Journal of Pediatric Surgery 50 (2015) 363–370

Contents lists available at ScienceDirect

Journal of Pediatric Surgery

journal homepage: www.elsevier.com/locate/jpedsurg

Original Articles

A screening algorithm for the efficient exclusion of biliary atresia in

infants with cholestatic jaundice
Tim Jancelewicz a,e,⁎, Rebecca Barmherzig a,e, Catherine T.-S. Chung b,f, Simon C. Ling c,f, Binita M. Kamath c,f,
Vicky L. Ng c,f, Joao Amaral d, Constance O’Connor c,f, Annie Fecteau a,e, Jacob C. Langer a,e
a
Division of Pediatric Surgery, The Hospital for Sick Children, Toronto, Ontario, Canada
b
Division of Pathology and Pediatric Laboratory Medicine, The Hospital for Sick Children, Toronto, Ontario, Canada
c
Division of Gastroenterology, Hepatology and Nutrition, The Hospital for Sick Children, Toronto, Ontario, Canada
d
Department of Diagnostic Imaging, Division of Image Guided Therapy, The Hospital for Sick Children, Toronto, Ontario, Canada
e
Department of Surgery, University of Toronto, Toronto, Ontario, Canada
f
Department of Pediatrics, University of Toronto, Toronto, Ontario, Canada

a r t i c l e i n f o a b s t r a c t

Article history: Background: Neonates with cholestasis may undergo many tests before biliary atresia (BA) or an alternative di-
Received 5 March 2014 agnosis is reached, and delayed intervention may worsen outcomes. An optimal diagnostic approach to reduce
Received in revised form 31 July 2014 risk, cost, and delay has yet to be defined. The purpose of this study was to develop an algorithm that rapidly
Accepted 16 August 2014 and accurately excludes BA for infants with cholestatic jaundice.
Methods: A single-center retrospective comparison of diagnostic workup was made between cholestatic infants
Key words:
with BA, and those without BA who underwent hepatobiliary iminodiacetic acid (HIDA) scan during admission.
Biliary atresia
Neonatal
Patients were born between 2000 and 2010 and those older than 100 days at assessment were excluded. Sensi-
Cholestasis tivity and specificity analysis of predictive variables was performed and an algorithm constructed.
Cholestatic jaundice Results: There were 45 BA and 167 non-BA patients. Some variables were 100% sensitive for the exclusion of BA:
conjugated bilirubin b2.5 mg/dL, gamma-glutamyl transpeptidase b 150 U/L, excretion on HIDA, or a normal per-
cutaneous cholangiogram. Clinical variables and ultrasound were less useful as screening tests owing to low
specificity and sensitivity, respectively. Liver biopsy was 98% sensitive and 84% specific in the diagnosis of BA.
An algorithm was constructed that rules out BA with a negative laparotomy rate of 3–22%.
Conclusion: We propose a screening algorithm for infants with conjugated hyperbilirubinemia that permits effi-
cient exclusion of BA with minimal invasive testing and with a low risk of negative laparotomy. This algorithm
now requires prospective evaluation to determine its diagnostic accuracy and its ability to reduce hospital
costs, patient morbidity, and time to Kasai portoenterostomy in patients with BA.
© 2015 Elsevier Inc. All rights reserved.

Biliary atresia (BA) is a common cause of neonatal conjugated
hyperbilirubinemia. Patients may undergo a wide range of tests before
BA or an alternative diagnosis is reached, and delay in diagnosis beyond
60–100 days of age may reduce longevity of the native liver and increase
morbidity in infants with BA, although the ideal surgical timing remains
controversial [1–3]. Late diagnosis also worsens outcomes in other con-
ditions associated with neonatal cholestasis, such as hypopituitarism,
galactosemia, and tyrosinemia [4,5]. The optimal diagnostic approach
to reduce risk, unnecessary testing, and delay prior to definitive diagno-
sis or surgery has yet to be completely defined.
The purpose of this study was to compare the diagnostic evaluation
of infants with BA to that of infants with both non–BA-associated chole-
stasis and an unclear diagnosis at presentation, in order to develop an
efficient approach for the exclusion of BA. We hypothesized that for
⁎ Corresponding author at: Le Bonheur Children's Hospital, 51 North Dunlap St. Suite
230, Memphis, TN, 38112. Tel.: +1 901 545 9973.
E-mail address: tjancele@uthsc.edu (T. Jancelewicz).
some patients there is superfluous workup performed prior to surgical
intervention, and our goal was to develop a streamlined algorithm
that allows rapid, objective, and accurate clinical assessment for infants
with cholestatic jaundice, and reduced time to operation for patients
with BA.
1. Patients and methods
This study was approved by the Research Ethics Board of The Hospi-
tal for Sick Children, University of Toronto (1000017943).
1.1. Patient selection and endpoints
A retrospective review was performed of all infants born between
January 1, 2000 and June 1, 2010 who underwent a Kasai
portoenterostomy for BA (confirmed at operation and by pathology)
and who were ≤ 100 days of age at time of initiation of workup. The
comparison cohort consisted of patients born during the same time

