ORIGINAL ARTICLE. See Editorial on Page 384

Transcription

1 LIVER TRANSPLANTATION 18: , 2012 ORIGINAL ARTICLE Comparison of the Sequential Organ Failure Assessment Score With the King s College Hospital Criteria and the Model for End-Stage Liver Disease Score for the Prognosis of Acetaminophen-Induced Acute Liver Failure Evangelos Cholongitas, 1,2 Eleni Theocharidou, 2 Panayota Vasianopoulou, 2 Alex Betrosian, 2 Steve Shaw, 3 David Patch, 2 James O Beirne, 3 Banwari Agarwal, 3 and Andrew K. Burroughs 2 1 Fourth Department of Internal Medicine, Medical School of Aristotle University, Hippokration General Hospital of Thessaloniki, Thessaloniki, Greece; and 2 Liver Transplantation and Hepatobiliary Unit and 3 Department of Intensive Care, Royal Free Hospital, London, United Kingdom Acetaminophen-induced acute liver failure (ALF) is a complex multiorgan illness. An assessment of the prognosis is essential for the accurate identification of patients for whom survival without liver transplantation (LT) is unlikely. The aims of this study were the comparison of prognostic models [King s College Hospital (KCH), Model for End-Stage Liver Disease, Sequential Organ Failure Assessment (SOFA), and Acute Physiology and Chronic Health Evaluation II (APACHE II)] and the identification of independent prognostic indicators of outcome. We evaluated consecutive patients with severe acetaminophen-induced ALF who were admitted to the intensive care unit. At admission, demographic, clinical, and laboratory parameters were recorded. The discriminative ability of each prognostic score at the baseline was evaluated with the area under the receiver operating characteristic curve (AUC). In addition, using a multiple logistic regression, we assessed independent factors associated with outcome. In all, 125 consecutive patients with acetaminophen-induced ALF were evaluated: 67 patients (54%) survived with conservative medical management (group 1), and 58 patients (46%) either died without LT (28%) or underwent LT (18%; group 2). Group 1 patients had significantly lower median APACHE II (10 versus 14) and SOFA scores (9 versus 12) than group 2 patients (P < 0.001). The independent indicators associated with death or LT were a longer prothrombin time (P ¼ 0.007), the inspiratory oxygen concentration (P ¼ 0.005), and the lactate level at 12 hours (P < 0.001). The KCH criteria had the highest specificity (83%) but the lowest sensitivity (47%), and the SOFA score had the best discriminative ability (AUC ¼ 0.79). In conclusion, for patients with acetaminophen-induced ALF, the SOFA score performed better than the other prognostic scores, and this reflected the presence of multiorgan dysfunction. A further evaluation of SOFA with the KCH criteria is warranted. Liver Transpl 18: , VC 2011 AASLD. Received September 14, 2011; accepted December 14, See Editorial on Page 384 Acute liver failure (ALF) is a complex multisystem illness that results after a catastrophic insult to the liver and manifests itself in the development of coagulopathy and encephalopathy within a short period of time. 1 Acetaminophen poisoning has become the most common cause of ALF in many countries. 1,2 Although many patients will recover fully with Additional Supporting Information may be found in the online version of this article. Abbreviations: ALF, acute liver failure; APACHE II, Acute Physiology and Chronic Health Evaluation II; AUC, area under the receiver operating characteristic curve; ICU, intensive care unit; INR, international normalized ratio; KCH, King s College Hospital; LT, liver transplantation; MELD, Model for End-Stage Liver Disease; NPV, negative predictive value; PPV, positive predictive value; RFH, Royal Free Hospital; SOFA, Sequential Organ Failure Assessment. Address reprint requests to Andrew K. Burroughs, F.R.C.P., F.Med.Sci., Liver Transplantation and Hepatobiliary Unit, Royal Free Hospital, Pond Street, Hampstead, London, United Kingdom NW3 2QG. Telephone: þ ; FAX: þ ; andrew.burroughs@royalfree.nhs.uk DOI /lt View this article online at wileyonlinelibrary.com. LIVER TRANSPLANTATION.DOI /lt. Published on behalf of the American Association for the Study of Liver Diseases VC 2011 American Association for the Study of Liver Diseases.

