Attaching a transparent cap to the endoscope tip can improve visualization by focusing the endoscopic field and reducing blind spots [30]. It can also facilitate clip deployment in anatomically challenging locations [31]. When the lesion moves with respiration, stabilizing the cap against the gastric wall may help maintain a steady position [25]. These advantages are particularly helpful in the duodenum, where luminal angulation and prominent folds can make identification of the bleeding source and precise hemostasis difficult.

A water-jet system can rapidly irrigate the blood, improve identification of the bleeding source, and help dislodge adherent clots using water pressure [25]. It can shorten the procedure time and confirm whether the bleeding vessel is adequately compressed when using hemostatic forceps.

An overtube is useful when repeated insertion of the endoscope is anticipated, such as in multiple-band ligation, when clots need to be removed, or when the lens requires cleaning. This may reduce the risk of aspiration. However, overtube placement can cause complications, including laryngopharyngeal or esophageal mucosal injury, perforation, vocal cord paralysis, and pneumomediastinum [32]. The endoscope should first be inserted, and the overtube should then be advanced over the endoscope. Adequate lubrication of the overtube tip is necessary, and the overtube should be gently inserted without resistance.

Carbon dioxide is rapidly absorbed into the bloodstream after intraluminal insufflation and can reduce post-procedural abdominal pain and distension. When used during hemostasis, it may reduce patient discomfort and belching [33]. It may also be advantageous if perforation occurs, which is a potential complication of electrocautery. However, room air is preferred for patients with chronic respiratory diseases, obstructive sleep apnea, moderate-to-severe obesity, or known carbon dioxide retention [25].

Certain gastric locations, including the lesser curvature or posterior wall of the body, the cardia, and the lesser curvature of the antrum, are technically demanding for endoscopic hemostasis because of limited access. A multibending endoscope has two bending sections at the distal end, which can facilitate a closer approach to difficult sites [25,34]. The two channels of this endoscope also provide practical advantages. One can be used to suction fluid or clots, while the other is used for the insertion of endoscopic devices [25].

Erythromycin is an antibiotic with a prokinetic effect that acts as an agonist of motilin receptors in the gastroduodenal smooth muscle. It has been used in UGIB before endoscopy to promote gastric emptying [38]. In a meta-analysis including eight studies, pre-endoscopic erythromycin improved visualization of the gastric mucosa and reduced the need for repeat endoscopy and the length of hospital stay [39]. US guidelines recommend its use to promote gastric emptying before endoscopy in UGIB [19], and European guidelines recommend selective use in patients with massive active bleeding [20]. However, injectable erythromycin is not available in Korea. The evidence for intravenous metoclopramide is inconsistent, and a meta-analysis found no significant effect on adequate mucosal visualization, transfusion requirements, length of stay, or the need for repeat endoscopy [40].

Nasogastric tube insertion and lavage have long been used in patients with suspected or confirmed UGIB, although evidence of its clinical benefits is limited. The rationale is to confirm the presence of UGIB, remove blood and clots, reduce aspiration risk during endoscopy, and improve the endoscopic field [41,42]. In a randomized controlled trial of 39 patients with UGIB, nasogastric lavage of up to a maximum of 15 L until the effluent became clear improved visualization compared with controls. However, it did not improve the identification of the bleeding source, hemostasis success, rebleeding, the need for repeat endoscopy, length of stay, or mortality [43]. A subsequent large retrospective study also found that nasogastric lavage did not improve mortality, length of stay, surgery rates, or transfusion requirements [41]. Although nasogastric lavage is commonly performed in emergency settings, it can cause substantial discomfort and carries risks of complications without clear clinical benefit. Therefore, routine nasogastric lavage is not recommended in patients with UGIB [44]. However, the diagnostic and decompressive effects of nasogastric tube placement may still be clinically relevant in selected situations.

When visualization is limited by clots in the gastric fundus, changing the patient’s position may allow clots to move inferiorly, improving the endoscopic field [35]. In practice, repositioning to a supine or right lateral decubitus position may be difficult, and aspiration risk may increase. In a report of patients in whom fundus inspection was difficult, elevating the upper body while maintaining a left lateral position allowed the clot to shift toward the gastric body under gravity in all 44 patients, thereby permitting fundus visualization and enabling detection of additional bleeding lesions [36]. Mori et al. [45] reported a case in which a newly developed U-shaped overtube was used to facilitate safe and effective repositioning to a right lateral decubitus position, allowing successful endoscopic hemostasis.

