Diagnostic challenges of inflammatory submucosal tumor-like lesions: a multicenter propensity score-matching analysis

Abstract

Background

Gastrointestinal submucosal tumors (SMTs) are frequently encountered in clinical practice, with diverse etiologies and overlapping endoscopic features. Inflammatory SMT-like lesions (I-SMTLs), though rare and benign, mimic SMTs endoscopically but are fundamentally inflammatory. Limited research exists on their specific diagnostic characteristics.

Methods

This multicenter retrospective study analyzed 59 I-SMTLs among 1,226 pathologically confirmed SMTs from five hospitals in China. The diagnostic efficacy of endoscopic ultrasound (EUS) for I-SMTLs was assessed, and comparison with non-inflammatory SMT-like lesions (non-I-SMTLs) was made using propensity score matching (PSM) balanced for sex, age, lesion location and size. Correlation between endoscopic and pathological characteristics were quantified using Cramer’s V coefficient.

Results

I-SMTLs accounted for 4.81% (59/1,226) of SMTs, with 98.31% (58/59) measuring under 2 cm. EUS demonstrated low sensitivity (3.39-4.00%) and a high misdiagnosis rate (96.61%) for I-SMTLs, with over 55% misdiagnosed as potentially malignant and 15% remaining inconclusive. In the PSM cohort (59 I-SMTLs, 118 non-I-SMTLs), I-SMTLs were more frequently located in submucosa (71.19% vs. 42.37%), exhibited mixed-echoic category (18.64% vs. 1.69%), heterogeneous echogenicity (62.71% vs. 21.19%) and unclear boundaries (42.37% vs. 13.56%). Correlation analysis showed cases with smooth muscle hyperplasia in pathology were more likely to arise from the muscularis propria or muscularis mucosae (Cramer’s V = 0.490), and cases with fibrous tissue alterations were more likely to exhibit mixed or hyperechoic patterns (Cramer’s V = 0.545) and heterogeneous echotexture (Cramer’s V = 0.336) on EUS.

Conclusions

I-SMTLs pose significant diagnostic challenges. Distinguishing EUS features, including submucosal location, mixed-echoic patterns, heterogeneous textures, and unclear boundaries, may improve diagnostic accuracy and facilitate appropriate clinical decision-making.

Background

Gastrointestinal submucosal tumors (SMTs) are a diverse group of lesions located beneath the mucosal layer of the gastrointestinal tract, often encountered incidentally by endoscopy [1]. The incidence of SMTs is gradually rising, with a reported prevalence of approximately 3.6% [1, 2]. The diverse etiology of SMTs encompasses malignant, potentially malignant, and benign conditions. Malignant or potentially malignant lesions, such as gastrointestinal stromal tumors (GISTs) and neuroendocrine tumors (NETs), comprising approximately 15% [3]. Inflammatory SMT-like lesions (I-SMTLs) are defined as benign lesions that resemble SMTs on endoscopy but are primarily characterized by chronic inflammatory changes without definitive tumor features in histopathology. Notably, 2.1% of resected SMTs lesions were pathologically confirmed as chronic inflammation [4]. A case report also highlighted eosinophilic colitis presenting with submucosal tumor-like lesions [5].

Endoscopic techniques, including white light endoscopy (WLE) and endoscopic ultrasound (EUS), are widely recommended for SMTs’ diagnosis [6,7,8], providing critical information on lesion location, size, color, surface characteristics, origin, echogenicity, echo heterogeneity, boundary, and so on. Accurate differentiation is crucial to avoid unnecessary procedures and optimize patient management. However, I-SMTLs can closely resemble neoplastic SMTs, leading to diagnostic uncertainty or potential errors [4, 9]. This poses a clinical decision-making challenge, as neoplastic SMTs often require resection or close monitoring, whereas inflammatory lesions typically only require follow-up or no treatment [6,7,8]. Although previous studies have noted that inflammatory lesions may mimic submucosal tumors [4, 5], were limited by small sample sizes and lacked systematic evaluation of their diagnostic features. As a result, the distinguishing characteristics of I-SMTLs remain insufficiently defined, underscoring the need for further investigation.

This multicenter retrospective study aimed to characterize the endoscopic and clinicopathological features of pathologically confirmed I-SMTLs and compare them with non-inflammatory SMT-like lesions (non-I-SMTLs). By leveraging a large, multicenter cohort and conducting a detailed analysis of EUS features, this study offers a more comprehensive understanding of I-SMTLs. These findings help bridge existing knowledge gaps, enhance clinical awareness, and support more accurate, evidence-based diagnostic decision-making.