2 406 CHOLONGITAS ET AL. LIVER TRANSPLANTATION, April 2012 appropriate medical management, mortality will be high in some cases unless liver transplantation (LT) is performed. 3 In fact, LT has significantly improved the survival of these patients and plays a central role in the modern management of ALF, which now accounts for 5% to 12% of all LT activity. 4 The determination of the prognosis for the accurate prediction of the outcome of ALF has immense value for determining the optimal management method. It is vital that irreversible ALF be recognized early so that life-threatening complications can be prevented. A variety of selection criteria are in use, but the King s College Hospital (KCH) criteria are the most widely used, and they have gained considerable acceptance. 5 The KCH criteria include clinical and biochemical data such as the age, duration of jaundice, bilirubin level, prothrombin time, arterial ph, and serum creatinine level, which are routinely available in clinical practice, and the etiology is taken into account (ie, acetaminophen versus non-acetaminophen). These parameters are available within a few hours of admission, and they help with early listing and referral to a specialist center. The Model for End-Stage Liver Disease (MELD) is a survival model based on a composite of 3 laboratory variables: the serum creatinine level, the serum bilirubin level, and the international normalized ratio (INR). 6 This model has been validated in several independent cohorts of patients with cirrhosis, and it has been adopted for donor liver allocation systems in the United States and more recently in several European, Asian, and South American countries. Only a few studies have evaluated the prognostic performance of MELD for patients with cirrhosis who are admitted to the intensive care unit (ICU), 7 but lately there has been a growing interest in its useasaprognosticmarkerforpatientswithalf. 8 The Acute Physiology and Chronic Health Evaluation II (APACHE II) score 9 and the Sequential Organ Failure Assessment (SOFA) score 10 are ICU-specific prognostic scores developed from general ICU populations. The SOFA score is used for grading organ dysfunction or failing organ systems (respiratory, hepatic, coagulation, cardiovascular, neurological, and renal systems). Each system is graded on a 0- to 4-point scale, and the total SOFA score is composed of the separate subscores (range ¼ 0-24; see the supporting information). Both the APACHE II score and the SOFA score have been validated in patients with cirrhosis and acute deterioration. 7 The APACHE II score has been evaluated in patients with ALF, and it might be helpful in identifying nonsurvivors among patients who do not fulfill the KCH criteria. 11 The SOFA score has been evaluated in only 1 study of patients with acetaminophen-induced ALF. 12 However, no study has compared the KCH criteria and the MELD score against the APACHE II and SOFA scores. The aims of this study were the identification of the factors predicting mortality in patients with acetaminophen-induced ALF and the comparison of the prognostic performances of the KCH criteria, the MELD score, the APACHE II score, and the SOFA score. PATIENTS AND METHODS Consecutive adult patients with acetaminopheninduced ALF who were admitted to the ICU of Royal Free Hospital (RFH) between January 1, 1993 and December 31, 2010 were evaluated. ALF was defined as the acute onset of coagulopathy and hepatic encephalopathy within 8 weeks of the initial symptoms in patients with no previous history of liver disease. The criteria for listing were based on the KCH criteria as long as the patient did not have significant contraindications for LT (ie, uncontrolled sepsis, cerebral edema, and rapidly escalating doses of inotropes or uncontrolled bleeding). All patients received medical therapy with N-acetylcysteine, intravenous antibiotics and antifungals, early hemofiltration, and intubation (for grade 2 or 3 encephalopathy). Liver support devices were not used. Fresh frozen plasma was not used routinely for the correction of clotting factors and was administered only for bleeding or before intracranial pressure bolt insertion. The following demographic and clinical variables were prospectively recorded for each patient on admission to the ICU: age, sex, transfer from the general ward of RFH or another hospital, and length of stay in the general ward of RFH or another hospital before ICU admission. On admission to the ICU, the following variables were recorded: white blood count, platelet count, creatinine