A Comparison of Logistic Regression Against Machine Learning Algorithms for Gastric Cancer Risk Prediction Within Real-World Clinical Data Streams.

PURPOSE: Noncardia gastric cancer (NCGC) is a leading cause of global cancer mortality, and is often diagnosed at advanced stages. Development of NCGC risk models within electronic health records (EHR) may allow for improved cancer prevention. There has been much recent interest in use of machine learning (ML) for cancer prediction, but few studies comparing ML with classical statistical models for NCGC risk prediction. METHODS: We trained models using logistic regression (LR) and four commonly used ML algorithms to predict NCGC from age-/sex-matched controls in two EHR systems: Stanford University and the University of Washington (UW). The LR model contained well-established NCGC risk factors (intestinal metaplasia histology, prior Helicobacter pylori infection, race, ethnicity, nativity status, smoking history, anemia), whereas ML models agnostically selected variables from the EHR. Models were developed and internally validated in the Stanford data, and externally validated in the UW data. Hyperparameter tuning of models was achieved using cross-validation. Model performance was compared by accuracy, sensitivity, and specificity. RESULTS: In internal validation, LR performed with comparable accuracy (0.732; 95% CI, 0.698 to 0.764), sensitivity (0.697; 95% CI, 0.647 to 0.744), and specificity (0.767; 95% CI, 0.720 to 0.809) to penalized lasso, support vector machine, K-nearest neighbor, and random forest models. In external validation, LR continued to demonstrate high accuracy, sensitivity, and specificity. Although K-nearest neighbor demonstrated higher accuracy and specificity, this was offset by significantly lower sensitivity. No ML model consistently outperformed LR across evaluation criteria. CONCLUSION: Drawing data from two independent EHRs, we find LR on the basis of established risk factors demonstrated comparable performance to optimized ML algorithms. This study demonstrates that classical models built on robust, hand-chosen predictor variables may not be inferior to data-driven models for NCGC risk prediction.

Endoscopic submucosal dissection for colorectal dysplasia in inflammatory bowel disease: a US multicenter study.

Background and study aims In patients with inflammatory bowel disease (IBD), endoscopically visible lesions with distinct borders can be considered for endoscopic resection. The role of endoscopic submucosal dissection (ESD) for these lesions is not well defined because of a paucity of data. We aimed to evaluate the outcomes of colorectal ESD of dysplastic lesions in patients with IBD across centers in the United States. Patients and methods This was a retrospective analysis of consecutive patients with IBD who were referred for ESD of dysplastic colorectal lesions at nine centers. The primary endpoints were the rates of en bloc resection and complete (R0) resection. The secondary endpoints were the rates of adverse events and lesion recurrence. Results A total of 45 dysplastic lesions (median size 30mm, interquartile range [IQR] 23 to 42 mm) in 41 patients were included. Submucosal fibrosis was observed in 73 %. En bloc resection was achieved in 43 of 45 lesions (96 %) and R0 resection in 34 of 45 lesions (76 %). Intraprocedural perforation occurred in one patient (2.4 %) and was treated successfully with clip placement. Delayed bleeding occurred in four patients (9.8 %). No severe intraprocedural bleeding or delayed perforation occurred. During a median follow-up of 18 months (IQR 13 to 37 months), local recurrence occurred in one case (2.6 %). Metachronous lesions were identified in 11 patients (31 %). Conclusions ESD, when performed by experts, is safe and effective for large, dysplastic colorectal lesions in patients with IBD. Despite the high prevalence of submucosal fibrosis, en bloc resection was achieved in nearly all patients with IBD undergoing ESD. Careful endoscopic surveillance is necessary to monitor for local recurrence and metachronous lesions after ESD.

Dedifferentiated Liposarcoma Presenting as Dysphagia and Weight Loss: Common Symptoms but an Uncommon Diagnosis.

A 76-year-old woman was referred for 6 months of progressively worsening dysphagia and unintentional weight loss. An esophagogastroduodenoscopy demonstrated an area of extrinsic compression in the lower esophagus measuring 7 cm in greatest dimension. Contrast-enhanced computed tomography revealed a solid homogeneous mass in the lower middle/posterior mediastinum, laterally displacing the esophagus. Endoscopic ultrasound-guided fine-needle biopsy showed a hypocellular infiltrate of pleomorphic cells in a loose collagenous matrix. By immunohistochemistry, neoplastic cells were negative for epithelial, vascular, neural, and melanocytic markers. Fluorescent in situ hybridization detected MDM2 amplification, compatible with a diagnosis of dedifferentiated liposarcoma.

