Topical hemostatic agents in the management of upper gastrointestinal bleeding: a meta-analysis

Abstract

Background and study aims  Novel topical hemostatic agents have shown promising results in treating patients with non-variceal upper gastrointestinal bleeding (NVUGIB). However, data are limited even in published meta-analyses as to their role, especially compared to conventional endoscopic approaches. The aim of this study was to perform a highly comprehensive systematic review assessing the effectiveness of topical hemostatic agents in UGIB in different clinical settings.

Methods  We performed a literature search of OVID MEDLINE, EMBASE, and ISI Web of Knowledge databases through September 2021. Studies assessing the efficacy of topical hemostatic agents in UGIB were included. Main outcomes were immediate hemostasis and overall rebleeding.

Results  A total of 980 citations were identified and 59 studies with a total of 3,417 patients were included in the analysis. Immediate hemostasis was achieved in 93 % (91 %; 94 %), with similar results according to etiology (NVUGIB vs. variceal), topical agent used, or treatment strategy (primary vs. rescue). The overall rebleeding rate was 18 % (15%; 21 %) with the majority of rebleeds occurring in the first 7 days. Among comparative studies, topical agents achieved immediate hemostasis more often than standard endoscopic modalities (OR 3.94 [1.73; 8.96), with non-different overall rebleeding odds (OR 1.06 [0.65; 1.74]). Adverse events occurred in 2 % (1 %; 3 %). Study quality was overall low to very low.

Conclusions  Topical hemostatic agents are effective and safe in the management of UGIB with favorable outcomes when compared to conventional endoscopic modalities across a variety of bleeding etiologies. This is especially true in novel subgroup analyses that assessed immediate hemostasis and rebleeding among RCTs and in malignant bleeding. Due to methodological limitations of available data, additional studies are needed to ascertain their effectiveness more confidently in the management of patients with UGIB.

Introduction

Upper gastrointestinal bleeding (UGIB) is a common emergency condition associated with significant morbidity and mortality ¹ . In the United States, UGIB results in more than 300,000 hospital admissions annually with significant associated costs ^{2 3} . Despite recent advances in the endoscopic management of both non-variceal upper gastrointestinal bleed (NVUGIB) and variceal upper gastrointestinal bleed, some patients fail conventional endoscopic therapy. Indeed, approximately 15 % of patients treated for UGIB fail primary endoscopic therapy, and re-bleeding can occur in up to 25 % of patients after initial successful therapy ⁴ with associated increased morbidity and mortality, especially in patients with variceal bleeding ⁵ .

Recently, novel topical endoscopic hemostatic agents were introduced for the management of UGIB ⁶ . These agents have shown promising results as a primary or salvage therapy while requiring less technical expertise ⁶ . Such agents include, among others the hemostatic powder TC-325 (Hemospray, Cook Medical, Bloomington, Indiana, United State), the starch derived polysaccharide hemostatic system (EndoClot, Endoclot Plus Inc., Santa Clara, California, United States), the biocompatible natural polymer UI-EWD (Nexpowder, NextBiomedical Co, Incheon, South Korea) and more recently the synthetic self-assembling peptide agent (PuraStat, 3D-Matrix, Europe Ltd., France) ^{7 8} ( Table 1 ). Previous studies have concluded that these hemostatic agents are effective and safe in the treatment of NVUGIB with low adverse event (AE) rates ^{9 10 11 12} . However, reported re-bleeding rates have been high (19 % at 72 hours) when used to treat NVUGIB ⁷ . As a result, the international guidelines for the management of ulcer bleeding suggested the hemostatic powder Tc-325 be used as temporizing measure when conventional endoscopic therapy fails but suggested against using it as a monotherapy ¹³ . In contradistinction, the more recent American College of Gastroenterology guidelines recently endorsed such a use ¹⁴ , bolstered by a recently published randomized trial exhibiting potentially limited external validity ^{15 16} . Unfortunately, preventing an adequate characterization of overall effectiveness, many of the published studies and systematic reviews did not completely assess the role of Tc-325 according to bleeding etiology which is thought to be a critical consideration when opting for such therapy. ¹⁶ Moreover, the published meta-analyses also failed to address other topical hemostatic agents in the management of UGIB or in certain situations such as variceal bleeding, even if data addressing these are sparse.

Table 1. Summary of included topical hemostatic agents.

Agent	Trade Name	Composition	Mechanism of action	Approved human application	Formulation
TC-325	Hemospray (Cook Medical, Winston-Salem, North Carolina, USA)	Granular mineral-based	Absorbs H 2 O Forms mechanical tamponade Activates clotting cascade	Nonvariceal gastrointestinal bleed	CO 2 pressurized handler canister (20 g)
EndClot	EndoClot (Endoclot Plus Inc., Santa Clara, California, United States)	Absorbable starch-based modified polysaccharide	Absorbs H 2 O and concentrates cells Activates clotting cascade	Upper and lower gastrointestinal bleeding	Pressurized air compressor
PuraStat	PuraStat (3D-Matrix, Europe Ltd., France)	Synthetic self-assembling peptide agent	Forms a gel coat that induces hemostasis Promotes healing	Bleeding from capillaries and oozing from capillaries of the gastrointestinal tract Prevention of delayed bleeding post colonic ESD	Prefilled syringe
UI-EWD	Nexpowder (NextBiomedical Co, Incheon, South Korea)	Biocompatible natural polymer	Forms adhesive hydrogel in the presence of water	Intended for upper gastrointestinal bleeding	Spraying device
CEGP-003	CEGP-003 (CGBio, Seongnam, South Korea)	Absorbable and adhesive macromolecules containing epidermal growth factor (EGF)	Forms adhesive gel upon contact with moist mucosa facilitating hemostasis Promotes wound healing	Treatment of upper gastrointestinal bleeding Prevention of bleeding post EMR/ESD	Spraying device

ESD, endoscopic submucosal dissection; EMR, endoscopic mucosal resection.

