Comparing intravenous and oral proton pump inhibitor therapy for bleeding peptic ulcers following endoscopic management: a systematic review and meta‐analysis

Abstract

Background and Aims

The efficacy of proton pump inhibitors (PPIs) has been demonstrated for bleeding peptic ulcers but the route of administration remains controversial. Several studies have demonstrated that high‐dose oral PPIs are as effective as intravenous PPIs in reducing recurrent bleeding. However, current guidelines recommend intravenous PPIs after endoscopic treatment. Previous data based on numbers that were too small to enable a firm conclusion to be reached suggested that oral and intravenous PPIs had equivalent efficacy. We undertook a meta‐analysis to compare oral and intravenous PPIs in patients with bleeding peptic ulcers after endoscopic management.

Methods

A literature search was undertaken using MEDLINE, EMBASE and the Cochrane Library, between 1990 and February 2016, to identify all randomized controlled trials (RCTs) that assessed the efficacy of PPIs administered by different routes. Nine RCTs, involving 1036 patients, were analysed. Outcomes were: recurrent bleeding, blood transfusion requirement, duration of hospital stay, a need for repeat endoscopy, surgery and 30‐day mortality.

Results

There were no differences in the rebleeding rates [odds ratio (OR) 0.93, 95% confidence interval (CI) 0.60, 1.46; P = 0.77], need for surgery (OR 0.77, 95% CI 0.25, 2.40; P = 0.65), need for repeat endoscopy (OR 0.69, 95% CI 0.39, 1.21; P = 0.19), need for blood transfusion [(MD) –0.03, 95% CI –0.26, 0.19; P = 0.76], duration of hospital stay (MD –0.61, 95% CI –1.45, 0.23; P = 0.16) or 30‐day mortality (OR 0.89, 95% CI 0.27, 2.43; P = 0.84) according to the route of administration.

Conclusions

Oral PPIs represent better value for money, with clinical efficacy equivalent to intravenous PPIs.

Tables of Links

TARGETS
Transporters
H+/K+‐ATPASE

LIGANDS
Esomeprazole	Pantoprazole
Omeprazole	Rabeprazole

These Tables list key protein targets and ligands in this article that are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 1, and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 2.

Introduction

Acute upper gastrointestinal bleeding represents a challenging problem with significant morbidity and mortality, despite advances in diagnosis and treatment 3. Haemostasis attained at endoscopy reduces recurrent bleeding, the need for surgery and mortality in patients with bleeding peptic ulcers 4, 5, 6.

Experimental data suggest that gastric acidity plays an important role in causing clot lysis in bleeding ulcers, thus impairing haemostasis 7. Furthermore, a pH of <6 inhibits platelet aggregation 8 and of <5 inhibits plasma coagulation 9, 10. It has been hypothesized that inhibiting acid production consolidates clot formation and reduces the risk of rebleeding 8.

Irreversible inhibition of the H⁺/K⁺‐ATPase proton pump, which varies between proton pump inhibitors (PPIs) 11, allows more effective, direct suppression of acid secretion. While the role of adjuvant intravenous PPIs after endoscopic therapy has been confirmed in managing bleeding peptic ulcers 12, 13, 14, the best route of administration remains uncertain 15, 16, 17.

A randomized controlled trial (RCT) comparing a frequent oral PPI with an intravenous (IV) PPI showed that the former achieved 24‐h pH control similar to that of an IV bolus infusion, although the IV PPI reached a mean pH of 6 approximately 1 h sooner than oral PPI 18.

Current guidelines recommend treatment with an IV bolus followed by continuous infusion of PPI in patients with high‐risk stigmata after successful endoscopic therapy 19, 20, while other guidelines state that a specific recommendation on the route of administration of PPIs cannot be made 21. However, as shown by 24‐h intragastric pH monitoring, oral PPIs may be able to replace IV PPIs in patients with bleeding peptic ulcers 18. Two studies from Asia showed that oral PPIs as an adjunct to endoscopic treatment reduced the risk of recurrent bleeding by approximately 14–50% 22, 23. Moreover, there was no difference between the same dose of various PPIs given either intravenously or orally with regard to increasing intragastric pH above 6 for 72 h after successful endoscopic haemostasis 24.

Administration of IV PPI is more expensive and more complex to manage compared with the oral route. Several recent RCTs compared the efficacy of oral and IV PPIs in peptic ulcer bleeding but most of these were based on a small sample size 25, 26, 27, 28, 29.

The only previous meta‐analysis undertaken specifically to compare oral and IV PPIs showed equivalence but, again, was based on RCTS with small numbers 30. A Cochrane meta‐analysis, including five RCTs, comparing different PPI regimens failed to show superiority of any particular route of administration 15. Since then, one further RCT has been published 31, showing no statistical difference between oral and intravenous PPIs for all clinical outcomes assessed.

We present an up‐to‐date systematic review and meta‐analysis comparing clinical outcomes with oral or IV PPIs in patients with bleeding peptic ulcers after endoscopic treatment, which included a trial 32 that was included in the meta‐analysis undertaken by Neumann et al. 15 but excluded from the meta‐analysis undertaken by Tsoi et al. 30. A further study 24 included in the National Institute for Health and Care Excellence guidelines 21 comparing the effects of oral with IV PPIs on intragastric pH after endoscopic treatment was also included.

Methods

The recommendations of the Preferred Reporting Items for Systematic Review and Meta‐analysis (PRISMA) were used 33. Methods of analysis and inclusion criteria were specified in advance and documented in a protocol according to Cochrane guidelines 34. A PRISMA checklist is provided in the Supporting Information.

Eligibility

Types of study

RCTs comparing oral vs. IV PPIs. No language or publication status restrictions were imposed.

Types of participant

Patients admitted with symptoms of bleeding peptic ulcers undergoing endoscopic therapy to control bleeding were considered.

Types of intervention

Trials comparing clinical outcomes after oral or IV PPIs following endoscopic treatment.

Outcomes

The primary outcome measure was recurrent bleeding. Secondary outcomes included: need for surgery, 30‐day mortality, need for a second endoscopic treatment, need for a blood transfusion and length of hospital stay.

Exclusions

Patients with bleeding from upper gastrointestinal neoplasms or patients already on PPI treatment.

