Medical information was collected from medical records. Demographic data, including race, sex, date of birth, and age at HPE, were collected for each patient. Dietary intake, nutritional indices, physical examination findings, and laboratory data were obtained from the records for each patient within one week of the time of HPE, transplant listing, clinical endpoint (liver transplant, death, or removal from transplant waiting list due to improved clinical status), and initiation of parenteral nutrition (PN). For each patient receiving PN, data including details on the composition of PN were collected at each monthly visit until the clinical endpoint.

Anthropometric data were abstracted from the medical record. Weight, length, and weight/length z-scores were calculated based on the Centers for Disease Control (CDC) growth charts. Triceps skin fold (TSF) thickness and mid arm circumference (MAC) were measured as part of the standard of care at our institution. MAC and TSF were collected using standard technique with Lange calipers by five clinical pediatric dieticians fully trained to measure anthropometrics.(14) MAC was obtained in a single measurement and documented in the medical record. TSF was obtained 3 times for each clinic visit and the average was recorded in the medical record by the clinical pediatric dietician. TSF and MAC z-scores were calculated using results of the United States Ten-State Nutritional Survey.(15) The use of nutritional supplementation (enteral or parenteral), type and amount of oral formula and IV PN formulation used were obtained from the medical record, and total kcal/kg/day was calculated based on the patient’s measured weight and intake. We included dietary intake in the analysis when it was completely documented in the medical record. Information on enteral intake and the composition of PN were collected. For each patient receiving PN, the mean glucose infusion rate (GIR), mean daily dose of intravenous lipids, and mean daily dose of amino acids were calculated. The median GIR, dose of intravenous lipids, and dose of amino acids over the study period were calculated based on aggregate averages for each patient receiving PN. Soy-oil-based intravenous lipid emulsion (Intralipid, Fresenius Kabi, IL, USA) was used in all patients; no patients received fish oil-based intravenous lipid emulsions in this cohort. Amino acid solutions were given as TrophAmine (B. Braun, Bethlehem, PA, USA) or Aminosyn (Hospira, Lake Forest, IL, USA). Standard chemistries and complete blood count were monitored monthly while on PN. Adjustments in PN formulation in cholestatic infants included removal of supplemental manganese and halving of the dose of copper supplementation.

The decision to initiate nutritional supplementation (both enteral and parenteral) was based on the clinical judgment of individual care providers. Enteral supplementation with nasogastric tube feedings was initiated when TSF and MAC z-scores were consistently falling despite maximal attempts with both increasing caloric density and volume of oral feedings. In some patients, z-scores <-2 SD below normal were the indication to initiate NG tube feedings In other cases, in order to prevent severe malnutrition, consistently falling z-scores were the indication to initiate NG feeds. Parenteral nutrition was started when maximal enteral feedings (oral and NG tube), based on both caloric density and volume of feeds, failed to achieve restitution of TSF and MAC z-scores.

Laboratory data obtained from the medical records included serum albumin, total and direct bilirubin, GGT, AST, ALT, alkaline phosphatase, white blood count (WBC), hemoglobin, hematocrit, prothrombin time, and INR. The Pediatric End-Stage Liver Disease (PELD) score was calculated at each time point based on laboratory results and growth parameters.(16)

The median age at PN initiation was 7.7 months (range: 1.9 to 26 months) and mean duration of PN supplementation was 86 ± 18 days. At the time of clinical endpoint (transplant, death, or removal from transplant waiting list), mean energy supplied by PN was 77 ± 5.5 kcal/kg/day. Median glucose infusion rate was 13.2 (range 6.1-17.2), median intravenous lipid dose was 2.7 gm/kg/day (range 1.7-3.9 gm/kg/day), and median intravenous amino acid dose was 2 gm/kg/day (range 0.7-3.2 gm/kg/day). Two patients received intravenous lipid only (mean 2.5 gm/kg/day).

Nutritional intake of patients receiving PN was compared to the non-PN group. Sixty nine percent of the cohort had complete documentation of dietary intake and were included in this analysis. The remaining 31% had incomplete data available for calculation of enteral intake (50% of these were breastfed, 20% were on a “regular diet,” 30% received an undocumented amount of formula) and were excluded from further analyses. The non-PN group was missing more dietary records than the PN group (40% vs. 17%, p = 0.01). Mean total energy intake (kcal/kg/day) in PN and non-PN groups at the time of HPE (143 ± 10 vs. 150 ± 10), transplant listing (121 ± 7 vs. 111 ± 7), and clinical endpoint (104 ± 7 vs. 103 ± 8) were similar (p=NS). Although total kcal/kg/day at the time of clinical endpoint was the same in both groups, subjects in the PN group received a majority (63%) of their energy intake from parenteral nutrition and 37% of their energy intake from enteral nutrition (Figure 1).