http://dx.doi.org/10.1016/j.jpedsurg.2014.08.014
0022-3468/© 2015 Elsevier Inc. All rights reserved.
364 T. Jancelewicz et al. / Journal of Pediatric Surgery 50 (2015) 363–370
period who did not have BA, who presented with cholestatic jaundice at
≤100 days of age, and who underwent hepatobiliary iminodiacetic acid
(HIDA) scan during their workup. As a selection criterion, HIDA scan
was considered a marker of clinician concern for a possible diagnosis
of BA, and was chosen to identify a subset of infants whose presentation
was clinically suspicious for extrahepatic obstruction.
Patients were excluded if records were incomplete (e.g. study re-
ports or laboratory values not available) or if they died before a diagno-
sis was reached. The primary endpoints were diagnostic operative
exploration with confirmation of BA, diagnosis of and/or treatment for
a non-BA cholestatic condition, or resolution of cholestasis without
targeted therapy or definitive diagnosis.
1.2. Data acquisition
Clinical data were gathered including patient demographics, birth
data, concomitant conditions, ultimate diagnosis, operative data, and rel-
evant medical history (age of onset of jaundice, age at time of workup for
cholestasis, and stool color as assessed by a physician at time of workup).
For serum laboratory measurements, the lowest levels seen within
the first 24 hours of initial admission (or, if already admitted, at time
of workup for cholestasis) were recorded, in order to subsequently gen-
erate stringent cutoff values; variables were conjugated bilirubin (CB),
unconjugated bilirubin (UB), gamma-glutamyl transpeptidase (GGT),
alkaline phosphatase (ALP), aspartate aminotransferase (AST), alanine
transaminase (ALT), and albumin.
Abdominal ultrasound (AUS) reports were reviewed and the follow-
ing findings, if noted, were documented: presence or absence of gall-
bladder (GB), size and shape of GB, intrahepatic or extrahepatic duct
dilation, liver echogenicity, triangular cord sign, size and number of
spleen(s), and any other salient diagnostic feature (e.g. presence of a
choledochal cyst).
HIDA scan reports were reviewed and results were recorded as ei-
ther “draining” or “non-draining”, in reference to the presence or ab-
sence of radiolabeled tracer in the small intestine within 24 hours of
injection. Number of days of pre-HIDA phenobarbital treatment was
also recorded. At our institution, HIDA scans were considered complete
by the radiologist only if radiotracer uptake was sufficient for interpre-
tation within 24 hours.
Percutaneous transhepatic cholecysto-cholangiogram (PC) reports
were recorded as “diagnostic of BA” (adequate cannulation with contrast
injection, and absent extrahepatic biliary drainage), “not BA” (patent ex-
trahepatic and intrahepatic bile ducts), or “inconclusive” if the biliary tree
could not be cannulated, or if a study could not be performed to the
radiologist's satisfaction and/or images could not be definitively
interpreted by the radiologist. Studies were performed using gallbladder
cannulation with a standard technique [6]. Intraoperative cholangiogram
findings were also reviewed but only used for endpoint determination.
For percutaneous core liver biopsy findings, the contemporaneous
pathologist's interpretation was categorized as “consistent with large
duct obstruction”, “not consistent with large duct obstruction”, or “inde-
terminate or cannot exclude BA.” Also recorded were degree of bile duct
proliferation (none, mild, moderate, or severe), stage of portal fibrosis
(stage 0–4), and the presence or absence of bile duct plugs, giant cell
transformation, hepatocellular ballooning, and portal inflammation
[7]. Intraoperative biopsy findings obtained at time of open cholangio-
gram and/or Kasai were also documented for endpoint determination,
but were not used as outcomes predictors.
1.3. Statistical analysis
For univariate comparison between cohorts and subgroups, Fisher's
exact test was used for categorical variables and the Wilcoxon rank-sum
test for continuous variables. Some laboratory values were dichoto-
mized into normal and abnormal values using cutoff thresholds
generated by us to enable use of these continuous variables as predic-
tors in subsequent analysis.
In order to generate a diagnostic algorithm, sensitivity and specific-
ity and positive and negative predictive values (PPV and NPV) were
used to generate some steps in the pathway. To assess for independent
predictors, multivariate logistic regression analysis was performed on
relevant significant variables using stepwise forward selection.
Statistics and graphing functions were performed using Stata (StataCorp
LP, College Station, Texas). A P value of b 0.05 was considered significant.
2. Results
2.1. Cohort characteristics
There were 45 BA and 167 non-BA patients, after exclusions for
missing information (2 BA and 9 non-BA patients). All BA patients
underwent Kasai portoenterostomy with diagnosis confirmed by intra-
operative findings and final pathology. In the non-BA group, seven (4%)
underwent diagnostic laparotomy with operative cholangiogram. Ulti-
mate diagnoses are shown in Table 1 and clinical characteristics at pre-
sentation are in Table 2. Among clinical features, by logistic regression,
presence of acholic stools yielded the highest odds ratio (OR) for BA.
With the exception of malrotation, heterotaxy, and polyspenia, the
presence of major concomitant diagnoses and comorbidities such as
prematurity was more common in non-BA patients (Tables 2 and 3).
2.2. Laboratory values
Laboratory values that were significantly different between groups in-
cluded UB, albumin, and GGT (Table 4). Several patients in the non-BA
group had low CB levels because the CB decreased within the first
24 hours after presentation, and the lowest value was taken for analysis
(as per our methods). Among BA patients, the lowest CB level seen was
2.7 mg/dL. BA patients had CB levels that fell within a relatively narrow
range as compared with the non-BA group (Fig. 1A), irrespective of age
at presentation (Fig. 1B). GGT levels were consistently higher among BA
patients, and increased with time (Fig. 1C and D). However, a few BA pa-
tients had only mildly elevated GGT levels; the lowest level seen in the BA
cohort was 150 international units (U) per liter (L) (normal 0–45 U/L).
2.3. Imaging studies
Among all patients who underwent AUS (Table 5), 64% had an ab-
normally “small”, “contracted”, or “absent” GB. One non-BA patient
had an ultrasound report that described the GB as “absent”. Other ultra-
sound features, including the triangular cord sign, were inconsistently
mentioned in radiologist reports and therefore were not useful as inde-
pendent predictors.
Table 1
Ultimate diagnoses in the non-BA cohort.
Non-BA patients
(N = 167)
Idiopathic neonatal hepatitis – INH (%) 97 (58)
Parenteral nutrition (PN) cholestasis 38 (23)
Alagille syndrome 8 (5)
Cytomegalovirus (CMV) hepatitis 6 (4)
Alpha-1 antitrypsin (A1AT) deficiency 5 (3)
Progressive familial intrahepatic cholestasis 3 (2)
Niemann–Pick type C 2 (1)
Congenital portosystemic shunt 2 (1)
Other [arthrogryposis–renal dysfunction–cholestasis (ARC) 6 (4)
syndrome, Budd–Chiari syndrome, neonatal sclerosing
cholangitis, citrin deficiency, choledochal cyst,
unspecified mitochondrial cytopathy]
BA – biliary atresia.
 T. Jancelewicz et al. / Journal of Pediatric Surgery 50 (2015) 363–370 365

Table 2
Cohort clinical characteristics.

