What Is Known
- Younger age at hepatoportoenterostomy and postoperative jaundice clearance are favorable prognostic factors associated with native liver survival in biliary atresia.
- Despite this, many infants who initially clear jaundice do develop progressive liver disease and ultimately require liver transplantation.
What Is New
- In the present study of multiple clinical and laboratory factors 3 months after hepatoportoenterostomy, serum total bilirubin, and albumin levels and center were independently associated with native liver survival.
- Infants with biliary atresia who cleared jaundice after portoenterostomy but had low serum albumin levels <35 g/L (3.5 mg/dL) 3 months postoperatively, had a high likelihood of requiring liver transplantation by 2 years of age, had poorer linear growth but did not have significant coagulopathy, suggesting a role for early nutritional intervention in this group.
Biliary atresia (BA), a progressive obliterative cholangiopathy of infancy, is the most frequent cause of chronic cholestatic liver disease in children and leading indication for pediatric liver transplantation (LT) (1).
Surgical management of BA is limited to hepatoportoenterostomy (HPE), which aims to restore bile flow by establishing continuity of intrahepatic bile ducts with the intestine, or primary LT. Approximately 80% of infants who undergo HPE will ultimately require LT (2,3), with native liver survival rates ranging between 46% and 65% at 2 years (4–6) and between 35% and 46% at 5 years (5,7,8). Younger age at HPE (5) and postoperative resolution of jaundice (7,9–14) have been associated with improved native liver survival; however, infants who initially clear jaundice may develop progressive liver disease and still require LT.
A major challenge for clinicians involved in the care of these patients is determining the optimal time to list BA infants for LT. We aimed to identify early prognostic factors that may be used to predict native liver survival in infants with BA who underwent HPE. We hypothesized that early clinical and laboratory measures of liver health in addition to serum bilirubin may enhance prognostication over the short and longer term.
Study Population and Design
We undertook a retrospective study using medical records of all infants with BA, who underwent HPE between January 1986 and June 2009 in 2 institutions in Sydney, Australia (Royal Alexandra Hospital for Children, RAHC and Sydney Children's Hospital, SCH) and 1 in Toronto, Canada (The Hospital for Sick Children). Data from the 2 Sydney institutions were considered together as 1 center because there was considerable crossover of care. The city of origin for each patient was deidentified as Center 1 and Center 2 for the purpose of this manuscript. The study was approved by the institutional review board at each site.
Key demographic, clinical, radiological, laboratory, and outcome data were extracted to a standard form in an electronic database by a single reviewer (S.N.). All admissions, outpatient encounters, and correspondence for the first 2 years of life were reviewed. Collected data points included sex; birthdate; date of HPE; routine laboratory values (see below) and growth parameters including weight and length at birth, time of HPE, and 3 and 6 months after HPE; extent of reported fibrosis on histopathology reports at the time of diagnosis or HPE; anatomical subtype of BA; presence of associated extrahepatic anomalies; date of first detection of ascites (clinically or sonographically); and date of any upper gastrointestinal hemorrhage requiring hospital admission. Age at first LT, death, and last follow-up encounter were recorded to determine the primary outcome of native liver survival at 2 years of age.
Laboratory values included total bilirubin (TB) and conjugated or direct serum bilirubin; albumin; alanine aminotransferase; gamma-glutamyltranspeptidase; platelet count; and cholesterol. The values were recorded from just before HPE and at 3 months after HPE (±3 weeks for the latter). If more than 1 value was available during this time range for 3 months ± 3 weeks after HPE, the most “healthy” value was recorded (eg, the lowest value for TB or highest value for albumin). Where TB was measured and reported by the laboratory, this was recorded. Alternatively, the sum of direct and indirect (or conjugated and unconjugated) fractions was used. Delta bilirubin was not recorded because it was rarely measured and reported. If an albumin infusion had been given in this time frame, the preinfusion albumin level was recorded. Weight and length measurements were converted to z scores using the Centers for Disease Control and Prevention 2000 growth data (available at http://www.cdc.gov/growthcharts/percentile_data_files.htm). Extrahepatic congenital anomalies were noted and BA structural malformation (BASM) syndrome (15) was noted if present. Ascites were determined to be present when clinically detectable, or described on sonographical or intraoperative reports as more than a trace or scant amount. Pathology reports on liver biopsies at the time of HPE were examined and categorized as describing either minimal fibrosis, bridging fibrosis, or cirrhosis (nodule formation). BA type was classified based on cholangiographic and/or surgical pathologic findings, using the Japanese Society of Pediatric Surgery criteria, defined by the most proximal level of extrahepatic bile duct obstruction (type I, common bile duct; type II, common hepatic duct; type III, the porta hepatis) (16).