Identification and selection of studies

A medical literature search was performed by two authors (AT and MS) using MEDLINE, EMBASE, Cochrane Library and the ISI Web of Knowledge from January 1990 to February 2016. The full‐text of each potentially eligible study was assessed by two authors (AT and RM) searching for references to relevant literature reviews to complement the computer searches. Keywords for the search were ‘melaena’, ‘haematemesis’, ‘peptic ulcer haemorrhage’, ‘proton pump inhibitors’, ‘PPI’, ‘oral PPI’, ‘IV PPI’, ‘randomized controlled trials’, ‘meta‐analysis’, ‘endoscopic treatment’ and ‘ulcer bleeding’. The search strategy is shown in Table 1. Each article was read and analysed by two members of the research team (AT and RM).

Table 1.

Search strategy Pubmed

1. randomized controlled trial.pt.

2. controlled clinical trial.pt.

3. randomized.ab.

4. meta‐analysis.pt.

5. placebo.ab.

6. drug therapy.fs.

7. randomly.ab.

8. trial.ab.

9. groups.ab.

10. or/1–8

11. (animals not (humans and animals)).sh.

12. 10 not 11

13. exp stomach/

14. stomach.mp.

15. gastr$.mp.

16. exp duodenum/

17. duoden$.mp.

18. peptic$.mp.

19. exp peptic ulcer/

20. (peptic adj5 ulcer$).mp.

21. (stomach adj5 ulcer$).mp.

22. (duoden$ adj5 ulcer$).mp.

23. (gastroduoden$ adj5 ulcer$).mp.

24. or/12–25

25. peptic ulcer haemorrhage.mp.

26. peptic ulcer haemorrhage.mp. [mp = title, original title, abstract, name of substance word, subject heading word]

27. gastrointestinal haemorrhage.mp.

28. gastrointestinal haemorrhage.mp. [mp = title, original title, abstract, name of substance word, subject heading word]

29. (bleed$ adj5 ulcer$).mp.

30. (rebleed$ adj5 ulcer$).mp.

31. (re‐bleed$ adj5 ulcer$).mp.

32. (recurrent adj5 bleed$ adj5 ulcer$).mp.

33. (acute adj5 bleed$ adj5 ulcer$).mp.

34. (gastrointestinal adj5 bleed$).mp.

35. (gastrointestinal adj5 rebleed$).mp.

36. (gastrointestinal adj5 re‐bleed$).mp.

37. (gastrointestinal adj5 hemorrhag$).mp.

38. (gastrointestinal adj5 haemorrhag$).mp.

39. (ulcer adj5 hemorrhag$).mp.

40. (ulcer adj5 haemorrhag$).mp.

41. (haemorrhagic adj3 gastritis).mp. [mp = title, original title, abstract, name of substance word, subject heading word]

42. (hemorrhagic adj3 gastritis).mp.

43. (haemorrhagic adj3 duodenitis).mp.

44. (hemorrhagic adj3 duodenitis).mp. [mp = title, original title, abstract, name of substance word, subject heading word]

45. exp melena/

46. melena.mp.

47. melaena.mp.

48. exp hematemesis/

49. haematemesis.mp.

50. hematemesis.mp. [mp = title, original title, abstract, name of substance word, subject heading word]

51. or/27–53

52. acid suppression treatment.mp.

53. exp omeprazole/

54. omeprazole.mp.

55. lansoprazole.mp.

56. pantoprazole.mp.

57. rabeprazole.mp.

58. esomeprazole.mp.

59. (proton adj5 pump adj5 inhibitor$).mp.

60. ppi$.mp. [mp = title, original title, abstract, name of substance word, subject heading word]

61. or/55–63

62 oral PPI

63 intravenous PPI

64 or/65–66

65 endoscopic haemostasis

66.10 and 24 and 52 and 62 and 65 and 66

Data extraction

All articles selected for relevance were evaluated independently two investigators (AT and MS). Extracted data were transcribed to specially prepared tables based on the Cochrane Review Group data extraction template. Two reviewers (AT and RM) included studies and extracted data, and disagreement was resolved by consensus.

The following data were extracted:

• Descriptive data: First author, year and type of publication, country of origin, study setting, number of patients, age and gender of patients, type and dose of PPIs used, Forrest classification, ulcer site, ulcer size, presence of comorbidity, Rockall score and type of endoscopic treatment used.

• Qualitative data: Random sequence generation, allocation concealment, blinding participant and personnel, blinding outcome assessment, incomplete outcome data, selective reporting and lost at follow‐up.

• Outcome data: Recurrent bleeding, and recurrent bleeding at 15 days and 30 days, need for a second endoscopic treatment, need for a blood transfusion, length of hospital stay, requirement for surgery and 30‐day mortality.

Study quality

The characteristics of studies were assessed independently by two investigators (AT and RM) and are reported in Table 2.

Table 2.

Study characteristics comparing ulcer patients receiving oral PPIs with those receiving IV PPIs