level, urea level, sodium level, potassium level, phosphate level, inspiratory oxygen concentration, arterial blood gas [ph and partial arterial pressures of oxygen and carbon dioxide] and acid-base values, arterial blood lactate level, albumin level, bilirubin level, and clotting profile (prothrombin time, INR, and activated partial thromboplastin time). The arterial lactate level was also recorded 4, 8, and 12 hours after the ICU admission. During the ICU stay, gastrointestinal bleeding episodes, the development of aspiration pneumonia, and the additional use of inotropes, mechanical ventilation, or hemofiltration were also recorded. The severity of liver disease was evaluated with the KCH criteria and the MELD score. APACHE II was used for the classification of illness severity, and the SOFA score was used for grading organ dysfunction or failing organ systems. The presence of a failing organ system on admission to the ICU was defined as a SOFA score of 3 or more points for any individual organ. All these scores were calculated as published and were evaluated soon after admission to the ICU after adequate volume resuscitation according to central venous pressure measurements. 5,6,9,10 Finally, the length of stay in the ICU and the general ward after discharge from the ICU was also recorded. The study protocol was approved by the RFH institutional committee. Statistical Analysis Univariate comparisons of demographic and baseline clinical factors between patients who died in the ICU or underwent transplantation and patients who remained alive were performed with Mann-Whitney U tests for continuous variables and with chi-square tests

3 LIVER TRANSPLANTATION, Vol. 18, No. 4, 2012 CHOLONGITAS ET AL. 407 for categorical variables. In order to identify ICU admission factors that were independently associated with mortality, a multivariate logistic regression analysis was performed. The discrimination ability of all published models to predict the outcomes of patients with ALF was evaluated with the area under the receiver operating characteristic curve (AUC). With the AUC, the true-positive and false-positive rates are shown on the vertical and horizontal axes, respectively. In this type of analysis, a model with an AUC between 0.7 and 0.8 is considered clinically useful, and a model with an AUC between 0.8 and 0.9 has excellent diagnostic accuracy. As the AUC approaches 1.0, the model approaches 100% sensitivity and specificity. 13 To test the calibration (ie, the degree of correspondence between the predicted and observed mortality rates), we used the Lemeshow- Hosmer test for goodness-of-fit testing. A high P value in this test (close to 1.0) is considered a sign of good calibration. 14 The Youden index (sensitivity þ specificity 1) 15 was used to select the best cutoff point for each prognostic model for calculating the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) and for correctly classifying patients. A P value 0.05 was considered statistically significant. RESULTS In all, 125 consecutive patients with severe acetaminophen-induced ALF were admitted to the ICU during the study period: 52 patients (41%) were men, and the mean age was years. Sixty-seven patients (54%) were transferred from general wards of RFH, and the remaining 46% were sent directly to the ICU from other hospitals. The median time from ingestion to the initiation of N-acetylcysteine treatment was 25 hours (range ¼ hours). Thirty-six patients (29%) fulfilled the KCH criteria. Thirty-six patients were listed for LT. The median waiting time for LT was 1 day (range ¼ 0-11 days). Thirteen of the 36 patients (36%) did not receive a liver graft and died on the waiting list (6 were removed from the waiting list because of disease progression). The median APACHE II, MELD, and SOFA scores on admission to the ICU were 12 (range ¼ 1-38), 36 (range ¼ 12-40), and 11 (range ¼ 2-19), respectively. On the first day of admission or during hospitalization, 15% of the patients had or developed aspiration pneumonia. Mechanical ventilation and cardiovascular support with inotropes were used for most of the patients (70% and 63%, respectively). The median ICU stay was 6 days (range ¼ days). During the ICU stay, sepsis and upper gastrointestinal bleeding developed in 52% and 6.5% of the patients, respectively. Renal support with hemofiltration was necessary for 58% of the patients during their ICU stay. Factors Associated With the Outcome: Univariate and Multivariate Analyses In our cohort, 67 patients (54%) survived with conservative