Methods

Patients

This retrospective multicenter study was conducted across five Chinese hospitals from January 2013 and August 2024, including the Third People’s Hospital of Chengdu, the First People’s Hospital of Liangshan Yi Autonomous Prefecture, the Suining Central Hospital, the First People’s Hospital of Chengdu, and Sichuan Mianyang 404 Hospital. Patients with suspected SMTs undergoing endoscopic or surgery resection following WLE and EUS examination were screened. Inclusion criteria were patients who received a definite pathological diagnosis. The exclusion criteria were patients with inconclusive pathological results or incomplete clinical data or combined with malignant tumors. Include patient demographics, endoscopic features, and histopathological results. A flowchart detailing participant recruitment is presented in Fig. 1.

Fig. 1

Flowchart of participant recruitment of the study

Endoscopic procedures

WLE and EUS were performed by endoscopic experts or trainees under experts’ supervision at all participating hospitals. Ultrasonic mini-probes (12/20-MHz, UM-2R/3R, Olympus, Japan; 12/20-MHz, IM-02P-202501, INNERMED, Shenzhen, China) and convex linear array echoendoscopes.

(EG-3870UTK, Pentax, Japan) were employed in the study.

The lesion’s location, size, color, surface features, originating layer, echogenicity, echo heterogeneity, and boundary were reviewed and recorded by three endoscopists (M.T. Yu, J. Li, and X.B. Sun), along with the initial EUS diagnosis. To minimize interobserver variability, a consensus approach was adopted: any discrepancies in interpretation were resolved through joint review and discussion of the EUS images until agreement was achieved. This process helped to standardize the evaluation and ensure consistency across cases.

Histopathology analysis

Resection lesions were immediately fixed in 10% neutral formalin and routinely embedded for histological examination. Immunohistochemistry was performed as necessary. Lesions that predominantly exhibit chronic inflammatory responses without definitive tumor characteristics were classified as I-SMTLs in histopathology. The classified histopathological features for I-SMTLs include inflammatory cell infiltration (either diffuse or focal), cellular changes (such as epithelial or glandular alterations), and stromal changes (including smooth muscle, vascular hyperplasia or fibrosis) [10,11,12], as detailed in Supplementary Table 1. In contrast, lesions displaying distinct tumor features, such as GISTs, NETs, or leiomyomas, were classified as non-I-SMTLs [9]. An experienced pathologist (Dr. Y.Y. Chen, > 10 years), blinded to clinical data, independently applied predefined diagnostic and classification criteria for I-SMTLs (Supplementary Table 1) to ensure standardized and objective assessment. The pathologist’s expertise further reinforced the reliability of the evaluation.

Outcomes

The study aimed to achieve three key outcomes: [1] evaluate the diagnostic efficacy of EUS in identifying I-SMTLs, with a focus on diagnostic challenges; [2] use propensity score-matching (PSM) analysis to compare I-SMTLs and non-I-SMTLs, enhancing diagnostic awareness among endoscopists; and [3] explore the correlation between endoscopic and pathological characteristics to deepen the understanding of I-SMTLs.

Propensity score matching (PSM)-based analysis

To minimize confounding, rare non-I-SMTL cases, which have distinct biological characteristics, were excluded from the PSM analysis. This approach focused the analysis on more representative lesions, thereby enhancing both the generalizability and clinical applicability of the findings.

PSM was employed to reduce selection bias and enhance the comparability between the I-SMTLs and non-I-SMTLs groups. Matching was performed at a 1:2 ratio based on key variables such as gender, age, lesion location, and size. A caliper value of 0.2 was chosen in accordance with established recommendation [13], as it offers an optimal balance between reducing bias and preserving sample size in PSM for our study.

Statistical analysis

Continuous variables with a normal distribution were reported as mean ± standard deviation and compared using the t-test. Non-normally distributed continuous variables were presented as median (interquartile range) and compared with the Wilcoxon test. The Kolmogorov–Smirnov test and Q–Q plots were used to assess normality. Categorical variables were expressed as frequencies (percentages) and compared using Pearson’s chi-square test or Fisher’s exact test. Diagnostic efficacy was evaluated by calculating sensitivity, specificity, and accuracy.

