Acute hemorrhage from upper gastrointestinal (UGI) ulcers is a major cause of morbidity and mortality, as well as a common medical emergency. The incidence of cardiovascular or cerebrovascular diseases is increasing along with the increasing population of elderly individuals in Japan, and such diseases are usually treated with low-dose aspirin (LDA). However, LDA is one of the most important risk factors for bleeding UGI ulcers,1 and the prevalence of bleeding peptic ulcers associated with LDA is gradually increasing.2 Although endoscopic hemostasis for UGI bleeding is generally acceptable, it can be difficult to completely achieve in some patients, and excessive hemorrhage from UGI ulcers can be fatal. Endoscopic treatment of UGI bleeding has recently advanced along with the administration of high-dose intravenous proton pump inhibitors (PPIs). Despite improvements in endoscopic hemostasis and pharmacological therapies, UGI ulcers rebleed in 10% to 20% of patients,3,4,5,6 and those without early endoscopic reintervention or definitive surgery might be at a high risk for mortality. In fact, mortality rates related to ulcer bleeding have remained essentially unchanged at 5% to 8% during the past three decades.4,7

Therefore, determining which factors are involved in rebleeding after an initial endoscopic hemostasis is extremely important for patients with bleeding UGI ulcers. In addition, understanding the factors that contribute to intractable or insufficient initial endoscopic hemostasis is needed to advance the management of such ulcers.

Here, we aimed to define which factors are associated with intractability to endoscopic hemostasis in patients with bleeding UGI ulcers.

The Institutional Review Board of Aichi Medical University Hospital approved this retrospective, single-center study, which was performed at the Department of Gastroenterology at Aichi Medical University Hospital in Japan. Between April 2000 and December 2010, 428 patients with hematemesis, melena, or both, due to bleeding UGI ulcers underwent emergency endoscopy within 24 hours of arrival at Aichi Medical University Hospital, and all were admitted thereafter. According to the Forrest classification,8 patients with the following bleeding types were indicated for endoscopic hemostasis: Ia, spurting; Ib, oozing; IIa, nonbleeding visible vessels; IIb, adherent blood clots. Ulcers with stages IIc (black base) or III (clear ulcer base) were excluded from endoscopic hemostasis. Only patients with active bleeding ulcers (Forrest types Ia and Ib) and ulcers with stigmata of a recent hemorrhage (Forrest types IIa and IIb) were ultimately enrolled. We excluded patients with bleeding from a nonulcer etiology (varices, hemorrhagic erosive gastritis, Mallory-Weiss tears, vascular ectasia, or malignancies), coagulopathy, or a history of gastrectomy.

Endoscopic hemostasis was achieved by injecting hypertonic saline-epinephrine solution (HSE)9 or ethanol, using argon plasma coagulation (APC),10 hemostatic clips,11,12 or both HSE and hemostatic clips, as needed. After the initial endoscopic hemostasis, patients received acid-suppressive treatment with histamine-2 blockers or PPIs. A scheduled second-look endoscopy was performed within 24 hours after the initial endoscopic hemostasis when symptoms that are considered to indicate rebleeding from UGI ulcers, such as recurrent hematemesis or hematochezia, fresh blood outflow in nasogastric aspirate, or circulatory instability such as shock, were absent. Emergency endoscopy was done immediately after the initial endoscopic hemostasis when a clinical suspicion of rebleeding was present. Forrest types Ia, Ib, IIa, and IIb at the scheduled or emergency endoscopy were defined as being intractable to endoscopic hemostasis, and a second endoscopic hemostatic procedure was scheduled. Endoscopic hemostasis of Forrest types IIc and III at the scheduled or emergency endoscopy was considered successful, and ulcers that were considered not to require further hemostasis were not treated endoscopically again. After a second attempt at hemostasis, a scheduled or emergency endoscopy was done by using the same strategy as in the second-look endoscopy. Patients underwent a third endoscopic hemostatic procedure if necessary. Endoscopic hemostasis was repeated until all exposed vessels with Forrest types Ia, Ib, IIa, or IIb changed to types IIc or III. Emergency surgery or transarterial embolization (TAE), or death due to vascular comorbidities, was also defined as intractable endoscopic hemostasis. Durable hemostasis was defined as a successful hemostatic procedure and the absence of bleeding indicators during hospitalization. We retrospectively documented the patients' age, sex, history of comorbid illnesses, history of UGI ulcers, smoking, alcohol consumption, LDA use with or without antithrombotic drugs, steroid drugs, and nonsteroidal anti-inflammatory drugs (NSAIDs), initial hemodynamic status including systolic blood pressure and heart rate, and initial hemoglobin and serum albumin levels. Hemodynamic instability diagnosed as a shock was defined as a systolic blood pressure of <90 mm Hg or a heart rate of >100 beats per minute. We evaluated the ulcer location, size, and number; size of exposed vessels on the bottom of the ulcer; and Forrest bleeding patterns. Bleeding ulcers were located in areas described as the upper (U), middle (M), and lower (L) parts of the stomach, the duodenal bulb (DB), and the second portion of the duodenum (DS), to determine factors associated with bleeding ulcers.

To determine the factors involved in intractability to the initial endoscopic hemostasis, we compared patients whose bleeding UGI ulcers were successfully treated with the hemostasis with those who were intractable.