Trial	Country	N	Intervention		Age (mean)	Gender (Male %)	Forrest classification (No. of patients receiving oral vs IV PPIs)
			Oral PPI	Intravenous PPI			I	II	III
Focareta et al. 2004 32	Italy	87	Esomeprazole 40 mg orally twice daily for 72 h, then 20 mg omeprazole or 40 mg esomeprazole orally per day for 4 weeks.	80 mg bolus, followed by 40 mg three times daily for 72 h, then 20 mg omeprazole or 40 mg esomeprazole orally per day for 4 weeks.	NR	NR	Spurting 13	Nonbleeding vessel 24	‐
							Oozing 17	Adherent clot 33
Jang et al. 2006 29	Korea	38	40 mg pantoprazole twice daily for 5 days	80 mg bolus of pantoprazole and 8 mg hr−1 infusion for 3 days, then 40 mg oral pantoprazole once daily for 8 weeks	59.3	76.3%	Spurting vessel (2 vs. 0)	Nonbleeding vessel (13 vs. 15)	–
							Oozing vessel (4 vs. 4)
Yilmaz et al. 2006 44	Turkey	211	40 mg omeprazole twice daily for 3 days, then 40 mg once daily for 6 weeks	80 mg bolus of omeprazole and 8 mg hr−1 infusion for 3 days, then 40 mg oral omeprazole once daily for 6 weeks	52.7	68.7%	Excluded	Flat spot (26 vs. 20)	Clean based (62 vs. 82)
								Old adherent clot (11 vs. 10)
Javid et al. 2009 24	India	90	80 mg bolus of omeprazole, pantoprazole or rabeprazole followed by 40 mg twice daily for 72 h	80 mg bolus of omeprazole, pantoprazole or rabeprazole followed by infusion of 8 mg h−1 for 72 h	35.6	NR	Spurting: 9 vs. 8	Visible vessel 26 vs. 27	_
							Oozing 10 vs. 10
Bajaj et al. 2007 25	USA	25	80 mg pantoprazole twice daily for 3 days, then 40 mg twice daily for 30 days	80 mg bolus of pantoprazole and 8 mg h−1 infusion for 3 days, then 40 mg oral pantoprazole twice daily for 30 days	63.0	64.0%	Active bleeding (3 vs. 4) 28%	Visible vessel (3 vs. 4) 36%	Clean based (4 vs. 5) 36%
								Red spot (2 vs. 0)
Tsai et al. 2009 26	Taiwan	156	20 mg rabeprazole twice daily for 3 days, then 20 mg once daily for 2 months	40 mg infusion of omeprazole twice a day for 3 days, then orally either 20 mg rabeprazole or 40 mg esomeprazole once daily for 2 months	68.7	72.4%	Spurting vessel (0 vs. 3)	Nonbleeding vessel (18 vs. 24)	–
							Oozing vessel (33 vs. 28)	Adherent clot (26 vs. 23)
Mostaghni et al. 2011 27	Iran	85	40 mg omeprazole twice daily for 3 days, then 20 mg once daily for 30 days	80 mg bolus of pantoprazole and 8 mg h−1 infusion for 2–3 days, then 20 mg oral omeprazole once daily for 30 days	59.4	74.1%	Active bleeding (6 vs. 7)	Visible vessel (25 vs. 26)	‐‐‐
							Oozing vessel (8 vs. 5)	Adherent clot (5 vs. 3)
Yen et al. 2012 28	Taiwan	100	30 mg lansoprazole four times daily for 3 days, then 30 mg once daily for 2 months	40 mg infusion of esomeprazole 4 times daily for 3 days, then 40 mg oral esomeprazole once daily for 2 months	63.9	71.0%	Spurting vessel (4 vs. 6)	Nonbleeding vessel (28 vs. 29)	–
							Oozing vessel (18 vs. 15)
Sung et al. 2014 31	Hong Kong	118	IV placebo bolus plus infusion for 72 h and 40 mg esomeprazole every 12 h	80 mg IV bolus of esomeprazole and infusion at 8 mg h−1 for 72 h and oral placebo every 12 h	64	74%	Spurting vessel (7 vs. 2)	Visible vessel (64 vs. 62)
							Oozing vessel (48 vs. 49)	Adherent clot (7 vs. 5)

IV, intravenous; NR, Not reported; PPI, proton pump inhibitor

Potential sources of bias were evaluated by the Cochrane Risk tool 35. The risk of bias tool evaluated the random sequence generation, allocation concealment, blinding of patients and investigators, blinding of outcome, incomplete outcome data, selective outcome reporting and other potential bias according to the Cochrane guidelines 35.

Study outcomes

The primary outcome was recurrent bleeding, defined as the failure to control bleeding after the first endoscopic therapy. Secondary outcomes included the mean volume of blood transfused, duration of hospital stay in days, need for a second endoscopic treatment or surgery and mortality within 30 days following endoscopic therapy.

Statistical analysis

Meta‐analysis

Data were synthesized using Review Manager software (version 5.1 for Windows, the Cochrane Collaboration, Oxford, UK). The odds ratio (OR) was used to analyse dichotomous data obtained from individual studies and the mean difference (MD) was used to calculate continuous variables. Recurrent bleeding, the need for surgery, the need for a second endoscopic treatment, and 30‐day mortality are summarized as ORs with 95% confidence intervals. Hospital stay and volume of blood transfused were calculated as mean differences according to the Cochrane guidelines 36. The random‐effects model according to the method of Der Simonian and Laird 37 was used as a conservative approach, considering variations between studies.

Analysis of heterogeneity

In order to assess heterogeneity visually, the Forest plot was inspected. Finally, to confirm and quantify the presence of inconsistency, the I ² statistic was used. Heterogeneity was significant at I ² > 50% 38. I ² has been shown to be preferable to the test of heterogeneity when assessing inconsistency between studies 39.

Publication bias

Publication bias was not assessed because of the low number of studies involved (<10 studies), according to the BMJ recommendation stating that the funnel plot and Egger test should only be included if the meta‐analysis includes 10 or more studies as, when there are too few studies, the power of the tests is too low to distinguish chance from real asymmetry 40. Despite a low power for detecting publication bias, with only nine eligible studies, a funnel plot (not shown) did hint at a possible lack of publication of small studies favouring IV over oral PPIs.

Sensitivity analysis

Sensitivity analysis was performed by removing each trial analysis in turn (to assess the influence of each individual study on the global analysis) and performing the meta‐analysis using only studies of higher quality (the leave‐one‐out‐method). Furthermore, we analysed trials using the fixed‐effects model or random‐effects model and compared trials of lower quality vs. higher quality, in addition to identifying an outlier study using a sequential algorithm, as previously described and recommended 41.

Subgroup analysis

Subgroup analysis, comparing high‐dose vs. nonhigh‐dose IV PPIs, high‐dose oral vs. high‐dose IV PPIs, patients with high‐risk stigmata and recurrent bleeding at 15 days and 30 days, was carried out using Revman software, Review Manager (RevMan) [Computer program]. Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014.

We selected the first 72 h as the period for calculating the cumulative dose of PPI because this period includes the majority of rebleeding episodes 42. Furthermore, studies were categorized according to the cumulative dose of PPI received within the first 72 h of treatment, as follows: low‐dose studies: ≤120 mg 72 h^–1 or high‐dose studies: ≥600 mg 72 h^–1. We selected a cut‐off above 240 mg 72 h^–1 as high‐dose oral PPI.

The risk of bias in individual studies, across studies and within studies according to the Cochrane recommendation 35 are provided. The meta‐regression was not performed because fewer than 10 trials were included in the meta‐analysis 36, 43.

Results

Study selection

The initial search identified 3325 abstracts after duplicates were removed. Of these, 3225 studies were excluded after evaluating title and abstract, 100 full‐text articles were assessed for eligibility and 91 were excluded as shown in the PRISMA flow chart (Figure 1). Finally, nine RCTs comprising eight full‐text articles 24, 25, 26, 27, 28, 29, 31, 44 and one in abstract form 32, published between 2004 and 2014, were included in the meta‐analysis. These included patients from the United States, Turkey, Iran, Taiwan, Hong Kong, Italy and Korea.

Figure 1.