The correlation between endoscopic and histopathologic characteristics was evaluated using Cramer’s V coefficient, a non-parametric statistic appropriate for categorical variables. It assesses the strength of association between two nominal variables without assuming causality or directionality. Given the limited sample size for some variables, Cramer’s V was chosen for its robustness and simplicity, avoiding the risk of overfitting associated with more complex models. Association strength was interpreted as: very strong (> 0.25), strong (> 0.15), moderate (> 0.10), weak (> 0.05), and very weak or none (≤ 0.05) [14].

To further explore the sources of between-group differences in categorical variables, the observed-minus-expected (O–E) value between the observed frequency (O) and the expected frequency (E) for key categorical variables was visualized. This approach provides an intuitive representation of how each category deviates from the expected distribution under the null hypothesis of independence. The observed frequency (O) refers to the actual count. The expected frequency (E) represents the theoretical count that would occur if the two variables were statistically independent. By comparing O–E value, this method helps identify the specific categories that contribute most to intergroup variation, enhancing the interpretability and potential clinical relevance of the results. Statistical analyses were performed using SPSS (version 29.0) and R software (version 4.2), with significance defined as a two-tailed p-value < 0.05.

Results

Patient demographics

A total of 1317 lesions resected endoscopically or surgically were retrospectively enrolled from five hospitals between January 2013 and August 2024. After excluding 91 lesions due to inconclusive pathology (n = 78), incomplete clinical data (n = 10), or concurrent early-stage malignancies (n = 3), 1226 lesions were included, comprising 4.81% (59/1226) I-SMTLs and 95.19% (1167/1226) non-I-SMTLs. The I-SMTL cohort consisted of 30 men and 29 women with a mean age of 52.41 ± 14.40 years (range, 19–87 years), while the non-I-SMTL cohort included 500 men and 667 women with a mean age of 53.18 ± 11.86 years (range, 16–92 years). Details are provided in Fig. 1; Table 1, and Supplementary Table 2.

Table 1 Comparison between I-SMTLs and non-I-SMTLs based on propensity score-matching cohort

Endoscopic features of I-SMTLs

I-SMTLs exhibited diverse endoscopic features. 98.31% of I-SMTLs were ≤ 20 mm in size. The most common sites were the rectum (37.29%), followed by stomach (27.12%), esophagus (16.95%), colon (15.25%), and duodenum (3.39%). 23.73% had a yellowish-white color on WLE, while 16.95% showed surface changes, including depression, hyperemia, vessel dilation (each 3 cases), and ulceration (1 case). EUS revealed that 71.19% of lesions were located in the submucosa, followed by the muscularis mucosa (11.86%), deep mucosa (10.17%), and muscularis propria (6.78%). Most lesions were hypoechoic (69.49%), with others showing mixed (18.64%), hyperechoic (8.47%), or isoechoic (3.39%,) patterns. Echotexture was heterogeneous in 62.71% of lesions, while 37.29% were homogeneous. Clear boundaries were observed in 57.63%, whereas 42.37% had unclear boundaries. Details are in Table 1.

Diagnostic effectiveness of EUS for I-SMTLs

Of the 59 I-SMTL cases, only two were correctly identified by EUS in accordance with postoperative pathology, resulting in a misdiagnosis rate of 96.61% (57/59) (Table 2). Additionally, a portion of cases were classified as “uncertain” diagnosis with a notably higher proportion in the I-SMTLs group compared to the non-I-SMTLs group (15.25% vs. 9.51%, as shown in Table 2).

Table 2 Confusion matrices comparing EUS diagnosis with histopathological results for I-SMTLs

To assess EUS diagnostic effectiveness, two methods were used. When excluding 120 EUS-diagnosed “uncertain” cases resulted in a sensitivity of 4.00% (2/50, 95% CI: 1.10–13.46%), minimizing interference and ensuring a more stable assessment focused on common, easily identifiable lesions. When treating these “uncertain” cases as misdiagnoses, sensitivity was 3.39% (2/59, 95% CI: 0.93–11.54%). Given the higher likelihood of misdiagnosis of I-SMTLs, this approach better reflects EUS performance in challenging cases (Table 3). No significant statistical difference was found between the two methods. Excluding “uncertain” cases from accurate diagnoses ensures a more accurate representation of clinical practice.