The approach to clinical care of infants with BA was similar in the 2 cities. HPE procedures were performed by pediatric surgeons with postoperative and ongoing care provided by expert multidisciplinary teams including pediatric gastroenterologists and dieticians. Postoperative corticosteroids were not used. Postoperative oral antibiotics were continued for a variable and often short period (typically <6 weeks) in center 1, except for the treatment of recurrent episodes of cholangitis. In center 2, ongoing postoperative antibiotic prophylaxis was routine care. Detailed nutritional support data were not uniformly available at 3 months after HPE.
Comparison of baseline variables between patient populations with BA in the 2 cities was made using chi-square and Student t test. Native liver survival rates were compared with the log-rank test. Univariate analysis of clinical and laboratory data evident at time of HPE and 3 months post-HPE associated with outcome at 2 years was undertaken using chi-square and Student t test. Variables with P < 0.10 and with <20% of data points missing were initially included in a multiple logistic regression model, then subjected to variable selection using a backward elimination process to select the best subset of predictors of native liver survival at the age of 2 years. Continuous covariables were examined using Pearson's correlation coefficient (or Spearman's rho if not normally distributed), with a value of >0.4 regarded as correlation. Variables were considered collinear if the variance inflation factor was >2. The performance of the regression model was assessed in the cohort using the area under the curve of receiver-operating characteristic plot (AUROC) for the model. Receiver operating characteristic (ROC) analysis of variables in the regression model was undertaken to determine optimal cut-off values (to maximize both sensitivity and specificity), and Kaplan-Meier curves for native liver survival were constructed using these cut-offs alone and in combination. Data analysis was undertaken using STATA Version 11 (StataCorp, College Station, TX).
Study Subject Characteristics
Two hundred thirty-five children with BA who underwent HPE were identified, of whom 18 patients were excluded because they either did not survive until 3 months post-HPE or did not have outcome data available at 2 years of age. Therefore, the final study cohort was composed of 217 subjects followed at 2 centers (110 from center 1; 107 from center 2). Table 1 summarizes the demographics and baseline data of the study population.
Patients with BA in center 1 were less often diagnosed with BASM (9.1% vs 19.6%, P = 0.02), and had a higher serum TB at HPE (155 vs 129 μmol/L, P < 0.001) compared with patients with BA in center 2. Median age at HPE was not significantly different between the 2 centers (63 vs 70 days, P = 0.134).
The combined group had a 2-year native liver survival rate of 54.8%. Overall patient survival was not significantly different between the centers (log rank P = 0.1213); however, 2-year native liver survival rates were different (61.8 in center 1 vs 47.6% in center 2, logrank P = 0.03). The 2 Sydney institutions did not have different 2-year native liver survival rates (logrank P > 0.05).
Native liver survival at 2 years of age was significantly associated with center, BA type, age at HPE, degree of histological fibrosis at diagnosis, development of ascites within 3 months, TB, albumin, and alanine aminotransferase concentrations 3 months post-HPE, and change in length z score in the first 3 months after HPE (Tables 2 and 3). Age at HPE was a significant factor when categorized either as a dichotomous variable (≤ or >45 days, ≤ or >60 days) or as a continuous variable.
Patients with type III BA fared worse; however, fewer than 10% of patients were type I or II, limiting the statistical power of analysis of outcome by BA type. Only 109 of 217 (49%) patients had complete length measurement data to calculate change between HPE and 3 months post-HPE. Of these patients, those with the better outcome at 2 years on average maintained their length z score during the first 3 months after HPE.
