Acute gastrointestinal bleeding is a common but serious medical emergency in clinical practice. It imposes a massive burden on global healthcare use each year and has a negative impact on patient survival.^1-3 Endoscopy is the gold standard for investigating the bleeding source and providing a direct view for therapeutic interventions. It was reported that approximately 90% of culprit lesion locations can be found under endoscopy in patients with upper gastrointestinal bleeding (UGIB).⁴ Furthermore, performance of endoscopic interventions dramatically reduced the risks of recurrent bleeding, need for surgery, and mortality.^5-7 In spite of numerous studies showing the efficiency of endoscopy for UGIB, one of the most crucial considerations in the entire procedure, the optimal timing of endoscopy performance remains unconfirmed.

In the management of UGIB, it is widely accepted that endoscopy should be performed within 24 hours of admission.^8,9 However, there is a scarcity of data to back up this recommendation, and the majority of studies that did assess the clinical benefits of endoscopy within 24 hours were of poor quality.^10-13 Another question that remains unanswered is whether endoscopy conducted within a shorter time frame such as 6 to 8 hours can improve the prognosis of patients who are at high risk of bleeding (e.g. Glasgow-Blatchford score, Rockall score, or hemodynamic parameters).^14-18 According to the European Society of Gastrointestinal Endoscopy in 2021, the risk of procedure-related adverse events is higher in patients who undergo early or emergent endoscopy (within 12 hours or 6 hours).¹⁹ However, the Asia-Pacific Working Group accepted that in patients with hemodynamic shock or instability, early endoscopy performed within 12 hours after admission, may provide benefits to patients after initial resuscitation and stabilization.²⁰ What is more, the American College of Gastroenterology has recently recommended both low- and high-risk patients should undergo endoscopy within 24 hours for economic effectiveness and clinical benefit.²¹

There have been a few systematic reviews and meta-analyses to look at these unsolved questions based on the present contradicting data. One review found that early endoscopy was safe and helpful in patients with acute UGIB, while others showed that it did not enhance clinical outcomes such as recurrent bleeding, mortality, or length of stay (LOS).^22-25 Furthermore, four recent randomized controlled studies (RCTs) with varied definitions of time have also yielded conflicting findings. Only one study by Lee et al.²⁶ concluded that the early procedure performed within 2 hours after admission was associated with a significant reduction in hospitalization, but there was a lack of consensus among the other three studies as to whether or not patients would benefit from very early procedures within 6 or 12 hours.^15,27,28 When comparing the outcomes of early and elective procedures, observational studies conducted in Western and Eastern countries have generated conflicting conclusions. The majority of Western studies found that early endoscopy was unnecessary.^17,18,29-32 However, in the East, the results remained more varied and uneven.^16,33 The following issues were the focus of our systematic review and meta-analysis: (1) Are procedures performed within 24 hours as generally accepted associated with better patient outcomes? (2) For patients at high risk of bleeding, does a shorter time frame, such as performing endoscopy within 6 to 12 hours, have a clinical benefit? (3) What is the difference in clinical outcomes between the East and West?

1. Literature search and data extraction

Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines were followed in the conduct of this review.³⁴ We used MEDLINE (OvidSP), Embase (OvidSP), and the Cochrane bibliographic database to look for papers published between January 1, 1980, and March 1, 2022. The search strategy was devised with the help of a medical librarian and was detailed in the Supplementary Material.

Two independent reviewers (B.L.Y. and J.W.) initially screened all publications by title and abstract. Observational cohort studies and randomized controlled trials reporting on the clinical outcomes of individuals with acute UGIB hospitalized for endoscopic assessment were included in this review. Studies were ineligible for consideration if they: (1) were case reports, reviews, re-analyses of public datasets, or conference abstracts; (2) did not compare the outcomes between early and nonearly arms; (3) only focused on variceal gastrointestinal bleeding; (4) were an irrelevant study on current clinical practice; (5) had a vague definition of the time frame for early endoscopy; and (6) were not reported in English. RCTs and observational cohort studies were pooled and analyzed separately. The data collected included the following: study and patient characteristics (e.g., first author, year of publication, study design, site/country, risk of bleeding, number of patients who underwent early endoscopy or nonearly endoscopy, time frame for early endoscopy) and outcomes. The data were rechecked if there was a discrepancy.

2. Study definitions and outcomes

Early endoscopy was defined as a patient with acute nonvariceal upper gastrointestinal bleeding (ANVUGIB) who underwent esophagogastroduodenoscopy within 24 hours after admission. Procedures performed after 24 hours were regarded as nonearly endoscopy which served as the comparator. We defined those procedures which were performed with shorter time frames ranging from 3 to 12 hours as very early endoscopy. A subgroup analysis was subsequently performed on studies that assessed the clinical outcomes of very early endoscopy. The bleeding risk was evaluated based on: (1) symptoms of overt UGIB (witnessed hematemesis, bloody nasogastric aspirate, or hematochezia) or hemodynamic instability (systolic blood pressure <100 mm Hg or heart rate >100 beats/min); (2) abnormal laboratory indices (hemoglobin <12 g/dL and urea >6.5 mmol/L); (3) risk factors such as age >60 years, ischemic heart disease, liver disease, cardiac failure, and metastatic malignancy, and (4) have high-risk stigmata on endoscopy (spurting, gushing, oozing bleeding, or non-bleeding visible vessel). Patients who met at least one of the standards above were regarded as high-risk bleeding populations.

