al. (23), medium- and high-risk gastric GISTs tended to appear with calcification and cystic changes on ultrasound. However, Xue et al. and Kim et al. (24,25) found no association between these features and malignant potential. In our study, although these features appeared to be significant in the univariate analysis, they were not independent predictors in the multivariate analysis. More objective quantitative criteria may help refine their significance. Tumor morphology has also been examined in this capacity (23). In our univariate analysis, irregular morphology correlated with high-risk GISTs (P<0.05), but it was not an independent predictor in the multivariate analysis, conflicting with some earlier reports (17,26). This discrepancy may stem from our limited sample size, suggesting that investigation with larger cohorts is needed.
We also constructed an efficient and easy-to-use prediction model, incorporating maximum diameter and echo uniformity, for predicting high-risk gastric GISTs. The SE and SP of the model were 76.4% and 97.3%, respectively, and the AUC was 0.934 (95% CI: 0.887–0.981; P<0.001), indicating that the model had good predictive power. In the regression model, a tumor diameter >5 cm (OR =0.028) was strongly predictive of high-risk GISTs, with high SE and SP. Similarly, heterogeneous echogenicity (OR =0.092) was significantly associated with a higher postoperative risk classification. Importantly, this prediction model proved capable of assessing the biological risk of gastric GISTs before surgery and thus may overcome the limitations of existing mainstream guidelines, which only recommend evaluation through mitotic count after surgery. Through our model, clinicians can preliminarily estimate the risk of gastric GISTs and then adopt appropriate treatment strategies. The two indicators used in our prediction model are very readily available in clinical practice, do not require tedious examinations, and do not increase the financial burden on patients. However, we acknowledge that our study is a retrospective validation of the 2017 Chinese consensus criteria, and any changes to standard practice need to be substantiated by data from prospective clinical trials. In this context, we hope our work can serve as a preliminary step toward the development of a decision-support tool and can inform future research and guideline updates.
Pathology is the gold standard for diagnosing gastric GISTs, with CD117 and DOG-1 levels being the characteristic immunohistochemical markers (27). It has been reported (28) that most gastric GISTs have a positive rate of 95% for CD117 and 98% for DOG-1, while in our study, these rates were 98.9% and 100%, respectively. Moreover, the proportion of mitotic figures exceeding 5/50 high-power fields (HPFs) was significantly higher in the high-risk group (58.2%) than in the low-risk group (8.1%; P<0.001). This finding is consistent with the risk classification criteria established by the 2017 consensus. In one study (29), the immunohistochemical indicators of 225 patients with GISTs were analyzed, and the results showed that CD117, CD34, DOG1, and Ki-67 were not correlated with the prognosis of patients with gastric GISTs. In our study, positivity of CD117, CD34, SMA, SDHB, vimentin, desmin, and S100 was not significantly associated with the risk level of GISTs after surgery (P>0.05), while Ki-67 index and mitotic count were (P<0.05). Ki-67 expression is closely related to tumor cell proliferation and growth (30), and the Ki-67 labeling index determined by immunohistochemical analysis can indicate the prognosis of tumor patients. Wang et al. (31) reported that patients with gastric GISTs and a high Ki-67 index may have a worse prognosis. In our study, univariate analysis identified the Ki-67 index as a high-risk factor (P<0.001), but it was not significant in multivariate regression analysis, indicating that while the Ki-67 index can be indicative of tumor malignancy and poor prognosis, it is not an independent predictor for evaluating the risk of invasion. It is worth noting that genetic testing was performed on 12 patients in our study, among whom 10 (10.9%) exhibited mutations in exon 11 of the KIT gene. Although the mutation status of KIT/PDGFRA genes was not directly correlated with ultrasound characteristics, this type of mutation may indirectly alter ultrasound manifestations by influencing tumor growth patterns. Future research should include larger sample sizes and integrate molecular characteristics such as gene mutation status to further determine its impact on the risk stratification and prognosis of patients with gastric GISTs.