Prisma flow chart

GORD, gastro‐oesophageal reflux disease; PPI, proton pump inhibitor

Characteristics of included trials

In the nine RCTs 24, 25, 26, 27, 28, 29, 31, 32, 44, 1036 subjects were allocated to receive oral PPIs (n = 518) or IV PPIs (n = 518). Four trials 25, 26, 29, 32 were also classified as having a high risk of performance and detection bias according to the Cochrane definition 35, 36. The number of patients in the individual trials ranged from 25 to 211, with a mean age of 60 years, with 71.1% male (Table 2). We did not evaluate or report significance tests of differences in baseline variables, in accordance with the CONSORT Statement 45 and other authors 46, 47 who have considered this method to be misleading.

The trial patients showed a wide range of bleeding stigmata. Patients with high‐risk stigmata (i.e. spurting, oozing or nonbleeding vessel) represented 65% in the oral group and 80% in the IV group. PPIs were prescribed intravenously for the first 3 days and then given orally for 1–2 months. Six trials used high‐dose IV PPIs (i.e. 600 mg 72 h^–1) 24, 25, 27, 29, 31, 44, whereas used low‐dose IV PPIs (i.e. 120 mg 72 h^–1) 26, 28, 32. Three trials used high‐dose oral PPIs 24, 25, 28, 31 (Table 2). Different PPIs were used in these studies, including pantoprazole 25, 27, 29, 48, omeprazole 26, 27, 44, rabeprazole 26, lansoprazole 28 and esomeprazole 28, 31, 32 (Table 2).

In all trials, endoscopic procedures were performed within 24 h of admission. Endoscopic treatment was not carried out in one trial 44; in one trial, only adrenaline was used 26; in another trial, adrenaline was used as the only approach in 66% of cases 27 (Table 3).

Table 3.

Baseline characteristics of included trials

Trial	N	NSAID Oral vs. IV	Systolic BP or shock admission Oral vs. IV	Ulcer site Oral vs. IV	Ulcer size >2 cm Oral vs. IV	HP positivity	Comorbidity Oral vs. IV	Rockall score (RS) Oral vs. IV	Endoscopic treatment
Focareta et al. 2004 32	87	NR	NR	NR	NR		NR	NR	Adrenaline or haemoclip or combination
Jang et al. 2006 29	38	7/19 (36.8%) vs. 10/19 (52.6%)	Shock 5/19 (26.3%) vs. 7/19 (36.8%)	Gastric 14/19 (73.7%) vs. 15/19 (78.9%) Duodenal 5/19 (26.3%) vs. 4/19 (21.1%)	2/19 (10.5%) vs. 6/19 (31.6%)	14/19 (73.7%) vs. 11/19 (57.9%)	7/19 (36.8%) vs. 11/19 (57.9%)	NR	Adrenaline + APC
Javid et al. 2009 24	90	4/45 vs. 4/45	Shock 2/45 vs. 1/45	NR	1.4 cm ± 0.7 vs. 1.2 cm ± 0.8	29/45 vs. 27/45	0/45 vs. 0/45	NR	Adrenaline + heater probe
Yilmaz et al. 2006 44	211	70.7% vs. 60.7%	Shock 5/99 vs. 6/99	Gastric 76/99 vs. 94/112, duodenal 33/99 vs. 18/112	1.06 ± 0.6 vs. 1.05 ± 0.4	56/99 (56.6%) vs. 63/112 (56.3%)	35.4% vs. 36.6%	RS < 3: 52.5% vs. 56.3% RS >8 19.1% vs. 18.7%	No treatment. Study design excluded patients with active bleeding
Bajaj et al. 2007 25	25	100% vs. 95%	106.8 ± 23 vs. 125.1 ± 32.6	Gastric 5/12 vs. 5/13, duodenal 5/12 vs. 4/13	NR	NR	Comorbidity score 2.2 ± 1.9 vs. 3.3 ± 2.6	4.5 ± 2.1 vs 5.3 ± 2.5	Adrenaline + clip or BICAP
Tsai et al. 2009 26	156	NR	Shock 16/78 (20%) vs. 21/78 (26.9%)	Gastric 50% vs. 53.8%, duodenal 47% vs. 41%	1.12 cm vs. 1.06 cm	51/78 (65.4%) vs. 48/78 (61.5%)	51% vs. 50%	5.3 vs. 5.4	Adrenaline 1:10:000 10 ml
Mostaghni, et al. 2011 27	85	19/44 (43%) vs. 17/41 (41%)	NR	Gastric 24/44 (54%) vs. 18/41 (44%), duodenal 17/44 (39%) vs. 20/41 (49%)	NR	NR	NR	NR	Adrenaline 1:10:000 5–30 ml/66% of cases ± APC
									APC 29/44 (32%)(oral); 12/41(30%) (IV
Yen et al. 2012 28	100	NR	Shock 18/50 (36%) vs. 21/50 (42%)	Gastric 18/50 (32) vs. 22/50 (44%), duodenal 32/50 (68%) vs. 28/50 (56%)	1.6 cm vs. 1.4 cm	25/50 (68.6%) vs. 26/50 (68.4%)	30/50 (60%) vs. 33/50 (66%)	5.3 (0.3) vs. 5.3 (0.3)	Heater probe and clip
Sung et al.2014 31	118	54/126 (42.9%) vs. 47/118 (39.8%)	BP < 100 31/126 (24.6%) vs. 26/118 (22%)	Gastric 40.7% vs. 42.1%	12.7% vs. 13.6%	65/126 (51.6%) vs. 62/118 (52.5%	NR	NR	Haemoclip or heater probe ± adrenaline

APC, argon plasma coagulator; BICAP, bipolar electrocoagulation; BP, blood pressure; HP, Helicobacter Pylori; IV, intravenous; NR, not reported; NSAIDs, Nonsteroid anti‐inflammatory drugs; RS, Rockall score

Study quality

The risk of bias, according to the Cochrane risk of bias tool, was not significant (Figure 2A,B).

Figure 2.

(A) Risk of bias graph: review of authors’ judgements about each risk of bias item presented as percentages across all included studies. (B) risk of bias summary: review of authors’ judgements about each risk of bias item for each included study. +, low risk ofbias; ?, unclear risk of bias; ‐, High risk of bias

Primary outcome

All trials reported recurrent bleeding (Table 1). Heterogeneity between trials was not statistically significant (I ² = 0%; P = 0.67) and the pooled results showed no significant differences concerning recurrent bleeding between oral and IV PPIs (OR 0.93, 95% CI 0.61, 1.46; P = 0.77) (Figure 3 and Table 4).

Figure 3.

Forest plot of recurrent bleeding. CI, confidence interval; IV, intravenous; M‐H, Mantel‐Haenszel; PPI, proton pump inhibitor

Table 4.