Real-World Data on the Impact of COVID-19 on Endoscopic Procedural Delays.

INTRODUCTION: The initial surge of the coronavirus disease 2019 (COVID-19) pandemic prompted national recommendations to delay nonurgent endoscopic procedures. The objective of this study was to provide real-world data on the impact of COVID-19 on endoscopic procedures in a safety-net healthcare system and cancer center affiliated with a tertiary academic center. METHODS: This retrospective cohort study used a combination of electronic health record data and a prospective data tool created to track endoscopy procedures throughout COVID-19 to describe patient and procedural characteristics of endoscopic procedures delayed during the initial COVID-19 surge. RESULTS: Of the 480 patients identified, the median age was 57 years (interquartile range 46-66), 55% (n = 262) were male, and 59% self-identified as white. Colonoscopy was the most common type of delayed procedure (49%), followed by combined esophagogastroduodenoscopy (EGD) and colonoscopy (22%), and EGD alone (20%). Colorectal cancer screening was the most common indication for delayed colonoscopy (35%), and evaluation of suspected bleeding (30%) was the most common indication for delayed combined EGD and colonoscopy. To date, 46% (223/480) of delayed cases have been completed with 12 colorectal, pancreatic, and stomach cancers diagnosed. Sociodemographic factors, procedure type, and sedation type were not significantly associated with endoscopy completion. The median time to endoscopy after delayed procedure was 88 days (interquartile range 63-119) with no differences by procedure type. DISCUSSION: To minimize potential losses to follow-up, delayed, or missed diagnoses and to reduce progression of gastrointestinal diseases, all efforts should be used to ensure follow-up in those whose endoscopic procedures were delayed because of COVID-19.

Esophageal granular cell tumor and eosinophils: a multicenter experience.

BACKGROUND: Esophageal granular cell tumor (eGCT) is rare, and the recent literature suggests a link between eosinophilic esophagitis (EoE) and eGCT. The aim of our study was to determine if EoE or other disorders associated with eosinophilia are consistently associated with eGCT. METHODS: We retrospectively searched pathology databases of three academic institutions from 1999 to 2018 for eGCTs. The archived slides and medical records were reviewed. RESULTS: From 294,855 esophagogastroduodenoscopy procedures, 45 patients (17 males and 28 females) with eGCTs were identified. The patients (30-73 years in age, median 50) had eGCT 0.2-2.0 cm in size (average 0.71). Thirteen had a history of gastroesophageal reflux disease, 5 had Barrett esophagus/goblet cell metaplasia and 1 had EoE. Thirty-four eGCTs had intralesional eosinophils (14 with peak > 10 eosinophils/400x hpf); of these, 21 also had eosinophils in lamina propria (9 with peak > 10 eosinophils/hpf). eGCT with atypical features (including nuclear enlargement and prominent nucleoli) were more likely to have increased eosinophils in non-epithelial compartments than those without atypia. Pleomorphism and spindled cells were seen in 3 eGCT cases (mean peak intralesional eosinophils: 43 per hpf); 2 of these had goblet cell metaplasia. We found no association between EoE and eGCT, p = 0.5966, (95% C.I. 0.0276, 6.5389, Fisher's exact test). Instead, most patients had gastroesophageal reflux disease or Barrett esophagus. CONCLUSION: Eosinophilia, common in eGCT and adjacent stroma, likely drives atypical/reactive histologic features, but a pathogenic relationship between eosinophil rich inflammatory conditions and eGCT has not yet been established.

Efficacy of Endoscopic Submucosal Dissection for Superficial Gastric Neoplasia in a Large Cohort in North America.