The primary outcome was the odds ratio for mortality in the included population. Secondary outcomes included recurrent bleeding, surgery, primary hemostasis, angiographic embolization, LOS, need for repeat endoscopy, and requirement for blood transfusion. We performed a subgroup analysis to investigate the difference in clinical outcomes between Eastern and Western populations. The East included South Korea, Taiwan, and Japan while the West included the United States, Canada, United Kingdom, Switzerland, and Poland, as determined by included studies.

3. Quality assessment

The quality of observational cohort studies was evaluated using the Newcastle-Ottawa Scale.³⁵ There are three domains in the Newcastle-Ottawa Scale (maximum 9 stars): selection (representativeness of cases, selection of controls, ascertainment of exposure, and demonstration that the outcome of interest was not existent at the beginning of the study); comparability (comparability of cohorts based on the design or analysis); and outcome (evaluation of outcome, length of follow-up necessary for outcomes to occur, appropriateness of follow-up of cohorts). The risk of bias assessment developed by the Cochrane group was used to evaluate RCT.³⁶ The tool consists of seven domains, which are sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessors, incomplete outcome data, selective outcome reporting, and other biases. Any disagreement was solved by consensus.

4. Statistical analysis and bias assessment

RevMan 5.4 software was used to conduct all statistical analyses in accordance with the Cochrane Collaboration methods. Data from observational cohort studies and RCTs were pooled for meta-analysis. Subgroup analysis was performed in terms of timing and patient population.

Continuous variables are provided as means and standard deviations (SD), whereas categorical variables are expressed as percentages. Online calculators were used to determine the sample mean and SD for studies that only provided the median and interquartile range.³⁷ The largest SD from other research was utilized for studies that did not publish their SD and for which the authors did not respond. Odds ratios (ORs) with 95% confidence intervals (CI) were used to compare dichotomous outcomes. Mean differences (MDs) were handled as continuous variables using the inverse variance approach. I² statistics were used to evaluate heterogeneity. When there was no statistically significant heterogeneity among studies (I²<50%, p>0.1), a fixed-effects model (Mantel-Haenszel method) was used to calculate the pooled estimates, otherwise a random effects model (DerSimonian and Laird method) was used. If the p-value was less than 0.05, it was thought to be statistically significant. The Mantel-Haenszel method with a fixed-effects model was used to conduct sensitivity studies when statistical heterogeneity was found. Patient populations, sample sizes, study designs, intervention types, and outcome definitions from various studies were all examined for any possible clinical heterogeneity. If at least 10 citations were found, publication bias was assessed using funnel plots for all comparisons.

1. Study characteristics

We found a total of 1,206 papers by searching databases, and 1,114 of those were discarded because of issues with the abstract, title, duplication, or the lack of full-text availability. The full texts of 92 papers were assessed afterward, and a further 67 publications were excluded (Fig. 1). Twenty-five studies (five RCTs and 20 observational cohort studies) were included in the meta-analysis (Table 1). For a timing cutoff of 24 hours, 13 observational cohort studies compared outcome parameters in patients with ANVUGIB.^{10-13,29-31,38-43} Seven observational cohort studies^{14,16-18,33,43,44} and two RCTs^15,28 included patients at high risk of bleeding, with time frames ranging from 6 to 12 hours.

Table 1 . Details and Characteristics of the Included Studies.

