文献速递消化急症脓毒症与休克重症综合管理

重新审视核心实践：肝硬化出血患者的抗生素预防策略

Revisiting a Core Practice: Antibiotic Prophylaxis in Cirrhotic Bleeds

RSS 订阅 · REBEL EM · 2026-09-14

一句话结论肝硬化上消化道出血患者使用短疗程（含不用）抗生素预防在死亡率上非劣效于指南推荐的5-7天疗程（非劣效概率97.3%），但感染率增加15.2%。

这项贝叶斯meta分析纳入14项RCT共1322例肝硬化上消化道出血患者，比较短疗程（2-3天或不用）与标准5-7天抗生素预防疗程。主要结果显示短疗程在全因死亡率上非劣效于标准疗程，绝对风险差0.9%（95% CrI: -2.6%至4.9%），非劣效概率97.3%。次要结果显示短疗程组细菌感染率增加15.2%（95% CrI: 5.0%至25.9%），早期再出血未明确达到非劣效（非劣效概率73.8%）。2004年后研究死亡率非劣效概率升至98.9%。现行5-7天抗生素预防指南推荐值得重新评估，但仍需高质量RCT确认。

要点

肝硬化上消化道出血患者短疗程抗生素（2-3天或不用）对比标准5-7天疗程，全因死亡率绝对风险差0.9%（95% CrI: -2.6%至4.9%），非劣效概率97.3%，死亡率事件83/665 vs 62/657
短疗程组细菌感染率增加15.2%（95% CrI: 5.0%至25.9%），非劣效概率仅2.4%，感染事件182/665 vs 78/657，但感染定义存在异质性且偏倚风险高
早期再出血风险差2.9%（95% CrI: -4.2%至10.0%），非劣效概率73.8%（91/545 vs 69/544），未明确达到非劣效标准（I²=71.7%）
2004年后（内镜治疗、血管活性药物、输血策略等现代治疗进步后）发表的研究显示短疗程死亡率非劣效概率提升至98.9%，而2004年前为72.4%
非劣效界值设定为5%绝对风险差；纳入14项RCT共1322例患者（12项比较抗生素vs不用，2项比较短疗程vs长疗程），无研究报告抗生素不良事件

适用人群与场景

肝硬化合并上消化道出血的成年患者
排除观察性研究、灰色/未发表文献、儿科人群、非系统性抗生素方案

诊断标准与关键阈值

原文未提及

处置流程与药物剂量

短疗程组：抗生素预防2-3天，或单次剂量至完全不用抗生素（0天）
标准疗程组：抗生素预防5-7天（国际指南推荐疗程）
原文未提供具体抗生素种类、剂量、给药途径或频次

红旗征与不能漏的情况

短疗程抗生素预防使细菌感染率增加15.2%（95% CrI: 5.0%至25.9%），需权衡死亡率非劣效与感染风险增加
早期再出血未明确达到非劣效（非劣效概率73.8%），存在临床不确定性

研究设计与关键结果

贝叶斯meta分析，纳入14项RCT共1322例患者（12项比较抗生素vs不用，2项比较短疗程vs长疗程）
主要终点全因死亡率：短疗程组83/665 vs 标准疗程组62/657，绝对风险差0.9%（95% CrI: -2.6%至4.9%），非劣效概率97.3%，I²=40.8%
次要终点早期再出血：91/545 vs 69/544，绝对风险差2.9%（95% CrI: -4.2%至10.0%），非劣效概率73.8%，I²=71.7%
次要终点细菌感染：182/665 vs 78/657，绝对风险差15.2%（95% CrI: 5.0%至25.9%），非劣效概率2.4%，I²=83.9%
无研究报告抗生素不良事件
2004年后研究亚组分析显示死亡率非劣效概率98.9%，2004年前为72.4%

证据类型、等级与局限

证据类型：系统评价与贝叶斯meta分析，纳入14项RCT
研究质量评估：使用Cochrane risk of bias 2工具
局限性：感染定义存在异质性并引入高偏倚风险；早期再出血结果异质性高（I²=71.7%）；感染率分析异质性高（I²=83.9%）
非劣效界值设定为5%绝对风险差（作者定义为「临床有意义」）
早期研究（2004年前）在内镜治疗、血管活性药物、输血策略等现代治疗进步前进行，方法学局限性包括偏倚风险高

对床旁实践的改变

现行肝硬化上消化道出血5-7天抗生素预防指南推荐值得重新评估，短疗程或不用抗生素在死亡率上可能非劣效
短疗程策略可能减少抗生素暴露，但需权衡细菌感染率增加15.2%的风险
2004年后现代治疗背景下（内镜、血管活性药物、输血策略），短疗程非劣效证据更强（98.9%）
仍需高质量RCT验证后才能做出改变实践的结论，目前证据尚不足以立即改变床旁常规5-7天预防方案
适用于肝硬化上消化道出血成年患者；不适用于儿科、非系统性抗生素方案或已确诊感染需治疗的患者

