King's Variceal Prediction Score

Abstract

Variceal haemorrhage can be a life-threatening complication of chronic liver disease in children. There is limited evidence about the optimal prophylactic management and selection criteria of children who will benefit from upper gastrointestinal endoscopy. Children presenting in our centre with suspected portal hypertension or gastrointestinal bleeding and undergoing their first oesophagogastroduodenoscopy between 2005 and 2012 were included. Clinical, biochemical, and radiological data were collected. A separate validation cohort from May 2013 to October 2014 was obtained. Data on 124 treatment-naïve patients were collected; 50% had biliary atresia. Thirty-five (28%) children presented with gastrointestinal bleeding and overall 79 (64%) had clinically significant (grade II-III) varices. Clinical prediction rule, aspartate aminotransferase-platelet ratio index, and varices prediction rule had at optimal cut-off sensitivity and specificity of 76% and 59%, 60% and 55%, and 80% and 59%, respectively. Logistic regression yielded a new prediction rule of (3 × albumin ([g/dL]) - (2 - equivalent adult spleen size [cm]). This King's variceal prediction score had a favourable areas under the curve of 0.772 (0.677-0.867) compared to clinical prediction rule 0.732 (0.632-0.832). At the optimal cut-off of 76 this yielded a sensitivity and specificity of 72% and 73% and a positive and negative predictive value of 82% and 60%, respectively. In the validation cohort areas under the curve was 0.818 (0.654-0.995) with sensitivity and specificity of 78% and 73%, respectively. Our new prediction score may be a useful tool in the selection of children with clinically significant varices eligible for a screening endoscopy.