This study involved several limitations that should be acknowledged. First, we employed a single-center, retrospective design, and the case selection might have been subject to potential bias. Future rigorous prospective studies are needed to validate the clinical utility of our prediction model. Second, the assessment capability of GFUS for small lesions is limited, which may affect the detection of early-stage disease. Third, variability is a common challenge in ultrasound examinations. To mitigate this limitation, future studies could adopt standardized systems such as the stomach ultrasound reporting and data system (Su-RADS), which translates subjective imaging features into objective categories to improve consistency (32,33). Looking forward, integrating multimodal imaging techniques (e.g., PET-CT) could help construct a more comprehensive risk prediction model and refine the imaging-pathology correlation system, thereby enhancing the accuracy of preoperative evaluation.
Conclusions
Melena as an initial symptom can help indicate the risk level of GISTs. Tumor size and echogenicity are independent predictors for risk stratification and have potential clinical value in assessing tumor behavior. In addition, the Ki-67 index and mitotic count can help provide a comprehensive evaluation of the disease and thus support the formulation of personalized treatment strategies.
Supplementary
The article’s supplementary files as
Acknowledgments
None.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Sichuan Cancer Hospital (No. SCCHEC-03-2014-013) and individual consent for this retrospective analysis was waived.
Footnotes
Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1306/rc
Funding: None.
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1306/coif). The authors have no conflicts of interest to declare.
Data Sharing Statement
Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1306/dss
References
- 1.Ren C, Wang S, Zhang S. Development and validation of a nomogram based on CT images and 3D texture analysis for preoperative prediction of the malignant potential in gastrointestinal stromal tumors. Cancer Imaging 2020;20:5. 10.1186/s40644-019-0284-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Liu JX, Zhou PY, Tang Z, Yuan W, Shen SS, Ren L, Xing Z, Fang Y, Gao XD, Xue AW, Shen KT, Hou YY. Comparison of prognostic prediction models for rectal gastrointestinal stromal tumor. Aging (Albany NY) 2020;12:11416-30. 10.18632/aging.103204 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Theiss L, Contreras CM. Gastrointestinal Stromal Tumors of the Stomach and Esophagus. Surg Clin North Am 2019;99:543-53. 10.1016/j.suc.2019.02.012 [DOI] [PubMed] [Google Scholar]
- 4.Nishida T, Yoshinaga S, Takahashi T, Naito Y. Recent Progress and Challenges in the Diagnosis and Treatment of Gastrointestinal Stromal Tumors. Cancers (Basel) 2021;13:3158. 10.3390/cancers13133158 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Nishimura J, Nakajima K, Omori T, Takahashi T, Nishitani A, Ito T, Nishida T. Surgical strategy for gastric gastrointestinal stromal tumors: laparoscopic vs. open resection. Surg Endosc 2007;21:875-8. 10.1007/s00464-006-9065-z [DOI] [PubMed] [Google Scholar]
- 6.Li J, Ye Y, Wang J, Zhang B, Qin S, Shi Y, He Y, Liang X, Liu X, Zhou Y, Wu X, Zhang X, Wang M, Gao Z, Lin T, Cao H, Shen L, Chinese Society Of Clinical Oncology Csco Expert Committee On Gastrointestinal Stromal Tumor . Chinese consensus guidelines for diagnosis and management of gastrointestinal stromal tumor. Chin J Cancer Res 2017;29:281-93. 10.21147/j.issn.1000-9604.2017.04.01 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Mastalier Manolescu BS, Popp CG, Popescu V, Andraş D, Zurac SA, Berceanu C, Petca AT. Novel perspectives on gastrointestinal stromal tumors (GISTs). Rom J Morphol Embryol 2017;58:339-50. [PubMed] [Google Scholar]