The aim of this systematic review and meta-analysis is thus to more completely characterize the effectiveness and safety of hemostatic agents in achieving and maintaining hemostasis when managing patients with varying etiologies of UGIB in order to better and more comprehensively guide clinician decisions and future societal recommendations.

Methods

Search strategy

A comprehensive literature search was performed, from the recorded start of databases to September 2021 using OVID MEDLINE, EMBASE, and ISI Web of Knowledge databases, with a combination of MeSH term and controlled vocabulary to identify studies related to: 1) hemostatic agent or powder and; 2) gastrointestinal bleeding (Supplementary Table 1). Abstracts presented at major gastroenterology conferences (ACG, CDDW, DDW, UEGW, APDW) in the past 5 years were also hand-searched. Additional relevant studies were identified from cross-referencing and hand-searches of references of retrieved articles.

Validity assessment, data abstraction and rating of evidence

Two reviewers (AA, MM) evaluated the eligibility of all identified citations independently, with a third (AB) resolving disagreements. Study quality was assessed using the Cochrane Risk of bias tool for randomized trials ¹⁷ , and the Ottawa-Newcastle criteria for observational studies ¹⁸ . We used the GRADE rating to characterize the certainty of evidence ¹⁹ .

Inclusion and exclusion criteria

Studies assessing any hemostatic agent in UGIB of any etiology were considered for inclusion. Both manuscripts and abstracts from major gastroenterology societies (limited to the previous 5 years) were considered for inclusion. Only studies published in English were included. We excluded case reports, studies with less than 10 patients, review articles, and non-human studies. In addition, we excluded studies reporting on the hemostatic agent “Ankaferd” due to its limited availability globally ⁶ . Studies that used hemostatic agents prophylactically to prevent GI bleeding were also excluded.

Study definitions

“Immediate hemostasis” was defined as no further bleeding at least 3 minutes after application of the hemostatic agent. “Rebleeding” was defined as evidence of recurrent UGIB manifested as overt gastrointestinal bleeding or drop in hemoglobin ≥ 2 g/dL after achieving immediate hemostasis ²⁰ . Rebleeding data were included only for patients following initial immediate hemostasis. “Technical success” was defined as successful use of the hemostatic agent during endoscopy without any technical problems (e. g. blockage of applications catheter). “Monotherapy” was defined as the use of a hemostatic agent alone, whereas “combination therapy” was defined as using the hemostatic agent in combination with adjunctive conventional endoscopic methods. “Rescue therapy” was defined as the use of a hemostatic agent when other conventional endoscopic methods had failed as evidenced by failure of immediate hemostasis. “Primary therapy” was defined as using the hemostatic agent as first-line endoscopic therapy for bleeding.

Outcomes

Immediate hemostasis and overall rebleeding (defined as any rebleeding during follow-up after the index endoscopy) were the outcomes of interest. Other outcomes assessed included 7-day and 30-day rebleeding, overall-mortality, bleeding-related mortality, technical success, and AEs related to application of the topical hemostatic agent. Data will be present in turn for a meta-analysis of proportions (purely descriptive) and in a meta-analysis assessing studies that compared hemostatic powders to a control therapeutic approach.

Sensitivity and subgroup analyses

Pre-planned possible subgroup and sensitivity analyses included assessments according to: type of hemostatic agent, type of therapies (primary therapy, rescue therapy, monotherapy, combination therapy), type of lesions (peptic ulcer disease (PUD), post endoscopic intervention, varices, malignant lesions), randomized-controlled trials alone; fully published articles alone, higher quality studies, year of publication, continent where the study was performed, performing a fixed rather than a random effect model (when appropriate), and when correcting for double-zero events.

Statistical analysis

Categorical estimates of outcomes were reported as proportions and 95 % confidence intervals (CI) using weighted random effects models. Continuous variables are reported as means and standard deviations medians were used if means were not available and standard deviations (SDs) were calculated or imputed when possible. For comparative studies, effect size was calculated with weighted mean differences (WMDs) for continuous variables. Odds ratios (ORs) are expressed for categorical variables. The DerSimonian and Laird method for random effect models was applied to all outcomes to determine corresponding overall effect sizes and their confidence intervals. Sensitivity analyses were performed using the Mantel-Haenszel method with random effect models; however fixed effects models were used when no statistical heterogeneity was noted. WMD were handled as continuous variables using the inverse variance approach. Presence of heterogeneity across studies was defined using a Chi-square test of homogeneity with a 0.10 significance level. The Higgins I ² statistic was calculated to quantify the proportion of variation in treatment effects attributable to between-study heterogeneity, with values of 25 %, 50 %, and 75% representing low, moderate, and high heterogeneity, respectively. When heterogeneity was noted, prediction intervals were calculated and added to the forest plot. The prediction interval calculates the 95 % of where the effect size will be if a new study is randomly added to the meta-analysis. In order to ensure that zero event trials did not significantly affect the heterogeneity or p-values, sensitivity analyses were performed where a continuity correction was added to each trial with zero events using the reciprocal of the opposite treatment arm size. For all comparisons. Publication bias was evaluated using funnel plots if at least three citations were identified. All statistical analyses were done using Revman 5.4 and Meta package in R version 2.13.0, (R Foundation for Statistical Computing, Vienna, Austria, 2008).

Results

Included studies, quality assessment and publication bias

We initially identified 980 citations. After review of abstracts, a total of 59 studies were included ( Fig. 1 ). One study was excluded since it reported results as per episode of bleeding rather than per patient outcomes ²¹ .

Fig. 1.

 PRISMA diagram of included studies. From: Page MJ, McKenzie JE, Bossuyt PM et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021; 372:n71.

