Safety and Efficacy of Sedation During Emergency Endoscopy for Upper Gastrointestinal Bleeding: A Propensity Score Matching Analysis

This retrospective chart review included patients who underwent emergency endoscopy and endoscopic hemostasis for UGIB at the National Hospital Organization Ureshino Medical Center from January 2016 to December 2021. Patients older than 20 years who fulfilled the following criteria were candidates for the study: (1) suspected UGIB requiring emergency endoscopy; (2) underwent endoscopic hemostasis within 24 h of symptom onset; (3) had a clear level of consciousness as well as stable respiratory and circulatory dynamics. Patients who had been endotracheally intubated or whose respiratory and circulatory status was sufficiently unstable to preclude emergency endoscopy and those with missing data were excluded. Written informed consent was obtained from patients who met the inclusion criteria after endoscopic hemostasis.

This study was conducted in accordance with the Declaration of Helsinki and the guidelines of the Consolidated Standards of Reporting Trials (CONSORT). The study protocol and the consent procedure were approved by the Ethics Review Committee of the National Hospital Organization Ureshino Medical Center (approval number 20-86).

The primary endpoint was the success rate of endoscopic hemostasis according to use of sedation during emergency endoscopy for UGIB. Treatment success was defined as controlled by endoscopic hemostasis and no rebleeding within 5 days after endoscopic hemostasis.

The secondary endpoints were rebleeding, procedure time, adverse events, and 30-day mortality after emergency endoscopy for UGIB. Rebleeding was defined as follows: (1) follow-up endoscopy identified recurrent UGIB or the stigmata of recent hemorrhage within 30 days or (2) melena and progressive anemia with a decrease in hemoglobin level greater than 2 g/dL and/or with a decrease in systolic blood pressure < 80 mmHg. The procedure time was defined as the time interval between insertion and removal of the endoscope [25, 26].

For urgent endoscopic treatment, each of the 17 endoscopists performed emergency endoscopy with or without sedation, based on their judgment. Thereby patients were divided into two groups: those who underwent emergency endoscopy with sedation (sedation group) and those who underwent the same procedure without sedation (nonsedation group). Sedatives used during emergency endoscopy were midazolam, diazepam, or propofol, and analgesics were pentazocine or pethidine hydrochloride [20, 27]. The type of sedative and analgesic drugs was selected by each endoscopist considering patients’ age and physical condition. Details of sedative procedures are provided in Online Appendix S1.

Vital signs including blood pressure, heart rate, and blood oxygen saturation were recorded before the introduction of sedatives. During the endoscopy, vital signs were monitored every 5 min. When oxygen saturation was below 92%, nasal oxygen supplementation (2 L/min) was initiated. When the vital signs fluctuated by 20% or more compared with baseline values, the endoscopic procedure was temporarily stopped until the recovery of those values. Flumazenil, an antagonist of benzodiazepines, was administered after endoscopy as necessary.

Endoscopic hemostasis was performed mainly by high-frequency soft coagulation, hemoclipping, or endoscopic variceal ligation (EVL) [26, 28‐30]. The choice of procedure was at the endoscopists’ discretion. Details of endoscopic hemostasis are presented in Online Appendix S2. Each procedure was repeated until hemostasis was endoscopically confirmed. Interventional radiology and/or surgery were implemented when endoscopic hemostasis was considered ineffective.

After endoscopic hemostasis was achieved, all patients were hospitalized and managed conventionally (fasting with peripheral parenteral nutrition and intravenous proton pump inhibitors). The indications for red blood cell (RBC) transfusion were a hemoglobin level < 6 g/dL on admission or a rapid drop in hemoglobin level > 2 g/dL in patients with hemoglobin levels < 10 g/dL at baseline. Follow-up endoscopy was performed on patients for whom it was judged necessary by the endoscopist 48–72 h after the initial endoscopy, and repeat endoscopic hemostasis was applied in cases where recurrent UGIB was detected. Interventional radiology and/or surgery were implemented when endoscopic hemostasis was considered ineffective.

Endoscopic hemostasis was performed by all 17 endoscopists, comprising 7 specialists and 10 trainees. Specialists were defined as endoscopists who had performed endoscopy for more than 5 years with experience in more than 40 endoscopic submucosal dissection procedures after mastering the required fundamental skills and knowledge [31, 32].

Clinical data collected in the retrospective cohort were age, gender, alcohol consumption, smoking habit, H. pylori infection, American Society of Anesthesiologists physical status, use of antithrombotic agents, nonsteroidal anti-inflammatory drugs (NSAIDs), or gastric acid secretion inhibitor, and comorbidity (including Charlson comorbidity score). H. pylori infection was diagnosed by the serum levels of anti-H. pylori antibodies, the urea breath test, or the rapid urease test. Patient status on admission, including initial symptoms, initial vital signs, initial laboratory data, and transfusion volume, was collected. Three scoring systems (Rockall score, Glasgow-Blatchford score, and AIMS65 score) have been reported to be useful in predicting mortality, rebleeding, need for transfusion, and hemostasis [33‐36]. These scoring systems were applied, on the basis of admission history, clinical and laboratory data, endoscopic findings, treatment, and clinical follow-up. Accumulated information concerning the cause of UGIB, endoscopic findings, and types of anesthesia drugs was also collected.

