Extrahepatic Anomalies in Infants With Biliary Atresia: Results of a Large Prospective North American Multicenter Study


  • Potential conflict of interest: Dr. Haber owns stock in Merck. Dr. Molleston received grants from Schering-Plough, Roche, Bristol-Myers Squibb, and Vertex. Dr. Murray owns stock in Merck and received grants from Roche and Gilead. Dr. Romero received grants from Bristol-Myers Squibb. Dr. Rosenthal consults for Merck, General Electric, and Ikaria and received grants from Bristol-Myers Squibb, Vertex, and Gilead. Dr. Schwarz received grants from Roche. Dr. Sokol consults and owns stock in Yasoo Health and also consults for Roche and Ikaria.


The etiology of biliary atresia (BA) is unknown. Given that patterns of anomalies might provide etiopathogenetic clues, we used data from the North American Childhood Liver Disease Research and Education Network to analyze patterns of anomalies in infants with BA. In all, 289 infants who were enrolled in the prospective database prior to surgery at any of 15 participating centers were evaluated. Group 1 was nonsyndromic, isolated BA (without major malformations) (n = 242, 84%), Group 2 was BA and at least one malformation considered major as defined by the National Birth Defects Prevention Study but without laterality defects (n = 17, 6%). Group 3 was syndromic, with laterality defects (n = 30, 10%). In the population as a whole, anomalies (either major or minor) were most prevalent in the cardiovascular (16%) and gastrointestinal (14%) systems. Group 3 patients accounted for the majority of subjects with cardiac, gastrointestinal, and splenic anomalies. Group 2 subjects also frequently displayed cardiovascular (71%) and gastrointestinal (24%) anomalies; interestingly, this group had genitourinary anomalies more frequently (47%) compared to Group 3 subjects (10%). Conclusion: This study identified a group of BA (Group 2) that differed from the classical syndromic and nonsyndromic groups and that was defined by multiple malformations without laterality defects. Careful phenotyping of the patterns of anomalies may be critical to the interpretation of both genetic and environmental risk factors associated with BA, allowing new insight into pathogenesis and/or outcome. (Hepatology 2013;58:1724–1731)


biliary atresia


central nervous system


gamma-glutamyl transpeptidase


inferior vena cava


left ventricle


pulmonary hypertension


right ventricle


splenic malformation


superior vena cava


total anomalous pulmonary venous return/partial anomalous pulmonary venous return


trachea-esophageal fistula


transposition of the great arteries


ventricular septal defect


white blood cells

The etiology of biliary atresia (BA) is unknown. In a large series of European infants reported by Davenport et al.,[1] infants with BA were catalogued by two different presentations: acquired/perinatal/nonsyndromic (∼90%) versus embryonal/syndromic (∼10%). Infants with splenic malformation (SM) and associated laterality defects were placed in the latter category, implying that the pathogenetic and developmental features of the two types of BA probably differed. In contrast, investigators from Taiwan described the BASM syndrome in only 0.7% of BA infants, whereas a total of 15.4% had other major congenital anomalies, suggesting different etiopathologies.[2] The number of potential etiologies that explain the pathogenesis of BA has expanded as the sophistication of scientific methods to detect them has evolved. The viral etiology hypothesis has been supported by a number of reports, such as the finding of cytomegalovirus in the livers of BA infants[3] and characterization of the rotavirus-induced murine model of BA.[4, 5] Other investigators have suggested an important role for primary immunologic dysfunction, possibly secondary to maternal microchimerism.[6] One hypothesis unifying the viral and immune dysfunction concept is that an in utero or perinatal viral infection may trigger an autoimmune attack on the biliary epithelium.[7] Still other groups have used new technologies to examine genetic susceptibility to BA. Leyva-Vega et al.[8] reported overlapping heterozygous deletions of chromosome 2q37.3 in two BA patients; the etiologic significance of these abnormalities is unclear. A genome-wide association study demonstrated a BA susceptibility locus on chromosome10q24.[9] Recent animal and human evidence support a role for epigenetic regulation of interferon-gamma signaling in BA.[10] Kohsaka et al.[11] found human JAG1 missense mutations in about 10% of their BA patients and noted an association of these mutations with a severe phenotype. Hartley et al.[12] suggested that the most likely etiopathogenetic explanation of BA is that there are multiple mechanisms of biliary injury leading ultimately to the one common phenotype of obliterative cholangiopathy.