Primary and secondary outcomes

Author	Publication type	Route	Recurrent bleeding Total	Further endoscopic therapy	Blood transfusion Mean (SD)/N	Hospital stay (Days) Mean (SD)/N	Surgery	Mortality
Focareta et al. 2004 32	Abstract	Oral	4/45	3/45	–	–	0/45	–
		IV	3/42	4/42	–	–	0/42	–
Jang et al. 2006 29	Full	Oral	1/19	1/19	1.2 (1.1)/19	–	0/19	0/19
	text	IV	2/19	1/19	1.2 (1.2)/19	–	0/19	1/19
Yilmaz et al. 2006 44	Full	Oral	5/99	–	2.1 (1.7)/99	4.5 (2.6)/99	2/99	2/99
	text	IV	7/112	–	1.9 (1.1)/112	4.6 (1.6)/112	3/112	3/112
Javid et al. 2009 24	Full	Oral	4/45	–	–	–	2/45	0/45
	text	IV	4/45	–	–	–	2/45	0/45
Bajaj et al. 2007 25	Full	Oral	0/12	0/12	3.6 (2.4)/12	5.2 (3.3)/12	NA	0/12
	text	IV	2/13	2/13	3.9 (3.7)/13	6.8 (4.8)/13		0/13
Tsai et al. 2009 26	Full	Oral	13/78	11/78	2.3 (4.9)/78	8.9 (5.4)/78	1/78	2/78
	text	IV	12/78	12/78	2.5 (6.8)/78	8.5 (5.0)/78	1/78	1/78
Mostaghni et al. 2011 27	Full	Oral	5/44	18/44	1.8 (1.3)/44	3.1 (1.5)/44	0/44	1/44
	text	Iv	4/41	24/41	2.9 (1.3)/41	3.6 (1.5)/41	0/41	1/41
Yen et al. 2012 28	Full	Oral	2/50	–	1.0(1.4)/50	1.8 (2.1)/50	0/50	0/50
	text	IV	2/50	–	1.6(2.1)/50	3.9 (1.4)/50	0/50	0/50
Sung et al. 2014 31	Full	Oral	8/126	–	1(3.2)/126	4.0 (3.2)/126	0/126	–
	text	IV	9/118	–	2(3)/118	4.0 (4)/118	1/118	–
No. of study	9		9	4	6	6	8	7
No. of participants			1036	350	615	821	1011	705
OR, 95% CI			OR 0.93, 95% CI 0.60, 1.46	OR 0.69, 95% CI 0.39, 1.21	MD –0.03 units, 95% CI –0.26, 0.19	MD −0.61 days, 95% CI –1.45, −0.23	OR 0.77, 95% CI 0.25, 2.40	OR 0.89, 95% CI 0.27, 2.93
P‐value			0.77	0.19	0.76	0.16	0.65	0.84
I2			0%	0%	0%	78%	0%	0%

CI, confidence interval; IV, intravenous; OR, odds ratio; MD, mean difference; NA, not available; SD, standard deviation

Secondary outcomes

Secondary outcomes are reported in Table 4, showing the need for a second endoscopic treatment (OR 0.69, 95% CI 0.39, 1.21; P = 0.19), a need for surgical intervention (OR 0.77, 95% CI 0.25, 2.40; P = 0.65) and 30‐day mortality (OR 0.89, 95% CI 0.27, 2.93; P = 0.84) (Figure 4 and Table 4). The mean number of transfusions (MD −0.03 units, 95% CI –0.26, 0.19 units; P = 0.76) was equivalent in both groups (Figure 5). There was no statistically significant difference among patients receiving oral or IV PPIs in terms of length of hospital stay (MD −0.61 days, 95% CI –1.45, 0.23 days; P = 0.16) (Figure 5).

Figure 4.

Forest plot of comparison between oral and IV PPIs for (A) need for surgery, (B) mortality and (C) EHT, need for a second endoscopic treatment. CI, confidence interval; IV, intravenous; M‐H, Mantel‐Haenszel; PPI, proton pump inhibitor

Figure 5.

Forest plot of comparison between oral and IV PPIs for (A) need for blood transfusion and (B) length of hospital stay. CI, confidence interval; IV, intravenous; M‐H, Mantel‐Haenszel; PPI, proton pump inhibitor

Exploring source of heterogeneity

Heterogeneity was found in the length of hospitalization (I ² = 78%; t² = 0.74, df = 5; P = 0.003) but a sensitivity analysis looking at higher‐quality trials did not show a difference.

We used both fixed‐ and random‐effects models for all outcomes, and these did not differ. On the contrary, for length of hospital stay we used the most appropriate random‐effects model, which considers variance between studies, because of the presence of heterogeneity (I ² = 78%), and this showed that the groups had an equal length of hospital stay (MD –0.61, 95% CI –1.45, –0.23; P = 0.16). We also took into consideration the clinical source of heterogeneity and statistical heterogeneity, as described previously 49.

To assess the reason for statistical heterogeneity, we performed a sequential algorithm, as described by Patsopoulos et al. 41 allowing the identification of an outlier study 28. The omission of this study reduced the estimated heterogeneity to 0%. This study had an unclear risk of bias in blinding because it did not address the outcome that could have justified the heterogeneity found. Finally, we analysed and compared primary and secondary outcomes (except for those data that were not available in the trials), subdividing trials into high quality (low/unclear risk of bias) and low quality (high risk of bias), as shown in Table 5.

Table 5.

Sensitivity analysis: high risk of bias vs. unclear/low risk of bias studies

Outcome or subgroup	Studies	Participants	Statistical method	Effect estimate
Rebleeding	9	1036	Odds ratio (M‐H, fixed, 95% CI)	0.92 [0.59, 1.43]
High risk of bias	5	391	Odds ratio (M‐H, fixed, 95% CI)	0.98 [0.53, 1.81]
Low/unclear risk of bias	4	645	Odds ratio (M‐H, fixed, 95% CI)	0.86 [0.46, 1.63]
Mortality	7	705	Odds ratio (M‐H, fixed, 95% CI)	0.88 [0.28, 2.77]
High risk of bias	4	304	Odds ratio (M‐H, fixed, 95% CI)	0.98 [0.22, 4.39]
Low/unclear risk of bias	3	401	Odds ratio (M‐H, fixed, 95% CI)	0.75 [0.12, 4.58]
Surgery	9	1011	Odds ratio (M‐H, fixed, 95% CI)	0.76 [0.25, 2.30]
High risk of bias	5	366	Odds ratio (M‐H, fixed, 95% CI)	1.00 [0.06, 16.28]
Low /unclear risk of bias	4	645	Odds ratio (M‐H, fixed, 95% CI)	0.72 [0.21, 2.43]
Blood transfusion	6	615	Mean difference (IV, random, 95% CI)	–0.03 [−0.26, 0.19]
High risk of bias	4	304	Mean difference (IV, random, 95% CI)	–0.07 [−0.37, 0.23]
Low/unclear risk of bias	2	311	Mean difference (IV, random, 95% CI)	–0.15 [−0.92, 0.63]
Hospital stay	6	821	Mean difference (IV, random, 95% CI)	–0.61 [−1.45, 0.23]
High risk of bias	3	266	Mean difference (IV, random, 95% CI)	–0.42 [−1.00, 0.16]
Low/unclear risk of bias	3	555	Mean difference (IV, random, 95% CI)	–0.76 [−2.20, 0.68]

