Abstract: Objective:Establish a histogram model for predicting the risks of biliary atresia (BA) in infants with cholestasis and gamma-glutamyl transferase (GGT) <300 U/L.Methods:Retrospective analysis was performed for 169 infants operated for cholestasis from January 2012 to January 2020.There were 90 boys and 79 girls with a body weight of (4.1±0.9) kg.A definite diagnosis with or without BA was made by intraoperative cholangiography.According to the level of GGT, they were divided into two groups of <150 U/L (very low GGT) and 150-300 U/L group (low GGT). Preterm birth and the last preoperative liver function parameters were recorded.And general profiles and liver function changes of BA and non-BA infants in very low GGT and low GGT groups were compared.Chi-square or t test of two independent samples was utilized for univariate analysis and multiple Logistic regression for examining the risk factors of BA.R software was utilized for line graph models of predicting BA.And model effect was examined by area under curve (AUC) and fit degree of receiver operator characteristic (ROC).Results:There were BA (n=77) and non-BA (n=92). There were 93 cases in very low GGT group (BA, n=31; non-BA, n=62) and 76 cases in low GGT group (BA, n=46; non-BA, n=30). Gender, body weight, preterm birth, alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TB), direct bilirubin (DB) and GGT had statistically significant differences ( P<0.05). And statistically significant differences existed in gender, body weight, premature delivery ALT/TB/DB in low GGT group ( P<0.05). Multivariate Logistic regression analysis revealed that body weight, gender and ALT/AST were independent risk factors for BA in very low GGT group (95% CI: 0.77-0.926, P<0.001). In low GGT group, gender, body weight, preterm birth and DB were independent risk factors for BA (95% CI: 0.870-0.985, P<0.001). Two groups of independent risk factors were used for modeling.The AUC of line diagram model was 0.864(female, body weight >3.4 kg, ALT >96 U/L, AST >107 U/L) in very low GGT group and 0.947 (term, female, body weight >4.2 kg, DB >101.5 μmol/L). And accuracy of two models in each group was 78.5% and 92.9% respectively. Conclusions:Differences exist in BA diagnostic parameters between GGT <150 U/L and 150-300 U/L cholestatic infants.It is worth promoting to establish a two-interval column graph prediction model using weight, gender, preterm birth and preoperative liver function.