Table 3 Diagnostic efficacy of endoscopic ultrasonography for I-SMTLs

From a lesion-based perspective, I-SMTLs were most frequently misdiagnosed as NETs (44.07%), predominantly in the rectum (Fig. 2A) and colon (Fig. 2B), followed by leiomyomas (13.56%) mainly in the esophagus (Fig. 2C), GISTs and ectopic pancreas (10.17% each) mainly in the stomach and duodenum (Fig. 2D-F), and granular cell tumor (GCT) in the esophagus and lipoma in the colon (1.69%), as shown in Table 4.

Fig. 2

Misdiagnosed inflammatory submucosal tumors -like lesions in various locations. A: misdiagnosed as rectal neuroendocrine; B: misdiagnosed as descending colonic neuroendocrine tumor; C: misdiagnosed as esophageal leiomyoma; D: misdiagnosed as gastric stromal tumor; E: misdiagnosed as gastric ectopic pancreas; F: misdiagnosed as duodenal ectopic pancreas

Table 4 EUS diagnosis of I-SMTLs by location

Comparison between I-SMTLs and non-I-SMTLs based on PSM cohort

A total of 48 rare cases, representing 22 pathological types (Supplementary Table 2), were removed in this comparison. The final analysis included 59 I-SMTLs and 118 PSM-matched non-I-SMTLs. Before PSM, significant differences were observed between I-SMTLs and non-I-SMTLs in lesion location, size, originating layer, echogenicity, echo heterogeneity, and borders (p<0.05, Table 1). A PSM cohort was constructed by matching cases based on sex, age, lesion location, and size, resulting in a total of 117 cases (59 I-SMTLs and 118 non-I-SMTLs, Supplementary Table 3). After PSM, significant differences in EUS features, including originating layer, echogenicity, echo heterogeneity, and borders (p<0.05), while lesion location, size, color and mucosal surface changes showed no significant differences.

Group variation sources were analyzed by visualizing the O-E value between the observed frequency (O) and the expected frequency (E). For the originating layer, the most significant discrepancy was observed in the submucosal layer, with an O - E value of 11.33 (Fig. 3A). Compared to non-I-SMLTs, I-SMTLs were more frequently located in submucosal layer (71.19% vs. 42.37%). Regarding echogenicity, the mixed-echoic category showed the largest O - E value of 6.67 (Fig. 3B), with I-SMTLs being more frequently observed in mixed-echoic lesions (18.64% vs. 1.69%). Furthermore, I-SMTLs were significantly more likely to exhibit heterogeneous echogenicity (62.71% vs. 21.19%, p < 0.001), corresponding to an O-E value of 16.33 (Fig. 3C), and unclear boundaries (42.37% vs. 13.56%, p < 0.001), with an O-E value of 11.33 (Fig. 3D), further emphasizing the distinct characteristics of this group.

Fig. 3

Observed-minus-expected (O-E) value between observed and expected frequencies of EUS features. A-D: O-E value of EUS features between I-SMTLs and non-I-SMTLs. A: shows the originating layer, B: shows echogenicity, C: shows echo heterogeneity, and D: shows boundary. Note: Green column: I-SMTLs group; Orange column: non-I-SMTLs group. E-G: O-E value of EUS features between groups with different pathological characteristics. E: shows Originating layer distribution in lesions with versus without smooth muscle hyperplasia; F: Echogenicity in lesions with versus without fibrous tissue alterations; G: Echo heterogeneity in lesions with versus without fibrous tissue alterations. Note: Green column: Presence of a specific pathological feature; Orange column: Absence of the feature. Positive values indicate observed > expected; negative values indicate observed < expected

Correlation between endoscopic and pathologic characteristics of I-SMTLs

In terms of pathological characteristics, I-SMTLs primarily show inflammatory cell infiltration, with 39 cases of diffuse and 21 cases of focal infiltration (out of 59). Among cellular proliferation, epithelial hyperplasia was observed in 12 cases, while glandular structural changes were noted in 11 cases. Stromal changes were dominated by fibrous tissue alterations (17 of 42 cases), followed by smooth muscle hyperplasia (9 cases) and vascular proliferation (6 cases) (Supplementary Table 4).