The Olympus EVIS X1 system includes a red dichromatic imaging (RDI) function and an image-enhanced endoscopy modality (Fig. 1). RDI can facilitate the identification of bleeding points and prominent submucosal vessels [50]. RDI may be considered when available.

Machine learning models developed for risk prediction in UGIB integrate a broad range of clinical variables, including demographics, vital signs, comorbidities, medication use, and laboratory results [51]. Representative methods include Random Survival Forests, Regularized Cox regression, XGBoost, and long short-term memory models [52]. These models have been reported to outperform conventional risk scores, including identifying patients at low risk, predicting mortality, assessing bleeding risk, and estimating transfusion requirements [51]. Wider clinical adoption will likely require further validation and cost-effectiveness evaluation.

A swallowed bleeding sensor is a device designed for the rapid and accurate detection of UGIB. It consists of an optical sensor capsule that the patient swallows and an external receiver that displays data in real time [51]. In a prospective study by Akiki et al. [53] involving 126 patients, the presence of blood was detected within 10 minutes with a sensitivity of 92.9% and specificity of 90.6%, without notable adverse events. The device may be useful when rapid triage of suspected UGIB is needed; however, it does not provide information on the bleeding location or volume, and additional cost-effectiveness studies are required [51].

Endoscopic ultrasound (EUS) and Doppler imaging for endoscopic hemostasis are advantageous for vascular assessment. They have been reported to improve success rates and cost-effectiveness in peptic ulcers or Dieulafoy lesions [54], and have high success rates in gastric varices [55]. However, in emergency endoscopy settings, these techniques may be difficult to perform. They also have limitations, including false-positive findings and difficulty in assessing blood flow in vessels deeper than 4 mm. They require operators with EUS expertise, and whether the published cost-effectiveness findings translate into Korean healthcare settings remains uncertain.

Diluted epinephrine can achieve rapid initial hemostasis and improve the visualization of active bleeding by tamponade from the injected volume, vasoconstriction, and subsequent thrombus formation [57,58]. Compared to pharmacotherapy alone, epinephrine injection provides benefits; however, its effect is often transient and is associated with higher rebleeding rates when used alone. Therefore, it is generally combined with other hemostatic modalities, such as mechanical therapy or thermal coagulation [58,59].

The injection catheter is introduced through the working channel and connected to a syringe. Diluted epinephrine, most commonly at a concentration of 1:10000 or 1:20000 in normal saline, is injected into four quadrants around the bleeding site, typically 0.5 to 2 mL per injection [20]. When the injection is effective, bleeding decreases or stops, and a vessel or any adherent clot may become more clearly visible, and the surrounding mucosa may appear blanched. Inadvertent intra-arterial epinephrine injections can cause marked hypertension [60].

Cyanoacrylate, a tissue adhesive widely known as histoacryl, is most commonly injected into bleeding gastric varices. In a meta-analysis comparing cyanoacrylate injection with band ligation for gastric variceal bleeding, initial hemostasis rates were similar, but rebleeding rates were lower (relative risk [RR] 0.6) [61]. Two randomized studies also reported better treatment outcomes with cyanoacrylate injections than with band ligation, without a difference in complication rates [62,63].

During the procedure, cyanoacrylate is often mixed in a 1:1 ratio with lipiodol to delay polymerization, and the injection volume is selected according to the varix size. To reduce the risk of damage to the endoscopic channel, a neutral oil, commonly olive oil, can be instilled in advance (approximately 5 mL), and relatively large-bore (21- to 22-gauge) and longer (5 mm) injection needles are used. After flushing the catheter with 1–2 mL of distilled water, the varix is punctured while maintaining an endoscope-to-varix distance of approximately 3–5 cm. Cyanoacrylate is injected over 4–5 seconds, followed by a distilled water flush to deliver any material remaining in the catheter, and the needle is then promptly withdrawn. Suction should be avoided because the polymerized cyanoacrylate can cause channel obstruction. A post-procedure radiograph can confirm the appropriate distribution of the injected material and assess complications, such as pulmonary or splenic embolic events [58].