medical management (group 1), and 58 patients (46%) either died without LT (35 or 28%) or underwent LT (23 or 18%; group 2). The causes of death were multiorgan failure (74%), sepsis (17%), cardiac/respiratory failure (4%), and other (5%). With respect to the length of stay in the ICU, group 1 (median ¼ 5 days, range ¼ 2-47 days) and group 2 (median ¼ 8 days, range ¼ days) did not significantly differ (P ¼ 0.10). Interestingly, there were no significant differences in survival over the years at our center. In particular, during the 3 time periods, the survival rates were similar: 55% (23/42) for , 55% (23/43) for , and 52% (21/40) for (P > 0.05). The KCH criteria were met less frequently in group 1 (9 patients or 13%) versus group 2 [27 patients (20diedwithoutLT,and7underwentLT)or46%,P < 0.001]. Group 1 patients had significantly lower median APACHE II (10 versus 14) and SOFA scores (9 versus 12) in comparison with group 2 patients (P < 0.001), whereas the median MELD scores were marginally significantly different between these 2 groups (35 versus 37, respectively, P ¼ 0.047). When each of the SOFA subscores (see the supporting information) was examined individually, the difference between groups 1 and group 2 was seen to arise from significant differences in the median cardiovascular (1 versus 3, P < 0.001), respiratory (1 versus 2, P < 0.001), and neurological subscores (1 versus 1.5, P ¼ 0.005). On admission, group 1 patients and group 2 patients had no significant differences with respect to the severity of encephalopathy. However, group 1 patients had less aspiration pneumonia than group 2 patients [4/67 (5.9%) versus 15/58 (26%), P ¼ 0.004], and they needed less cardiovascular support with inotropes [28/67 (42%) versus 51/58 (88%), P < 0.001], mechanical ventilation [34/67 (51%) versus 54/58 (93%), P < 0.001], and renal hemofiltration [31/67 (46%) versus 42/58 (72%), P ¼ 0.003] during the ICU stay (Table 1). As for the on-admission laboratory variables, group 1 patients had, in comparison with group 2 patients, significantly lower prothrombin times [62 seconds (range ¼ seconds) versus 77 seconds (range ¼ seconds), P ¼ 0.01], partial thromboplastin times [69 seconds (range ¼ seconds) versus 109 seconds (range ¼ seconds), P < 0.001], INRs ( versus , P ¼ 0.02), and arterial lactate levels ( versus mmol/l at the baseline, P ¼ 0.01; versus mmol/l 4 hours after ICU admission, P ¼ 0.02; versus mmol/l 8 hours after ICU admission, P ¼ 0.003; and versus mmol/l 12 hours after ICU admission, P < 0.001). However, group 1 patients needed lower inspiratory oxygen concentrations (0.40% % versus 0.50% %, P < 0.001). Finally, group 1 patients had significantly higher arterial ph values on admission ( versus , P ¼ 0.005; Table 1). Factors Associated With the Outcome: A Multivariate Logistic Regression Analysis Three variables were independently associated with outcome (survival versus death or LT): the

4 408 CHOLONGITAS ET AL. LIVER TRANSPLANTATION, April 2012 TABLE 1. Clinical and Laboratory Characteristics of Patients With Acetaminophen-Induced ALF on Admission to the ICU Variable Group 1 (n ¼ 67 or 54%) Group 2 (n ¼ 58 or 46%) P Value Age (years)* Male sex [n (%)] 25 (37) 27 (47) 0.29 Transfer from RFH [n (%)] 38 (57) 29 (50) 0.56 ICU stay (days) 5 (2-47) 8 (1-103) 0.10 On admission to ICU [n (%)] Inotropic support 26 (39) 36 (62) Mechanical ventilation 29 (43) 49 (84) <0.001 During ICU stay [n (%)] Aspiration pneumonia 4 (6) 15 (26) Gastrointestinal bleeding 2 (3) 6 (10) Mechanical ventilation 34 (51) 54 (93) <0.001 Inotropic support 28 (42) 51 (88) <0.001 Hemofiltration 31 (46) 42 (72) Albumin (g/l)* White blood count (10 9 /L) 13 (0.4-83) 14 (3-38) 0.15 Platelet count (10 9 /L) 103 (29-340) 88 (14-240) 0.09 Bilirubin (lmol/l) 94 (9-280) 143 (17-750) 0.11 Prothrombin time (seconds) 62 (24-120) 77 (40-140) 0.01 Activated partial thromboplastin 69 (35-150) 109 (55-250) <0.001 time (seconds) INR* Creatinine (mg/dl) 2.5 ( ) 2.6 ( ) 0.12 Urea (mmol/l) 17 (12-25) 18 (12-28) 0.82 Sodium (mmol/l)* Potassium (mmol/l)* Phosphate (mmol/l) 0.7 ( ) 0.8 ( ) 0.7 Lactate (mmol/l)* Baseline hours hours hours <0.001 Inspiratory oxygen concentration (%)* <0.001 ph* NOTE: Group 1 consisted of patients who survived with conservative medical management, and group 2 consisted of patients who died without LT or underwent LT. *The data are presented as means and standard deviations. The data are presented as medians and ranges. prothrombin time (odds ratio ¼ 1.025, 95% confidence interval ¼ , P ¼ 0.007), the inspiratory oxygen