All patients (N = 212) BA (N = 45) Non-BA (N = 167) OR 95% CI

Male sex (%) 126 (59) 16 (36) 110 (66) 0.3 0.1–0.6
Median GA birth, weeks (IQR) 38 (35–39) 39 (38–40) 38 (35–38) 1.4 1.2–1.7
Mean birth weight, kg (SD) 2.8 (1.0) 3.2 (0.7) 2.7 (1.1) 1.0 1.0–1.0
Acholic stools reported (%) 72 (36) 40 (89) 32 (21) 31 11–84
Median age at presentation, days (IQR) 28 (2–56) 61 (44–71) 17 (2–42) 1.0 1.0–1.1
Median age jaundice onset, days (IQR) 2 (1–14) 2 (1–17) 2 (1–14) – –
Other major comorbid diagnoses (%) 81 (38) 5 (11) 76 (46) 0.1 0.1–0.4

BA – biliary atresia; GA – gestational age; OR – odds ratio; CI – confidence interval.

Table 3 which led to subsequent laparotomy with an eventual diagnosis of idi-

List of the most common patient comorbidities. opathic neonatal hepatitis.
All patients BA Non-BA P
(N = 212) (N = 45) (N = 167) 2.4. Liver biopsy
Prematurity (%)⁎ 55 (26) 1 (2) 54 (32) b0.001
Cardiac anomaly (%)⁎⁎ 30 (14) 1 (2) 29 (17) b0.001 Percutaneous liver biopsy was obtained in 45% of all patients, in 89%
Trisomy 21 (%) 14 (7) 0 (0) 14 (8) - of BA patients, and in 34% of non-BA patients (univariate analysis
Intestinal atresia (%) 11 (5) 0 (0) 11 (7) -
Table 6). There were no major complications in any patient (i.e. hemor-
Polysplenia (%) 6 (3) 5 (11) 1 (1) 0.002
Heterotaxy (%) 6 (3) 4 (9) 2 (1) 0.03 rhage requiring transfusion or visceral injury). Corrected median age at
Malrotation (%) 4 (2) 2 (4) 2 (1) 0.2 biopsy was only slightly higher for BA patients (69 days) than non-BA
⁎ GA b 37 at birth. patients (56 days, P = 0.05). Operative biopsy was obtained in 14 BA
⁎⁎ Including ventricular septal defect, patent ductus arteriosus, or other major structural patients (31%) and three non-BA patients (2%). Twelve BA patients
anomalies. (27%) and one non-BA patient (b1%) had both a percutaneous and
operative biopsy.
Nearly all patients had an interpretable HIDA scan (Table 5). Four BA The pathologist's interpretation was consistent with large duct ob-
patients (9%) went to surgery without a HIDA scan, and 6 non-BA pa- struction in 98% of BA patients and 16% of non-BA patients. Moderate
tients (4%) had a HIDA scan that the radiologist deemed incomplete to severe bile duct proliferation, stage 2–4 portal fibrosis, bile duct
owing to poor radiotracer uptake. One-quarter of non-BA patients had plugs, and portal inflammation were significantly more common
a nondraining scan, and no BA patient had a draining scan. All but six pa- among BA patients. Mild bile duct proliferation was seen in 3 BA pa-
tients received phenobarbital prior to HIDA scan; there was no differ- tients (8%), and those patients underwent biopsy at 32, 50, and 58
ence in the number of days of phenobarbital treatment between BA corrected days of age.
and non-BA patients (median 5 days, P = 0.09), and between patients By multivariate analysis using the four significant pathologic criteria,
with draining scans and patients with nondraining scans (median moderate to severe bile duct proliferation was the most significant inde-
5 days, P = 0.5). For non-BA patients older than 6 weeks at presenta- pendent predictor of BA, with an OR of 14 (95% confidence interval [CI]
tion, 90% of HIDA scans showed drainage, versus 68% of scans in those 3–62, P = 0.001). Stage 2–4 portal fibrosis was also an independent pre-
less than 6 weeks of age. dictor of BA (OR 9, CI 2–36, P = 0.002). When added as a variable into
PC was attempted in 47 patients (22%). Date of PC versus date of bi- the regression model, the pathologist's final interpretation of “consis-
opsy was not analyzed, though nearly all patients undergoing PC had al- tent with large duct obstruction” was a better independent predictor
ready undergone biopsy (Fig. 2). PC was more likely to be inconclusive for BA than any single pathologic finding or combination of findings
among BA patients (9 BA vs. 3 non-BA patients, or 75% BA). Correspond- (OR 90, CI 6–1416, P = 0.001).
ingly, a small or absent GB was noted on AUS in 10 of the 12 (83%) in-
conclusive PCs. Operative cholangiogram was performed in 35 BA 2.5. Subgroup analysis
patients (78%) and 7 non-BA patients (4%); both a preoperative and
an operative cholangiogram were attempted in 11 BA patients (24%) Fig. 2 demonstrates the diagnostic pathways observed for patients
and three non-BA patients (2%). There were no false positive or false who had a nondraining HIDA scan, which is a clinical picture that was
negative results, although one non-BA patient underwent operative particularly concerning for BA and which included 41 patients (91%)
cholangiogram despite a PC showing drainage into the duodenum from the BA cohort (four BA patients underwent Kasai without HIDA
(see Fig. 2); for this patient, the percutaneous study did not opacify scan). As shown, there was no consistent diagnostic pathway employed,
the intrahepatic ducts, causing concern for isolated hepatic duct atresia, and depended on clinician preference at the time. One non-BA patient
with both a nonsuspicious biopsy and a cholangiogram that showed
drainage into the duodenum underwent a diagnostic laparotomy
Table 4 owing to concern for isolated hepatic duct atresia (diagnosis was idio-
Median serum laboratory values at presentation (IQR). pathic neonatal hepatitis). Of the nine patients in the subgroup with
All patients BA Non-BA P an inconclusive cholangiogram, all underwent laparotomy; seven had
(N = 212) (N = 45) (N = 167) BA (78%) and two did not.
GGT (IU/L) 290 (157–621) 612 (485–1057) 203 (119–452) b0.001
Albumin (g/L) 32 (26–36) 35 (30–38) 30 (24–36) b0.001 2.6. Algorithm generation
UB (mg/dl) 2.2 (1.3–8.4) 1.8 (1.3–2.3) 2.6 (1.3–10.5) 0.02
CB (mg/dl) 5.0 (3.2–7.0) 4.6 (4–5.7) 5.1 (2.6–7.6) 0.7
Most steps in the diagnostic algorithm (Fig. 3) were defined by the
ALP (IU/L) 547 (400–731) 540 (408–699) 550 (394–746) 0.9
AST (IU/L) 197 (113–380) 198 (144–308) 195 (108–408) 0.9 identification of diagnostic tests that were highly (100%) sensitive
ALT (IU/L) 133 (70–256) 126 (89–193) 135 (61–256) 0.6 and/or specific for BA, or for its definitive exclusion (Table 7).
BA – biliary atresia; GGT – gamma-glutamyl transpeptidase; UB – unconjugated bilirubin;
The lowest GGT and CB levels seen in a patient with BA (150 U/L and
CB – conjugated bilirubin; ALP – alkaline phosphatase; AST – aspartate aminotransferase; 2.7 mg/dL) were used to generate threshold values that excluded a BA
ALT – alanine transaminase. diagnosis. A slightly lower CB cutoff value (2.5 mg/dL) was used to
366 T. Jancelewicz et al. / Journal of Pediatric Surgery 50 (2015) 363–370