Treatment outcomes and adverse events were compared between the two groups using propensity score matching analysis. This method was applied to adjust significant differences in the baseline characteristics of the patients and reduce the influence of possible confounding factors [37]. The two groups were matched at a 1:1 ratio (133 patients in each group) with adjustment for 13 covariates (age, gender, H. pylori infection, use of antithrombotic agents, use of NSAIDs, chronic liver damage, Rockall score, Glasgow-Blatchford score, AIMS65 score, systolic blood pressure, diastolic blood pressure, hemoglobin level, and peptic ulcer) to minimize inherent bias. These 13 covariates were selected on the basis of the opinions of seven expert endoscopists. This model yielded a C statistic of 0.705, indicating a preferable ability for comparison between two groups. The caliper width of propensity score matching was 0.20.

Categorical data were expressed as a number (percentage), and chi-squared test was used to identify differences between the two groups. Numerical data for normal distribution were expressed as the mean ± standard deviation, and Student’s t-test was used to determine differences between the two groups. Numerical data for a skewed distribution were expressed as median interquartile range, and the Mann–Whitney U-test was applied. Levels of significance for all comparisons made were reported, whether significant or not, with P values or confidence intervals. A P value of < 0.05 was statistically significant for each test. Survival was analyzed using Kaplan–Meier plots with log-rank tests. The Cox proportional hazard model was then used to adjust confounding factors in the analysis of 30-day mortality from UGIB. All statistical analyses were performed with JMP version 13.0.0 (SAS Institute, Tokyo, Japan).

During the period between January 2016 and December 2021, we underwent emergency endoscopy for UGIB in 402 patients. Among them, endotracheal intubation was applied in five patients, respiratory and circulatory status was unstable in three patients, and missing data were found in five patients. Thirteen patients were thus excluded, and 389 patients were assessed as eligible for the present study.

Of the 389 patients, 171 patients (44.0%) received sedation during emergency endoscopy, and the remaining 218 patients (56.0%) did not receive sedation during the procedure (Fig. 1). Baseline patient characteristics are presented in Table 1. The sedation group was significantly younger with more male patients, and had H. pylori infection and chronic liver disease. The use of antithrombotic agents and NSAIDs was higher in the sedation group. Three scoring systems (Rockall score, Glasgow-Blatchford score, and AIMS65 score) were significantly higher for the sedation group.

UGIB patient status on admission is presented in Table 2. Systolic and diastolic blood pressure and hemoglobin levels were significantly higher in the sedation group. Initial laboratory data other than hemoglobin level did not differ between the sedation and nonsedation groups. There were no significant differences between the two groups in the number of patients who received RBC transfusions and the total volume of RBC transfused.

Table 3 demonstrates the endoscopic findings of UGIB. The main cause of bleeding was peptic ulcers (49.1% in the sedation group and 59.6% in the nonsedation group, P = 0.04). The number and size of peptic ulcers did not differ significantly between the two groups.

Administered dosages of sedatives are presented in Table S1. One hundred and thirteen patients (66.1%) were sedated using midazolam, for whom the mean dose of midazolam was 4.2 ± 1.8 mg. In addition, 54 (31.6%) and 4 (2.3%) patients underwent sedation using diazepam and propofol.

Table 4 compares the clinical characteristics between the two groups before and after propensity score matching. Before propensity score matching, mean age, gender, H. pylori infection, use of antithrombotic agents, use of NSAIDs, chronic liver damage, Rockall score, Glasgow-Blatchford score, AIMS65 score, systolic and diastolic blood pressure, hemoglobin level, and peptic ulcer were significantly different between the two groups. Propensity score matching subsequently created 133 matched pairs in the present study and averaged the differences in 13 covariates.

Table 5 compares the treatment outcomes after propensity score matching between the two groups. The success rate of endoscopic hemostasis was similar in both groups, with a significantly shorter procedure time in the sedation group than in the nonsedation group (17.6 ± 10.0 versus 20.2 ± 10.2 min, P = 0.04). The main modality of hemostasis was soft coagulation between the two groups. Adverse events such as aspiration pneumonia, rebleeding, and 30-day mortality did not differ significantly between the two groups.

In the survival analysis, the Kaplan–Meier plot demonstrated no impact of sedation on mortality within 30 days (Fig. S1). Table 6 lists the risk factors for 30-day mortality from UGIB. The Cox proportional hazard model demonstrated that RBC transfusion and rebleeding were risk factors for 30-day mortality from UGIB: hazard ratio (HR) [95% confidence interval] 4.45 [1.50–19.11], P < 0.01; and HR 3.30 [1.14–8.36], P = 0.03. Nevertheless, sedation endoscopy was not associated with 30-day mortality after adjusting for potential confounders: HR 1.39 [0.64–3.02], P = 0.40.