The presence of associated extrahepatic malformations (BASM) was not associated with adverse 2-year outcome. This was also true when the narrower definition of BA splenic malformation syndrome of Davenport (17) was used (P = 0.666; data not shown). Weight z score at 3 months after HPE and the change in weight z score from time of HPE were also not significantly associated with native liver survival at 2 years.
BA type was not included as an independent variable due to small sample size of nontype III patients. TB at 3 months after HPE rather than the delta value (change in TB from time of HPE to 3 months later) was chosen because there were fewer missing values in the former, and both were correlated and significantly associated with outcome. Backwards elimination arrived at the final logistic model involving center, TB, and albumin at 3 months post-HPE (Table 4).
Center, TB, and albumin at 3 months post-HPE were the 3 variables with significant and independent predictive capability of native liver survival at 2 years of age. TB and albumin were not highly correlated (r = −0.3987), nor were other variables in the initial multivariate regression model. The AUROC of a logistic regression model based on these parameters was 0.95. Removing albumin from this model reduced the AUROC to 0.94. The AUROCs between these 2 models were not significantly different (P = 0.8470).
Predictive Model Based on Receiver Operating Characteristic Analysis of Total Bilirubin and Albumin
Because albumin did not appear to contribute a great deal of predictive power when combined with TB in a logistic regression model but did seem strongly and independently associated with outcome, we investigated whether albumin would be predictive in patients with lower TB levels 3 months post-HPE. ROC analysis of both variables to predict 2-year native liver survival showed optimal cut-off values (to maximize both sensitivity and specificity) of 74 μmol/L for TB and 35 g/L for albumin (Fig. 1).
A TB level >74 μmol/L at 3 months post-HPE predicts death or transplant at 2 years with sensitivity of 81% (95% CI: 72–89), specificity of 82% (95% CI: 73–88), positive likelihood ratio of 4.39 (95% CI: 2.92–6.60), and negative likelihood ratio of 0.23 (0.15–0.36). An albumin level <35 g/L at 3 months post-HPE predicts death or transplant at 2 years with sensitivity of 73% (95% CI: 63–82), specificity of 70% (95% CI: 60–79), positive likelihood ratio of 2.44 (95% CI: 1.76–3.38), and negative likelihood ratio of 0.38 (0.26–0.56). Assuming a pre-test probability of requiring transplantation by 2 years of approximately 55% (4–6), the post-test probability given a TB of >74 μmol/L at 3 months post-HPE is 84.3% (95% CI: 78.1, 89.0). Albumin level <35 g/L at 3 months post-HPE would lead to a post-test probability of 74.9% (95% CI: 68.3, 80.5).
Patients were assigned to 1 of 3 prognostic groups based on these parameters: TB ≤74 μmol/L and albumin >35 g/L; TB ≤74 μmol/L and albumin ≤35 g/L; and TB >74 μmol/L. Group 3 fared poorest, with a 2-year native liver survival rate of 20.2%; those in group 2 had an intermediate prognosis, with 2-year native liver survival of 74.4%. Infants in group 1 had the best prognosis with 92.6% surviving with native liver at 2 years of age. Native liver survival, beyond 2 years was also significantly different between these 3 groups (log rank P < 0.001; Fig. 2). Groups 1 and 2 had similar TB levels and weight z scores; however, group 2 had significantly lower length z scores 3 months post-HPE (Table 5). Of those with coagulation data available, the proportion with international normalized ratio (INR) >1.2 was similar between groups 1 and 2 (3/35 vs 3/33, P = NS).
Serum TB and albumin levels at 3 months post-HPE were independently associated with native liver survival at 2 years of age on multivariate regression analysis. A cut-off TB value of 74 μmol/L at 3 months post-HPE defines a poor prognostic group. Albumin at 3 months post-HPE can improve prognostication for infants who achieve a TB ≤74 μmol/L: those with albumin levels ≤35 g/L have an intermediate prognosis and also show early signs of growth failure with significantly lower length z scores than those with albumin levels >35 g/L who fare much better. The differences in 2-year native liver survival between these 3 groups were durable into the longer term.
Identification of BA infants who are likely to require early LTis advantageous for a number of reasons. Firstly, early aggressive nutritional management of these infants may improve pre- and post-LT outcomes. Better nutritional status in infants with BA has been associated with reduced pre- and post-LT morbidity and mortality and better neurodevelopmental outcomes (18–23). Secondly, early referral to a pediatric transplant center to facilitate evaluation and timely listing may result in the patient receiving a graft in a more optimal timeframe. Thirdly, parents and families may be more accurately counseled regarding the infant's prognosis and be more psychologically prepared for the transplantation process, which may improve parental and family functioning (24–26).
This is the largest study examining multiple early clinical and laboratory prognostic variables in infants with BA after HPE toward the derivation of optimum predictive cut-off values for these variables. A large number of potential prognostic factors have previously been examined, but rarely in combination. Decline in serum bilirubin, reflecting effective bile drainage has long been regarded as a clinical sign of successful HPE and is indeed associated with improved native liver survival (7,9,11–14). A postoperative TB level <20 μmol/L (1.2 mg/dL) at any stage predicted a better prognosis in Taiwanese infants (7). A large Japanese study found TB <17 μmol/L (1 mg/dL) at 3 months post-HPE was associated with improved native liver survival (11). A multicenter study from the United States of 104 patients found TB levels <34 μmol/L (2 mg/dL) to be associated with better native liver survival (9). Rodeck et al (10) found postoperative TB and excretion on hepatobiliary scintigraphic scan to be the strongest predictors of favorable outcome in 24 children in a multivariate Cox regression analysis (10). This small study also found that the optimal TB cut-off to predict event-free survival was 57 μmol/L (3.3 mg/dL) at 6 weeks post-HPE. With the exception of our study, this is the only other study to our knowledge that has used ROC analysis to determine an optimum cut-off TB level to predict outcome. In our cohort, the optimum derived cut-off of 74 μmol/L for TB at 3 months post-HPE has high positive predictive value (almost 80% of infants above this threshold required transplantation within the first 2 years of life). In those with TB less than this threshold, serum albumin defined 2 distinct prognostic groups.
Albumin measures a different aspect of liver health compared to TB, reflecting synthetic capabilities rather than conjugating and excretory function. Whilst albumin did not seem to strengthen a logistic regression predictive model that already contained TB at 3 months, it appeared to be particularly discriminating in infants with TB ≤74 μmol/L. Other factors influence albumin synthesis such as malnutrition, and albumin loss through enteropathy or proteinuria. We cannot be certain as to the degree each may have contributed to hypoalbuminemia in our cohort. Of note the intermediate prognostic group (group 2; TB ≤74 μmol/L, albumin ≤35 g/L) had significantly lower length z scores than the good prognostic group (group 1; TB ≤74 μmol, albumin >35 g/L) despite both groups having similar weight z scores and TB levels. This suggests that group 2 may have either been undernourished or that the low albumin level and poor length growth were both indicators of failing liver function despite relatively preserved excretion of bilirubin. The similar weight z scores 3 months post-HPE between groups 1 and 2 may be accounted for by organomegaly or edema, and group 2 did have a higher incidence of ascites by 3 months. Thus length z scores may be more reliable early indicators of liver health than weight z scores in infants with BA. Indeed, we found that preservation of length z scores in the first 3 months post-HPE was strongly associated with 2-year native liver survival in univariate analysis. Unfortunately, missing data point for length precluded its inclusion into the multivariate analysis. Growth failure post-HPE has also been associated with 2-year native liver survival by DeRusso et al (20).
Coagulation factor levels, whether measured directly or functionally such as with a prothrombin time or INR provide an alternative measure of liver synthetic capacity. In view of the variability in measurement of prothrombin time and INR between different laboratories in patients with liver disease (27), missing values (only 132/217 had an INR measurement at 3 months post-HPE recorded) and the confounding effect of vitamin K malabsorption, available measures of coagulation function were not analyzed as part of the predictive model. Despite the missing data, the similar low proportion of patients in groups 1 and 2 with INR >1.2 suggests that liver synthetic dysfunction may not be responsible for the differences in serum albumin between these groups. Clinicians should not be reassured solely by a declining bilirubin level. Implementation of escalating aggressive nutritional management (alongside early referral to a liver transplant center) should not be delayed in infants with BA with mild hypoalbuminemia.
Older age at HPE, bridging fibrosis or cirrhosis on initial liver biopsy, and the development of ascites were associated with poor outcome. These parameters, however, did not remain significant in multivariate analysis. The association between center and outcome is not easily explainable. Although the center 2 patients had an older age at HPE and a higher proportion with BASM than those from center 1, age was not significantly associated with outcome in the multivariate model and BASM was not significant on univariate analysis. Other possible explanations may include differences in surgical technique, transplant decision-making behavior, and patient ethnicity. In any case, this factor was retained in the final multivariate model and thus controlling for center, bilirubin and albumin remained the only 2 factors found to be associated with outcome.
The present study has a number of recognized limitations. Being retrospective, incomplete data limited the number of patients that could be used in analysis of each variable. This was particularly the case with serial growth measurements. We also had incomplete information on the extent of nutritional support provided post-HPE. A limitation of collecting laboratory data across centers and time periods is that methods used for analysis and reporting are not uniform. This was encountered with recording TB. Some laboratories analyzed and reported TB (which includes serum delta bilirubin level), whereas other laboratories report conjugated and unconjugated fractions, which clinical teams may often sum in the setting of absence of a true TB level being reported. The latter method does not account for delta bilirubin, which makes up a variable but significant component of TB in cholestatic infants and even those with recent but resolving cholestasis (28,29). The center 2 laboratory reported TB until 1994 when only serum conjugated and unconjugated fractions were reported. For the purposes of the present study, these fractions were added for a surrogate TB value (thus underestimating a true total). The center 1 laboratory analyzed and reported a true TB throughout the period of study. Thus our derived cut-point for TB at 3 months post-HPE is likely to be conservative. We would recommend that prospective studies standardize and report the precise fractions of bilirubin that are analyzed. It may be more precise to analyze conjugated or direct bilirubin fractions for such studies but not use these 2 interchangeably. We chose TB as this was the most consistent measure available across laboratories and time periods.
We cannot exclude the possibility that low serum albumin levels in group 2 may have influenced a decision to transplant earlier, resulting in reduced native liver survival in this group. Nonetheless, this group did have objective signs of poor liver health such as growth failure and ascites, and is likely to have received attention to nutritional care comparable to group 1.
In addition to serum TB level, other markers of liver health at 3 months post-HPE such as serum albumin and length z scores are strong predictors of both short- and long-term native liver survival in a large cohort of BA infants. Serum TB and albumin levels 3 months after HPE may be useful in defining 3 distinct prognostic groups. Serum albumin <35 g/L in BA infants who are no longer jaundiced at 3 months post-HPE should be regarded as a poor prognostic indicator and immediately prompt clinicians toward aggressive nutritional management in this group.
The authors thank Alessandra Bisquera and Patrick McElduff, Clinical Research Design, Information Technology and Statistical Support, Hunter Medical Research Institute, Newcastle, Australia, for support with statistical analysis.
1. Magee JC, Krishnan SM, Benfield MR, et al. Pediatric transplantation in the United States, 1997–2006. Am J Transplant
2008; 8 (pt 2):935–945.
2. Laurent J, Gauthier F, Bernard O, et al. Long-term outcome after surgery for biliary atresia. Study of 40 patients surviving for more than 10 years. Gastroenterology
3. Lykavieris P, Chardot C, Sokhn M, et al. Outcome in adulthood of biliary atresia: a study of 63 patients who survived for over 20 years with their native liver. Hepatology
4. Schreiber RA, Barker CC, Roberts EA, et al. Biliary atresia: the Canadian experience. J Pediatr
5. Serinet MO, Wildhaber BE, Broue P, et al. Impact of age at Kasai operation on its results in late childhood and adolescence: a rational basis for biliary atresia screening. Pediatrics
6. Davenport M, Caponcelli E, Livesey E, et al. Surgical outcome in biliary atresia: etiology affects the influence of age at surgery. Ann Surg
7. Hung PY, Chen CC, Chen WJ, et al. Long-term prognosis of patients with biliary atresia: a 25 year summary. J Pediatr Gastroenterol Nutr
8. Davenport M, Ong E, Sharif K, et al. Biliary atresia in England and Wales: results of centralization and new benchmark. J Pediatr Surg
9. Shneider BL, Brown MB, Haber B, et al. A multicenter study of the outcome of biliary atresia in the United States, 1997 to 2000. J Pediatr
10. Rodeck B, Becker AC, Gratz KF, et al. Early predictors of success of Kasai operation in children with biliary atresia. Eur J Pediatr Surg
11. Ohhama Y, Shinkai M, Fujita S, et al. Early prediction of long-term survival and the timing of liver transplantation
after the Kasai operation. J Pediatr Surg
12. Lai HS, Chen WJ, Chen CC, et al. Long-term prognosis and factors affecting biliary atresia from experience over a 25 year period. Chang Gung Med J
13. Toyoki Y, Hakamada K, Narumi S, et al. Timing for orthotopic liver transplantation
in children with biliary atresia: a single-center experience. Transplant Proc
14. Yanchar NL, Shapiro AM, Sigalet DL. Is early response to portoenterostomy predictive of long-term outcome for patients with biliary atresia? J Pediatr Surg
15. Guttman OR, Roberts EA, Schreiber RA, et al. Biliary atresia with associated structural malformations in Canadian infants. Liver Int
16. Kasai M, Sawaguchi S, Akiyama H. A proposal of new classification of biliary atresia. J Jpn Soc Pediatr Surg
17. Davenport M, Savage M, Mowat AP, et al. Biliary atresia splenic malformation syndrome: an etiologic and prognostic subgroup. Surgery
18. Chin SE, Shepherd RW, Cleghorn GJ, et al. Survival, growth and quality of life in children after orthotopic liver transplantation
: a 5 year experience. J Paediatr Child Health
19. Shepherd RW, Chin SE, Cleghorn GJ, et al. Malnutrition in children with chronic liver disease accepted for liver transplantation
: clinical profile and effect on outcome. J Paediatr Child Health
20. DeRusso PA, Ye W, Shepherd R, et al. Growth failure and outcomes in infants with biliary atresia: a report from the Biliary Atresia Research Consortium. Hepatology
21. Wayman KI, Cox KL, Esquivel CO. Neurodevelopmental outcome of young children with extrahepatic biliary atresia 1 year after liver transplantation
. J Pediatr
22. Stewart SM, Hiltebeitel C, Nici J, et al. Neuropsychological outcome of pediatric liver transplantation
23. Moukarzel AA, Najm I, Vargas J, et al. Effect of nutritional status on outcome of orthotopic liver transplantation
in pediatric patients. Transplant Proc
24. Tarbell SE, Kosmach B. Parental psychosocial outcomes in pediatric liver and/or intestinal transplantation: pretransplantation and the early postoperative period. Liver Transpl Surg
25. Kennard BD, Petrik K, Stewart SM, et al. Identifying factors in post-operative successful adaptation to pediatric liver transplantation
. Soc Work Health Care
26. Young GS, Mintzer LL, Seacord D, et al. Symptoms of posttraumatic stress disorder in parents of transplant recipients: incidence, severity, and related factors. Pediatrics
2003; 111 (pt 1):e725–e731.
27. Lisman T, Van Leeuwen Y, Adelmeijer J, et al. Interlaboratory variability in assessment of the model of end-stage liver disease score. Liver Int
28. Ito F, Ando H, Watanabe Y, et al. Serum bilirubin fractions in cholestatic pediatric patients: determination with Micronex high-performance liquid chromatography. J Pediatr Surg
29. Ye W, Rosenthal P, Magee JC, et al. Factors determining delta-bilirubin levels in infants with biliary atresia. J Pediatr Gastroenterol Nutr