concentration (odds ratio ¼ 23.75, 95% confidence interval ¼ , P ¼ 0.005), and the lactate level at 12 hours (odds ratio ¼ 1.96, 95% confidence interval ¼ , P < 0.001). Prognostic Factors Associated With Mortality: Receiver Operating Characteristics and Calibration Table 2 shows the sensitivity, specificity, PPV, NPV and diagnostic accuracy at the best cutoff points for the entire cohort. According to the AUC values, the SOFA score had the best discriminative accuracy for outcome (AUC ¼ 0.79); it was followed by the APACHE II score (AUC ¼ 0.72), but the difference was not significant (P ¼ 0.23; Fig. 1). As for the goodness of fit (as measured by the Lemeshow-Hosmer test), the calibration of SOFA (v 2 ¼ 5.1, P ¼ 0.73) was superior to the calibration of APACHE II (v 2 ¼ 10.2, P ¼ 0.23) and the KCH criteria (v 2 ¼ 3.9, P ¼ 0.12). Seventytwo patients (58%) were transferred directly (ie, <24 hours) from general wards to the ICU; 35 patients were transferred after they had stayed more than 24 hours in general wards, and data were not available for 18 patients. For these 3 groups of patients, the SOFA score always had superior discriminative accuracy (AUC ¼ 0.76, AUC ¼ 0.78, and AUC ¼ 0.83, respectively) in comparison with the other prognostic scores (eg, AUC ¼ 0.66, AUC ¼ 0.72, and AUC ¼ 0.80, respectively, for APACHE II). Interestingly, only cardiovascular and respiratory dysfunction or failure, as defined by the SOFA score (see the supporting information), had similar discriminative ability (AUC ¼ 0.75 and AUC ¼ 0.71, respectively). In addition, the presence of cardiovascular failure (ie, a

5 LIVER TRANSPLANTATION, Vol. 18, No. 4, 2012 CHOLONGITAS ET AL. 409 TABLE 2. Prediction of Outcomes for 125 Consecutive Patients With Acetaminophen-Induced ALF Who Survived (n 5 67) or Died Without LT or Underwent LT (n 5 58) AUC Cutoff Point Sensitivity (%) Specificity (%) PPV NPV KCH APACHE II SOFA MELD Lactate (mmol/l) NOTE: The KCH, MELD, APACHE II, and SOFA scores were measured on admission; the lactate level was measured 12 hours after admission. SOFA score 3 points for this system) showed very good performance (AUC ¼ 0.75). Table 3 shows the sensitivity, specificity, PPV, NPV, and diagnostic accuracy at the best cutoff points for patients who survived (n ¼ 67) and patients who died without LT (n ¼ 35). According to the AUCs, the SOFA score had the best discriminative accuracy for mortality (AUC ¼ 0.84), and it was followed by the APACHE II score (AUC ¼ 0.76); the AUC for the MELD score was only The SOFA score had the highest PPV (0.84), but the KCH criteria had the highest specificity (86%); the MELD score had the best NPV (0.83). For the goodness of fit measured by the Lemeshow- Hosmer test, the calibration of SOFA (v 2 ¼ 4.8, P ¼ 0.67) was superior to the calibration of the KCH criteria (v 2 ¼ 2.7, P ¼ 0.10). Finally, in the subgroup of patients listed for LT (n ¼ 36), the SOFA score had the highest PPV (0.95) and the best discriminative Figure 1. AUCs for SOFA, MELD, APACHE II, and KCH for 125 consecutive patients with acetaminophen-induced ALF who were admitted to the ICU. ability (AUC ¼ 0.79), and it was followed by the APACHE II score (AUC ¼ 0.67), the KCH criteria (AUC ¼ 0.66), and the MELD score (AUC ¼ 0.62). Arterial Lactate and Prognostic Scores Table 3 also shows the prognostic performance of the lactate level on its own at 12 hours in the group of patients who either survived or died without LT (n ¼ 102). The lactate level at 12 hours had very good discriminative ability (AUC ¼ 0.80), excellent sensitivity (98%), and an excellent NPV (0.95). In this group of patients, the SOFA score had the best correlation with the lactate level at 12 hours [r 2 (Spearman) ¼ 0.42, P < 0.001] in comparison with the lactate level at other time points (the baseline and 4 and 8 hours after admission). In addition, the lactate level at 12 hours had the best correlation with SOFA in comparison with the other prognostic scores, and it correlated significantly with the cardiovascular subscore (r 2 ¼ 0.39, P < 0.001) and the respiratory subscore (r 2 ¼ 0.35, P ¼ 0.001) but not with any of the other SOFA subscores. The results were similar when the entire cohort of patients (n ¼ 125) was evaluated. In the group of patients who either survived or died without LT (n ¼ 102), the discriminative accuracy of the KCH criteria was not significantly improved when they were combined with the arterial lactate level at 12 hours [with the concentration of 4.7 mmol/l used as the cutoff point (see Table 3); AUC for the KCH criteria