Invited Commentary 

Antibiotic Prophylaxis in Variceal Hemorrhage—Time to Stop?

Catherine Mezzacappa, Guadalupe Garcia-Tsao

JAMA Intern Med 2025;185;(10):1204-1205. doi:10.1001/jamainternmed.2025.3841

In their systematic review and bayesian meta-analysis published in this issue of JAMA Internal Medicine, Prosty et al¹ assess the evidence supporting the universally recommended practice of administering prophylactic broad-spectrum antibiotics to patients with cirrhosis presenting with upper gastrointestinal hemorrhage. They conclude that antibiotic prophylaxis may not be warranted in all patients with cirrhosis. Using bayesian meta-analysis, they pooled data from 14 trials: 12 trials compared no antibiotic prophylaxis to any prophylaxis, and 2 compared shorter vs longer durations of prophylaxis. The authors compared no prophylaxis and shorter prophylaxis vs longer courses of antibiotics (median [range] prophylaxis duration, 7 [1-10] days). Applying a noninferiority margin of 5%, they found a 97.3% probability of noninferiority for all-cause mortality (ie, a high likelihood that no or shorter prophylaxis is not worse than longer prophylaxis), a 73.8% probability of noninferiority for rebleeding (suggesting a moderate chance that shorter or no prophylaxis is not worse than longer prophylaxis), and a significantly higher risk of bacterial infections (risk difference, 15.2%). However, the included trials¹ were heterogeneous with respect to patient populations, interventions studied, outcomes ascertained, and duration of follow-up. Importantly, 9 of the 14 trials¹ (64%) were conducted more than 20 years ago.

In the intervening decades, our understanding of cirrhosis has evolved: cirrhosis is no longer considered a uniform disease. In compensated cirrhosis, where the patient has not yet developed any decompensating event, the median survival exceeds 10 years. If the patient develops any complication (ie, ascites, variceal hemorrhage, or encephalopathy), the patient has decompensated, and median survival decreases to approximately 1.5 years. Further decompensation, marked by complications of the complications, such as refractory ascites, recurrent variceal hemorrhage, or hepatorenal syndrome, shortens survival to mere months.² A patient with variceal hemorrhage can present at any of these stages of cirrhosis and experience markedly different outcomes: the patient with variceal hemorrhage as the first decompensating event will be much more likely to survive than the patient with variceal hemorrhage in the setting of refractory ascites and/or acute kidney injury. This clinical heterogeneity was reflected in trials included in the meta-analysis¹: the trial by Gupta et al,³ which reported no differences in short-term rebleeding, infection, or mortality, included only patients with Child-Turcotte-Pugh A (compensated) cirrhosis; and the study by Lo et al,⁴ which included patients with all stages of cirrhosis, found bacterial infections occurred only in patients with decompensated cirrhosis.

The rationale for antibiotic prophylaxis is based on the theory that variceal hemorrhage increases bacterial translocation which, in turn, worsens the hemodynamic status of the patient with cirrhosis, leading to an increase in portal pressure and rebleeding. Consequently, nearly all the trials included in this meta-analysis were designed to compare a primary outcome of rebleeding and had short durations of follow-up. The potential harms of antibiotic overuse—an increased number of infections and infections with resistant organisms over time—were not captured and could be associated with higher mortality.

In addition, trials of antibiotic prophylaxis performed in the 1990s were informed by and conducted in a clinical context of much higher mortality in patients with cirrhosis and upper gastrointestinal hemorrhage. Management of variceal hemorrhage has greatly improved since that time. Use of vasoactive agents, restrictive transfusion strategies, new endoscopic treatments, and preemptive transjugular intrahepatic portosystemic shunt have decreased mortality related to variceal hemorrhage. For example, in-hospital mortality in an advanced liver center⁵decreased from 42.6% to 14.5% from 1980 to 2000, as did rebleeding (from 47% to 13%) and bacterial infections (from 38% to 14%). Antibiotic prophylaxis was among several interventions credited with these improved outcomes.⁵ Therefore, vasoactive medications, restrictive transfusion, prompt endoscopy, and prophylactic antibiotics became a widely accepted bundled approach to improve outcomes for patients with cirrhosis and upper gastrointestinal hemorrhage.

Now, based on the findings of Prosty et al,¹ it may be time to unbundle our treatments and assess whether antibiotics continue to provide benefit in patients with cirrhosis and upper gastrointestinal bleeding, and if so, in which patients. This unbundling should be considered through carefully planned clinical trials. Future trials must be carefully designed to capture both the potential benefits and potential harms of antibiotic prophylaxis. It is essential that these trials be stratified by cirrhosis stage (ie, compensated, decompensated, further decompensated). Also, in the context of an advanced chronic illness (cirrhosis) and the potential for synergism between disease progression and treatment toxicity (antibiotic resistance), all-cause mortality will be a more meaningful outcome than cause-specific mortality.⁶

Although most of the clinical trials included in this meta-analysis¹ used 7 days of prophylaxis in the longer antibiotic duration group, there is mounting evidence that shorter durations of antibiotics are efficacious against most infections, including intra-abdominal infections (4 days) and pneumonia (3-5 days).⁷ A 2- or 3-day course of ceftriaxone may be sufficient to treat a range of subclinical or early infections. Therefore, to measure the effect of antibiotic prophylaxis, it will be necessary to not only compare duration of prophylaxis, but to include a comparison to patients receiving no antibiotic prophylaxis. Given the heterogeneity of risk in cirrhosis, patients who present with upper gastrointestinal bleeding, but who are otherwise compensated, are the most appropriate first population in whom to assess whether antibiotic prophylaxis can be safely withheld.

The findings of this systematic review and meta-analysis¹ highlight the paucity of high-quality evidence to support the practice of administering prophylactic broad-spectrum antibiotics to patients with cirrhosis presenting with upper gastrointestinal hemorrhage. The current level of evidence is also inadequate to answer whether it is time to stop this practice, which has become the standard of care. Modern trials of shorter duration and no antibiotic prophylaxis should be designed in specific patient populations to compare sequelae of antibiotic prophylaxis, including subsequent infections and all-cause mortality.