Author (year)	Types of study	Country	Study population	High-risk bleeding*	Time frames for early endoscopy (1)	Time frames for early endoscopy (2)	Sample size (early)	Sample size (non-early)
Cooper et al. (2009)10	Retrospective cohort study	USA	UGIB	NA	24		1,854	738
Wierzchowski et al. (2013)11	Retrospective cohort study	Poland	UGIB	NA	24		295	187
Chak et al. (2001)12	Retrospective cohort study	USA	UGIB	High-risk	24		166	36
Kim et al. (2018)13	Retrospective cohort study	South Korea	UGIB	High-risk	24		187	66
Cho et al. (2018)14	Retrospective cohort study	South Korea	UGIB	High-risk	6		571	390
Lau et al. (2020)15	Randomized controlled trial	Hong Kong	UGIB	High-risk	6		258	258
Tai et al. (2007)16	Retrospective cohort study	Taiwan	UGIB	High-risk	8		88	101
Targownik et al. (2007)17	Retrospective cohort study	Canada	UGIB	High-risk	6		77	92
Kumar et al. (2017)18	Retrospective cohort study	USA	UGIB	High-risk	12		Total (n=89), high-risk (n=41)	Total (n=272), high-risk (n=80)
Lee et al. (1999)26	Randomized controlled trial	USA	UGIB	NA	2		56	54
Bjorkman et al. (2004)27	Randomized controlled trial	USA	UGIB	NA	6		47	46
Lin et al. (1996)28	Randomized controlled trial	Taiwan	UGIB	High-risk	12		Total (n=162), high-risk (n=53)	Total (n=163), high-risk (n=54)
Iqbal et al. (2018)29	Retrospective cohort study	USA	UGIB	NA	24		33	146
Jairath et al. (2012)30	Prospective cohort study	UK	UGIB	NA	12	24	<12 hr (n=834), 12-24 hr (n=1,190)	>24 hr (n=2,158)
Saleem et al. (2020)31	Retrospective cohort study	USA	UGIB	NA	12	24	<12 hr (n=61), 12–24 hr (n=94)	>24 hr (n=96)
Schacher et al. (2005)32	Retrospective cohort study	Switzerland	UGIB	NA	3		43	38
Ahn et al. (2016)33	Retrospective cohort study	South Korea	UGIB	High-risk	8		Total (n=60), high-risk (n=18)	Total (n=98), high-risk (n=36)
El-Dallal et al. (2021)38	Retrospective cohort study	USA	UGIB	NA	24		24,830	96,005
Siau et al. (2019)39	Retrospective cohort study	UK	UGIB	NA	24		205	143
Garg et al. (2017)40	Retrospective cohort study	USA	UGIB	NA	24		870,159	631,412
Sarin et al. (2009)41	Retrospective cohort study	UK	UGIB	NA	6	24	<6 hr (n=72), 6–24 hr (n=198)	>24 hr (n=232)
Cooper et al. (1999)42	Retrospective cohort study	USA	UGIB	NA	24		583	326
Guo et al. (2021)43	Retrospective cohort study	Hong Kong	UGIB	High-risk	6	24	<6 hr (n=1,008), 6–24 hr (n=3,865)	>24 hr (n=1,601)
Horibe et al. (2021)44	Retrospective cohort study	Japan	UGIB	High-risk	6		115	115
Whorwell et al. (1981)45	Randomized controlled trial	UK	UGIB	NA	6		54	46

UGIB, upper gastrointestinal bleeding; NA, not available..

*The risk of bleeding was evaluated and confirmed based on (1) symptoms of overt UGIB (witnessed hematemesis, bloody nasogastric aspirate, or hematochezia) or hemodynamic instability (systolic blood pressure <100 mm Hg or heart rate >100 beats/min); (2) abnormal laboratory indices (hemoglobin <12 g/dL and urea >6.5 mmol/L); (3) risk factors including age >60 years, ischemic heart disease, liver disease, cardiac failure, and metastatic malignancy and (4) have high-risk stigmata on endoscopy. Patients with at least one of the standards above were regarded as high-risk bleeding populations and included in the subgroup analysis..

Figure 1. PRISMA flow diagram.
PRIMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; GIB, gastrointestinal bleeding.

2. Primary outcome

The principal results in the included studies were shown in Table 2. The meta-analysis of 13 observational cohort studies found no significant difference in mortality rates when endoscopy was performed within 24 hours of admission compared to that performed after 24 hours (OR, 1.13; 95% CI, 0.83 to 1.54; p=0.44; I²=96%) (Fig. 2A).^{10-13,28-31,38-43} Three observational cohort studies of patients at high-risk of bleeding reported outcomes within 24 hours and it demonstrated no significant difference in mortality rates between early and nonearly endoscopy arms (OR, 0.87; 95% CI, 0.69 to 1.09; p=0.21, I²=26%).^12,13,43

Table 2 . Principal Results of the Included Studies (Early vs Non-Early).