Given that there well may be more than two forms of BA, we believe that a critical reappraisal of the anomalies associated with BA could provide useful clues as to the etiopathogenesis of the disease and have followed the guidelines which the Center for Disease Control used in the National Birth Defects Prevention Network. This network was established in 1997 in order provide uniform reporting of birth defects that might then be linked to a common etiology. Major birth defects were defined by the following criteria: “a) considered to be a major defect (affecting survival, requiring substantial medical care, or resulting in marked physiological or psychological impairment); b) usually identifiable in the first 6 weeks of life (may be extended for some defects); and c) consistently classifiable.”[13] In an attempt to identify environmental causes of a given disease process (such as toxins or viral infections) as well as genetic causes, the Center for Disease Control has followed the principle that careful homogenous case definition is the optimum way to identify risk factors and that etiologies of disease conditions such as biliary atresia are likely to be distinct for isolated cases without other major birth defects (our Group 1: no major anomalies), cases associated with other major birth defects but not syndromes (our Group 2: major anomalies without laterality defects) versus cases associated with stereotypical anomalies (our Group 3: major anomalies with laterality defects.[14]

The purpose of this study was to use data from the large prospective multicenter study of BA of the North American Childhood Liver Disease Research Network (ChiLDREN) to perform a detailed analysis of congenital anomalies associated with BA. A sub-aim was to determine if certain demographic variables were associated with the subgroups of BA.

Patients and Methods


Infants with suspected BA were enrolled into a prospective longitudinal study of cholestasis in infancy (PROBE: Clinicaltrials.gov NCT00061828) prior to diagnostic surgery at any of 15 centers participating in ChiLDREN. The diagnosis of BA was confirmed by intraoperative cholangiogram and surgical exploration prior to Kasai hepatoportenterostomy. In addition, the central Pathology Committee of the network supported the diagnosis of BA by blinded review of liver biopsies, coupled with examination of the biliary remnants in cases where the biopsy was uncertain. Determination of each associated anomaly was made from information gathered at the time of surgery, by review of imaging and other clinical studies, and by physical examination. When a discrepancy was identified (e.g., no mention of polysplenia made on ultrasound versus polysplenia noted at the time of surgery), a three-person adjudication committee determined the credibility of evidence. After review of all the data collected on a patient, infants were assigned to one of three groups. Group 1 was isolated BA (without major malformations and with a single spleen), Group 2 was BA without laterality defects but with other congenital malformations, including at least one malformation considered major as defined by the National Birth Defects Prevention Study.[13] Group 3 was BA with one or more laterality defects. These defects included splenic abnormalities (asplenia, polysplenia, right-sided spleen, or a double spleen), cardiovascular anomalies (dextrocardia, mesocardia, total or partial anomalous pulmonary venous return [TAPVR/PAPVR], absent or interrupted inferior vena cava [IVC], anomalous/bilateral superior vena cava [SVC]), and/or preduodenal portal vein and gastrointestinal anomalies (“abdominal heterotaxy,” midline/transverse liver, right-sided stomach, intestinal malrotation, and anomalous or annular pancreas).[15]

All children in this study were enrolled between May 29, 2004 and November 1, 2010. Written informed consent was obtained from the parent/legal guardian of each patient and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected by approval by the Institutional Review Committees at each site.

Demographic and Clinical Variables

Extensive demographic information was collected prospectively for each subject. This information included maternal age, paternal age, parity, and fetal exposure to drugs (including prescribed, over-the-counter, recreational, and herbal preparations) and gestational diabetes. Location of the home during the pregnancy was categorized as rural, urban, or suburban. Family history included the presence of autoimmune diseases among the primary family and first-degree relatives and the frequency of autoimmunity in family members was calculated (percent of patients with at least one first-degree relative with autoimmune disease). Both mothers and fathers were asked if any first-degree relatives had one or more of the autoimmune diseases listed in Table 1. Information collected about the child included birth weight, birth length, and sequential laboratory tests from the time of presentation to the evaluation by the specialist. All laboratory tests are reported as measured except that globulin was inferred by subtraction of albumin from total protein.

