Prediction of esophageal varices in biliary atresia: Derivation of the “varices prediction rule”, a novel noninvasive predictor

Background and aimsLiver fibrosis and cirrhosis are nearly inevitable following Kasai portoenterostomy (KPE) for biliary atresia (BA), though the formation of varices is not. We sought to assess the value of noninvasive indices of portal hypertension (PHT), in predicting significant esophageal varices, and to develop a novel prediction model through regression modeling.MethodsThis is a retrospective, observational study with analysis of routine biochemical and ultrasound data. Five indices were examined: AST-to-platelet ratio index (APRi); hepatic artery resistance index (HARI); spleen size z score (SSAZ); platelet count-to-SSAZ ratio (P/SSAZ); and clinical prediction rule (CPR) [(0.75 × platelets)/(SSAZ + 5)] + (2.5 × albumin), each at specific time points following KPE (6 months, 1 year and 2 years). Significant varices were defined as grade ≥ 2 at endoscopy (screening or following a gastrointestinal bleed). AUROC was calculated for all indices. Univariate analysis was used to assess variables' suitability for inclusion in a subsequent multivariate logistic regression model to generate a predictive index. Data are quoted as median (range). P values ≤ 0.05 were regarded as significant.Results195 infants (median age at KPE of 55 days [11–216]) were analyzed. 42 (22%) had significant varices (median time to first presentation of varices of 1.20 [0.20–6.40] years). CPR and APRi (AUROCs ranging from 0.73–0.80 and 0.69–0.83 respectively) performed best overall. Multiple logistic regression modeling yielded a novel predictor at 6 months post-KPE: the Varices Prediction Rule (albumin × platelets/1000) (AUROC 0.75, sensitivity 86%, sensitivity 71%).ConclusionsNoninvasive indices such as CPR, APRi and now VPR can provide a tool for stratifying BA patients for elective endoscopy and possible preemptive intervention.