Author (year)	Mortality, %	Recurrent bleeding, %	Surgery, %	Blood transfusion, %	Primary hemostasis, %	Angiographic embolization, %	Repeat endoscopy, %	Length of stay, mean±SD, day
Cooper et al. (2009)10	6.2 vs 7.3	NA	1.2 vs 3.4	NA	31.8 vs 32.9	NA	NA	4.78±2.75 vs 7.19±4.74
Wierzchowski et al. (2013)11	6.78 vs 9.09	12.2 vs 18.2	2.37 vs 6.42	NA	88.5 vs 40.1	NA	NA	4.3±11.91 vs 4.79±27.22
Chak et al. (2001)12	4.2 vs 5.6	16.9 vs 19.4	6.0 vs 5.6	NA	NA	NA	NA	6.31±3.8 vs 7±2.97
Kim et al. (2018)13	46.5 vs 39.4	20.3 vs 25.6	1.6 vs 0	NA	32.1 vs 12.1	14.4 vs 15.2	20.3 vs 15.2	52.0±48.3 vs 57.4±58.8
Cho et al. (2018)14	1.6 vs 3.8	11.4 vs 9.0	1.1 vs 0	NA	69.5 vs 53.3	2.8 vs 0.5	NA	6.7±7.8 vs 6.4±7.5
Lau et al. (2020)15	ANVUGIB: 7.3 vs 6.3	ANVUGIB: 9.9 vs 6.7	ANVUGIB: 0.96 vs 0	ANVUGIB: 90.6 vs 91.6	ANVUGIB: 57.9 vs 45.2	ANVUGIB: 0.86 vs 0.84	NA	ANVUGIB: 7.64±7.63 vs 6.81±6.61
	AVUGIB: 24.0 vs 10.5	AVUGIB: 16.0 vs 15.8	AVUGIB: 0 vs 5.3	AVUGIB: 80.0 vs 78.9	AVUGIB: 80.0 vs 89.5	AVUGIB: 4.0 vs 0		AVUGIB: 8.88±7.60 vs 7.95±4.14
Tai et al. (2007)16	1 vs 6	7 vs 10	0 vs 0	74 vs 78	95 vs 95	2 (2) vs 4 (4), p=0.687	33 vs 50	5.1±5 vs 6.0±7.7
Targownik et al. (2007)17	8 vs 6	9 vs 8	8 vs 2	NA	53 vs 37	NA	10 v s 15	4±15.16 vs 4±12.76
Kumar et al. (2017)18	Total: 4 vs 1, high-risk: 5 vs 1	Total: 11 vs 4, high-risk: 12 vs 6	Total: 4 vs 0, high-risk: 2 vs 0	Total: 85 vs 75, high-risk: 98 vs 99	NA	Total: 6 vs 1, high-risk: 7 vs 2	Total: 10 vs 1, high-risk: 7 vs 2	Total: 4.3±4.2 vs 4.0±4.0, high-risk: NA
Lee et al. (1999)26	0 vs 3.7	3.6 vs 5.6	3.6 vs 1.9	NA	NA	NA	7.1 vs 7.4	2.35±3.6 vs 2.43±0.86
Bjorkman et al. (2004)27	0 vs 0	NA	2.1 vs 2.2	40.4 vs 32.6	NA	0 vs 0	8.5 vs 2.2	3.98±3.89 vs 3.26±3.17
Lin et al. (1996)28	3.8 vs 1.8	5.7 vs 9.2	5.7 vs 7.4	NA	43.4 vs 42.6	NA	NA	4.7±4.4 vs 4.2±6
Iqbal et al. (2018)29	6.8 vs 6.1	8.3 vs 12.1	NA	NA	NA	NA	NA	6.1±11.91 vs 6.0±27.22
Jairath et al. (2012)30	<12 hr: 9.0	<12 hr: 19.7	7.6 vs 3.3 vs 2.2	<12 hr: 49.2	NA	NA	NA	<12 hr: 11.46±15.16
	12–24 hr: 6.3	12–24 hr: 10.9		12–24 hr: 37.1				12–24 hr: 8.04±8.96
	>24 hr: 5.4	>24 hr: 8.8		>24 hr: 31.7				>24 hr: 15.37±27.22
Saleem et al. (2020)31	<12 hr: 1.64	NA	<12 hr: 3.28	<12 hr: 54.10	<12 hr: 35.71	<12 hr: 4.92	<12 hr: 26.23	NA
	12–24 hr: 1.06		12–24 hr: 2.13	12–24 hr: 48.94	12–24 hr: 33.33;	12–24 hr: 2.13	12–24 hr: 4.26
	>24 hr: 1.04		>24 hr:1.04	>24 hr:50.0	>24 hr:77.78	>24 hr:1.04	>24 hr:2.08
Schacher et al. (2005)32	0 vs 0	14 vs 16	9.3 vs 7.9	NA	76.7 vs 47.4	NA	NA	5.9±15.16 vs 5.1±12.76
Ahn et al. (2016)33	Total: 1.7 vs 2.0, high-risk: 0 vs 5.6	Total: 6.7 vs 5.1, high-risk: 11.1 vs 13.9	Total: 0 vs 1.0, high-risk: 0 vs 2.8	NA	Total: 71.1 vs 35.7, high-risk:61.1 vs 41.7	Total: 5 vs 2.0, high-risk:11.1 vs 2.8	Total: 3.3 vs 3.1, high-risk:5.6 vs 8.3	Total: 7.3±5.2 vs 9.1±9.4, high-risk: 7.7±5.3 vs 12.0±9.9
El-Dallal et al. (2021)38	2.5 vs 1.8	NA	1.43 (0.89–2.31)*, p=0.14	NA	NA	1.43 (0.93–2.19)*, p=0.11	NA	3.93±9.41 vs 5.1±8.58
Siau et al. (2019)39	7.8 vs 4.9	12.2 vs 3.9	NA	NA	NA	NA	NA	4 vs 5
Garg et al. (2017)40	3.0 vs 4.2	NA	NA	54.8 vs 51.3	43.2 vs 51.3	NA	NA	4.6±0.02 vs 8.5±0.08
Sarin et al. (2009)41	<6 hr: 22.2	NA	<6 hr: 12.5	<6 hr: 81.9	<6 hr: 52.8	NA	<6 hr: 22.2	<6 hr: 5.2±12
	6–24 hr: 25.8		6–24 hr: 14.6	6–24 hr: 15.7	6–24 hr: 58.6		6–24 hr: 21.2	6–24 hr: 5.0±10.6
	>24 hr: 9.9		>24 hr: 5.2	>24 hr: 75.9	>24 hr: 22.4		>24 hr: 8.6	>24 hr: 4.3±11
Cooper et al. (1999)42	3.7 vs 3.4	NA	NA	NA	NA	NA	NA	5.0 vs 6.4
Guo et al. (2021)43	<6 hr: 6.2	NA	NA	NA	NA	<6 hr: 1.215 (1.107–1.334)†, p<0.001;	NA	<6 hr: 7.28±12.07
	6–24 hr: 4.3							6–24 hr: 6.49±9.01
	>24 hr: 5.8					>24 hr: 1.040 (0.944–1.145)†, p=0.426		>24 hr: 6.96±11.75
Horibe et al. (2021)44	2.61 vs 7.83	NA	NA	NA	NA	NA	NA	NA
Whorwell et al. (1981)45	11.1 vs 15.2	NA	29.6 vs 28.3	NA	NA	NA	NA	12.0±7.6 vs 12.0±6.6