(A) (B)

50
50

conjugated bilirubin (mg/dL)

conjugated bilirubin (mg/dL)
p=0.7

40
40

30
30

20
20
10

10
0

0
(-)BA (+)BA 0 20 40 60 80 100
age at presentation (days)

(C) (D)
3,000

3000
p=<0.001
2,000

2000
GGT (IU/L)
GGT (IU/L)
1,000

1000
0

(-)BA (+)BA
0 20 40 60 80 100
age at presentation (days)

Fig. 1. Serum laboratory values for CB (mg/dL) and GGT (IU/L). The lowest value seen in the first 24 hours of presentation was used. A and C: CB and GGT box plots for (+)BA and (−)BA
patients, median with 25–75% intraquartile range (IQR, boxes) and the highest and lowest data points within 1.5 IQR (bars). Dashed lines represent upper limits of normal (CB: 0.5 mg/dL;
GGT: 45 IU/L). B and D: CB and GGT values plotted against age at presentation; open circles are (−)BA patients (dashed line is the trending mean), and triangles are (+)BA patients (solid
line is the trending mean).

include a larger proportion of the non-BA population. Using these levels,
80 patients (38%) did not meet the threshold lab criteria and did not
have BA, with a sensitivity and NPV of 100% (CI 0.92–1). For comparison,
the presence of acholic stools as a predictor was only 89% sensitive for
BA (CI 0.77–0.95) with an NPV of 96% (CI 0.91–0.98).
A draining HIDA scan excluded 120 non-BA patients (72%) with an
NPV and sensitivity of 100% (CI 0.91–1) but specificity of only 75% (CI
0.67–0.81, Table 7). Absence of GB on AUS was highly specific but not
sensitive for BA; in contrast, PC was 100% sensitive and 100% specific
when it could be completed successfully.
A pathologic interpretation favoring large duct obstruction was
highly sensitive for BA but was not as specific. As an isolated test, liver
biopsy did not reliably exclude BA unless bile duct proliferation was
completely absent, which occurred in only 20 (36%) of non-BA patients.
In order to assess the utility of diagnostic tests in the subgroup of pa-
tients with clinical findings highly suspicious for BA, sensitivity/specific-
ity calculations were regenerated on the group of patients with above-
threshold CB/GGT, and a nondraining HIDA scan (Table 7, right). In
that subgroup, absence of GB on AUS was 100% specific for BA (CI
0.87–1), and there was a decrease in the specificity of a pathologic diag-
nosis that favored BA (78%, CI 0.55–0.91). The specificity for BA of all
clinical predictive variables, such as acholic stools and absence of co-
morbidities, was consistently low. PC, when completed successfully,
remained a perfect test for BA.
Based on these data, the diagnostic algorithm was constructed
(Fig. 3). Workup for other sources of elevated CB should proceed con-
currently or sequentially with BA exclusion using this algorithm accord-
ing to individual patient circumstances, as it does not address pathways
for the identification of non-BA causes of cholestasis. The initial workup
of patients with cholestatic jaundice includes an AUS, screening for pos-
sible metabolic and genetic conditions, and laboratory tests and cultures
(see Fig. 3 legend for details) [5]. After the initial workup, many patients
(N = 80, 38% in our cohort) will be excluded from a diagnosis of BA and
another diagnosis should be pursued (HIDA scan is not necessary in

Table 5
Imaging findings.

All patients (N = 212) BA (N = 45) Non-BA (N = 167) P

Abdominal ultrasound performed (%) 192 (90) 45 (100) 147 (88) NC

Gallbladder absent (%) 15 (8) 14 (31) 1 (1) b0.001
Gallbladder absent, contracted, or small (%) 123 (64) 38 (84) 85 (58) 0.001
HIDA scan completed (%) 202 (95) 41 (91) 161 (96) NC
No duodenal drainage seen (%) 82 (41) 41 (100) 41 (25) b0.001
Percutaneous cholangiogram attempt (%) 47 (22) 19 (42) 28 (17) b0.001
Percutaneous cholangiogram completed (%) 35 (74) 10 (53) 25 (89) b0.001
Diagnostic of BA (%) 10/35 (29) 10/10 (100) 0/25 (0) b0.001

BA – biliary atresia; NC – not calculated, irrelevant; HIDA – hepatobiliary iminodiacetic acid.

 T. Jancelewicz et al. / Journal of Pediatric Surgery 50 (2015) 363–370 367

Fig. 2. Outcomes for patients who had a nondraining HIDA scan (N = 82). Most (64/82, 78%) underwent percutaneous liver biopsy; 29/82 (35%) underwent percutaneous cholangiogram. 48/
82 (59%) underwent diagnostic laparotomy, yielding a 15% (7/48) negative laparotomy rate. One patient with a cholangiogram that did not show proximal duct filling, and who had a biopsy
interpretation not consistent with BA, underwent laparotomy and did not have BA. Another patient underwent laparotomy despite a biopsy inconsistent with BA, and did not have BA.