Measures of muscle mass (MAC) and adipose tissue stores (TSF) are believed to better reflect nutritional status than weight and weight/length measurements in infants and children with chronic liver disease because of the contribution of an enlarged liver and/or spleen and the presence of ascites and fluid retention that may inflate the weight for calculation of the growth parameters.(2, 8-13) Therefore, MAC and TSF z-scores were used as primary measures of nutritional status. At the time of HPE, MAC (-1.4 ± 0.2 vs. -1.2 ± 0.3) and TSF (-1.4 ± 0.1 vs. -1.4 ± 0.2) z-scores were similar in PN vs. non-PN groups, p=NS (Figure 2). At the time of transplant listing, the PN group had lower MAC z-scores compared with the non-PN group (-1.7 ± 0.2 vs. -0.8 ± 0.3, p=0.01). Between listing and clinical endpoint, MAC z-scores improved in the PN group (-1.7 ± 0.2 to -1.4 ± 0.2, p=0.02) but worsened in the non-PN group (-0.8 ± 0.3 to -1.3 ± 0.2, p = 0.03). Following transplant listing, the PN group had an average monthly rate of change in MAC z-score of +0.30 z-score units per month while the non-PN group had a monthly change in rate of -0.15 z-score units per month (p=0.0002). At the time of transplant listing, the PN group had lower TSF z-scores compared with the non-PN group (-2.2 ± 0.2 vs. -1.5 ± 0.3, p=0.001). Between listing and clinical endpoint, TSF z-scores improved significantly in the PN group (-2.2 ± 0.2 to -1.8 ± 0.2, p =0.006) and were unchanged in the non-PN group (-1.5 ± 0.2 to -1.6 ± 0.1, p=NS). The average monthly rate of change in TSF z-score from listing to clinical endpoint was +0.25 and +0.02 z-score units respectively for PN and non-PN groups (p=0.04). At clinical endpoint, there was no significant difference between the two groups with respect to MAC and TSF z-score. Similar patterns of change were seen in weight, length, and weight/length z-scores; however, the differences between groups were not statistically significant.

At the time of HPE, the PN and non-PN groups had similar baseline laboratory studies (Table 2). At the time of transplant listing, patients on PN had lower mean serum albumin (2.8 vs. 3.2 g/dL, p=0.03), GGT (528 vs. 940 units/L, p =0.04), and platelet counts (202 vs. 330 × 10³/μL, p = 0.001). At the time of clinical endpoint, mean serum albumin (2.8 vs. 3.1 g/dL, p=NS) was similar in the PN and the non-PN group, however the PN group had higher serum total bilirubin (21.9 vs. 13.8 mg/dL, p=0.04), more prolonged prothrombin time (18.5 vs. 14.6 sec, p=0.01), and lower GGT (210 vs. 502 units/L, p=0.0007).

Weight, length, TSF, and MAC z-scores were compared to serum albumin, total bilirubin, and prothrombin time at the time of transplant listing. TSF z-score was weakly associated with serum albumin (R²=0.34, p = 0.05) and weight z-score was weakly associated with prothrombin time (R²=0.32, p = 0.03). All other associations were not significant.

Medical outcomes were compared between the PN and non-PN groups from time of HPE until clinical endpoint. The PN group had a higher incidence of GI bleeding (60.0% vs. 14%, p = 0.002) and development of ascites (76.0% vs. 36.4%, p=0.009) than the non-PN group. There was no significant association between presence of a central line and the incidence of a positive blood culture (48% of patients with a central line developed bacteremia compared to 20% of patients without a central line, p = 0.09). There was no significant association between PN use and incidence of a positive blood culture (52% of PN group developed bacteremia compared with 41% of non-PN group, p = 0.89).

Malnutrition, a significant problem for children with biliary atresia, can be difficult to assess as ascites, organomegaly, and peripheral edema may confound interpretation of their weight and weight/length measurements.(2, 13) Weight z-scores and weight/length percentiles have been shown to overestimate nutritional status in patients with biliary atresia. MAC (a measure of muscle mass) and TSF (a measure of adipose stores) more accurately assess the state of malnutrition in these patients and were utilized as the primary markers of nutritional status in this study.(2, 8-13)

Decreased oral intake, early satiety, fat malabsorption, and increased energy expenditure due to a hypermetabolic state all likely contribute to malnutrition in biliary atresia patients.(17) As mean energy intake (111 kcal/kg/day) in the PN groups was well below estimated caloric needs (131 kcal/kg/day), with signs of progressive malnutrition (lower MAC and TSF z-scores), parenteral nutrition was initiated to improve nutritional status while the patients awaited liver transplantation.(3) Within the PN group, 64% had already failed to show improved nutritional status following the initiation of NG tube feeding supplementation, thus PN was considered the only viable option. With the addition of PN, the total energy intake was maintained in the PN group at the same level as the non-PN group at clinical endpoint. PN administration reversed the trend of falling TSF z-scores in the PN group and restored values to those observed in the non-PN group by the time of the clinical endpoint. At transplant listing and time of the clinical endpoint, although total energy intake was similar in both groups, MAC and TSF z-scores had improved only in the PN group. It is likely that the PN group would have had continued deterioration of nutritional status following transplant listing had PN not been instituted.

