Journal of Pediatric Gastroenterology & Nutrition:
Shneider, Benjamin L.
Department of Pediatrics, University of Pittsburgh School of Medicine, Pittsburgh, PA.
Address correspondence and reprint requests to Benjamin L. Shneider, MD, Department of Pediatrics, Children's Hospital of Pittsburgh, 4401 Penn Av, Pittsburgh, PA 15224 (e-mail: email@example.com).
Received 7 September, 2011
Accepted 19 September, 2011
The author reports no conflicts of interest.
See “Pediatric End-stage Liver Disease Score in Acute Liver Failure to Assess Poor Prognosis” by Sanchez and D’Agostino on page 193.
One of the more vexing questions in pediatric hepatology is deciding whether to proceed with liver transplantation in a child with acute liver failure (ALF). Decisions are often made in the middle of the night, while many of your colleagues are fast asleep. Choosing to proceed with transplantation in effect cannot be “wrong” because there is no opportunity to assess the outcome independent of transplant. It is impossible to determine whether the patient in question may have survived without a transplant. Many have presumed that if a decision was made to undertake something as dramatic as liver transplantation, it must have been necessary (more about this later). Opting against transplant is fraught with the risk of death, although it allows for the possibility of spontaneous survival and avoidance of the known risk of morbidity and mortality incumbent with liver transplantation. Auxiliary and hepatocyte transplantations have evolved as approaches to finesse this complicated medical decision making. Scoring systems have been developed, evaluated, and revised in an effort to aid the lonely clinician in the middle of the night.
Sanchez and D’Agostino have taken on this difficult and important question (1). In their study, pediatric end-stage liver disease (PELD) score was assessed at the time of hospitalization and analyzed as a possible predictor of outcome in ALF. On initial assessment, it appears that there may be some value in this readily available score, but the story is more complicated, much more complicated. The underlying etiologies for ALF in the present study are not necessarily applicable to all of the centers, in particular the high prevalence of hepatitis A. One of the problems with developing a predictive model for ALF in children is the remarkably complicated range of potential etiologies of the clinical syndrome. It is not at all surprising that pediatric ALF is difficult to model reliably, perhaps impossible. Accurate models require extremely large numbers of patients and, by corollary, a multicenter approach. Sanchez and D’Agostino's study uses laboratory data from the time of admission. ALF is a moving target, and the time of admission varies widely based upon local practice. For many centers, patients are admitted elsewhere and transferred days later. A dynamic scoring system is required.
Perhaps the most important issue in all of the modeling done to date is the endpoint. In this case, and in many others, the predicted outcome is death or transplant versus spontaneous survival. This is a fundamental flaw and limitation. Many have presumed that the decision to proceed with transplantation is accurate and should be equivalent to death. When one considers the difficulties of the decision in the middle of the night, it is expected that most individuals would err on the side of caution and proceed with transplant. Previous studies have suggested that total bilirubin, prolongation of prothrombin time, and degree of encephalopathy predict poor outcome. Clinicians use this information to decide about transplant; thus, it is not surprising that these parameters predict outcome when transplant is included. Total bilirubin and prothrombin time/international normalized ratio are key elements of PELD, and thus the present study is, at least in part, a self-fulfilling prophecy. The more essential analysis involves spontaneous survival and death, in which transplant is excluded. When one reanalyzes the data in Table 1 of the present analysis, the difference in PELD score between spontaneous survival and death is not significant (PELD score: survivors 31.3 ± 10.3 [n = 11], nonsurvivors 39.8 ± 9.5 [n = 6]). One wonders what the receiver operating characteristics curve would look like.
Overall, this important analysis is food for thought for our community. More investigation is required; that investigation needs to be multicentered and prospective. Ultimately, no model will be perfect. We need to make the difficult decision as to where we should optimize the predictive capacity—to not overuse transplantation or to minimize the risk of spontaneous death.
1. Sanchez MC, D’Agostino DE. Pediatric end-stage liver disease score in acute liver failure to assess poor prognosis. J Pediatr Gastroenterol Nutr 2012;54:193–6.