these patients). In contrast, the 62% of patients with no obvious diagno-
sis and with a high GGT/CB should undergo HIDA scan, to exclude those
with drainage into the duodenum (NPV 100%). With a nondraining
scan, the initial AUS should be reexamined for absence of a GB, which
should prompt operative cholangiogram and Kasai for BA given the
100% PPV for that finding in this setting.
For patients still without a disposition, namely those with a “normal”
or “small” (not “absent”) GB on AUS and a nondraining HIDA scan (N =
54, Fig. 3), the patient's clinical status should be reassessed, and one of
two options may be considered. If the diagnosis is unclear (e.g. owing
to comorbidities or a history of normal stools), patients with no contra-
indication should ideally undergo PC, since it is 100% sensitive and spe-
cific when completed successfully (Option 1, Fig. 3). If PC is inconclusive,
the next step should still be diagnostic laparotomy and operative
cholangiogram since most of these patients will have BA [7/9 (78%), or
a negative laparotomy rate of 22%]. Alternatively, if after a nondraining
HIDA scan, a diagnosis of BA is clinically suspected, proceeding directly
to operative cholangiogram without PC is reasonable (Option 2, Fig. 3)
because the negative laparotomy rate will be low: of the 37 patients
with high GGT/CB, a nondraining HIDA, and acholic stools, 36 had BA
and one did not, yielding a PPV of 97% (CI 0.86–1), or, in other words,
a negative laparotomy rate of 3%.
When a percutaneous liver biopsy was performed in patients with
high GGT/CB and a nondraining HIDA (N = 40), there were 3 discordant
cases (8%; two non-BA patients had a biopsy favoring large duct ob-
struction, and one BA patient had a nonsuspicious biopsy). In addition,
none of these 40 patients had absent bile duct proliferation on biopsy,
which could have excluded BA based on the NPV of 100% for that

Table 6
Percutaneous (preoperative) liver biopsy findings (univariate analysis).

All patients (N = 212) BA (N = 45) Non-BA (N = 167) OR 95% CI

Percutaneous liver biopsy obtained (%) 96 (45) 40 (89) 56 (34)

Corrected age at biopsy⁎, days (IQR) 62 (45–75) 69 (50–79) 56 (44–70)
Pathologist final diagnosis (%):
Consistent with large duct obstruction 48 (50) 39 (98) 9 (16) 204 25–1678
Indeterminate or cannot exclude BA 21 (22) 0 (0) 21 (54) 34 4–263
Inconsistent with large duct obstruction 27 (28) 1 (3) 26 (30) 0.03 0–0.2
Bile duct proliferation (%):
Any degree 70 (73) 39 (98) 31 (55) 31 4–245
Moderate or severe 28 (29) 21 (53) 7 (13) 8 3–21
Present but severity not stated 21 (22) 15 (38) 6 (11) 5 2–14
Mild or minimal 21 (22) 3 (8) 18 (32) 0.2 0.0–0.6
Absent 20 (21) 0 (0) 20 (36) NC NC
Not mentioned in report 6 (6) 1 (3) 5 (9) 0.3 0.0–2.3
Portal fibrosis (%):
Stage 2–4 38 (40) 31 (78) 7 (13) 24 8–71
Stage 1 43 (45) 7 (18) 36 (64) 0.1 0–0.3
Absent 13 (14) 1 (3) 12 (21) 0.1 0–0.8
Not mentioned in report 2 (2) 1 (3) 1 (2) 0.4 0.1–23
Bile duct plugs (%) 60 (63) 34 (85) 26 (46) 7 3–19
Portal inflammation (%) 36 (38) 21 (53) 15 (27) 3 1–6
Giant cell transformation (%) 57 (59) 21 (53) 36 (64) 0.7 0.3–1.5
Hepatocellular ballooning (%) 13 (14) 6 (15) 7 (13) 1.2 0.4–4.0

BA – biliary atresia NC – not calculable owing to perfect prediction.

⁎ Age corrected if gestational age ≤36 weeks at birth.
368 T. Jancelewicz et al. / Journal of Pediatric Surgery 50 (2015) 363–370

Fig. 3. Proposed algorithm for exclusion of BA in infants under 100 days of age who present with cholestasis. See text for confidence intervals. N values for each step in the pathway refer to
the number of patients from the study cohort who matched those characteristics. ⁎Workup for other sources of elevated CB should proceed concurrently/sequentially with BA exclusion, as
the algorithm does not address the identification of non-BA causes of cholestasis. Standard workup includes: (1) ultrasound to exclude choledochal cyst, bile duct perforation, or other
anomalies; (2) laboratory tests including liver and coagulation profiles, ammonia, electrolytes, glucose, urinalysis, TORCH workup, cultures; and (3) screening for cystic fibrosis, hypothy-
roidism, galactosemia, and other metabolic and genetic conditions [5]. †PPV increases to 100% with the presence of acholic stools and/or absence of comorbidities.

predictor. Therefore, percutaneous biopsy was not a useful discrimina-
tor prior to surgery for patients at this stage of workup, and in fact
could be misleading; hence it has not been included as a required step
at this point in the algorithm. Liver biopsy, however, is an important
test for the confirmation of other causes of cholestasis and should be ob-
tained if indicated when an alternate diagnosis is being pursued in par-
allel or in sequence with the BA algorithm.
3. Discussion
Newborns with cholestatic jaundice may undergo a battery of initial
tests, as the clinician attempts to rapidly exclude BA while simultaneously
working up alternate causes. We hypothesized that in many cases there
are superfluous tests performed in these infants which might delay defin-
itive treatment for BA, and that it is possible to definitively exclude BA
without invasive studies. We have shown that a diagnosis of BA can be ef-
ficiently excluded if a simple series of sequential clinical observations are
made. The algorithm we have developed has only four major decision
points (GGT and CB, AUS, HIDA scan, and PC), and could lead to a shorter
time to therapy, fewer expensive and potentially harmful tests, and a low
negative laparotomy rate (3–22%). Prospective validation of this algo-
rithm is now required prior to application to all patient populations.
In 2004, the North American Society for Pediatric Gastroenterology,
Hepatology and Nutrition (NASPGHAN) published an algorithm for the
workup of infants with cholestasis [1]. This algorithm appropriately fo-
cuses on avoiding unhelpful tests and advocates expeditious triage of
patients with high suspicion for BA and other time-dependent diagno-
ses. We studied the subgroup of patients with definitive evidence of
biliary obstruction by preselecting patients who underwent HIDA
scan, which itself is a test not often recommended owing to low speci-
ficity [8]. Our algorithm might reasonably be seen as a detailed supple-
ment to the later steps provided in the NASPGHAN algorithm. Omission
of cholestatic infants whose diagnostic work-up did not include a HIDA
scan introduces an unquantified selection bias. However, we believe
that this comparison group represents cholestatic infants who generat-
ed a diagnostic quandary for the clinician, since by its nature a HIDA
scan is typically obtained in an attempt to exclude extrahepatic biliary
obstruction from the differential diagnosis.
Our major finding was that the decision to operate on a patient with
cholestasis can be made reliably without the use of percutaneous liver
biopsy, and in some cases without PC. Both of these tests are invasive
and therefore associated with possible morbidity. In addition, the time
required to arrange, process, and interpret a liver biopsy introduces
delay in diagnosis and decision to proceed to laparotomy. In many
cases, the clinical picture of BA will be convincing and the AUS and
HIDA scan, along with laboratory findings, will be enough evidence to
decide whether or not to proceed with diagnostic laparotomy or lapa-
roscopy, at which time a liver biopsy can be obtained.
Ultrasound is an operator-dependent modality that has some utility in
the diagnosis of BA, including the triangular cord sign and various
methods of evaluating GB size, shape, and contraction [9–14]. We found
inconsistency in the use of these findings by our radiologists, and the
only reliable predictor was absence of a GB, probably because it is a less
subjective feature of AUS. Therefore, we found that initial AUS serves
two roles in the workup of BA: identification of alternate causes of extra-
hepatic obstruction (e.g. choledochal cyst), and identification of which
 T. Jancelewicz et al. / Journal of Pediatric Surgery 50 (2015) 363–370 369