For the outcomes in the (solely descriptive) meta-analysis of proportions, moderate to high heterogeneity was present for immediate hemostasis, rebleeding (overall, 7 days and 30 days) and overall mortality. No significant heterogeneity was noted in the portion of the meta-analysis addressing between-group comparative results. No publication bias was observed (data available upon request).

Study quality was overall low-very low with the Ottawa Newcastle Scores (NOS) ranging from 4 to 8 stars out of a possible score of 9, with a mean of 5.9 ± 0.8 for observational studies. Assessing the individual domains of the NOS confirmed the low studies’ quality (Supplementary Table 2). The Cochrane risk bias tool revealed a high potential for performance bias across studies for randomized trials since all were single-blinded (Supplementary Fig. 1). The grading of the evidence was performed in studies that included a comparative arm and was found to be “very low” for all outcomes (Supplementary Table 3).

Patient and study characteristics

Overall, 59 studies were included (n = 3417 patients). Hemospray (TC-325) was the sole intervention in 44 studies ^{15 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64} , EndoClot in five ^{65 66 67 68 69} , PuraStat in four ^{70 71 72 73} , Nexpowder in three ^{74 75 76} and CEGP-003 ⁷⁷ in one. In addition, two studies included both Hemospray and EndoClot ^{78 79} . Seven were RCTs, and 15 were prospective while 37 were retrospective cohort studies. Furthermore, 44 were fully published articles, while the remainder were sole abstracts ( Table 2 ). Ten studies included a comparison with standard endoscopic therapy ^{15 24 30 39 42 47 52 65 69 77} . Reports recruited patients between 2009 and 2021. Most of the bleeding lesions were classified as oozing bleeding (Forrest Ib). There were 11 studies ^{26 27 28 37 38 49 50 57 60 64 78} (n = 124 patients) that addressed patients with variceal bleeding. Full study and patient characteristics are shown in Table 2 .

Table 2. Patient and study characteristics.

