Hyperbilirubinemia at hospitalization predicts nosocomial infection in decompensated cirrhosis: Data from ATTIRE...

To identify clinical characteristics and serological biomarkers that predicted subsequent nosocomial infection in ATTIRE trial patients.

We identified 360 patients at hospitalization without infection and not prescribed antibiotics and compared clinical characteristics between those who subsequently developed a nosocomial infection and not. In a 68-patient subcohort, we compared plasma biomarkers of bacterial translocation, infection, and inflammation at hospitalization between those who developed a nosocomial infection and not. In a 56-patient subcohort, we investigated plasma lipidomic profiles in those who did and did not develop nosocomial infection using Lipotype Shotgun platform analysis and multivariate statistical techniques. To further investigate lipid pathways, we compared outcomes in patients taking statins or not at hospitalization.

Serum bilirubin >188 µmol/L at hospitalization predicted subsequent nosocomial infection in univariate and multivariate analyses, with 80% specificity. The most common nosocomial infections were respiratory tract (29%) and those developing infection had significantly greater 28 and 90-day mortality than those not (p=9.34E-05 and 0.014). Serological biomarkers of bacterial translocation, infection, and inflammation did not predict subsequent infection. Partial least squares discriminatory analyses identified cholesterol esters (CEs) (CE.18.1.2, CE.18.1.0, and CE.16.0.0) as important predictors of infection but provided only a small improvement in predictive ability over bilirubin alone. RNA-sequencing analyses suggest this is mediated by a downregulation of the cellular cholesterol esterification enzyme sterol O-acyltransferase 1. Statin use was not associated with nosocomial infection prevention.

In ATTIRE, elevated serum bilirubin at hospitalization was the only clinical characteristic that predicted subsequent development of nosocomial infection. Considering the rising incidence of antimicrobial resistance, these data could be used to limit antibiotic prophylaxis or aid trial design for investigating use in high-risk patients.