In summary, patients who received prophylactic antibiotics:

Benefits in NNT

  • 4.6% reduction in risk of death in patients receiving antibiotics (compared to placebo or no treatment)
  • 23% reduction in hospital-acquired bacterial infections in patients receiving antibiotics (compared to placebo or no treatment)
  • 1 in 22 death prevented
  • 1 in 4 infection prevented

Harms in NNT

    No one was harmed
    No one was harmed

View As:

Source: Chavez-tapia NC, Barrientos-gutierrez T, Tellez-avila FI, Soares-weiser K, Uribe M. Antibiotic prophylaxis for cirrhotic patients with upper gastrointestinal bleeding. Cochrane Database Syst Rev. 2010;(9):CD002907.

Study Population: 1241 patients with cirrhosis hospitalized with upper gastrointestinal bleeding from 12 trials comparing prophylactic antibiotics to no antibiotic prophylaxis

Efficacy Endpoints: Mortality, infection during hospitalization (spontaneous peritoneal peritonitis, bacteremia, pneumonia, urinary tract infection)

Harm Endpoints: Not Reported

Narrative: Cirrhotic patients often develop bleeding from gastric or esophageal varices that occur secondary to portal hypertension. Gastrointestinal (GI) bleeding is fatal in approximately 20% of these episodes and bacterial infections are an important contributor to this mortality. Patients with cirrhosis are also known to have impaired immune function and also at higher risk of translocation of bacteria from the gut into the bloodstream.1 Therefore, the administration of prophylactic antibiotics during the bleeding event might help prevent such infections.

The Cochrane systematic review discussed here included 12 trials (n = 1241) involving cirrhotic patients with upper GI bleeding. Of the 12 included trials, only 1 was placebo controlled, the other 11 examined antibiotics vs. no intervention.2 These trials enrolled adult patients with cirrhosis and upper GI bleeding regardless of the severity or etiology of the cirrhosis. They excluded patients who had bacterial infections at the time of admission, positive blood cultures or who underwent surgery in the first 12-24 hours of hospitalization. Among the included trials, the length of follow up for determining mortality endpoint ranged from in-hospital to 90 days. The analysis demonstrated a clear decrease in overall rate of hospital-acquired bacterial infections, with marked reductions in nosocomial bacteremia, pneumonia, spontaneous bacterial peritonitis and urinary tract infections (Odds ratio [OR]: 0.36, 95%CI, 0.27 to 0.49; Absolute risk difference [ARD]: 23%; Number-needed-to-treat [NNT]: 4). With the exception of pneumonia, all of the infections were confirmed by cultures. The trials also noted an overall decrease in mortality (OR: 0.79, 95%CI, 0.63 to 0.98; ARD: 4.6%; NNT: 22). The choice of antibiotic regimen appeared to have no effect, although all antibiotics used in these trials were chosen because of their activity against gram negative organisms (the most common infecting agents for the targeted infection types). The most common antibiotics used in the trials were quinolones followed by cephalosporins. The subgroup analysis showed more benefits from cephalosporins than quinolones for reducing bacterial infections.

One study that was published after the Cochrane systematic review, reported a retrospective analysis of 381 patients with cirrhosis and variceal upper GI bleeding. This represented one of the most relevant studies on this topic since the completion of the most recent Cochrane systematic review, even though its retrospective nature would preclude it from being included in an updated Cochrane review on this topic. It found that antibiotic prophylaxis was associated with a lower risk of infection (OR: 0.37; 95% CI 0.31 to 0.74) but no significant change in overall mortality.3 However, subgroup analysis did find a mortality benefit that was severity-dependent: in patients with Child-Pugh class C (i.e. more severe cirrhosis) antibiotics reduced 6-week mortality by approximately 50% (from 62% in those not exposed to antibiotics to 35% in those exposed to antibiotics). In patients with Child-Pugh class B (i.e. less severe cirrhosis), the drop in mortality after using antibiotics was from 7% in nonexposed to 5% in exposed groups. The mortality in patients with Child-Pugh class A (i.e. mild cirrhosis) was negligible regardless of antibiotic administration.3 These findings generally support the conclusions of the systematic review discussed here.

Caveats: We should also note that while this review is over a decade old, a review of the literature did not reveal any new trials that would have impacted the conclusions of this review. None of the included trials reported harms or adverse effects associated with administration of antibiotics. The authors of the Cochrane systematic review themselves make a point to note that, “Adverse events, quality of life, and the economic impact of the intervention were not explored in the trials included, remaining important areas of uncertainty and requiring further data to establish an evidence‐based conclusion.” Furthermore, only 1 trial was placebo controlled, which introduces a significant risk of bias. Perhaps more importantly, as noted by the Cochrane review authors, the data for decreased mortality in the intervention group was not as compelling as it was for preventing infection. Many of the included trials were not powered to determine a mortality benefit. Furthermore, trial sequential analysis (a statistical tool used to evaluate the strength of results found during meta-analysis) found that the 12 included trials were not enough to produce a definitive conclusion regarding the survival benefit. This indicates that a large, high quality, methodologically rigorous randomized trial has the potential to trump the results of this review, as it may well generate results that are in disagreement with the results of this review (for all outcomes including infection rates). We would like to see a trial like this performed, and we believe that given the poor quality of existing trial data there is clinical equipoise adequate to perform such a trial.

The trials included in the meta-analysis suffered from various methodological limitations and therefore the produced evidence is subject to bias. None of the included trials were rated as low risk of bias. The heterogeneity for the mortality endpoint was low but it was significant for bacterial infection. Regardless of these limitations, as this review represents the best available data, and given the reported benefits of reducing hospital-acquired infections and the possible decrease in mortality, it seems appropriate to recommend this intervention in cirrhotic patients with upper GI bleeding. However, the adverse events associated with antibiotics such as allergic reactions, rash, gastrointestinal upset, clostridium difficile infections, antibiotic resistance, etc. should be balanced against these benefits on a case by case basis.

Based on the existing evidence, we have assigned a color recommendation of Green (benefits outweigh harms), as the reported benefits are significant and clinically relevant. As noted above, further data would be welcome to better characterize the degree of benefit, and assess any adverse events associated with antibiotic prophylaxis.

The original manuscript was published in Academic Emergency Medicine as part of the partnership between and AEM.

Author: Daniel Runde, MD
Supervising Editor: Shahriar Zehtabchi, MD

Published/Updated: August 24, 2020

  1. The Title Bar

    The title bar is color-coded with our overall recommendation.

    • Green: Benefits outweigh risks.
    • Yellow: Unclear risk/benefit profile.
    • Red: Benefits do not outweigh risks.
    • Black: Obvious harms, no clear benefits.
  2. Tip content...