Study Country	Intervention	Comparison	Sample size	Design	Study period	Indication	Etiology	Bleeding severity
Arena et al. 2017 23 Italy	Hemospray	None	15	Retrospective cohort	2014–2015	Primary therapy	Malignancy (100 %)	Oozing (100 %)
Appleby et al. 2017 22 UK Abstract	Hemospray	None	32	Retrospective cohort	2014–2017	Primary therapy Rescue therapy	PUD (56 %) Malignancy (9 %) Dieulafoy lesion (28 %) Others (7 %)	NR
Baracat et al. 2020 24 Brazil	Hemospray	Hemoclips	39	RCT	2015–2017	Primary therapy	PUD (43.6 %) Malignancy (12.8 %) Post intervention (10.3 %) Dieulafoy lesion (5.1 %) MalloryWeiss tear (7.7 %) Others (20.5 %)	Spurting (10.3 %) Oozing (89.7 %)
Bestari et al. 2020 26 Indonesia	Hemospray	None	30	Retrospective cohort	2016–2019	Primary therapy Rescue therapy	PUD (56.7 %) Malignancy (26.7 %) Variceal bleed (10 %) Portal hypertensive gastropathy (6.7 %)	NR
Becq et al. 2020 25 France	Hemospray	None	152	Retrospective cohort	2015–2018	Primary therapy Rescue therapy	PUD (47.7 %) Malignancy (22.2 %) Esophagitis (12.4 %) Other (17.7 %)	Oozing (84.3 %) spurting (11.1 %) None active (2.6 %) Undetermined (2.0 %)
Chahal et al. 2020 28 Canada	Hemospray	None	73	Retrospective cohort	2014–2018	Primary therapy Rescue therapy	PUD (67.1 %) Malignancy (5.5 %) Variceal bleed (5.5 %) MWT (8.2 %) Other (13.7 %)	For PUD (n = 49) Forrest Ia (14.3 %) Forrest Ib (53.1 %) Forrest IIa (20.4 %) Forrest IIb (14.3 %) Forrest III (4.1 %)
Cahyadi et al. 2017 27 Germany	Hemospray	None	52	Retrospective cohort	2013–2017	Primary therapy Rescue therapy	PUD (34.6 %) Malignancy (19.2 %) Post procedure (30.7 %) Dieulafoy lesion (1.9 %) GAVE (1.9 %) Variceal bleed (3.8 %) Other (7.6 %)	Peptic ulcer only, n = 18 Forrest Ib (38.9 %) Forrest IIa (33.3 %) Forrest IIb (22.2 %) Forrest IIc (5.6 %)
Chen et al. 2015 29 Canada	Hemospray	None	56	Retrospective cohort	2011–2013	Primary therapy Rescue therapy	Benign NVUGIB (37.5 %) Malignancy (33.9 %) Post intervention (28.6 %)	For PUD (n = 13) Forrest Ia (23.1 %) Forrest Ib (76.9 %)
Chen et al. 2020 30 Canada	Hemospray	CHP	17	RCT	2014–2016	Primary therapy	Malignancy (100 %)	Spurting (5 %) Oozing (95 %)
De Santiago et al. 2019 60 Spain	Hemospray	None	219	Retrospective cohort	2011–2018	Primary therapy Rescue therapy	PUD (33.3 %) Malignancy (21.0 %) Post-procedure (14.2 %) postsurgical (6.4 %) Dieulafoy lesion (3.7 %) Variceal bleed (3.2 %) GAVE (0.9 %) Others (17.4 %)	Peptic ulcer only, n = 73 Forrest Ia (24.7 %) Forrest Ib (64.4 %) Forrest IIa (5.5 %) Forrest IIb (5.5 %)
Disney et al. 2015 31 UK Abstract	Hemospray	None	19	Retrospective cohort	NR	Primary therapy Rescue therapy	Malignancy (100 %)	Oozing (100 %)
Giles et al. 2016 32 New Zeland	Hemospray	None	36	Retrospective cohort	2013–2016	Rescue Therapy	PUD (66.7 %) Malignant (5.6 %) post intervention (16.7 %) Others (11.0 %)	Spurting (13.9 %) Oozing (69.4 %) Visible vessel (5.6 %) Adherent clot (5.6 %) Clean base (5.6 %)
Gupta et al. 2018 33 UK Abstract	Hemospray	None	45	Retrospective cohort	2013–2017	Primary therapy Rescue therapy	PUD (41 %) Malignancy (15 %) Post-intervention (15 %) Other (29 %)	Forrest Ia (15 %) Forrest Ib (46 %) Forrest IIa (15 %) Forrest IIb (15 %) Forrest IIc (9 %)
Haddara et al. 2016 34 France	Hemospray	None	202	Retrospective cohort	2013–2015	Primary therapy Rescue therapy	PUD (37.1 %) Malignant (30.2 %) Post-intervention (17.3 %) Dieulafoy lesion (1.5 %) Others (3.9 %)	For PUD (n = 75) Forrest Ia (20.0 %) Forrest Ib (57.3 %) Forrest II (18.7) Unclassified (4.0 %)
Hagel et al. 2017 35 Germany	Hemospray	None	33	Retrospective cohort	2013–2014	Rescue Therapy	PUD (48.5 %) Malignancy (12.1 %) Post intervention (12.1 %) Diffuse bleeding (18.2 %) Others (9.1 %)	NR
Holster et al. 2015 36 Netherland	Hemospray	None	16	Retrospective cohort	2011–2012	Primary therapy Rescue therapy	PUD (56.0 %) Malignant (13.0 %) Others (31.0 %)	For PUD (n = 9) Forrest Ia (55.5 %) Forrest Ib (44.4 %)
Hussein et al. 2021 (1) 62 UK	Hemospray	None	105	Prospective study	2016–2020	Primary therapy Rescue therapy	Malignancy (100 %)	Spurting (6 %) Oozing (77 %) Visible vessel/adherent clot (12 %)
Hussein et al. 2021 (2) 61 UK	Hemospray	None	202	Prospective cohort	2016–2019	Primary therapy Rescue therapy	PUD (100 %)	Forrest Ia (19.0 %) Forrest Ib (58.0 %) Forrest IIa (12.0 %) Forrest IIb (10.0 %)
Hussein et al. 2021 (3) 64 UK Abstract	Hemospray	None	12	Prospective study	2016–2019	Primary therapy Rescue therapy	Esophageal varices (83.3 %) Gastric varices (16.7 %)	NR
Hussein et al. 2020 63 UK	Hemospray	None	73	Prospective cohort	2016–2019	Primary therapy Rescue therapy	Post-procedure (100 %)	NR
Ibrahim et al. 2015 37 Belgium	Hemospray	None	30	Prospective cohort	2013–2014	Primary Therapy	Esophageal varices (83.4 %) Gastric varices (10 %) Duodenal varices (6.6 %)	Spurting (43.4 %) Fibrin plug/red streaks (56.6 %)
Ibrahim et al. 2019 38 Belgium	Hemospray	Early elective endoscopy	86	RCT	2014–2016	Primary therapy	Variceal bleed (100 %)	Spurting (16.3 %) Blood in stomach (83.7 %)
Kwek et al. 2017 39 Singapore	Hemospray	Conventional endoscopic therapy	20	RCT	2013–2015	Primary therapy	PUD (100 %)	Forrest Ia (5.0 %) Forrest Ib (35.0 %) Forrest IIa (50.0) Forrest IIb (10.0 %)
Lau et al. 2020 15 China	Hemospray	Conventional endoscopic therapy	224	RCT	2015–2018	Primary therapy	PUD (58 %) Malignancy (14.3 %) Others (27.7 %)	NR
Leblanc et al. 2013 40 France	Hemospray	None	17	Retrospective cohort	2011–2012	Primary therapy Rescue therapy	Post intervention (70.6 %) Malignancy (29.4 %)	For post-intervention (n = 12) pulsatile (8.3 %) oozing (91.7 %)
Malik et al. 2015 41 UK Abstract	Hemospray	None	19	Retrospective cohort	2013–2014	Primary therapy Rescue therapy	PUD (53 %) Malignancy (21 %) Post intervention (21 %) Other (5 %)	For PUD (n = 10) Forrest Ia (0 %) Forrest Ib (60 %) Forrest IIa (30 %) Forrest IIb (10 %)
Martins et al. 2019 42 Brazil Abstract	Hemospray	CHP	36	RCT	2016–2017	Primary therapy	Malignancy (100 %)	Active bleeding (63.9 %)
Masci et al. 2014 43 Italy	Hemospray	None	13	Prospective cohort	NR	Primary therapy	PUD (100 %)	Forrest Ia (35.7 %) Forrest Ib (64.3 %)
Meng et al. 2018 44 Canada	Hemospray	None	25	Retrospective study	2010–2016	Primary therapy Rescue therapy	Malignancy (100 %)	Forrest Ia (8.0 %) Forrest Ib (76.0 %) unclassified (16.0 %)
