As Harold Conn told us many years ago, ‘nature abhors a vacuum’. Something always rushes in to fill the gap left by the absentee.
When we alter the flora of patients with cirrhosis, we must expect an alteration in the bacteria that eventually infect such patients. We saw this when we gave rats with cirrhosis oral norfloxacin (1). Their gut flora changed from the usual Gram-negative bacteria to a predominance of enterococci. The undiluted stool of treated rats was sterile on a Gram-negative selective agar plate – this astonished me. The stool of a coprophagic creature contained no Gram-negative bacteria – because of selective intestinal decontamination and replacement of Gram negatives with Gram positives. When these treated rats developed spontaneous bacterial peritonitis, it was exclusively because of Gram-positive cocci, usually Enterococci (1).
This same phenomenon has been reported in patients with cirrhosis (2). The flora of bacterial infections in the Liver Unit in Barcelona (where selective intestinal decontamination was introduced and popularized) changed from 70 to 80% Gram negative to 53% Gram positive (2). The authors attributed this change to norfloxacin and invasive procedures.
We have been warned that the liberal use of antibiotics over the past 70 years could lead to selection of bacteria for which we have ‘no reliable treatment’ (3). Meanwhile there is no interest among drug companies in developing new antibiotics. Drugs that are used short-term yield little profit compared with ‘block-buster’ drugs that patients take for the rest of their lives, such as statins.
The study reported in this issue of Liver International provides more evidence for this concern – selection of/and infection by Gram-positive flora in patients with liver disease (4). This prospective study specifically focused on S. aureus. Therefore we have no denominator to determine the prevalence of infections by this organism compared with other bacteria. However, in a series of patients with spontaneous bacterial peritonitis that I reported 20 years ago, there was not a single episode because of S. aureus (5).
As the authors of the current paper point out, cirrhosis is one of the most common immunodeficiency states worldwide. Bacterial infections are expected and may cause the deaths of up to 25% of patients with cirrhosis (6). This figure may be an underestimate because of the probability that many patients die at home before they have a chance to be admitted to a hospital. Bacterial infections regularly start a fatal multi-organ failure syndrome in these patients.
The authors of the current paper were infectious diseases specialists and molecular biologists. No hepatologist is listed among the authors. Therefore we have no information on severity of liver disease such as model for endstage liver disease score, presence or absence of ascites, etc. Chronic liver disease was associated with higher mortality; this is not surprising. ‘Primary bacteraemia’ was the most common infection reported. In view of the presence of indwelling vascular and bladder catheters in many patients, some readers may label these infections as episodes of ‘line sepsis’ or ‘catheter-related sepsis’ rather than ‘primary’ infections.
What messages can we ‘take home’ from this paper? I think we should limit the use and the duration of use of vascular and bladder catheters in patients with cirrhosis. I also think that we should limit use of selective intestinal decontamination to data-supported subgroups of patients with cirrhosis, i.e. those who have survived spontaneous bacterial peritonitis, those with transfusion-requiring gut hemorrhage, those with an ascitic fluid total protein <1 g/dl, and those who meet the complicated inclusion criteria for primary prophylaxis (7, 8). These criteria include: ascitic fluid total protein <1.5 g/dl and advanced liver failure (Child–Pugh score ≥9 points with serum bilirubin ≥3 mg/dl) OR impaired renal function serum creatinine ≥1.2 mg/dl, or blood urea nitrogen ≥25 mg/dl or serum sodium <130 mEq/L) (8).