Hemostasis with clips is achieved by firmly grasping the target vessel and surrounding tissue to provide mechanical compression [64]. Through-the-scope (TTS) clips are delivered through the working channel of the endoscope either as a preloaded device or by mounting the clip immediately before use. A wide range of clip designs is available, differing in metal composition, jaw length, angulation, rotatability, and ability to reopen and close the clips. For hemostasis, a relatively short clip with jaws angled at 135° may be advantageous, particularly in firm or fibrotic lesions where secure anchoring is needed. The characteristics of commonly used TTS clips in Korea are summarized in Table 2.

In a meta-analysis by Sung et al. [65], TTS clipping, alone or in combination, achieved superior hemostasis compared with epinephrine injection alone. A major advantage of TTS clipping is less tissue injury. The limitations include the need for a skilled assistant and suboptimal first placement that can worsen bleeding and complicate subsequent therapy [64]. Technical difficulty can also be encountered in areas such as the gastric fundus, lesser curvature of the gastric body, posterior wall of the duodenal bulb, lesions with marked fibrosis, or large-caliber exposed vessels [66,67].

Accurate clip placement generally benefits from maintaining a perpendicular orientation towards the wall of the target site. Gentle suction during deployment can bring the target closer to the endoscope tip and may be helpful. When retroflexion is required, introducing a clip device into the channel before retroflexion may be helpful. Moving the endoscope while the clip is open may cause mucosal injury or clip deformation; therefore, caution is warranted [64].

An over-the-scope (OTS) clip is a large-diameter clip (8.5–14 mm) mounted on a cap at the endoscope tip. It has a clawlike configuration and is deployed in a manner similar to band ligation. Tissue is suctioned into the cap, and the clip is released by rotating an external hand wheel connected to the device [67]. After deployment, the final diameter is 14.6–21 mm [59]. OTS clips are typically made of nitinol, a shape-memory alloy composed of nickel and titanium, which provides strong elastic recoil and enables thick tissue capture over a broad area [58]. They may be useful for large ulcers (>2 cm), large vessels, and chronic ulcers [67].

In a prospective randomized study of rebleeding after initial endoscopic hemostasis, OTS clips reduced further bleeding compared with standard therapy (15.2% vs. 57.6%) [68]. In a meta-analysis evaluating OTS clips as first-line therapy for nonvariceal UGIB, they reduced bleeding (odds ratio [OR] 0.32, 95% confidence interval [CI] 0.17 to 0.59), but no difference was observed in 30-day mortality (OR 0.60, 95% CI 0.23 to 1.56) [1]. US guidelines recommend OTS clips for rebleeding [19], and European guidelines recommend it as primary therapy in ulcers larger than 2 cm or with severe fibrosis [20].

The limitations of OTS clips include the need to withdraw the endoscope to mount the device, increased endoscope tip diameter, difficulty with subsequent therapy if misdeployment occurs, and higher costs [64,67]. For effective deployment, close apposition between the cap and lesion is required. Grasping devices may be used to capture tissue into the cap before suction [64]. Long-term data are needed to clarify the durability and potential adverse effects. Under the current Korean National Health Insurance system, OTS clips are only reimbursed for the closure of gastrointestinal perforations.

Band ligation is primarily used for esophageal varices; however, it has also been applied to vascular ectasia, Dieulafoy lesions, Mallory–Weiss tears, and ulcer bleeding with limited fibrosis. A hood containing a rubber band is mounted on the endoscope tip. After approaching, the target is suctioned into the cap, an assistant insufflates air through a tube connected to the device using a syringe, and the band is released. It provides mechanical compression and a tamponade effect [20,25]. Effective therapy depends on adequate suction and perpendicular positioning of the cap to the lesion [25], and suction is maintained gently until the endoscopic view becomes fully occupied by mucosa (the “red-out” sign). Earlier devices required single-band loading; however, multiband ligation systems are now available, allowing repeated ligation without removing the endoscope.

Covered self-expandable metallic stents (SEMS) have been studied for the treatment of refractory esophageal variceal bleeding. In a multicenter randomized controlled trial involving 28 patients with refractory variceal bleeding, SEMS demonstrated better outcomes than balloon tamponade using a Sengstaken–Blakemore tube in terms of mortality, bleeding control, serious procedure-related complications, and transfusion requirements [69]. Stents can also be maintained for longer periods and cause less patient discomfort. However, under the current Korean National Health Insurance System, SEMS use for esophageal variceal bleeding is not reimbursed.

