End-Stage Liver Disease Models and Outcomes in Pediatric Patients Supported With Short-Term Continuous-Flow Ventricular Assist Devices.

Short-term continuous-flow ventricular assist devices (STCF-VADs) are increasingly being utilized in pediatrics. End-stage liver disease (ELD) models have been associated with outcomes in adult patients on mechanical circulatory support. We sought to determine the relationship between outcomes in children on STCF-VADs and three ELD models: model for end-stage liver disease-excluding international normalized ratio (MELD-XI; all) and MELD-XI (> 1 year), PELD, and a novel score, PedMELD-XI. All patients (< 19 years) supported with STCF-VADs, between June 2009 and December 2016 were included. The MELD-XI, PELD, and PedMELD-XI scores were calculated and their association with adverse events and a composite measure of death, major bleeding, and neurologic dysfunction was analyzed. Of 32 patients, median age was 0.57 years (interquartile range [IQR], 0.10-4.43), median weight was 7.15 kg (IQR, 3.68-16.53), 53.1% had congenital heart disease, and 53.1% were male. In total, 78.1% patients experienced an adverse event (78.1% a major bleed, 25.0% neurologic dysfunction, and 15.6% death). The median MELD-XI score was 11.17 (IQR, 9.44-30.01), MELD-XI (>1 year) 9.44 (IQR, 9.44-24.33), PELD 6.00 (IQR, 4.00-13.75), and PedMELD-XI -14.91 (IQR, -18.85 to -12.25). A higher MELD-XI for all ages (13.80 vs. 9.44, p = 0.037) and less negative PedMELD-XI (-14.16 vs. -19.34, p = 0.028) scores were significantly associated with bleeding and the composite outcome. PedMELD-XI was significantly associated with death (-12.87 vs. -16.84, p = 0.041) while a trend was seen for increased MELD-XI in all ages being associated with death (31.52 vs. 10.11, p = 0.051). Last, there was no association with the models and neurologic events. MELD-XI and PedMELD-XI were significantly associated with major bleeding and the composite endpoints with PedMELD-XI also being associated with death. These results suggest that ELD models can be used to predict outcomes in this specific patient population, however, further analysis in a larger population is required.