Description of the condition

Liver cirrhosis is the advanced stage of liver fibrosis, and histologically, it is defined as the presence of regenerative nodules surrounded by extensive fibrosis (scarring of the liver) (Schuppan 2009; Moon 2016). Even though the definition of liver cirrhosis is histological and the gold standard for diagnosing liver cirrhosis is liver biopsy, most of the time diagnosis is made only from clinical history, clinical features, and laboratory abnormalities. The absence of standardised criteria for diagnosing liver cirrhosis when liver biopsy is not performed, often leads to an unclear boundary between liver disease, liver fibrosis, and liver cirrhosis (Ratib 2017).

In 2001, cirrhosis was the 14^th cause of death in the world and the 10^th in low‐income countries (Sarin 2012). In most low‐income countries and in some high‐income countries, such as the UK, the risk of death related to cirrhosis has increased not only because of increased alcohol consumption and obesity, but also because of previous underreported mortality (Ratib 2017). A systematic analysis with data from 187 countries reported an increase in deaths due to liver cirrhosis from 676,000 in 1980 to over 1 million in 2010 (2% of all deaths) (Mokdadd 2014). In the USA, the estimated annual direct healthcare costs amounted up to $2.5 billion ($2500 million), while the annual indirect costs related to loss of work productivity and reduction in health‐related quality of life were $10.6 billion ($10,600 million) (Neff 2011).

Liver cirrhosis is commonly recognised as resulting from either 1) a result of toxic, infectious, immunopathological, or vascular insult that chronically damages the liver, or 2) a result of the accumulation of substances due to an inborn error of metabolism (i.e. iron in haemochromatosis, copper in Wilson's disease, or some inborn errors of metabolism of carbohydrates, amino acids, organic acids, and others) (Wiegand 2013). The most common specific causes of liver cirrhosis are chronic hepatitis B infection, alcohol, chronic hepatitis C infection, and non‐alcoholic fatty liver disease (Hsiang 2015).

One third of people with liver cirrhosis will remain asymptomatic (Scaglione 2015). Of those presenting with symptoms, non‐specific and general manifestations, such as anorexia, fatigue, weight loss, and muscle wasting are the most common. The most reported clinical features are jaundice (yellow discolouration of skin and eyes), ascites (accumulation of fluid in the abdominal cavity), spider angioma (arterioles with tiny radiating vessels visible in the skin), irregular and hard liver on palpation, enlarged spleen, caput medusae (prominent veins radiating from umbilicus), Cruveilhier Baumgarten syndrome (epigastric vascular murmur), reddening of the palms that spares the central portion, white bands on nails, Dupuytren’s contracture (fibrosis and contraction of the palms), gynaecomastia (proliferation of glandular male breast tissue), and loss of male hair pattern (Schuppan 2009).

Model for End Stage Liver Disease (MELD) score and the Child‐Pugh score are among the most‐used scoring systems for prognosis of liver cirrhosis. MELD score uses bilirubin, creatinine, international normalised ratio, and sodium levels to predict short‐term mortality in people with liver cirrhosis and is used for people on the waiting list for liver transplantation (Wiesner 2003; Kim 2008). Child‐Pugh score uses presence of ascites, presence of encephalopathy, and serum bilirubin, albumin and international normalised ratio levels, and was originally designed to predict mortality during surgery, but it is now used to assess severity and predict mortality for longer periods than MELD score (one and two years) (Child 1964). Fibrosis can also be classified histologically, using Laennec staging system with five grades: F0 (no fibrosis) to F4 (advanced fibrosis), depending on the amount of fibrotic tissue, and with a sub‐classification that divides stage 4 in A to C, depending on the width of fibrous septa and the amount of nodules detected on the tissue sample (Bedossa 1996; Kim 2011).

Liver cirrhosis increases resistance through the hepatic sinusoids, causing an increase in portal blood flow which leads to endothelial dysfunction with imbalance of vasodilator and vasoconstrictor factors, and a distortion of the hepatic vascular bed (Iwakiri 2012). Vasodilator factors and changes in hepatic circulation lead to arterial peripheric vasodilatation, mainly in the splanchnic area that contributes to a hyperdynamic state with decreased effective blood volume (Hilzenrat 2012). The presence of clinically significant portal hypertension (portal pressure gradient ≥ 10 mmHg) promotes the formation of collateral portosystemic circulation, portal hypertensive gastropathy, gastric varices, and oesophageal varices (Bosch 2008).

Variceal development and progression depends on the severity of portal hypertension and severity of liver disease (Garcia‐Tsao 2007). Kovolak 2007 reported the presence of varices in 42% of patients with Child‐Pugh class A and in 71.9% of patients with Child‐Pugh B or C cirrhosis.