CI, confidence interval; IV, intravenous; M‐H, Mantel‐Haenszel

Subgroup analyses

In the subgroup of trials comparing high‐dose (i.e. ≥600 mg 72 h^–1) with low‐dose (i.e. ≤120 mg 72 h^–1) IV PPIs 25, 26, 27, 28, 29, 31, 32, 44, no statistically significant difference was shown in terms of recurrent bleeding (OR 0.96, 95% CI 0.60, 1.63; P = 0.73), the need for surgery (OR 0.68, 95% CI 0.17, 2.70; P = 0.53), 30‐day mortality (OR 0.89, 95% CI 0.27, 2.93; P = 0.58) or length of hospital stay (mean difference −0.61 days, 95% CI –1.45, 0.23; P = 0.19) (Table 6). The volume of blood transfused was not statistically significant (MD 0.20 units, 95% CI –0.45, 0.85; P = 0.28).

Table 6.

Subgroup analysis

Author	Recurrent bleeding 15 days Oral vs IV	Recurrent bleeding 30 days Oral vs IV	High‐dose IV vs. nonhigh‐dose IV PPI	High‐dose oral vs high‐dose IV PPI	High‐risk stigmata oral vs. IV
Focareta et al. 2004 32	NR	NR	Rebleeding 4/45 vs 3/42 Surgery 0/45 vs 0/42 Mortality : NR LOS: NR Blood transfusion : NR	NR	NR
Jang et al. 2006 29	NR	1/19 vs. 2/19	Rebleeding 1/19 vs 2/19 Surgery 0/19 vs 0/19 Mortality 0/19 vs 1/19 LOS : NR Blood transfusion 1.2(0.55) /19 vs 1.2(0.6)/19	NR	Rebleeding 1/19 vs. 2/19 Surgery 0/19 vs. 1/19 Mortality 0/19 vs. 0/19
Yilmaz et al. 2006 44	NR	5/99 vs. 7/112	Rebleeding 5/99 vs. 7/112 Surgery 2/99 vs. 3/112 Mortality 2/99 vs. 3/112 LOS 4.5 (±2.6) vs. 4.6 (±1.6) Blood transfusion 2.1 (±1.7) vs. 1.9 (±1.1)	NR	NR
Bajaj et al. 2007 25	NR	0/12 vs 2/13	Rebleeding 0/12 vs. 2/13 Surgery 0/12 vs. 1/13 Mortality 0/12 vs. 0/13 LOS 5.2 (3.3) vs. 6.8 (4.8) Blood transfusion 3.6 (±3.4) vs. 3.9 (±3.7)	Rebleeding 0/12 vs. 2/13 Surgery 0/12 vs. 0/13 Mortality 0/12 vs. 0/13 LOS 5.2 (3.3) vs. 6.8 (4.8) Blood transfusion 3.6 (2.4) vs. 3.9(3.7)	NR
Tsai et al. 2009 26	13/78 vs. 12/78		Rebleeding 3/78 vs 12/78 Surgery 1/78 vs 1/78 Motality 2/78 vs 1/78 LOS 8.9(5.4)/78 vs 8.5 (4.9)/78 Blood transfusion 2.3(4.9)/78 vs 2.5(6.8)/78	NR	Rebleeding 13/78 vs. 12/78 Surgery 1/78 vs. 1/78 Mortality 2/78 vs. 1/78
Javid et al. 2009 24	NR	NR	NR	Rebleeding 4/45 vs. 4/45 Surgery 2/45 vs. 2/45 Mortality 0/45 vs 0/45 LOS NR Blood transfusion NR	NR
Mostaghni et al. 2011 27	NR	5/44 vs. 4/41	Rebleeding 5/44 vs 4/44 Surgery 0/44 vs 0/44 Mortality 1/44 vs 1/41 LOS 3.1 (1.5)/44 vs 3.6(1.5)/ 44 Blood transfusion 1.8(1.3)/44 vs 2(1.3)/44	NR	NR
Yen et al. 2012 28	2/50 vs. 2/50	NR	Rebleeding 2/50 vs 2/50 Surgery 0/50 vs 0/50 Mortality 0/50 vs 0/50 LOS 1.8(2.1)/50 vs 3.9(1.4)/50 Blood transfusion 1(1.4)/50 vs 1.6(2.1)/50	Rebleeding 2/50 vs. 2/50 Surgery 0/50 vs. 0/50 Mortality 0/50 vs. 0/50 LOS 1.8 (2.1) vs. 3.9 (1.4) Blood transfusion 1 (1.4) vs. 1.6 (2.1)	Rebleeding 2/50 vs. 2/50 Surgery 0/50 vs. 0/50 Mortality 0/50 vs. 0/50
Sung et al. 2014 31	11/126 (8.7%) vs. 9/118 (7.7%) Excluding F 1b	11/126 (8.7%) vs. 9/118 (7.7%) Excluding F 1b	Rebleeding 8/126 vs 9/118 Surgery 0/126 vs 0/118 Mortality 0/126 vs 0(/118 LOS 4(4.7)/126 vs 4(4.5)/118	Rebleeding 8/126 vs. 9/118 Surgery 0/126 vs. 1/118 Mortality 0/126 vs. 0/118 LOS 4 (3.2) vs. 4 (4) Blood transfusion 1 (3) vs. 2 (3)	Rebleeding 8/126 vs. 9/118 Surgery 0/126 vs. 1/118 Mortality 0/126 vs. 0/118
OR, 95% CI	OR 1.1, 95% CI 0.61, 2.03; P = 0.72	OR 0.97, 95% CI 0.52, 1.83; P = 0.93	Rebleeding OR 0.96, 95% CI 0.60, 1.55 Surgery OR 0.68, 95% CI 0.17, 2.70 Mortality OR 0.89, 95% CI 0.27, 2.93 LOS MD –0.61, 95% CI –1.45, 0.23 Blood transfusion MD 0.20, 95% CI –0.45, 0.85	Rebleeding OR 0.82, 95% CI 0.39, 1.7 P = 0.59 Surgery OR 0.72, 95% CI 0.13, 3.9; P = 0.70 Mortality: not estimable LOS MD −0.12 days, 95% CI –1.0, 0.76; P = 0.79 Blood transfusion MD −0.76 units, 95% CI –1.27, –0.26; P = 0.003	Rebleeding OR 0.93, 95% CI 0.51, 1.70 Surgery OR 0.60, 95% CI 0.07, 4.96 Mortality OR 1.05, 95% CI 0.15, 7.32