and a 12-hour lactate level > 4.7 mmol/l ¼ 0.73]. Similarly, on the basis of the regression coefficients of the MELD score and the lactate level at 12 hours, a new prognostic score was derived, but its discriminative ability (AUC ¼ 0.65) was not superior to that of the standard MELD score (AUC ¼ 0.61, P > 0.05). Lastly, the arterial lactate level at 12 hours was graded on a 0- to 4-point scale (<1.6, , , and >4.5 mmol/l, respectively), which was based on its concentration distribution (median ¼ 3.2 mmol/l, 25th percentile ¼ 1.6 mmol/l, 75th percentile ¼ 4.5 mmol/l). A new SOFA-lactate score was derived through the incorporation of the arterial lactate level at 12 hours into the SOFA score, and it ranged from 0 to 28 points. However, this new prognostic score

6 410 CHOLONGITAS ET AL. LIVER TRANSPLANTATION, April 2012 TABLE 3. Prediction of Outcomes for 102 Consecutive Patients With Acetaminophen-Induced ALF Who Survived (n 5 67) or Died Without LT (n 5 35) AUC Cutoff Point Sensitivity (%) Specificity (%) PPV NPV KCH APACHE II SOFA MELD Lactate (mmol/l) NOTE: The KCH, MELD, APACHE II, and SOFA scores were measured on admission; the lactate level was measured 12 hours after admission. (AUC ¼ 0.85) was not superior to the simple SOFA score (AUC ¼ 0.84, P > 0.05). DISCUSSION Acetaminophen is the single most common cause of drug-induced liver injury. 1 The majority of acetaminophen overdose cases are uncomplicated, but a small proportion of patients (approximately 2%), will develop evidence of ALF. 16 This group of patients remains a major clinical challenge with respect to prognosis. Among the several sets of selection criteria for ALF, the KCH criteria have gained considerable acceptance. However, the KCH criteria are not optimal because they have limited sensitivity and provide a late identification of the patients who require LT. 16 Indeed, a recent meta-analysis showed that the KCH criteria had excellent specificity (94.6%) but low sensitivity (58.2%), and this raises further questions about changes or improvements to the KCH criteria. 17 Our study confirmed these findings: the KCH criteria had the best specificity, but this occurred at the cost of sensitivity. Indeed, although the application of the KCH criteria solely on the first day of admission (as in this study) may have affected their performance, they had the lowest sensitivity and a limited ability to discriminate between patients who survived with conservative medical management and patients who died without LT (AUC ¼ 0.71; Table 3). It should be mentioned that apart from discriminative accuracy, both the PPV and the NPV of a prognostic score are important determinants of its clinical usefulness. For example, in our cohort, we found that the KCH criteria had the lowest NPV, and the MELD score had the highest NPV; this indicates that the former has greater applicability for predicting death rather than spontaneous survival. It has, therefore, been interesting to evaluate more accurate prognostic markers that could improve the sensitivity and reduce the mortality rate for patients failing to meet the KCH criteria. ALF is considered a systematic disease with multiorgan system dysfunction; thus, multiorgan failure scores such as the APACHE II and SOFA scores might be applicable and appropriate as early prognostic markers for patients with ALF. Indeed, the APACHE II score has been suggested as a supplement for the KCH criteria. 11 Similarly to Larson et al. 2 (who studied 22 tertiary care centers in the United States), we found the APACHE II score to have higher sensitivity than the KCH criteria. The SOFA score, which was previously used for patients with acute deterioration from chronic liver disease, 7,18 was investigated in 1 study of 101 patients with acetaminophen-induced ALF, 12 and a second study 19 evaluated the SOFA score in 87 patients with ALF [only 8 of these patients (9%) had acetaminophen-induced ALF]. However, the SOFA score had not been evaluated in comparison with the APACHE II score, the KCH criteria, and the MELD score. In addition, in our ALF cohort, the SOFA score had the best discriminative ability in comparison with the other prognostic scores (APACHE II, KCH, and MELD; Fig. 1). The predictive accuracy of the SOFA score likely reflects multiorgan dysfunction, which is predominant in patients with ALF, 20 and it is much simpler than APACHE II. This is supported by the fact that among the SOFA components, the cardiovascular (AUC ¼ 0.78) and respiratory components (AUC ¼ 0.75) had very good discriminative ability, which was