Min et al. 2018 45 UK Abstract	Hemospray	None	48	Retrospective cohort	2016–2017	Primary therapy Rescue therapy	PUD (69.0 %) Malignancy (2.0 %) Post- intervention (4.0 %) GAVE (4.0 %) Esophagitis (4.0 %) Others (16.0 %)	NR
Nasr et al. 2015 46 UK Abstract	Hemospray	None	26	Retrospective cohort	2013–2015	Primary therapy Rescue therapy	NVUGIB (65.4 %) Post-intervention (34.6 %)	NR
Paoluzi et al. 2021 79 Italy	Hemospray EndoClot	CHP	108	Retrospective cohort	2017–2019	Primary therapy Rescue therapy	PUD (66.7 %) Malignancy (21.3 %) Post-intervention (6.4 %) Others (5.5 %)	For PUD (n = 72) Forrest IA (20.8 %) Forrest IB (79.2 %) For malignancy (n = 23) oozing (100 %)
Pittayanon et al. 2016 47 Thailand	Hemospray	CHP	20	Prospective cohort	2014–2015	Primary therapy	Malignancy (100 %)	oozing blood (100 %
Pittayanon et al. 2018 48 Canada	Hemospray	None	79	Retrospective cohort	2011–2016	Primary therapy Rescue therapy	Malignancy (100 %)	Adherent clot (4.5 %) Blood oozing (94.3 %) Blood spurting (1.1 %)
Prentice et al. 2018 49 UK Abstract	Hemospray	None	47	Retrospective cohort	2014–2017	Primary therapy Rescue therapy	NVUGIB (78.7 %) Variceal bleed (21.3 %)	NR
Ramirez-Polo et al. 2019 50 Mexico	Hemospray	None	81	Retrospective cohort	2015–2017	Primary therapy	PUD (17.3 %) Malignancy (43.2 %) post-procedure (14.8 %) Dieulafoy lesion (6.1 %) Variceal bleed (2.5 %) Postsurgical (2.5 %) Other (13.6 %)	NR
Shivaji et al. 2018 51 UK Abstract	Hemospray	None	45	Retrospective cohort	2013–2017	Primary therapy Rescue therapy	PUD (41.0 %) Malignancy (15.0 %) Post-intervention (15.0 %) Others (29.0 %)	Peptic ulcer only, n = 22 Forrest Ia (15.0 %) Forrest Ib (46.0 %) Forrest IIa (15.0 %) Forrest IIb (15.0 %) Forrest IIc (9.0 %)
Sinha et al. 2016 52 UK	Hemospray	CHP	40	Retrospective cohort	2013–2015	Rescue therapy	PUD (65.0 %) Esophageal ulcer (20 %) Others (15.0 %)	Forrest Ia (60.0 %) Forrest Ib (40.0 %)
Smith et al. 2014 53 UK	Hemospray	None	63	Prospective cohort	2011	Primary therapy Rescue therapy	PUD (57.1 %) Malignancy (12.6 %) Post-procedure (15.9 %) Dieulafoy lesion (3.2 %) MWT (3.2 %) GAVE (3.2 %) Other (5.8 %)	Peptic ulcer only, n = 30 Forrest Ia (36.7 %) Forrest Ib (53.3 %) Unclassified (10 %)
Sulz et al. 2014 54 Switzerland	Hemospray	None	15	Prospective cohort	2013	Primary therapy Rescue therapy	PUD (25.0 %) Malignancy (18.8 %) Post-intervention (12.5 %) Others (43.8 %)	Spurting (18.8 %) oozing (81.2 %)
Sung et al. 2011 55 Hong Kong	Hemospray	None	20	Prospective cohort	2009–2010	Primary Therapy	PUD (100 %)	Forrest Ia (5 %) Forrest Ib (95 %)
Thayalasekaran et al. 2017 56 UK Abstract	Hemospray	None	44	Retrospective cohort	2014–2016	Primary therapy Rescue therapy	PUD (65.9 %) Malignancy (2.3 %) Variceal bleed (6.8 %) MWT (4.5 %) Post-intervention (12.3 %) Others (8.2 %)	Forrest Ia (26.5 %) Forrest Ib (44.1 %) Forrest IIa (23.5 %) Forrest IIb (5.9 %)
Vitali et al. 2019 78 Germany	Hemospray	EndoClot	127	Prospective cohort	2013–2017	Primary therapy Rescue therapy	PUD (38.6 %) Malignancy (11.8 %) Variceal bleed (10.2 %) Reflux esophagitis (9.4 %) Others (30.0 %)	NR
Weaver et al. 2019 57 USA Abstract	Hemospray	None	12	Prospective cohort	2018–2019	Primary therapy Rescue therapy	PUD (50.0 %) MWT (16.7 %) Variceal bleed (8.3 %) GAVE (8.3 %) Post-intervention (16.7 %)	Peptic ulcer only, n = 6 Forrest Ia (50.0 %) Forrest Ib (50.0 %)
Widlak et al. 2015 58 UK Abstract	Hemospray	None	48	Retrospective cohort	2013–2015	Primary therapy Rescue therapy	PUD (60.5 %) Other (27 %) Post banding ulcer (12.5 %)	NR
Yau et al. 2014 59 Canada	Hemospray	None	19	Retrospective cohort	2012–2013	Primary therapy Rescue therapy	PUD (63.2 %) Dieulafoy lesion (10.5 %) Mucosal erosion (5.3 %) Angiodysplasia (5.3 %) Post-intervention (10.6 %) Unknown (5.3 %)	Spurting (21.1 %) oozing (57.9 %) No active bleeding (21.1 %)
Branchi et al. 2021 72 Germany	PuraStat	None	78	Prospective cohort	2017–2018	Primary therapy Rescue therapy	PUD Post-intervention Malignancy Vascular lesions	spurting (6 %) Oozing (69 %) Visible vessel (14 %) Adherent clot (5 %) Flat pigmented spot (5 %)
de Nucci et al. 2020 70 Italy	PuraStat	None	41	Retrospective cohort	2017–2019	Rescue therapy	PUD (46.3 %) Malignancy (7.3 %) Post-intervention (39.0 %) Other (7.4 %)	Spurting (19.5 %) Oozing (80.5 %)
Labianca et al. 2021 73 Italy Abstract	PuraStat	None	15	Retrospective cohort	2018–2020	Primary therapy Rescue therapy	PUD (86.7 %) Post-intervention (13.3 %)	NR
Subramaniam et al. 2019 71 UK	PuraStat	None	44	Retrospective cohort	2016–2017	Primary therapy	Post-intervention (100 %)	NR
Beg et al. 2015 65 UK	EndoClot	CHP	130	Retrospective cohort	2012–2014	Rescue therapy	PUD (90 %) Esophageal ulcer (4.6 %) MWT (3.8 %) Other (1.6 %)	NR
Hagel et al. 2020 68 Germany	EndoClot	None	22	Retrospective cohort	2015–2020	Primary therapy Rescue therapy	PUD (40.9 %) Malignancy (40.9 %) Others (18.2 %)	NR
Kim et al. 2018 66 South Korea	EndoClot	None	12	Retrospective cohort	2016–2017	Primary therapy	Malignancy (100 %)	Forrest Ib (100 %)
Park et al. 2019 69 South Korea	EndoClot	CHP	176	Retrospective cohort	2012–2017	Primary therapy Rescue therapy	PUD (68.8 %) Malignancy (21.6 %) Post-intervention (5.7 %) Radiation gastritis (3.9 %)	Forrest Ia (9.7 %) Forrest Ib (54.5 %) Forrest IIa (35.8 %)
Prei et al. 2016 67 Germany	EndClot	None	58	Prospective cohort	2012–2014	Primary therapy Rescue therapy	PUD (46.6 %) Malignancy (17.2 %) Post-intervention (10.3 %) Esophagitis (10.3 %) Other (15.6 %)	Forrest Ia (5.2 %) Forrest Ib (65.5 %) Forrest IIa (10.3 %) Forrest IIb (6.9 %) Forrest IIc (12.1 %)
Park et al. 2019(1) 74 South Korea	Nexpowder	None	17	Prospective cohort	2016–2017	Rescue therapy	PUD (29.4 %) Malignancy (23.5 %) Post-intervention (41.2 %) Other (5.9 %)	Forrest Ia (11.8 %) Forrest Ib (88.2 %)
Park et al 2019 (2) 75 South Korea Abstract	Nexpowder	None	56	Retrospective cohort	NR	Primary therapy	PUD (14.3 %) Malignancy (1.8 %) Post-intervention (82.1 %) Other (1.8 %)	NR
Shin et al. 2021 76 South Korea	Nexpowder	None	41	Retrospective cohort	2016–2019	Primary therapy Rescue therapy	Malignancy (100 %)	Forrest Ia (7.3 %) Forrest Ib (92.7 %)
Bang et al. 2018 77 South Korea	CEGP-003	Epinephrine injection	72	RCT	2014–2015	Primary therapy	PUD (22.2 %) Post-intervention (77.8 %)	Forrest Ib (83.3 %) Forrest IIa (6.9 %) Forrest IIb (9.8 %)