Table 1. Autoimmune Diseases in First-Degree Relatives About Which Parents Were Queried
Autoimmune liver disease
Primary biliary cirrhosis
Primary sclerosing cholangitis
Autoimmune hepatitis
Autoimmune and connective tissue disease
Systemic lupus erythematosus
Raynaud's syndrome
Rheumatoid arthritis
Multiple sclerosis
Sjogren's syndrome
Autoimmune endocrine diseases
Insulin-dependent diabetes in subjects <30 years old
Thyroid disease including hypothyroid, goiter, thyrotoxicosis, and thyroid disease type unknown
Autoimmune gastrointestinal diseases
Ulcerative colitis
Crohn's disease
Other unspecified autoimmune disease

Analytic Methods

Descriptive data were summarized by means and SDs for continuous variables and as percentages for categorical variables. The data were summarized overall, as well as within each of the three BA groups. In addition to the descriptive analyses, several factors were evaluated for differences across the BA groups. For the continuous variables, analysis of variance was used to assess overall differences among the groups. Where the F-test reached statistical significance (P < 0.1), pairwise comparisons were made for the three BA groups to ascertain specific differences. The categorical variables were assessed by chi-square tests where evidence of general association (P < 0.1) was further explored through pairwise comparisons of the three groups. All analyses were performed using SAS (SAS Institute, 2008, SAS/STAT 9.2 User's Guide, Cary, NC).


Three Distinct BA Groups Identified

The majority of patients with BA were within Group 1, isolated BA without associated major malformations (242/289, 84%). Group 2, BA without laterality defects but with at least one major malformation, encompassed 17 of the 289 BA patients (6%), and Group 3, BA with one or more laterality defects, encompassed 30 of 289 patients (10%). Table 2 summarizes the most common major and minor anomalies reported by system in all 289 subjects and in each of the three groups. Overall, anomalies were most prevalent in the cardiovascular (16% of subjects), and gastrointestinal (14%) systems and splenic anomalies (7%). Group 3 patients with laterality defects accounted for the majority of subjects with cardiac, gastrointestinal, and splenic anomalies. Splenic anomalies were noted in 70% of Group 3 patients.

Table 2. Frequency of Congenital Anomalies in 289 Biliary Atresia Patients by Group
Any AnomalyTotalGroup 1 (Without Major Anomalies)Group 2 (Major Anomalies Without Laterality Defects)Group 3 (Laterality Defects)
 289 242 17 30 
Splenic anomaly217.300002170.0

Group 2 subjects, while also displaying significant cardiovascular (71%) and gastrointestinal (24%) anomalies, also had significant genitourinary (47%) anomalies that were uncommon in Group 3 subjects. The most common genitourinary defects found in this group were cystic kidney and hydronephrosis. Four of the Group 2 patients had vertebral and rib anomalies but only one had a major musculoskeletal anomaly (longitudinal limb deficiency). The cardiovascular anomalies in Group 2 included aortic arch abnormalities, aortic coarctation, atrial septal defects, patent ductus arteriosus, patent foramen ovale, pulmonary artery stenosis, pulmonary valvular stenosis, Tetralogy of Fallot, transposition of the great vessels, and ventricular septal defect. Gastrointestinal anomalies included duodenal/jejuna atresia, esophageal atresia, and imperforate anus. Supporting Table S1 summarizes the distribution of the systems with at least one reported anomaly for the 47 individual patients in Groups 2 and 3. Supporting Table S2 summarizes the distribution of specific genitourinary anomalies across all three groups.

Demographic Variables Associated With BA Groups

Analysis of demographic variables between groups revealed significant differences in the age at first evaluation, with Group 1 having a later age at evaluation compared to Group 3 (Table 3). Recreational drug use during pregnancy was reported more commonly in Group 3 compared to Group 1. There was no difference between the three groups for mother's or father's age, gender, race, history of familial autoimmune disease, z-scores for birth weight or length, or rural versus urban location. For gestational age, the difference across the three groups was significant (F test P = 0.0912). Subsequent pairwise comparison revealed Group 1 infants tended to be slightly older than Group 3 infants (P = 0.0512). The mean maternal age was 29.2 ± 6.0 years and the mean paternal age was 31.9 ± 7.0 years. The incidence of gestational diabetes was increased in Group 3 compared to Group 1. Interestingly, the incidence of an autoimmune disease in first-degree relatives was substantial: 44% overall, with no difference between groups. Sixty-three percent of the whole population of BA infants was white, without differences between the three groups. The race/ethnicity distribution was relatively even across groups but the small sample size makes it difficult to compare anything other than white versus nonwhite.