significantly better than that of the other SOFA components (eg, the liver or neurological system). Interestingly, the SOFA score had similar discriminative accuracy for the 72 patients admitted from general wards to the ICU in less than 24 hours (AUC ¼ 0.76) and for the 35 patients who had a later transfer (AUC ¼ 0.78, P ¼ 0.32), but its performance was always superior to that of the other prognostic scores. The prognostic performance of the MELD score for patients with ALF has been evaluated in a few recent studies. Kremers et al. 21 were the first to show that the MELD score may be useful in prioritizing nonacetaminophen ALF candidates for liver allocation. For patients with non-acetaminophen ALF, a recent study found that the MELD score had excellent prognostic performance (AUC ¼ 0.96) in the setting of ALF, 22 but this finding has not been confirmed in other studies. 23 For patients with acetaminopheninduced ALF, Rossaro et al. 24 found that a MELD score < 30 had the greatest applicability for predicting spontaneous survival, whereas a MELD score 30 did not accurately predict death or LT (NPV ¼ 82%, PPV ¼ 52%). In our acetaminophen-induced ALF

7 LIVER TRANSPLANTATION, Vol. 18, No. 4, 2012 CHOLONGITAS ET AL. 411 cohort, we confirmed the latter findings because the MELD score had the highest sensitivity and the highest NPV, but it had the lowest PPV and the lowest discriminative ability (Tables 2 and 3). Although the precise mechanisms of hyperlactatemia in severe liver dysfunction have not been fully determined, 25 the serum lactate level can be used to predict mortality in patients with acetaminopheninduced ALF, 26 and its prognostic impact for patients with ALF has been recognized for several years. 27 In fact, hyperlactatemia may reflect not only a severely reduced hepatic clearance of lactate but also an overproduction of lactate in the systemic tissues. The latter could explain the finding that arterial lactate had a stronger association with the SOFA score versus the other prognostic scores and particularly with the dysfunction or failure of the cardiovascular and respiratory systems; both can lead to systemic tissue hypoxia. Thus, our results confirm those of Schmidt and Larsen, 12 who found that lactate was significantly associated with the SOFA score. However, our study is the first in which the correlation of lactate with other prognostic scores has been assessed. In addition, it is the first in which serial lactate measurements have been evaluated as prognostic markers for patients with severe acetaminophen-induced ALF. In our cohort, the serum lactate level 12 hours after admission had the best correlation with the SOFA score. In addition, similarly to the study of 83 ALF patients in Birmingham, 27 the lactate level at 12 hours was independently associated with the outcome, and it remained an independent risk factor in a logistic regression analysis that included the KCH criteria. Interestingly, the discriminative accuracy of the KCH criteria was not significantly improved when they were combined with the arterial lactate level at 12 hours. The latter had better discriminative ability than the KCH criteria. Thus, arterial lactate levels that are persistently elevated despite adequate fluid resuscitation (based on the central venous pressure) seem to be associated with a poor prognosis. On the basis of these findings, we believe that serum lactate is clearly a useful prognostic marker, and further work on clinically useful cutoff values and their validation will improve the management of ALF. 20 Finally, the serum creatinine levels did not significantly differ between the 2 groups, and this was also true when we compared the patients who died and the patients who survived without LT (2.5 versus 3.1 mg/dl, P ¼ 0.07). However, the need for dialysis during the ICU stay was significantly different between the 2 groups, but it was not an independent factor in a multivariate analysis. In conclusion, our study has confirmed that for patients with severe acetaminophen-induced ALF, the SOFA score performs better than the other prognostic scores, and this reflects the presence of multiorgan dysfunction. The further evaluation of the SOFA score is warranted in high-quality, prospective studies between several large centers. In addition, serum lactate after adequate fluid resuscitation could be a useful prognostic marker, and further work on clinically useful cutoff values and their validation would improve the management of ALF. REFERENCES 1. Bernal W, Auzinger G, Dhawan A, Wendon J. Acute