ANVUGIB, acute nonvariceal upper gastrointestinal bleeding; AVUGIB, acute variceal upper gastrointestinal bleeding; NA, not available..

*Odds ratio (95% confidence interval); ^†Hazard ratio (95% confidence interval)..

Figure 2. Forest plot comparing early versus nonearly endoscopy for (A) mortality, (B) recurrent bleeding, (C) surgery, and (D) primary hemostasis in acute nonvariceal upper gastrointestinal bleeding patients.
M-H, Mantel-Haenszel method; IV, interval variable; CI, confidence interval; SE, standard error.

Subgroup analysis of studies with very early endoscopy ranging from 3 to 12 hours was subsequently conducted. For observational cohort studies, the mortality rate was significantly higher in the very early endoscopy group within 12 hours (OR, 1.66; 95% CI, 1.27 to 2.18; p<0.001; I²=0%) (Table 3).^18,30,31 One study included patients with high bleeding risk who underwent endoscopy within 12 hours and no significant reduction in mortality rate was observed in the very early endoscopy group.¹⁸ For studies with time frames ranging from 6 to 8 hours, it also failed to show any superiority of very early endoscopy with respect to lower mortality either in high-risk patients (n=4; OR, 0.74; 95% CI, 0.34 to 1.60; p=0.44; I²=72%) or in patients at all bleeding risks (Table 3).^{14,16,17,33,43,44}

Table 3 . Subgroup Analysis for Outcomes Based on Different Cutoff Times for Endoscopy.

Outcome assessed	No. of studies	Estimate	95% CI	p-value	I2
Endoscopy ≤12 hr vs >12 hr
Mortality	3	1.66	1.27 to 2.18	<0.001	0
Recurrent bleeding	2	2.36	1.93 to 2.88	<0.001	0
Surgery	3	3.24	2.35 to 4.47	<0.001	0.17
Angiographic embolization	2	2.16	0.49 to 9.46	0.310	0.71
Length of stay	2	–0.83	–3.07 to 1.41	0.470	0.89
Primary hemostasis	1	2.29	1.25 to 4.17	0.007	NA
Repeat endoscopy	2	9.39	4.36 to 20.20	<0.001	0
Blood transfusion	3	1.81	1.48 to 2.22	<0.001	0.13
Endoscopy ≤8 hr vs >8 hr
Mortality	2	0.35	0.07 to 1.74	0.200	0
Recurrent bleeding	2	1.06	0.35 to 3.19	0.910	0
Surgery	2	0.54	0.02 to 13.40	0.700	NA
Angiographic embolization	2	1.15	0.33 to 4.00	0.830	0.27
Length of stay	2	–1.25	–2.68 to 0.17	0.090	0
Primary hemostasis	2	4.13	2.58 to 6.61	<0.001	0
Repeat endoscopy	2	1.11	0.24 to 5.09	0.890	0
Blood transfusion	1	0.79	0.40 to 1.54	0.480	NA
Endoscopy ≤6 hr vs >6 hr
Mortality	5	0.89	0.52 to 1.53	0.680	0.64
Recurrent bleeding	2	1.29	0.86 to 1.93	0.210	0
Surgery	3	1.59	0.81 to 3.12	0.180	0
Angiographic embolization	1	5.59	1.28 to 24.46	0.020	NA
Length of stay	4	0.51	–0.09 to 1.10	0.100	0
Primary hemostasis	3	1.94	1.56 to 2.42	<0.001	0
Repeat endoscopy	3	1.22	1.11 to 1.33	<0.001	0.31
Blood transfusion	2	3.21	1.07 to 9.61	0.040	0.57
Endoscopy ≤3 hr vs >3 hr
Mortality	1	NA	NA	NA	NA
Recurrent bleeding	1	0.86	0.25 to 2.95	0.820	NA
Angiographic embolization	1	1.06	0.34 to 3.35	0.920	NA
Length of stay	1	0.80	–5.28 to 6.88	0.800	NA
Primary hemostasis	1	3.67	1.42 to 9.50	0.007	NA