CI, confidence interval; IV, intravenous; LOS, length of hospital stay; NR, not reported; OR, odds ratio; PPI, proton pump inhibitor; F 1b, Forest classification oozing bleeding

In the subgroup of trials considering low‐dose IV PPIs (120 mg 72 h^–1) 26, 28, no statistically significant difference was shown in terms of recurrent bleeding (OR 1.08, 95% CI 0.49, 2.38; P = 0.84), the need for surgery (OR 1.0, 95% CI 0.06, 16.28; P = 1.0), 30‐day mortality (OR 2.0, 95% CI 0.18, 22.8; P = 0.57), need for blood transfusion (MD –0.55, 95% CI –1.21, 0.10; P = 0.10) and length of hospital stay (MD –0.96, 95% CI –3.40, 1.48; P = 0.44). In the subgroup analysis, we also compared studies considering high‐dose oral PPIs (>240 mg 72 h^–1) vs. high‐dose IV PPIs (>600 mg 72 h^–1); no statistically significant difference was found for any of the outcomes analysed, except for the need for blood transfusion, which favoured oral PPIs (MD –0.76 units, 95% CI –1.27, 0.26; P = 0.003) (Figure 6). Length of hospital stay was significantly shorter with oral treatment (MD −1.34 days, 95% CI –1.88, 1.79; P = 0.00001; I ² = 84%) but when the outlying trial 28 was excluded, no statistically significant difference was observed (MD −0.12 days, 95% CI –1.0, 0.76; P = 0.79; I ² = 0%).

Figure 6.

Blood transfusion requirement with high‐dose oral vs. high‐dose intravenous PPI. CI, confidence interval; IV, intravenous; PPI, proton pump inhibitor; SD, standard deviation

Four trials 23, 25, 26, 28 reported outcomes in 538 patients with high‐risk stigmata (i.e. spurting, oozing or nonbleeding vessel); recurrent bleeding was similar between the oral and IV PPI groups (OR 0.93, 95% CI 0.51, 1.70). Three trials 26, 28, 31 reported recurrent bleeding after 15 days in 500 patients, with no statistically significance difference observed between the two groups (OR 1.11, 95% CI 0.61, 2.03; P = 0.72) and no heterogeneity (I ² = 0%).

Five trials 25, 27, 29, 31, 44 reported recurrent bleeding after 30 days, showing no statistically significance difference between oral and IV PPIs (OR 0.93, 95% CI 0.51, 1.71; P = 0.82). No significant difference was observed in the volume of blood transfused (mean difference: −0.31 units, 95% CI –0.79, 0.18), length of hospital stay (mean difference −0.96 days, 95% CI –3.40, 1.48), requirement for surgery (OR 0.60, 95% CI 0.07, 4.96) or 30‐day mortality (OR 1.05, 95% CI 0.15, 7.32).

Discussion

The present meta‐analysis included nine RCTs comparing treatment efficacy between oral and IV PPIs in patients with bleeding peptic ulcers following endoscopic treatment. Oral PPIs were found to be equivalent to IV PPIs in regard to the risk of recurrent bleeding, need for blood transfusion, need for repeated endoscopy, length of hospital stay, need for surgery and overall mortality.

A RCT demonstrated that high‐dose IV PPI (80 mg bolus followed by a continuous infusion of 8 mg h⁻¹ for 72 h) was more effective than a standard IV dose of PPI (40 mg bolus twice a day for 72 h) in reducing rebleeding, blood transfusion requirement and length of hospital stay 50. By contrast, subgroup analysis showed that high‐dose IV PPIs were equivalent to low‐dose IV PPIs for all outcomes considered. A subgroup analysis comparing a high‐dose oral PPI to a high‐dose IV PPI demonstrated no statistically significance difference for any of the outcomes considered, except for the need for a blood transfusion, which favoured the high‐dose oral PPI.

In the present meta‐analysis, oral PPIs showed similar efficacy compared with IV PPIs. One possible explanation might be that both oral and IV PPIs display equal efficacy in maintaining intragastric pH >6 over 24 h 18.

Previous studies showed that the inhibition of gastric acid secretion with 40 mg oral pantoprazole compared with a similar IV dose yielded a similar result, with comparable rebleeding rates in bleeding peptic ulcers 51, 52. Two previous studies from Asia showed that oral PPIs after endoscopic treatment reduced the risk of recurrent bleeding to between 17% and 50% 22, 23, although these results may not be applicable generally to Caucasian populations because of underlying differences in pharmacodynamic profiles – factors that may favour the acid‐suppressive effect of a given dose of a proton‐pump inhibitor in Asian patients 17.

Some endoscopic stigmata (e.g. spurting, oozing and nonbleeding vessel) are at high risk of recurrent bleeding 53, 54. The present meta‐analysis included some low‐risk patients with Forrest classification IIc or III. These patients may have a lower risk of recurrent bleeding, which could explain the comparable efficacy of oral and IV PPIs.

Our analysis showed that oral and IV PPIs are associated with an equal need for blood transfusions, in agreement with the results of the meta‐ analysis by Tsoi et al. 30; however, that meta‐analysis showed that oral PPIs are associated with a reduced hospital stay compared with IV PPIs 30. There may have been methodological and clinical reasons for the difference between this finding and that of the present report.

The RCTs included in the meta‐analysis by Tsoi et al. 30 that assessed such outcomes were heterogeneous (I ² = 82%), and the authors used a fixed‐effects model instead of a random‐effects one, despite the apparent heterogeneity between studies (I ² = 82%), and this could have influenced the result. Our meta‐analysis evaluated the same outcome by including the same RCTs, with the exception of our addition of the more recent trial by Sung et al. 31, but used the random method described by Der Simonian and Laird 37 as a conservative approach and based on the heterogeneity of the studies.

Another important clinical consideration is that, theoretically, even though the use of oral PPIs would allow an early discharge, in reality, patients with a high risk of rebleeding would not be discharged early, to avoid medico‐legal problems.