Endoscopic purse-string sutures using clips and a detachable snare have been reported as a rescue therapy for nonvariceal UGIB refractory to conventional methods, with promising success rates in small studies [70,71]. This technique uses TTS clips and a detachable snare without requiring specialized equipment. It can be applied to large ulcers and is relatively cost-effective. The limitations include a longer procedure time when multiple clips are required, the need for an experienced endoscopist, and difficulty in anatomically challenging locations.

Argon plasma coagulation (APC) is a representative noncontact modality. The ionized argon gas generated at the probe tip delivers a high-energy monopolar current to the tissue, resulting in dehydration and coagulation [56]. As tissue dehydrates, electrical conductivity decreases. For this reason, APC has a limited effect on large vessels or deeper tissue and is mainly used for superficial vessels [25,56,59]. In a meta-analysis that included two randomized controlled trials, APC monotherapy for peptic ulcer bleeding showed comparable efficacy to other endoscopic modalities such as heater probe, hemostatic forceps, and clipping (97% vs. 95%) [72]. In addition, combination therapy using epinephrine injection followed by APC did not differ in initial hemostasis or rebleeding compared with other combination approaches [73-75].

For the procedure, an APC catheter is introduced through the endoscope channel and positioned perpendicular to the target area, followed by activation using the foot pedal [25]. The optimal probe-to-tissue distance is reported to be 2–8 mm, and coagulation becomes ineffective if the probe is too far. The coagulation effect can be adjusted through settings such as power (W), argon gas flow rate (L/min), distance to the tissue, and application time [20,56]. As argon gas can cause luminal distension and may obscure the field, intermittent suction is often required. Caution is required in the duodenum because of the risk of gas-related distension and perforation. The risk of perforation is influenced by power settings, application time, and probe-to-tissue distance.

In monopolar electrocautery, the active electrode makes contact with the target lesion, and a grounding pad attached to the patient serves as the return electrode. A grounding pad is not used for the bipolar electrocautery. These contact methods achieve hemostasis by generating heat or delivering currents, leading to tissue edema and activation of the coagulation cascade [56].

Hemostatic forceps resemble biopsy forceps and achieve hemostasis through direct contact with the tissue. Soft coagulation uses a lower voltage, which may reduce the risk of perforation. In a study of 50 patients with nonvariceal UGIB, bipolar hemostatic forceps showed better hemostatic outcomes than clips (100% vs. 78%) [76]. In high-risk peptic ulcers (Forrest Ia, Ib, and IIa), monopolar hemostatic forceps were associated with higher hemostasis success, shorter procedure times, and lower rebleeding rates compared with TTS clips [77]. A recent meta-analysis also reported that monopolar hemostatic forceps provided higher hemostasis success and lower rebleeding rates than TTS clips, APC, or heater probes [78].

The device is rotatable and is useful for treating spurting bleeding or exposed vessels. Hemostasis can be achieved either by opening the forceps to grasp a vessel or by keeping it closed and in contact with it. Gentle traction on the grasped tissue may reduce deep-tissue injury. The coagulation efficacy may decrease if blood, clots, or water remain.

Endoscopic findings in bleeding peptic ulcers are commonly classified using the Forrest classification (Table 4 and Fig. 2), which is associated with rebleeding risk and prognosis, and guides endoscopic therapy. Endoscopic treatment is recommended for spurting (Forrest Ia), oozing (Forrest Ib), and a non-bleeding visible vessel (Forrest IIa) [18-20]. In a meta-analysis including [19] randomized controlled trials, Laine and McQuaid [86] reported that endoscopic therapy was associated with a lower risk of rebleeding in Forrest I lesions (RR 0.29) and Forrest IIa lesions (RR 0.49). Forrest IIb lesions remain debated. In a meta-analysis by Laine and McQuaid [86], endoscopic treatment did not reduce the risk of persistent bleeding or rebleeding in Forrest IIb lesions. A recent meta-analysis performed by Beran et al. [87] concluded that endoscopic therapy for Forrest IIb lesions was associated with lower rates of rebleeding, mortality, and surgery than conservative therapy alone. However, another meta-analysis conducted by Tassone et al. [88] described that endoscopic hemostasis reduced recurrent bleeding from peptic ulcers with adherent clots compared with medical management alone, but there was no difference in the rates of mortality or need for surgery. Further large-scale randomized controlled trials are required to validate these findings. The current American and European guidelines neither recommend nor oppose endoscopic therapy for Forrest IIb lesions. They recommended vigorous irrigation to remove the adherent clot [19-20]. Immediate endoscopic hemostasis should be performed if a high-risk stigma is identified. Removal of an adherent clot can clearly identify the characteristics of the bleeding lesion but can also convert the lesion into spurting or oozing bleeding, which may complicate subsequent hemostasis. The decision for endoscopic therapy in Forrest IIb lesions should consider multiple factors, including the endoscopist’s judgment and the availability of surgical or interventional radiology backup [18-20]. High-dose PPI therapy with follow-up endoscopy can be considered, or if endoscopic therapy is pursued, diluted epinephrine can be injected around the lesion, followed by cautious attempts at clot removal. Forrest IIc and III lesions have a low risk of rebleeding and generally do not require endoscopic treatment.