BACKGROUND & AIMS: Endoscopic submucosal dissection (ESD) is a widely accepted treatment option for superficial gastric neoplasia in Asia, but there are few data on outcomes of gastric ESD from North America. We aimed to evaluate the safety and efficacy of gastric ESD in North America. METHODS: We analyzed data from 347 patients who underwent gastric ESD at 25 centers, from 2010 through 2019. We collected data on patient demographics, lesion characteristics, procedure details and related adverse events, treatment outcomes, local recurrence, and vital status at the last follow up. For the 277 patients with available follow-up data, the median interval between initial ESD and last clinical or endoscopic evaluation was 364 days. The primary endpoint was the rate of en bloc and R0 resection. Secondary outcomes included curative resection, rates of adverse events and recurrence, and gastric cancer-related death. RESULTS: Ninety patients (26%) had low-grade adenomas or dysplasia, 82 patients (24%) had high-grade dysplasia, 139 patients (40%) had early gastric cancer, and 36 patients (10%) had neuroendocrine tumors. Proportions of en bloc and R0 resection for all lesions were 92%/82%, for early gastric cancers were 94%/75%, for adenomas and low-grade dysplasia were 93%/ 92%, for high-grade dysplasia were 89%/ 87%, and for neuroendocrine tumors were 92%/75%. Intraprocedural perforation occurred in 6.6% of patients; 82% of these were treated successfully with endoscopic therapy. Delayed bleeding occurred in 2.6% of patients. No delayed perforation or procedure-related deaths were observed. There were local recurrences in 3.9% of cases; all occurred after non-curative ESD resection. Metachronous lesions were identified in 14 patients (6.9%). One of 277 patients with clinical follow up died of metachronous gastric cancer that occurred 2.5 years after the initial ESD. CONCLUSIONS: ESD is a highly effective treatment for superficial gastric neoplasia and should be considered as a viable option for patients in North America. The risk of local recurrence is low and occurs exclusively after non-curative resection. Careful endoscopic surveillance is necessary to identify and treat metachronous lesions.

Type II achalasia is associated with a comparably favorable outcome following per oral endoscopic myotomy.

Per oral endoscopic myotomy (POEM) is a safe and effective treatment for esophageal motility disorder in treatment-naïve patients as well as salvage therapy. Though type II achalasia, compared to other subtypes, is reported to have a more favorable outcome with pneumatic dilation (PD) or Heller myotomy (HM), it is unclear whether achalasia subtype predicts symptom response to POEM. We aimed to evaluate whether type II achalasia is associated with a comparably favorable outcome following POEM. We performed a retrospective review of patients with esophageal motility disorder who were referred for POEM from April 2014 to June 2017. The main outcome was clinical success based on Eckardt score ≤3 and its association with subtype and safety. A total of 63 patients (mean age 51 years [SD 15]; 63% male) underwent a total of 68 POEMs with median of 263 days follow-up. Of these, 45 (71.3%) patients were type II achalasia. In all, 29 (46%) patients were treatment-naïve and 34 (54%) patients had previous endoscopic or surgical therapy including botulinum toxin injection in 16 (25%), PD in 10 (16%), both botulinum toxin injection and PD in 8 (13%) and HM in 3 (5%). Technical success was 100% and clinical success was achieved in 51 (81%) patients. The rate of clinical success was higher in patients with type II achalasia compared to the other subtypes (88.9% vs. 61.1% [P = 0.028]) and type II achalasia patients required fewer redo POEM (2.2% vs. 22.2% [P = 0.021]). Multivariate logistic regression analysis demonstrated the positive prediction of clinical success for type II achalasia following POEM (P = 0.046). As observed with PD and HM, type II achalasia was associated with a favorable clinical outcome following POEM.

Endoscopic submucosal dissection (ESD) for Barrett's esophagus (BE)-related early neoplasia after standard endoscopic management is feasible and safe.

Background and study aims There is little data on the feasibility and safety of endoscopic submucosal dissection (ESD) as a salvage treatment for Barrett's esophagus (BE)-related neoplasia after standard endoscopic treatments. Patients and methods A multicenter retrospective analysis on patients who underwent ESD for BE was performed. The primary endpoint was effectiveness of obtaining en-bloc resection in salvage as compared to non-salvage treatments. Results Median age was 71 (IQR 55 - 79) years. Twelve (37%) of 32 patients underwent salvage ESD. Median resection time was 100 (IQR 60 - 136) minutes. En-bloc resection was achieved in 31 patients (97 %). Complete R0 resection was obtained in 75 % in the salvage group and 80 % in the non-salvage group ( P  = 1.00). In seven patients (22 %), the pre-ESD diagnosis was upgraded on post-ESD histopathology (1 low-grade dysplasia to high grade dysplasia [HGD], 4 HGD to early esophageal carcinoma (EAC), and 2 intramucosal EAC to invasive EAC). No perforations occurred in either group. Two late adverse events occurred, both in the salvage group ( P  = 0.133). Delayed bleeding occurred in a patient who had just resumed warfarin and stricture occurred in a patient who had a circumferential resection requiring serial dilation and stent placement. Conclusions Our cohort study demonstrated that ESD as salvage therapy for BE related neoplasia is feasible and safe, achieving similar high rates of en-bloc resection and complete R0 resection as in treatment-naïve patients. Referral to an expert center performing ESD should be considered for patients with recurrence or progression following endoscopic mucosal resection or ablation therapy.