CI, confidence interval; NA, not available..

RCTs involving patients with ANVUGIB demonstrated no significant difference in mortality rates during time periods ranging from 2 to 12 hours, according to the pooled estimates (OR, 0.97; 95% CI, 0.55 to 1.72; p=0.92; I²=0%).^15,26-28,45 There was no significant difference in mortality rates among individuals with high-risk UGIB in two RCTs (Fig. 3A).^15,28

Figure 3. Forrest plot comparing early versus nonearly endoscopy for (A) mortality, (B) recurrent bleeding, (C) surgery, (D) primary hemostasis, and (E) length of stay in high-risk acute nonvariceal upper gastrointestinal bleeding patients from randomized controlled studies.
M-H, Mantel-Haenszel method; CI, confidence interval; SD, standard deviation; IV, interval variable.

3. Secondary outcomes

According to the pooled data of observational cohort studies, recurrent bleeding, need for surgery, angiographic embolization, repeat endoscopy, and blood transfusion had no significant difference among individuals who received early versus nonearly endoscopy (Figs 2B, 2C, 4A, 4C, 4D). In terms of the LOS, there was a significant difference between the early endoscopy and nonearly endoscopy groups (MD, −2.25; 95% CI, −3.42 to −1.08; p<0.001, I²=94%) (Fig. 4B). In the early endoscopy group, primary hemostasis was required more frequently than in the nonearly endoscopy group (OR, 2.95; 95% CI, 1.75 to 4.98; p<0.001, I²=96%) (Fig. 2D). Subgroup analysis of three observational cohort studies that included patient with high risk of bleeding demonstrated no significant difference in recurrent bleeding (OR, 0.77; 95% CI, 0.45 to 1.31; p=0.34; I²=0%), need for surgery need (OR, 1.31; 95% CI, 0.33 to 5.21; p=0.70; I²=0%), repeat endoscopy (OR, 1.05; 95% CI, 0.95 to 1.15; p=0.37; I²=0%) and LOS (MD, −0.18; 95% CI, −0.85 to 0.48; p=0.59; I²=0%).^12,13,43 Primary hemostasis in patients with high bleeding risk was only assessed in one observational cohort study,¹³ which found that patients with early endoscopy were associated with a higher need for primary hemostasis.

Figure 4. Forest plot comparing early versus nonearly endoscopy for (A) angiographic embolization, (B) length of stay, (C) repeat endoscopy, and (D) blood transfusion in acute nonvariceal upper gastrointestinal bleeding patients.
SE, standard error; IV, interval variable; CI, confidence interval; M-H, Mantel-Haenszel method.

Leave-one-out meta-analysis revealed that the heterogeneity of recurrent bleeding was lower after removing the study by Jairath et al.³⁰ The sources of heterogeneity may include the larger sample size (over 2,000 cases) and the prospective study design. Additionally, the significant heterogeneity in the need for surgery decreased after removing two studies that only included elderly patients.^10,38 The source of heterogeneity may be the age difference between patients in pooled studies.

Subgroup analysis of studies with time frames shorter than 24-hour, such as 3, 6, 8, and 12 hours, was performed, and the results were described in Table 3. For observational cohort studies, endoscopic procedures within 12 hours were significantly associated with a higher rate of recurrent bleeding, higher need for surgery, and higher requirement for blood transfusion and repeat endoscopy (p_s<0.05).^18,30,31 LOS was not significantly changed, and primary hemostasis was required in more patients with early endoscopy ranging from 3 to 12 hours (Table 3). Repeat endoscopy and blood transfusion were required in more patients with ANVUGIB who underwent early endoscopy within 6 hours (Table 3).^{14,17,41,43,44} Only one study looked at people with a high risk of bleeding who had an early endoscopy within 12 hours,¹⁸ and no significant differences were found in any clinical parameters. Angiographic embolization and repeat endoscopy were assessed in one and two studies in patients with high risk of bleeding, respectively. It showed that angiographic embolization was more often necessary in patients who underwent very early endoscopy within 6 hours (OR, 5.59; 95% CI, 1.28 to 24.46; p=0.02). and repeat endoscopy (OR, 1.21; 95% CI, 1.10 to 1.32; p<0.001; I²=43%).^13,14,43 Pooled results of five RCTs failed to show a significant difference in outcomes.^15,26-28,45 Two RCTs looked at patients with a high risk of bleeding, and a meta-analysis showed that patients who had a very early endoscopy within 6 to 12 hours were much more likely to have primary hemostasis (Fig. 3).^15,28

