Patients and methods. 12 studies including 2132 cirrhotic admitted in ICU were analyzed after selection of original articles and response to a standardized questionnaire by the corresponding authors. The prognostic performance of 177 variables (including reason for admission, organ check details replacement therapy, and composite prognostic scores)were analyzed for each period according to the method
of Der Simonian and Laird (708 pooled analysis). Results. In-ICU, in-hospital, 3 and 6 month-survival were 55% (range 28-66%), 45% (27-61%), 23% (13-37%) and 20% (13-35%), respectively. In-ICU survival was better in recent studies (>yr2000) (OR=1.36,p=0.036), in centers containing a liver transplant program (OR=1.82,p<0.0001) in patients admitted for variceal bleeding (OR=1.80, p<0.0001, PPV=0.67), with MELD<13 (OR=4.31, p<0.0001, PPV=0.86), albumin>35g/L (OR=3.97, p<0.0001, PPV=0.77), INCB024360 clinical trial <2 organ failures (OR=4.81, p<0.0001, VPP=0.70). In-ICU mortality was significantly associated with
23 variables and better predicted by: a CLIF-SOFA>22 (OR=5.94, p=0.005, PPV=1); >5 organ failures (OR=10.87, p<0.0001, PPV=0.98); SOFA>19 (OR=14.46, p<0.0001, PPV=0.97), a fungemia (OR=4.61, p=0.0005,PPV=0.87), ARDS (OR=4.48,p<0.0001,PPV=0.81), refractory oli- guria (OR=9.17, p<0.0001, VPP=0.79), a MELD>35 (OR=5.43, p<0.0001,PPV=0.77), this website sepsis-induced hypotension (OR=5.75, p<0.0001, PPV=0.77), an increased SOFA at d3 (OR=4.57,p<0.0001,PPV=0.72), positive
blood cultures (OR=2.15, p=0.004, PPV=0.73), infection with GN Bacilii (OR=2.24,p<0.0001,PPV=0.70),and the use of MARS (OR=2.04,p=0.0081,PPV=0.64). The mSOFA, APACHE, alcohol consumption, direct admission in ICU, HRS, nosocomial infection or infection with GP cocci had no impact. The results were heterogeneous for the Pugh, creatinine, the use of intubation or norepinephrine. SBP was associated only with in-hospital mortality. Patients who received TIPS had better in-hospital, 3 and 6 month survival. The Pugh, MELD, SOFA, the presence of SIRS, bacterial infection or ARDS, the need for haemodi-alysis kept an impact on 3 and 6 month survival. Conclusion The prognostic performance of general ICU scores decreases over the long-term, unlike the Pugh and MELD scores. Some events can be considered alone and have an excellent predictive value for short-term and long-term prognosis, as well as composite scores. Disclosures: Constantine J.