Utility of Liver Transplant Risk Models

Pediatric liver transplantation (LT) is a highly specialized practice requiring surgical expertise and medical sophistication regarding preoperative optimization, perioperative management, and immunosuppression. Contemporary results have yielded very high success rates with long-term survival approaching 90% beyond a decade.1 Numerous teams have proposed predictive models for both short- and long-term survival that have varying utility. The value of early prediction is for the education of families regarding the probability of success and perhaps in eliminating futile transplants in those unlikely to benefit. However, models generated in transplanted patients chosen by conventional criteria may not be useful for selection and in the education of these families. Models based on age, disease condition, and comorbidities would likely be useful in selection. This is opposed to variables that emerge in the setting of technical complications, which are only helpful if there is the possibility to avoid these events (such as advising less experienced centers in adopting best surgical practices).2 Additionally, variables such as infection or intestinal perforation, although important predictors of outcomes, are presumably a consequence of surgical technique, or underlying patient risk, and are not useful in preoperative prediction. Long-term failures, such as recurrent malignancy, are intrinsic to the cancer indications but are not relevant to overall selection decisions and, therefore, not as useful for models. Outcomes associated with immunosuppression practice are subject to substantial variability based on center practice both with respect to chronic rejection and the risk of posttransplant lymphoproliferative syndromes.3 A recent report concerning a large series aggregated from the Organ Procurement and Transplantation Network identified a number of variables generally characterizing the severity of liver disease and underlying condition in a predictive model.4 In this issue of Transplantation, the authors presented an analysis of risk for pediatric LT generated from an experience of 3004 transplants, likely the largest in the world.5 The authors have selected 7 variables for the risk score out of 19 predictors identified in univariate analysis. The model performs well in their cohort and is replicated in the validation set. However, in a single-center analysis, such as used in the article here, numerous unmeasured clinical activities contribute to consistency over time (other than experience and protocol modifications) that limits the generalizability of the results. Although the authors acknowledge the limitation of a single-center model, they do not fully address this concern. There are no tables or figures detailing the composition of their cohort outside of nationality and blood type. Additionally, they note that “the duration of hospitalization, intensive care unit (ICU) stay, and the use of rituximab may vary due to social and economic factors,” but they do not comment on why these particular factors are affected and in what way. There has been literature in both the United States and Italy that demonstrated those with disadvantaged socioeconomic statuses have decreased access to living donor LT and higher waitlist mortality.6,7 Although living donation is often the primary avenue to transplantation in Asian countries, barriers faced by these families should be commented upon. Although many costs incurred during transplant may be covered by national/social support, it is important to note whether extended ICU stays or more advanced treatments result in increased out-of-pocket costs from patient families. A large part of this study claims that 1 of the most important factors in posttransplant care is infection, including ventilator-induced lung injury (VILI). VILI is usually a complication of prolonged intubation and, as a consequence, prolonged ICU stay.8 It is well known that early extubation after pediatric LT leads to decreased ICU stay and thus decreased VILI.9 Thus, it would be important to comment on how social and economic factors can affect ICU stay in this study as commented by the authors. Is underlying lung disease an issue in predicting prolonged ventilation or is this a sign of other technical complications? It would also be useful to comment on how sputum cultures were obtained and if these cultures occurred only in intubated patients or extubated as well, as this was used as 1 of the 7 factors to normalize the model and pretransplant sputum culture was not collected. The validation of sputum cultures from sputum is concerning for decreased sensitivity and sensitivity as well as being affected by prior antimicrobial use.10 Finally, blood loss has long been recognized as a predictor in LT and may be a consequence of patient condition or surgical technique, background information that is not detailed in the model used here. In summary, the authors present a vast experience that has informed their practice and demonstrated excellent outcomes.11 However, the value of predictive models across centers is best ascertained using multicenter data sets, where the impact of individual practices, both measured and unmeasured, can best be understood.