Table 3. Clinical and Demographic Characteristics of 289 Infants With Biliary Atresia
ParameterNMean ± SD or N (%)P Value vs. Group 2P Value vs. Group 3
  1. a

    Significant at α = 0.1 for differences across BA groups.. Pairwise comparisons were assessed.

  2. BA Group 1, n = 242: infants without major anomalies.

  3. BA Group 2, n = 17: infants with major anomalies without laterality defects.

  4. BA Group 3, n = 30: infants with laterality defects.

Gender (male vs. female) (P = 0.4063)289135 (46.7%)
BA Group 1242109 (45.0%)
BA Group 21710 (58.8%)
BA Group 33016 (53.3%)
Birth weight kg (P = 0.4079)2693.15 ± 0.57
BA Group 12233.16 ± 0.54
BA Group 2163.25 ± 0.56
BA Group 3303.03 ± 0.78
Birth length cm (P = 0.2858)25249.8 ± 3.6
BA Group 121149.9 ± 3.6
BA Group 21549.2 ± 2.6
BA Group 32648.8 ± 4.0
Gestational age in weeks (P = 0.0912)a27138.2 ± 2.2
BA Group 122638.3 ± 2.20.24480.0512
BA Group 21737.6 ± 2.10.7471
BA Group 32837.4 ± 2.30.7471
Age at first evaluation (days) (P = 0.0397a)28868.1 ± 36.3
BA Group 124270.4 ± 37.60.21810.0204
BA Group 21658.9 ± 27.80.6688
BA Group 33054.2 ± 24.70.6688
Recreational drug use during pregnancy (P = 0.0964a)27610 (3.6%)
BA Group 12306 (2.6%)0.43250.0321
BA Group 2171 (5.9%)0.6041
BA Group 3293 (10.3%)0.6041
Rural vs. urban (P = 0.4650)280230 (82.1%)
BA Group 1234191 (81.6%)
BA Group 22622 (84.6%)
BA Group 32017 (85.0%)
Gestational diabetes (yes vs. no) (P < 0.0948a)27932 (11.5%)
BA Group 123223 (9.9%)0.80640.0298
BA Group 2172 (11.8%)0.3328
BA Group 3307 (23.3%)0.3328

Clinical Laboratory Variables Associated With BA Groups

Table 4 reports select clinical and laboratory variables that were prospectively collected. While total bilirubin did not differ across the three groups, there was a difference in direct bilirubin across groups (F test P = 0.0693). Group 1 infants tended to have a higher direct bilirubin values compared to Group 2 and Group 3, although neither of these pairwise comparison reached significance at the P = 0.05 threshold (P = 0.0999 and P = 0.0654, respectively). Gamma-glutamyl transpeptidase (GGTP) was similar across the groups. Alkaline phosphatase was significantly higher in Group 1 compared to Group 2. After adjusting for age at first evaluation, these laboratory differences across the groups remained (data not shown).

Table 4. Selected Laboratory Characteristics of 289 Infants With Biliary Atresia
ParameterNMean ± SDP Value vs. Group 2P Value vs. Group 3
  1. a

    Significant at α = 0.1. Pairwise comparisons were assessed.