Family history of colorectal cancer and prevalence of advanced colorectal neoplasia in asymptomatic screened populations in different age groups.

BACKGROUND AND AIMS: The clinical significance of a family history (FH) of colorectal cancer (CRC) in first-degree relatives (FDRs) in CRC screening stratified by different age groups of screened individuals is not fully understood. We investigated the relationship between FH and the presence of advanced colorectal neoplasia (ACN) in screened individuals in different age groups. METHODS: Data from screened individuals aged 40 to 54 years (n = 2263) and 55 to 69 years (n = 2621) who underwent their first-ever screening colonoscopy were analyzed. The relationship between FH and ACN was examined, and a multivariate logistic regression analysis incorporating other baseline characteristics was performed. RESULTS: Among individuals aged 40 to 54 years, the prevalence of ACN was significantly higher in 249 individuals with affected FDRs than in those without (5.6% vs 1.6%; P < .01), with an adjusted odds ratio of 3.7 (95% confidence interval, 1.9-7.0; P < .01); the prevalence was particularly high in those having FDRs with CRC mortality (7.3%). Among individuals aged 55 to 69 years, the prevalence of ACN was not significantly different between 291 individuals with affected FDRs and those without (5.8% vs 5.8%; P = .95); however, individuals with 2 FDRs with CRC and mortality showed a high prevalence of ACN (17.4% and 42.9%, respectively). CONCLUSIONS: An FH of CRC in FDRs was associated with a higher prevalence of ACN in younger individuals, with a particularly high impact of FH of CRC mortality. In contrast, the impact of FH was weaker in older individuals except those having 2 FDRs with CRC or mortality.

Per oral endoscopic myotomy as salvage therapy in patients with achalasia refractory to endoscopic or surgical therapy is technically feasible and safe: Systematic review and meta-analysis.

BACKGROUNDS AND AIMS: Per oral endoscopic myotomy (POEM) has been reported as an effective and safe salvage therapy for achalasia but there is limited composite data. We performed a systematic review and meta-analysis of studies that reported the rates of clinical success and adverse events among patients who underwent POEM after failed conventional endoscopic or surgical therapy. METHODS: Electronic literature search was conducted from inception through December 2018 for articles reporting the efficacy and safety of POEM in patients with achalasia who failed endoscopic or surgical therapy. Primary outcome was the pooled estimated rates of clinical success, defined as Eckardt score ≤ 3 after POEM. Secondary outcomes were procedural time, the rates of POEM-related gastroesophageal reflux disease (GERD) and procedure-related adverse events. RESULTS: Seven studies reporting outcomes on 487 patients met our criteria. Pooled estimated rate of clinical success of POEM was 88% (95% confidence interval (CI) 79-94%). Mean procedural time was 64 minutes (95% CI 44-85 minutes). POEM-related GERD was found in 20% (95% CI 16-24%) of patients. Estimated incidence of overall adverse events was 10% (95% CI 5-18%) with individual risk of bleeding, mucosotomy, pneumothorax, pneumoperitoneum hydrothorax/mediastinitis, and subcutaneous emphysema ranging from 1 to 4%. CONCLUSIONS: Per oral endoscopic myotomy after failed endoscopic or surgical therapy in patients with achalasia is an effective and safe treatment. Further long-term follow-up studies in a larger number of patients are warranted to validate the sustainable efficacy of POEM for achalasia.

Safety and efficacy of endoscopic submucosal dissection for rectal neoplasia: a multicenter North American experience.