liver failure. Lancet 2010;376: Larson AM, Polson J, Fontana RJ, Davern TJ, Lalani E, Hynan LS, et al.; for Acute Liver Failure Study Group. Acetaminophen-induced acute liver failure: results of a United States multicenter, prospective study. Hepatology 2005;42: Brandsaeter B, H ockerstedt K, Friman S, Ericzon BG, Kirkegaard P, Isoniemi H, et al. Fulminant hepatic failure: outcome after listing for highly urgent liver transplantation 12 years experience in the Nordic countries. Liver Transpl 2002;8: O Grady JG. Acute liver failure. Postgrad Med J 2005;81: O Grady JG, Alexander GJ, Hayllar KM, Williams R. Early indicators of prognosis in fulminant hepatic failure. Gastroenterology 1989;97: Wiesner R, Edwards E, Freeman R, Harper A, Kim R, Kamath P, et al.; for United Network for Organ Sharing Liver Disease Severity Score Committee. Model for End- Stage Liver Disease (MELD) and allocation of donor livers. Gastroenterology 2003;124: Cholongitas E, Senzolo M, Patch D, Kwong K, Nikolopoulou V, Leandro G, et al. Risk factors, Sequential Organ Failure Assessment and Model for End-Stage Liver Disease scores for predicting short term mortality in cirrhotic patients admitted to intensive care unit. Aliment Pharmacol Ther 2006;23: Schmidt LE, Larsen FS. MELD score as a predictor of liver failure and death in patients with acetaminopheninduced liver injury. Hepatology 2007;45: Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med 1985;13: Vincent JL, Moreno R, Takala J, Willatts S, De Mendonça A, Bruining H, et al.; for Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. The SOFA (Sepsis-Related Organ Failure Assessment) score to describe organ dysfunction/failure. Intensive Care Med 1996;22: Mitchell I, Bihari D, Chang R, Wendon J, Williams R. Earlier identification of patients at risk from acetaminophen-induced acute liver failure. Crit Care Med 1998;26: Schmidt LE, Larsen FS. Prognostic implications of hyperlactatemia, multiple organ failure, and systemic inflammatory response syndrome in patients with acetaminophen-induced acute liver failure. Crit Care Med 2006;34: Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148: Lemeshow S, Hosmer DW Jr. A review of goodness of fit statistics for use in the development of logistic regression models. Am J Epidemiol 1982;115: Youden WJ. Index for rating diagnostic tests. Cancer 1950;3: Chun LJ, Tong MJ, Busuttil RW, Hiatt JR. Acetaminophen hepatotoxicity and acute liver failure. J Clin Gastroenterol 2009;43: Craig DG, Ford AC, Hayes PC, Simpson KJ. Systematic review: prognostic tests of paracetamol-induced acute liver failure. Aliment Pharmacol Ther 2010;31:

8 412 CHOLONGITAS ET AL. LIVER TRANSPLANTATION, April Wehler M, Kokoska J, Reulbach U, Hahn EG, Strauss R. Short-term prognosis in critically ill patients with cirrhosis assessed by prognostic scoring systems. Hepatology 2001;34: Kitzberger R, Funk GC, Holzinger U, Miehsler W, Kramer L, Kaider A, et al. Severity of organ failure is an independent predictor of intracranial hypertension in acute liver failure. Clin Gastroenterol Hepatol 2009;7: Blei AT. Selection for acute liver failure: have we got it right? Liver Transpl 2005;11(suppl 2):S30-S Kremers WK, van IJperen M, Kim WR, Freeman RB, Harper AM, Kamath PS, Wiesner RH. MELD score as a predictor of pretransplant and posttransplant survival in OPTN/UNOS status 1 patients. Hepatology 2004;39: Yantorno SE, Kremers WK, Ruf AE, Trentadue JJ, Podesta LG, Villamil FG. MELD is superior to King s College and Clichy s criteria to assess prognosis in fulminant hepatic failure. Liver Transpl 2007;13: Dhiman RK, Jain S, Maheshwari U, Bhalla A, Sharma N, Ahluwalia J, et al. Early indicators of prognosis in fulminant hepatic failure: an assessment of the Model for End-Stage Liver Disease (MELD) and King s College Hospital criteria. Liver Transpl 2007; 13: Rossaro L, Chambers CC, Polson J, Bowlus CL, Hynan LS, Fontana RJ, et al. Performance of MELD in predicting outcome in acute liver failure [abstract]. Gastroenterology 2005;128:A Almenoff PL, Leavy J, Weil MH, Goldberg NB, Vega D, Rackow EC. Prolongation of the half-life of lactate after maximal exercise in patients with hepatic dysfunction. Crit Care Med 1989;17: Bernal W, Donaldson N, Wyncoll D, Wendon J. Blood lactate as an early predictor of outcome in paracetamolinduced acute liver failure: a cohort study. Lancet 2002; 359: Macquillan GC, Seyam MS, Nightingale P, Neuberger JM, Murphy N. Blood lactate but not serum phosphate levels can predict patient outcome in fulminant hepatic failure. Liver Transpl 2005;11: