Secondary Logo

Journal Logo

Original Articles: Hepatology

Factors Determining δ-Bilirubin Levels in Infants With Biliary Atresia

Ye, Wen*; Rosenthal, Philip; Magee, John C.; Whitington, Peter F.§ for the Childhood Liver Disease Research and Education Network.

Author Information
Journal of Pediatric Gastroenterology and Nutrition: May 2015 - Volume 60 - Issue 5 - p 659-663
doi: 10.1097/MPG.0000000000000690
  • Free


What Is Known/What Is New?

What Is Known

  • Conjugated bilirubin (Bc) in plasma of patients with cholestasis reacts to form the bilirubin–albumin conjugate called δ-bilirubin (Bδ).
  • Bδ is 1 component of direct-reacting bilirubin.
  • Bδ has a long half-life in plasma; thus, Bδ and direct bilirubin may remain elevated long after biliary obstruction is cleared.
  • What Is New
  • The level of Bδ increases for approximately 50 days after a failed Kasai operation, after which Bδ plateaus.
  • After Kasai operation, persistent elevated Bδ is accompanied by elevated Bc and is thus related to continuing cholestasis.

The measurement of serum bilirubin is perhaps the single most valuable tool in the clinical evaluation of infants with liver disease. The majority of circulating serum bilirubin consists of 2 species. First is unconjugated bilirubin (Bu), which is essentially insoluble in water and is tightly bound to albumin and other plasma proteins. The other is conjugated bilirubin (Bc), which consists of an array of sugar (mainly glucuronide) ester monoconjugates and diconjugates of bilirubin. Within the context of neonatology, the clinical utility of measuring Bu comes in the management of infants with routine physiologic jaundice, those with increased bilirubin production, and those with defective conjugation, and the accuracy of measurement specifically relates to assessing risk for bilirubin encephalopathy (1,2). Serum Bc levels reflect the reflux of bilirubin from hepatocytes and perhaps bile after conjugation; thus, the clinical utility of measuring Bc is an indicator of cholestatic liver disease (3). Indeed, an elevated Bc (or direct bilirubin) is the entryway into the pathway leading to the diagnosis of biliary atresia (BA) (4).

Using slide film technology, it is possible in the clinical chemistry laboratory to precisely and independently measure total bilirubin (TB), Bu, and Bc in serum samples (5–7). In doing so, it may be found that certain sera contain a third bilirubin species referred to as δ-bilirubin (Bδ) because its value equals TB − (Bu + Bc). Bδ consists of bilirubin conjugates covalently esterified to albumin and other serum proteins, so called biliprotein, and arises by spontaneous transesterification of bilirubin glucuronide esters to exposed carboxyl groups on various serum proteins (mainly albumin). Because this occurs only when Bc is present in excess, significant amounts of circulating Bδ are observed only in patients with cholestasis. In studies involving children, it was found that Bδ insignificantly contributed to TB in patients with predominantly Bu and in infants 1 month old or younger, whereas Bδ may contribute >50% of TB in older children with cholestasis (8,9).

BA is a liver disease of human infants, occurring with an incidence of 1 in 12 to 18,000 live births in populations around the world. Its etiology and pathogenesis are unknown, but the characteristic endpoint of the process is fibroinflammatory obliteration of part or all of the extrahepatic biliary system and obstructive cholestasis. Clinical cholestasis typically appears within 2 months of birth. The only known effective therapy for BA is the Kasai portoenterostomy (KPE), in which the extrahepatic bile duct remnants are surgically removed and bile drainage is attempted by anastomosing the biliary limb of a Roux-en-Y to the liver hilum. Outcome after KPE is not good: approximately 50% of patients do not achieve adequate bile drainage and require liver transplantation for end-stage cirrhosis usually by 2 years of age. Cholestasis persists in patients with failure of drainage. Although it is well known that the formation of Bδ is dependent on the presence of Bc in plasma, the dynamics of the process over time are not well understood. Because the transesterification process is bidirectional, we hypothesized that a dynamic equilibrium would be established over time, in which the proportion of TB in serum comprising Bδ would become stable. The study of a large and well-characterized cohort of infants with BA and failed KPE permitted us to study the dynamics of Bδ formation over time.


Study Patients and Bilirubin Assays

The patients were infants with BA enrolled in the Prospective Study of Infants and Children with Cholestasis (PROBE: NCT00061828) of the Childhood Liver Disease Research and Education Network and its predecessor, the Biliary Atresia Research Consortium, from June 1, 2004 to March 1, 2012. Informed consent was obtained from the study participant's parents or guardians, and the protocol was carried out under institutional review board approval. Baseline data were collected at the time of KPE, including demographics, medical history, physical examination, surgical information, and laboratory measurements. Follow-up visits for data collection occurred at 1, 2, 3, and 6 months, after KPE, and then at ages 12 and 18 months. Thereafter, annual follow-up was performed on each patient's birth date, until age 9 or loss to follow-up. Serum bilirubin measurements obtained during the first 2 years of life before liver transplant (if any) in which TB, Bu, and Bc were reported as being measured independently were used in the present analysis. Each such measurement was considered as an “observation”; values from individual patients obtained at different time points were considered to be repeated observations separated in time. The time interval was computed from the dates of the observations.

Measurements were made in the clinical chemistry laboratories of the various clinical centers. All but 1 of the centers used the Vitros Integrated Systems (Ortho Clinical Diagnostics, Rochester, NY), whereas the 1 center used Dimension Vista 3000T (Siemens, Malvern, PA).

Data Analysis and Statistics

Because the precision of bilirubin measurements is poor at low values and Bδ is considered to develop only in individuals with elevated Bc, observations with TB < 1.3 mg/dL, Bc = 0, or Bu = 0 were excluded from the analysis. Descriptive data were summarized as the mean, standard deviation, and range for continuous variables and as percentages for categorical variables. For continuous variables with skewed distributions, median, quartiles, and range are reported. To characterize how Bδ level is related with TB, Bc, and serum albumin, locally weighted scatterplot smoothing (10) regression was used to explore function form of the relation using all of the observations.

Nonenzymatic transesterification is a chemical reaction whose progress is determined by the amount of the reactants Bc and albumin, temperature, and time. In the present study of BA, Bc, serum albumin, and time varied enough to contribute to variability in Bδ levels and were studied as independent variables. We computed the time interval between KPE and the date of the observation as an approximation of length of exposure to TB. To study how Bc and serum albumin contribute to Bδ variability over time, we constrained the analysis to observations from patients with poor drainage after KPE (defined as no TB ≤ 2.0 reported before 3 months after KPE). The rationale for excluding observations from patients with good drainage is because these infants achieved low level of Bc within 3 months after KPE and therefore did not have long-term exposure to cholestasis. A linear mixed effects model was used to test the association between Bδ and the factors of interest (ie, Bc, time since KPE, and serum albumin). This model takes account of the correlation between repeated measures from the same patients. Interaction terms between time and Bc were tested to access potential interaction between these 2 factors. Model diagnostics on residuals and influential points were examined to assure model fit and find final range of data to include in the analysis.

In addition, we also estimated mean time trajectories of Bδ as a fraction of TB (Bδ/TB) in the poor drainage groups. To adjust for potential bias caused by informative censoring of the repeated TB measurements because of transplant and death, a joint model of longitudinal and survival data with penalized splines was used for the present analysis (11). All of the analyses were performed using SAS/STAT (SAS Institute, Cary, NC) and R (12).


A total of 350 infants with BA were enrolled from June 1, 2004 to March 1, 2012. Among these 350 patients, 140 patients had 1 or more bilirubin measurements that were recorded as TB, Bu, and Bc (a total of 557 observations). After excluding observations with TB < 1.3 mg/dL (195 instances), Bc = 0 (18), Bu = 0 (3), and (Bu + Bc) > TB (as being physiologically impossible and probably in error for technical or recording reasons) (5), 334 observations from 129 patients were used for the analyses of the present study. The median of number of observations per patient is 2 (range: 1–10). Table 1 shows the demographics and other characteristics of this cohort of 129 patients.

Characteristics of the study cohort

The median (quartile, range) of the 4 bilirubin species among the 334 observations is 7.5 mg/dL (quartile: 4.7–10.5, range: 1.3–36.3) for TB, 3.6 mg/dL (quartile: 1.8–5.4, range: 0.1–27.0) for Bc, 1.1 mg/dL (quartile: 0.7–1.6, range: 0.1–11.0) for Bu, and 2.3 mg/dL (quartile: 1.5–3.7, range: 0.0–12.1) for Bδ.

All of the observations are plotted as (Bu + Bc) versus TB in Figure 1A. In this plot, Bδ for each observation equals the vertical distance from the fitted line of (Bu + Bc) to the line of identity at each level of TB. This shows the general relation between Bδ and TB expected in cholestasis, in which (Bc + Bu) levels are elevated and give rise to Bδ. It can be seen that in this cohort of patients with BA, the regression line for (Bu + Bc) observations falls increasingly below the line of identity with increasing TB, reflecting larger Bδ values at higher levels of TB.

Bδ and other bilirubin species. A, Dot–dash line is the 45° line that indicates the complete identity of values on horizontal and vertical axes. Solid line is the estimated mean of Bc + Bu using LOESS regression. B, Solid line is the estimated mean of Bδ using LOESS regression. Dotted lines are the 95% confidence interval of the predicted mean. Dashed lines are the 95% prediction interval. Bδ = δ-bilirubin; Bc = conjugated bilirubin; Bu = unconjugated bilirubin; LOESS = locally weighted scatterplot smoothing; TB = total bilirubin.

The relation between Bδ and Bc is shown more clearly by plotting Bδ values calculated for each sample versus Bc, as shown in Figure 1B. These data show the expected relation between these individual bilirubin analytes. Because Bδ is derived from Bc by spontaneous transesterification in plasma, Bδ tends to increase with increasing Bc. The data show this to be the case in BA. The regression line demonstrates that on average, for every 1 mg/dL increase in Bc between values of 2 and 10 mg/dL, Bδ increases by approximately 0.37 mg/dL. Only approximately 60% of the variability in Bδ, however, is explained by the variation of Bc (Pearson regression coefficient R2 = 0.59, P < 0.001), which suggests that there may be other factors that determine how much Bδ exists in plasma in patients with BA. In addition, the change in Bδ with change in Bc >10 mg/dL flattens, suggesting saturation kinetics for the transesterification reaction. The determinants of steady-state levels of Bδ, however, are numerous and cannot be gleaned from the present data.

Among the 334 observations, there are only 15 observations from 9 patients obtained between 1 and 2 years after KPE. In addition, although 97% of the Bc values are between 0 and 15 mg/dL, there are 8 observations with Bc value between 15 and 27 mg/dL. Some of these observations were shown to be influential points when conducting model diagnostics with the initial model using all of the 214 observations from 65 patients with poor drainage (1 patient had missing serum albumin value and was excluded from the analysis). To avoid biased results driven by these extreme values, we excluded observations with time >365 days or Bc > 15 mg/dL. A total of 195 observations from 62 patients with poor drainage were included in the final analysis. As demonstrated in Figure 2A, there is no significant interaction between time and Bc (P = 0.55), but both time and Bc are significantly associated with Bδ. An increase of 1 mg/dL in Bc is related to approximately 0.36 mg/dL increase in Bδ (P < 0.0001); every 100 days of exposure to Bc is associated with approximately 1.0 mg/dL increase in Bδ (P < 0.0001) given the same level of Bc. Albumin levels were not found to be significantly related to Bδ (P = 0.89). The range of albumin concentrations in the cohort, however, was narrow, which limits the conclusions that can be made from this observation.

Bδ, Bδ/TB, and time in patients with poor drainage after Kasai surgery. A, Circles and solid line, around time of Kasai (2 weeks before to 1 day after Kasai); triangles and dashed line, 2 days to 3 months after Kasai; cross and dotted line, 3 months to 6 months after Kasai; x and dashed and dotted line, 6 months to 1 year after Kasai. All of the lines are estimated mean of Bδ using linear regression by time windows. B, Grey dotted shows individual trajectories. Solid line is the estimated mean of Bδ/TB using LOESS regression. Dashed line is the estimated mean of Bδ/TB using joint model of longitudinal and survival data for adjusting informative dropout. Bδ = δ-bilirubin; Bc = conjugated bilirubin; LOESS = locally weighted scatterplot smoothing; TB = total bilirubin.

The data shown in Figure 2 demonstrate that in BA, the length of time cholestasis exists in part determines the level of Bδ in patients with poor drainage after KPE. In Figure 2A, Bδ values are plotted against Bc grouped for various windows of time. The data show that the regression lines for groups generally are displaced upward for later windows of time (ie, longer exposure). Another way to look at the effect of time is to determine whether the contribution of Bδ to TB increases with time in each individual patient (shown in Fig. 2B). The data show that patients with BA with poor drainage have increasing amounts of Bδ, expressed as a fraction of TB, with time after KPE. The estimated mean trajectory using joint modeling, which adjusted for repeated measurements and informative dropout, shows a plateau approximately 50 days post-KPE and with Bδ approximately 40% of TB. Estimated mean trajectory using locally weighted scatterplot smoothing regression leads to a similar conclusion. Figure 3 demonstrates that in the limited range of serum albumin levels, the albumin level does not influence the level of Bδ.

Bδ and albumin. Black solid line is the predicted mean value of Bδ using LOESS regression. Dotted lines are the 95% confidence intervals of the predicted mean of Bδ. Dashed lines are the 95% prediction interval of Bδ. Bδ = δ-bilirubin; LOESS = locally weighted scatterplot smoothing.


We took the opportunity to study a large and well-characterized cohort of infants with BA and failed KPE whose serum bilirubin measurements were performed using slide technology, thus providing the ability to assess the dynamics of Bδ over time. This unique data set permitted us to establish the indicators that affect Bδ concentration with some certainty. We found that the concentration of Bc has the greatest influence on Bδ concentration, which is to be expected because Bδ is the product of Bc, the excretion of which is impaired in hepatobiliary disease. The prospective collection of observations over time permitted an accurate assessment of the effect of length of exposure on Bδ concentration, which has not been reported. We found that Bδ concentration as a fraction of TB increased with time of exposure up to approximately 50 days, after which a steady state is achieved. A model was constructed for Bδ dynamics in BA in which Bc and time are the important contributing factors to its formation and steady-state concentration.

Our findings have relevance in regard to the clinical assessment of serum bilirubin in the follow-up in infants with BA. Many, if not most, clinical laboratories measure serum bilirubin as “total” and “direct” using chemical methodology, as opposed to the TB, Bu, and Bc measurements provided by slide technology. The approaches are equally useful in clinical medicine, but the differences in output may be confusing to clinicians. Our findings may help clinicians’ understanding of serum bilirubin measurements obtained in infants with BA and perhaps other cholestatic diseases. TB is equivalent to “total” bilirubin from clinical analyzers using chemical methodology. Bc, however, is not equivalent to “direct” bilirubin, which is actually roughly equivalent to Bc + Bδ (or Bc + TB − [Bu + Bc]). So, a clinician following an infant with BA using measurements of “TB” and “direct” bilirubin will find that “direct” bilirubin will contribute a much greater percentage of “TB” than a clinician using measurements from a laboratory using slide technology will find Bc contributing to TB. And, if a clinician who typically follows “direct” bilirubin measurements gets results from a laboratory using slide technology and finds a Bc that is much lower than the last observed “direct” bilirubin, she or he should not interpret this as showing improved bilirubin excretion. Indeed, it has been shown that the fraction of TB constituted by Bδ increases considerably after relief of biliary obstruction in adults with obstructive jaundice (13), and thus the “direct” bilirubin level would remain elevated longer than would Bc. The 2 approaches are equally useful clinically, but not equivalent unless the differences in chemical species are understood. The present study helps to explain the differences expected in a follow-up of patients with BA.

Ours is the most comprehensive study of the determinants of serum Bδ reported to date. Some studies have encompassed larger numbers of patients, but with an array of disorders including hemolysis, Gilbert syndrome, and hepatobiliary disease (8,9,14). None of the previous reports investigated a single population with a single disease as the present study did. Focusing on infants with BA restricts any variability in Bδ related to the disease state and age. Furthermore, the study cohort was prospectively acquired and followed for 2 years, with close ongoing monitoring of data quality. One shortfall of the present study is the lack of knowledge as to when in the course of events cholestasis occurred. Hyperbilirubinemia may be present at birth in some patients with BA (15). Thus, in a few patients with BA, cholestasis may have existed and Bδ may have been formed in utero (16). Most patients with BA, however, do not develop clinical cholestasis until well after birth, typically at 4 to 6 weeks of age. Our data did not provide a time for onset of cholestasis in individual patients. Because we needed to select a consistent and secure start time for our analysis and because no patient had serum bilirubin measurement before the time of KPE, we chose that time to begin our analysis. The data presented in Figure 2 support the choice of start time in that they indicate the beginning of Bδ accumulation at or just before KPE.

Our data show a linear increase in Bδ with increasing Bc up to a Bc of 10 mg/dL. Upwards of that point, however, Bδ concentration appeared to plateau, but this observation must be viewed cautiously because there were few observations with Bc > 10 mg/dL. It seems unlikely that the transesterification reaction would have reached saturation at this Bc concentration given that the concentration of albumin did not affect Bδ concentration. More data in the Bc range >10 mg/dL are required to make more confirmatory conclusions.

The increase in Bδ/TB with increasing time exposure appears to be linear up to an exposure time of 50 days, after which it plateaus. This is probably a true phenomenon because we have adjusted for the potential bias caused by informative dropout (patients with a high value of Bδ/TB tended to have transplants earlier and consequently had a shorter follow-up and less number of observations). This makes physiologic sense also. The plateau would occur where the rate of formation of Bδ equals the rate of degradation. The rate of formation is determined by the Bc concentration, whereas the rate of degradation is determined by the half-life of albumin, which is approximately 20 days. Whereas the Bc concentration varied substantially among patients, the half-life of albumin would not be expected to vary among individuals. These determinants would lead to a plateau of Bδ (or steady-state condition) that would vary somewhat in Bδ concentration but not in time of reaching a steady state.

In conclusion, Bδ levels in infants with BA increase with increasing levels of Bc and longer duration of cholestasis. Understanding the relation among Bδ, Bc, TB, and direct-reacting bilirubin levels can help in interpretation of the clinical extent of cholestasis in infants and children with BA, assisting in the diagnosis and management of these infants.


1. Flaherman VJ, Kuzniewicz MW, Escobar GJ, et al. Total serum bilirubin exceeding exchange transfusion thresholds in the setting of universal screening. J Pediatr 2012; 160:796–800.e1.
2. Gartner LM, Herrarias CT, Sebring RH. Practice patterns in neonatal hyperbilirubinemia. Pediatrics 1998; 101:25–31.
3. Arvan D, Shirey TL. Conjugated bilirubin: a better indicator of impaired hepatobiliary excretion than direct bilirubin. Ann Clin Lab Sci 1985; 15:252–259.
4. Moyer V, Freese DK, Whitington PF, et al. Guideline for the evaluation of cholestatic jaundice in infants: recommendations of the North American Society for Pediatric Gastroenterology, Hepatology and Nutrition. J Pediatr Gastroenterol Nutr 2004; 39:115–128.
5. Rosenthal P, Keefe MT, Henton D, et al. Ektachem and unconjugated bilirubin measurements. J Clin Chem Clin Biochem 1989; 27:829.
6. Wu TW, Dappen GM, Powers DM, et al. The Kodak Ektachem clinical chemistry slide for measurement of bilirubin in newborns: principles and performance. Clin Chem 1982; 28:2366–2372.
7. Wu TW, Dappen GM, Spayd RW, et al. The Ektachem clinical chemistry slide for simultaneous determination of unconjugated and sugar-conjugated bilirubin. Clin Chem 1984; 30:1304–1309.
8. Brett EM, Hicks JM, Powers DM, et al. Delta bilirubin in serum of pediatric patients: correlations with age and disease. Clin Chem 1984; 30:1561–1564.
9. Rosenthal P, Henton D, Felber S, et al. Distribution of serum bilirubin conjugates in pediatric hepatobiliary diseases. J Pediatr 1987; 110:201–205.
10. Cleveland WS. Robust locally weighted regression and smoothing scatter-plots. J Am Stat Assoc 1979; 74:827–836.
11. Rizopoulos D. Joint Models for Longitudinal and Time-to-Event Data, With Applications in R. Boca Raton: Chapman & Hall/CRC Press; 2012.
12. R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2011.
13. Kozaki N, Shimizu S, Higashijima H, et al. Significance of serum delta-bilirubin in patients with obstructive jaundice. J Surg Res 1998; 79:61–65.
14. Weiss JS, Gautam A, Lauff JJ, et al. The clinical importance of a protein-bound fraction of serum bilirubin in patients with hyperbilirubinemia. N Engl J Med 1983; 309:147–150.
15. Harpavat S, Finegold MJ, Karpen SJ. Patients with biliary atresia have elevated direct/conjugated bilirubin levels shortly after birth. Pediatrics 2011; 128:e1428–e1433.
16. Ostrea EM Jr, Ongtengco EA, Tolia VA, et al. The occurrence and significance of the bilirubin species, including delta bilirubin, in jaundiced infants. J Pediatr Gastroenterol Nutr 1988; 7:511–516.

biliary atresia; δ-bilirubin; conjugated bilirubin

© 2015 by European Society for Pediatric Gastroenterology, Hepatology, and Nutrition and North American Society for Pediatric Gastroenterology,