Clarifying the role of C-reactive protein as a bacterial infection predictor in decompensated cirrhosis.

INTRODUCTION: Bacterial infections are frequent in cirrhosis and may induce other deleterious complications. Ultrasensitive C-reactive protein (US-CRP), like other acute-phase proteins, is often considered useful in predicting bacterial infection in decompensated cirrhosis. However, US-CRP's reliability remains inconclusive, as inflammation in cirrhosis causes US-CRP synthesis independently of infection. The aim of this study was to clarify US-CRP's role as an infection predictor in decompensated cirrhosis.

PATIENTS AND METHODS: This was a prospective single-center study with systematic inclusion of cirrhotic patients admitted because of decompensation.

RESULTS: A total of 118 patients were enrolled, of whom 47 (39.8%) had an overt infection, defined by clinical and laboratory/imaging criteria. Within those, 17 had infection confirmed by culture bacterial identification. Escherichia coli was the most frequent isolated bacteria. Seventeen patients had spontaneous bacterial peritonitis, but only four (23.5%) had positive ascitic fluid cultures. US-CRP levels were significantly higher in cases of overt infection and positive culture groups than the no infection group (median: 4.14 and 6.40 vs. 1.11 mg/dl, P<0.0001 for both). When considering both overt infection and positive culture groups, the US-CRP values of area under the curve as an infection predictor were, respectively, 0.824 and 0.870, P<0.0001 for both, with associated cutoff values of 2.40 and 3.92 mg/dl, and sensitivity and specificity of 78.7/74.6 and 82.4/79.2%, respectively.

CONCLUSION: The ideal US-CRP infection confirmatory cutoff is probably situated between 2.40 and 3.92 mg/dl. However, as infection is somewhat concealed and hazardous in cirrhotic patients, if not considered with lower US-CRP levels according to specific clinic scenarios, it should be carefully considered, at least, if US-CRP is greater than 2.40 mg/dl (0.5 mg/dl normal upper cutoff).