Therefore, oral PPIs are ideally indicated for patients discharged from accident and emergency departments with a low risk of endoscopic stigmata, and also as a replacement for IV PPIs in hospitalized patients, as a cost‐saving exercise.

Theoretically, the analysis of each subgroup of bleeding stigmata could help to improve our understanding of the differences in treatment efficacy between oral and IV PPIs, but the relevant data were not extractable from the original publications due to the limited sample size.

In the present meta‐analysis, we could only select for analysis subgroup trials on high‐risk stigmata patients, in whom recurrent bleeding in those using oral and IV PPIs was similar.

A Cochrane review by Neumann et al. 15 included five RCTs, four of which were classified as having a high risk of bias and the other 44 as having an unclear risk of bias. The 72‐h cumulative dose of PPI was higher in the IV than in the oral regimen, and three of the studies used high‐dose PPI (≥600 mg 72 h^–1) and two used low‐dose PPI (200 mg 72 h^–1). Neumann et al. concluded that there was no difference between oral and IV PPIs in the outcomes analysed. Furthermore, they performed a meta‐regression showing that the route of PPI administration was not significantly associated with treatment effects 15. However, it should be pointed out that the use of meta‐regression should generally not be considered when there are fewer than 10 studies in a meta‐analysis 36, 43.

There were some clinical and methodological limitations in the meta‐analysis by Tsoi et al. 30, as summarized in Table 7. First, the study was likely to have been underpowered owing to the low number of patients included (only 615) for a rare rebleeding rate as the primary outcome because of the underpowered RCT included as pointed out by the same authors.

Table 7.

Limitations of the meta‐analysis by Tsoi et al. 30 according to the PRISMA statement for reporting systematic review and meta‐analysis by Liberati et al. 33 and Cochrane guidelines

Limits	Description
Method of identifying all relevant information	A search strategy was not provided
Heterogeneity	Assessed as I 2 but not reported in the results
Study quality	Risk of bias: data not presented as required by Cochrane
PRISMA checklist	Not reported

PRISMA, Preferred Reporting Items for Systematic Review and Meta‐analysis

Secondly, the lack of blinding in most of the trials included in the meta‐analysis by Tsoi et al. 30 was also a key limitation due to the fact that all of the included trials reported a short follow‐up period, which might have resulted in an observer bias among physicians and patients. As a result, the overall quality scores of included trials were weak. Thirdly, individual trials included patients reported in trials someone have a high risk stigmata of rebleeding compared to others, and the results may not be applicable to patients who are at high risk of rebleeding. In addition, the statistical data obtained from the available literature cannot be used in a meta‐analysis of individual patients by different stigmata. Moreover, endoscopic therapies were applied heterogeneously among the trials; patients were selectively treated by epinephrine injection alone, thermocoagulation or haemoclips. A meta‐analysis proved that the application of haemoclips is superior to injection alone but comparable to thermocoagulation for definitive haemostasis 6 so the use of epinephrine injection alone should be avoided.

In addition, available and selected trials in the meta‐analysis by Tsoi et al. used different pharmacokinetic characteristics, including pantoprazole, omeprazole, rabeprazole, lansoprazole and esomeprazole. The pharmaceutical formulas of these PPIs are similar but not identical, so they can potentially have different efficacies. Subgroup analysis based on different PPIs can be assessed but the available sample size is too small. Finally, IV PPIs require continuous infusion in hospital for at least 3 days and therefore cannot be compared with oral PPIs in terms of the duration of hospitalization. IV PPIs have been shown to reduce recurrent bleeding after endoscopic treatment of bleeding peptic ulcers 12, 14. The study by Tsoi et al. 30 showed that the efficacy of oral PPIs is similar to that of IV PPIs, but the conclusions that can be drawn from this study are limited by an insufficient sample size 30.

In view of the scarcity of economic resources, the cost of medical treatment is of considerable importance for hospital budgets. The cost of administering PPIs is determined by the nature of the pharmacotherapy; therefore, IV PPIs are far more expensive than orally administered forms 55 because they require dedicated IV lines, nursing care and continuous clinical monitoring, contributing to a longer hospitalization and higher costs. Conversely, the use of oral PPIs allows cost savings for healthcare systems as such patients can be discharged earlier.

A cost‐effectiveness analysis performed by Spiegel et al. 56, using a decision analysis to measure the clinical and economic outcomes of three different modes of PPI administration, showed that oral PPIs had a cost‐effectiveness advantage over the IV route. However, these findings were not consistent with previous studies, which showed that IV PPIs were likely to be both more efficacious and cost effective when compared with oral PPIs 57, 58.

Our systematic review, although adding a new trial 31 and a trial not previously reported 24 in other systematic reviews 15, 30, had some limitations. Fifty per cent of the trials included in our meta‐analysis were at a high risk of performance and detection bias. Furthermore, the sample size in some of the RCTs included was too small, resulting in studies that were underpowered to demonstrate a statistically significant difference between the two groups (oral vs. IV), leading to unreliable conclusions which would have limited the strength of our meta‐analysis. To solve the problem of underpowering of the meta‐analysis, a non‐inferiority trial could be carried out, in which the maximum mean differences would be prespecified; however, the cost of such a trial could not be justified.

Finally, it should be pointed out that a systematic review of RCTs represents the highest level in the hierarchy of studies 59, 60. Until future studies produce different data, the present study provides the best evidence to date.

In conclusion, the present meta‐analysis showed that oral and IV PPIs have a similar efficacy after endoscopic treatment in controlling recurrent bleeding, the requirement for surgery and mortality in patients with peptic ulcer bleeding from different stigmata. Oral PPIs probably present a cost‐saving approach in hospital administration. A subgroup analysis showed that the need for blood transfusion was significantly lower in the oral group when comparing high‐dose oral PPIs to high‐dose IV PPIs.

Large, well‐designed RCTs are needed to allow a definitive conclusion to be drawn.

Competing Interests

There are no competing interests to declare.

I am grateful to my teacher, Professor Graeme Alexander, Consultant Hepatologist, Addenbrooke's Hospital, Cambridge and Royal Free Hospital, London, UK, and President of the British Association for the Study of the Liver (BASL), for his invaluable help in revising this article.

Tringali, A. , Manta, R. , Sica, M. , Bassotti, G. , Marmo, R. , and Mutignani, M. (2017) Comparing intravenous and oral proton pump inhibitor therapy for bleeding peptic ulcers following endoscopic management: a systematic review and meta‐analysis. Br J Clin Pharmacol, 83: 1619–1635. doi: 10.1111/bcp.13258.