We found that endoscopic hemostasis was a useful initial approach to treating bleeding UGI ulcers, and we identified the risk factors for intractability to this procedure. Among 428 patients with UGI bleeding ulcers treated with immediate endoscopic hemostasis, 354 acquired durable hemostasis, and 74 (17.3%) did not. We used multivariate analysis of data from these groups to evaluate factors associated with intractability to endoscopic hemostasis. The results identified age ≥70 years, shock on admission, hemoglobin <8.0 mg/dL, serum albumin <3.3 g/dL, exposed vessels with a diameter of ≥2 mm on the bottom of ulcers, and Forrest types Ia and Ib bleeds as risk factors for intractability to endoscopic hemostasis.

Several studies have examined risk factors for intractability to endoscopic hemostasis. Rockall et al.13 found that comorbidities such as heart failure, ischemic heart disease, and renal failure were factors indicating a high risk for rebleeding. A systematic review of 10 articles14 revealed that predictors for rebleeding after endoscopic hemostasis comprised hemodynamic instability, comorbidity, active bleeding, and large ulcers. We found here that hemodynamic instability presenting as shock and active bleeding presenting as Forrest types Ia and Ib bleeding were predictors of intractability to endoscopic hemostasis. Our results therefore agreed with the reported findings.14 However, comorbid illness was not a risk factor for intractability to endoscopic hemostasis, which differed from previous findings.14,15 The reason for this might be associated with the strategies used to achieve hemostasis and that the number of hemostatic approaches was limited in the published clinical trial.14 We found that patients aged ≥70 years and hemoglobin <8.0 mg/dL were also independent risk factors for intractability to endoscopic hemostasis. These findings are broadly compatible with those of three other studies.16,17,18 Previous evaluations of risk factors among patients with nonvariceal UGI bleeding, by using multivariate analysis, included age >65 years16 and low initial hemoglobin values among the clinical predictors of increased risk for persistent or recurrent bleeding.17,18 Many patients are administered with LDA as an antithrombotic therapy for the secondary prevention of cerebral and/or myocardial vascular events. However, LDA intake is considered an important risk factor for UGI bleeding, as the relative risk is about 2-fold higher and it is not related to either the formulation (plain or entericcoated) or dosage of LDA.19 Among 119 of our patients who were administered with LDA, 26 (21.8%) were intractable to the initial endoscopic hemostasis. However, LDA, steroids, and NSAIDs were not risk factors for intractability to endoscopic hemostasis, which agreed with previous findings.20,21,22 Serum albumin <3.3 g/dL was a predictor of intractable endoscopic hemostasis. Hypoalbuminemia might reflect a reduction in serum coagulation factors. Both hypoalbuminemia and a loss of serum coagulation factors might delay mucosal healing after injury and result in intractability to endoscopic hemostasis. We believe that serum albumin <3.3 g/dL is a novel predictor of intractable endoscopic hemostasis for UGI bleeding ulcers. Vessels with a diameter of ≥2 mm exposed on the bottom of ulcers also predicted intractable endoscopic hemostasis. Precisely treating large exposed vessels with hemostasis is generally difficult, and HSE injection, APC, and hemostatic clips might be insufficient. Thus, larger exposed vessels might not disappear so rapidly after endoscopic hemostasis, resulting in intractability. Although a systematic review related endoscopic hemostatic intractability with bleeding location, such as posterior duodenal or lesser gastric curve ulcers,14 we found that bleeding locations did not significantly differ between a successful initial and an intractable endoscopic hemostasis. That might be because both the skill of those implementing the endoscopic hemostatic procedures and the devices used for endoscopic hemostasis have improved recently. Therefore, mature endoscopic hemostasis would reduce the likelihood of a relation between bleeding location and endoscopic hemostatic intractability.

Since various factors contribute to intractability, careful observation after endoscopic hemostasis is needed, particularly for patients at a high risk for incomplete hemostasis. Our patients with Forrest types IIc and III never underwent additional further endoscopic hemostasis, and none of them developed symptoms of rebleeding during hospitalization. These results indicated that types IIc and III vessels do not require treatment with endoscopic hemostasis. Diagnosing vessels on UGI ulcers based on the Forrest classification might avoid ineffective endoscopic treatment and unnecessary observational endoscopy, and thus reduce medical care expenses.

Recently, TAE has been considered as an alternative to salvage surgery. Some retrospective studies have compared TAE with surgery in patients for whom endoscopic treatment failed.23,24,25,26,27 All of these studies found that TAE reduced the number of complications and the need for surgery without increasing overall mortality. Although several studies have demonstrated that TAE is useful for treating acute hemorrhage from UGI ulcers,28,29,30 the choice of TAE or surgery after a failed endoscopic treatment depended on the discretion of the operating surgeon or physician. Moreover, the selection of TAE or surgery after a failed endoscopic treatment might also be affected by institutional TAE limitations. We attempted TAE in only one patient for whom a third endoscopic hemostasis failed. This was a limitation of treatment for bleeding UGI ulcers at Aichi Medical University Hospital.

In conclusion, age ≥70 years, shock on hospital admission, hemoglobin <8.0 mg/dL, serum albumin <3.3 g/dL, exposed vessels with a diameter of ≥2 mm on the ulcer bottom, and Forrest types Ia and Ib bleeds were identified as independent risk factors associated with an intractable initial endoscopic hemostasis in patients with peptic UGI bleeds. Additional studies are required to validate our results and document the impact of initial risk stratification for intractability to endoscopic hemostasis for the emergency management of UGI bleeding.