  2. BA Group 1, n = 242: infants without major anomalies.

  3. BA Group 2, n = 17: infants with major anomalies without laterality defects.

  4. BA Group 3, n = 30: infants with laterality defects.

Total bilirubin (P = 0.4777)2778.43 ± 3.48  
BA Group 12318.53 ± 3.54  
BA Group 2178.29 ± 3.44  
BA Group 3297.70 ± 3.01  
Direct bilirubin (P = 0.0693)a1675.68 ± 2.29  
BA Group 11385.87 ± 2.220.09990.0758
BA Group 2124.74 ± 2.49 0.9184
BA Group 3174.83 ± 2.510.9184 
GGTP (P = 0.3921)262686.8 ± 533.4  
BA Group 1217692.2 ± 530.6  
BA Group 216802.0 ± 669.5  
BA Group 329582.4 ± 470.2  
Alkaline phosphatase (P = 0.0010)a272589.6 ± 315.5  
BA Group 1226608.3 ± 316.60.00070.0654
BA Group 217342.5 ± 131.6 0.1056
BA Group 329495.6 ± 316.00.1056 
Total protein (P = 0.0140)a2136.00 ± 0.82
BA Group 11766.07 ± 0.810.11210.0094
BA Group 2155.73 ± 0.960.6263
BA Group 3225.60 ± 0.590.6263
Albumin (P = 0.0003)a2683.56 ± 0.54
BA Group 12243.62 ± 0.510.00260.0034
BA Group 2173.22 ± 0.660.5895
BA Group 3273.30 ± 0.570.5895
ALT (P = 0.0299)a276150 ± 112
BA Group 1230153 ± 1010.01170.5534
BA Group 21783 ± 500.0141
BA Group 329166 ± 1890.1041
WBC (P = 0.0121)a26713.67 ± 4.54
BA Group 122313.41 ± 4.330.79980.0034
BA Group 21613.11 ± 3.910.0369
BA Group 32816.06 ± 5.760.0369
Platelets (P = 0.0416)a266432.0 ± 178.1
BA Group 1224426.1 ± 165.20.35110.0238
BA Group 215382.1 ± 172.90.0278
BA Group 327508.0 ± 256.00.0278

Total protein and albumin levels were higher in Group 1 compared to Group 3. Alanine aminotransferase was lower in Group 2. Group 3 was characterized by higher white blood cell counts and platelet counts versus the other two groups.


In this prospective North American multicenter study of BA, we identified three groups of BA patients. The most common was isolated BA, the perinatal or acquired form of BA without associated major malformations (Group 1). A second group was identified whereby not only gastrointestinal and cardiac anomalies were associated with BA in the absence of laterality defects, but also findings of genitourinary anomalies (Group 2). The most frequent renal anomalies reported in Group 2 were cystic kidneys or hydronephrosis. The observation that as many as 16% of children with BA may have heart disease and 3% may have renal anomalies makes differentiation from Alagille syndrome difficult. Likewise, the fact that infants with BA may occasionally have cystic kidneys may make differentiation from infants with polycystic liver-kidney disease a bit of a challenge, although cholestasis is rare in the latter condition. The incidence of clinically significant hydronephrosis in otherwise healthy newborns is ∼1 in 600 live births (0.17%).[16, 17] The incidence of hydronephrosis in BA patients in this study (all within Group 2) was 3 in 289 (1%), an almost 10-fold greater incidence compared to the general population. There is scant recent literature on genitourinary and musculoskeletal abnormalities associated with BA. A case report described an infant with BASM, sacro-coccygeal agenesis, clubfoot, and ano-urinary incontinence.[18] A BA patient with anorectal agenesis and a complicated urogenital malformation was also described.[9] It is known that many genitourinary anomalies are associated with concurrent vertebral segmentation anomalies.[20]

In our study of Group 2 patients with genitourinary and musculoskeletal abnormalities, a similar association to that previously reported in the literature is suggested. In addition, some in Group 1 had clinically insignificant rib or vertebral defects. Twenty years ago Carmi et al.[21] reported that one-third of their 51 BA patients with major anomalies had laterality defects but two-thirds had cardiac, genitourinary, and musculoskeletal defects not associated with laterality defects. Our report confirms their findings, extends the spectrum of renal anomalies observed, and also strongly reinforces the authors' suggestion that there is etiologic heterogeneity in BA.

In a large study from England the incidence of splenic anomalies was 10.2%,[1] almost identical to the incidence identified in this study. The investigators from England also reported similar rates of intestinal malrotation, absent or interrupted IVC, and preduodenal portal vein in patients with splenic anomalies. Fifteen percent of the patients in their series with laterality defects were born to mothers with diabetes and this association was not found in their BA patients without laterality defects. Gestational diabetes was observed in 9.9%, 11.8%, and 23.3% of our infants in Groups 1, 2, and 3. Interestingly, the English study also found a female predominance of 2:1 in patients with splenic anomalies, a finding that was not identified in our North American cohort. As noted in the article by Davenport et al.,[1] in addition to “BASM,” another term for infants with BA and stereotypical syndromic abdominal and vascular anomalies is “biliary atresia laterality sequence.” Given that only 70% of our patients with laterality defects actually had splenic anomalies, the latter term might be preferable in the future to “BASM” to describe this stereotypical group of infants.