Depending on the lesion characteristics, clip monotherapy, electrocautery-based therapy alone, or combination therapy with epinephrine injections may be used. Epinephrine injection alone is not recommended because of high rebleeding rates [19]. When combined with clips or electrocautery-based methods, rebleeding can be reduced [86].

Bleeding from Mallory–Weiss tears usually stops spontaneously, and massive bleeding is uncommon. However, patients receiving antiplatelet agents or anticoagulants, or those with bleeding diathesis, may develop persistent and substantial bleeding. If ongoing bleeding is present, the therapeutic options include clipping, band ligation, epinephrine injection, and APC [1].

A Dieulafoy lesion refers to an abnormally large submucosal artery (approximately 1–5 mm) that fails to taper and becomes exposed [89]. Ulceration is typically limited to the vessel apex, resulting in a small mucosal defect with a disproportionately large exposed vessel. Endoscopic findings include spurting arterial bleeding, a pulsatile exposed vessel, or an adherent clot at a small mucosal defect (often 3 mm or less), with otherwise normal surrounding mucosa. Endoscopic therapy generally follows principles similar to those for peptic ulcer bleeding, and clipping or band ligation can be effective [90].

In malignant tumor bleeding, endoscopic therapy is often a bridge to definitive treatment or palliative hemostasis. Tumor bleeding is characterized by a high rebleeding rate, limited durability of the hemostatic effect, frequent coagulopathy, and a friable tumor surface with diffuse oozing [1,81,91]. When ulcerative lesions are present, the principles used for peptic ulcer bleeding can be applied [91]. APC is frequently used, but rebleeding is common [92,93]. Topical hemostatic powders have attracted attention because they can be applied without direct contact and cover a broad surface area. Karna et al. [94] performed a meta-analysis of topical powder therapy for malignant tumor bleeding, including 16 studies and 530 patients. The initial hemostasis rate was 94.1%, and late rebleeding occurred in 11.4% of patients, which was lower than previously reported rates (28.3% to 80%). Hemostatic powder may be more effective than conventional endoscopic methods for malignant tumor bleeding; however, further research is needed to determine whether it can replace existing treatments.

Endoscopy may reveal active variceal bleeding, either spurting or oozing, or the bleeding may have stopped with the stigmata of a recent hemorrhage (Fig. 3). The Korean Association for the Study of the Liver, as well as American Association for the Study of Liver Disease and European Baveno VII consensus recommend endoscopic band ligation as the treatment of choice for esophageal variceal bleeding [21-23].

The Sarin classification is commonly used for gastric varices [55]. Gastroesophageal varices type 1 (GOV1) refers to esophageal varices that extend along the lesser curvature at the cardia, whereas GOV2 refers to extension along the greater curvature at the cardia. Isolated gastric varices type 1 (IGV1) are confined to the fundus, whereas IGV2 refers to the varices located in the distal stomach. For gastric variceal bleeding, cyanoacrylate injection is most commonly used [95]. There are reports of favorable outcomes with EUS-guided cyanoacrylate injection or coil embolization [96,97]. However, evidence remains limited, and these approaches require EUS expertise, which may restrict widespread adoption in routine clinical practice.

Duodenal variceal bleeding (Fig. 4) should be considered and carefully evaluated in patients with cirrhosis presenting with melena or hematochezia but without esophageal or gastric variceal bleeding. The endoscopic treatment included cyanoacrylate injection, band ligation, and clipping [98,99].