指南或共识出处

Prosty C, et al. Prophylactic Antibiotics for Upper Gastrointestinal Bleeding in Patients With Cirrhosis: A Systematic Review and Bayesian Meta-Analysis. JAMA Internal Medicine. 2025;185(10):1194–1203. PMID: 39271740
研究遵循PRISMA指南与Cochrane手册，方案注册于PROSPERO (CRD4202459444)
检索数据库：MEDLINE、CENTRAL、Embase，截至2024年9月25日

以上提炼依据论文摘要（全文未开放获取，或许可不允许翻译转载）；全文请点「阅读原文」。

原文图文 · 图 28 张

REBEL Rundown

Key Points

Short vs. standard antibiotic prophylaxis in cirrhotic UGIB: This Bayesian meta-analysis of 14 RCTs (n=1,322) compared shorter courses — including no prophylaxis — against the guideline-recommended 5–7 days in cirrhotic patients with upper GI bleeding.

Non-inferior mortality with shorter or no prophylaxis: Shorter antibiotic courses demonstrated a 97.3% probability of non-inferiority for mortality compared to standard 5–7 day regimens.

Similar rates of rebleeding and bacterial infection: Early rebleeding and infection rates did not meaningfully differ between shorter and standard prophylaxis durations.

Current guidelines warrant reappraisal: The 5–7 day antibiotic recommendation for cirrhotic UGIB deserves reconsideration — but high-quality RCTs are still needed before practice-changing conclusions can be drawn.

Introduction

Upper gastrointestinal bleeding (UGIB) is a complication in liver cirrhosis patients that can ultimately be life-threatening. There are notable risks of rebleeding, infections and mortality. For many years, the international guidelines recommended 5 to 7 days of prophylactic antibiotics for cirrhotic patients with UGIB. This is based on older trials conducted before major advances that were introduced after 2004, e.a. endoscopic therapies, vasoactive medications, transfusion strategies, and supportive care. 1

However, many of these earlier studies were limited by methodological weaknesses, heterogeneity in infection definitions, and a high risk of bias, raising uncertainty about the true benefit of prolonged prophylaxis.1,2

The authors of this paper sought to compare shorter durations of antibiotic prophylaxis, (including 0 days) with the current recommendation of longer courses via 5–7 days in patients with cirrhosis and upper gastrointestinal bleeding. Their paper also evaluates if shorter antibiotic durations were overall noninferior in preventing all‑cause mortality, early rebleeding, and bacterial infections.

Paper

Prosty C, et al. Prophylactic Antibiotics for Upper Gastrointestinal Bleeding in Patients With Cirrhosis: A Systematic Review and Bayesian Meta-Analysis. JAMA Internal Medicine. 2025;185(10):1194–1203. PMID: 39271740

Previously Covered On REBEL

Core Cast 39.0 – Upper GI Bleed

Cast Ep82: Timing of Endoscopy for UGIB

Core Content: Spontaneous Bacterial Peritonitis

Core Cast 22.0 – Decompensated Liver Disease

TEG-Guided Resuscitation of Patients with Cirrhosis and Non-Variceal Bleeding

What They Did

CLINICAL QUESTION

In adults with cirrhosis and upper gastrointestinal bleeding, does shorter‑duration or no antibiotic prophylaxis, compared with longer‑duration prophylaxis, lead to better clinical outcomes such as reduced mortality, infection, or adverse events?

STUDY DESIGN

Systematic review, Bayesian meta-analysis of 14 randomized clinical trials (RCTs)

Conducted according to the protocol registered with PROSPERO (CRD4202459444)

Followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline and the Cochrane Handbook

Databases utilized for gathering papers e.a. MEDLINE, CENTRAL, Embase

Study quality was assessed using the Cochrane risk of bias tool for randomized trials

Noninferiority Margin: The 5% absolute risk difference margin chosen as “clinically meaningful”. Larger differences beyond the margin would be clinically significant and ultimately favor longer antibiotic duration therapy.

POPULATION

Inclusion Criteria:

Patients with cirrhosis that are presenting with an upper GI bleed, comparing longer durations (5-7 days) vs shorter (2-3 days) and no antibiotics or systemic antibiotic regimens

Exclusion Criteria:

Observational studies,

Gray/unpublished literature

Pediatric cohorts

Non-systemic antibiotic regimens

INTERVENTION & COMPARATOR

Intervention Group:

Shorter-duration antibiotic prophylaxis, typically 2–3 days. Also includes no antibiotic prophylaxis, ranging from a single dose to completely avoiding antibiotic use for 0 days.

Comparator Group:

Longer-duration antibiotic prophylaxis, typically 5-7 days.