In the correlation analysis, smooth muscle hyperplasia was significantly associated with the originating layer (Cramer’s V = 0.490, p = 0.004). Cases with smooth muscle hyperplasia were more likely to arise from the muscularis propria or muscularis mucosae (55.56%, 5/9), as reflected by higher-than-expected frequencies in the hyperplasia group: muscularis propria (O − E = + 2.39) and muscularis mucosa (O − E = + 0.93) (Fig. 3E). Fibrous tissue alterations demonstrated strong correlations with echogenicity (Cramer’s V = 0.545, p < 0.001) and heterogeneous echotexture (Cramer’s V = 0.336, p = 0.022). Cases with fibrous tissue alterations were more likely to exhibit mixed or hyperechoic patterns (64.71%, 11/17) with an O-E value of 4.83 (Fig. 3F) and heterogeneous echotexture (88.24%, 15/17) with an O-E value of 4.34 (Fig. 3G). Details are provided in Table 5. The endoscopic and pathological characteristics of a I-SMTL misdiagnosed as a rectal NET are shown in Fig. 4A.

Table 5 Correlation between endoscopic and histopathologic characteristics of I-SMTLs

Fig. 4

Correlation between endoscopic and histopathologic characteristics in inflammatory submucosal tumors -like lesions. A1-2: A lesion misdiagnosed as a rectal neuroendocrine tumor, presenting as a submucosal mass with mixed echogenicity and a heterogeneous echotexture on EUS. A3-4: Hematoxylin and Eosin (HE) staining reveals diffuse inflammatory cell infiltration, fibrous tissue proliferation (black arrow), and inflammatory granulation tissue (blue arrow). B1-2: A collagenous nodule in the transverse colon, presenting as a submucosal arcuate hyperechoic band with posterior acoustic shadowing on EUS. B3-4: HE staining reveals collagen fiber proliferation, forming a dense, mesh-like structure with nodular formations

Representative cases analysis

Of the 59 I-SMTL cases, only two were accurately diagnosed. The first, a gastric collagenous nodule, was characterized by a submucosal, arcuate hyperechoic band with posterior acoustic shadowing on EUS (Fig. 5A). Similar endoscopic findings were also presented in a case of a collagenous nodule in the transverse colon (Fig. 4B1-2), and the pathological findings show significant proliferation of collagen fibers, forming a dense, mesh-like structure with nodular formations (Fig. 4B3-4). The second case showed ulceration and purulent discharge, and mixed-echo submucosal lesion with unclear margins, calcifications, and non-echoic mass on EUS (Fig. 5B). Pathology revealed chronic inflammation with erosion. Another gastric lesion with erosion on surface, misdiagnosed as a GIST, appeared hypoechoic in the muscularis propria with inhomogeneous echoes extending beyond the serosa (Fig. 5C). Pathology confirmed a foreign body with gastric perforation. These cases, along with a schistosomal infection (Fig. 5D), had confirmed inflammation etiologies. Additionally, two lymphoglandular complexes (LGCs) displayed typical EUS findings as high-echo or mixed-echo lesions in the submucosal layer, accompanied by calcifications (Figs. 5E and F). Pathologically, LGCs are lymphoid nodules containing intestinal mucosa, located near the submucosa or muscularis mucosae [11]. In addition, one patient underwent a combined assessment using EUS, EUS elastography (EUS-E), and contrast-enhanced EUS (CE-EUS). Histopathology revealed a necrotic focus within the muscularis propria of the gastric antrum, along with chronic inflammatory cell infiltration and surrounding smooth muscle hyperplasia and hypertrophy. The corresponding endoscopic features are shown in Fig. 6.

Fig. 5

Representative cases of inflammatory submucosal tumors -like lesions. A: Collagenous nodule; B: Chronic inflammation with erosion; C: Foreign body with gastric perforation; D: Schistosomal infection; E and F: Lymphoglandular complexes

Fig. 6

Endoscopic features of an inflammatory submucosal tumor-like lesion. A–B: White-light endoscopy shows a submucosal bulge in the anterior wall of the gastric antrum with a smooth surface and intact mucosa. C: Endoscopic ultrasonography (EUS) shows a hypoechoic lesion originating from the muscularis propria, with heterogeneous internal echotexture, a central patchy hyperechoic focus, and well-defined margins. D: EUS elastography shows predominantly heterogeneous blue coloration (white arrow), indicating increased tissue stiffness. E: Contrast-enhanced EUS (CE-EUS) shows no significant contrast uptake within the lesion (white arrow), suggesting hypo-vascularity

Discussion

This multicenter study identified the proportion of I-SMTLs among resected SMTs was 4.81%, exceeding the previously documented 2.1% [4]. Currently, the diagnostic accuracy for I-SMTLs is not optimistic, with a limited diagnostic sensitivity (approximately 4%) and a high misdiagnosis rate (96.61%). Distinguishing EUS features, including submucosal location, mixed-echoic patterns, heterogeneous textures, and unclear boundaries, may improve diagnostic accuracy. To the best of our knowledge, this study is the first to highlight the diagnostic challenges of the overlooked condition of I-SMTLs, emphasizing EUS features to improve differentiation from non-I-SMTLs.