4. East-West outcome differences

For observational cohort studies including Western population with a time frame of 24-hour, pooled results demonstrated shorter LOS (MD, −1.96; 95% CI, −3.31 to −0.60; p=0.005; I²=100%), higher likelihood of requiring repeat endoscopy (OR, 6.96; 95% CI, 1.59 to 30.50; p=0.01; I²=73%) and blood transfusion (OR, 1.61; 95% CI, 1.43 to 1.82; p<0.001; I²=67%). Two studies from the East were included.^13,43 Only one study showed that endoscopy performed within 24 hours was related to a greater need for primary hemostasis (OR, 3.43; 95% CI, 1.54 to 7.63; p=0.003).¹³

For RCTs, pooled analysis of studies from populations either in the West or East did not reveal significant differences in mortality rates, need for surgery, or LOS (all p>0.05). The need for primary hemostasis was only reported in studies from the East, and pooled results demonstrated a significantly higher need for primary hemostasis (OR, 1.53; 95% CI, 1.10 to 2.12; p=0.01; I²=19%) in patients who underwent early endoscopy within 6 or 12 hours.^15,28

5. Quality assessment and risk of publication bias

It remained unclear whether patients or endoscopists were blinded to the timing of treatment in all included RCTs. However, there was a low risk of incomplete outcome data and other potential biases.

Five of the twenty observational studies had a potential for participant comparability and confounding characteristics that needed to be taken into consideration. They did not make any comparisons between the two groups’ baseline characteristics. The Newcastle-Ottawa Scale yielded an average score of 7.4±1.0 stars (range 6 to 9 stars; the highest-quality studies are given 9 stars). Tables 4 and 5 detail the quality of these included studies. Statistical heterogeneity was observed in mortality, recurrent bleeding, surgery, LOS, primary hemostasis, blood transfusion,n and repeat endoscopy for ANVUGIB. Visual inspection of the funnel plot uncovered some asymmetry in the data (Supplementary Fig. 1).

Table 4 . Assessment of the Methodological Quality (Cochrane Risk of Bias for Randomized Controlled Trials).

Author (year)	Sequence generation	Allocation concealment	Blinding of participants and personnel	Blinding of outcome assessors	Incomplete outcome data	Selective outcome reporting	Other
Lau et al. (2020)15	Low	Unclear	Unclear	Unclear	Low	Low	Low
Lee et al. (1999)26	Unclear	Unclear	Unclear	Unclear	Low	Low	Low
Bjorkman et al. (2004)27	Low	Low	Unclear	Unclear	Low	Unclear	Low
Lin et al. (1996)28	Unclear	Low	Unclear	Unclear	Low	Low	Low
Whorwell et al. (1981)45	Unclear	Unclear	Unclear	Unclear	Low	Unclear	Low

Table 5 . Assessment of the Methodological Quality (Newcastle-Ottawa for Observational Studies).

Author (year)	Selection					Comparability		Outcome			Total
	S1	S2	S3	S4		C		O1	O2	O3
Cooper et al. (2009)10	-	*	*	*		**		*	*	*	8
Wierzchowski et al. (2013)11	*	*	*	*		-		*	*	*	7
Chak et al. (2001)12	*	*	*	*		-		*	*	*	7
Kim et al. (2018)13	-	*	*	*		**		*	*	*	8
Cho et al. (2018)14	*	*	*	*		**		*	*	*	9
Tai et al. (2007)16	*	*	*	*		**		*	*	*	9
Targownik et al. (2007)17	-	*	*	*		-		*	*	*	6
Kumar et al. (2017)18	-	*	*	*		**		*	*	*	8
Iqbal et al. (2018)29	-	*	*	*		**		*	*	-	7
Jairath et al. (2012)30	*	*	*	*		-		*	*	*	7
Saleem et al. (2020)31	*	*	*	-		**		*	*	*	8
Schacher et al. (2005)32	*	*	*	*		**		*	*	*	9
Ahn et al. (2016)33	*	*	*	*		**		*	*	*	9
El-Dallal et al. (2021)38	-	-	*	*		**		*	*	*	7
Siau et al. (2019)39	*	*	*	-		*		*	*	*	7
Garg et al. (2017)40	*	*	*	-		*		*	*	*	7
Sarin et al. (2009)41	*	*	*	-		-		*	*	*	6
Cooper et al. (1999)42	*	*	*	-		*		*	*	-	6
Guo et al. (2021)43	*	*	*	-		*		*	*	*	7
Horibe et al. (2021)44	*	*	*	-		*		*	*	-	6

A study can receive up to four stars for Selection (S1=representativeness of the exposed cohort, S2=selection of the nonexposed cohort, S3=ascertainment of exposure, and S4=demonstration that the outcome of interest was not present at the start of the study), two stars for Comparability (C=comparability of cohorts based on design or analysis), and three stars for Outcome (O1=assessment of outcome, O2=follow-up enough for outcomes to occur, O3=adequacy of follow-up of cohorts)..