The Canadian Pediatric Hepatology Research group has recently reported their analysis of 382 infants with BA and the associated anomalies.[22] Forty-four (13%) had associated anomalies, only 25 (6.5%) of which were associated with SM. The authors concluded that BA infants with anomalies demonstrated a spectrum of laterality defects and suggested that the meaning of the acronym BASM be modified to “biliary atresia structural malformation.” Our conclusions are somewhat similar in that a total of 16% of our infants were in the anomaly Groups 2 and 3. On the other hand, the main difference between our observations and those of the Canadian group was that Group 2 infants frequently exhibited major birth defects of the genitourinary and/or gastrointestinal systems, not considered part of defective lateralization, suggesting that this group may represent a different etiopathogenesis than Groups 1 and 3.

Group 3 infants were younger at the time of initial evaluation compared to Group 1. The associated anomalies in Group 3, especially the cardiac lesions associated with murmurs or cyanosis, probably brought the patient to medical attention sooner than the infants with isolated cholestasis.

An unexpected finding was the high incidence of autoimmunity in first-degree relatives of all BA groups (average 44%). The occurrence of autoimmune diseases in relatives provides circumstantial evidence that a candidate disease (i.e., BA) may be autoimmune in nature.[23] The incidence of autoimmunity in first-degree relatives is much higher than that found in the general population, where autoimmunity rates vary from 2.5%-9%.[26, 27] Importantly, the incidence of autoimmunity in first-degree relatives of BA patients was similar to the rate of 37%-43% identified in autoimmune hepatitis[26] and 25.5% in type-1 diabetes mellitus.[25] This intriguing finding of autoimmunity in first-degree relatives of BA patients warrants further investigation. The fact that there was no difference in autoimmunity rates between the three groups suggests that the autoimmune hypothesis of BA may be relevant to the pathogenesis of all types of BA and is a clue to be pursued in further studies. It is also possibly that the high incidence simply resulted from our rigorous questionnaire containing a long list of autoimmune diseases and not being of pathogenetic significance. We agree that the lack of differences between groups is not only at variance to explain the autoimmune hypothesis of Group 1 but also at variance to explain the genetic, nonimmune hypothesis of syndromic BA, in which lower familial rates of autoimmune disease would have been expected. Possible explanations include the possibility that this hypothesis is incorrect versus the immune dysregulation hypothesized for Group 1 BA[27] being atypical from the usual types of familial autoimmune diseases.

The analysis of laboratory tests revealed no difference in total bilirubin across the groups, although infants in Group 1 had higher alkaline phosphatase levels and they also tended to have higher direct bilirubin values. The significance of this observation is uncertain. Group 1 infants tended to be older at the time of initial evaluation and thus could be hypothesized to have a longer duration of obstruction. We explored this possibility by adjusting for age at first evaluation and the laboratory differences across the groups remained, suggesting age alone was not responsible. Group 1 infants had higher total serum albumin levels compared to Groups 2 and 3. It has been reported that newborns have lower albumin levels that increase with age.[28] Both Groups 2 and 3 were younger at the time of evaluation compared to Group 1 and the younger age at presentation may explain the lower albumin levels. Furthermore, it is possible that increased protein and albumin losses could be associated with some of the anomalies present in Groups 2 and 3. Specifically, intestinal atresias could lead to intestinal protein loss and renal anomalies could result in urinary protein loss. Finally, higher total white cell counts and platelet counts were identified in Group 3 compared to the others. The hemodynamics within the spleen in polysplenia are most likely altered and it is theorized that decreased filtration through the splenic venules would be associated with decreased trapping and removal of white cells and platelets.

In summary, BA is a heterogeneous disease that is composed of at least three subgroups. This study identified a group that was defined by multiple malformations including genitourinary anomalies, reinforcing a similar report by Carmi et al. in 1993.[21] Future investigations are indicated to determine if each of these subtypes is associated with unique predisposition or etiology.