patients may be optimal candidates for PC versus operative cholangio-

Sensitivity for BA, % Specificity for BA, %

gram owing to the presence or absence of a GB.

78 (0.55–0.91)
68 (0.48–0.83)
65 (0.46–0.81)

100 (0.87–1)
100 (0.65–1)
HIDA scan was useful mainly to exclude a subset of non-BA patients,

Patients with GGT N150 U/L, CB N2.5 mg/dL, and no drainage on HIDA, N = 67 (32%)
and not to diagnose BA. This observation has been demonstrated in

(95% CI)
multiple prior reports [15,16], and well-summarized in a recent meta-
analysis that yielded a pooled sensitivity of 98.7% (range 98.1–99.2%)

-
and specificity of 70.4% (range 68.5–72.2%) of a nondraining HIDA
scan for BA [17]. Premedication with a drug that augments hepatocyte

88 (0.75–0.95)
88 (0.74–0.95)

32 (0.2–0.47)
function and bile flow (such as phenobarbital) may improve specificity

97 (0.86–1)
100 (0.7–1)
by increasing the likelihood of drainage in patients with conditions such
(95% CI)
as neonatal hepatitis, but we did not see a significant effect from
-
premedication in our study; further work is necessary to assess the util-

-
ity of premedication for HIDA, as currently it represents the greatest

48 (0.35–0.61)

93 (0.70–0.99)
77 (0.57–0.9)
77 (0.57–0.9)
contributor to diagnostic delay in the algorithm (median 5 days). The

100 (0.7–1)
specificity of HIDA decreases with younger age of patients at time of
(95% CI)
NPV, %

study [18], and indeed we saw that younger patients were more likely
to have a nondraining scan. Importantly, some studies have suggested
-

that premature infants may have different results when tested not only
80 (0.66–0.89)

90 (0.76–0.96)
Sensitivity for BA, % Specificity for BA, % N (% of tested) PPV, % (95% CI)

82 (0.68–0.9)

using HIDA scan, but also using GGT [19,20], and possibly liver biopsy
100 (0.77–1)
100 (0.65–1)

[21]. Thus, further study is required to assess whether our algorithm is ap-
plicable to premature infants in general, as our report is underpowered to
examine this subgroup (only one BA patient was premature).
-

Cholangiogram remains the gold standard for exclusion and diagno-

sis of BA, but we found it was often difficult or impossible to complete
percutaneously in patients with BA owing to the small or absent GB.
13 (19)

39 (72)
44 (67)
45 (67)

9 (56)

In fact the inability to percutaneously cannulate the biliary tree is highly

suggestive of BA, and our interventional radiologists (JA) recommend
-

proceeding to diagnostic laparotomy after an inconclusive PC. Since it

84 (0.72–0.91)
33 (0.26–0.42)
79 (0.72–0.85)
46 (0.38–0.53)
75 (0.67–0.81)

was reported in BA patients in 1988, PC has become a valuable preoper-

99 (0.96–1)
100 (0.87–1)

ative test in this setting, and its safety and ease-of-use have improved
[6,22]. Though invasive, improved technique by interventional radiolo-
(95% CI)

gists has made this modality possible even in patients with a small GB,
and PC may exclude BA in up to 47% of infants studied [23]. In some se-
lected centers, endoscopic retrograde cholangiopancreatography
89 (0.77–0.95)
89 (0.77–0.95)

(ERCP) is done as an alternative to PC and there are data supporting

31 (0.2–0.46)

98 (0.87–1)
100 (0.92–1)

100 (0.91–1)

100 (0.72–1)

its feasibility and accuracy in the pediatric population [24,25]. ERCP

may be an acceptable alternative to PC in the workup of BA but this
(95% CI)

was not assessed in the current study, since ERCP is rarely performed
at our institution.
Liver biopsy has been a central component of the workup of patients
96 (0.91–0.98)
94 (0.86–0.97)

82 (0.76–0.87)
NPV, % (95% CI)

50 (0.39–0.61) 100 (0.97–1)

100 (0.72–1)
98 (0.89–1)
34 (0.27-0.43) 100 (0.92-1)

with cholestasis [21], but has been reported to be inconsistently useful

in the diagnosis of BA [26]. A pitfall in the interpretation of liver biopsies
in this clinical setting is that BA and many other cholestatic conditions
often show overlapping histologic features [27]. Furthermore, biopsies
Predictive values and sensitivity/specificity of selected significant variables for BA.

taken early in the course of disease may show only mild or subtle chang-
56 (0.44–0.66)
31 (0.23–0.39)
N (% of tested) PPV, % (95% CI)