This was the most comprehensive systematic review and meta-analysis yet performed examining the outcomes of early endoscopy in patients diagnosed with ANVUGIB. We found that: (1) early endoscopy performed within 24 hours did not lead to a lower mortality rate. Patients who received endoscopy within 24 hours had a higher rate of primary hemostasis and shorter LOS; (2) patients at high risk of bleeding who received very early endoscopy ranging from 3 to 12 hours were more likely to undergo primary hemostasis, and (3) discrepancies exist between studies from populations in the East and West.

In this review, early endoscopy performed within 24 hours had no obvious clinical benefit except for the reduction of LOS among patients with ANVUGIB. Furthermore, the mortality rate was found to be higher in very early endoscopy within 12 hours. These findings were based on the investigation of the overall population without regard to the bleeding risk. Since recently, there has been rising attention on the question of whether early endoscopy is necessary when it comes to high-risk bleeding patients. According to our subgroup analysis, there was no significant difference in mortality if high-risk individuals received endoscopy within 24 hours or shorter timeframes (3 to 12 hours).^{12-14,16-18,33,43,44} What’s more, Lau et al.¹⁵ conducted an RCT to compare endoscopy done within 6 hours with that performed later among 516 high-risk patients. Similarly, no clinical benefit on mortality was observed. But it is noteworthy that this study has excluded patients with persistent hypotensive shock and additionally included patients with variceal bleeding. Additionally, the American College of Gastroenterology updated its guidelines in 2021 and recommended that patients with high-risk bleeding features should be resuscitated and given attention to other active comorbidities rather than undergoing endoscopy at a very early time after admission.²¹ However, this recommendation was made in terms of low-quality observational studies. The impact of confounders and selection bias cannot be avoided.

Clinical outcomes in early endoscopy within 24 hours remained inconsistent between East and West populations. Only one study from the East with a 24-hour timeframe was included, and it found that early endoscopy was associated with a significantly higher rate of primary hemostasis.¹³ However, pooled results of studies from Western populations failed to show any significant difference in primary hemostasis. We postulated that the difference could be explained by both patient- and endoscopist-related factors. On one hand, the increased risk of Helicobacter pylori and genetic differences may make Asians more susceptible to gastrointestinal bleeding,⁴⁶ resulting in a greater need for hemostasis treatment in the East. On the other hand, Tang et al.⁴⁷ conducted a survey to investigate the differences in endoscopic hemostasis practices between the East and West and reported that Asian endoscopists may be more familiar with primary hemostasis than Western endoscopists due to its frequent use in the treatment of endoscopy-associated complications. Furthermore, while studies from the West showed a significantly shorter LOS in the early timing group, the advantage may have been offset by a significantly higher rate of repeat endoscopy and blood transfusion requirement.

We gathered all the information we could find by contacting research authors and sorting trials rigorously by study design (observational cohort studies and RCTs) and bleeding risks. A variety of time frames for endoscopy were retrieved, and a subgroup analysis was conducted to investigate shorter time frames ranging from 3 to 12 hours separately from endoscopy conducted within 24 hours. Finally, we compared the clinical outcomes in the East and West.

However, there are several limitations to this review. First, most of the studies considered in this review had an observational cohort design. Several secondary outcomes exhibited high heterogeneity, which we speculated may be due to differences in the population’s bleeding risk. A subgroup analysis of patients with a high bleeding risk was used to try to account for this. Second, the publication time of included studies was spread out across a period of more than two decades. Since then, endoscopic instruments have advanced and guidelines have been updated. We conducted subgroup analyses for each outcome by eliminating the results from the oldest observational cohort study and the RCT. Except for primary hemostasis in observational cohort studies by eliminating a study by Cooper et al. (p=0.001),¹⁰ no significant difference in other outcomes was identified.

In conclusion, early endoscopy within 24 hours does not appear to significantly reduce the mortality rates of patients with ANVUGIB. It is warranted to explore if very early endoscopy within 12 hours can provide clinical benefit for patients at high risk of bleeding.

Supplementary materials can be accessed at https://doi.org/10.5009/gnl220291.

No potential conflict of interest relevant to this article was reported.

Study concept and design: L.B., W.J. Data acquisition: L.B., W.J. Data analysis and interpretation: L.B., W.J., R.C., Y.D., L.M. Drafting of the manuscript: L.B., W.J. Critical revision of the manuscript for important intellectual content: S.Z. Statistical analysis: L.B. study supervision: S.Z. Approval of final manuscript: all authors.