At transplant listing, albumin levels and platelet counts were significantly lower in the PN compared to the non-PN group, reflecting malnutrition and more severe liver disease. Among patients receiving PN, albumin levels stabilized or improved after listing. Platelet counts, however, continued to fall, and serum bilirubin and prothrombin time rose after transplant listing in the PN group, suggesting progression of chronic liver failure and portal hypertension. Moreover, at clinical endpoint the serum bilirubin and prothrombin time were higher and platelet counts significantly lower in the PN vs. non-PN group. Thus, despite worsening liver function, PN administration was successful in improving or stabilizing nutritional status in these infants with end-stage biliary atresia.

Malnourished biliary atresia patients are at higher risk of poor pre-transplant outcomes.(5, 18) Length and weight z-scores >-2 are associated with better outcomes in biliary atresia patients.(5) Once listed for liver transplant, malnutrition (length or weight z-scores <-2) is associated with increased need for intensive care unit monitoring pre- transplant and mortality.(16) Decreasing weight and length z-scores and serum albumin are also risk factors for death pre- transplant.(6) In our PN group, TSF, MAC and albumin levels were decreasing prior to PN administration, and stabilized/increased following PN. Prior to transplant, the PN group had a higher incidence of GI bleeding and ascites compared with the non-PN group. Although this most likely represents more advanced liver disease and portal hypertension in the PN group, parenteral nutrition, with subsequent expansion of vascular volume, may have also contributed. The administration of PN was not, however, associated with increased bacteremia or pre-transplant mortality. In summary, based on expected worse pre-transplant outcomes in BA patients who are severely malnourished,(5, 6, 16) we speculate that the pre-transplant complications of liver disease would have been even worse in the PN group had they not achieved improved nutritional status provided by PN administration.

Malnourished biliary atresia patients are also at increased risk for significant post-transplant complications. Growth failure at the time of liver transplantation confers increased risk for graft failure and post-transplant death.(6, 7, 18) In our study, PN improved MAC and TSF z-scores to > -2 at clinical endpoint. Although weight z-scores also improved, these findings are difficult to interpret given the frequency of ascites and fluid retention in this population. Following transplant, both PN and non-PN groups had similar clinical outcomes (days in ICU post-transplant, need for re-transplant, graft and patient survival). We postulate that without PN driven improvements in nutritional status, this group would have had poorer post-transplant outcomes compared to the non-PN group.(6, 7, 18)

Although this study addresses the importance of nutrition in biliary atresia, there are several limitations. The retrospective nature of this study resulted in incomplete nutritional data at all time points for some subjects (e.g., those that were breast fed). Although bacteremia rates were similar, we were unable to normalize bacteremia events to number of days with intravenous catheters in place because of inadequate documentation of line removal dates in the medical record. In addition, although pre-transplant mortality was similar between the PN and non-PN group, the possibility of a Type 2 error exists. We acknowledge that the PN and non-PN groups were not clinically equivalent (the PN group had more advanced disease and more severe portal hypertension), thus limiting to some extent the comparison of outcomes of these two groups. However, the more advanced liver disease in the PN group would have biased this group towards worse clinical outcomes, which were not observed.

It should be noted that, despite improved nutritional status in patients receiving PN, their liver disease appeared to progress more rapidly after PN initiation (based on higher bilirubin, lower platelets, longer prothrombin time, and higher calculated PELD) in the setting of a significantly lower GGT, compared to the non-PN group. PN-associated liver disease is well-documented in premature infants and infants with intestinal failure requiring PN supplementation(19) but has not been described in BA patients. Our PN patients received doses of soy-based intravenous lipids in the range implicated in the pathogenesis of liver injury in infants with intestinal failure.(20-22) Thus, it is possible that PN may have contributed to the more rapid progression of cholestasis in the PN group. Lipid sparing techniques and use of omega-3 fatty acid-based intravenous lipid preparations, shown to reverse PN-associated cholestasis(20-22), might be considered for BA patients in the future to minimize this possibility.