81 (0.68–0.9)
93 (0.7–0.99)
100 (0.87–1)

es of BA, thus making it difficult to confirm a diagnosis of BA. In addition,

percutaneous liver biopsies may not be readily available at all centers,
All patients (N = 212)

and multiple biopsies may be required to secure an adequate sample,

leading to increased risk to the patient. The NASPGHAN review found
that 50–99% of patients with BA are correctly identified with biopsy,
and BA is incorrectly suspected from the biopsy in 0–46% [1]. Patholo-
10 (30)
48 (50)
132 (62)
72 (36)
131 (62)
82 (41)

gists participating in the Biliary Atresia Research Consortium found

15 (8)

that approximately 90% of BA patients were diagnosed as having ob-

structive findings by the pathologists, but a diagnosis of “no obstruc-
Perc. biopsy – BA favored by pathologist

tion” or “indeterminate” was favored in 14% of biopsy readings from

Perc. cholangiogram – diagnostic of BA
HIDA – no drainage within 24 hours

BA patients, and obstructive changes were favored in 21–50% of read-

GGT N150 IU/L and CB N2.5 mg/dL

ings from non-BA patients [28]. We found that percutaneous liver biop-
Ultrasound – absent gallbladder

sy reports almost always favored large duct obstruction in infants with

BA, but rarely excluded BA in patients with other causes of cholestasis,
owing primarily to the presence of variable degrees of bile duct prolifer-
Comorbidities absent

ation in most cholestatic patients (with and without BA). While highly
sensitive, in our study, biopsy rarely prevented a trip to the operating
Acholic stools

room even when key features such as bile duct proliferation were not
reported on pathologic review. We therefore suggest that in order to
Table 7