PUD, peptic ulcer disease; RCT, randomized controlled trial; CHP, conventional hemostatic procedures; MWT, Mallory-Weiss tearing; GAVE, gastric antral vascular ectasia; NVUGIB, non-variceal upper gastrointestinal bleeding.

Immediate hemostasis and overall rebleeding

Immediate hemostasis was reported in 59 studies (n = 2,919 patients) with a pooled success rate of 93 % (91 %; 94 %) ( Fig. 2 , Table 3 ). Overall rebleeding was included in 58 studies (n = 2,696) with a pooled overall rebleeding rate of 18 % (15 %; 21%) ( Fig. 3 , Table 3 ). Rebleeding rates at 7 and 30 days were 17 % (14 %; 20 %) and 21 % (17 %; 27 %), respectively ( Table 3 ).

Fig. 2.

 Forrest plot of primary outcome by proportion: immediate hemostasis.

Table 3. Primary and secondary outcome for proportions.

	No. studies	No. patients	Proportion (95 % CI)	P value for heterogeneity	I 2
Primary outcome
Immediate hemostasis (overall UGIB)	59	2919	0.93 (0.91; 0.94)	< 0.01	67 %
Overall rebleeding	58	2696	0.18 (0.15; 0.21)	< 0.01	69 %
Rebleeding 7 days	42	1943	0.17 (0.14; 0.20)	< 0.01	55 %
Rebleeding 30 days	34	1692	0.21 (0.17; 0.26)	< 0.01	75 %
Secondary outcome
Overall mortality	45	2245	0.15 (0.12; 0.19)	< 0.01	64 %
Bleeding-related mortality	34	1563	0.05 (0.04; 0.07)	0.42	3 %
Technical success	52	2392	0.97 (0.96; 0.98)	0.99	0 %
Adverse events	45	2111	0.02 (0.01; 0.03)	0.99	0 %

UGIB, upper gastrointestinal bleeding.

Fig. 3.

 Forrest plot of primary outcome by proportion: overall rebleeding.

In the 10 studies comparing the topical hemostatic agent to a control intervention (n = 797), immediate hemostasis was more significantly achieved with the former (odds ratio [OR] = 3.94 (1.73; 8.96) ( Table 4 and Fig. 4 ). Among the comparative studies, the overall rebleeding risk was not significantly different between topical agents and conventional endoscopic therapy (OR = 1.06; 0.65–1.74, 10 studies, n = 775); this was also the case for rebleeding at 7 and 30 days (OR = 0.97; 0.43–2.16) and OR = 0.75; 0.39–1.45, respectively) ( Table 4 and Fig. 5 ).

Table 4. Primary and secondary outcome for comparative studies (comparing topical agent vs. conventional endoscopic therapy).

	No. studies	No. patients	Odds ratio (95 % CI)	P value for heterogeneity	I 2
Primary outcome
Immediate hemostasis (overall UGIB)	10	797	3.94 (1.73; 8.96)	0.79	0 %
Overall rebleeding	10	775	1.06 (0.65; 1.74)	0.58	0 %
Rebleeding 7 days	6	356	0.97 (0.43; 2.16)	0.51	0 %
Rebleeding 30 days	7	649	0.75 (0.39; 1.45)	0.24	25 %
Secondary outcome (UGIB)
Overall mortality	9	621	1.05 (0.64; 1.70)	0.79	0 %
Bleeding-related mortality	5	347	0.62 (0.14; 2.71)	0.45	0 %
Technical success	9 1	573	0.30 (0.01; 8.33)	–	–
Adverse events	5 2	185	–	–	–

¹Only one study with estimable data.