Background and aims Rectal lesions traditionally represent the first lesions approached during endoscopic submucosal dissection (ESD) training in the West. We evaluated the safety and efficacy of rectal ESD in North America. Methods This is a multicenter retrospective analysis of rectal ESD between January 2010 and September 2018 in 15 centers. End points included: rates of en bloc resection, R0 resection, adverse events, comparison of pre- and post-ESD histology, and factors associated with failed resection. Results In total, 171 patients (median age 63 years; 56 % men) underwent rectal ESD (median size 43 mm). En bloc resection was achieved in 141 cases (82.5 %; 95 %CI 76.8-88.2), including 24 of 27 (88.9 %) with prior failed endoscopic mucosal resection (EMR). R0 resection rate was 74.9 % (95 %CI 68.4-81.4). Post-ESD bleeding and perforation occurred in 4 (2.3 %) and 7 (4.1 %), respectively. Covert submucosal invasive cancer (SMIC) was identified in 8.6 % of post-ESD specimens. There was one case (1/120; 0.8 %) of recurrence at a median follow-up of 31 weeks; IQR: 19-76 weeks). Older age and higher body mass index (BMI) were predictors of failed R0 resection, whereas submucosal fibrosis was associated with a higher likelihood of both failed en bloc and R0 resection. Conclusion Rectal ESD in North America is safe and is associated with high en bloc and R0 resection rates. The presence of submucosal fibrosis was the main predictor of failed en bloc and R0 resection. ESD can be considered for select rectal lesions, and serves not only to establish a definitive tissue diagnosis but also to provide curative resection for lesions with covert advanced disease.

Additional value of linked color imaging in colonoscopy: a retrospective study.

Background and study aims Linked color imaging (LCI), a newly developed optical modality, enhances mucosal surface contrast. We aimed to evaluate the efficacy and feasibility of insertion-phase LCI in terms of additional benefit of colorectal polyp detection over that obtained with white light imaging (WLI). Patients and methods We consecutively enrolled eligible patients from November 2017 to June 2018. During colonoscopy, LCI or WLI was alternatively applied on scope insertion and LCI was applied on scope withdrawal. Patients were divided into two groups according to the protocolized difference of imaging modality used in the scope insertion phase (LCI and WLI groups). Group differences in clinical outcomes were evaluated. Results A total of 138 patients were enrolled in this study, with equal numbers of patients assigned to the LCI and WLI groups. Most of the lesions located in the proximal colon were detected during the withdrawal phase, without a difference in proportions between the two groups. However, in the LCI group, eight of 49 lesions (16 %) located in the sigmoid and rectosigmoid colon were only detected during the insertion phase, and no such lesions (0 %) were detected during the insertion phase in the WLI group ( P  = 0.045). Conclusions This study showed the efficacy and feasibility of LCI in improving colorectal polyp detection in the sigmoid colon, especially during insertion. Further studies are warranted to validate the results of our single-center study.

Incidence of Advanced Colorectal Neoplasia in Individuals With Untreated Diminutive Colorectal Adenomas Diagnosed by Magnifying Image-Enhanced Endoscopy.

OBJECTIVES: Because of the increasing number of detected diminutive colorectal adenomas, the "diagnose-and-do-not-resect" approach has recently attracted attention as an alternative to resection. We evaluated the cumulative incidence of advanced colorectal neoplasia (ACN) in individuals with untreated diminutive adenomas and compared this incidence in individuals without adenomas. METHODS: Data from 1,378 individuals who underwent first screening colonoscopy (CS) and at least one follow-up CS without polypectomy were analyzed. Patients with no adenomas or with only nonadvanced diminutive adenomas (<5 mm) diagnosed by magnifying image-enhanced endoscopy were scheduled to undergo a follow-up CS within 5 years after the initial CS without treatment. The participants were divided into 2 groups: those with untreated diminutive adenomas (group A) and those with no adenomas (group B). The cumulative incidence of ACN and the hazard ratio were assessed using Gray's test and the Fine and Gray model. RESULTS: During the median follow-up period of 60.9 months, 21 ACNs were detected. The 5-year cumulative incidences of ACN in group A (n = 361) and group B (n = 1,017) were 1.4% (95% confidence interval [CI]: 0.5-3.4) and 0.8% (95% CI: 0.3-1.7), respectively, without a statistically significant difference (P = 0.23). No ACNs developed from unresected adenomas. The smoking status was significantly associated with the incidence of ACN, and the hazard ratio for ACN in group A vs group B adjusted for smoking status was 1.43 (95% CI: 0.52-3.90; P = 0.48). DISCUSSION: The low 5-year cumulative incidence of ACN suggests the potential to adopt the "diagnose-and-do-not-resect" strategy as an alternative option for diminutive adenomas not requiring excessive surveillance.