shorten time to intervention, when the diagnosis is in doubt, liver biop-

sy should be obtained only after BA is excluded either by clinical history
370 T. Jancelewicz et al. / Journal of Pediatric Surgery 50 (2015) 363–370
or, if necessary, by PC or operative exploration, since a “positive” biopsy
is extremely likely in these circumstances, and “negative” biopsies were
occasionally ignored in our population. This approach disagrees with
the NASPGHAN committee's report, which recommended biopsy prior
to surgical intervention, with repeat biopsy possibly necessary if the
first attempt was early in the course of the disease. Because of the inher-
ent selection bias in our study, however, prospective validation of this
algorithm is required to completely assess the utility of liver biopsy
and all other tests in the workup of BA.
One other area of potential controversy is the exclusion of patients
with low CB and GGT in the algorithm. CB levels are elevated in all pa-
tients with BA soon after birth [29]; in keeping with those data, we
found no BA patients with a normal CB level, even in those who present-
ed early. Therefore, we emphasize that a normal CB at any time in the
clinical course effectively rules out BA and should prevent further diag-
nostic consideration in that direction. If direct bilirubin (DB) is mea-
sured rather than CB, the level will be slightly higher but a cutoff of
2.5 mg/dL shouldn't result in the inclusion of any normal patients in
the algorithm, as a DB or CB N2.0 mg/dL is almost always pathologic
[30]. The use of a normal GGT level as an exclusion criterion is more con-
troversial, since the GGT level varies somewhat with patient age [20].
However, numerous studies have demonstrated the sensitivity and
low specificity of elevated GGT in the discrimination of BA [19,31,32],
and a recent report of over 600 patients supports its usefulness in the as-
sessment of cholestatic infants [33]. All BA patients in our cohort had a
high GGT, even if a high-normal cutoff of 100 U/L was used; in fact
this number (99.5 U/L) was reported as the highest value seen in a
study of infant reference-range GGT levels [34]. For the algorithm we
used the lowest value seen in a BA patient (150 U/L).
Interestingly, an algorithm nearly identical to our own (including
the same cutoff GGT value of 150 U/L) was reported by Arora et al in
2001 in a prospective study of 132 patients [32]. All patients with chole-
stasis underwent a workup protocol to establish cholestasis and rule out
alternative diagnoses, and all patients without a clear diagnosis then
underwent HIDA scan and GGT testing. Using their algorithm, they re-
ported a negative laparotomy rate of 22%. The study represents a more
inclusive population than ours, but unfortunately 2/3 of the BA patients
were older than 12 weeks at presentation, and PC and liver biopsy were
not routinely performed. Nonetheless, the authors were able to exclude
BA with relative efficiency.
It should be mentioned that our algorithm is probably most useful at
a tertiary referral center where pediatric gastroenterology and pediatric
surgery consultation is available. There are two major limitations to this
study: it is retrospective, and patient selection may limit the applicabil-
ity of our results. With regard to possible selection bias, we chose our
comparison cohort based on the use of a HIDA scan, which should
have selected for patients in whom the clinical picture was ambiguous,
since this test is not universally ordered on all patients with cholestasis
at our institution. The ideal study would prospectively follow all pa-
tients with cholestasis using this algorithm.
4. Conclusion
We propose a screening algorithm for infants with conjugated
hyperbilirubinemia that permits efficient and timely exclusion of BA
with minimal invasive testing and with a low risk of negative laparotomy.
This algorithm now requires prospective evaluation to determine its diag-
nostic accuracy and its ability to reduce hospital costs, patient morbidity,
and time to diagnosis and Kasai portoenterostomy in patients with BA.
References
[1] 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 Gas-
troenterology, Hepatology and Nutrition. J Pediatr Gastroenterol Nutr 2004;39:115–28.
[2] Chardot C, Carton M, Spire-Bendelac N, et al. Is the Kasai operation still indicated in
children older than 3 months diagnosed with biliary atresia? J Pediatr 2001;138:
224–8.
[3] Wong KK, Chung PH, Chan IH, et al. Performing Kasai portoenterostomy beyond
60 days of life is not necessarily associated with a worse outcome. J Pediatr
Gastroenterol Nutr 2010;51:631–4.
[4] Benchimol EI, Walsh CM, Ling SC. Early diagnosis of neonatal cholestatic jaundice:
test at 2 weeks. Can Fam Physician 2009;55:1184–92.
[5] Brumbaugh D, Mack C. Conjugated hyperbilirubinemia in children. Pediatr Rev
2012;33:291–302.
[6] Meyers RL, Book LS, O'Gorman MA, et al. Percutaneous cholecysto-cholangiography
in the diagnosis of obstructive jaundice in infants. J Pediatr Surg 2004;39:16–8.
[7] Batts KP, Ludwig J. Chronic hepatitis. An update on terminology and reporting. Am J
Surg Pathol 1995;19:1409–17.
[8] Jensen MK, Biank VF, Moe DC, et al. HIDA, percutaneous transhepatic cholecysto-
cholangiography and liver biopsy in infants with persistent jaundice: can a combina-
tion of PTCC and liver biopsy reduce unnecessary laparotomy? Pediatr Radiol 2012;
42:32–9.
[9] Abramson SJ, Treves S, Teele RL. The infant with possible biliary atresia: evaluation
by ultrasound and nuclear medicine. Pediatr Radiol 1982;12:1–5.
[10] Ikeda S, Sera Y, Akagi M. Serial ultrasonic examination to differentiate biliary atresia
from neonatal hepatitis—special reference to changes in size of the gallbladder. Eur J
Pediatr 1989;148:396–400.
[11] Park WH, Choi SO, Lee HJ. The ultrasonographic “triangular cord” coupled with gall-
bladder images in the diagnostic prediction of biliary atresia from infantile
intrahepatic cholestasis. J Pediatr Surg 1999;34:1706–10.
[12] Tan Kendrick AP, Phua KB, Ooi BC, et al. Biliary atresia: making the diagnosis by the
gallbladder ghost triad. Pediatr Radiol 2003;33:311–5.
[13] Kanegawa K, Akasaka Y, Kitamura E, et al. Sonographic diagnosis of biliary atresia in
pediatric patients using the “triangular cord” sign versus gallbladder length and con-
traction. AJR Am J Roentgenol 2003;181:1387–90.
[14] Park WH, Choi SO, Lee HJ. Technical innovation for noninvasive and early diagnosis
of biliary atresia: the ultrasonographic “triangular cord” sign. J Hepatobiliary
Pancreat Surg 2001;8:337–41.
[15] Lin WY, Lin CC, Changlai SP, et al. Comparison technetium of Tc-99m disofenin
cholescintigraphy with ultrasonography in the differentiation of biliary atresia
from other forms of neonatal jaundice. Pediatr Surg Int 1997;12:30–3.
[16] Kirks DR, Coleman RE, Filston HC, et al. An imaging approach to persistent neonatal
jaundice. AJR Am J Roentgenol 1984;142:461–5.
[17] Kianifar HR, Tehranian S, Shojaei P, et al. Accuracy of hepatobiliary scintigraphy for
differentiation of neonatal hepatitis from biliary atresia: systematic review and
meta-analysis of the literature. Pediatr Radiol 2013;43:905–19.
[18] Siu LY, Wong KN, Li KW, et al. Outcome of hepatobiliary scanning: preterm versus
full-term cholestatic infants. J Paediatr Child Health 2013;49:E46–51.
[19] Rendon-Macias ME, Villasis-Keever MA, Castaneda-Mucino G, et al. Improvement in
accuracy of gamma-glutamyl transferase for differential diagnosis of biliary atresia
by correlation with age. Turk J Pediatr 2008;50:253–9.
[20] Yamagiwa I, Iwafuchi M, Obata K, et al. Pre-operative time course changes in liver
function tests in biliary atresia: its usefulness in the discrimination of biliary atresia
in early infancy. Acta Paediatr Jpn 1996;38:506–12.
[21] Ovchinsky N, Moreira RK, Lefkowitch JH, et al. Liver biopsy in modern clinical prac-
tice: a pediatric point-of-view. Adv Anat Pathol 2012;19:250–62.
[22] Treem WR, Grant EE, Barth KH, et al. Ultrasound guided percutaneous
cholecystocholangiography for early differentiation of cholestatic liver disease in in-
fants. J Pediatr Gastroenterol Nutr 1988;7:347–52.
[23] Nwomeh BC, Caniano DA, Hogan M. Definitive exclusion of biliary atresia in infants
with cholestatic jaundice: the role of percutaneous cholecysto-cholangiography.
Pediatr Surg Int 2007;23:845–9.
[24] Chardot C. Endoscopic retrograde cholangiopancreatography in patients with neo-
natal cholestasis: an additional diagnostic tool for selected indications. J Pediatr
Gastroenterol Nutr 2009;49:380–1.
[25] Petersen C, Meier PN, Schneider A, et al. Endoscopic retrograde cholangiopancreaticography
prior to explorative laparotomy avoids unnecessary surgery in patients suspected for biliary
atresia. J Hepatol 2009;51:1055–60.
[26] Senyuz OF, Yesildag E, Emir H, et al. Diagnostic laparoscopy in prolonged jaundice. J
Pediatr Surg 2001;36:463–5.
[27] Whitington PF. Fortune telling in biliary atresia: what is in the tea leaves? Liver
Transpl 2009;15:829–30.
[28] Russo P, Magee JC, Boitnott J, et al. Design and validation of the biliary atresia re-
search consortium histologic assessment system for cholestasis in infancy. Clin
Gastroenterol Hepatol 2011;9:357–362.e2.
[29] Harpavat S, Finegold MJ, Karpen SJ. Patients with biliary atresia have elevated direct/
conjugated bilirubin levels shortly after birth. Pediatrics, 2015;128:e1428-e1433.
[30] Davis AR, Rosenthal P, Escobar GJ, et al. Interpreting conjugated bilirubin levels in
newborns. J Pediatr 2011;158:562–565.e1.
[31] Tang KS, Huang LT, Huang YH, et al. Gamma-glutamyl transferase in the diagnosis of
biliary atresia. Acta Paediatr Taiwan 2007;48:196–200.
[32] Arora NK, Kohli R, Gupta DK, et al. Hepatic technetium-99m-mebrofenin
iminodiacetate scans and serum gamma-glutamyl transpeptidase levels interpreted
in series to differentiate between extrahepatic biliary atresia and neonatal hepatitis.
Acta Paediatr 2001;90:975–81.
[33] Sun S, Chen G, Zheng S, et al. Analysis of clinical parameters that contribute to the
misdiagnosis of biliary atresia. J Pediatr Surg 2013;48:1490–4.
[34] Huang YC, Chen HL, Tsai KS, et al. Serum gamma-glutamyl transpeptidase activity
and bile acids in normal Taiwanese infants. Acta Paediatr Taiwan 2002;43:245–8.