In this study, using a second independent validation cohort of children with PALF at a single institution, the previously derived LIU score was shown to be strongly predictive of death or LT (C index 86.3). In the validation set, the second and third quartiles showed overlap when compared to the training set of data, however the lowest and highest quartiles clearly reflected patients with low and high likelihood of death or LT. The predictive value was better for the LIU score using INR rather than PT, which may be explained by the manner in which the INR is derived. INR is a mathematical manipulation of the PT meant to standardize this test across different laboratories taking into account different clotting reagents and techniques. Because the PT measured on a day to day basis can vary and techniques used in different countries vary considerably, the INR standardizes this measurement and thus limits variation in the calculated score.
We also performed analysis by calculating LIU scores on the days that each variable peaked, using the corresponding daily values for the other variables, as opposed to using the cumulative peak values. The LIU scores using INR calculated on these individual variable peak days had only a slightly lower C index (75 – 86) than the overall peak LIU score. This suggests that the LIU score has the potential to be used on a daily basis, reflecting the risk for death or need for liver transplantation at any given point in time during hospitalization.
Both the King’s College Hospital Criteria and the Model for End-Stage Liver Disease (MELD) score have been evaluated as prognostic indices in adult ALF with mixed results (13
). Several other laboratory variables not included in these have been studied in adult ALF including alpha-fetoprotein (19
), phosphorus (21
), ammonia (22
), lactate (23
), and the APACHE II score (24
). Adding serum sodium values to the MELD score has also received attention for predicting survival in chronic liver failure (25
). We analyzed admission sodium values and did not find a correlation to outcome in PALF.
Following validation of the LIU score, we applied the LIU score to admission values to see if this was as predictive of outcome. We found a lower C index (79.3) when the LIU score was applied to admission values. We then derived a new admission LIU score that used only total bilirubin and PT/INR(C index 83.7), which appears promising although is slightly less predictive of outcome than the peak value LIU score. The poorer performance of the aLIU was believed to be due to variability in the duration and severity of PALF at the time of admission or transfer of patients to our hospital. Thus, admission laboratory values may be reflective of a later time point in the disease process in certain patients, most likely those transferred from other hospitals.
Our study included children with ALF caused by acetaminophen overdose, oncologic disorders, and sepsis/ischemia-reperfusion injury. As shown in our previous study, the LIU score appears to be valid in these patients. Acetaminophen overdose patients often have a better survival than other PALF patients; however, this was reflected by lower quartile LIU scores in acetaminophen survivors. Analysis of the LIU score excluding oncology and sepsis/ischemia-reperfusion injury patients did not significantly change the C index (data not shown).
The distribution of diagnoses we report differ slightly than those reported by the PALF Study Group, in which we have less indeterminate causes and more sepsis-ischemia reperfusion injury, infectious, and hematologic-oncologic etiologies (1
). This is likely due to our balance of primary versus outside referrals as shown in , in which we would expect primary referrals to have a slightly higher rate of sepsis-ischemia reperfusion injury and hematologic-oncologic etiologies than outside referrals. Other studies likely included patients that were primarily referred to centers for liver transplantation evaluation and thus may represent a more severely ill and selected population skewed toward indeterminate causes of PALF.
In conclusion, despite the limitations of a single center study, the current study validates the LIU score in a second set of patients with PALF and lends merit to this scoring system. This score could be useful as a dynamic biomarker to predict outcome and need for liver transplantation, for stratification of patients for clinical trials, and as a surrogate marker of outcome. The aLIU score was less predictive of outcome, which was not surprising given the variable clinical status of patients at the time of admission or transfer to our hospital. Further prospective analysis of the predictive value of the LIU score in additional independent sets of PALF patients will provide guidance for the clinical and research value of these two scoring systems in the clinical care and research setting of PALF.