I-SMTLs are rare benign conditions that often mimic true neoplastic SMTs. Over 40% of I-SMTLs exhibit yellowish-white discoloration or mucosal surface changes on WLE. On EUS, over 70% are located in the submucosa, nearly 70% are hypoechoic, and over 60% exhibit heterogeneous echotexture. In fact, over 55% of I-SMTLs misclassified as potentially malignant conditions (e.g., GISTs, NETs, and GCT) and more than 15% yielding inconclusive results. Misclassification of I-SMTLs as potentially malignant lesions may lead to unnecessary endoscopic or surgical interventions, increasing procedural risks and healthcare costs. The presumption of malignancy can also cause significant psychological distress in patients. Inaccurate diagnosis may further delay appropriate treatment of true malignancies. Such misdiagnoses may contribute to overtreatment in cases where a conservative approach would have sufficed, highlighting the importance of improving diagnostic accuracy to ensure appropriate clinical management.

To improve the comparability of endoscopic features between I-SMTLs and non-I-SMTLs, a PSM cohort was established, matching cases by sex, age, lesion location, and size to minimize sample imbalance and reduce confounding bias. The PSM analysis revealed that significant differences were observed in EUS, but not WLE, features between the two groups. Compared to non-I-SMLTs, I-SMTLs were more frequently located in submucosal layer (71.19% vs. 42.37%), had mixed echogenicity (18.64% vs. 1.69%), heterogeneous echogenicity (62.71% vs. 21.19%), and unclear boundaries (42.37% vs. 13.56%), further highlighting their distinct EUS features. These findings suggest that endoscopists should prioritize EUS features when distinguishing inflammatory SMTs. To further illustrate these typical findings, representative endoscopic images of I-SMTLs are provided in Figs. 2 and 5. In one representative case in Fig. 6, EUS and EUS-E mimicked features of GISTs [15], but CE-EUS showed no significant enhancement, suggesting hypo-vascularity, which unlike the typical hyper-vascular pattern of GISTs [16]. The absence of enhancement corresponded to a necrotic focus on pathology. Although based on a single case, the findings suggest that advanced EUS techniques may aid in differentiating I-SMTLs from other subepithelial lesions, warranting further study.

This study further explored the correlation between the pathological features of I-SMTLs and endoscopic characteristics to deepen endoscopists’ understanding of I-SMTLs. Chronic inflammation is a multifaceted process involving an initial inflammatory response to tissue damage, followed by regeneration and repair [10]. On EUS, I-SMTLs predominantly originate from the submucosal layer (71.19%), though 18.44% arise from the muscularis mucosae or muscularis propria. This may result from smooth muscle hyperplasia, which thickens the muscularis layer and leads a muscular origin appearance (Cramer’s V = 0.490). Meanwhile, fibrous tissue alterations can be visualized on EUS through changes in echogenicity and echo heterogeneity, with significant correlations observed (Cramer’s V = 0.545 and 0.336, respectively). Lesion with fibrotic changes frequently exhibit mixed or hyperechoic patterns (64.71%) and heterogeneous echotexture (88.24%). The PSM cohort also confirmed significant differences in echogenicity and echo heterogeneity between I-SMTLs and non-I-SMTLs. A deeper understanding of these correlation can enhance endoscopists’ recognition their diverse EUS features of I-SMTLs, facilitating more accurate diagnoses. A potential limitation is the lack of a standardized pathological classification for I-SMTLs. The criteria adopted in the study are exploratory, informed by clinical experience, relevant histopathological features, and existing literature [10,11,12]. Although some interobserver variability is expected, the proposed framework offers a practical starting point for I-SMTLs. Future studies may refine diagnostic markers, incorporate blinded multi-pathologist assessments, and assess interobserver agreement to enhance consistency and applicability.