²All were double zero-event.

Fig. 4.

 Forrest plot of primary outcome for comparative studies: immediate hemostasis.

Fig. 5.

 Forrest plot of primary outcome for comparative studies: overall rebleeding.

Other outcomes

Pooled proportions for overall mortality (45 studies, n = 2,245) and bleeding-related mortality (34 studies; n = 1,563) were 15% (12 %; 19 %) and 5 % (4 %; 7 %), respectively. In terms of technical success (52 studies, n = 2,392) and AEs (45 studies, n = 2,111), analyses yielded pooled proportions of 97 % (96 %; 98 %) and 2% (1 %; 3 %), respectively.

In studies that included a comparison group, none of these outcomes differed significantly when comparing the topical agent to the control hemostatic modality(ies) (table 4).

Sensitivity and subgroup analyses

Results for sensitivity analyses for immediate hemostasis and rebleeding are described in Appendix 1–4. Results confirm the robustness of the findings described in the primary analysis.

When analyzed according to etiology of UGIB, immediate hemostasis rates were high for all etiologies of UGIB including PUD (91 % [88 %; 93 %]) malignancy (94 % [91 %; 95 %]), post-endoscopic intervention (90 % [85 %; 93 %]), and a variceal etiology of bleeding (87 % [79 %; 92 %]). The risk of rebleeding for NVUGIB was 22 % (18 %; 27 %), for variceal bleeding was 23 % (10 %; 45 %), and for malignant bleeding, in particular, was 24% (19 %; 30 %) (Appendix 2).

When grouped according to indication, immediate hemostasis rates were similar when these agents were used as primary or rescue therapy (93 %, [89 %; 95 %] and 90 %, [85 %; 93%], respectively); these rates varied between 86 % to 99 % according to the topical agent assessed. The risk of rebleeding was 25 % (20 %; 30 %) when the topical agent was used as rescue therapy, and 18 % (14 %; 22 %) when used as primary therapy.

When studied according to topical agent studied, immediate hemostasis among the different agents was (Hemospray 93 % [91 %; 95 %]), EndoClot (87 % [75 %; 94 %]), PuraStat (86 % [80 %; 90 %]) Nexpowder (96 % [91 %; 99 %]) and CEGP-003 (99% [81 %; 100 %]).The risk of rebleeding was Hemospray 20 % (17 %; 24 %), EndoClot (10 % [6 %; 16 %]), PuraStat (7 % [1 %; 27%]) Nexpowder (8 % [3 %; 20 %]) and CEGP-003 (9 % [3 %; 23%]).

The subgroup analysis of sole RCTs yielded similar outcomes with a proportion of immediate hemostasis of 94 % (89 %; 97 %), and a rebleeding rate of 26 % (11 %; 48 %). Among Sensitivity analysis using comparative RCTs data confirmed that immediate hemostasis was more frequently achieved with topical hemostatic agents compared to conventional endoscopic therapy (3.62 [1.29; 10.11]), with similar odds of rebleeding (1.27 [0.62;2.59]). Topical agents were more effective when used as primary therapy (3.83 [1.59,9.24]) and when used for malignant lesions (14.74 [2.16,100.61]) (Appendix 3 and 4).

Discussion

To our knowledge, this is the largest, most granular and up-to-date meta-analysis that has characterized the efficacy and safety of different topical hemostatic agents according to various UGIB etiologies. Results confirm the effectiveness and safety of topical hemostatic agents when used to treat UGIB, however with a low to very low certainty of evidence for the data that these observations are based on, as is also the case for argon plasma coagulation and soft monopolar electrocoagulation for which there is also less published evidence ¹⁴ .

Overall, topical hemostatic agents achieve high immediate hemostasis rates (93 %, (91 %; 94 %)). This conclusion is true regardless of the cause of UGIB, including peptic ulcer disease, malignancy, post-endoscopic intervention, and even variceal bleeding. These impressive results remain robust whether the agent is used for primary therapy or rescue therapy. Furthermore, these results are comparable to published effectiveness for other conventional endoscopic modalities such as clips and thermocoagulation (98.5 % and 94.5 %, respectively) ⁸⁰ . Direct comparisons between topical hemostatic agents and conventional endoscopic modalities were only available in 10 studies (n = 797), with topical agents achieving greater immediate hemostasis (OR 3.94 (1.73; 8.96)). This conclusion remained true when including only data derived from RCT (OR 4.01 (1.52; 10.60). This latter subgroup analysis yields a very important summary result previously not reported by other systematic reviews and for which the small number of patients is counterbalanced by the study quality and magnitude of treatment effect ^{10 11 12} . When different topical hemostatic agents were analyzed separately, all studied compounds appeared effective, although the number of patients studied varied greatly.