OUTCOMES

Primary Outcome:

All-cause mortality

Secondary Outcomes:

Early rebleeding

Bacterial infections

Results:

Databases searched: Embase, MEDLINE, and CENTRAL.

Search timeframe: From database inception through September 25, 2024.

Search terms: Focused on gastrointestinal bleeding and prophylactic antibiotics.

Study type sought: Randomized clinical trials evaluating antibiotic prophylaxis strategies in patients with cirrhosis and upper gastrointestinal bleeding.

Initial records identified: 621 records were screened.

Full-text articles reviewed: 18 articles underwent full-text review.

Studies included: 14 randomized clinical trials were included in the final systematic review and Bayesian meta-analysis.

Total participants: 1322 patients were included across the 14 trials.

Types of comparisons:12 RCTs compared antibiotic prophylaxis with no prophylaxis.

2 RCTs compared shorter-duration prophylaxis with longer-duration prophylaxis.

Critical Results

Shorter and no antibiotic durations showed 97.3% probability of noninferiority for all-cause mortality, compared to longer antibiotic duration

Absolute risk difference for mortality was within the 5% noninferiority margin: 0.9% (95% CrI: -2.6% to 4.9%)

Reports showcase a 15.2% increase in bacterial infections via shorter duration antibiotics

Infection definitions were heterogeneous, ultimately introducing high risk of bias

There was no reports made of adverse events from antibiotics

Post-2004 studies regarding modern advancements revealed stronger evidence for noninferiority of shorter antibiotic durations

PRIMARY OUTCOMES:

All-cause mortalityShorter durations, including no prophylaxis, were noninferior to longer durations.

Events: 83/665 vs 62/657

Risk difference: 0.9%

95% credible interval: −2.6% to 4.9%

Probability of noninferiority: 97.3%

Heterogeneity: I² = 40.8%

Secondary outcome: Early rebleedingShorter durations did not clearly meet noninferiority.

Events: 91/545 vs 69/544

Risk difference: 2.9%

95% credible interval: −4.2% to 10.0%

Probability of noninferiority: 73.8%

Heterogeneity: I² = 71.7%

Secondary outcome: Bacterial infectionsShorter durations were associated with more study-defined bacterial infections.

Events: 182/665 vs 78/657

Risk difference: 15.2%

95% credible interval: 5.0% to 25.9%

Probability of noninferiority: 2.4%

Heterogeneity: I² = 83.9%

Probability of noninferiority: 81.0%

Adverse eventsNo included studies reported adverse events.

Era-based subgroup: Studies published after 2004Probability of noninferiority for mortality increased to 98.9% in post-2004 studies.

Pre-2004 studies had a lower probability of noninferiority: 72.4%.

The authors attributed this partly to lower modern baseline mortality and smaller between-group differences.

Strengths

Clinical Question: The question highlighted in this article is sensible, relevant and defined well. It covers the pressing factors that are central to today’s practice such as standard practice of antibiotics prophylaxis, clinical practice guidelines depending on updated evidence and more

Randomized Controlled Trials: There was a major focus on RCTs, with 14 total used in this study.

To increase accuracy, the Cochran Risk of Bias 2 tool was used by reviewers on an independent basis

Database Search: Comprehensive search via databases like MEDLINE, CENTRAL, Embase was used, lead up to September 25th 2024 in order to gather the most from relevant studies

Modern Advancements: With advances of upper GI bleed management methods, a subgroup analysis of trials after 2004 was highlighted to look at more current practices.

Patient-centered Outcomes: this includes all-cause mortality, early rebleeding, and bacterial infections. There is emphasis on actively applying medical decision making decisions

Transparency of Articles: the selection of articles in this meta-analysis appeared to be reproducible and transparent with their data

Use of Reviewers for Papers: It is noted that 2 independent reviewers were able to review papers

Limitations

Quality of studies: Cochrane ROB2 model, studies were found to be of low to moderate quality

Definitions and Terms: Broad definitions used to identify types of infections in the studies; this introduced risk for bias

Post-2004 Overview: Overview of data from studies prior to 2004 may not reflect modern practices for upper GI bleed management

Specific Variables and Types: Looking at specific antibiotics, types of bleeds and their sources, management protocols that may guide the clinician to a better decision from a medical standpoint, but was not elaborated in this study. No specific recommendations given overall.

