Objective Prolonged neonatal jaundice is common and usually benign; however, assessment of bilirubin fractions is recommended to determine the need for further assessment for congenital liver disease, particularly biliary atresia. The direct (conjugated) bilirubin thresholds currently used are variable and poorly evidenced. Hence, we aimed to delineate direct bilirubin levels in disease-free neonates with prolonged jaundice.

Methods We performed a retrospective cohort analysis of split bilirubin levels, and subsequent follow-up, for all neonates initially assessed in our prolonged neonatal jaundice clinic over 2 years. We plotted centile charts for total, direct and direct–total bilirubin ratio levels against age at sampling. The association was assessed using linear regression analysis.

Results Data were collected for 420 neonates (501 blood samples) across an age range of 10–70 days. No significant liver disease was found. For each day of older age, total bilirubin fell by 3.72 µmol/L (95% CI 2.46 to 5.00) and direct bilirubin fell by 0.39 µmol/L (0.18 to 0.59). The ratio between the two did not change significantly (−0.0006 to +0.0034). The 95th centile for direct bilirubin was stable at ~25 µmol/L. Direct–total bilirubin ratio was very variable with some 95th centiles >30%.

Conclusions In a clinically relevant population of disease-free neonates with prolonged jaundice both the total and the direct bilirubin decreased with age. The absolute direct bilirubin is more useful clinically than the direct–total bilirubin ratio. Our results support National Institute for Health and Care Excellence guidance that conjugated bilirubin >25 µmol/L, or even more stringent criteria, constitutes an appropriate threshold for further investigation for neonatal liver disease.