The major concern about the use of topical hemostatic agents expressed in previous studies has related to the risk of rebleeding. We found that the overall risk of rebleeding was 18 % (15 %; 21 %) among non-comparative studies. This risk is higher compared to conventional endoscopic approaches ^{80 81} . This elevated risk likely represents the inclusion of difficult-to-treat lesions (e. g. large ulcers or difficult position) in which these agents are best used as temporizing strategy until a more definitive endoscopic therapy can be employed. It may also relate to the mechanism of action of some of these agents that form a mechanical barrier to stop bleeding but washes off while a risk of rebleeding may persist (within 24 hours as in the case of Tc-325 and peptic ulcers for which the high-risk period post-hemostasis extends to at least 72 hours ²⁹ ). However, studies that directly compared topical agents with conventional endoscopic modalities failed to show a difference in overall rebleeding between the two interventions (OR 1.06 [0.65; 1.74]). Similar conclusions were reached when RCT data only were included, even though the point estimate of overall rebleeding was actually greater for the topical agents, but with a large confidence interval owing to small patient numbers (1.27 [0.62; 2.59]). This observation is congruent with the hypothesis that the risk of rebleeding when using topical agents may have been over-estimated from previous studies, possibly reflecting selection and/or reporting bias. However, it is important to once again emphasize that RCTs comparing topical agents to conventional endoscopic modalities with regards to assessing rebleeding are few, with many exhibiting limitations. For NVUGIB, there were six RCTs with varying sample sizes. The main limitations with all these studies are two-fold: First, Kwek et al. ³⁹ included a large proportion (60 %) of patients with non-bleeding lesions (Forrest IIa and IIb) – lesions to which Hemospray cannot adhere and thus are not indicated for its use. Second, the remaining studies (Baracat et al. ²⁴ , Chen et al. ²⁹ , Martins et al. ⁴² and Bang et al. ⁷⁷ ) were small (20–72 patients in each study) and included a large proportion of patients with oozing stigmata (Forrest Ib). This category of lesions has recently been shown to exhibit a lower risk of re-bleeding than previously thought, possibly reflecting the inclusion of lower risk lesions than “true” Forrest Ib ⁸² . By far, the largest RCT assessing Hemospray, was recently published by Lau et al. ¹⁵ This study randomized 224 patients with NVUGIB to Hemospray or conventional endoscopic therapy. It concluded that Hemospray was non-inferior to conventional endoscopic hemostasis. However, the study was limited by the inclusion of a large proportion of Forrest Ib lesions. Furthermore, this study exhibited a marked imbalance in the malignant bleeding subgroup which were over-represented in the Hemospray arm and may have biased the results in favor of the topical agent ¹⁶ . Only one study assessed variceal bleeding (Ibrahim et al. ³⁸ ) limiting the conclusions specifically addressing this subgroup of patients.

When the risk of rebleeding was stratified according to the agent used, it was numerically greater with Hemospray (20 % [17 %; 24 %]) compared to other agents (EndoClot 10 % [6 %; 16%], PuraStat 7 % [1 %; 27 %], or Nexpowder 8 % [3 %; 20 %]) and CEGP-003 (9 % [3 %; 23 %]). This difference may not be statistically significant (there are very few direct comparisons) and may only reflect low numbers of patients treated with the other agents or could alternately be a result of the inclusion of higher-risk lesions in the Hemospray studies. Indeed, only one study directly compared different topical hemostatic agents (Hemospray vs. EndoClot) ⁷⁷ , concluding similar effectiveness, as well as rebleeding rates and safety profiles when used for UGIB of different etiologies in 127 patients.

When stratified by etiology, the risk of rebleeding was similar in NVUGIB, malignancy-related lesions, as well as in variceal bleeding (Appendix 2b). Even though the risk of rebleeding was numerically greater when the topical agent was used for rescue (25 % [20 %; 30 %]) compared to primary therapy (18 % [14 %; 22 %]); this difference is likely to be non-significant as suggested by marked overlapping confidence intervals. Given the ease of use of these topical agents, they comprise an attractive option to manage UGIB in emergency situations. These compounds may also prove to be the preferred endoscopic modalities when managing diffuse bleeding from fragile surfaces such as in malignancy, especially considering the difficulty and limited success when using conventional endoscopic options in such a scenario ¹ . This is supported by the current meta-analysis results as we showed that topical agents were more effective in achieving immediate hemostasis than conventional endoscopic therapy when used for malignancy-related bleeding (OR = 14.74 [2.16;100.61]). This, too, is an important finding that bears important clinical implications and that has not been clearly identified previously using summary data.

One of the major strengths of the topical hemostatic agents is their remarkable safety profile. Indeed, the overall AE rate seen with these agents was 2 % (1;3 %). Some of the reported AEs were serious but remain rare, such as the three cases of perforations among 2111 patients. Another important advantage when using these agents is the high technical success rate noted (97 %, 97;98 %). The main technical problem reported was catheter blockage due to premature activation of the powder inside the delivery catheter. This problem was encountered with Hemospray only, and can be avoided by ensuring the endoscope channel and actual delivery catheter are cleared of any liquid (e. g. blood or fluid) before inserting the catheter.

Our study has a number of strengths. It is the largest and most comprehensive meta-analysis performed to-date assessing the effectiveness and safety of different topical hemostatic agents in the management of UGIB. It included 59 studies from different regions of the world published over the past decade, and assessed different agents, hence increasing generalizability. In addition, we assessed the effectiveness of these agents in UGIB of different etiologies (e. g. peptic ulcer disease, malignancy, variceal bleed) and in different settings (primary vs. rescue therapy), further increasing external validity. The main limitation of our analysis is the overall low to very low certainty of the evidence but such limitation exists also with many endoscopic therapeutic modalities yet have not deterred consideration for use in authoritative guidelines with adequate characterization of the data ¹⁴ . In addition to limitations in study design, the studies also display a lack of follow-up information in some reports which may have introduced reporting bias. Nevertheless, this systematic review addresses many of the limitations of previously published works that included small number of studies, limited inclusion criteria (limiting generalizability), and no formal assessment of the quality of the evidence using validated instruments such as the GRADE assessment tool ^{10 11 12} .

Conclusions

In conclusion, topical hemostatic agents are effective and safe in managing UGIB of different etiologies even when used as primary therapy and monotherapy. Patients treated with these agents should be monitored closely in the first few days due to a higher risk of rebleeding. More complete and novel subgroup analyses suggest these agents yield the highest immediate hemostatic rates, and are particularly effective in malignant bleeding, even when compared to other modalities. Future well designed studies should further compare hemostatic topical agents to conventional endoscopic modalities, and among each other to better inform recommendations as to their use, especially when chosen as sole first-line agent.