Variceal haemorrhage is the consequence of variceal rupture, which is explained by an excessive wall tension. Wall tension in oesophageal and gastro‐oesophageal varices is determined mainly by three factors: 1) increased portal pressure, 2) increased size of the vessel, and 3) decreased wall thickness (Laplace’s law). Therefore, the associated risk factors for developing variceal bleeding are portal pressure, the size of the vessel, and the endoscopic characteristics of the vessel that translate into thinner walls (red marks) (Hilzenrat 2012).

Upper gastrointestinal bleeding comprises haemorrhage from the oesophagus to the ligament of Treitz, recently redefined as a bleeding source that is potentially visualised by upper endoscopy (Kim 2014). Upper gastrointestinal bleeding remains one of the most severe decompensating events in people with cirrhosis, and the second most frequent severe decompensating event after ascites (EASL 2018). Variceal haemorrhage causes 70% of the upper gastrointestinal bleeding events in people with ascites and cirrhosis (Mallet 2017). Risks of death within six weeks due to variceal haemorrhage ranges from 15% to 25%, and is particularly high when associated with acute kidney injury or infections (D'Amico 2014; EASL 2018).

People with upper gastrointestinal bleeding are at high risk of developing bacterial infections (Garcia‐Tsao 2017). Up to 50% of active upper gastrointestinal bleeding events are related to the presence of bacterial infections, and infections are considered an independent predictor of failure to control bleeding and mortality (Goulis 1998).

The association between infection and variceal bleeding in people with cirrhosis is reciprocal and its implications have been extensively studied (Bernard 1995; Goulis 1998; Augustin 2009). It is not always possible to define if a bleeding event leads to the development of infection or vice versa (Lee 2014). If variceal bleeding leads to infection, then it could be due to instrumental procedures, endoscopic procedures, bacterial translocation, or placement of catheters (Blaise 1994). When infection is present, release of toxins occur leading to an activation of mediators that damage vessels and enhance the haemostatic impairment present in most people with cirrhosis, which might lead to failure to control a bleeding haemorrhage and re‐bleeding events (Lee 2014).

The most common bacterial infections developed in people with cirrhosis are caused by Gram‐negative bacteria, producing spontaneous bacterial peritonitis (25%), urinary tract infections (20%), pneumonia (15%), and bacteraemia (12%) (Fernandez 2002).

Description of the intervention

The prophylactic use of oral or intravenous antibiotics for people with cirrhosis and gastrointestinal bleeding has been recommended in consensus guidelines (Tipathi 2015; Garcia‐Tsao 2017; EASL 2018). The recommended drugs are mainly intravenous cephalosporins (ceftriaxone 1 g every 24 hours for seven days) or oral quinolones (norfloxacin 400 mg every 12 hours for seven days) (Garcia‐Tsao 2009). Glycopeptides, aminoglycosides, macrolides, and rifaximine have also been administered (Rimola 1985; Altraif 2011; Huang 2018).

Cephalosporins (e.g. cefalexin, cefalotin, ceftriaxone, cefotaxime, cefazolin, cefixime, cefepime, ceftolozane) are part of the beta‐lactam family, and the mechanism of action is to disrupt synthesis of the bacterial wall (Török 2009). There are five generations of cephalosporins: the first generation (e.g. cefalexin, cefazolin, cefalotin ) covers mostly gram‐positive bacteria, the second generation (e.g. cefaclor, cefuroxime, and cefotetan) has less activity for gram‐positive but greater activity for gram‐negative than the first generation, and the third generation (e.g. cefotaxime, ceftriaxone, cefixime) works mostly against gram‐negative bacteria (Marshall 1999). The fourth and fifth generation of cephalosporins (e.g. cefepime, ceftaroline, ceftolozane) are considered broad spectrum antibiotics, acting against gram‐negative and gram‐positive bacteria. In addition, some of them have antipseudomonal activity (Török 2009). Cephalosporins can be administered orally or intravenously. A large retrospective study with data from 352,661 people who had received cephalosporins reported rash as the most common adverse event, followed by anaphylactoid reaction (Shi 2013).

Quinolones (e.g. norfloxacin, ciprofloxacin, levofloxacin, ofloxacin) prevent bacterial replication by preventing DNA winding and duplication (Aldred 2014). Quinolones, available for both oral or intravenous administration, work against both gram‐negative and gram‐positive bacteria. Most common adverse effects are mild gastrointestinal symptoms, but other adverse effects are described, mainly involving the musculoskeletal system (specifically risk of tendon rupture) and arrhythmia conditions such as QT prolongation (Norrby 1991). Nevertheless, serious adverse events have been uncommonly reported, and quinolones are considerably cheaper than cephalosporins (Xu 2011).