No Adverse Effects Mentioned: lack of adverse event reports in these studies highlights an overall incomplete analysis, given that there are always side effects to any antibiotic use

Small Groups to Compare: specifically, comparing longer duration antibiotics which only was highlighted in 2 RCTs vs shorter duration antibiotics via 12 RCTs

Publication Bias: the funnel plot asymmetry illustrated in this paper is suggestive of this

Exclusion of Observational Trials: though this potentially increases quality of papers included, it introduces reporting bias

Discussion

Heterogeneity:

A major discussion point in this meta-analysis is the heterogeneity across the included trials, seen through the I² statistic; this factor quantifies how much of the variability in effect estimates is due to differences between studies. In this review, heterogeneity arises because the trials vary across variables. This includes antibiotic choice, duration of prophylaxis, infection risk at baseline, upper GI bleeding management, and period of care in which the studies were conducted. These differences may complicate the assumption that all studies are estimating the same underlying treatment effect, making it challenging to gather results confidently or generalize findings to modern practice.

Noninferiority Requires Clinical Judgment

Noninferiority is not entirely statistical. Rather, it shows a value judgment about what difference clinicians and patients are able to accept. Regarding this paper, selecting a small noninferiority margin, e.a. 5%, would be very difficult to meet given the available data, whereas selecting a wider margin makes it easier to conclude that a shorter or alternative strategy is “not worse.” Even when a treatment satisfies the statistical margin, clinical differences may still stand; those differences can influence outcomes such as infection rates, mortality, or antibiotic use with consequences.

Risk of Bias in Infection Outcomes

Bias is a major limitation in this meta-analysis due to infection outcomes being highly susceptible to measurement bias. Several of the included RCTs scored poorly on risk‑of‑bias assessments, raising concerns about reliability of the reported effects. Overall, these issues make it difficult to fully trust or utilize the gathered results, even if there were consistent statistical findings.

Mortality Benefit Is Not Clearly Supported

Another important discussion point is that this meta-analysis challenges the presumed mortality benefit behind current guideline recommendations. The review found a high probability that shorter-duration antibiotics, including no prophylaxis, were noninferior for all-cause mortality. However, this finding should be interpreted with caution because the certainty of evidence was low and several trials had important limitations. This doesn’t prove antibiotics are useless. It implies the evidence of mortality benefit supporting routine 5–7 days of prophylaxis is less robust than what guidelines outline.

Patient Selection May Be the Future Direction

Rather than asking, “should all patients receive antibiotics or not?”, one could ask “which patients with cirrhosis and upper GI bleeding benefit enough from antibiotics to justify the risks?” In higher-risk patients, such as those with advanced cirrhosis, shock, ascites, suspected SBP, ICU-level illness, or high baseline infection risk, may still benefit. Lower-risk patients, especially those with Child-Pugh A disease and controlled bleeding, may be candidates for shorter therapy or no prophylaxis.

Author's Conclusion

“The findings of this systematic review and bayesian meta-analysis do not support the purported mortality benefit driving guideline recommendations for antibiotic prophylaxis in patients with cirrhosis and upper gastrointestinal bleeds. Although prophylaxis reduced reported infections, methodological concerns regarding infection definitions introduced high risk of bias. Higher-quality RCTs are needed to determine the benefit and optimal duration of antibiotic prophylaxis in the modern era of advanced interventions. Until these studies are available, clinicians should be aware that the current guideline recommendations are not based on high-quality evidence.”

Our Conclusion

We support the authors conclusion on needing more RCTs and data to support what would be the preferred period of time to have patients on antibiotic prophylaxis. In a situation where prevention of infection is prioritized (vs prevention of rebleeding), having longer duration of antibiotic use continues to be the current practice. However, the duration of antibiotic administration is still debatable on a case-by-case basis.

Clinical Bottom Line

This meta-analysis challenges routine 5–7 day antibiotic prophylaxis for all cirrhotic UGIB patients. Shorter courses may be reasonable, especially in lower-risk patients, but the evidence is limited by heterogeneity, bias, poor infection definitions, and lack of harm data.

References

Prosty et al. 2025

Prophylactic antibiotics for upper gastrointestinal bleeding in patients with cirrhosis: a systematic review and Bayesian meta-analysis. JAMA Intern Med. 2025

PMID: 40788637

Medical Dialogues 2025

JAMA study questions routine prophylactic antibiotics use for upper‑GI bleeding in cirrhosis. 2025.

Link is Here

Gao et al. 2022

Prophylactic antibiotics on patients with cirrhosis and upper gastrointestinal bleeding: a meta-analysis. PLoS One. 2022;

PMID: 36548353

Mansoor et al. 2021

Antibiotic prophylaxis for upper gastrointestinal bleed in liver cirrhosis: less may be more. Open Forum Infect Dis. 2021

PMID: 35467310

Post Peer Reviewed By: Marco Propersi DO, (X/IG: @Marco_Propersi) and Mark Ramzy, DO (X/IG: @MRamzyDO)

Meet the Authors

Kevin Tomecsek, MD

Clerkship Director and Assistant Program Director

AdventHealth, Orlando, FL

Azalia Samsam, DO

PGY-3 Emergency Medicine Resident

AdventHealth, Orlando, FL

Showing Slide 1 of 2

Your Deep-Dive Starts Here