Non-Invasive Predictors for the First Variceal Hemorrhage in Children With Biliary Atresia After Kasai Portoenterostomy.

Background/aim: : Portal hypertension and variceal hemorrhage (VH) are significant complications in biliary atresia (BA). The study aims to evaluate risk factors and noninvasive markers that predict actual VH for the first time in children with BA without prior endoscopic surveillance or treatment.

Methods: Retrospective review was performed of patients diagnosed with BA from 1989 to 2016 at a single center. Primary outcome was the first episode of VH. Patients were stratified into VH and non-VH groups according to the development of VH, and laboratory and ultrasonographic data were analyzed at 2 time points: pre-VH and the last follow-up. Existing indices, varices prediction rule (VPR), and aspartate aminotransferase (AST)-platelet ratio index (APRI) were also applied retrospectively to evaluate their performance in prediction of VH in our cohort.

Results: Seventy-two patients were included; 16 patients developed the first VH at median age of 5.5 years. On univariate analysis, serum albumin ( = 0.034), AST ( = 0.017), hemoglobin ( = 0.019), platelet count ( = <0.001), spleen size Z-score ( = <0.001), and rate of splenic enlargement ( = 0.006) were associated with VH. On multivariable regression analysis, only platelet count was independently predictive ( = 0.041). The optimal cutoff values for prediction of the first VH were platelet count ≤100 × 10/L (sensitivity 75.0%, specificity 80.4%, positive predictive value [PPV] 52.2%, negative predictive value [NPV] 91.8%), VPR ≤3.0 (sensitivity 81.3%, specificity 85.7%, PPV 61.9%, NPV 94.1%), and APRI ≥3.0 (sensitivity 81.3%, specificity 76.8%, PPV 50.0%, NPV 93.5%).

Conclusions: Platelet count <100 × 10/L and VPR <3.0 are simple, reproducible and effective noninvasive markers in predicting the first episode of acute VH in children with BA and may be used in pediatrics for the selection of patients to undergo primary prophylactic endoscopic therapy.