- 8.Li LM, Feng LY, Chen XH, Liang P, Li J, Gao JB. Gastric heterotopic pancreas and stromal tumors smaller than 3 cm in diameter: clinical and computed tomography findings. Cancer Imaging 2018;18:26. 10.1186/s40644-018-0161-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Gu M, Ghafari S, Nguyen PT, Lin F. Cytologic diagnosis of gastrointestinal stromal tumors of the stomach by endoscopic ultrasound-guided fine-needle aspiration biopsy: cytomorphologic and immunohistochemical study of 12 cases. Diagn Cytopathol 2001;25:343-50. 10.1002/dc.10003 [DOI] [PubMed] [Google Scholar]
- 10.Eckardt AJ, Adler A, Gomes EM, Jenssen C, Siebert C, Gottschalk U, Koch M, Röcken C, Rösch T. Endosonographic large-bore biopsy of gastric subepithelial tumors: a prospective multicenter study. Eur J Gastroenterol Hepatol 2012;24:1135-44. 10.1097/MEG.0b013e328356eae2 [DOI] [PubMed] [Google Scholar]
- 11.Shi H, Yu XH, Guo XZ, Guo Y, Zhang H, Qian B, Wei ZR, Li L, Wang XC, Kong ZX. Double contrast-enhanced two-dimensional and three-dimensional ultrasonography for evaluation of gastric lesions. World J Gastroenterol 2012;18:4136-44. 10.3748/wjg.v18.i31.4136 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Xu R, Wang Y, Yin X. Application valve of oral contrast-enhanced ultrasound in gastric disease screening in the community. Chinese J ultrasound Med 2021;37:153-6. [Google Scholar]
- 13.Gao M, Miao L, Ge H, Zhao B, Xue H. The correlation analysis between CEUS characteristic and benign and malignant potential of gastric stromal tumors. Chinese J ultrasound Med 2017;33:184-6. [Google Scholar]
- 14.Liang X, Zhou Q, Ke X, Han L, Zhou J. Comparison on CT signs of gastric stromal tumor and inflammatory fibrous polyp. Chin J Med Imaging Technol 2019;35:376-80. [Google Scholar]
- 15.Kim GH. Systematic Endoscopic Approach for Diagnosing Gastric Subepithelial Tumors. Gut Liver 2022;16:19-27. 10.5009/gnl20296 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Blay JY, Kang YK, Nishida T, von Mehren M. Gastrointestinal stromal tumours. Nat Rev Dis Primers 2021;7:22. 10.1038/s41572-021-00254-5 [DOI] [PubMed] [Google Scholar]
- 17.Tang X, Guo J, Qian Q, Zhang X, Chen Z, Xue E, Lin L. Multivariate logistic regression analysis of biorisk prediction of gastric gastrointestinal stromal tumor by transabdominal ultrasonography. Oncoradiology 2022;31:607-11. [Google Scholar]
- 18.Ultrasound Professional Committee of China Medicine Education Association . Chinese Expert Consensus on Gastric Contrast Ultrasound for Scanning Technique and Imaging Acquisition. J Cancer Control Treat 2020;33:817-27. [Google Scholar]
- 19.Mullady DK, Tan BR. A multidisciplinary approach to the diagnosis and treatment of gastrointestinal stromal tumor. J Clin Gastroenterol 2013;47:578-85. 10.1097/MCG.0b013e3182936c87 [DOI] [PubMed] [Google Scholar]
- 20.Takahashi K, Nihei T, Aoki Y, Konno N, Nakagawa M, Munakata A, Okawara K, Ohtani H, Kashimura H. Gastric gastrointestinal stromal tumor with predominant cystic formation diagnosed by endoscopic ultrasound-fine needle aspiration. Clin J Gastroenterol 2020;13:359-64. 10.1007/s12328-019-01058-7 [DOI] [PubMed] [Google Scholar]
- 21.Wang L, Shi R, Dong N, Feng J, Huang X. Analyzing the correlation between endoscopic ultrasound features and immunohistochemical results of gastric stromal tumors with different risk levels. Modern Interventional Diagnosis and Treatment in Gastroenterology 2024;29:281-6. [Google Scholar]
- 22.Charville GW, Longacre TA. Surgical Pathology of Gastrointestinal Stromal Tumors: Practical Implications of Morphologic and Molecular Heterogeneity for Precision Medicine. Adv Anat Pathol 2017;24:336-53. 10.1097/PAP.0000000000000166 [DOI] [PubMed] [Google Scholar]
- 23.Chen T, Xu L, Dong X, Li Y, Yu J, Xiong W, Li G. The roles of CT and EUS in the preoperative evaluation of gastric gastrointestinal stromal tumors larger than 2 cm. Eur Radiol 2019;29:2481-9. 10.1007/s00330-018-5945-6 [DOI] [PubMed] [Google Scholar]
