CALCULATION OF BILIARY ATRESIA PROGNOSTIC INDEX USING A MULTIVARIATE LINEAR-MODEL

The purpose of this study was to determine the relative value of liver function markers in predicting the magnitude of morbidity and to develop a quantitative estimate of the prognostic risk using a multivariate regression model in patients who have been operated on for biliary atresia. The study sample consisted of 37 patients who had undergone a Kasai portoenterostomy; the jaundice disappeared in 32 and persisted in five. A computer-based stepwise regression procedure produced the linear predictive models by the equation: biliary atresia prognostic index (BAPI) = 9.2 Cu:Zn + 1.0 ZTT + 3.2 TB - 0.05 ChE + 9.9 for infants under 1 year of age, and BAPI = 10.3 Cu:Zn + 0.8 ZTT + 0.03 GGTP - 0.12 ChE + 25.6 for children over 1 year of age. In validation of these models, the indexes fluctuated from -17 to 122, and the degree of morbidity increased linearly with the increase in BAPI. Postoperatively the patients were classified into four categories according to the dynamics of their postoperative course: A (BAPI < 25), successful cases that should not require liver transplantation (40.5%); B (25 less than or equal to BAPI less than or equal to 50), improved cases that have extended survival with their native liver (29.7%); C (50 < BAPI less than or equal to 75), cases that improved in terms of disappearance of jaundice but ultimately will require liver transplantation (8.1%); and D (BAPI > 75), cases that require early referral for transplantation (21.6%). (The percentages indicate the distribution rate of patients at the time of final follow-up evaluation.) These models allow quantification of the risk of morbidity from progressive liver cirrhosis in the individual patient, permitting the clinician to consider whether such patients should be considered for liver transplantation. Copyright (C) 1995 by W.B. Saunders Company.