The study enrolled consecutive patients undergoing EMR for gastrointestinal lesions at least 1 cm in diameter at Zhengzhou Central Hospital between January 2025 and March 2026. Eligibility required clear vascular structures and blood-flow signals at the lesion base on EUS, including a vessel at least 1 mm in diameter. Patients with suspected deep infiltration or metastasis, severe organ insufficiency, pregnancy or lactation, or an epinephrine allergy were excluded. Upper- and lower-gastrointestinal-tract lesions were pooled despite differences in morphology, histopathology, procedural technique and bleeding risk.

Treatment was not randomized. At the discretion of the operating endoscopist, patients could receive targeted epinephrine injection when color Doppler imaging showed a clearly delineated feeding vessel; the remaining patients received non-targeted saline injection. The injection group received 1:20,000 epinephrine at multiple points adjacent to feeding vessels under real-time EUS guidance, using a 19-gauge needle and a total dose of 1.0–3.0 mL. Controls underwent the same EUS assessment but received conventional submucosal saline without epinephrine. Both groups then underwent snare EMR and routine hemoclip closure of the wound surface.

The primary outcome was visible oozing or overt active bleeding from the wound during the procedure, as documented in endoscopy reports and re-reviewed for the analysis. Bleeding severity was not routinely graded, and the endpoint did not require additional hemostatic intervention. Secondary outcomes included delayed bleeding within 14 days, procedure time, hospital stay, en bloc resection and endoscopy-related adverse events. Follow-up continued for 14 days or until April 2026, and no patients were reported lost to follow-up.

Multivariable logistic regression adjusted for hypertension and Yamada lesion morphology. Targeted epinephrine injection was associated with lower odds of immediate bleeding, with an odds ratio of 0.112, 95% confidence interval 0.028–0.443 and P = 0.002. Firth penalized regression produced a similar estimate of OR 0.14, 95% CI 0.03–0.46, P = 0.0008. Propensity-score matching gave an OR of 0.143, 95% CI 0.026–0.785, while inverse probability of treatment weighting produced an OR of 0.081, 95% CI 0.018–0.359. Measured covariates were balanced after the propensity-based adjustments, with all post-adjustment standardized mean differences below 0.10.

Only 13 bleeding events occurred in the cohort, giving the multivariable model an events-per-variable ratio of approximately 2.6. The authors cautioned that adjusted odds ratios may be exaggerated and that the analyses were exploratory. Residual confounding by indication could not be excluded, particularly because intraoperative vascular density was not quantified in the medical records. Because every patient received EUS assessment, the study could not separate the contributions of EUS guidance, epinephrine and targeted injection.

Procedure time and hospital stay did not differ significantly between groups. Median procedure time was 25 minutes in the injection group, with an interquartile range of 16–45 minutes, and 26 minutes in the control group, with an interquartile range of 18–53 minutes; P was 0.871. Median hospital stay was five days in both groups, P = 0.751. En bloc resection was achieved in 100% of both groups. These observations show no measured increase in procedure time or stay, although the study was not designed to assess cost-effectiveness.

Two delayed bleeding events occurred among controls, or 3.8%, and none occurred in the injection group. No statistical comparison was performed because the cohort was underpowered for this endpoint, and routine clip closure in every patient prevented assessment of delayed bleeding independently of epinephrine. Endoscopy-related adverse events occurred in two injection-group patients, or 6.5%, and five controls, or 9.4%, with no significant between-group difference. The injection-group events were mild laboratory abnormalities. No perforation, severe epinephrine-related cardiopulmonary reaction or conversion to surgery occurred, but the absence of such events among 31 treated patients cannot establish safety.

The study was retrospective, non-randomized and limited to one high-volume center, where all operators had experience with at least 300 EMR procedures and at least 100 EUS puncture procedures. Upper- and lower-tract lesions were analyzed together, bleeding severity was not systematically recorded, and follow-up was restricted to 14 days. A history of hypertension also appeared inversely associated with bleeding, but the authors considered this unexpected, biologically unexplained and potentially affected by residual confounding or statistical instability. The findings require confirmation in prospective randomized studies that separately evaluate EUS guidance and epinephrine.

Subject of Research: A retrospective cohort study of EUS-guided targeted epinephrine injection during endoscopic mucosal resection in 84 patients with gastrointestinal lesions.

Article Title: EUS-guided targeted epinephrine injection and immediate intraoperative bleeding during endoscopic mucosal resection: a retrospective cohort study (with video)

Article References: Wang, W., Rong, A., Liao, C., Chen, H., Han, J., & Liang, Q. (2026). EUS-guided targeted epinephrine injection and immediate intraoperative bleeding during endoscopic mucosal resection: a retrospective cohort study (with video). Frontiers in Medicine, 13, Article 1985945. https://doi.org/10.3389/fmed.2026.1985945

Image Credits: AI Generated

DOI: 10.3389/fmed.2026.1985945

Keywords: endoscopic mucosal resection, endoscopic ultrasound, epinephrine, intraoperative bleeding, retrospective cohort study, gastrointestinal lesions, propensity score analysis, hemostasis, EUS-guided, targeted, injection, immediate