Clinical significance of colorectal polyp detection on colonoscopy insertion.

Background: Colorectal lesions are generally evaluated during the withdrawal phase of colonoscopy. Minimising the risk of missed lesions is crucial to determine an appropriate future surveillance colonoscopy interval. Objective: This study aimed to evaluate the clinical significance of detecting sigmoid colon lesions during the insertion phase. Methods: This retrospective study included 172 consecutive patients undergoing colonoscopy between October 2017 and April 2018. The total number of detected polyps, mean polyps per procedure, mean polyps per positive procedure, and histological and clinical characteristics of detected lesions were recorded. The primary endpoint was the difference in sigmoid colon polyp detection rates during insertion and withdrawal. Results: A total of 172 colonoscopies were performed for each patient and 322 lesions were detected. Sixty-two (19%) polyps were detected during insertion, 312 (97%) during withdrawal, and 52 (16%) during both insertion and withdrawal. Although all polyps except for those in the sigmoid colon could be detected during withdrawal, 10 of 87 (11%) polyps in the sigmoid colon could only be detected during insertion. Conclusions: In this study, attempts to detect polyps, even in the insertion phase, showed the clinical significance to decrease the risk of missed adenomatous polyps in the sigmoid colon.

Endoscopic Resection of Duodenal Carcinoid Tumors: A Single-Center Comparison Between Simple Polypectomy and Endoscopic Mucosal Resection.

OBJECTIVES: Endoscopic resection is preferred for duodenal carcinoids less than 20 mm; however, the efficacy of simple polypectomy has not been compared with advanced endoscopic resection techniques. METHODS: We performed a retrospective review of 33 patients who underwent endoscopic duodenal carcinoid resection (10 simple, 23 endoscopic mucosal resection) at the Hospital of the University of Pennsylvania between January 1, 2006, and June 15, 2017. The primary outcomes were resection margin positivity and local tumor recurrence. RESULTS: There were no significant differences in demographics or tumor functionality. Lesions managed with simple polypectomy had smaller median gross specimen size (6.0 mm vs 8.0 mm, P = 0.043). There was no significant difference in pathology resection margins between simple polypectomy and endoscopic mucosal resection (86% vs 68% positive, P = 0.64). Local recurrence on surveillance endoscopy was also similar (14.3% vs 17.7%, respectively; P = 1.000), with median time to recurrence 2.3 months (interquartile range, 1.2-5.4 months). The median follow-up time in patients without local recurrence was 21.4 months (interquartile range, 7.1-39.6 months). CONCLUSIONS: Simple polypectomy may be adequate treatment of small duodenal carcinoids, although further studies are needed for validation and to define the upper limits of tumor size that can be managed with this technique.

EUS-guided fiducial placement for GI malignancies: a systematic review and meta-analysis.

BACKGROUND AND AIMS: Image-guided radiotherapy (IGRT) allows the delivery of radiation with high precision to a target lesion while minimizing toxicity to surrounding tissues. EUS provides excellent visualization of GI tumors and consequently is being used for fiducial placement with increased frequency. Our goal was to perform a systematic review and meta-analysis of studies evaluating the technical aspects, safety, and efficacy of EUS fiducial placement for IGRT in GI malignancies. METHODS: A systematic literature search was carried out in the following databases: Medline, PubMed, Embase, Web of Science, and Cochrane Library, using Medical Subject Headings terms combined with text words. A random effects model was used to determine pooled proportions of technical success, migration, and adverse event rates. Heterogeneity was assessed using the I2 statistic. Publication bias was assessed visually using a funnel plot and by the Begg and Egger tests. RESULTS: Nine full articles and 5 abstracts reporting on 1155 patients, 49% from a single study by Dhadham et al, were included in the meta-analysis. The pooled rate of technical success was 98% (95% confidence interval [CI], 96-99). Moderate heterogeneity (I2 = 34.18) was present, which appeared to be due to variable sample sizes. Publication bias was present, suggesting that studies with less-substantial outcomes may have not been reported (Begg test, P = .87; Egger test, P < .01). Pooled rates for fiducial migration and adverse events were 3% (95% CI, 1.0-8.0) and 4% (95% CI, 3-7), respectively. CONCLUSIONS: Our meta-analysis showed that EUS-guided insertion of gold fiducials for IGRT is technically feasible and safe. Further controlled studies assessing its long-term effectiveness in GI malignancies are needed.