- 24.Xue A, Yuan W, Gao X, Fang Y, Shu P, Xu C, Li H, Xu Y, Song Q, Hou Y, Shen K. Gastrointestinal stromal tumors (GISTs) with remarkable cystic change: a specific subtype of GISTs with relatively indolent behaviors and favorable prognoses. J Cancer Res Clin Oncol 2019;145:1559-68. 10.1007/s00432-019-02853-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Kim MN, Kang SJ, Kim SG, Im JP, Kim JS, Jung HC, Song IS. Prediction of risk of malignancy of gastrointestinal stromal tumors by endoscopic ultrasonography. Gut Liver 2013;7:642-7. 10.5009/gnl.2013.7.6.642 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Yang Z. Preoperative prediction model study for high-risk gastrointestinal stromal tumors. Master’s thesis: Air Force Medical University; 2020. doi: 10.27002/d.cnki.gsjyu.2020.000210. [DOI] [Google Scholar]
- 27.Unk M, Jezeršek Novaković B, Novaković S. Molecular Mechanisms of Gastrointestinal Stromal Tumors and Their Impact on Systemic Therapy Decision. Cancers (Basel) 2023;15:1498. 10.3390/cancers15051498 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Schaefer IM, DeMatteo RP, Serrano C. The GIST of Advances in Treatment of Advanced Gastrointestinal Stromal Tumor. Am Soc Clin Oncol Educ Book 2022;42:1-15. 10.1200/EDBK_351231 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Zhang J, Cao X, Liu L, Jiao W, Guo C. Correlation between Ki-67proliferation index and prognosis and risk classification of gastric stromal tumor. Henan Medical Research 2022;31:2527-31. [Google Scholar]
- 30.Lin C, Sui C, Tao T, Guan W, Zhang H, Tao L, Wang M, Wang F. Prognostic analysis of 2-5 cm diameter gastric stromal tumors with exogenous or endogenous growth. World J Surg Oncol 2023;21:139. 10.1186/s12957-023-03006-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Wang JP, Liu L, Li ZA, Wang Q, Wang XY, Lin J. Ki-67 labelling index is related to the risk classification and prognosis of gastrointestinal stromal tumours: a retrospective study. Gastroenterol Hepatol 2021;44:103-14. 10.1016/j.gastrohep.2020.05.022 [DOI] [PubMed] [Google Scholar]
- 32.Gastrointestinal Ultrasound Section, Ultrasound Medicine Specialized Committee of National Health Commission Capacity Building and Continuing Education ; National Major Chronic Disease Prevention and Control Innovation and Integration Pilot Project Management Committee; Gastrointestinal Ultrasound Collaborative Group, Institute of Ultrasound Medicine, Tongji University; Gastrointestinal Ultrasound Section, Ultrasound Specialized Committee, China Medical Education Association; Gastrointestinal Ultrasound Expert Committee, Chinese Ultrasound Medicine Training Project; Abdominal Specialized Committee, Chinese Association of Ultrasound in Medicine and Engineering; Ultrasound Medicine Branch of Sino-Japan Medical Science and Technology Exchange Association; Gastrointestinal Ultrasound Branch of Beijing Association of Holistic Integrative Medicine; Ultrasound Medicine Branch of China National Health Association; Chinese Anti-Cancer Association Science Popularization Committee; Ultrasound Medicine Branch of Cross-Straits Medicine Exchange Association; Engineering Research Center of Digestive Endoscopy Minimally Invasive Diagnostic and Therapeutic Technologies, Ministry of Education. Chinese Expert Consensus on the Clinical Application of Ultrasound Screening for Gastric Cancer(2025 Edition). Zhongguo Yi Xue Ke Xue Yuan Xue Bao 2025;47:679-701. [DOI] [PubMed] [Google Scholar]
- 33.Liu Z, Ren W, Guo J, Zhao Y, Sun S, Li Y, Liu Z. Preliminary opinion on assessment categories of stomach ultrasound report and data system (Su-RADS). Gastric Cancer 2018;21:879-88. 10.1007/s10120-018-0798-x [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
The article’s supplementary files as
Data Availability Statement
Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1306/dss