The diagnostic complexity of I-SMTLs poses significant challenges to clinical decision-making. In this study, 98.31% of I-SMTLs measured less than 2 cm, making follow-up the preferred but still controversial strategy in current guidelines [7, 8, 17,18,19]. This issue is further compounded by the fact that EUS frequently underestimates lesion size [1, 20], which complicates accurate evaluation of tumor dimensions and size-based treatment strategies. Additionally, the overlap between I-SMTLs and malignant features, such as NETs, often shifts clinical decisions toward endoscopic resection [7, 17]. Approximately 56% of lesions were located outside the esophagus and stomach, where the European Society of Gastrointestinal Endoscopy (ESGE) guidelines emphasize the need for an accurate pathological diagnosis to guide management [8]. EUS-guided fine-needle aspiration/biopsy (EUS-FNA/B) is initially recommended as the preferred biopsy method for SMTs, but lesions smaller than 2 cm pose technical challenges and low diagnostic accuracy [7, 8, 21, 22]. A network meta-analysis showed that for lesions < 20 mm, mucosal incision-assisted biopsy (MIAB) had the highest sample adequacy(surface under cumulative ranking [SUCRA] score 0.86) and diagnostic accuracy (SUCRA score 0.91), significantly outperforming EUS-FNB, EUS-FNA, and bite-on-bite biopsy [21], with diagnostic accuracy reaching 88.2%-89% in supporting studies [22, 23]. A key limitation is that I-SMTLs are nonspecific chronic inflammatory lesions, and even with adequate MIAB samples, a diagnosis of ‘chronic inflammation’ may be clinically ambiguous. This raises concern for false negatives, potentially prompting unnecessary resections despite benign pathology. Thus, while MIAB may help avoid resection in some cases, its findings should be interpreted cautiously and within the full clinical and endoscopic context. These limitations contribute to uncertainty in determining the optimal approach—whether to pursue surveillance, biopsy, or resection. The American College of Gastroenterology (ACG) Clinical Guideline states that small lesions in the deep mucosa or submucosa may be resected endoscopically for definitive diagnosis [24]. Additionally, study have reported that resection of gastric submucosal tumors ≤ 20 mm in patients who can afford it can help reduce anxiety levels while yielding similar clinical outcomes [25]. In this context, when EUS features raise suspicion of an I-SMTL in a small lesion, a low threshold for tissue diagnosis (via mucosal incision biopsy or endoscopic resection) is recommended, further validation in prospective studies is warranted.

This study has several limitations. First, the retrospective design inherently limits the study, as it may introduce selection bias and depends on the accuracy of historical records. Future prospective or registry-based studies are needed to validate these findings. Second, inclusion was limited to resected and pathologically confirmed lesions, which may introduce selection bias. Benign-appearing lesions that were not resected were excluded, potentially leading to a misestimation of the true prevalence of I-SMTLs among all submucosal lesions. In addition, management outcomes for unresected I-SMTLs were not assessed. Thus, the findings are primarily applicable to lesions warranting resection. Third, the cohort was drawn from tertiary medical centers within a single country, limiting generalizability due to potential geographic differences in inflammatory etiologies and clinical practices. Validation in more diverse populations or through international collaboration is warranted. Fourth, the small sample size for analyzing the correlation between EUS features and histopathological changes in I-SMTLs may limit the statistical power and robustness of the results. These findings should be considered preliminary and require validation in larger cohorts. Fifth, although PSM was used to reduce baseline differences, unmeasured confounders may remain. Subtle variations in patient demographics, EUS methodologies across centers, and endoscopists’ thresholds for resection may have influenced the results. Finally, this study focuses on diagnostic features and short-term outcomes, but lacks long-term follow-up. The natural history of I-SMTLs remains unclear, particularly regarding whether unresected lesions remain stable, resolve spontaneously, or progress to malignancy. Clarifying these aspects is essential to determine whether surveillance may represent a safe management strategy.

Conclusion

I-SMTLs present notable diagnostic challenges. Recognizing EUS features, including submucosal location, mixed-echoic patterns, heterogeneous textures, and unclear boundaries, may improve diagnostic accuracy and support appropriate clinical decision-making. Large-scale and multicenter prospective studies are needed to validate these findings. Future research should also assess whether advanced imaging techniques can enhance diagnostic precision. In addition, understanding the natural history and long-term outcomes of I-SMTLs, may help optimize treatment strategies.