Glycopeptides (e.g. vancomycin, teicoplanin, ramoplanin) have a similar mechanism of action as the beta‐lactam antibiotics, inhibiting bacterial wall synthesis of gram‐positive bacteria, but their use is limited to infections caused by beta‐lactam resistant bacteria because of the great risk of toxicity affecting kidneys and inner ears, and the high risk of hypersensitivity reactions (Török 2009). Glycopeptides were designed only for intravenous administration, but in recent years, oral administration was approved for the treatment of clostridium difficile‐associated diarrhoea (FDA 2018).

Aminoglycosides (e.g. streptomycin, gentamicin, neomycin) inhibit protein synthesis of gram‐negative bacteria (Mingeot‐Leclercq 1999). The associated adverse effects are inner ear toxicity and renal toxicity. Aminoglycosides can be administrated intravenously or orally (Mingeot‐Leclercq 1999).

Macrolides (e.g. erythromycin, clarithromycin, azithromycin) inhibit growth of bacteria by affecting protein synthesis (Hamilton‐Miller 1973). The main adverse effects are arrhythmias, reversible deafness, and allergic reactions (Periti 1993; Jespersen 2006; Winkel 2011; Winkel 2015; Mosholder 2018). Erythromycin is known to have a prokinetic effect, and, therefore, it is used in states of gastrointestinal hypomotility (Peeters 1993). Macrolides can be administrated intravenously or orally (Periti 1993).

Rifaximine is a macrocyclic drug derived from rifamycin (Yu 2013). Rifaximine interferes with bacterial replication binding to RNA polymerase and interfering in bacterial DNA transcription (Yu 2013). Because of its poor oral absorption, it is used for gastrointestinal infections (Yu 2013). The lack of systemic absorption diminishes systemic adverse effects. Recently, some studies report that rifaximin inhibits the activation of NF‐kB and reduces the expression of pro‐inflammatory cytokines, downregulating inflammatory response (Ponziani 2017).

How the intervention might work

The different antibiotics, due to their mechanisms of action, are expected to kill or assist in killing the bacteria in people with weakened immune system due to liver cirrhosis. This could have beneficial effects on the course of the infection and, more importantly, on the patients, and it could even assist in stopping the upper gastrointestinal bleeding.

Why it is important to do this review

If people with cirrhosis and variceal bleeding do not receive antibiotic prophylaxis, then 30% to 48% of them will develop a bacterial infection (Bernard 1995; Moon 2016; Mallet 2017). Prophylactic use of antibiotics during an episode of upper gastrointestinal bleeding in people with cirrhosis is considered the standard of care, and it is encouraged in European, American, and UK guidelines (Tipathi 2015; Garcia‐Tsao 2017; EASL 2018). However, these recommendations are based on evidence from a meta‐analysis by Bernard and colleagues, including only five randomised clinical trials, with few patients and few events (Bernard 1999).

The first Cochrane Review on antibiotic prophylaxis in people with cirrhosis was published in 2002, and it included eight trials comparing antibiotics versus placebo or no antibiotic in 864 randomised participants (Soares‐Weiser 2002). Clinically significant beneficial effects were found on mortality (risk ratio (RR) 0.73, 95% confidence interval (CI) 0.55 to 0.95), and in the incidence of bacterial infections (RR 0.40, 95% CI 0.32 to 0.51). The review was updated in 2010, and included 12 trials with 1241 participants (Chavez‐Tapia 2010). The updated review confirmed a decrease in all‐cause‐mortality (RR 0.79, 95% CI 0.63 to 0.98) and bacterial infections (RR 0.36, 95% CI 0.27 to 0.49), and also found a decrease in the proportion of re‐bleeding events (RR 0.53, 95% CI 0.38 to 0.74). No serious adverse events were reported. However, all the included trials were at high risk of bias and uncertainty was observed regarding missing data. Also, the cumulative Z curve did not reach the required information size and the boundaries for benefit, futility, or harm were not crossed in Trial Sequential Analysis (Chavez‐Tapia 2010).

A number of randomised clinical trials have been published since the last search for trials in June 2010 (Chavez‐Tapia 2010), and the additional data need to be implemented in this review update, for which we will also follow the most recent Cochrane methodology. As the review by Chavez‐Tapia 2010 encompasses two comparisons, the current review will only assess antibiotic prophylaxis versus no intervention or placebo in people with cirrhosis and variceal bleeding. This is why we will publish it as a separate, new review. We will also prepare another separate review with the remaining comparisons, in the Chavez‐Tapia 2010 review, on different types of antibiotics for prophylaxis in people with cirrhosis and variceal bleeding (Sanchez‐Jimenez 2018).